Improved compatibility for automatically enabling huge pages on Linux systems without NUMA support.

feat: initial haiku os support

.github/ISSUE_TEMPLATE/bug_report.md

@@ -17,6 +17,9 @@ Steps to reproduce the behavior.
 A clear and concise description of what you expected to happen.
 
 **Required data**
+ - XMRig version
+    - Either the exact link to a release you downloaded from https://github.com/xmrig/xmrig/releases
+    - Or the exact command lines that you used to build XMRig
  - Miner log as text or screenshot
  - Config file or command line (without wallets)
  - OS: [e.g. Windows]

.gitignore

@@ -4,3 +4,4 @@ scripts/deps
 /CMakeLists.txt.user
 /.idea
 /src/backend/opencl/cl/cn/cryptonight_gen.cl
+.vscode

CHANGELOG.md

@@ -1,3 +1,52 @@
+# v6.24.0
+- [#3671](https://github.com/xmrig/xmrig/pull/3671) Fixed detection of L2 cache size for some complex NUMA topologies.
+- [#3674](https://github.com/xmrig/xmrig/pull/3674) Fixed ARMv7 build.
+- [#3677](https://github.com/xmrig/xmrig/pull/3677) Fixed auto-config for AMD CPUs with less than 2 MB L3 cache per thread.
+- [#3678](https://github.com/xmrig/xmrig/pull/3678) Improved IPv6 support: the new default settings use IPv6 equally with IPv4.
+
+# v6.23.0
+- [#3668](https://github.com/xmrig/xmrig/issues/3668) Added support for Windows ARM64.
+- [#3665](https://github.com/xmrig/xmrig/pull/3665) Tweaked auto-config for AMD CPUs with < 2 MB L3 cache per thread.
+
+# v6.22.3
+- [#3605](https://github.com/xmrig/xmrig/pull/3605) CUDA backend: added missing RandomX dataset update.
+- [#3646](https://github.com/xmrig/xmrig/pull/3646) Optimized auto-config for AMD CPUs with less than 2 MB L3 cache per thread.
+- [#3652](https://github.com/xmrig/xmrig/pull/3652) Fixed possible crash when submitting RandomX benchmark.
+- [#3662](https://github.com/xmrig/xmrig/pull/3662) Fixed OpenCL kernel compilation error on some platforms.
+
+# v6.22.2
+- [#3569](https://github.com/xmrig/xmrig/pull/3569) Fixed corrupted API output in some rare conditions.
+- [#3571](https://github.com/xmrig/xmrig/pull/3571) Fixed number of threads on the new Intel Core Ultra CPUs.
+
+# v6.22.1
+- [#3531](https://github.com/xmrig/xmrig/pull/3531) Always reset nonce on RandomX dataset change.
+- [#3534](https://github.com/xmrig/xmrig/pull/3534) Fixed threads auto-config on Zen5.
+- [#3535](https://github.com/xmrig/xmrig/pull/3535) RandomX: tweaks for Zen5.
+- [#3539](https://github.com/xmrig/xmrig/pull/3539) Added Zen5 to `randomx_boost.sh`.
+- [#3540](https://github.com/xmrig/xmrig/pull/3540) Detect AMD engineering samples in `randomx_boost.sh`.
+
+# v6.22.0
+- [#2411](https://github.com/xmrig/xmrig/pull/2411) Added support for [Yada](https://yadacoin.io/) (`rx/yada` algorithm).
+- [#3492](https://github.com/xmrig/xmrig/pull/3492) Fixed `--background` option on Unix systems.
+- [#3518](https://github.com/xmrig/xmrig/pull/3518) Possible fix for corrupted API output in rare cases.
+- [#3522](https://github.com/xmrig/xmrig/pull/3522) Removed `rx/keva` algorithm.
+- [#3525](https://github.com/xmrig/xmrig/pull/3525) Added Zen5 detection.
+- [#3528](https://github.com/xmrig/xmrig/pull/3528) Added `rx/yada` OpenCL support.
+
+# v6.21.3
+- [#3462](https://github.com/xmrig/xmrig/pull/3462) RandomX: correct memcpy size for JIT initialization.
+
+# v6.21.2
+- The dependencies of all prebuilt releases have been updated. Support for old Ubuntu releases has been dropped.
+- [#2800](https://github.com/xmrig/xmrig/issues/2800) Fixed donation with GhostRider algorithm for builds without KawPow algorithm.
+- [#3436](https://github.com/xmrig/xmrig/pull/3436) Fixed, the file log writer was not thread-safe.
+- [#3450](https://github.com/xmrig/xmrig/pull/3450) Fixed RandomX crash when compiled with fortify_source.
+
+# v6.21.1
+- [#3391](https://github.com/xmrig/xmrig/pull/3391) Added support for townforge (monero fork using randomx).
+- [#3399](https://github.com/xmrig/xmrig/pull/3399) Fixed Zephyr mining (OpenCL).
+- [#3420](https://github.com/xmrig/xmrig/pull/3420) Fixed segfault in HTTP API rebind.
+
 # v6.21.0
 - [#3302](https://github.com/xmrig/xmrig/pull/3302) [#3312](https://github.com/xmrig/xmrig/pull/3312) Enabled keepalive for Windows (>= Vista).
 - [#3320](https://github.com/xmrig/xmrig/pull/3320) Added "built for OS/architecture/bits" to "ABOUT".

CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.1)
+cmake_minimum_required(VERSION 3.10)
 project(xmrig)
 
 option(WITH_HWLOC           "Enable hwloc support" ON)
@@ -95,7 +95,7 @@ set(HEADERS_CRYPTO
     src/crypto/common/VirtualMemory.h
    )
 
-if (XMRIG_ARM)
+if (XMRIG_ARM OR XMRIG_RISCV)
     set(HEADERS_CRYPTO "${HEADERS_CRYPTO}" src/crypto/cn/CryptoNight_arm.h)
 else()
     set(HEADERS_CRYPTO "${HEADERS_CRYPTO}" src/crypto/cn/CryptoNight_x86.h)
@@ -162,7 +162,7 @@ if (XMRIG_OS_WIN)
         src/crypto/common/VirtualMemory_win.cpp
         )
 
-    set(EXTRA_LIBS ws2_32 psapi iphlpapi userenv)
+    set(EXTRA_LIBS ws2_32 psapi iphlpapi userenv dbghelp)
 elseif (XMRIG_OS_APPLE)
     list(APPEND SOURCES_OS
         src/App_unix.cpp
@@ -240,7 +240,10 @@ add_executable(${CMAKE_PROJECT_NAME} ${HEADERS} ${SOURCES} ${SOURCES_OS} ${HEADE
 target_link_libraries(${CMAKE_PROJECT_NAME} ${XMRIG_ASM_LIBRARY} ${OPENSSL_LIBRARIES} ${UV_LIBRARIES} ${EXTRA_LIBS} ${CPUID_LIB} ${ARGON2_LIBRARY} ${ETHASH_LIBRARY} ${GHOSTRIDER_LIBRARY})
 
 if (WIN32)
+    if (NOT ARM_TARGET)
         add_custom_command(TARGET ${CMAKE_PROJECT_NAME} POST_BUILD COMMAND ${CMAKE_COMMAND} -E copy_if_different "${CMAKE_SOURCE_DIR}/bin/WinRing0/WinRing0x64.sys" $<TARGET_FILE_DIR:${CMAKE_PROJECT_NAME}>)
+    endif()
+
     add_custom_command(TARGET ${CMAKE_PROJECT_NAME} POST_BUILD COMMAND ${CMAKE_COMMAND} -E copy_if_different "${CMAKE_SOURCE_DIR}/scripts/benchmark_1M.cmd" $<TARGET_FILE_DIR:${CMAKE_PROJECT_NAME}>)
     add_custom_command(TARGET ${CMAKE_PROJECT_NAME} POST_BUILD COMMAND ${CMAKE_COMMAND} -E copy_if_different "${CMAKE_SOURCE_DIR}/scripts/benchmark_10M.cmd" $<TARGET_FILE_DIR:${CMAKE_PROJECT_NAME}>)
     add_custom_command(TARGET ${CMAKE_PROJECT_NAME} POST_BUILD COMMAND ${CMAKE_COMMAND} -E copy_if_different "${CMAKE_SOURCE_DIR}/scripts/pool_mine_example.cmd" $<TARGET_FILE_DIR:${CMAKE_PROJECT_NAME}>)
@@ -249,5 +252,5 @@ if (WIN32)
 endif()
 
 if (CMAKE_CXX_COMPILER_ID MATCHES Clang AND CMAKE_BUILD_TYPE STREQUAL Release AND NOT CMAKE_GENERATOR STREQUAL Xcode)
-    add_custom_command(TARGET ${PROJECT_NAME} POST_BUILD COMMAND ${CMAKE_STRIP} ${CMAKE_PROJECT_NAME})
+    add_custom_command(TARGET ${PROJECT_NAME} POST_BUILD COMMAND ${CMAKE_STRIP} "$<TARGET_FILE:${CMAKE_PROJECT_NAME}>")
 endif()

README.md

@@ -10,7 +10,7 @@
 XMRig is a high performance, open source, cross platform RandomX, KawPow, CryptoNight and [GhostRider](https://github.com/xmrig/xmrig/tree/master/src/crypto/ghostrider#readme) unified CPU/GPU miner and [RandomX benchmark](https://xmrig.com/benchmark). Official binaries are available for Windows, Linux, macOS and FreeBSD.
 
 ## Mining backends
-- **CPU** (x86/x64/ARMv7/ARMv8)
+- **CPU** (x86/x64/ARMv7/ARMv8/RISC-V)
 - **OpenCL** for AMD GPUs.
 - **CUDA** for NVIDIA GPUs via external [CUDA plugin](https://github.com/xmrig/xmrig-cuda).
 

cmake/asm.cmake

@@ -1,4 +1,4 @@
-if (WITH_ASM AND NOT XMRIG_ARM AND CMAKE_SIZEOF_VOID_P EQUAL 8)
+if (WITH_ASM AND NOT XMRIG_ARM AND NOT XMRIG_RISCV AND CMAKE_SIZEOF_VOID_P EQUAL 8)
     set(XMRIG_ASM_LIBRARY "xmrig-asm")
 
     if (CMAKE_C_COMPILER_ID MATCHES MSVC)

cmake/cpu.cmake

@@ -21,6 +21,19 @@ if (NOT VAES_SUPPORTED)
     set(WITH_VAES OFF)
 endif()
 
+# Detect RISC-V architecture early (before it's used below)
+if (CMAKE_SYSTEM_PROCESSOR MATCHES "^(riscv64|riscv|rv64)$")
+    set(RISCV_TARGET 64)
+    set(XMRIG_RISCV ON)
+    add_definitions(-DXMRIG_RISCV)
+    message(STATUS "Detected RISC-V 64-bit architecture (${CMAKE_SYSTEM_PROCESSOR})")
+elseif (CMAKE_SYSTEM_PROCESSOR MATCHES "^(riscv32|rv32)$")
+    set(RISCV_TARGET 32)
+    set(XMRIG_RISCV ON)
+    add_definitions(-DXMRIG_RISCV)
+    message(STATUS "Detected RISC-V 32-bit architecture (${CMAKE_SYSTEM_PROCESSOR})")
+endif()
+
 if (XMRIG_64_BIT AND CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|AMD64)$")
     add_definitions(-DRAPIDJSON_SSE2)
 else()
@@ -29,6 +42,47 @@ else()
     set(WITH_VAES OFF)
 endif()
 
+# Disable x86-specific features for RISC-V
+if (XMRIG_RISCV)
+    set(WITH_SSE4_1 OFF)
+    set(WITH_AVX2 OFF)
+    set(WITH_VAES OFF)
+
+    # default build uses the RV64GC baseline
+    set(RVARCH "rv64gc")
+
+    # for native builds, enable Zba and Zbb if supported by the CPU
+    if(ARCH STREQUAL "native")
+        enable_language(ASM)
+
+        try_run(RANDOMX_ZBA_RUN_FAIL
+            RANDOMX_ZBA_COMPILE_OK
+            ${CMAKE_CURRENT_BINARY_DIR}/
+            ${CMAKE_CURRENT_SOURCE_DIR}/src/crypto/randomx/tests/riscv64_zba.s
+            COMPILE_DEFINITIONS "-march=rv64gc_zba")
+
+        if (RANDOMX_ZBA_COMPILE_OK AND NOT RANDOMX_ZBA_RUN_FAIL)
+            set(RVARCH "${RVARCH}_zba")
+            message(STATUS "RISC-V zba extension detected")
+        endif()
+
+        try_run(RANDOMX_ZBB_RUN_FAIL
+            RANDOMX_ZBB_COMPILE_OK
+            ${CMAKE_CURRENT_BINARY_DIR}/
+            ${CMAKE_CURRENT_SOURCE_DIR}/src/crypto/randomx/tests/riscv64_zbb.s
+            COMPILE_DEFINITIONS "-march=rv64gc_zbb")
+
+        if (RANDOMX_ZBB_COMPILE_OK AND NOT RANDOMX_ZBB_RUN_FAIL)
+            set(RVARCH "${RVARCH}_zbb")
+            message(STATUS "RISC-V zbb extension detected")
+        endif()
+    endif()
+
+    message(STATUS "Using -march=${RVARCH}")
+endif()
+
+add_definitions(-DRAPIDJSON_WRITE_DEFAULT_FLAGS=6) # rapidjson::kWriteNanAndInfFlag | rapidjson::kWriteNanAndInfNullFlag
+
 if (ARM_V8)
     set(ARM_TARGET 8)
 elseif (ARM_V7)
@@ -36,9 +90,9 @@ elseif (ARM_V7)
 endif()
 
 if (NOT ARM_TARGET)
-    if (CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64|arm64|armv8-a)$")
+    if (CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64|arm64|ARM64|armv8-a)$")
         set(ARM_TARGET 8)
-    elseif (CMAKE_SYSTEM_PROCESSOR MATCHES "^(armv7|armv7f|armv7s|armv7k|armv7-a|armv7l|armv7ve)$")
+    elseif (CMAKE_SYSTEM_PROCESSOR MATCHES "^(armv7|armv7f|armv7s|armv7k|armv7-a|armv7l|armv7ve|armv8l)$")
         set(ARM_TARGET 7)
     endif()
 endif()
@@ -49,7 +103,7 @@ if (ARM_TARGET AND ARM_TARGET GREATER 6)
 
     message(STATUS "Use ARM_TARGET=${ARM_TARGET} (${CMAKE_SYSTEM_PROCESSOR})")
 
-    if (ARM_TARGET EQUAL 8)
+    if (ARM_TARGET EQUAL 8 AND (CMAKE_CXX_COMPILER_ID MATCHES GNU OR CMAKE_CXX_COMPILER_ID MATCHES Clang))
         CHECK_CXX_COMPILER_FLAG(-march=armv8-a+crypto XMRIG_ARM_CRYPTO)
 
         if (XMRIG_ARM_CRYPTO)

cmake/flags.cmake

@@ -26,8 +26,13 @@ if (CMAKE_CXX_COMPILER_ID MATCHES GNU)
         set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${ARM8_CXX_FLAGS}")
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${ARM8_CXX_FLAGS} -flax-vector-conversions")
     elseif (ARM_TARGET EQUAL 7)
-        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=armv7-a -mfpu=neon")
+        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=armv7-a -mfpu=neon -flax-vector-conversions")
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=armv7-a -mfpu=neon -flax-vector-conversions")
+    elseif (XMRIG_RISCV)
+        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=${RVARCH}")
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=${RVARCH}")
+        
+        add_definitions(-DHAVE_ROTR)
     else()
         set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -maes")
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -maes")
@@ -41,6 +46,8 @@ if (CMAKE_CXX_COMPILER_ID MATCHES GNU)
         else()
             set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static -Wl,--large-address-aware")
         endif()
+    elseif(CMAKE_SYSTEM_NAME STREQUAL "Haiku")
+        set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static-libgcc")
     else()
         set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static-libgcc -static-libstdc++")
     endif()
@@ -63,10 +70,10 @@ elseif (CMAKE_CXX_COMPILER_ID MATCHES MSVC)
 elseif (CMAKE_CXX_COMPILER_ID MATCHES Clang)
 
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall")
-    set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} -Ofast -funroll-loops -fmerge-all-constants")
+    set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} -funroll-loops -fmerge-all-constants")
 
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -fexceptions -fno-rtti -Wno-missing-braces")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -fexceptions -fno-rtti")
-    set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast -funroll-loops -fmerge-all-constants")
+    set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -funroll-loops -fmerge-all-constants")
 
     if (ARM_TARGET EQUAL 8)
         set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${ARM8_CXX_FLAGS}")
@@ -74,6 +81,11 @@ elseif (CMAKE_CXX_COMPILER_ID MATCHES Clang)
     elseif (ARM_TARGET EQUAL 7)
         set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mfpu=neon -march=${CMAKE_SYSTEM_PROCESSOR}")
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mfpu=neon -march=${CMAKE_SYSTEM_PROCESSOR}")
+    elseif (XMRIG_RISCV)
+        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=${RVARCH}")
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=${RVARCH}")
+        
+        add_definitions(-DHAVE_ROTR)
     else()
         set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -maes")
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -maes")
@@ -84,10 +96,9 @@ elseif (CMAKE_CXX_COMPILER_ID MATCHES Clang)
         endif()
     endif()
 
-    if (BUILD_STATIC)
+    if ((WIN32 AND ARM_TARGET) OR BUILD_STATIC)
         set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static")
     endif()
-
 endif()
 
 if (NOT WIN32)

cmake/os.cmake

@@ -17,10 +17,13 @@ else()
         set(XMRIG_OS_LINUX ON)
     elseif(CMAKE_SYSTEM_NAME STREQUAL FreeBSD OR CMAKE_SYSTEM_NAME STREQUAL DragonFly)
         set(XMRIG_OS_FREEBSD ON)
+    elseif(CMAKE_SYSTEM_NAME STREQUAL OpenBSD)
+        set(XMRIG_OS_OPENBSD ON)
+    elseif(CMAKE_SYSTEM_NAME STREQUAL "Haiku")
+        set(XMRIG_OS_HAIKU ON)
     endif()
 endif()
 
-
 if (XMRIG_OS_WIN)
     add_definitions(-DWIN32 -DXMRIG_OS_WIN)
 elseif(XMRIG_OS_APPLE)
@@ -44,6 +47,10 @@ elseif(XMRIG_OS_UNIX)
         add_definitions(-DXMRIG_OS_LINUX)
     elseif (XMRIG_OS_FREEBSD)
         add_definitions(-DXMRIG_OS_FREEBSD)
+    elseif (XMRIG_OS_OPENBSD)
+        add_definitions(-DXMRIG_OS_OPENBSD)
+    elseif (XMRIG_OS_HAIKU)
+        add_definitions(-DXMRIG_OS_HAIKU)
     endif()
 endif()
 

cmake/randomx.cmake

@@ -1,4 +1,18 @@
 if (WITH_RANDOMX)
+    include(CheckSymbolExists)
+
+    if (WIN32)
+        check_symbol_exists(_aligned_malloc "stdlib.h" HAVE_ALIGNED_MALLOC)
+        if (HAVE_ALIGNED_MALLOC)
+            add_compile_definitions(HAVE_ALIGNED_MALLOC)
+        endif()
+    else()
+        check_symbol_exists(posix_memalign "stdlib.h" HAVE_POSIX_MEMALIGN)
+        if (HAVE_POSIX_MEMALIGN)
+            add_compile_definitions(HAVE_POSIX_MEMALIGN)
+        endif()
+    endif()
+
     add_definitions(/DXMRIG_ALGO_RANDOMX)
     set(WITH_ARGON2 ON)
 
@@ -48,7 +62,7 @@ if (WITH_RANDOMX)
              src/crypto/randomx/jit_compiler_x86_static.asm
              src/crypto/randomx/jit_compiler_x86.cpp
             )
-    elseif (WITH_ASM AND NOT XMRIG_ARM AND CMAKE_SIZEOF_VOID_P EQUAL 8)
+    elseif (WITH_ASM AND NOT XMRIG_ARM AND NOT XMRIG_RISCV AND CMAKE_SIZEOF_VOID_P EQUAL 8)
         list(APPEND SOURCES_CRYPTO
              src/crypto/randomx/jit_compiler_x86_static.S
              src/crypto/randomx/jit_compiler_x86.cpp
@@ -66,6 +80,13 @@ if (WITH_RANDOMX)
         else()
             set_property(SOURCE src/crypto/randomx/jit_compiler_a64_static.S PROPERTY LANGUAGE C)
         endif()
+    elseif (XMRIG_RISCV AND CMAKE_SIZEOF_VOID_P EQUAL 8)
+        list(APPEND SOURCES_CRYPTO
+             src/crypto/randomx/jit_compiler_rv64_static.S
+             src/crypto/randomx/jit_compiler_rv64.cpp
+            )
+        # cheat because cmake and ccache hate each other
+        set_property(SOURCE src/crypto/randomx/jit_compiler_rv64_static.S PROPERTY LANGUAGE C)
     else()
         list(APPEND SOURCES_CRYPTO
              src/crypto/randomx/jit_compiler_fallback.cpp
@@ -102,7 +123,7 @@ if (WITH_RANDOMX)
             )
     endif()
 
-    if (WITH_MSR AND NOT XMRIG_ARM AND CMAKE_SIZEOF_VOID_P EQUAL 8 AND (XMRIG_OS_WIN OR XMRIG_OS_LINUX))
+    if (WITH_MSR AND NOT XMRIG_ARM AND NOT XMRIG_RISCV AND CMAKE_SIZEOF_VOID_P EQUAL 8 AND (XMRIG_OS_WIN OR XMRIG_OS_LINUX))
         add_definitions(/DXMRIG_FEATURE_MSR)
         add_definitions(/DXMRIG_FIX_RYZEN)
         message("-- WITH_MSR=ON")

doc/ALGORITHMS.md

@@ -13,7 +13,6 @@ Option `coin` useful for pools without [algorithm negotiation](https://xmrig.com
 | Name | Memory | Version | Description | Notes |
 |------|--------|---------|-------------|-------|
 | `kawpow` | - | 6.0.0+ | KawPow (Ravencoin) | GPU only |
-| `rx/keva` | 1 MB | 5.9.0+ | RandomKEVA (RandomX variant for Keva). |  |
 | `astrobwt` | 20 MB | 5.8.0+ | AstroBWT (Dero). |  |
 | `cn-pico/tlo` | 256 KB | 5.5.0+ | CryptoNight-Pico (Talleo). |  |
 | `rx/sfx` | 2 MB | 5.4.0+ | RandomSFX (RandomX variant for Safex). |  |

doc/CHANGELOG_OLD.md

@@ -256,7 +256,7 @@
 
 # v2.8.0
 - **[#753](https://github.com/xmrig/xmrig/issues/753) Added new algorithm [CryptoNight variant 2](https://github.com/xmrig/xmrig/issues/753) for Monero fork, thanks [@SChernykh](https://github.com/SChernykh).**
-  - Added global and per thread option `"asm"` and and command line equivalent.
+  - Added global and per thread option `"asm"` and command line equivalent.
 - **[#758](https://github.com/xmrig/xmrig/issues/758) Added SSL/TLS support for secure connections to pools.**
   - Added per pool options `"tls"` and `"tls-fingerprint"` and command line equivalents.
 - [#767](https://github.com/xmrig/xmrig/issues/767) Added config autosave feature, same with GPU miners.  

doc/RISCV_PERF_TUNING.md

@@ -0,0 +1,365 @@
+# RISC-V Performance Optimization Guide
+
+This guide provides comprehensive instructions for optimizing XMRig on RISC-V architectures.
+
+## Build Optimizations
+
+### Compiler Flags Applied Automatically
+
+The CMake build now applies aggressive RISC-V-specific optimizations:
+
+```cmake
+# RISC-V ISA with extensions
+-march=rv64gcv_zba_zbb_zbc_zbs
+
+# Aggressive compiler optimizations
+-funroll-loops              # Unroll loops for ILP (instruction-level parallelism)
+-fomit-frame-pointer        # Free up frame pointer register (RISC-V has limited registers)
+-fno-common                 # Better code generation for global variables
+-finline-functions          # Inline more functions for better cache locality
+-ffast-math                 # Relaxed FP semantics (safe for mining)
+-flto                       # Link-time optimization for cross-module inlining
+
+# Release build additions
+-minline-atomics            # Inline atomic operations for faster synchronization
+```
+
+### Optimal Build Command
+
+```bash
+mkdir build && cd build
+cmake -DCMAKE_BUILD_TYPE=Release ..
+make -j$(nproc)
+```
+
+**Expected build time**: 5-15 minutes depending on CPU
+
+## Runtime Optimizations
+
+### 1. Memory Configuration (Most Important)
+
+Enable huge pages to reduce TLB misses and fragmentation:
+
+#### Enable 2MB Huge Pages
+```bash
+# Calculate required huge pages (1 page = 2MB)
+# For 2 GB dataset: 1024 pages
+# For cache + dataset: 1536 pages minimum
+sudo sysctl -w vm.nr_hugepages=2048
+```
+
+Verify:
+```bash
+grep HugePages /proc/meminfo
+# Expected: HugePages_Free should be close to nr_hugepages
+```
+
+#### Enable 1GB Huge Pages (Optional but Recommended)
+
+```bash
+# Run provided helper script
+sudo ./scripts/enable_1gb_pages.sh
+
+# Verify 1GB pages are available
+cat /sys/kernel/mm/hugepages/hugepages-1048576kB/nr_hugepages
+# Should be: >= 1 (one 1GB page)
+```
+
+Update config.json:
+```json
+{
+    "cpu": {
+        "huge-pages": true
+    },
+    "randomx": {
+        "1gb-pages": true
+    }
+}
+```
+
+### 2. RandomX Mode Selection
+
+| Mode | Memory | Init Time | Throughput | Recommendation |
+|------|--------|-----------|-----------|-----------------|
+| **light** | 256 MB | 10 sec | Low | Testing, resource-constrained |
+| **fast** | 2 GB | 2-5 min* | High | Production (with huge pages) |
+| **auto** | 2 GB | Varies | High | Default (uses fast if possible) |
+
+*With optimizations; can be 30+ minutes without huge pages
+
+**For RISC-V, use fast mode with huge pages enabled.**
+
+### 3. Dataset Initialization Threads
+
+Optimal thread count = 60-75% of CPU cores (leaves headroom for OS/other tasks)
+
+```json
+{
+    "randomx": {
+        "init": 4
+    }
+}
+```
+
+Or auto-detect (rewritten for RISC-V):
+```json
+{
+    "randomx": {
+        "init": -1
+    }
+}
+```
+
+### 4. CPU Affinity (Optional)
+
+Pin threads to specific cores for better cache locality:
+
+```json
+{
+    "cpu": {
+        "rx/0": [
+            { "threads": 1, "affinity": 0 },
+            { "threads": 1, "affinity": 1 },
+            { "threads": 1, "affinity": 2 },
+            { "threads": 1, "affinity": 3 }
+        ]
+    }
+}
+```
+
+### 5. CPU Governor (Linux)
+
+Set to performance mode for maximum throughput:
+
+```bash
+# Check current governor
+cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
+
+# Set to performance (requires root)
+echo performance | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
+
+# Verify
+cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
+# Should output: performance
+```
+
+## Configuration Examples
+
+### Minimum (Testing)
+```json
+{
+    "randomx": {
+        "mode": "light"
+    },
+    "cpu": {
+        "huge-pages": false
+    }
+}
+```
+
+### Recommended (Balanced)
+```json
+{
+    "randomx": {
+        "mode": "auto",
+        "init": 4,
+        "1gb-pages": true
+    },
+    "cpu": {
+        "huge-pages": true,
+        "priority": 2
+    }
+}
+```
+
+### Maximum Performance (Production)
+```json
+{
+    "randomx": {
+        "mode": "fast",
+        "init": -1,
+        "1gb-pages": true,
+        "scratchpad_prefetch_mode": 1
+    },
+    "cpu": {
+        "huge-pages": true,
+        "priority": 3,
+        "yield": false
+    }
+}
+```
+
+## CLI Equivalents
+
+```bash
+# Light mode
+./xmrig --randomx-mode=light
+
+# Fast mode with 4 init threads
+./xmrig --randomx-mode=fast --randomx-init=4
+
+# Benchmark
+./xmrig --bench=1M --algo=rx/0
+
+# Benchmark Wownero variant (1 MB scratchpad)
+./xmrig --bench=1M --algo=rx/wow
+
+# Mine to pool
+./xmrig -o pool.example.com:3333 -u YOUR_WALLET -p x
+```
+
+## Performance Diagnostics
+
+### Check if Vector Extensions are Detected
+
+Look for `FEATURES:` line in output:
+```
+ * CPU:       ky,x60 (uarch ky,x1)
+ * FEATURES:  rv64imafdcv zba zbb zbc zbs
+```
+
+- `v`: Vector extension (RVV) ✓
+- `zba`, `zbb`, `zbc`, `zbs`: Bit manipulation ✓
+- If missing, make sure build used `-march=rv64gcv_zba_zbb_zbc_zbs`
+
+### Verify Huge Pages at Runtime
+
+```bash
+# Run xmrig with --bench=1M and check output
+./xmrig --bench=1M
+
+# Look for line like:
+# HUGE PAGES   100%  1 / 1 (1024 MB)
+```
+
+- Should show 100% for dataset AND threads
+- If less, increase `vm.nr_hugepages` and reboot
+
+### Monitor Performance
+
+```bash
+# Run benchmark multiple times to find stable hashrate
+./xmrig --bench=1M --algo=rx/0
+./xmrig --bench=10M --algo=rx/0
+./xmrig --bench=100M --algo=rx/0
+
+# Check system load and memory during mining
+while true; do free -h; grep HugePages /proc/meminfo; sleep 2; done
+```
+
+## Expected Performance
+
+### Hardware: Orange Pi RV2 (Ky X1, 8 cores @ ~1.5 GHz)
+
+| Config | Mode | Hashrate | Init Time |
+|--------|------|----------|-----------|
+| Scalar (baseline) | fast | 30 H/s | 10 min |
+| Scalar + huge pages | fast | 33 H/s | 2 min |
+| RVV (if enabled) | fast | 70-100 H/s | 3 min |
+
+*Actual results depend on CPU frequency, memory speed, and load*
+
+## Troubleshooting
+
+### Long Initialization Times (30+ minutes)
+
+**Cause**: Huge pages not enabled, system using swap
+**Solution**:
+1. Enable huge pages: `sudo sysctl -w vm.nr_hugepages=2048`
+2. Reboot: `sudo reboot`
+3. Reduce mining threads to free memory
+4. Check available memory: `free -h`
+
+### Low Hashrate (50% of expected)
+
+**Cause**: CPU governor set to power-save, no huge pages, high contention
+**Solution**:
+1. Set governor to performance: `echo performance | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor`
+2. Enable huge pages
+3. Reduce number of mining threads
+4. Check system load: `top` or `htop`
+
+### Dataset Init Crashes or Hangs
+
+**Cause**: Insufficient memory, corrupted huge pages
+**Solution**:
+1. Disable huge pages temporarily: set `huge-pages: false` in config
+2. Reduce mining threads
+3. Reboot and re-enable huge pages
+4. Try light mode: `--randomx-mode=light`
+
+### Out of Memory During Benchmark
+
+**Cause**: Not enough RAM for dataset + cache + threads
+**Solution**:
+1. Use light mode: `--randomx-mode=light`
+2. Reduce mining threads: `--threads=1`
+3. Increase available memory (kill other processes)
+4. Check: `free -h` before mining
+
+## Advanced Tuning
+
+### Vector Length (VLEN) Detection
+
+RISC-V vector extension variable length (VLEN) affects performance:
+
+```bash
+# Check VLEN on your CPU
+cat /proc/cpuinfo | grep vlen
+
+# Expected values:
+# - 128 bits (16 bytes) = minimum
+# - 256 bits (32 bytes) = common
+# - 512 bits (64 bytes) = high performance
+```
+
+Larger VLEN generally means better performance for vectorized operations.
+
+### Prefetch Optimization
+
+The code automatically optimizes memory prefetching for RISC-V:
+
+```
+scratchpad_prefetch_mode: 0 = disabled (slowest)
+scratchpad_prefetch_mode: 1 = prefetch.r (default, recommended)
+scratchpad_prefetch_mode: 2 = prefetch.w (experimental)
+```
+
+### Memory Bandwidth Saturation
+
+If experiencing memory bandwidth saturation (high latency):
+
+1. Reduce mining threads
+2. Increase L2/L3 cache by mining fewer threads per core
+3. Enable cache QoS (AMD Ryzen): `cache_qos: true`
+
+## Building with Custom Flags
+
+To build with custom RISC-V flags:
+
+```bash
+mkdir build && cd build
+cmake -DCMAKE_BUILD_TYPE=Release \
+      -DCMAKE_C_FLAGS="-march=rv64gcv_zba_zbb_zbc_zbs -O3 -funroll-loops -fomit-frame-pointer" \
+      ..
+make -j$(nproc)
+```
+
+## Future Optimizations
+
+- [ ] Zbk* (crypto) support detection and usage
+- [ ] Optimal VLEN-aware algorithm selection
+- [ ] Per-core memory affinity (NUMA support)
+- [ ] Dynamic thread count adjustment based on thermals
+- [ ] Cross-compile optimizations for various RISC-V cores
+
+## References
+
+- [RISC-V Vector Extension Spec](https://github.com/riscv/riscv-v-spec)
+- [RISC-V Bit Manipulation Spec](https://github.com/riscv/riscv-bitmanip)
+- [RISC-V Crypto Spec](https://github.com/riscv/riscv-crypto)
+- [XMRig Documentation](https://xmrig.com/docs)
+
+---
+
+For further optimization, enable RVV intrinsics by replacing `sse2rvv.h` with `sse2rvv_optimized.h` in the build.

scripts/build.hwloc.sh

@@ -1,8 +1,8 @@
-#!/bin/bash -e
+#!/bin/sh -e
 
 HWLOC_VERSION_MAJOR="2"
-HWLOC_VERSION_MINOR="9"
+HWLOC_VERSION_MINOR="12"
-HWLOC_VERSION_PATCH="0"
+HWLOC_VERSION_PATCH="1"
 
 HWLOC_VERSION="${HWLOC_VERSION_MAJOR}.${HWLOC_VERSION_MINOR}.${HWLOC_VERSION_PATCH}"
 

scripts/build.hwloc1.sh

@@ -1,4 +1,4 @@
-#!/bin/bash -e
+#!/bin/sh -e
 
 HWLOC_VERSION="1.11.13"
 

scripts/build.libressl.sh

@@ -1,4 +1,4 @@
-#!/bin/bash -e
+#!/bin/sh -e
 
 LIBRESSL_VERSION="3.5.2"
 

scripts/build.openssl.sh

@@ -1,6 +1,6 @@
-#!/bin/bash -e
+#!/bin/sh -e
 
-OPENSSL_VERSION="1.1.1s"
+OPENSSL_VERSION="1.1.1u"
 
 mkdir -p deps
 mkdir -p deps/include
@@ -8,7 +8,7 @@ mkdir -p deps/lib
 
 mkdir -p build && cd build
 
-wget https://www.openssl.org/source/openssl-${OPENSSL_VERSION}.tar.gz -O openssl-${OPENSSL_VERSION}.tar.gz
+wget https://openssl.org/source/old/1.1.1/openssl-${OPENSSL_VERSION}.tar.gz -O openssl-${OPENSSL_VERSION}.tar.gz
 tar -xzf openssl-${OPENSSL_VERSION}.tar.gz
 
 cd openssl-${OPENSSL_VERSION}

scripts/build.openssl3.sh

@@ -1,6 +1,6 @@
-#!/bin/bash -e
+#!/bin/sh -e
 
-OPENSSL_VERSION="3.0.7"
+OPENSSL_VERSION="3.0.16"
 
 mkdir -p deps
 mkdir -p deps/include
@@ -8,7 +8,7 @@ mkdir -p deps/lib
 
 mkdir -p build && cd build
 
-wget https://www.openssl.org/source/openssl-${OPENSSL_VERSION}.tar.gz -O openssl-${OPENSSL_VERSION}.tar.gz
+wget https://github.com/openssl/openssl/releases/download/openssl-${OPENSSL_VERSION}/openssl-${OPENSSL_VERSION}.tar.gz -O openssl-${OPENSSL_VERSION}.tar.gz
 tar -xzf openssl-${OPENSSL_VERSION}.tar.gz
 
 cd openssl-${OPENSSL_VERSION}

scripts/build.uv.sh

@@ -1,6 +1,6 @@
-#!/bin/bash -e
+#!/bin/sh -e
 
-UV_VERSION="1.44.2"
+UV_VERSION="1.51.0"
 
 mkdir -p deps
 mkdir -p deps/include
@@ -8,10 +8,10 @@ mkdir -p deps/lib
 
 mkdir -p build && cd build
 
-wget https://github.com/libuv/libuv/archive/v${UV_VERSION}.tar.gz -O v${UV_VERSION}.tar.gz
+wget https://dist.libuv.org/dist/v${UV_VERSION}/libuv-v${UV_VERSION}.tar.gz -O v${UV_VERSION}.tar.gz
 tar -xzf v${UV_VERSION}.tar.gz
 
-cd libuv-${UV_VERSION}
+cd libuv-v${UV_VERSION}
 sh autogen.sh
 ./configure --disable-shared
 make -j$(nproc || sysctl -n hw.ncpu || sysctl -n hw.logicalcpu)

scripts/build_deps.sh

@@ -1,5 +1,5 @@
-#!/bin/bash -e
+#!/bin/sh -e
 
 ./build.uv.sh
 ./build.hwloc.sh
-./build.openssl.sh
+./build.openssl3.sh

scripts/enable_1gb_pages.sh

@@ -1,4 +1,4 @@
-#!/bin/bash -e
+#!/bin/sh -e
 
 # https://xmrig.com/docs/miner/hugepages#onegb-huge-pages
 

scripts/generate_cl.js

@@ -6,7 +6,6 @@ const fs = require('fs');
 const path = require('path');
 const { text2h, text2h_bundle, addIncludes } = require('./js/opencl');
 const { opencl_minify } = require('./js/opencl_minify');
-const cwd = process.cwd();
 
 
 function cn()
@@ -50,7 +49,6 @@ function rx()
         'randomx_constants_monero.h',
         'randomx_constants_wow.h',
         'randomx_constants_arqma.h',
-        'randomx_constants_keva.h',
         'randomx_constants_graft.h',
         'aes.cl',
         'blake2b.cl',
@@ -77,18 +75,24 @@ function kawpow()
     fs.writeFileSync('kawpow_dag_cl.h', text2h(kawpow_dag, 'xmrig', 'kawpow_dag_cl'));
 }
 
+for (let i = 0; i < 2; i++) {
+    if (fs.existsSync('src/backend/opencl/cl/OclSource.h')) {
+        break;
+    }
 
-process.chdir(path.resolve('src/backend/opencl/cl/cn'));
+    process.chdir('..');
+}
 
+process.chdir(path.resolve('src/backend/opencl/cl'));
+
+const cwd = process.cwd();
+
+process.chdir(path.resolve(cwd, 'cn'));
 cn();
 cn_r();
 
-process.chdir(cwd);
+process.chdir(path.resolve(cwd, 'rx'));
-process.chdir(path.resolve('src/backend/opencl/cl/rx'));
-
 rx();
 
-process.chdir(cwd);
+process.chdir(path.resolve(cwd, 'kawpow'));
-process.chdir(path.resolve('src/backend/opencl/cl/kawpow'));
-
 kawpow();

scripts/randomx_boost.sh

@@ -8,11 +8,11 @@ else
 	modprobe msr allow_writes=on
 fi
 
-if grep -E 'AMD Ryzen|AMD EPYC' /proc/cpuinfo > /dev/null;
+if grep -E 'AMD Ryzen|AMD EPYC|AuthenticAMD' /proc/cpuinfo > /dev/null;
 	then
 	if grep "cpu family[[:space:]]\{1,\}:[[:space:]]25" /proc/cpuinfo > /dev/null;
 		then
-			if grep "model[[:space:]]\{1,\}:[[:space:]]97" /proc/cpuinfo > /dev/null;
+			if grep "model[[:space:]]\{1,\}:[[:space:]]\(97\|117\)" /proc/cpuinfo > /dev/null;
 				then
 					echo "Detected Zen4 CPU"
 					wrmsr -a 0xc0011020 0x4400000000000
@@ -28,6 +28,14 @@ if grep -E 'AMD Ryzen|AMD EPYC' /proc/cpuinfo > /dev/null;
 					wrmsr -a 0xc001102b 0x2000cc10
 					echo "MSR register values for Zen3 applied"
 				fi
+		elif grep "cpu family[[:space:]]\{1,\}:[[:space:]]26" /proc/cpuinfo > /dev/null;
+			then
+				echo "Detected Zen5 CPU"
+				wrmsr -a 0xc0011020 0x4400000000000
+				wrmsr -a 0xc0011021 0x4000000000040
+				wrmsr -a 0xc0011022 0x8680000401570000
+				wrmsr -a 0xc001102b 0x2040cc10
+				echo "MSR register values for Zen5 applied"
 		else
 			echo "Detected Zen1/Zen2 CPU"
 			wrmsr -a 0xc0011020 0

src/3rdparty/argon2/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.1)
+cmake_minimum_required(VERSION 3.10)
 
 project(argon2 C)
 set(CMAKE_C_STANDARD 99)
@@ -35,7 +35,7 @@ if (CMAKE_C_COMPILER_ID MATCHES MSVC)
     add_feature_impl(xop     ""              HAVE_XOP)
     add_feature_impl(avx2    "/arch:AVX2"    HAVE_AVX2)
     add_feature_impl(avx512f "/arch:AVX512F" HAVE_AVX512F)
-elseif (NOT XMRIG_ARM AND CMAKE_SIZEOF_VOID_P EQUAL 8)
+elseif (NOT XMRIG_ARM AND NOT XMRIG_RISCV AND CMAKE_SIZEOF_VOID_P EQUAL 8)
     function(add_feature_impl FEATURE GCC_FLAG DEF)
         add_library(argon2-${FEATURE} STATIC arch/x86_64/lib/argon2-${FEATURE}.c)
         target_include_directories(argon2-${FEATURE} PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/../../)

src/3rdparty/epee/README.md

@@ -1 +1 @@
-epee -  is a small library of helpers, wrappers, tools and and so on, used to make my life easier.
+epee -  is a small library of helpers, wrappers, tools and so on, used to make my life easier.

src/3rdparty/hwloc/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.1)
+cmake_minimum_required(VERSION 3.10)
 project (hwloc C)
 
 include_directories(include)

src/3rdparty/hwloc/NEWS

@@ -1,5 +1,5 @@
 Copyright © 2009 CNRS
-Copyright © 2009-2022 Inria.  All rights reserved.
+Copyright © 2009-2025 Inria.  All rights reserved.
 Copyright © 2009-2013 Université Bordeaux
 Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
 Copyright © 2020 Hewlett Packard Enterprise.  All rights reserved.
@@ -17,6 +17,214 @@ bug fixes (and other actions) for each version of hwloc since version
 0.9.
 
 
+Version 2.12.1
+--------------
+* Add hwloc-calc's --default-nodes option to hwloc-bind and hwloc-info.
+* Improve the --best-memattr "default" fallback, try to use "default"
+  memory nodes, and add verbose messages and warnings if some
+  performance info are incomplete or missing.
+  Thanks to Antoine Morvan for the report.
+* Fix CPU and memory binding on different locations,
+  thanks to Antoine Morvan for the report.
+* Add HWLOC_LOCAL_NUMANODE_FLAG_INTERSECT_LOCALITY and enable it by
+  default in hwloc-calc --local-memory for finding local NUMA nodes
+  that do not exactly match input locations.
+  Thanks to Antoine Morvan for the report.
+* Fix a possible crash in the x86 backend when Qemu is configured to
+  expose multicore/thread CPUs that are actually single-core/thread.
+  Thanks to Georg Pfuetzenreuter.
+
+
+Version 2.12.0
+--------------
+* Add hwloc_topology_get_default_nodeset() for the set of default
+  NUMA nodes.
+  - hwloc-calc now has --default-nodes option.
+* Rework oneAPI LevelZero support to use zesInit() and avoid the need
+  to set ZES_ENABLE_SYSMAN=1 in the environment.
+  - zesDriverGetDeviceByUuidExp() is now required in the L0 runtime.
+  - ZES/Sysman variants were added in hwloc/levelzero.h to specifically
+    handle ZES/Sysman device handles.
+* Fix the locality of AMD GPU partitions, thanks to Edgar Leon for
+  reporting and debugging the issue.
+* Better detect Cray Slingshot NICs, thanks to Edgar Leon.
+* Add support for Die objects and Module groups on Windows.
+* Only filter-out Dies that are identical to their Packages
+  when it applies to all Dies.
+* Improve hwloc-calc to handle CPU-less NUMA nodes or platforms with
+  heterogeneous memory without requiring --nodeset-output.
+* hwloc-calc now accepts counting/listing cpukinds and memory tiers
+  with -N and -I cpukind/memorytier.
+* The systemd-dbus-api output of hwloc-calc has changed, and
+  --nodeset-output-format was added, to support NUMA node outputs.
+  Thanks to Pierre Neyron.
+* Update NVLink bandwidth and CUDA capabilities up to NVIDIA Blackwell.
+* Fix some NUMA syscalls on Linux for platforms with old libc headers.
+* Some minor fixes in distances.
+
+
+Version 2.11.2
+--------------
+* Add missing CPU info attrs on aarch64 on Linux.
+* Use ACPI CPPC on Linux to get better information about cpukinds,
+  at least on AMD CPUs.
+* Fix crash when manipulating cpukinds after topology
+  duplication, thanks to Hadrien Grasland for the report.
+* Fix missing input target checks in memattr functions,
+  thanks to Hadrien Grasland for the report.
+* Fix a memory leak when ignoring NUMA distances on FreeBSD.
+* Fix build failure on old Linux distributions without accessat().
+* Fix non-Windows importing of XML topologies and CPUID dumps exported
+  on Windows.
+* hwloc-calc --cpuset-output-format systemd-dbus-api now allows
+  to generate AllowedCPUs information for systemd slices.
+  See the hwloc-calc manpage for examples. Thanks to Pierre Neyron.
+* Some fixes in manpage EXAMPLES and split them into subsections.
+
+
+Version 2.11.1
+--------------
+* Fix bash completions, thanks Tavis Rudd.
+
+
+Version 2.11.0
+--------------
+* API
+  + Add HWLOC_MEMBIND_WEIGHTED_INTERLEAVE memory binding policy on
+    Linux 6.9+. Thanks to Honggyu Kim for the patch.
+    - weighted_interleave_membind is added to membind support bits.
+    - The "weighted" policy is added to the hwloc-bind tool.
+  + Add hwloc_obj_set_subtype(). Thanks to Hadrien Grasland for the report.
+* GPU support
+  + Don't hide the GPU NUMA node on NVIDIA Grace Hopper.
+  + Get Intel GPU OpenCL device locality.
+  + Add bandwidths between subdevices in the LevelZero XeLinkBandwidth
+    matrix.
+  + Fix PCI Gen4+ link speed of NVIDIA GPU obtained from NVML,
+    thanks to Akram Sbaih for the report.
+* Windows support
+  + Fix Windows support when UNICODE is enabled, several hwloc features
+    were missing, thanks to Martin for the report.
+  + Fix the enabling of CUDA in Windows CMake build,
+    Thanks to Moritz Kreutzer for the patch.
+  + Fix CUDA/OpenCL test source path in Windows CMake.
+* Tools
+  + Option --best-memattr may now return multiple nodes. Additional
+    configuration flags may be given to tweak its behavior.
+  + hwloc-info has a new --get-attr option to get a single attribute.
+  + hwloc-info now supports "levels", "support" and "topology"
+    special keywords for backward compatibility for hwloc 3.0.
+  + The --taskset command-line option is superseded by the new
+    --cpuset-output-format which also allows to export as list.
+  + hwloc-calc may now import bitmasks described as a list of bits
+    with the new "--cpuset-input-format list".
+* Misc
+  + The MemoryTiersNr info attribute in the root object now says how many
+    memory tiers were built. Thanks to Antoine Morvan for the report.
+  + Fix the management of infinite cpusets in the bitmap printf/sscanf
+    API as well as in command-line tools.
+  + Add section "Compiling software on top of hwloc's C API" in the
+    documentation with examples for GNU Make and CMake,
+    thanks to Florent Pruvost for the help.
+
+
+Version 2.10.0
+--------------
+* Heterogeneous Memory core improvements
+  + Better heuristics to identify the subtype of memory such as HBM,
+    DRAM, NVM, CXL-DRAM, etc.
+  + Build memory tiers, i.e. sets of NUMA nodes with the same subtype
+    and similar performance.
+    - NUMA node tier ranks are exposed in the new MemoryTier info
+      attribute (starts from 0 for highest bandwidth tier)..
+  + See the new Heterogeneous Memory section in the documentation.
+* API
+  + Add hwloc_topology_free_group_object() to discard a Group created
+    by hwloc_topology_alloc_group_object().
+* Linux backend
+  + Fix cpukinds on NVIDIA Grace to report identical cores even if they
+    actually have very small frequency differences.
+    Thanks to John C. Linford for the report.
+  + Add CXLDevice attributes to CXL DAX objects and NUMA nodes to show
+    which PCI device implements which window.
+  + Ignore buggy memory-side caches and memory attributes when fake NUMA
+    emulation is enabled on the Linux kernel command-line.
+  + Add more info attributes in MemoryModule Misc objects,
+    thanks to Zubiao Xiong for the patch.
+  + Get CPUModel and CPUFamily info attributes on LoongArch platforms.
+* x86 backend
+  + Add support for new AMD CPUID leaf 0x80000026 for better detection
+    of Core Complex and Die on Zen4 processors.
+  + Improve Zhaoxin CPU topology detection.
+* Tools
+  + Input locations and many command-line options (e.g. hwloc-calc -I -N -H,
+    lstopo --only) now accept filters such as "NUMA[HBM]" so that only
+    objects are that type and subtype are considered.
+    - NUMA[tier=1] is also accepted for selecting NUMA nodes depending
+      on their MemoryTier info attribute.
+  + Add --object-output to hwloc-calc to report the type as a prefix to
+    object indexes, e.g. Core:2 instead of 2 in the output of -I.
+  + hwloc-info --ancestor and --descendants now accepts kinds of objects
+    instead of single types.
+    - The new --first option only shows the first matching object.
+  + Add --children-of-pid to hwloc-ps to show a hierarchy of processes.
+    Thanks to Antoine Morvan for the suggestion.
+  + Add --misc-from to lstopo to add Misc objects described in a file.
+    - To be combined with the new hwloc-ps --lstopo-misc for a customizable
+      lstopo --top replacement.
+* Misc
+  + lstopo may now configure the layout of memory object placed above,
+    for instance with --children-order memory:above:vert.
+  + Fix XML import from memory or stdin when using libxml2 2.12.
+  + Fix installation failures when configuring with --target,
+    thanks to Clement Foyer for the patch.
+  + Fix support for 128bit pointer architectures.
+  + Remove Netloc.
+
+
+Version 2.9.3
+-------------
+* Handle Linux glibc allocation errors in binding routines (CVE-2022-47022).
+* Fix hwloc-calc when searching objects on heterogeneous memory platforms,
+  thanks to Antoine Morvan for the report.
+* Fix hwloc_get_next_child() when there are some memory-side caches.
+* Don't crash if the topology is empty because Linux cgroups are wrong.
+* Improve some hwloc-bind warnings in case of command-line parsing errors.
+* Many documentation improvements all over the place, including:
+  + hwloc_topology_restrict() and hwloc_topology_insert_group() may reorder
+    children, causing the logical indexes of objects to change.
+
+
+Version 2.9.2
+-------------
+* Don't forget L3i when defining filters for multiple levels of caches
+  with hwloc_topology_set_cache/icache_types_filter().
+* Fix object total_memory after hwloc_topology_insert_group_object().
+* Fix the (non-yet) exporting in synthetic description for complex memory
+  hierarchies with memory-side caches, etc.
+* Fix some default size attributes when building synthetic topologies.
+* Fix size units in hwloc-annotate.
+* Improve bitmap reallocation error management in many functions.
+* Documentation improvements:
+  + Better document return values of functions.
+  + Add "Error reporting" section (in hwloc.h and in the doxygen doc).
+  + Add FAQ entry "What may I disable to make hwloc faster?"
+  + Improve FAQ entries "Why is lstopo slow?" and
+    "I only need ..., why should I use hwloc?"
+  + Clarify how to deal with cpukinds in hwloc-calc and hwloc-bind
+    manpages.
+
+
+Version 2.9.1
+-------------
+* Don't forget to apply object type filters to "perflevel" caches detected
+  on recent Mac OS X releases, thanks to Michel Lesoinne for the report.
+* Fix a failed assertion in hwloc_topology_restrict() when some NUMA nodes
+  are removed because of HWLOC_RESTRICT_FLAG_REMOVE_CPULESS but no PUs are.
+  Thanks to Mark Grondona for reporting the issue.
+* Mark HPE Cray Slingshot NICs with subtype "Slingshot".
+
+
 Version 2.9.0
 -------------
 * Backends
@@ -61,6 +269,14 @@ Version 2.8.0
     file from the documentation.
 
 
+Version 2.7.2
+-------------
+* Fix a crash when LevelZero devices have multiple subdevices,
+  e.g. on PonteVecchio GPUs, thanks to Jonathan Peyton.
+* Fix a leak when importing cpukinds from XML,
+  thanks to Hui Zhou.
+
+
 Version 2.7.1
 -------------
 * Workaround crashes when virtual machines report incoherent x86 CPUID

src/3rdparty/hwloc/README

@@ -1,4 +1,8 @@
-Introduction
+This is a truncated and poorly-formatted version of the documentation main page.
+See https://www.open-mpi.org/projects/hwloc/doc/ for more.
+
+
+hwloc Overview
 
 The Hardware Locality (hwloc) software project aims at easing the process of
 discovering hardware resources in parallel architectures. It offers
@@ -8,66 +12,450 @@ high-performance computing (HPC) applications, but is also applicable to any
 project seeking to exploit code and/or data locality on modern computing
 platforms.
 
-hwloc is actually made of two subprojects distributed together:
+hwloc provides command line tools and a C API to obtain the hierarchical map of
+key computing elements within a node, such as: NUMA memory nodes, shared
+caches, processor packages, dies and cores, processing units (logical
+processors or "threads") and even I/O devices. hwloc also gathers various
+attributes such as cache and memory information, and is portable across a
+variety of different operating systems and platforms.
 
-  * The original hwloc project for describing the internals of computing nodes.
+hwloc primarily aims at helping high-performance computing (HPC) applications,
- It is described in details starting at section Hardware Locality (hwloc)
+but is also applicable to any project seeking to exploit code and/or data
- Introduction.
+locality on modern computing platforms.
-  * The network-oriented companion called netloc (Network Locality), described
- in details starting with section Network Locality (netloc).
 
-See also the Related pages tab above for links to other sections.
+hwloc supports the following operating systems:
 
-Netloc may be disabled, but the original hwloc cannot. Both hwloc and netloc
+  * Linux (with knowledge of cgroups and cpusets, memory targets/initiators,
-APIs are documented after these sections.
+ etc.) on all supported hardware, including Intel Xeon Phi, ScaleMP vSMP,
+ and NumaScale NumaConnect.
+  * Solaris (with support for processor sets and logical domains)
+  * AIX
+  * Darwin / OS X
+  * FreeBSD and its variants (such as kFreeBSD/GNU)
+  * NetBSD
+  * HP-UX
+  * Microsoft Windows
+  * IBM BlueGene/Q Compute Node Kernel (CNK)
 
-Installation
+Since it uses standard Operating System information, hwloc's support is mostly
+independant from the processor type (x86, powerpc, ...) and just relies on the
+Operating System support. The main exception is BSD operating systems (NetBSD,
+FreeBSD, etc.) because they do not provide support topology information, hence
+hwloc uses an x86-only CPUID-based backend (which can be used for other OSes
+too, see the Components and plugins section).
 
-hwloc (https://www.open-mpi.org/projects/hwloc/) is available under the BSD
+To check whether hwloc works on a particular machine, just try to build it and
-license. It is hosted as a sub-project of the overall Open MPI project (https:/
+run lstopo or lstopo-no-graphics. If some things do not look right (e.g. bogus
-/www.open-mpi.org/). Note that hwloc does not require any functionality from
+or missing cache information), see Questions and Bugs.
-Open MPI -- it is a wholly separate (and much smaller!) project and code base.
-It just happens to be hosted as part of the overall Open MPI project.
 
-Basic Installation
+hwloc only reports the number of processors on unsupported operating systems;
+no topology information is available.
 
-Installation is the fairly common GNU-based process:
+For development and debugging purposes, hwloc also offers the ability to work
+on "fake" topologies:
 
-shell$ ./configure --prefix=...
+  * Symmetrical tree of resources generated from a list of level arities, see
-shell$ make
+ Synthetic topologies.
-shell$ make install
+  * Remote machine simulation through the gathering of topology as XML files,
+ see Importing and exporting topologies from/to XML files.
 
-hwloc- and netloc-specific configure options and requirements are documented in
+hwloc can display the topology in a human-readable format, either in graphical
-sections hwloc Installation and Netloc Installation respectively.
+mode (X11), or by exporting in one of several different formats, including:
+plain text, LaTeX tikzpicture, PDF, PNG, and FIG (see Command-line Examples
+below). Note that some of the export formats require additional support
+libraries.
 
-Also note that if you install supplemental libraries in non-standard locations,
+hwloc offers a programming interface for manipulating topologies and objects.
-hwloc's configure script may not be able to find them without some help. You
+It also brings a powerful CPU bitmap API that is used to describe topology
-may need to specify additional CPPFLAGS, LDFLAGS, or PKG_CONFIG_PATH values on
+objects location on physical/logical processors. See the Programming Interface
-the configure command line.
+below. It may also be used to binding applications onto certain cores or memory
+nodes. Several utility programs are also provided to ease command-line
+manipulation of topology objects, binding of processes, and so on.
 
-For example, if libpciaccess was installed into /opt/pciaccess, hwloc's
+Bindings for several other languages are available from the project website.
-configure script may not find it be default. Try adding PKG_CONFIG_PATH to the
-./configure command line, like this:
 
-./configure PKG_CONFIG_PATH=/opt/pciaccess/lib/pkgconfig ...
+Command-line Examples
 
-Running the "lstopo" tool is a good way to check as a graphical output whether
+On a 4-package 2-core machine with hyper-threading, the lstopo tool may show
-hwloc properly detected the architecture of your node. Netloc command-line
+the following graphical output:
-tools can be used to display the network topology interconnecting your nodes.
 
-Installing from a Git clone
+[dudley]
 
-Additionally, the code can be directly cloned from Git:
+Here's the equivalent output in textual form:
 
-shell$ git clone https://github.com/open-mpi/hwloc.git
+Machine
-shell$ cd hwloc
+  NUMANode L#0 (P#0)
-shell$ ./autogen.sh
+  Package L#0 + L3 L#0 (4096KB)
+ L2 L#0 (1024KB) + L1 L#0 (16KB) + Core L#0
+   PU L#0 (P#0)
+   PU L#1 (P#8)
+ L2 L#1 (1024KB) + L1 L#1 (16KB) + Core L#1
+   PU L#2 (P#4)
+   PU L#3 (P#12)
+  Package L#1 + L3 L#1 (4096KB)
+ L2 L#2 (1024KB) + L1 L#2 (16KB) + Core L#2
+   PU L#4 (P#1)
+   PU L#5 (P#9)
+ L2 L#3 (1024KB) + L1 L#3 (16KB) + Core L#3
+   PU L#6 (P#5)
+   PU L#7 (P#13)
+  Package L#2 + L3 L#2 (4096KB)
+ L2 L#4 (1024KB) + L1 L#4 (16KB) + Core L#4
+   PU L#8 (P#2)
+   PU L#9 (P#10)
+ L2 L#5 (1024KB) + L1 L#5 (16KB) + Core L#5
+   PU L#10 (P#6)
+   PU L#11 (P#14)
+  Package L#3 + L3 L#3 (4096KB)
+ L2 L#6 (1024KB) + L1 L#6 (16KB) + Core L#6
+   PU L#12 (P#3)
+   PU L#13 (P#11)
+ L2 L#7 (1024KB) + L1 L#7 (16KB) + Core L#7
+   PU L#14 (P#7)
+   PU L#15 (P#15)
 
-Note that GNU Autoconf >=2.63, Automake >=1.11 and Libtool >=2.2.6 are required
+Note that there is also an equivalent output in XML that is meant for exporting
-when building from a Git clone.
+/importing topologies but it is hardly readable to human-beings (see Importing
+and exporting topologies from/to XML files for details).
 
-Nightly development snapshots are available on the web site, they can be
+On a 4-package 2-core Opteron NUMA machine (with two core cores disallowed by
-configured and built without any need for Git or GNU Autotools.
+the administrator), the lstopo tool may show the following graphical output
+(with --disallowed for displaying disallowed objects):
+
+[hagrid]
+
+Here's the equivalent output in textual form:
+
+Machine (32GB total)
+  Package L#0
+ NUMANode L#0 (P#0 8190MB)
+ L2 L#0 (1024KB) + L1 L#0 (64KB) + Core L#0 + PU L#0 (P#0)
+ L2 L#1 (1024KB) + L1 L#1 (64KB) + Core L#1 + PU L#1 (P#1)
+  Package L#1
+ NUMANode L#1 (P#1 8192MB)
+ L2 L#2 (1024KB) + L1 L#2 (64KB) + Core L#2 + PU L#2 (P#2)
+ L2 L#3 (1024KB) + L1 L#3 (64KB) + Core L#3 + PU L#3 (P#3)
+  Package L#2
+ NUMANode L#2 (P#2 8192MB)
+ L2 L#4 (1024KB) + L1 L#4 (64KB) + Core L#4 + PU L#4 (P#4)
+ L2 L#5 (1024KB) + L1 L#5 (64KB) + Core L#5 + PU L#5 (P#5)
+  Package L#3
+ NUMANode L#3 (P#3 8192MB)
+ L2 L#6 (1024KB) + L1 L#6 (64KB) + Core L#6 + PU L#6 (P#6)
+ L2 L#7 (1024KB) + L1 L#7 (64KB) + Core L#7 + PU L#7 (P#7)
+
+On a 2-package quad-core Xeon (pre-Nehalem, with 2 dual-core dies into each
+package):
+
+[emmett]
+
+Here's the same output in textual form:
+
+Machine (total 16GB)
+  NUMANode L#0 (P#0 16GB)
+  Package L#0
+ L2 L#0 (4096KB)
+   L1 L#0 (32KB) + Core L#0 + PU L#0 (P#0)
+   L1 L#1 (32KB) + Core L#1 + PU L#1 (P#4)
+ L2 L#1 (4096KB)
+   L1 L#2 (32KB) + Core L#2 + PU L#2 (P#2)
+   L1 L#3 (32KB) + Core L#3 + PU L#3 (P#6)
+  Package L#1
+ L2 L#2 (4096KB)
+   L1 L#4 (32KB) + Core L#4 + PU L#4 (P#1)
+   L1 L#5 (32KB) + Core L#5 + PU L#5 (P#5)
+ L2 L#3 (4096KB)
+   L1 L#6 (32KB) + Core L#6 + PU L#6 (P#3)
+   L1 L#7 (32KB) + Core L#7 + PU L#7 (P#7)
+
+Programming Interface
+
+The basic interface is available in hwloc.h. Some higher-level functions are
+available in hwloc/helper.h to reduce the need to manually manipulate objects
+and follow links between them. Documentation for all these is provided later in
+this document. Developers may also want to look at hwloc/inlines.h which
+contains the actual inline code of some hwloc.h routines, and at this document,
+which provides good higher-level topology traversal examples.
+
+To precisely define the vocabulary used by hwloc, a Terms and Definitions
+section is available and should probably be read first.
+
+Each hwloc object contains a cpuset describing the list of processing units
+that it contains. These bitmaps may be used for CPU binding and Memory binding.
+hwloc offers an extensive bitmap manipulation interface in hwloc/bitmap.h.
+
+Moreover, hwloc also comes with additional helpers for interoperability with
+several commonly used environments. See the Interoperability With Other
+Software section for details.
+
+The complete API documentation is available in a full set of HTML pages, man
+pages, and self-contained PDF files (formatted for both both US letter and A4
+formats) in the source tarball in doc/doxygen-doc/.
+
+NOTE: If you are building the documentation from a Git clone, you will need to
+have Doxygen and pdflatex installed -- the documentation will be built during
+the normal "make" process. The documentation is installed during "make install"
+to $prefix/share/doc/hwloc/ and your systems default man page tree (under
+$prefix, of course).
+
+Portability
+
+Operating System have varying support for CPU and memory binding, e.g. while
+some Operating Systems provide interfaces for all kinds of CPU and memory
+bindings, some others provide only interfaces for a limited number of kinds of
+CPU and memory binding, and some do not provide any binding interface at all.
+Hwloc's binding functions would then simply return the ENOSYS error (Function
+not implemented), meaning that the underlying Operating System does not provide
+any interface for them. CPU binding and Memory binding provide more information
+on which hwloc binding functions should be preferred because interfaces for
+them are usually available on the supported Operating Systems.
+
+Similarly, the ability of reporting topology information varies from one
+platform to another. As shown in Command-line Examples, hwloc can obtain
+information on a wide variety of hardware topologies. However, some platforms
+and/or operating system versions will only report a subset of this information.
+For example, on an PPC64-based system with 8 cores (each with 2 hardware
+threads) running a default 2.6.18-based kernel from RHEL 5.4, hwloc is only
+able to glean information about NUMA nodes and processor units (PUs). No
+information about caches, packages, or cores is available.
+
+Here's the graphical output from lstopo on this platform when Simultaneous
+Multi-Threading (SMT) is enabled:
+
+[ppc64-with]
+
+And here's the graphical output from lstopo on this platform when SMT is
+disabled:
+
+[ppc64-with]
+
+Notice that hwloc only sees half the PUs when SMT is disabled. PU L#6, for
+example, seems to change location from NUMA node #0 to #1. In reality, no PUs
+"moved" -- they were simply re-numbered when hwloc only saw half as many (see
+also Logical index in Indexes and Sets). Hence, PU L#6 in the SMT-disabled
+picture probably corresponds to PU L#12 in the SMT-enabled picture.
+
+This same "PUs have disappeared" effect can be seen on other platforms -- even
+platforms / OSs that provide much more information than the above PPC64 system.
+This is an unfortunate side-effect of how operating systems report information
+to hwloc.
+
+Note that upgrading the Linux kernel on the same PPC64 system mentioned above
+to 2.6.34, hwloc is able to discover all the topology information. The
+following picture shows the entire topology layout when SMT is enabled:
+
+[ppc64-full]
+
+Developers using the hwloc API or XML output for portable applications should
+therefore be extremely careful to not make any assumptions about the structure
+of data that is returned. For example, per the above reported PPC topology, it
+is not safe to assume that PUs will always be descendants of cores.
+
+Additionally, future hardware may insert new topology elements that are not
+available in this version of hwloc. Long-lived applications that are meant to
+span multiple different hardware platforms should also be careful about making
+structure assumptions. For example, a new element may someday exist between a
+core and a PU.
+
+API Example
+
+The following small C example (available in the source tree as ``doc/examples/
+hwloc-hello.c'') prints the topology of the machine and performs some thread
+and memory binding. More examples are available in the doc/examples/ directory
+of the source tree.
+
+/* Example hwloc API program.
+*
+* See other examples under doc/examples/ in the source tree
+* for more details.
+*
+* Copyright (c) 2009-2016 Inria. All rights reserved.
+* Copyright (c) 2009-2011 Universit?eacute; Bordeaux
+* Copyright (c) 2009-2010 Cisco Systems, Inc. All rights reserved.
+* See COPYING in top-level directory.
+*
+* hwloc-hello.c
+*/
+#include "hwloc.h"
+#include <errno.h>
+#include <stdio.h>
+#include <string.h>
+static void print_children(hwloc_topology_t topology, hwloc_obj_t obj,
+int depth)
+{
+char type[32], attr[1024];
+unsigned i;
+hwloc_obj_type_snprintf(type, sizeof(type), obj, 0);
+printf("%*s%s", 2*depth, "", type);
+if (obj->os_index != (unsigned) -1)
+printf("#%u", obj->os_index);
+hwloc_obj_attr_snprintf(attr, sizeof(attr), obj, " ", 0);
+if (*attr)
+printf("(%s)", attr);
+printf("\n");
+for (i = 0; i < obj->arity; i++) {
+print_children(topology, obj->children[i], depth + 1);
+}
+}
+int main(void)
+{
+int depth;
+unsigned i, n;
+unsigned long size;
+int levels;
+char string[128];
+int topodepth;
+void *m;
+hwloc_topology_t topology;
+hwloc_cpuset_t cpuset;
+hwloc_obj_t obj;
+/* Allocate and initialize topology object. */
+hwloc_topology_init(&topology);
+/* ... Optionally, put detection configuration here to ignore
+some objects types, define a synthetic topology, etc....
+The default is to detect all the objects of the machine that
+the caller is allowed to access. See Configure Topology
+Detection. */
+/* Perform the topology detection. */
+hwloc_topology_load(topology);
+/* Optionally, get some additional topology information
+in case we need the topology depth later. */
+topodepth = hwloc_topology_get_depth(topology);
+/*****************************************************************
+* First example:
+* Walk the topology with an array style, from level 0 (always
+* the system level) to the lowest level (always the proc level).
+*****************************************************************/
+for (depth = 0; depth < topodepth; depth++) {
+printf("*** Objects at level %d\n", depth);
+for (i = 0; i < hwloc_get_nbobjs_by_depth(topology, depth);
+i++) {
+hwloc_obj_type_snprintf(string, sizeof(string),
+hwloc_get_obj_by_depth(topology, depth, i), 0);
+printf("Index %u: %s\n", i, string);
+}
+}
+/*****************************************************************
+* Second example:
+* Walk the topology with a tree style.
+*****************************************************************/
+printf("*** Printing overall tree\n");
+print_children(topology, hwloc_get_root_obj(topology), 0);
+/*****************************************************************
+* Third example:
+* Print the number of packages.
+*****************************************************************/
+depth = hwloc_get_type_depth(topology, HWLOC_OBJ_PACKAGE);
+if (depth == HWLOC_TYPE_DEPTH_UNKNOWN) {
+printf("*** The number of packages is unknown\n");
+} else {
+printf("*** %u package(s)\n",
+hwloc_get_nbobjs_by_depth(topology, depth));
+}
+/*****************************************************************
+* Fourth example:
+* Compute the amount of cache that the first logical processor
+* has above it.
+*****************************************************************/
+levels = 0;
+size = 0;
+for (obj = hwloc_get_obj_by_type(topology, HWLOC_OBJ_PU, 0);
+obj;
+obj = obj->parent)
+if (hwloc_obj_type_is_cache(obj->type)) {
+levels++;
+size += obj->attr->cache.size;
+}
+printf("*** Logical processor 0 has %d caches totaling %luKB\n",
+levels, size / 1024);
+/*****************************************************************
+* Fifth example:
+* Bind to only one thread of the last core of the machine.
+*
+* First find out where cores are, or else smaller sets of CPUs if
+* the OS doesn't have the notion of a "core".
+*****************************************************************/
+depth = hwloc_get_type_or_below_depth(topology, HWLOC_OBJ_CORE);
+/* Get last core. */
+obj = hwloc_get_obj_by_depth(topology, depth,
+hwloc_get_nbobjs_by_depth(topology, depth) - 1);
+if (obj) {
+/* Get a copy of its cpuset that we may modify. */
+cpuset = hwloc_bitmap_dup(obj->cpuset);
+/* Get only one logical processor (in case the core is
+SMT/hyper-threaded). */
+hwloc_bitmap_singlify(cpuset);
+/* And try to bind ourself there. */
+if (hwloc_set_cpubind(topology, cpuset, 0)) {
+char *str;
+int error = errno;
+hwloc_bitmap_asprintf(&str, obj->cpuset);
+printf("Couldn't bind to cpuset %s: %s\n", str, strerror(error));
+free(str);
+}
+/* Free our cpuset copy */
+hwloc_bitmap_free(cpuset);
+}
+/*****************************************************************
+* Sixth example:
+* Allocate some memory on the last NUMA node, bind some existing
+* memory to the last NUMA node.
+*****************************************************************/
+/* Get last node. There's always at least one. */
+n = hwloc_get_nbobjs_by_type(topology, HWLOC_OBJ_NUMANODE);
+obj = hwloc_get_obj_by_type(topology, HWLOC_OBJ_NUMANODE, n - 1);
+size = 1024*1024;
+m = hwloc_alloc_membind(topology, size, obj->nodeset,
+HWLOC_MEMBIND_BIND, HWLOC_MEMBIND_BYNODESET);
+hwloc_free(topology, m, size);
+m = malloc(size);
+hwloc_set_area_membind(topology, m, size, obj->nodeset,
+HWLOC_MEMBIND_BIND, HWLOC_MEMBIND_BYNODESET);
+free(m);
+/* Destroy topology object. */
+hwloc_topology_destroy(topology);
+return 0;
+}
+
+hwloc provides a pkg-config executable to obtain relevant compiler and linker
+flags. See Compiling software on top of hwloc's C API for details on building
+program on top of hwloc's API using GNU Make or CMake.
+
+On a machine 2 processor packages -- each package of which has two processing
+cores -- the output from running hwloc-hello could be something like the
+following:
+
+shell$ ./hwloc-hello
+*** Objects at level 0
+Index 0: Machine
+*** Objects at level 1
+Index 0: Package#0
+Index 1: Package#1
+*** Objects at level 2
+Index 0: Core#0
+Index 1: Core#1
+Index 2: Core#3
+Index 3: Core#2
+*** Objects at level 3
+Index 0: PU#0
+Index 1: PU#1
+Index 2: PU#2
+Index 3: PU#3
+*** Printing overall tree
+Machine
+  Package#0
+ Core#0
+   PU#0
+ Core#1
+   PU#1
+  Package#1
+ Core#3
+   PU#2
+ Core#2
+   PU#3
+*** 2 package(s)
+*** Logical processor 0 has 0 caches totaling 0KB
+shell$
 
 Questions and Bugs
 
@@ -80,6 +468,20 @@ www.open-mpi.org/community/lists/hwloc.php).
 
 There is also a #hwloc IRC channel on Libera Chat (irc.libera.chat).
 
+History / Credits
+
+hwloc is the evolution and merger of the libtopology project and the Portable
+Linux Processor Affinity (PLPA) (https://www.open-mpi.org/projects/plpa/)
+project. Because of functional and ideological overlap, these two code bases
+and ideas were merged and released under the name "hwloc" as an Open MPI
+sub-project.
+
+libtopology was initially developed by the Inria Runtime Team-Project. PLPA was
+initially developed by the Open MPI development team as a sub-project. Both are
+now deprecated in favor of hwloc, which is distributed as an Open MPI
+sub-project.
 
 
-See https://www.open-mpi.org/projects/hwloc/doc/ for more hwloc documentation.
+
+See https://www.open-mpi.org/projects/hwloc/doc/ for more hwloc documentation,
+actual links to related pages, images, etc.

src/3rdparty/hwloc/VERSION

@@ -8,8 +8,8 @@
 # Please update HWLOC_VERSION* in contrib/windows/hwloc_config.h too.
 
 major=2
-minor=9
+minor=12
-release=0
+release=1
 
 # greek is used for alpha or beta release tags.  If it is non-empty,
 # it will be appended to the version number.  It does not have to be
@@ -22,7 +22,7 @@ greek=
 
 # The date when this release was created
 
-date="Dec 14, 2022"
+date="May 12, 2025"
 
 # If snapshot=1, then use the value from snapshot_version as the
 # entire hwloc version (i.e., ignore major, minor, release, and
@@ -41,7 +41,6 @@ snapshot_version=${major}.${minor}.${release}${greek}-git
 # 2. Version numbers are described in the Libtool current:revision:age
 # format.
 
-libhwloc_so_version=21:1:6
+libhwloc_so_version=25:0:10
-libnetloc_so_version=0:0:0
 
 # Please also update the <TargetName> lines in contrib/windows/libhwloc.vcxproj

src/3rdparty/hwloc/include/hwloc/autogen/config.h

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2022 Inria.  All rights reserved.
+ * Copyright © 2009-2025 Inria.  All rights reserved.
  * Copyright © 2009-2012 Université Bordeaux
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -11,10 +11,10 @@
 #ifndef HWLOC_CONFIG_H
 #define HWLOC_CONFIG_H
 
-#define HWLOC_VERSION "2.9.0"
+#define HWLOC_VERSION "2.12.1"
 #define HWLOC_VERSION_MAJOR 2
-#define HWLOC_VERSION_MINOR 9
+#define HWLOC_VERSION_MINOR 12
-#define HWLOC_VERSION_RELEASE 0
+#define HWLOC_VERSION_RELEASE 1
 #define HWLOC_VERSION_GREEK ""
 
 #define __hwloc_restrict

src/3rdparty/hwloc/include/hwloc/bitmap.h

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2022 Inria.  All rights reserved.
+ * Copyright © 2009-2024 Inria.  All rights reserved.
  * Copyright © 2009-2012 Université Bordeaux
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -50,9 +50,10 @@ extern "C" {
  * hwloc_bitmap_free(set);
  * \endcode
  *
- * \note Most functions below return an int that may be negative in case of
+ * \note Most functions below return 0 on success and -1 on error.
- * error. The usual error case would be an internal failure to realloc/extend
+ * The usual error case would be an internal failure to realloc/extend
  * the storage of the bitmap (\p errno would be set to \c ENOMEM).
+ * See also \ref hwlocality_api_error_reporting.
  *
  * \note Several examples of using the bitmap API are available under the
  * doc/examples/ directory in the source tree.
@@ -83,7 +84,13 @@ typedef const struct hwloc_bitmap_s * hwloc_const_bitmap_t;
  */
 HWLOC_DECLSPEC hwloc_bitmap_t hwloc_bitmap_alloc(void) __hwloc_attribute_malloc;
 
-/** \brief Allocate a new full bitmap. */
+/** \brief Allocate a new full bitmap.
+ *
+ * \returns A valid bitmap or \c NULL.
+ *
+ * The bitmap should be freed by a corresponding call to
+ * hwloc_bitmap_free().
+ */
 HWLOC_DECLSPEC hwloc_bitmap_t hwloc_bitmap_alloc_full(void) __hwloc_attribute_malloc;
 
 /** \brief Free bitmap \p bitmap.
@@ -106,73 +113,134 @@ HWLOC_DECLSPEC int hwloc_bitmap_copy(hwloc_bitmap_t dst, hwloc_const_bitmap_t sr
  * Bitmap/String Conversion
  */
 
-/** \brief Stringify a bitmap.
+/** \brief Stringify a bitmap in the default hwloc format.
+ *
+ * <b>Note that if the bitmap is a CPU or nodeset, it contains physical indexes.</b>
+ *
+ * Print the bits set inside a bitmap as a comma-separated list of hexadecimal 32-bit blocks.
+ * A bitmap containing bits 1, 33, 34, and all from 64 to 95 is printed as <tt>"0xffffffff,0x00000006,0x00000002"</tt>.
  *
  * Up to \p buflen characters may be written in buffer \p buf.
  *
  * If \p buflen is 0, \p buf may safely be \c NULL.
  *
  * \return the number of characters that were actually written if not truncating,
- * or that would have been written (not including the ending \\0).
+ * or that would have been written (not including the ending \c \0).
+ * \return -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_snprintf(char * __hwloc_restrict buf, size_t buflen, hwloc_const_bitmap_t bitmap);
 
-/** \brief Stringify a bitmap into a newly allocated string.
+/** \brief Stringify a bitmap into a newly allocated string in the default hwloc format.
  *
- * \return -1 on error.
+ * <b>Note that if the bitmap is a CPU or nodeset, it contains physical indexes.</b>
+ *
+ * Print the bits set inside a bitmap as a comma-separated list of hexadecimal 32-bit blocks.
+ * A bitmap containing bits 1, 33, 34, and all from 64 to 95 is printed as <tt>"0xffffffff,0x00000006,0x00000002"</tt>.
+ *
+ * \return the number of characters that were written (not including the ending \c \0).
+ * \return -1 on error, for instance with \p errno set to \c ENOMEM on failure to allocate the output string.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_asprintf(char ** strp, hwloc_const_bitmap_t bitmap);
 
-/** \brief Parse a bitmap string and stores it in bitmap \p bitmap.
+/** \brief Parse a bitmap string as the default hwloc format and stores it in bitmap \p bitmap.
+ *
+ * <b>Note that if the bitmap is a CPU or nodeset, the input string must contain physical indexes.</b>
+ *
+ * The input string should be a comma-separared list of hexadecimal 32-bit blocks.
+ * String <tt>"0xffffffff,0x6,0x2"</tt> is parsed as a bitmap containing all bits between 64 and 95,
+ * and bits 33, 34 and 1.
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_sscanf(hwloc_bitmap_t bitmap, const char * __hwloc_restrict string);
 
 /** \brief Stringify a bitmap in the list format.
+ *
+ * <b>Note that if the bitmap is a CPU or nodeset, it contains physical indexes.</b>
  *
  * Lists are comma-separated indexes or ranges.
  * Ranges are dash separated indexes.
- * The last range may not have an ending indexes if the bitmap is infinitely set.
+ * A bitmap containing bits 1, 33, 34, and all from 64 to 95 is printed as <tt>"1,33-34,64-95"</tt>.
+ * The last range may not have an ending index if the bitmap is infinitely set.
  *
  * Up to \p buflen characters may be written in buffer \p buf.
  *
  * If \p buflen is 0, \p buf may safely be \c NULL.
  *
  * \return the number of characters that were actually written if not truncating,
- * or that would have been written (not including the ending \\0).
+ * or that would have been written (not including the ending \c \0).
+ * \return -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_list_snprintf(char * __hwloc_restrict buf, size_t buflen, hwloc_const_bitmap_t bitmap);
 
 /** \brief Stringify a bitmap into a newly allocated list string.
  *
- * \return -1 on error.
+ * <b>Note that if the bitmap is a CPU or nodeset, it contains physical indexes.</b>
+ *
+ * Lists are comma-separated indexes or ranges.
+ * Ranges are dash separated indexes.
+ * A bitmap containing bits 1, 33, 34, and all from 64 to 95 is printed as <tt>"1,33-34,64-95"</tt>.
+ * The last range may not have an ending index if the bitmap is infinitely set.
+ *
+ * \return the number of characters that were written (not including the ending \c \0).
+ * \return -1 on error, for instance with \p errno set to \c ENOMEM on failure to allocate the output string.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_list_asprintf(char ** strp, hwloc_const_bitmap_t bitmap);
 
 /** \brief Parse a list string and stores it in bitmap \p bitmap.
+ *
+ * <b>Note that if the bitmap is a CPU or nodeset, the input string must contain physical indexes.</b>
+ *
+ * Lists are comma-separated indexes or ranges.
+ * Ranges are dash separated indexes.
+ * String <tt>"1,33-34,64-95"</tt> is parsed as a bitmap containing bits 1, 33, 34, and all from 64 to 95.
+ * The last range may not have an ending index if the bitmap is infinitely set.
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_list_sscanf(hwloc_bitmap_t bitmap, const char * __hwloc_restrict string);
 
 /** \brief Stringify a bitmap in the taskset-specific format.
  *
- * The taskset command manipulates bitmap strings that contain a single
+ * <b>Note that if the bitmap is a CPU or nodeset, it contains physical indexes.</b>
+ *
+ * The taskset program manipulates bitmap strings that contain a single
  * (possible very long) hexadecimal number starting with 0x.
+ * A bitmap containing bits 1, 33, 34, and all from 64 to 95 is printed as </tt>"0xffffffff0000000600000002"</tt>.
  *
  * Up to \p buflen characters may be written in buffer \p buf.
  *
  * If \p buflen is 0, \p buf may safely be \c NULL.
  *
  * \return the number of characters that were actually written if not truncating,
- * or that would have been written (not including the ending \\0).
+ * or that would have been written (not including the ending \c \0).
+ * \return -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_taskset_snprintf(char * __hwloc_restrict buf, size_t buflen, hwloc_const_bitmap_t bitmap);
 
 /** \brief Stringify a bitmap into a newly allocated taskset-specific string.
  *
- * \return -1 on error.
+ * <b>Note that if the bitmap is a CPU or nodeset, it contains physical indexes.</b>
+ *
+ * The taskset program manipulates bitmap strings that contain a single
+ * (possible very long) hexadecimal number starting with 0x.
+ * A bitmap containing bits 1, 33, 34, and all from 64 to 95 is printed as <tt>"0xffffffff0000000600000002"</tt>.
+ *
+ * \return the number of characters that were written (not including the ending \c \0).
+ * \return -1 on error, for instance with \p errno set to \c ENOMEM on failure to allocate the output string.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_taskset_asprintf(char ** strp, hwloc_const_bitmap_t bitmap);
 
 /** \brief Parse a taskset-specific bitmap string and stores it in bitmap \p bitmap.
+ *
+ * <b>Note that if the bitmap is a CPU or nodeset, the input string must contain physical indexes.</b>
+ *
+ * The taskset program manipulates bitmap strings that contain a single
+ * (possible very long) hexadecimal number starting with 0x.
+ * String <tt>"0xffffffff0000000600000002"</tt> is parsed as a bitmap containing all bits between 64 and 95,
+ * and bits 33, 34 and 1.
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_taskset_sscanf(hwloc_bitmap_t bitmap, const char * __hwloc_restrict string);
 
@@ -279,6 +347,7 @@ HWLOC_DECLSPEC int hwloc_bitmap_to_ulongs(hwloc_const_bitmap_t bitmap, unsigned
  * When called on the output of hwloc_topology_get_topology_cpuset(),
  * the returned number is large enough for all cpusets of the topology.
  *
+ * \return the number of unsigned longs required.
  * \return -1 if \p bitmap is infinite.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_nr_ulongs(hwloc_const_bitmap_t bitmap) __hwloc_attribute_pure;
@@ -305,21 +374,23 @@ HWLOC_DECLSPEC int hwloc_bitmap_isfull(hwloc_const_bitmap_t bitmap) __hwloc_attr
 
 /** \brief Compute the first index (least significant bit) in bitmap \p bitmap
  *
- * \return -1 if no index is set in \p bitmap.
+ * \return the first index set in \p bitmap.
+ * \return -1 if \p bitmap is empty.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_first(hwloc_const_bitmap_t bitmap) __hwloc_attribute_pure;
 
 /** \brief Compute the next index in bitmap \p bitmap which is after index \p prev
  *
- * If \p prev is -1, the first index is returned.
+ * \return the first index set in \p bitmap if \p prev is \c -1.
- *
+ * \return the next index set in \p bitmap if \p prev is not \c -1.
  * \return -1 if no index with higher index is set in \p bitmap.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_next(hwloc_const_bitmap_t bitmap, int prev) __hwloc_attribute_pure;
 
 /** \brief Compute the last index (most significant bit) in bitmap \p bitmap
  *
- * \return -1 if no index is set in \p bitmap, or if \p bitmap is infinitely set.
+ * \return the last index set in \p bitmap.
+ * \return -1 if \p bitmap is empty, or if \p bitmap is infinitely set.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_last(hwloc_const_bitmap_t bitmap) __hwloc_attribute_pure;
 
@@ -327,28 +398,29 @@ HWLOC_DECLSPEC int hwloc_bitmap_last(hwloc_const_bitmap_t bitmap) __hwloc_attrib
  * indexes that are in the bitmap).
  *
  * \return the number of indexes that are in the bitmap.
- *
  * \return -1 if \p bitmap is infinitely set.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_weight(hwloc_const_bitmap_t bitmap) __hwloc_attribute_pure;
 
 /** \brief Compute the first unset index (least significant bit) in bitmap \p bitmap
  *
- * \return -1 if no index is unset in \p bitmap.
+ * \return the first unset index in \p bitmap.
+ * \return -1 if \p bitmap is full.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_first_unset(hwloc_const_bitmap_t bitmap) __hwloc_attribute_pure;
 
 /** \brief Compute the next unset index in bitmap \p bitmap which is after index \p prev
  *
- * If \p prev is -1, the first unset index is returned.
+ * \return the first index unset in \p bitmap if \p prev is \c -1.
- *
+ * \return the next index unset in \p bitmap if \p prev is not \c -1.
  * \return -1 if no index with higher index is unset in \p bitmap.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_next_unset(hwloc_const_bitmap_t bitmap, int prev) __hwloc_attribute_pure;
 
 /** \brief Compute the last unset index (most significant bit) in bitmap \p bitmap
  *
- * \return -1 if no index is unset in \p bitmap, or if \p bitmap is infinitely set.
+ * \return the last index unset in \p bitmap.
+ * \return -1 if \p bitmap is full, or if \p bitmap is not infinitely set.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_last_unset(hwloc_const_bitmap_t bitmap) __hwloc_attribute_pure;
 
@@ -428,6 +500,8 @@ HWLOC_DECLSPEC int hwloc_bitmap_not (hwloc_bitmap_t res, hwloc_const_bitmap_t bi
 /** \brief Test whether bitmaps \p bitmap1 and \p bitmap2 intersects.
  *
  * \return 1 if bitmaps intersect, 0 otherwise.
+ *
+ * \note The empty bitmap does not intersect any other bitmap.
  */
 HWLOC_DECLSPEC int hwloc_bitmap_intersects (hwloc_const_bitmap_t bitmap1, hwloc_const_bitmap_t bitmap2) __hwloc_attribute_pure;
 

src/3rdparty/hwloc/include/hwloc/cuda.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2010-2021 Inria.  All rights reserved.
+ * Copyright © 2010-2023 Inria.  All rights reserved.
  * Copyright © 2010-2011 Université Bordeaux
  * Copyright © 2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -42,6 +42,9 @@ extern "C" {
 /** \brief Return the domain, bus and device IDs of the CUDA device \p cudevice.
  *
  * Device \p cudevice must match the local machine.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if device information could not be found.
  */
 static __hwloc_inline int
 hwloc_cuda_get_device_pci_ids(hwloc_topology_t topology __hwloc_attribute_unused,
@@ -87,6 +90,9 @@ hwloc_cuda_get_device_pci_ids(hwloc_topology_t topology __hwloc_attribute_unused
  *
  * This function is currently only implemented in a meaningful way for
  * Linux; other systems will simply get a full cpuset.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if device information could not be found.
  */
 static __hwloc_inline int
 hwloc_cuda_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_unused,

src/3rdparty/hwloc/include/hwloc/cudart.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2010-2021 Inria.  All rights reserved.
+ * Copyright © 2010-2023 Inria.  All rights reserved.
  * Copyright © 2010-2011 Université Bordeaux
  * Copyright © 2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -43,6 +43,9 @@ extern "C" {
 /** \brief Return the domain, bus and device IDs of the CUDA device whose index is \p idx.
  *
  * Device index \p idx must match the local machine.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if device information could not be found.
  */
 static __hwloc_inline int
 hwloc_cudart_get_device_pci_ids(hwloc_topology_t topology __hwloc_attribute_unused,
@@ -84,6 +87,9 @@ hwloc_cudart_get_device_pci_ids(hwloc_topology_t topology __hwloc_attribute_unus
  *
  * This function is currently only implemented in a meaningful way for
  * Linux; other systems will simply get a full cpuset.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if device information could not be found.
  */
 static __hwloc_inline int
 hwloc_cudart_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_unused,

src/3rdparty/hwloc/include/hwloc/diff.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2013-2020 Inria.  All rights reserved.
+ * Copyright © 2013-2024 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -222,6 +222,8 @@ enum hwloc_topology_diff_apply_flags_e {
 HWLOC_DECLSPEC int hwloc_topology_diff_apply(hwloc_topology_t topology, hwloc_topology_diff_t diff, unsigned long flags);
 
 /** \brief Destroy a list of topology differences.
+ *
+ * \return 0.
  */
 HWLOC_DECLSPEC int hwloc_topology_diff_destroy(hwloc_topology_diff_t diff);
 
@@ -233,6 +235,8 @@ HWLOC_DECLSPEC int hwloc_topology_diff_destroy(hwloc_topology_diff_t diff);
  * This identifier is usually the name of the other XML file
  * that contains the reference topology.
  *
+ * \return 0 on success, -1 on error.
+ *
  * \note the pointer returned in refname should later be freed
  * by the caller.
  */
@@ -246,10 +250,17 @@ HWLOC_DECLSPEC int hwloc_topology_diff_load_xml(const char *xmlpath, hwloc_topol
  * This identifier is usually the name of the other XML file
  * that contains the reference topology.
  * This attribute is given back when reading the diff from XML.
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_topology_diff_export_xml(hwloc_topology_diff_t diff, const char *refname, const char *xmlpath);
 
 /** \brief Load a list of topology differences from a XML buffer.
+ *
+ * Build a list of differences from the XML memory buffer given
+ * at \p xmlbuffer and of length \p buflen (including an ending \c \0).
+ * This buffer may have been filled earlier with
+ * hwloc_topology_diff_export_xmlbuffer().
  *
  * If not \c NULL, \p refname will be filled with the identifier
  * string of the reference topology for the difference file,
@@ -257,6 +268,8 @@ HWLOC_DECLSPEC int hwloc_topology_diff_export_xml(hwloc_topology_diff_t diff, co
  * This identifier is usually the name of the other XML file
  * that contains the reference topology.
  *
+ * \return 0 on success, -1 on error.
+ *
  * \note the pointer returned in refname should later be freed
  * by the caller.
   */
@@ -271,9 +284,11 @@ HWLOC_DECLSPEC int hwloc_topology_diff_load_xmlbuffer(const char *xmlbuffer, int
  * that contains the reference topology.
  * This attribute is given back when reading the diff from XML.
  *
- * The returned buffer ends with a \0 that is included in the returned
+ * The returned buffer ends with a \c \0 that is included in the returned
  * length.
  *
+ * \return 0 on success, -1 on error.
+ *
  * \note The XML buffer should later be freed with hwloc_free_xmlbuffer().
  */
 HWLOC_DECLSPEC int hwloc_topology_diff_export_xmlbuffer(hwloc_topology_diff_t diff, const char *refname, char **xmlbuffer, int *buflen);

src/3rdparty/hwloc/include/hwloc/distances.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2010-2022 Inria.  All rights reserved.
+ * Copyright © 2010-2025 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -28,18 +28,18 @@ extern "C" {
 
 /** \brief Matrix of distances between a set of objects.
  *
- * This matrix often contains latencies between NUMA nodes
+ * The most common matrix contains latencies between NUMA nodes
  * (as reported in the System Locality Distance Information Table (SLIT)
  * in the ACPI specification), which may or may not be physically accurate.
  * It corresponds to the latency for accessing the memory of one node
  * from a core in another node.
- * The corresponding kind is ::HWLOC_DISTANCES_KIND_FROM_OS | ::HWLOC_DISTANCES_KIND_FROM_USER.
+ * The corresponding kind is ::HWLOC_DISTANCES_KIND_MEANS_LATENCY | ::HWLOC_DISTANCES_KIND_FROM_USER.
  * The name of this distances structure is "NUMALatency".
- * Others distance structures include and "XGMIBandwidth", "XGMIHops",
- * "XeLinkBandwidth" and "NVLinkBandwidth".
  *
  * The matrix may also contain bandwidths between random sets of objects,
  * possibly provided by the user, as specified in the \p kind attribute.
+ * Others common distance structures include and "XGMIBandwidth", "XGMIHops",
+ * "XeLinkBandwidth" and "NVLinkBandwidth".
  *
  * Pointers \p objs and \p values should not be replaced, reallocated, freed, etc.
  * However callers are allowed to modify \p kind as well as the contents
@@ -70,11 +70,10 @@ struct hwloc_distances_s {
  * The \p kind attribute of struct hwloc_distances_s is a OR'ed set
  * of kinds.
  *
- * A kind of format HWLOC_DISTANCES_KIND_FROM_* specifies where the
+ * Each distance matrix may have only one kind among HWLOC_DISTANCES_KIND_FROM_*
- * distance information comes from, if known.
+ * specifying where distance information comes from,
- *
+ * and one kind among HWLOC_DISTANCES_KIND_MEANS_* specifying
- * A kind of format HWLOC_DISTANCES_KIND_MEANS_* specifies whether
+ * whether values are latencies or bandwidths.
- * values are latencies or bandwidths, if applicable.
  */
 enum hwloc_distances_kind_e {
   /** \brief These distances were obtained from the operating system or hardware.
@@ -131,6 +130,8 @@ enum hwloc_distances_kind_e {
  *
  * Each distance matrix returned in the \p distances array should be released
  * by the caller using hwloc_distances_release().
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int
 hwloc_distances_get(hwloc_topology_t topology,
@@ -140,6 +141,8 @@ hwloc_distances_get(hwloc_topology_t topology,
 /** \brief Retrieve distance matrices for object at a specific depth in the topology.
  *
  * Identical to hwloc_distances_get() with the additional \p depth filter.
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int
 hwloc_distances_get_by_depth(hwloc_topology_t topology, int depth,
@@ -149,6 +152,8 @@ hwloc_distances_get_by_depth(hwloc_topology_t topology, int depth,
 /** \brief Retrieve distance matrices for object of a specific type.
  *
  * Identical to hwloc_distances_get() with the additional \p type filter.
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int
 hwloc_distances_get_by_type(hwloc_topology_t topology, hwloc_obj_type_t type,
@@ -162,6 +167,8 @@ hwloc_distances_get_by_type(hwloc_topology_t topology, hwloc_obj_type_t type,
  * The name of the most common structure is "NUMALatency".
  * Others include "XGMIBandwidth", "XGMIHops", "XeLinkBandwidth",
  * and "NVLinkBandwidth".
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int
 hwloc_distances_get_by_name(hwloc_topology_t topology, const char *name,
@@ -171,7 +178,12 @@ hwloc_distances_get_by_name(hwloc_topology_t topology, const char *name,
 /** \brief Get a description of what a distances structure contains.
  *
  * For instance "NUMALatency" for hardware-provided NUMA distances (ACPI SLIT),
- * or NULL if unknown.
+ * or \c NULL if unknown.
+ *
+ * \return the constant string with the name of the distance structure.
+ *
+ * \note The returned name should not be freed by the caller,
+ * it belongs to the hwloc library.
  */
 HWLOC_DECLSPEC const char *
 hwloc_distances_get_name(hwloc_topology_t topology, struct hwloc_distances_s *distances);
@@ -215,17 +227,24 @@ enum hwloc_distances_transform_e {
   HWLOC_DISTANCES_TRANSFORM_LINKS = 1,
 
   /** \brief Merge switches with multiple ports into a single object.
-   * This currently only applies to NVSwitches where GPUs seem connected to different
+   *
-   * separate switch ports in the NVLinkBandwidth matrix. This transformation will
+   * This currently only applies to NVSwitches where GPUs seem connected
-   * replace all of them with the same port connected to all GPUs.
+   * to different switch ports. Switch ports must be objects with subtype
-   * Other ports are removed by applying ::HWLOC_DISTANCES_TRANSFORM_REMOVE_NULL internally.
+   * "NVSwitch" as in the NVLinkBandwidth matrix.
+   *
+   * This transformation will replace all ports with only the first one,
+   * now connected to all GPUs. Other ports are removed by applying
+   * ::HWLOC_DISTANCES_TRANSFORM_REMOVE_NULL internally.
    * \hideinitializer
    */
   HWLOC_DISTANCES_TRANSFORM_MERGE_SWITCH_PORTS = 2,
 
   /** \brief Apply a transitive closure to the matrix to connect objects across switches.
-   * This currently only applies to GPUs and NVSwitches in the NVLinkBandwidth matrix.
+   *
-   * All pairs of GPUs will be reported as directly connected.
+   * All pairs of GPUs will be reported as directly connected instead GPUs being
+   * only connected to switches.
+   *
+   * Switch ports must be objects with subtype "NVSwitch" as in the NVLinkBandwidth matrix.
    * \hideinitializer
    */
   HWLOC_DISTANCES_TRANSFORM_TRANSITIVE_CLOSURE = 3
@@ -252,6 +271,8 @@ enum hwloc_distances_transform_e {
  *
  * \p flags must be \c 0 for now.
  *
+ * \return 0 on success, -1 on error for instance if flags are invalid.
+ *
  * \note Objects in distances array \p objs may be directly modified
  * in place without using hwloc_distances_transform().
  * One may use hwloc_get_obj_with_same_locality() to easily convert
@@ -272,6 +293,7 @@ HWLOC_DECLSPEC int hwloc_distances_transform(hwloc_topology_t topology, struct h
 
 /** \brief Find the index of an object in a distances structure.
  *
+ * \return the index of the object in the distances structure if any.
  * \return -1 if object \p obj is not involved in structure \p distances.
  */
 static __hwloc_inline int
@@ -289,6 +311,7 @@ hwloc_distances_obj_index(struct hwloc_distances_s *distances, hwloc_obj_t obj)
  * The distance from \p obj1 to \p obj2 is stored in the value pointed by
  * \p value1to2 and reciprocally.
  *
+ * \return 0 on success.
  * \return -1 if object \p obj1 or \p obj2 is not involved in structure \p distances.
  */
 static __hwloc_inline int
@@ -340,6 +363,8 @@ typedef void * hwloc_distances_add_handle_t;
  * Otherwise, it will be copied internally and may later be freed by the caller.
  *
  * \p kind specifies the kind of distance as a OR'ed set of ::hwloc_distances_kind_e.
+ * Only one kind of meaning and one kind of provenance may be given if appropriate
+ * (e.g. ::HWLOC_DISTANCES_KIND_MEANS_BANDWIDTH and ::HWLOC_DISTANCES_KIND_FROM_USER).
  * Kind ::HWLOC_DISTANCES_KIND_HETEROGENEOUS_TYPES will be automatically set
  * according to objects having different types in hwloc_distances_add_values().
  *
@@ -374,8 +399,8 @@ hwloc_distances_add_create(hwloc_topology_t topology,
  *
  * \p flags must be \c 0 for now.
  *
- * \return \c 0 on success.
+ * \return 0 on success.
- * \return \c -1 on error.
+ * \return -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_distances_add_values(hwloc_topology_t topology,
                                               hwloc_distances_add_handle_t handle,
@@ -386,7 +411,8 @@ HWLOC_DECLSPEC int hwloc_distances_add_values(hwloc_topology_t topology,
 /** \brief Flags for adding a new distances to a topology. */
 enum hwloc_distances_add_flag_e {
   /** \brief Try to group objects based on the newly provided distance information.
-   * This is ignored for distances between objects of different types.
+   * Grouping is only performed when the distances structure contains latencies,
+   * and when all objects are of the same type.
    * \hideinitializer
    */
   HWLOC_DISTANCES_ADD_FLAG_GROUP = (1UL<<0),
@@ -411,8 +437,8 @@ enum hwloc_distances_add_flag_e {
  *
  * On error, the temporary distances structure and its content are destroyed.
  *
- * \return \c 0 on success.
+ * \return 0 on success.
- * \return \c -1 on error.
+ * \return -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_distances_add_commit(hwloc_topology_t topology,
                                               hwloc_distances_add_handle_t handle,
@@ -433,18 +459,24 @@ HWLOC_DECLSPEC int hwloc_distances_add_commit(hwloc_topology_t topology,
  *
  * If these distances were used to group objects, these additional
  * Group objects are not removed from the topology.
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_distances_remove(hwloc_topology_t topology);
 
 /** \brief Remove distance matrices for objects at a specific depth in the topology.
  *
  * Identical to hwloc_distances_remove() but only applies to one level of the topology.
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_distances_remove_by_depth(hwloc_topology_t topology, int depth);
 
 /** \brief Remove distance matrices for objects of a specific type in the topology.
  *
  * Identical to hwloc_distances_remove() but only applies to one level of the topology.
+ *
+ * \return 0 on success, -1 on error.
  */
 static __hwloc_inline int
 hwloc_distances_remove_by_type(hwloc_topology_t topology, hwloc_obj_type_t type)
@@ -458,6 +490,8 @@ hwloc_distances_remove_by_type(hwloc_topology_t topology, hwloc_obj_type_t type)
 /** \brief Release and remove the given distance matrice from the topology.
  *
  * This function includes a call to hwloc_distances_release().
+ *
+ * \return 0 on success, -1 on error.
  */
 HWLOC_DECLSPEC int hwloc_distances_release_remove(hwloc_topology_t topology, struct hwloc_distances_s *distances);
 

src/3rdparty/hwloc/include/hwloc/export.h

@@ -55,7 +55,7 @@ enum hwloc_topology_export_xml_flags_e {
  *
  * \p flags is a OR'ed set of ::hwloc_topology_export_xml_flags_e.
  *
- * \return -1 if a failure occured.
+ * \return 0 on success, or -1 on error.
  *
  * \note See also hwloc_topology_set_userdata_export_callback()
  * for exporting application-specific object userdata.
@@ -91,7 +91,7 @@ HWLOC_DECLSPEC int hwloc_topology_export_xml(hwloc_topology_t topology, const ch
  *
  * \p flags is a OR'ed set of ::hwloc_topology_export_xml_flags_e.
  *
- * \return -1 if a failure occured.
+ * \return 0 on success, or -1 on error.
  *
  * \note See also hwloc_topology_set_userdata_export_callback()
  * for exporting application-specific object userdata.
@@ -145,13 +145,15 @@ HWLOC_DECLSPEC void hwloc_topology_set_userdata_export_callback(hwloc_topology_t
  * that were given to the export callback.
  *
  * Only printable characters may be exported to XML string attributes.
- * If a non-printable character is passed in \p name or \p buffer,
- * the function returns -1 with errno set to EINVAL.
  *
  * If exporting binary data, the application should first encode into
  * printable characters only (or use hwloc_export_obj_userdata_base64()).
  * It should also take care of portability issues if the export may
  * be reimported on a different architecture.
+ *
+ * \return 0 on success.
+ * \return -1 with errno set to \c EINVAL if a non-printable character is
+ * passed in \p name or \b buffer.
  */
 HWLOC_DECLSPEC int hwloc_export_obj_userdata(void *reserved, hwloc_topology_t topology, hwloc_obj_t obj, const char *name, const void *buffer, size_t length);
 
@@ -165,8 +167,14 @@ HWLOC_DECLSPEC int hwloc_export_obj_userdata(void *reserved, hwloc_topology_t to
  * This function may only be called from within the export() callback passed
  * to hwloc_topology_set_userdata_export_callback().
  *
+ * The name must be made of printable characters for export to XML string attributes.
+ *
  * The function does not take care of portability issues if the export
  * may be reimported on a different architecture.
+ *
+ * \return 0 on success.
+ * \return -1 with errno set to \c EINVAL if a non-printable character is
+ * passed in \p name.
  */
 HWLOC_DECLSPEC int hwloc_export_obj_userdata_base64(void *reserved, hwloc_topology_t topology, hwloc_obj_t obj, const char *name, const void *buffer, size_t length);
 

src/3rdparty/hwloc/include/hwloc/gl.h

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2012 Blue Brain Project, EPFL. All rights reserved.
- * Copyright © 2012-2021 Inria.  All rights reserved.
+ * Copyright © 2012-2023 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -102,7 +102,8 @@ hwloc_gl_get_display_osdev_by_name(hwloc_topology_t topology,
  * Retrieves the OpenGL display port (server) in \p port and device (screen)
  * in \p screen that correspond to the given hwloc OS device object.
  *
- * \return \c -1 if none could be found.
+ * \return 0 on success.
+ * \return -1 if none could be found.
  *
  * The topology \p topology does not necessarily have to match the current
  * machine. For instance the topology may be an XML import of a remote host.

src/3rdparty/hwloc/include/hwloc/glibc-sched.h

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2020 Inria.  All rights reserved.
+ * Copyright © 2009-2023 Inria.  All rights reserved.
  * Copyright © 2009-2011 Université Bordeaux
  * Copyright © 2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -52,6 +52,8 @@ extern "C" {
  * that takes a cpu_set_t as input parameter.
  *
  * \p schedsetsize should be sizeof(cpu_set_t) unless \p schedset was dynamically allocated with CPU_ALLOC
+ *
+ * \return 0.
  */
 static __hwloc_inline int
 hwloc_cpuset_to_glibc_sched_affinity(hwloc_topology_t topology __hwloc_attribute_unused, hwloc_const_cpuset_t hwlocset,
@@ -80,6 +82,9 @@ hwloc_cpuset_to_glibc_sched_affinity(hwloc_topology_t topology __hwloc_attribute
  * that takes a cpu_set_t  as input parameter.
  *
  * \p schedsetsize should be sizeof(cpu_set_t) unless \p schedset was dynamically allocated with CPU_ALLOC
+ *
+ * \return 0 on success.
+ * \return -1 with errno set to \c ENOMEM if some internal reallocation failed.
  */
 static __hwloc_inline int
 hwloc_cpuset_from_glibc_sched_affinity(hwloc_topology_t topology __hwloc_attribute_unused, hwloc_cpuset_t hwlocset,
@@ -95,7 +100,8 @@ hwloc_cpuset_from_glibc_sched_affinity(hwloc_topology_t topology __hwloc_attribu
   cpu = 0;
   while (count) {
     if (CPU_ISSET_S(cpu, schedsetsize, schedset)) {
-      hwloc_bitmap_set(hwlocset, cpu);
+      if (hwloc_bitmap_set(hwlocset, cpu) < 0)
+        return -1;
       count--;
     }
     cpu++;
@@ -107,7 +113,8 @@ hwloc_cpuset_from_glibc_sched_affinity(hwloc_topology_t topology __hwloc_attribu
   assert(schedsetsize == sizeof(cpu_set_t));
   for(cpu=0; cpu<CPU_SETSIZE; cpu++)
     if (CPU_ISSET(cpu, schedset))
-      hwloc_bitmap_set(hwlocset, cpu);
+      if (hwloc_bitmap_set(hwlocset, cpu) < 0)
+        return -1;
 #endif /* !CPU_ZERO_S */
   return 0;
 }

src/3rdparty/hwloc/include/hwloc/levelzero.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2021 Inria.  All rights reserved.
+ * Copyright © 2021-2024 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -32,7 +32,8 @@ extern "C" {
 /** \defgroup hwlocality_levelzero Interoperability with the oneAPI Level Zero interface.
  *
  * This interface offers ways to retrieve topology information about
- * devices managed by the Level Zero API.
+ * devices managed by the Level Zero API, both for main Core devices (ZE API)
+ * and the Sysman devices (ZES API).
  *
  * @{
  */
@@ -44,8 +45,7 @@ extern "C" {
  * the Level Zero device \p device.
  *
  * Topology \p topology and device \p device must match the local machine.
- * The Level Zero must have been initialized with Sysman enabled
+ * The Level Zero library must have been initialized with zeInit().
- * (ZES_ENABLE_SYSMAN=1 in the environment).
  * I/O devices detection and the Level Zero component are not needed in the
  * topology.
  *
@@ -55,6 +55,12 @@ extern "C" {
  *
  * This function is currently only implemented in a meaningful way for
  * Linux; other systems will simply get a full cpuset.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if device information could not be found.
+ *
+ * \note zeDevicePciGetPropertiesExt() must be supported, or the entire machine
+ * locality will be returned.
  */
 static __hwloc_inline int
 hwloc_levelzero_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_unused,
@@ -64,8 +70,7 @@ hwloc_levelzero_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_un
   /* If we're on Linux, use the sysfs mechanism to get the local cpus */
 #define HWLOC_LEVELZERO_DEVICE_SYSFS_PATH_MAX 128
   char path[HWLOC_LEVELZERO_DEVICE_SYSFS_PATH_MAX];
-  zes_pci_properties_t pci;
+  ze_pci_ext_properties_t pci;
-  zes_device_handle_t sdevice = device;
   ze_result_t res;
 
   if (!hwloc_topology_is_thissystem(topology)) {
@@ -73,7 +78,65 @@ hwloc_levelzero_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_un
     return -1;
   }
 
-  res = zesDevicePciGetProperties(sdevice, &pci);
+  pci.stype =  ZE_STRUCTURE_TYPE_PCI_EXT_PROPERTIES;
+  pci.pNext = NULL;
+  res = zeDevicePciGetPropertiesExt(device, &pci);
+  if (res != ZE_RESULT_SUCCESS) {
+    errno = EINVAL;
+    return -1;
+  }
+
+  sprintf(path, "/sys/bus/pci/devices/%04x:%02x:%02x.%01x/local_cpus",
+          pci.address.domain, pci.address.bus, pci.address.device, pci.address.function);
+  if (hwloc_linux_read_path_as_cpumask(path, set) < 0
+      || hwloc_bitmap_iszero(set))
+    hwloc_bitmap_copy(set, hwloc_topology_get_complete_cpuset(topology));
+#else
+  /* Non-Linux systems simply get a full cpuset */
+  hwloc_bitmap_copy(set, hwloc_topology_get_complete_cpuset(topology));
+#endif
+  return 0;
+}
+
+/** \brief Get the CPU set of logical processors that are physically
+ * close to the Level Zero Sysman device \p device
+ *
+ * Store in \p set the CPU-set describing the locality of
+ * the Level Zero device \p device.
+ *
+ * Topology \p topology and device \p device must match the local machine.
+ * The Level Zero library must have been initialized with Sysman enabled
+ * with zesInit().
+ * I/O devices detection and the Level Zero component are not needed in the
+ * topology.
+ *
+ * The function only returns the locality of the device.
+ * If more information about the device is needed, OS objects should
+ * be used instead, see hwloc_levelzero_get_device_osdev().
+ *
+ * This function is currently only implemented in a meaningful way for
+ * Linux; other systems will simply get a full cpuset.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if device information could not be found.
+ */
+static __hwloc_inline int
+hwloc_levelzero_get_sysman_device_cpuset(hwloc_topology_t topology __hwloc_attribute_unused,
+                                         zes_device_handle_t device, hwloc_cpuset_t set)
+{
+#ifdef HWLOC_LINUX_SYS
+  /* If we're on Linux, use the sysfs mechanism to get the local cpus */
+#define HWLOC_LEVELZERO_DEVICE_SYSFS_PATH_MAX 128
+  char path[HWLOC_LEVELZERO_DEVICE_SYSFS_PATH_MAX];
+  zes_pci_properties_t pci;
+  ze_result_t res;
+
+  if (!hwloc_topology_is_thissystem(topology)) {
+    errno = EINVAL;
+    return -1;
+  }
+
+  res = zesDevicePciGetProperties(device, &pci);
   if (res != ZE_RESULT_SUCCESS) {
     errno = EINVAL;
     return -1;
@@ -98,17 +161,90 @@ hwloc_levelzero_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_un
  * \return \c NULL if none could be found.
  *
  * Topology \p topology and device \p dv_ind must match the local machine.
+ * The Level Zero library must have been initialized with zeInit().
  * I/O devices detection and the Level Zero component must be enabled in the
  * topology. If not, the locality of the object may still be found using
  * hwloc_levelzero_get_device_cpuset().
  *
+ * \note If the input ZE device is actually a subdevice, then its parent
+ * (root device) is actually translated, i.e. the main hwloc OS device
+ * is returned instead of one of its children.
+ *
+ * \note The corresponding hwloc PCI device may be found by looking
+ * at the result parent pointer (unless PCI devices are filtered out).
+ *
+ * \note zeDevicePciGetPropertiesExt() must be supported.
+ */
+static __hwloc_inline hwloc_obj_t
+hwloc_levelzero_get_device_osdev(hwloc_topology_t topology, ze_device_handle_t device)
+{
+  ze_pci_ext_properties_t pci;
+  ze_result_t res;
+  hwloc_obj_t osdev;
+
+  if (!hwloc_topology_is_thissystem(topology)) {
+    errno = EINVAL;
+    return NULL;
+  }
+
+  pci.stype = ZE_STRUCTURE_TYPE_PCI_EXT_PROPERTIES;
+  pci.pNext = NULL;
+  res = zeDevicePciGetPropertiesExt(device, &pci);
+  if (res != ZE_RESULT_SUCCESS) {
+    errno = EINVAL;
+    return NULL;
+  }
+
+  osdev = NULL;
+  while ((osdev = hwloc_get_next_osdev(topology, osdev)) != NULL) {
+    hwloc_obj_t pcidev;
+
+    if (strncmp(osdev->name, "ze", 2))
+      continue;
+
+    pcidev = osdev;
+    while (pcidev && pcidev->type != HWLOC_OBJ_PCI_DEVICE)
+      pcidev = pcidev->parent;
+    if (!pcidev)
+      continue;
+
+    if (pcidev
+      && pcidev->type == HWLOC_OBJ_PCI_DEVICE
+      && pcidev->attr->pcidev.domain == pci.address.domain
+      && pcidev->attr->pcidev.bus == pci.address.bus
+      && pcidev->attr->pcidev.dev == pci.address.device
+      && pcidev->attr->pcidev.func == pci.address.function)
+      return osdev;
+
+    /* FIXME: when we'll have serialnumber, try it in case PCI is filtered-out */
+  }
+
+  return NULL;
+}
+
+/** \brief Get the hwloc OS device object corresponding to Level Zero Sysman device
+ * \p device.
+ *
+ * \return The hwloc OS device object that describes the given Level Zero device \p device.
+ * \return \c NULL if none could be found.
+ *
+ * Topology \p topology and device \p dv_ind must match the local machine.
+ * The Level Zero library must have been initialized with Sysman enabled
+ * with zesInit().
+ * I/O devices detection and the Level Zero component must be enabled in the
+ * topology. If not, the locality of the object may still be found using
+ * hwloc_levelzero_get_device_cpuset().
+ *
+ * \note If the input ZES device is actually a subdevice, then its parent
+ * (root device) is actually translated, i.e. the main hwloc OS device
+ * is returned instead of one of its children.
+ *
  * \note The corresponding hwloc PCI device may be found by looking
  * at the result parent pointer (unless PCI devices are filtered out).
  */
 static __hwloc_inline hwloc_obj_t
-hwloc_levelzero_get_device_osdev(hwloc_topology_t topology, ze_device_handle_t device)
+hwloc_levelzero_get_sysman_device_osdev(hwloc_topology_t topology, zes_device_handle_t device)
 {
-  zes_device_handle_t sdevice = device;
   zes_pci_properties_t pci;
   ze_result_t res;
   hwloc_obj_t osdev;
@@ -118,20 +254,25 @@ hwloc_levelzero_get_device_osdev(hwloc_topology_t topology, ze_device_handle_t d
     return NULL;
   }
 
-  res = zesDevicePciGetProperties(sdevice, &pci);
+  res = zesDevicePciGetProperties(device, &pci);
   if (res != ZE_RESULT_SUCCESS) {
-    /* L0 was likely initialized without sysman, don't bother */
     errno = EINVAL;
     return NULL;
   }
 
   osdev = NULL;
   while ((osdev = hwloc_get_next_osdev(topology, osdev)) != NULL) {
-    hwloc_obj_t pcidev = osdev->parent;
+    hwloc_obj_t pcidev;
 
     if (strncmp(osdev->name, "ze", 2))
       continue;
 
+    pcidev = osdev;
+    while (pcidev && pcidev->type != HWLOC_OBJ_PCI_DEVICE)
+      pcidev = pcidev->parent;
+    if (!pcidev)
+      continue;
+
     if (pcidev
       && pcidev->type == HWLOC_OBJ_PCI_DEVICE
       && pcidev->attr->pcidev.domain == pci.address.domain

src/3rdparty/hwloc/include/hwloc/linux-libnuma.h

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2017 Inria.  All rights reserved.
+ * Copyright © 2009-2023 Inria.  All rights reserved.
  * Copyright © 2009-2010, 2012 Université Bordeaux
  * See COPYING in top-level directory.
  */
@@ -50,6 +50,8 @@ extern "C" {
  * This function may be used before calling set_mempolicy, mbind, migrate_pages
  * or any other function that takes an array of unsigned long and a maximal
  * node number as input parameter.
+ *
+ * \return 0.
  */
 static __hwloc_inline int
 hwloc_cpuset_to_linux_libnuma_ulongs(hwloc_topology_t topology, hwloc_const_cpuset_t cpuset,
@@ -84,6 +86,8 @@ hwloc_cpuset_to_linux_libnuma_ulongs(hwloc_topology_t topology, hwloc_const_cpus
  * This function may be used before calling set_mempolicy, mbind, migrate_pages
  * or any other function that takes an array of unsigned long and a maximal
  * node number as input parameter.
+ *
+ * \return 0.
  */
 static __hwloc_inline int
 hwloc_nodeset_to_linux_libnuma_ulongs(hwloc_topology_t topology, hwloc_const_nodeset_t nodeset,
@@ -119,6 +123,9 @@ hwloc_nodeset_to_linux_libnuma_ulongs(hwloc_topology_t topology, hwloc_const_nod
  * This function may be used after calling get_mempolicy or any other function
  * that takes an array of unsigned long as output parameter (and possibly
  * a maximal node number as input parameter).
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if failing an internal reallocation.
  */
 static __hwloc_inline int
 hwloc_cpuset_from_linux_libnuma_ulongs(hwloc_topology_t topology, hwloc_cpuset_t cpuset,
@@ -130,7 +137,8 @@ hwloc_cpuset_from_linux_libnuma_ulongs(hwloc_topology_t topology, hwloc_cpuset_t
   while ((node = hwloc_get_next_obj_by_depth(topology, depth, node)) != NULL)
     if (node->os_index < maxnode
 	&& (mask[node->os_index/sizeof(*mask)/8] & (1UL << (node->os_index % (sizeof(*mask)*8)))))
-      hwloc_bitmap_or(cpuset, cpuset, node->cpuset);
+      if (hwloc_bitmap_or(cpuset, cpuset, node->cpuset) < 0)
+        return -1;
   return 0;
 }
 
@@ -142,6 +150,9 @@ hwloc_cpuset_from_linux_libnuma_ulongs(hwloc_topology_t topology, hwloc_cpuset_t
  * This function may be used after calling get_mempolicy or any other function
  * that takes an array of unsigned long as output parameter (and possibly
  * a maximal node number as input parameter).
+ *
+ * \return 0 on success.
+ * \return -1 with errno set to \c ENOMEM if some internal reallocation failed.
  */
 static __hwloc_inline int
 hwloc_nodeset_from_linux_libnuma_ulongs(hwloc_topology_t topology, hwloc_nodeset_t nodeset,
@@ -153,7 +164,8 @@ hwloc_nodeset_from_linux_libnuma_ulongs(hwloc_topology_t topology, hwloc_nodeset
   while ((node = hwloc_get_next_obj_by_depth(topology, depth, node)) != NULL)
     if (node->os_index < maxnode
 	&& (mask[node->os_index/sizeof(*mask)/8] & (1UL << (node->os_index % (sizeof(*mask)*8)))))
-      hwloc_bitmap_set(nodeset, node->os_index);
+      if (hwloc_bitmap_set(nodeset, node->os_index) < 0)
+        return -1;
   return 0;
 }
 
@@ -184,7 +196,7 @@ hwloc_nodeset_from_linux_libnuma_ulongs(hwloc_topology_t topology, hwloc_nodeset
  * This function may be used before calling many numa_ functions
  * that use a struct bitmask as an input parameter.
  *
- * \return newly allocated struct bitmask.
+ * \return newly allocated struct bitmask, or \c NULL on error.
  */
 static __hwloc_inline struct bitmask *
 hwloc_cpuset_to_linux_libnuma_bitmask(hwloc_topology_t topology, hwloc_const_cpuset_t cpuset) __hwloc_attribute_malloc;
@@ -209,7 +221,7 @@ hwloc_cpuset_to_linux_libnuma_bitmask(hwloc_topology_t topology, hwloc_const_cpu
  * This function may be used before calling many numa_ functions
  * that use a struct bitmask as an input parameter.
  *
- * \return newly allocated struct bitmask.
+ * \return newly allocated struct bitmask, or \c NULL on error.
  */
 static __hwloc_inline struct bitmask *
 hwloc_nodeset_to_linux_libnuma_bitmask(hwloc_topology_t topology, hwloc_const_nodeset_t nodeset) __hwloc_attribute_malloc;
@@ -231,6 +243,9 @@ hwloc_nodeset_to_linux_libnuma_bitmask(hwloc_topology_t topology, hwloc_const_no
  *
  * This function may be used after calling many numa_ functions
  * that use a struct bitmask as an output parameter.
+ *
+ * \return 0 on success.
+ * \return -1 with errno set to \c ENOMEM if some internal reallocation failed.
  */
 static __hwloc_inline int
 hwloc_cpuset_from_linux_libnuma_bitmask(hwloc_topology_t topology, hwloc_cpuset_t cpuset,
@@ -241,7 +256,8 @@ hwloc_cpuset_from_linux_libnuma_bitmask(hwloc_topology_t topology, hwloc_cpuset_
   hwloc_bitmap_zero(cpuset);
   while ((node = hwloc_get_next_obj_by_depth(topology, depth, node)) != NULL)
     if (numa_bitmask_isbitset(bitmask, node->os_index))
-      hwloc_bitmap_or(cpuset, cpuset, node->cpuset);
+      if (hwloc_bitmap_or(cpuset, cpuset, node->cpuset) < 0)
+        return -1;
   return 0;
 }
 
@@ -249,6 +265,9 @@ hwloc_cpuset_from_linux_libnuma_bitmask(hwloc_topology_t topology, hwloc_cpuset_
  *
  * This function may be used after calling many numa_ functions
  * that use a struct bitmask as an output parameter.
+ *
+ * \return 0 on success.
+ * \return -1 with errno set to \c ENOMEM if some internal reallocation failed.
  */
 static __hwloc_inline int
 hwloc_nodeset_from_linux_libnuma_bitmask(hwloc_topology_t topology, hwloc_nodeset_t nodeset,
@@ -259,7 +278,8 @@ hwloc_nodeset_from_linux_libnuma_bitmask(hwloc_topology_t topology, hwloc_nodese
   hwloc_bitmap_zero(nodeset);
   while ((node = hwloc_get_next_obj_by_depth(topology, depth, node)) != NULL)
     if (numa_bitmask_isbitset(bitmask, node->os_index))
-      hwloc_bitmap_set(nodeset, node->os_index);
+      if (hwloc_bitmap_set(nodeset, node->os_index) < 0)
+        return -1;
   return 0;
 }
 

src/3rdparty/hwloc/include/hwloc/linux.h

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2021 Inria.  All rights reserved.
+ * Copyright © 2009-2023 Inria.  All rights reserved.
  * Copyright © 2009-2011 Université Bordeaux
  * See COPYING in top-level directory.
  */
@@ -38,6 +38,8 @@ extern "C" {
  * The behavior is exactly the same as the Linux sched_setaffinity system call,
  * but uses a hwloc cpuset.
  *
+ * \return 0 on success, -1 on error.
+ *
  * \note This is equivalent to calling hwloc_set_proc_cpubind() with
  * HWLOC_CPUBIND_THREAD as flags.
  */
@@ -52,6 +54,8 @@ HWLOC_DECLSPEC int hwloc_linux_set_tid_cpubind(hwloc_topology_t topology, pid_t
  * The behavior is exactly the same as the Linux sched_getaffinity system call,
  * but uses a hwloc cpuset.
  *
+ * \return 0 on success, -1 on error.
+ *
  * \note This is equivalent to calling hwloc_get_proc_cpubind() with
  * ::HWLOC_CPUBIND_THREAD as flags.
  */
@@ -62,6 +66,8 @@ HWLOC_DECLSPEC int hwloc_linux_get_tid_cpubind(hwloc_topology_t topology, pid_t
  * The CPU-set \p set (previously allocated by the caller)
  * is filled with the PU which the thread last ran on.
  *
+ * \return 0 on success, -1 on error.
+ *
  * \note This is equivalent to calling hwloc_get_proc_last_cpu_location() with
  * ::HWLOC_CPUBIND_THREAD as flags.
  */
@@ -72,6 +78,8 @@ HWLOC_DECLSPEC int hwloc_linux_get_tid_last_cpu_location(hwloc_topology_t topolo
  * Might be used when reading CPU set from sysfs attributes such as topology
  * and caches for processors, or local_cpus for devices.
  *
+ * \return 0 on success, -1 on error.
+ *
  * \note This function ignores the HWLOC_FSROOT environment variable.
  */
 HWLOC_DECLSPEC int hwloc_linux_read_path_as_cpumask(const char *path, hwloc_bitmap_t set);

src/3rdparty/hwloc/include/hwloc/memattrs.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2019-2022 Inria.  All rights reserved.
+ * Copyright © 2019-2025 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -54,6 +54,15 @@ extern "C" {
  * Attribute values for these nodes, if any, may then be obtained with
  * hwloc_memattr_get_value() and manually compared with the desired criteria.
  *
+ * Memory attributes are also used internally to build Memory Tiers which provide
+ * an easy way to distinguish NUMA nodes of different kinds, as explained
+ * in \ref heteromem.
+ *
+ * Beside tiers, hwloc defines a set of "default" nodes where normal memory
+ * allocations should be made from (see hwloc_topology_get_default_nodeset()).
+ * This is also useful for dividing the machine into a set of non-overlapping
+ * NUMA domains, for instance for binding tasks per domain.
+ *
  * \sa An example is available in doc/examples/memory-attributes.c in the source tree.
  *
  * \note The API also supports specific objects as initiator,
@@ -65,7 +74,10 @@ extern "C" {
  * @{
  */
 
-/** \brief Memory node attributes. */
+/** \brief Predefined memory attribute IDs.
+ * See ::hwloc_memattr_id_t for the generic definition of IDs
+ * for predefined or custom attributes.
+ */
 enum hwloc_memattr_id_e {
   /** \brief
    * The \"Capacity\" is returned in bytes (local_memory attribute in objects).
@@ -74,6 +86,8 @@ enum hwloc_memattr_id_e {
    *
    * No initiator is involved when looking at this attribute.
    * The corresponding attribute flags are ::HWLOC_MEMATTR_FLAG_HIGHER_FIRST.
+   *
+   * Capacity values may not be modified using hwloc_memattr_set_value().
    * \hideinitializer
    */
   HWLOC_MEMATTR_ID_CAPACITY = 0,
@@ -89,6 +103,8 @@ enum hwloc_memattr_id_e {
    *
    * No initiator is involved when looking at this attribute.
    * The corresponding attribute flags are ::HWLOC_MEMATTR_FLAG_HIGHER_FIRST.
+
+   * Locality values may not be modified using hwloc_memattr_set_value().
    * \hideinitializer
    */
   HWLOC_MEMATTR_ID_LOCALITY = 1,
@@ -169,15 +185,26 @@ enum hwloc_memattr_id_e {
 
   /* TODO persistence? */
 
-  HWLOC_MEMATTR_ID_MAX /**< \private Sentinel value */
+  HWLOC_MEMATTR_ID_MAX /**< \private
+                        * Sentinel value for predefined attributes.
+                        * Dynamically registered custom attributes start here.
+                        */
 };
 
 /** \brief A memory attribute identifier.
- * May be either one of ::hwloc_memattr_id_e or a new id returned by hwloc_memattr_register().
+ *
+ * hwloc predefines some commonly-used attributes in ::hwloc_memattr_id_e.
+ * One may then dynamically register custom ones with hwloc_memattr_register(),
+ * they will be assigned IDs immediately after the predefined ones.
+ * See \ref hwlocality_memattrs_manage for more information about
+ * existing attribute IDs.
  */
 typedef unsigned hwloc_memattr_id_t;
 
 /** \brief Return the identifier of the memory attribute with the given name.
+ *
+ * \return 0 on success.
+ * \return -1 with errno set to \c EINVAL if no such attribute exists.
  */
 HWLOC_DECLSPEC int
 hwloc_memattr_get_by_name(hwloc_topology_t topology,
@@ -223,6 +250,16 @@ enum hwloc_local_numanode_flag_e {
    */
   HWLOC_LOCAL_NUMANODE_FLAG_SMALLER_LOCALITY = (1UL<<1),
 
+  /** \breif Select NUMA nodes whose locality intersects the given cpuset.
+   * This includes larger and smaller localities as well as localities
+   * that are partially included.
+   * For instance, if the locality is one core of both packages, a NUMA node
+   * local to one package is neither larger nor smaller than this locality,
+   * but it intersects it.
+   * \hideinitializer
+   */
+  HWLOC_LOCAL_NUMANODE_FLAG_INTERSECT_LOCALITY = (1UL<<3),
+
   /** \brief Select all NUMA nodes in the topology.
    * The initiator \p initiator is ignored.
    * \hideinitializer
@@ -247,6 +284,8 @@ enum hwloc_local_numanode_flag_e {
  * or the number of nodes that would have been stored if there were
  * enough room.
  *
+ * \return 0 on success or -1 on error.
+ *
  * \note Some of these NUMA nodes may not have any memory attribute
  * values and hence not be reported as actual targets in other functions.
  *
@@ -266,7 +305,57 @@ hwloc_get_local_numanode_objs(hwloc_topology_t topology,
                               hwloc_obj_t *nodes,
                               unsigned long flags);
 
-
+/** \brief Return the set of default NUMA nodes
+ *
+ * In machines with heterogeneous memory, some NUMA nodes are considered
+ * the default ones, i.e. where basic allocations should be made from.
+ * These are usually DRAM nodes.
+ *
+ * Other nodes may be reserved for specific use (I/O device memory, e.g. GPU memory),
+ * small but high performance (HBM), large but slow memory (NVM), etc.
+ * Buffers should usually not be allocated from there unless explicitly required.
+ *
+ * This function fills \p nodeset with the bits of NUMA nodes considered default.
+ *
+ * It is guaranteed that these nodes have non-intersecting CPU sets,
+ * i.e. cores may not have multiple local NUMA nodes anymore.
+ * Hence this may be used to iterate over the platform divided into separate
+ * NUMA localities, for instance for binding one task per NUMA domain.
+ *
+ * Any core that had some local NUMA node(s) in the initial topology should
+ * still have one in the default nodeset. Corner cases where this would be
+ * wrong consist in asymmetric platforms with missing DRAM nodes, or topologies
+ * that were already restricted to less NUMA nodes.
+ *
+ * The returned nodeset may be passed to hwloc_topology_restrict() with
+ * ::HWLOC_RESTRICT_FLAG_BYNODESET to remove all non-default nodes from
+ * the topology. The resulting topology will be easier to use when iterating
+ * over (now homogeneous) NUMA nodes.
+ *
+ * The heuristics for finding default nodes relies on memory tiers and subtypes
+ * (see \ref heteromem) as well as the assumption that hardware vendors list
+ * default nodes first in hardware tables.
+ *
+ * \p flags must be \c 0 for now.
+ *
+ * \return 0 on success.
+ * \return -1 on error.
+ *
+ * \note The returned nodeset usually contains all nodes from a single memory
+ * tier, likely the DRAM one.
+ *
+ * \note The returned nodeset is included in the list of available nodes
+ * returned by hwloc_topology_get_topology_nodeset(). It is strictly smaller
+ * if the machine has heterogeneous memory.
+ *
+ * \note The heuristics may return a suboptimal set of nodes if hwloc could
+ * not guess memory types and/or if some default nodes were removed earlier
+ * from the topology (e.g. with hwloc_topology_restrict()).
+ */
+HWLOC_DECLSPEC int
+hwloc_topology_get_default_nodeset(hwloc_topology_t topology,
+                                   hwloc_nodeset_t nodeset,
+                                   unsigned long flags);
 
 /** \brief Return an attribute value for a specific target NUMA node.
  *
@@ -274,8 +363,16 @@ hwloc_get_local_numanode_objs(hwloc_topology_t topology,
  * (it does not have the flag ::HWLOC_MEMATTR_FLAG_NEED_INITIATOR),
  * location \p initiator is ignored and may be \c NULL.
  *
+ * \p target_node cannot be \c NULL. If \p attribute is ::HWLOC_MEMATTR_ID_CAPACITY,
+ * \p target_node must be a NUMA node. If it is ::HWLOC_MEMATTR_ID_LOCALITY,
+ * \p target_node must have a CPU set.
+ *
  * \p flags must be \c 0 for now.
  *
+ * \return 0 on success.
+ * \return -1 on error, for instance with errno set to \c EINVAL if flags
+ * are invalid or no such attribute exists.
+ *
  * \note The initiator \p initiator should be of type ::HWLOC_LOCATION_TYPE_CPUSET
  * when refering to accesses performed by CPU cores.
  * ::HWLOC_LOCATION_TYPE_OBJECT is currently unused internally by hwloc,
@@ -307,7 +404,10 @@ hwloc_memattr_get_value(hwloc_topology_t topology,
  *
  * \p flags must be \c 0 for now.
  *
- * If there are no matching targets, \c -1 is returned with \p errno set to \c ENOENT;
+ * \return 0 on success.
+ * \return -1 with errno set to \c ENOENT if there are no matching targets.
+ * \return -1 with errno set to \c EINVAL if flags are invalid,
+ * or no such attribute exists.
  *
  * \note The initiator \p initiator should be of type ::HWLOC_LOCATION_TYPE_CPUSET
  * when refering to accesses performed by CPU cores.
@@ -323,10 +423,6 @@ hwloc_memattr_get_best_target(hwloc_topology_t topology,
                               hwloc_obj_t *best_target, hwloc_uint64_t *value);
 
 /** \brief Return the best initiator for the given attribute and target NUMA node.
- *
- * If the attribute does not relate to a specific initiator
- * (it does not have the flag ::HWLOC_MEMATTR_FLAG_NEED_INITIATOR),
- * \c -1 is returned and \p errno is set to \c EINVAL.
  *
  * If \p value is non \c NULL, the corresponding value is returned there.
  *
@@ -340,96 +436,22 @@ hwloc_memattr_get_best_target(hwloc_topology_t topology,
  * The returned initiator should not be modified or freed,
  * it belongs to the topology.
  *
+ * \p target_node cannot be \c NULL.
+ *
  * \p flags must be \c 0 for now.
  *
- * If there are no matching initiators, \c -1 is returned with \p errno set to \c ENOENT;
+ * \return 0 on success.
+ * \return -1 with errno set to \c ENOENT if there are no matching initiators.
+ * \return -1 with errno set to \c EINVAL if the attribute does not relate to a specific initiator
+ * (it does not have the flag ::HWLOC_MEMATTR_FLAG_NEED_INITIATOR).
  */
 HWLOC_DECLSPEC int
 hwloc_memattr_get_best_initiator(hwloc_topology_t topology,
                                  hwloc_memattr_id_t attribute,
-                                 hwloc_obj_t target,
+                                 hwloc_obj_t target_node,
                                  unsigned long flags,
                                  struct hwloc_location *best_initiator, hwloc_uint64_t *value);
 
-/** @} */
-
-
-/** \defgroup hwlocality_memattrs_manage Managing memory attributes
- * @{
- */
-
-/** \brief Return the name of a memory attribute.
- */
-HWLOC_DECLSPEC int
-hwloc_memattr_get_name(hwloc_topology_t topology,
-                       hwloc_memattr_id_t attribute,
-                       const char **name);
-
-/** \brief Return the flags of the given attribute.
- *
- * Flags are a OR'ed set of ::hwloc_memattr_flag_e.
- */
-HWLOC_DECLSPEC int
-hwloc_memattr_get_flags(hwloc_topology_t topology,
-                        hwloc_memattr_id_t attribute,
-                        unsigned long *flags);
-
-/** \brief Memory attribute flags.
- * Given to hwloc_memattr_register() and returned by hwloc_memattr_get_flags().
- */
-enum hwloc_memattr_flag_e {
-  /** \brief The best nodes for this memory attribute are those with the higher values.
-   * For instance Bandwidth.
-   */
-  HWLOC_MEMATTR_FLAG_HIGHER_FIRST = (1UL<<0),
-  /** \brief The best nodes for this memory attribute are those with the lower values.
-   * For instance Latency.
-   */
-  HWLOC_MEMATTR_FLAG_LOWER_FIRST = (1UL<<1),
-  /** \brief The value returned for this memory attribute depends on the given initiator.
-   * For instance Bandwidth and Latency, but not Capacity.
-   */
-  HWLOC_MEMATTR_FLAG_NEED_INITIATOR = (1UL<<2)
-};
-
-/** \brief Register a new memory attribute.
- *
- * Add a specific memory attribute that is not defined in ::hwloc_memattr_id_e.
- * Flags are a OR'ed set of ::hwloc_memattr_flag_e. It must contain at least
- * one of ::HWLOC_MEMATTR_FLAG_HIGHER_FIRST or ::HWLOC_MEMATTR_FLAG_LOWER_FIRST.
- */
-HWLOC_DECLSPEC int
-hwloc_memattr_register(hwloc_topology_t topology,
-                       const char *name,
-                       unsigned long flags,
-                       hwloc_memattr_id_t *id);
-
-/** \brief Set an attribute value for a specific target NUMA node.
- *
- * If the attribute does not relate to a specific initiator
- * (it does not have the flag ::HWLOC_MEMATTR_FLAG_NEED_INITIATOR),
- * location \p initiator is ignored and may be \c NULL.
- *
- * The initiator will be copied into the topology,
- * the caller should free anything allocated to store the initiator,
- * for instance the cpuset.
- *
- * \p flags must be \c 0 for now.
- *
- * \note The initiator \p initiator should be of type ::HWLOC_LOCATION_TYPE_CPUSET
- * when referring to accesses performed by CPU cores.
- * ::HWLOC_LOCATION_TYPE_OBJECT is currently unused internally by hwloc,
- * but users may for instance use it to provide custom information about
- * host memory accesses performed by GPUs.
- */
-HWLOC_DECLSPEC int
-hwloc_memattr_set_value(hwloc_topology_t topology,
-                        hwloc_memattr_id_t attribute,
-                        hwloc_obj_t target_node,
-                        struct hwloc_location *initiator,
-                        unsigned long flags,
-                        hwloc_uint64_t value);
-
 /** \brief Return the target NUMA nodes that have some values for a given attribute.
  *
  * Return targets for the given attribute in the \p targets array
@@ -460,6 +482,8 @@ hwloc_memattr_set_value(hwloc_topology_t topology,
  * NUMA nodes with hwloc_get_local_numanode_objs() and then look at their attribute
  * values.
  *
+ * \return 0 on success or -1 on error.
+ *
  * \note The initiator \p initiator should be of type ::HWLOC_LOCATION_TYPE_CPUSET
  * when referring to accesses performed by CPU cores.
  * ::HWLOC_LOCATION_TYPE_OBJECT is currently unused internally by hwloc,
@@ -491,12 +515,16 @@ hwloc_memattr_get_targets(hwloc_topology_t topology,
  * The returned initiators should not be modified or freed,
  * they belong to the topology.
  *
+ * \p target_node cannot be \c NULL.
+ *
  * \p flags must be \c 0 for now.
  *
  * If the attribute does not relate to a specific initiator
  * (it does not have the flag ::HWLOC_MEMATTR_FLAG_NEED_INITIATOR),
  * no initiator is returned.
  *
+ * \return 0 on success or -1 on error.
+ *
  * \note This function is meant for tools and debugging (listing internal information)
  * rather than for application queries. Applications should rather select useful
  * NUMA nodes with hwloc_get_local_numanode_objs() and then look at their attribute
@@ -508,6 +536,131 @@ hwloc_memattr_get_initiators(hwloc_topology_t topology,
                              hwloc_obj_t target_node,
                              unsigned long flags,
                              unsigned *nr, struct hwloc_location *initiators, hwloc_uint64_t *values);
+
+/** @} */
+
+
+/** \defgroup hwlocality_memattrs_manage Managing memory attributes
+ *
+ * Memory attribues are identified by an ID (::hwloc_memattr_id_t)
+ * and a name. hwloc_memattr_get_name() and hwloc_memattr_get_by_name()
+ * convert between them (or return error if the attribute does not exist).
+ *
+ * The set of valid ::hwloc_memattr_id_t is a contigous set starting at \c 0.
+ * It first contains predefined attributes, as listed
+ * in ::hwloc_memattr_id_e (from \c 0 to \c HWLOC_MEMATTR_ID_MAX-1).
+ * Then custom attributes may be dynamically registered with
+ * hwloc_memattr_register(). They will get the following IDs
+ * (\c HWLOC_MEMATTR_ID_MAX for the first one, etc.).
+ *
+ * To iterate over all valid attributes
+ * (either predefined or dynamically registered custom ones),
+ * one may iterate over IDs starting from \c 0 until hwloc_memattr_get_name()
+ * or hwloc_memattr_get_flags() returns an error.
+ *
+ * The values for an existing attribute or for custom dynamically registered ones
+ * may be set or modified with hwloc_memattr_set_value().
+ *
+ * @{
+ */
+
+/** \brief Return the name of a memory attribute.
+ *
+ * The output pointer \p name cannot be \c NULL.
+ *
+ * \return 0 on success.
+ * \return -1 with errno set to \c EINVAL if the attribute does not exist.
+ */
+HWLOC_DECLSPEC int
+hwloc_memattr_get_name(hwloc_topology_t topology,
+                       hwloc_memattr_id_t attribute,
+                       const char **name);
+
+/** \brief Return the flags of the given attribute.
+ *
+ * Flags are a OR'ed set of ::hwloc_memattr_flag_e.
+ *
+ * The output pointer \p flags cannot be \c NULL.
+ *
+ * \return 0 on success.
+ * \return -1 with errno set to \c EINVAL if the attribute does not exist.
+ */
+HWLOC_DECLSPEC int
+hwloc_memattr_get_flags(hwloc_topology_t topology,
+                        hwloc_memattr_id_t attribute,
+                        unsigned long *flags);
+
+/** \brief Memory attribute flags.
+ * Given to hwloc_memattr_register() and returned by hwloc_memattr_get_flags().
+ */
+enum hwloc_memattr_flag_e {
+  /** \brief The best nodes for this memory attribute are those with the higher values.
+   * For instance Bandwidth.
+   */
+  HWLOC_MEMATTR_FLAG_HIGHER_FIRST = (1UL<<0),
+  /** \brief The best nodes for this memory attribute are those with the lower values.
+   * For instance Latency.
+   */
+  HWLOC_MEMATTR_FLAG_LOWER_FIRST = (1UL<<1),
+  /** \brief The value returned for this memory attribute depends on the given initiator.
+   * For instance Bandwidth and Latency, but not Capacity.
+   */
+  HWLOC_MEMATTR_FLAG_NEED_INITIATOR = (1UL<<2)
+};
+
+/** \brief Register a new memory attribute.
+ *
+ * Add a new custom memory attribute.
+ * Flags are a OR'ed set of ::hwloc_memattr_flag_e. It must contain one of
+ * ::HWLOC_MEMATTR_FLAG_HIGHER_FIRST or ::HWLOC_MEMATTR_FLAG_LOWER_FIRST but not both.
+ *
+ * The new attribute \p id is immediately after the last existing attribute ID
+ * (which is either the ID of the last registered attribute if any,
+ * or the ID of the last predefined attribute in ::hwloc_memattr_id_e).
+ *
+ * \return 0 on success.
+ * \return -1 with errno set to \c EINVAL if an invalid set of flags is given.
+ * \return -1 with errno set to \c EBUSY if another attribute already uses this name.
+ */
+HWLOC_DECLSPEC int
+hwloc_memattr_register(hwloc_topology_t topology,
+                       const char *name,
+                       unsigned long flags,
+                       hwloc_memattr_id_t *id);
+
+/** \brief Set an attribute value for a specific target NUMA node.
+ *
+ * If the attribute does not relate to a specific initiator
+ * (it does not have the flag ::HWLOC_MEMATTR_FLAG_NEED_INITIATOR),
+ * location \p initiator is ignored and may be \c NULL.
+ *
+ * The initiator will be copied into the topology,
+ * the caller should free anything allocated to store the initiator,
+ * for instance the cpuset.
+ *
+ * \p target_node cannot be \c NULL.
+ *
+ * \p attribute cannot be ::HWLOC_MEMATTR_FLAG_ID_CAPACITY or
+ * ::HWLOC_MEMATTR_FLAG_ID_LOCALITY.
+ *
+ * \p flags must be \c 0 for now.
+ *
+ * \note The initiator \p initiator should be of type ::HWLOC_LOCATION_TYPE_CPUSET
+ * when referring to accesses performed by CPU cores.
+ * ::HWLOC_LOCATION_TYPE_OBJECT is currently unused internally by hwloc,
+ * but users may for instance use it to provide custom information about
+ * host memory accesses performed by GPUs.
+ *
+ * \return 0 on success or -1 on error.
+ */
+HWLOC_DECLSPEC int
+hwloc_memattr_set_value(hwloc_topology_t topology,
+                        hwloc_memattr_id_t attribute,
+                        hwloc_obj_t target_node,
+                        struct hwloc_location *initiator,
+                        unsigned long flags,
+                        hwloc_uint64_t value);
+
 /** @} */
 
 #ifdef __cplusplus

src/3rdparty/hwloc/include/hwloc/nvml.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2012-2021 Inria.  All rights reserved.
+ * Copyright © 2012-2023 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -51,6 +51,9 @@ extern "C" {
  *
  * This function is currently only implemented in a meaningful way for
  * Linux; other systems will simply get a full cpuset.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if device information could not be found.
  */
 static __hwloc_inline int
 hwloc_nvml_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_unused,

src/3rdparty/hwloc/include/hwloc/opencl.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2012-2021 Inria.  All rights reserved.
+ * Copyright © 2012-2023 Inria.  All rights reserved.
  * Copyright © 2013, 2018 Université Bordeaux.  All right reserved.
  * See COPYING in top-level directory.
  */
@@ -41,6 +41,15 @@ extern "C" {
  */
 /* Copyright (c) 2008-2018 The Khronos Group Inc. */
 
+/* needs "cl_khr_pci_bus_info" device extension, but not strictly required for clGetDeviceInfo() */
+typedef struct {
+    cl_uint pci_domain;
+    cl_uint pci_bus;
+    cl_uint pci_device;
+    cl_uint pci_function;
+} hwloc_cl_device_pci_bus_info_khr;
+#define HWLOC_CL_DEVICE_PCI_BUS_INFO_KHR 0x410F
+
 /* needs "cl_amd_device_attribute_query" device extension, but not strictly required for clGetDeviceInfo() */
 #define HWLOC_CL_DEVICE_TOPOLOGY_AMD 0x4037
 typedef union {
@@ -69,15 +78,28 @@ typedef union {
 /** \brief Return the domain, bus and device IDs of the OpenCL device \p device.
  *
  * Device \p device must match the local machine.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if device information could not be found.
  */
 static __hwloc_inline int
 hwloc_opencl_get_device_pci_busid(cl_device_id device,
                                unsigned *domain, unsigned *bus, unsigned *dev, unsigned *func)
 {
 	hwloc_cl_device_topology_amd amdtopo;
+	hwloc_cl_device_pci_bus_info_khr khrbusinfo;
 	cl_uint nvbus, nvslot, nvdomain;
 	cl_int clret;
 
+	clret = clGetDeviceInfo(device, HWLOC_CL_DEVICE_PCI_BUS_INFO_KHR, sizeof(khrbusinfo), &khrbusinfo, NULL);
+	if (CL_SUCCESS == clret) {
+		*domain = (unsigned) khrbusinfo.pci_domain;
+		*bus = (unsigned) khrbusinfo.pci_bus;
+		*dev = (unsigned) khrbusinfo.pci_device;
+		*func = (unsigned) khrbusinfo.pci_function;
+		return 0;
+	}
+
 	clret = clGetDeviceInfo(device, HWLOC_CL_DEVICE_TOPOLOGY_AMD, sizeof(amdtopo), &amdtopo, NULL);
 	if (CL_SUCCESS == clret
 	    && HWLOC_CL_DEVICE_TOPOLOGY_TYPE_PCIE_AMD == amdtopo.raw.type) {
@@ -126,6 +148,9 @@ hwloc_opencl_get_device_pci_busid(cl_device_id device,
  * This function is currently only implemented in a meaningful way for
  * Linux with the AMD or NVIDIA OpenCL implementation; other systems will simply
  * get a full cpuset.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if the device could not be found.
  */
 static __hwloc_inline int
 hwloc_opencl_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_unused,

src/3rdparty/hwloc/include/hwloc/openfabrics-verbs.h

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2021 Inria.  All rights reserved.
+ * Copyright © 2009-2023 Inria.  All rights reserved.
  * Copyright © 2009-2010 Université Bordeaux
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -57,6 +57,9 @@ extern "C" {
  *
  * This function is currently only implemented in a meaningful way for
  * Linux; other systems will simply get a full cpuset.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if device information could not be found.
  */
 static __hwloc_inline int
 hwloc_ibv_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_unused,

src/3rdparty/hwloc/include/hwloc/plugins.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2013-2022 Inria.  All rights reserved.
+ * Copyright © 2013-2024 Inria.  All rights reserved.
  * Copyright © 2016 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
  */
@@ -26,7 +26,7 @@ struct hwloc_backend;
 
 
 
-/** \defgroup hwlocality_disc_components Components and Plugins: Discovery components
+/** \defgroup hwlocality_disc_components Components and Plugins: Discovery components and backends
  *
  * \note These structures and functions may change when ::HWLOC_COMPONENT_ABI is modified.
  *
@@ -90,18 +90,6 @@ struct hwloc_disc_component {
   struct hwloc_disc_component * next;
 };
 
-/** @} */
-
-
-
-
-/** \defgroup hwlocality_disc_backends Components and Plugins: Discovery backends
- *
- * \note These structures and functions may change when ::HWLOC_COMPONENT_ABI is modified.
- *
- * @{
- */
-
 /** \brief Discovery phase */
 typedef enum hwloc_disc_phase_e {
   /** \brief xml or synthetic, platform-specific components such as bgq.
@@ -164,7 +152,7 @@ struct hwloc_disc_status {
    */
   unsigned excluded_phases;
 
-  /** \brief OR'ed set of hwloc_disc_status_flag_e */
+  /** \brief OR'ed set of ::hwloc_disc_status_flag_e */
   unsigned long flags;
 };
 
@@ -313,6 +301,64 @@ struct hwloc_component {
   void * data;
 };
 
+/** \brief Make sure that plugins can lookup core symbols.
+ *
+ * This is a sanity check to avoid lazy-lookup failures when libhwloc
+ * is loaded within a plugin, and later tries to load its own plugins.
+ * This may fail (and abort the program) if libhwloc symbols are in a
+ * private namespace.
+ *
+ * \return 0 on success.
+ * \return -1 if the plugin cannot be successfully loaded. The caller
+ * plugin init() callback should return a negative error code as well.
+ *
+ * Plugins should call this function in their init() callback to avoid
+ * later crashes if lazy symbol resolution is used by the upper layer that
+ * loaded hwloc (e.g. OpenCL implementations using dlopen with RTLD_LAZY).
+ *
+ * \note The build system must define HWLOC_INSIDE_PLUGIN if and only if
+ * building the caller as a plugin.
+ *
+ * \note This function should remain inline so plugins can call it even
+ * when they cannot find libhwloc symbols.
+ */
+static __hwloc_inline int
+hwloc_plugin_check_namespace(const char *pluginname __hwloc_attribute_unused, const char *symbol __hwloc_attribute_unused)
+{
+#ifdef HWLOC_INSIDE_PLUGIN
+  void *sym;
+#ifdef HWLOC_HAVE_LTDL
+  lt_dlhandle handle = lt_dlopen(NULL);
+#else
+  void *handle = dlopen(NULL, RTLD_NOW|RTLD_LOCAL);
+#endif
+  if (!handle)
+    /* cannot check, assume things will work */
+    return 0;
+#ifdef HWLOC_HAVE_LTDL
+  sym = lt_dlsym(handle, symbol);
+  lt_dlclose(handle);
+#else
+  sym = dlsym(handle, symbol);
+  dlclose(handle);
+#endif
+  if (!sym) {
+    static int verboseenv_checked = 0;
+    static int verboseenv_value = 0;
+    if (!verboseenv_checked) {
+      const char *verboseenv = getenv("HWLOC_PLUGINS_VERBOSE");
+      verboseenv_value = verboseenv ? atoi(verboseenv) : 0;
+      verboseenv_checked = 1;
+    }
+    if (verboseenv_value)
+      fprintf(stderr, "Plugin `%s' disabling itself because it cannot find the `%s' core symbol.\n",
+	      pluginname, symbol);
+    return -1;
+  }
+#endif /* HWLOC_INSIDE_PLUGIN */
+  return 0;
+}
+
 /** @} */
 
 
@@ -422,64 +468,6 @@ HWLOC_DECLSPEC int hwloc_obj_add_children_sets(hwloc_obj_t obj);
  */
 HWLOC_DECLSPEC int hwloc_topology_reconnect(hwloc_topology_t topology, unsigned long flags __hwloc_attribute_unused);
 
-/** \brief Make sure that plugins can lookup core symbols.
- *
- * This is a sanity check to avoid lazy-lookup failures when libhwloc
- * is loaded within a plugin, and later tries to load its own plugins.
- * This may fail (and abort the program) if libhwloc symbols are in a
- * private namespace.
- *
- * \return 0 on success.
- * \return -1 if the plugin cannot be successfully loaded. The caller
- * plugin init() callback should return a negative error code as well.
- *
- * Plugins should call this function in their init() callback to avoid
- * later crashes if lazy symbol resolution is used by the upper layer that
- * loaded hwloc (e.g. OpenCL implementations using dlopen with RTLD_LAZY).
- *
- * \note The build system must define HWLOC_INSIDE_PLUGIN if and only if
- * building the caller as a plugin.
- *
- * \note This function should remain inline so plugins can call it even
- * when they cannot find libhwloc symbols.
- */
-static __hwloc_inline int
-hwloc_plugin_check_namespace(const char *pluginname __hwloc_attribute_unused, const char *symbol __hwloc_attribute_unused)
-{
-#ifdef HWLOC_INSIDE_PLUGIN
-  void *sym;
-#ifdef HWLOC_HAVE_LTDL
-  lt_dlhandle handle = lt_dlopen(NULL);
-#else
-  void *handle = dlopen(NULL, RTLD_NOW|RTLD_LOCAL);
-#endif
-  if (!handle)
-    /* cannot check, assume things will work */
-    return 0;
-#ifdef HWLOC_HAVE_LTDL
-  sym = lt_dlsym(handle, symbol);
-  lt_dlclose(handle);
-#else
-  sym = dlsym(handle, symbol);
-  dlclose(handle);
-#endif
-  if (!sym) {
-    static int verboseenv_checked = 0;
-    static int verboseenv_value = 0;
-    if (!verboseenv_checked) {
-      const char *verboseenv = getenv("HWLOC_PLUGINS_VERBOSE");
-      verboseenv_value = verboseenv ? atoi(verboseenv) : 0;
-      verboseenv_checked = 1;
-    }
-    if (verboseenv_value)
-      fprintf(stderr, "Plugin `%s' disabling itself because it cannot find the `%s' core symbol.\n",
-	      pluginname, symbol);
-    return -1;
-  }
-#endif /* HWLOC_INSIDE_PLUGIN */
-  return 0;
-}
-
 /** @} */
 
 
@@ -645,6 +633,19 @@ HWLOC_DECLSPEC struct hwloc_obj * hwloc_pci_find_parent_by_busid(struct hwloc_to
  */
 HWLOC_DECLSPEC struct hwloc_obj * hwloc_pci_find_by_busid(struct hwloc_topology *topology, unsigned domain, unsigned bus, unsigned dev, unsigned func);
 
+
+/** @} */
+
+
+
+
+/** \defgroup hwlocality_components_distances Components and Plugins: distances
+ *
+ * \note These structures and functions may change when ::HWLOC_COMPONENT_ABI is modified.
+ *
+ * @{
+ */
+
 /** \brief Handle to a new distances structure during its addition to the topology. */
 typedef void * hwloc_backend_distances_add_handle_t;
 

src/3rdparty/hwloc/include/hwloc/rename.h

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
- * Copyright © 2010-2022 Inria.  All rights reserved.
+ * Copyright © 2010-2025 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -176,6 +176,7 @@ extern "C" {
 
 #define hwloc_topology_insert_misc_object HWLOC_NAME(topology_insert_misc_object)
 #define hwloc_topology_alloc_group_object HWLOC_NAME(topology_alloc_group_object)
+#define hwloc_topology_free_group_object HWLOC_NAME(topology_free_group_object)
 #define hwloc_topology_insert_group_object HWLOC_NAME(topology_insert_group_object)
 #define hwloc_obj_add_other_obj_sets HWLOC_NAME(obj_add_other_obj_sets)
 #define hwloc_topology_refresh HWLOC_NAME(topology_refresh)
@@ -209,6 +210,7 @@ extern "C" {
 
 #define hwloc_obj_get_info_by_name HWLOC_NAME(obj_get_info_by_name)
 #define hwloc_obj_add_info HWLOC_NAME(obj_add_info)
+#define hwloc_obj_set_subtype HWLOC_NAME(obj_set_subtype)
 
 #define HWLOC_CPUBIND_PROCESS HWLOC_NAME_CAPS(CPUBIND_PROCESS)
 #define HWLOC_CPUBIND_THREAD HWLOC_NAME_CAPS(CPUBIND_THREAD)
@@ -231,6 +233,7 @@ extern "C" {
 #define HWLOC_MEMBIND_FIRSTTOUCH HWLOC_NAME_CAPS(MEMBIND_FIRSTTOUCH)
 #define HWLOC_MEMBIND_BIND HWLOC_NAME_CAPS(MEMBIND_BIND)
 #define HWLOC_MEMBIND_INTERLEAVE HWLOC_NAME_CAPS(MEMBIND_INTERLEAVE)
+#define HWLOC_MEMBIND_WEIGHTED_INTERLEAVE HWLOC_NAME_CAPS(MEMBIND_WEIGHTED_INTERLEAVE)
 #define HWLOC_MEMBIND_NEXTTOUCH HWLOC_NAME_CAPS(MEMBIND_NEXTTOUCH)
 #define HWLOC_MEMBIND_MIXED HWLOC_NAME_CAPS(MEMBIND_MIXED)
 
@@ -406,8 +409,10 @@ extern "C" {
 #define hwloc_local_numanode_flag_e HWLOC_NAME(local_numanode_flag_e)
 #define HWLOC_LOCAL_NUMANODE_FLAG_LARGER_LOCALITY HWLOC_NAME_CAPS(LOCAL_NUMANODE_FLAG_LARGER_LOCALITY)
 #define HWLOC_LOCAL_NUMANODE_FLAG_SMALLER_LOCALITY HWLOC_NAME_CAPS(LOCAL_NUMANODE_FLAG_SMALLER_LOCALITY)
+#define HWLOC_LOCAL_NUMANODE_FLAG_INTERSECT_LOCALITY HWLOC_NAME_CAPS(LOCAL_NUMANODE_FLAG_INTERSECT_LOCALITY)
 #define HWLOC_LOCAL_NUMANODE_FLAG_ALL HWLOC_NAME_CAPS(LOCAL_NUMANODE_FLAG_ALL)
 #define hwloc_get_local_numanode_objs HWLOC_NAME(get_local_numanode_objs)
+#define hwloc_topology_get_default_nodeset HWLOC_NAME(topology_get_default_nodeset)
 
 #define hwloc_memattr_get_name HWLOC_NAME(memattr_get_name)
 #define hwloc_memattr_get_flags HWLOC_NAME(memattr_get_flags)
@@ -559,6 +564,7 @@ extern "C" {
 
 /* opencl.h */
 
+#define hwloc_cl_device_pci_bus_info_khr HWLOC_NAME(cl_device_pci_bus_info_khr)
 #define hwloc_cl_device_topology_amd HWLOC_NAME(cl_device_topology_amd)
 #define hwloc_opencl_get_device_pci_busid HWLOC_NAME(opencl_get_device_pci_ids)
 #define hwloc_opencl_get_device_cpuset HWLOC_NAME(opencl_get_device_cpuset)
@@ -595,7 +601,9 @@ extern "C" {
 /* levelzero.h */
 
 #define hwloc_levelzero_get_device_cpuset HWLOC_NAME(levelzero_get_device_cpuset)
+#define hwloc_levelzero_get_sysman_device_cpuset HWLOC_NAME(levelzero_get_sysman_device_cpuset)
 #define hwloc_levelzero_get_device_osdev HWLOC_NAME(levelzero_get_device_osdev)
+#define hwloc_levelzero_get_sysman_device_osdev HWLOC_NAME(levelzero_get_sysman_device_osdev)
 
 /* gl.h */
 
@@ -714,6 +722,8 @@ extern "C" {
 #define hwloc__obj_type_is_dcache HWLOC_NAME(_obj_type_is_dcache)
 #define hwloc__obj_type_is_icache HWLOC_NAME(_obj_type_is_icache)
 
+#define hwloc__pci_link_speed HWLOC_NAME(_pci_link_speed)
+
 /* private/cpuid-x86.h */
 
 #define hwloc_have_x86_cpuid HWLOC_NAME(have_x86_cpuid)
@@ -807,6 +817,8 @@ extern "C" {
 #define hwloc_topology_setup_defaults HWLOC_NAME(topology_setup_defaults)
 #define hwloc_topology_clear HWLOC_NAME(topology_clear)
 
+#define hwloc__reconnect HWLOC_NAME(_reconnect)
+
 #define hwloc__attach_memory_object HWLOC_NAME(insert_memory_object)
 
 #define hwloc_get_obj_by_type_and_gp_index HWLOC_NAME(get_obj_by_type_and_gp_index)

src/3rdparty/hwloc/include/hwloc/rsmi.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2012-2021 Inria.  All rights reserved.
+ * Copyright © 2012-2023 Inria.  All rights reserved.
  * Copyright (c) 2020, Advanced Micro Devices, Inc. All rights reserved.
  * Written by Advanced Micro Devices,
  * See COPYING in top-level directory.
@@ -55,6 +55,9 @@ extern "C" {
  *
  * This function is currently only implemented in a meaningful way for
  * Linux; other systems will simply get a full cpuset.
+ *
+ * \return 0 on success.
+ * \return -1 on error, for instance if device information could not be found.
  */
 static __hwloc_inline int
 hwloc_rsmi_get_device_cpuset(hwloc_topology_t topology __hwloc_attribute_unused,

src/3rdparty/hwloc/include/hwloc/shmem.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2013-2018 Inria.  All rights reserved.
+ * Copyright © 2013-2023 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -48,6 +48,8 @@ extern "C" {
  * This length (in bytes) must be used in hwloc_shmem_topology_write()
  * and hwloc_shmem_topology_adopt() later.
  *
+ * \return the length, or -1 on error, for instance if flags are invalid.
+ *
  * \note Flags \p flags are currently unused, must be 0.
  */
 HWLOC_DECLSPEC int hwloc_shmem_topology_get_length(hwloc_topology_t topology,
@@ -74,9 +76,10 @@ HWLOC_DECLSPEC int hwloc_shmem_topology_get_length(hwloc_topology_t topology,
  * is not. However the caller may also allocate it manually in shared memory
  * to share it as well.
  *
- * \return -1 with errno set to EBUSY if the virtual memory mapping defined
+ * \return 0 on success.
+ * \return -1 with errno set to \c EBUSY if the virtual memory mapping defined
  * by \p mmap_address and \p length isn't available in the process.
- * \return -1 with errno set to EINVAL if \p fileoffset, \p mmap_address
+ * \return -1 with errno set to \c EINVAL if \p fileoffset, \p mmap_address
  * or \p length aren't page-aligned.
  */
 HWLOC_DECLSPEC int hwloc_shmem_topology_write(hwloc_topology_t topology,
@@ -112,14 +115,16 @@ HWLOC_DECLSPEC int hwloc_shmem_topology_write(hwloc_topology_t topology,
  *
  * \note This function takes care of calling hwloc_topology_abi_check().
  *
- * \return -1 with errno set to EBUSY if the virtual memory mapping defined
+ * \return 0 on success.
+ *
+ * \return -1 with errno set to \c EBUSY if the virtual memory mapping defined
  * by \p mmap_address and \p length isn't available in the process.
  *
- * \return -1 with errno set to EINVAL if \p fileoffset, \p mmap_address
+ * \return -1 with errno set to \c EINVAL if \p fileoffset, \p mmap_address
  * or \p length aren't page-aligned, or do not match what was given to
  * hwloc_shmem_topology_write() earlier.
  *
- * \return -1 with errno set to EINVAL if the layout of the topology structure
+ * \return -1 with errno set to \c EINVAL if the layout of the topology structure
  * is different between the writer process and the adopter process.
  */
 HWLOC_DECLSPEC int hwloc_shmem_topology_adopt(hwloc_topology_t *topologyp,

src/3rdparty/hwloc/include/private/cpuid-x86.h

@@ -11,6 +11,22 @@
 #ifndef HWLOC_PRIVATE_CPUID_X86_H
 #define HWLOC_PRIVATE_CPUID_X86_H
 
+/* A macro for annotating memory as uninitialized when building with MSAN
+ * (and otherwise having no effect). See below for why this is used with
+ * our custom assembly.
+ */
+#ifdef __has_feature
+#define HWLOC_HAS_FEATURE(name) __has_feature(name)
+#else
+#define HWLOC_HAS_FEATURE(name) 0
+#endif
+#if HWLOC_HAS_FEATURE(memory_sanitizer) || defined(MEMORY_SANITIZER)
+#include <sanitizer/msan_interface.h>
+#define HWLOC_ANNOTATE_MEMORY_IS_INITIALIZED(ptr, len) __msan_unpoison(ptr, len)
+#else
+#define HWLOC_ANNOTATE_MEMORY_IS_INITIALIZED(ptr, len)
+#endif
+
 #if (defined HWLOC_X86_32_ARCH) && (!defined HWLOC_HAVE_MSVC_CPUIDEX)
 static __hwloc_inline int hwloc_have_x86_cpuid(void)
 {
@@ -71,12 +87,18 @@ static __hwloc_inline void hwloc_x86_cpuid(unsigned *eax, unsigned *ebx, unsigne
   "movl %k2,%1\n\t"
   : "+a" (*eax), "=m" (*ebx), "=&r"(sav_rbx),
     "+c" (*ecx), "=&d" (*edx));
+  /* MSAN does not recognize the effect of the above assembly on the memory operand
+   * (`"=m"(*ebx)`). This may get improved in MSAN at some point in the future, e.g.
+   * see https://github.com/llvm/llvm-project/pull/77393. */
+  HWLOC_ANNOTATE_MEMORY_IS_INITIALIZED(ebx, sizeof *ebx);
 #elif defined(HWLOC_X86_32_ARCH)
   __asm__(
   "mov %%ebx,%1\n\t"
   "cpuid\n\t"
   "xchg %%ebx,%1\n\t"
   : "+a" (*eax), "=&SD" (*ebx), "+c" (*ecx), "=&d" (*edx));
+  /* See above. */
+  HWLOC_ANNOTATE_MEMORY_IS_INITIALIZED(ebx, sizeof *ebx);
 #else
 #error unknown architecture
 #endif

src/3rdparty/hwloc/include/private/misc.h

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2019 Inria.  All rights reserved.
+ * Copyright © 2009-2024 Inria.  All rights reserved.
  * Copyright © 2009-2012 Université Bordeaux
  * Copyright © 2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -186,7 +186,7 @@ hwloc_ffsl_from_ffs32(unsigned long x)
 /**
  * flsl helpers.
  */
-#ifdef __GNUC_____
+#ifdef __GNUC__
 
 #  if (__GNUC__ >= 4) || ((__GNUC__ == 3) && (__GNUC_MINOR__ >= 4))
 #    define hwloc_flsl(x) ((x) ? (8*sizeof(long) - __builtin_clzl(x)) : 0)
@@ -573,4 +573,35 @@ typedef SSIZE_T ssize_t;
 #  endif
 #endif
 
+static __inline float
+hwloc__pci_link_speed(unsigned generation, unsigned lanes)
+{
+  float lanespeed;
+  /*
+   * These are single-direction bandwidths only.
+   *
+   * Gen1 used NRZ with 8/10 encoding.
+   * PCIe Gen1 = 2.5GT/s signal-rate per lane x 8/10    =  0.25GB/s data-rate per lane
+   * PCIe Gen2 = 5  GT/s signal-rate per lane x 8/10    =  0.5 GB/s data-rate per lane
+   * Gen3 switched to NRZ with 128/130 encoding.
+   * PCIe Gen3 = 8  GT/s signal-rate per lane x 128/130 =  1   GB/s data-rate per lane
+   * PCIe Gen4 = 16 GT/s signal-rate per lane x 128/130 =  2   GB/s data-rate per lane
+   * PCIe Gen5 = 32 GT/s signal-rate per lane x 128/130 =  4   GB/s data-rate per lane
+   * Gen6 switched to PAM with with 242/256 FLIT (242B payload protected by 8B CRC + 6B FEC).
+   * PCIe Gen6 = 64 GT/s signal-rate per lane x 242/256 =  8   GB/s data-rate per lane
+   * PCIe Gen7 = 128GT/s signal-rate per lane x 242/256 = 16   GB/s data-rate per lane
+   */
+
+  /* lanespeed in Gbit/s */
+  if (generation <= 2)
+    lanespeed = 2.5f * generation * 0.8f;
+  else if (generation <= 5)
+    lanespeed = 8.0f * (1<<(generation-3)) * 128/130;
+  else
+    lanespeed = 8.0f * (1<<(generation-3)) * 242/256; /* assume Gen8 will be 256 GT/s and so on */
+
+  /* linkspeed in GB/s */
+  return lanespeed * lanes / 8;
+}
+
 #endif /* HWLOC_PRIVATE_MISC_H */

src/3rdparty/hwloc/include/private/netloc.h

@@ -1,578 +0,0 @@
-/*
- * Copyright © 2014 Cisco Systems, Inc.  All rights reserved.
- * Copyright © 2013-2014 University of Wisconsin-La Crosse.
- *                         All rights reserved.
- * Copyright © 2015-2017 Inria.  All rights reserved.
- *
- * $COPYRIGHT$
- *
- * Additional copyrights may follow
- * See COPYING in top-level directory.
- *
- * $HEADER$
- */
-
-#ifndef _NETLOC_PRIVATE_H_
-#define _NETLOC_PRIVATE_H_
-
-#include <hwloc.h>
-#include <netloc.h>
-#include <netloc/uthash.h>
-#include <netloc/utarray.h>
-#include <private/autogen/config.h>
-
-#define NETLOCFILE_VERSION 1
-
-#ifdef NETLOC_SCOTCH
-#include <stdint.h>
-#include <scotch.h>
-#define NETLOC_int SCOTCH_Num
-#else
-#define NETLOC_int int
-#endif
-
-/*
- * "Import" a few things from hwloc
- */
-#define __netloc_attribute_unused __hwloc_attribute_unused
-#define __netloc_attribute_malloc __hwloc_attribute_malloc
-#define __netloc_attribute_const __hwloc_attribute_const
-#define __netloc_attribute_pure __hwloc_attribute_pure
-#define __netloc_attribute_deprecated __hwloc_attribute_deprecated
-#define __netloc_attribute_may_alias __hwloc_attribute_may_alias
-#define NETLOC_DECLSPEC HWLOC_DECLSPEC
-
-
-/**********************************************************************
- * Types
- **********************************************************************/
-
-/**
- * Definitions for Comparators
- * \sa These are the return values from the following functions:
- *     netloc_network_compare, netloc_dt_edge_t_compare, netloc_dt_node_t_compare
- */
-typedef enum {
-    NETLOC_CMP_SAME    =  0,  /**< Compared as the Same */
-    NETLOC_CMP_SIMILAR = -1,  /**< Compared as Similar, but not the Same */
-    NETLOC_CMP_DIFF    = -2   /**< Compared as Different */
-} netloc_compare_type_t;
-
-/**
- * Enumerated type for the various types of supported networks
- */
-typedef enum {
-    NETLOC_NETWORK_TYPE_ETHERNET    = 1, /**< Ethernet network */
-    NETLOC_NETWORK_TYPE_INFINIBAND  = 2, /**< InfiniBand network */
-    NETLOC_NETWORK_TYPE_INVALID     = 3  /**< Invalid network */
-} netloc_network_type_t;
-
-/**
- * Enumerated type for the various types of supported topologies
- */
-typedef enum {
-    NETLOC_TOPOLOGY_TYPE_INVALID = -1, /**< Invalid */
-    NETLOC_TOPOLOGY_TYPE_TREE    = 1,  /**< Tree */
-} netloc_topology_type_t;
-
-/**
- * Enumerated type for the various types of nodes
- */
-typedef enum {
-    NETLOC_NODE_TYPE_HOST    = 0, /**< Host (a.k.a., network addressable endpoint - e.g., MAC Address) node */
-    NETLOC_NODE_TYPE_SWITCH  = 1, /**< Switch node */
-    NETLOC_NODE_TYPE_INVALID = 2  /**< Invalid node */
-} netloc_node_type_t;
-
-typedef enum {
-    NETLOC_ARCH_TREE    =  0,  /* Fat tree */
-} netloc_arch_type_t;
-
-
-/* Pre declarations to avoid inter dependency problems */
-/** \cond IGNORE */
-struct netloc_topology_t;
-typedef struct netloc_topology_t netloc_topology_t;
-struct netloc_node_t;
-typedef struct netloc_node_t netloc_node_t;
-struct netloc_edge_t;
-typedef struct netloc_edge_t netloc_edge_t;
-struct netloc_physical_link_t;
-typedef struct netloc_physical_link_t netloc_physical_link_t;
-struct netloc_path_t;
-typedef struct netloc_path_t netloc_path_t;
-
-struct netloc_arch_tree_t;
-typedef struct netloc_arch_tree_t netloc_arch_tree_t;
-struct netloc_arch_node_t;
-typedef struct netloc_arch_node_t netloc_arch_node_t;
-struct netloc_arch_node_slot_t;
-typedef struct netloc_arch_node_slot_t netloc_arch_node_slot_t;
-struct netloc_arch_t;
-typedef struct netloc_arch_t netloc_arch_t;
-/** \endcond */
-
-/**
- * \struct netloc_topology_t
- * \brief Netloc Topology Context
- *
- * An opaque data structure used to reference a network topology.
- *
- * \note Must be initialized with \ref netloc_topology_construct()
- */
-struct netloc_topology_t {
-    /** Topology path */
-    char *topopath;
-    /** Subnet ID */
-    char *subnet_id;
-
-    /** Node List */
-    netloc_node_t *nodes; /* Hash table of nodes by physical_id */
-    netloc_node_t *nodesByHostname; /* Hash table of nodes by hostname */
-
-    netloc_physical_link_t *physical_links; /* Hash table with physcial links */
-
-    /** Partition List */
-    UT_array *partitions;
-
-    /** Hwloc topology List */
-    char *hwlocpath;
-    UT_array *topos;
-    hwloc_topology_t *hwloc_topos;
-
-    /** Type of the graph */
-    netloc_topology_type_t type;
-};
-
-/**
- * \brief Netloc Node Type
- *
- * Represents the concept of a node (a.k.a., vertex, endpoint) within a network
- * graph. This could be a server or a network switch. The \ref node_type parameter
- * will distinguish the exact type of node this represents in the graph.
- */
-struct netloc_node_t {
-    UT_hash_handle hh;       /* makes this structure hashable with physical_id */
-    UT_hash_handle hh2;      /* makes this structure hashable with hostname */
-
-    /** Physical ID of the node */
-    char physical_id[20];
-
-    /** Logical ID of the node (if any) */
-    int logical_id;
-
-    /** Type of the node */
-    netloc_node_type_t type;
-
-    /* Pointer to physical_links */
-    UT_array *physical_links;
-
-    /** Description information from discovery (if any) */
-    char *description;
-
-    /**
-     * Application-given private data pointer.
-     * Initialized to NULL, and not used by the netloc library.
-     */
-    void * userdata;
-
-    /** Outgoing edges from this node */
-    netloc_edge_t *edges;
-
-    UT_array *subnodes; /* the group of nodes for the virtual nodes */
-
-    netloc_path_t *paths;
-
-    char *hostname;
-
-    UT_array *partitions; /* index in the list from the topology */
-
-    hwloc_topology_t hwlocTopo;
-    int hwlocTopoIdx;
-};
-
-/**
- * \brief Netloc Edge Type
- *
- * Represents the concept of a directed edge within a network graph.
- *
- * \note We do not point to the netloc_node_t structure directly to
- * simplify the representation, and allow the information to more easily
- * be entered into the data store without circular references.
- * \todo JJH Is the note above still true?
- */
-struct netloc_edge_t {
-    UT_hash_handle hh;       /* makes this structure hashable */
-
-    netloc_node_t *dest;
-
-    int id;
-
-    /** Pointers to the parent node */
-    netloc_node_t *node;
-
-    /* Pointer to physical_links */
-    UT_array *physical_links;
-
-    /** total gbits of the links */
-    float total_gbits;
-
-    UT_array *partitions; /* index in the list from the topology */
-
-    UT_array *subnode_edges; /* for edges going to virtual nodes */
-
-    struct netloc_edge_t *other_way;
-
-    /**
-     * Application-given private data pointer.
-     * Initialized to NULL, and not used by the netloc library.
-     */
-    void * userdata;
-};
-
-
-struct netloc_physical_link_t {
-    UT_hash_handle hh;       /* makes this structure hashable */
-
-    int id; // TODO long long
-    netloc_node_t *src;
-    netloc_node_t *dest;
-    int ports[2];
-    char *width;
-    char *speed;
-
-    netloc_edge_t *edge;
-
-    int other_way_id;
-    struct netloc_physical_link_t *other_way;
-
-    UT_array *partitions; /* index in the list from the topology */
-
-    /** gbits of the link from speed and width */
-    float gbits;
-
-    /** Description information from discovery (if any) */
-    char *description;
-};
-
-struct netloc_path_t {
-    UT_hash_handle hh;       /* makes this structure hashable */
-    char dest_id[20];
-    UT_array *links;
-};
-
-
-/**********************************************************************
- *        Architecture structures
- **********************************************************************/
-struct netloc_arch_tree_t {
-    NETLOC_int num_levels;
-    NETLOC_int *degrees;
-    NETLOC_int *cost;
-};
-
-struct netloc_arch_node_t {
-    UT_hash_handle hh;       /* makes this structure hashable */
-    char *name; /* Hash key */
-    netloc_node_t *node; /* Corresponding node */
-    int idx_in_topo; /* idx with ghost hosts to have complete topo */
-    int num_slots; /* it is not the real number of slots but the maximum slot idx */
-    int *slot_idx; /* corresponding idx in slot_tree */
-    int *slot_os_idx; /* corresponding os index for each leaf in tree */
-    netloc_arch_tree_t *slot_tree; /* Tree built from hwloc */
-    int num_current_slots; /* Number of PUs */
-    NETLOC_int *current_slots; /* indices in the complete tree */
-    int *slot_ranks; /* corresponding MPI rank for each leaf in tree */
-};
-
-struct netloc_arch_node_slot_t {
-    netloc_arch_node_t *node;
-    int slot;
-};
-
-struct netloc_arch_t {
-    netloc_topology_t *topology;
-    int has_slots; /* if slots are included in the architecture */
-    netloc_arch_type_t type;
-    union {
-        netloc_arch_tree_t *node_tree;
-        netloc_arch_tree_t *global_tree;
-    } arch;
-    netloc_arch_node_t *nodes_by_name;
-    netloc_arch_node_slot_t *node_slot_by_idx; /* node_slot by index in complete topo */
-    NETLOC_int num_current_hosts; /* if has_slots, host is a slot, else host is a node */
-    NETLOC_int *current_hosts; /* indices in the complete topology */
-};
-
-/**********************************************************************
- * Topology Functions
- **********************************************************************/
-/**
- * Allocate a topology handle.
- *
- * User is responsible for calling \ref netloc_detach on the topology handle.
- * The network parameter information is deep copied into the topology handle, so the
- * user may destruct the network handle after calling this function and/or reuse
- * the network handle.
- *
- * \returns NETLOC_SUCCESS on success
- * \returns NETLOC_ERROR upon an error.
- */
-netloc_topology_t *netloc_topology_construct(char *path);
-
-/**
- * Destruct a topology handle
- *
- * \param topology A valid pointer to a \ref netloc_topology_t handle created
- * from a prior call to \ref netloc_topology_construct.
- *
- * \returns NETLOC_SUCCESS on success
- * \returns NETLOC_ERROR upon an error.
- */
-int netloc_topology_destruct(netloc_topology_t *topology);
-
-int netloc_topology_find_partition_idx(netloc_topology_t *topology, char *partition_name);
-
-int netloc_topology_read_hwloc(netloc_topology_t *topology, int num_nodes,
-        netloc_node_t **node_list);
-
-#define netloc_topology_iter_partitions(topology,partition) \
-    for ((partition) = (char **)utarray_front(topology->partitions); \
-            (partition) != NULL; \
-            (partition) = (char **)utarray_next(topology->partitions, partition))
-
-#define netloc_topology_iter_hwloctopos(topology,hwloctopo) \
-    for ((hwloctopo) = (char **)utarray_front(topology->topos); \
-            (hwloctopo) != NULL; \
-            (hwloctopo) = (char **)utarray_next(topology->topos, hwloctopo))
-
-#define netloc_topology_find_node(topology,node_id,node) \
-    HASH_FIND_STR(topology->nodes, node_id, node)
-
-#define netloc_topology_iter_nodes(topology,node,_tmp) \
-    HASH_ITER(hh, topology->nodes, node, _tmp)
-
-#define netloc_topology_num_nodes(topology) \
-    HASH_COUNT(topology->nodes)
-
-/*************************************************/
-
-
-/**
- * Constructor for netloc_node_t
- *
- * User is responsible for calling the destructor on the handle.
- *
- * Returns
- *   A newly allocated pointer to the network information.
- */
-netloc_node_t *netloc_node_construct(void);
-
-/**
- * Destructor for netloc_node_t
- *
- * \param node A valid node handle
- *
- * Returns
- *   NETLOC_SUCCESS on success
- *   NETLOC_ERROR on error
- */
-int netloc_node_destruct(netloc_node_t *node);
-
-char *netloc_node_pretty_print(netloc_node_t* node);
-
-#define netloc_node_get_num_subnodes(node) \
-    utarray_len((node)->subnodes)
-
-#define netloc_node_get_subnode(node,i) \
-    (*(netloc_node_t **)utarray_eltptr((node)->subnodes, (i)))
-
-#define netloc_node_get_num_edges(node) \
-    utarray_len((node)->edges)
-
-#define netloc_node_get_edge(node,i) \
-    (*(netloc_edge_t **)utarray_eltptr((node)->edges, (i)))
-
-#define netloc_node_iter_edges(node,edge,_tmp) \
-    HASH_ITER(hh, node->edges, edge, _tmp)
-
-#define netloc_node_iter_paths(node,path,_tmp) \
-    HASH_ITER(hh, node->paths, path, _tmp)
-
-#define netloc_node_is_host(node) \
-    (node->type == NETLOC_NODE_TYPE_HOST)
-
-#define netloc_node_is_switch(node) \
-    (node->type == NETLOC_NODE_TYPE_SWITCH)
-
-#define netloc_node_iter_paths(node, path,_tmp) \
-    HASH_ITER(hh, node->paths, path, _tmp)
-
-int netloc_node_is_in_partition(netloc_node_t *node, int partition);
-
-/*************************************************/
-
-
-/**
- * Constructor for netloc_edge_t
- *
- * User is responsible for calling the destructor on the handle.
- *
- * Returns
- *   A newly allocated pointer to the edge information.
- */
-netloc_edge_t *netloc_edge_construct(void);
-
-/**
- * Destructor for netloc_edge_t
- *
- * \param edge A valid edge handle
- *
- * Returns
- *   NETLOC_SUCCESS on success
- *   NETLOC_ERROR on error
- */
-int netloc_edge_destruct(netloc_edge_t *edge);
-
-char * netloc_edge_pretty_print(netloc_edge_t* edge);
-
-void netloc_edge_reset_uid(void);
-
-int netloc_edge_is_in_partition(netloc_edge_t *edge, int partition);
-
-#define netloc_edge_get_num_links(edge) \
-    utarray_len((edge)->physical_links)
-
-#define netloc_edge_get_link(edge,i) \
-    (*(netloc_physical_link_t **)utarray_eltptr((edge)->physical_links, (i)))
-
-#define netloc_edge_get_num_subedges(edge) \
-    utarray_len((edge)->subnode_edges)
-
-#define netloc_edge_get_subedge(edge,i) \
-    (*(netloc_edge_t **)utarray_eltptr((edge)->subnode_edges, (i)))
-
-/*************************************************/
-
-
-/**
- * Constructor for netloc_physical_link_t
- *
- * User is responsible for calling the destructor on the handle.
- *
- * Returns
- *   A newly allocated pointer to the physical link information.
- */
-netloc_physical_link_t * netloc_physical_link_construct(void);
-
-/**
- * Destructor for netloc_physical_link_t
- *
- * Returns
- *   NETLOC_SUCCESS on success
- *   NETLOC_ERROR on error
- */
-int netloc_physical_link_destruct(netloc_physical_link_t *link);
-
-char * netloc_link_pretty_print(netloc_physical_link_t* link);
-
-/*************************************************/
-
-
-netloc_path_t *netloc_path_construct(void);
-int netloc_path_destruct(netloc_path_t *path);
-
-
-/**********************************************************************
- *        Architecture functions
- **********************************************************************/
-
-netloc_arch_t * netloc_arch_construct(void);
-
-int netloc_arch_destruct(netloc_arch_t *arch);
-
-int netloc_arch_build(netloc_arch_t *arch, int add_slots);
-
-int netloc_arch_set_current_resources(netloc_arch_t *arch);
-
-int netloc_arch_set_global_resources(netloc_arch_t *arch);
-
-int netloc_arch_node_get_hwloc_info(netloc_arch_node_t *arch);
-
-void netloc_arch_tree_complete(netloc_arch_tree_t *tree, UT_array **down_degrees_by_level,
-        int num_hosts, int **parch_idx);
-
-NETLOC_int netloc_arch_tree_num_leaves(netloc_arch_tree_t *tree);
-
-
-/**********************************************************************
- *        Access functions of various elements of the topology
- **********************************************************************/
-
-#define netloc_get_num_partitions(object) \
-    utarray_len((object)->partitions)
-
-#define netloc_get_partition(object,i) \
-    (*(int *)utarray_eltptr((object)->partitions, (i)))
-
-
-#define netloc_path_iter_links(path,link) \
-    for ((link) = (netloc_physical_link_t **)utarray_front(path->links); \
-            (link) != NULL; \
-            (link) = (netloc_physical_link_t **)utarray_next(path->links, link))
-
-/**********************************************************************
- *        Misc functions
- **********************************************************************/
-
-/**
- * Decode the network type
- *
- * \param net_type A valid member of the \ref netloc_network_type_t type
- *
- * \returns NULL if the type is invalid
- * \returns A string for that \ref netloc_network_type_t type
- */
-static inline const char * netloc_network_type_decode(netloc_network_type_t net_type) {
-    if( NETLOC_NETWORK_TYPE_ETHERNET == net_type ) {
-        return "ETH";
-    }
-    else if( NETLOC_NETWORK_TYPE_INFINIBAND == net_type ) {
-        return "IB";
-    }
-    else {
-        return NULL;
-    }
-}
-
-/**
- * Decode the node type
- *
- * \param node_type A valid member of the \ref netloc_node_type_t type
- *
- * \returns NULL if the type is invalid
- * \returns A string for that \ref netloc_node_type_t type
- */
-static inline const char * netloc_node_type_decode(netloc_node_type_t node_type) {
-    if( NETLOC_NODE_TYPE_SWITCH == node_type ) {
-        return "SW";
-    }
-    else if( NETLOC_NODE_TYPE_HOST == node_type ) {
-        return "CA";
-    }
-    else {
-        return NULL;
-    }
-}
-
-ssize_t netloc_line_get(char **lineptr, size_t *n, FILE *stream);
-
-char *netloc_line_get_next_token(char **string, char c);
-
-int netloc_build_comm_mat(char *filename, int *pn, double ***pmat);
-
-#define STRDUP_IF_NOT_NULL(str) (NULL == str ? NULL : strdup(str))
-#define STR_EMPTY_IF_NULL(str) (NULL == str ? "" : str)
-
-
-#endif // _NETLOC_PRIVATE_H_

src/3rdparty/hwloc/include/private/private.h

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009      CNRS
- * Copyright © 2009-2022 Inria.  All rights reserved.
+ * Copyright © 2009-2025 Inria.  All rights reserved.
  * Copyright © 2009-2012, 2020 Université Bordeaux
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
  *
@@ -245,6 +245,12 @@ struct hwloc_topology {
    * temporary variables during discovery
    */
 
+  /* set to 1 at the beginning of load() if the filter of any cpu cache type (L1 to L3i) is not NONE,
+   * may be checked by backends before querying caches
+   * (when they don't know the level of caches they are querying).
+   */
+  int want_some_cpu_caches;
+
   /* machine-wide memory.
    * temporarily stored there by OSes that only provide this without NUMA information,
    * and actually used later by the core.
@@ -296,6 +302,9 @@ extern void hwloc__reorder_children(hwloc_obj_t parent);
 extern void hwloc_topology_setup_defaults(struct hwloc_topology *topology);
 extern void hwloc_topology_clear(struct hwloc_topology *topology);
 
+#define _HWLOC_RECONNECT_FLAG_KEEPSTRUCTURE (1UL<<0)
+extern int hwloc__reconnect(struct hwloc_topology *topology, unsigned long flags);
+
 /* insert memory object as memory child of normal parent */
 extern struct hwloc_obj * hwloc__attach_memory_object(struct hwloc_topology *topology, hwloc_obj_t parent,
                                                       hwloc_obj_t obj, const char *reason);
@@ -420,7 +429,7 @@ extern void hwloc_internal_memattrs_need_refresh(hwloc_topology_t topology);
 extern void hwloc_internal_memattrs_refresh(hwloc_topology_t topology);
 extern int hwloc_internal_memattrs_dup(hwloc_topology_t new, hwloc_topology_t old);
 extern int hwloc_internal_memattr_set_value(hwloc_topology_t topology, hwloc_memattr_id_t id, hwloc_obj_type_t target_type, hwloc_uint64_t target_gp_index, unsigned target_os_index, struct hwloc_internal_location_s *initiator, hwloc_uint64_t value);
-extern int hwloc_internal_memattrs_guess_memory_tiers(hwloc_topology_t topology);
+extern int hwloc_internal_memattrs_guess_memory_tiers(hwloc_topology_t topology, int force_subtype);
 
 extern void hwloc_internal_cpukinds_init(hwloc_topology_t topology);
 extern int hwloc_internal_cpukinds_rank(hwloc_topology_t topology);
@@ -477,6 +486,7 @@ extern char * hwloc_progname(struct hwloc_topology *topology);
 #define HWLOC_GROUP_KIND_INTEL_DIE			104	/* no subkind */
 #define HWLOC_GROUP_KIND_S390_BOOK			110	/* subkind 0 is book, subkind 1 is drawer (group of books) */
 #define HWLOC_GROUP_KIND_AMD_COMPUTE_UNIT		120	/* no subkind */
+#define HWLOC_GROUP_KIND_AMD_COMPLEX                    121     /* no subkind */
 /* then, OS-specific groups */
 #define HWLOC_GROUP_KIND_SOLARIS_PG_HW_PERF		200	/* subkind is group width */
 #define HWLOC_GROUP_KIND_AIX_SDL_UNKNOWN		210	/* subkind is SDL level */

src/3rdparty/hwloc/include/private/xml.h

@@ -19,13 +19,14 @@ HWLOC_DECLSPEC int hwloc__xml_verbose(void);
 typedef struct hwloc__xml_import_state_s {
   struct hwloc__xml_import_state_s *parent;
 
-  /* globals shared because the entire stack of states during import */
+  /* globals shared between the entire stack of states during import */
   struct hwloc_xml_backend_data_s *global;
 
   /* opaque data used to store backend-specific data.
    * statically allocated to allow stack-allocation by the common code without knowing actual backend needs.
+   * libxml is 3 ptrs. nolibxml is 3 ptr + one int.
    */
-  char data[32];
+  char data[4 * SIZEOF_VOID_P];
 } * hwloc__xml_import_state_t;
 
 struct hwloc__xml_imported_v1distances_s {
@@ -74,8 +75,9 @@ typedef struct hwloc__xml_export_state_s {
 
   /* opaque data used to store backend-specific data.
    * statically allocated to allow stack-allocation by the common code without knowing actual backend needs.
+   * libxml is 1 ptr. nolibxml is 1 ptr + 2 size_t + 3 ints.
    */
-  char data[40];
+  char data[6 * SIZEOF_VOID_P];
 } * hwloc__xml_export_state_t;
 
 HWLOC_DECLSPEC void hwloc__xml_export_topology(hwloc__xml_export_state_t parentstate, hwloc_topology_t topology, unsigned long flags);

src/3rdparty/hwloc/src/bind.c

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2020 Inria.  All rights reserved.
+ * Copyright © 2009-2024 Inria.  All rights reserved.
  * Copyright © 2009-2010, 2012 Université Bordeaux
  * Copyright © 2011-2015 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -287,6 +287,7 @@ static __hwloc_inline int hwloc__check_membind_policy(hwloc_membind_policy_t pol
       || policy == HWLOC_MEMBIND_FIRSTTOUCH
       || policy == HWLOC_MEMBIND_BIND
       || policy == HWLOC_MEMBIND_INTERLEAVE
+      || policy == HWLOC_MEMBIND_WEIGHTED_INTERLEAVE
       || policy == HWLOC_MEMBIND_NEXTTOUCH)
     return 0;
   return -1;

src/3rdparty/hwloc/src/bitmap.c

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2020 Inria.  All rights reserved.
+ * Copyright © 2009-2024 Inria.  All rights reserved.
  * Copyright © 2009-2011 Université Bordeaux
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -245,6 +245,7 @@ int hwloc_bitmap_copy(struct hwloc_bitmap_s * dst, const struct hwloc_bitmap_s *
 /* Strings always use 32bit groups */
 #define HWLOC_PRIxSUBBITMAP		"%08lx"
 #define HWLOC_BITMAP_SUBSTRING_SIZE	32
+#define HWLOC_BITMAP_SUBSTRING_FULL_VALUE 0xFFFFFFFFUL
 #define HWLOC_BITMAP_SUBSTRING_LENGTH	(HWLOC_BITMAP_SUBSTRING_SIZE/4)
 #define HWLOC_BITMAP_STRING_PER_LONG	(HWLOC_BITS_PER_LONG/HWLOC_BITMAP_SUBSTRING_SIZE)
 
@@ -261,6 +262,7 @@ int hwloc_bitmap_snprintf(char * __hwloc_restrict buf, size_t buflen, const stru
   const unsigned long accum_mask = ~0UL;
 #else /* HWLOC_BITS_PER_LONG != HWLOC_BITMAP_SUBSTRING_SIZE */
   const unsigned long accum_mask = ((1UL << HWLOC_BITMAP_SUBSTRING_SIZE) - 1) << (HWLOC_BITS_PER_LONG - HWLOC_BITMAP_SUBSTRING_SIZE);
+  int merge_with_infinite_prefix = 0;
 #endif /* HWLOC_BITS_PER_LONG != HWLOC_BITMAP_SUBSTRING_SIZE */
 
   HWLOC__BITMAP_CHECK(set);
@@ -279,6 +281,9 @@ int hwloc_bitmap_snprintf(char * __hwloc_restrict buf, size_t buflen, const stru
       res = size>0 ? (int)size - 1 : 0;
     tmp += res;
     size -= res;
+#if HWLOC_BITS_PER_LONG > HWLOC_BITMAP_SUBSTRING_SIZE
+    merge_with_infinite_prefix = 1;
+#endif
   }
 
   i=(int) set->ulongs_count-1;
@@ -294,16 +299,24 @@ int hwloc_bitmap_snprintf(char * __hwloc_restrict buf, size_t buflen, const stru
   }
 
   while (i>=0 || accumed) {
+    unsigned long value;
+
     /* Refill accumulator */
     if (!accumed) {
       accum = set->ulongs[i--];
       accumed = HWLOC_BITS_PER_LONG;
     }
+    value = (accum & accum_mask) >> (HWLOC_BITS_PER_LONG - HWLOC_BITMAP_SUBSTRING_SIZE);
 
-    if (accum & accum_mask) {
+#if HWLOC_BITS_PER_LONG > HWLOC_BITMAP_SUBSTRING_SIZE
+    if (merge_with_infinite_prefix && value == HWLOC_BITMAP_SUBSTRING_FULL_VALUE) {
+      /* first full subbitmap merged with infinite prefix */
+      res = 0;
+    } else
+#endif
+    if (value) {
       /* print the whole subset if not empty */
-        res = hwloc_snprintf(tmp, size, needcomma ? ",0x" HWLOC_PRIxSUBBITMAP : "0x" HWLOC_PRIxSUBBITMAP,
+      res = hwloc_snprintf(tmp, size, needcomma ? ",0x" HWLOC_PRIxSUBBITMAP : "0x" HWLOC_PRIxSUBBITMAP, value);
-		     (accum & accum_mask) >> (HWLOC_BITS_PER_LONG - HWLOC_BITMAP_SUBSTRING_SIZE));
       needcomma = 1;
     } else if (i == -1 && accumed == HWLOC_BITMAP_SUBSTRING_SIZE) {
       /* print a single 0 to mark the last subset */
@@ -323,6 +336,7 @@ int hwloc_bitmap_snprintf(char * __hwloc_restrict buf, size_t buflen, const stru
 #else
     accum <<= HWLOC_BITMAP_SUBSTRING_SIZE;
     accumed -= HWLOC_BITMAP_SUBSTRING_SIZE;
+    merge_with_infinite_prefix = 0;
 #endif
 
     if (res >= size)
@@ -362,7 +376,8 @@ int hwloc_bitmap_sscanf(struct hwloc_bitmap_s *set, const char * __hwloc_restric
 {
   const char * current = string;
   unsigned long accum = 0;
-  int count=0;
+  int count = 0;
+  int ulongcount;
   int infinite = 0;
 
   /* count how many substrings there are */
@@ -383,9 +398,20 @@ int hwloc_bitmap_sscanf(struct hwloc_bitmap_s *set, const char * __hwloc_restric
     count--;
   }
 
-  if (hwloc_bitmap_reset_by_ulongs(set, (count + HWLOC_BITMAP_STRING_PER_LONG - 1) / HWLOC_BITMAP_STRING_PER_LONG) < 0)
+  ulongcount = (count + HWLOC_BITMAP_STRING_PER_LONG - 1) / HWLOC_BITMAP_STRING_PER_LONG;
+  if (hwloc_bitmap_reset_by_ulongs(set, ulongcount) < 0)
     return -1;
-  set->infinite = 0;
+
+  set->infinite = 0; /* will be updated later */
+
+#if HWLOC_BITS_PER_LONG != HWLOC_BITMAP_SUBSTRING_SIZE
+  if (infinite && (count % HWLOC_BITMAP_STRING_PER_LONG) != 0) {
+    /* accumulate substrings of the first ulong that are hidden in the infinite prefix */
+    int i;
+    for(i = (count % HWLOC_BITMAP_STRING_PER_LONG); i < HWLOC_BITMAP_STRING_PER_LONG; i++)
+      accum |= (HWLOC_BITMAP_SUBSTRING_FULL_VALUE << (i*HWLOC_BITMAP_SUBSTRING_SIZE));
+  }
+#endif
 
   while (*current != '\0') {
     unsigned long val;
@@ -544,6 +570,9 @@ int hwloc_bitmap_taskset_snprintf(char * __hwloc_restrict buf, size_t buflen, co
   ssize_t size = buflen;
   char *tmp = buf;
   int res, ret = 0;
+#if HWLOC_BITS_PER_LONG == 64
+  int merge_with_infinite_prefix = 0;
+#endif
   int started = 0;
   int i;
 
@@ -563,6 +592,9 @@ int hwloc_bitmap_taskset_snprintf(char * __hwloc_restrict buf, size_t buflen, co
       res = size>0 ? (int)size - 1 : 0;
     tmp += res;
     size -= res;
+#if HWLOC_BITS_PER_LONG == 64
+    merge_with_infinite_prefix = 1;
+#endif
   }
 
   i=set->ulongs_count-1;
@@ -582,7 +614,11 @@ int hwloc_bitmap_taskset_snprintf(char * __hwloc_restrict buf, size_t buflen, co
     if (started) {
       /* print the whole subset */
 #if HWLOC_BITS_PER_LONG == 64
+      if (merge_with_infinite_prefix && (val & 0xffffffff00000000UL) == 0xffffffff00000000UL) {
+        res = hwloc_snprintf(tmp, size, "%08lx", val & 0xffffffffUL);
+      } else  {
         res = hwloc_snprintf(tmp, size, "%016lx", val);
+      }
 #else
       res = hwloc_snprintf(tmp, size, "%08lx", val);
 #endif
@@ -599,6 +635,9 @@ int hwloc_bitmap_taskset_snprintf(char * __hwloc_restrict buf, size_t buflen, co
       res = size>0 ? (int)size - 1 : 0;
     tmp += res;
     size -= res;
+#if HWLOC_BITS_PER_LONG == 64
+    merge_with_infinite_prefix = 0;
+#endif
   }
 
   /* if didn't display anything, display 0x0 */
@@ -679,6 +718,10 @@ int hwloc_bitmap_taskset_sscanf(struct hwloc_bitmap_s *set, const char * __hwloc
       goto failed;
 
     set->ulongs[count-1] = val;
+    if (infinite && tmpchars != HWLOC_BITS_PER_LONG/4) {
+      /* infinite prefix with partial substring, fill remaining bits */
+      set->ulongs[count-1] |= (~0ULL)<<(4*tmpchars);
+    }
 
     current += tmpchars;
     chars -= tmpchars;

src/3rdparty/hwloc/src/components.c

@@ -94,8 +94,7 @@ static hwloc_dlhandle hwloc_dlopenext(const char *_filename)
 {
   hwloc_dlhandle handle;
   char *filename = NULL;
-  (void) asprintf(&filename, "%s.so", _filename);
+  if (asprintf(&filename, "%s.so", _filename) < 0)
-  if (!filename)
     return NULL;
   handle = dlopen(filename, RTLD_NOW|RTLD_LOCAL);
   free(filename);

src/3rdparty/hwloc/src/cpukinds.c

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2020-2022 Inria.  All rights reserved.
+ * Copyright © 2020-2024 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -50,6 +50,7 @@ hwloc_internal_cpukinds_dup(hwloc_topology_t new, hwloc_topology_t old)
     return -1;
   new->cpukinds = kinds;
   new->nr_cpukinds = old->nr_cpukinds;
+  new->nr_cpukinds_allocated = old->nr_cpukinds;
   memcpy(kinds, old->cpukinds, old->nr_cpukinds * sizeof(*kinds));
 
   for(i=0;i<old->nr_cpukinds; i++) {

src/3rdparty/hwloc/src/diff.c

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2013-2022 Inria.  All rights reserved.
+ * Copyright © 2013-2023 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -411,6 +411,30 @@ int hwloc_topology_diff_build(hwloc_topology_t topo1,
           }
         }
 
+        if (!err) {
+          /* cpukinds */
+          if (topo1->nr_cpukinds != topo2->nr_cpukinds)
+            goto roottoocomplex;
+          for(i=0; i<topo1->nr_cpukinds; i++) {
+            struct hwloc_internal_cpukind_s *ic1 = &topo1->cpukinds[i];
+            struct hwloc_internal_cpukind_s *ic2 = &topo2->cpukinds[i];
+            unsigned j;
+            if (!hwloc_bitmap_isequal(ic1->cpuset, ic2->cpuset)
+                || ic1->efficiency != ic2->efficiency
+                || ic1->forced_efficiency != ic2->forced_efficiency
+                || ic1->ranking_value != ic2->ranking_value
+                || ic1->nr_infos != ic2->nr_infos)
+              goto roottoocomplex;
+            for(j=0; j<ic1->nr_infos; j++) {
+              struct hwloc_info_s *info1 = &ic1->infos[j], *info2 = &ic2->infos[j];
+              if (strcmp(info1->name, info2->name)
+                  || strcmp(info1->value, info2->value)) {
+                goto roottoocomplex;
+              }
+            }
+          }
+        }
+
 	return err;
 
  roottoocomplex:

src/3rdparty/hwloc/src/distances.c

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2010-2022 Inria.  All rights reserved.
+ * Copyright © 2010-2025 Inria.  All rights reserved.
  * Copyright © 2011-2012 Université Bordeaux
  * Copyright © 2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -624,8 +624,8 @@ void * hwloc_distances_add_create(hwloc_topology_t topology,
     return NULL;
   }
   if ((kind & ~HWLOC_DISTANCES_KIND_ALL)
-      || hwloc_weight_long(kind & HWLOC_DISTANCES_KIND_FROM_ALL) != 1
+      || hwloc_weight_long(kind & HWLOC_DISTANCES_KIND_FROM_ALL) > 1
-      || hwloc_weight_long(kind & HWLOC_DISTANCES_KIND_MEANS_ALL) != 1) {
+      || hwloc_weight_long(kind & HWLOC_DISTANCES_KIND_MEANS_ALL) > 1) {
     errno = EINVAL;
     return NULL;
   }
@@ -699,7 +699,7 @@ hwloc_distances_add_commit(hwloc_topology_t topology,
   }
 
   /* in case we added some groups, see if we need to reconnect */
-  hwloc_topology_reconnect(topology, 0);
+  hwloc__reconnect(topology, 0);
 
   return 0;
 
@@ -1387,19 +1387,12 @@ static __hwloc_inline int is_nvswitch(hwloc_obj_t obj)
 }
 
 static int
-hwloc__distances_transform_merge_switch_ports(hwloc_topology_t topology,
+hwloc__distances_transform_merge_switch_ports(struct hwloc_distances_s *distances)
-                                              struct hwloc_distances_s *distances)
 {
-  struct hwloc_internal_distances_s *dist = hwloc__internal_distances_from_public(topology, distances);
   hwloc_obj_t *objs = distances->objs;
   hwloc_uint64_t *values = distances->values;
   unsigned first, i, j, nbobjs = distances->nbobjs;
 
-  if (strcmp(dist->name, "NVLinkBandwidth")) {
-    errno = EINVAL;
-    return -1;
-  }
-
   /* find the first port */
   first = (unsigned) -1;
   for(i=0; i<nbobjs; i++)
@@ -1435,20 +1428,13 @@ hwloc__distances_transform_merge_switch_ports(hwloc_topology_t topology,
 }
 
 static int
-hwloc__distances_transform_transitive_closure(hwloc_topology_t topology,
+hwloc__distances_transform_transitive_closure(struct hwloc_distances_s *distances)
-                                              struct hwloc_distances_s *distances)
 {
-  struct hwloc_internal_distances_s *dist = hwloc__internal_distances_from_public(topology, distances);
   hwloc_obj_t *objs = distances->objs;
   hwloc_uint64_t *values = distances->values;
   unsigned nbobjs = distances->nbobjs;
   unsigned i, j, k;
 
-  if (strcmp(dist->name, "NVLinkBandwidth")) {
-    errno = EINVAL;
-    return -1;
-  }
-
   for(i=0; i<nbobjs; i++) {
     hwloc_uint64_t bw_i2sw = 0;
     if (is_nvswitch(objs[i]))
@@ -1467,8 +1453,8 @@ hwloc__distances_transform_transitive_closure(hwloc_topology_t topology,
         if (is_nvswitch(objs[k]))
           bw_sw2j += values[k*nbobjs+j];
 
-      /* bandwidth from i to j is now min(i2sw,sw2j) */
+      /* bandwidth from i to j now gets indirect bandwidth too, min(i2sw,sw2j) */
-      values[i*nbobjs+j] = bw_i2sw > bw_sw2j ? bw_sw2j : bw_i2sw;
+      values[i*nbobjs+j] += bw_i2sw > bw_sw2j ? bw_sw2j : bw_i2sw;
     }
   }
 
@@ -1476,7 +1462,7 @@ hwloc__distances_transform_transitive_closure(hwloc_topology_t topology,
 }
 
 int
-hwloc_distances_transform(hwloc_topology_t topology,
+hwloc_distances_transform(hwloc_topology_t topology __hwloc_attribute_unused,
                           struct hwloc_distances_s *distances,
                           enum hwloc_distances_transform_e transform,
                           void *transform_attr,
@@ -1495,13 +1481,13 @@ hwloc_distances_transform(hwloc_topology_t topology,
   case HWLOC_DISTANCES_TRANSFORM_MERGE_SWITCH_PORTS:
   {
     int err;
-    err = hwloc__distances_transform_merge_switch_ports(topology, distances);
+    err = hwloc__distances_transform_merge_switch_ports(distances);
     if (!err)
       err = hwloc__distances_transform_remove_null(distances);
     return err;
   }
   case HWLOC_DISTANCES_TRANSFORM_TRANSITIVE_CLOSURE:
-    return hwloc__distances_transform_transitive_closure(topology, distances);
+    return hwloc__distances_transform_transitive_closure(distances);
   default:
     errno = EINVAL;
     return -1;

src/3rdparty/hwloc/src/pci-common.c

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2009-2022 Inria.  All rights reserved.
+ * Copyright © 2009-2024 Inria.  All rights reserved.
  * See COPYING in top-level directory.
  */
 
@@ -886,36 +886,12 @@ hwloc_pcidisc_find_linkspeed(const unsigned char *config,
 			     unsigned offset, float *linkspeed)
 {
   unsigned linksta, speed, width;
-  float lanespeed;
 
   memcpy(&linksta, &config[offset + HWLOC_PCI_EXP_LNKSTA], 4);
   speed = linksta & HWLOC_PCI_EXP_LNKSTA_SPEED; /* PCIe generation */
   width = (linksta & HWLOC_PCI_EXP_LNKSTA_WIDTH) >> 4; /* how many lanes */
-  /*
-   * These are single-direction bandwidths only.
-   *
-   * Gen1 used NRZ with 8/10 encoding.
-   * PCIe Gen1 = 2.5GT/s signal-rate per lane x 8/10    =  0.25GB/s data-rate per lane
-   * PCIe Gen2 = 5  GT/s signal-rate per lane x 8/10    =  0.5 GB/s data-rate per lane
-   * Gen3 switched to NRZ with 128/130 encoding.
-   * PCIe Gen3 = 8  GT/s signal-rate per lane x 128/130 =  1   GB/s data-rate per lane
-   * PCIe Gen4 = 16 GT/s signal-rate per lane x 128/130 =  2   GB/s data-rate per lane
-   * PCIe Gen5 = 32 GT/s signal-rate per lane x 128/130 =  4   GB/s data-rate per lane
-   * Gen6 switched to PAM with with 242/256 FLIT (242B payload protected by 8B CRC + 6B FEC).
-   * PCIe Gen6 = 64 GT/s signal-rate per lane x 242/256 =  8   GB/s data-rate per lane
-   * PCIe Gen7 = 128GT/s signal-rate per lane x 242/256 = 16   GB/s data-rate per lane
-   */
 
-  /* lanespeed in Gbit/s */
+  *linkspeed = hwloc__pci_link_speed(speed, width);
-  if (speed <= 2)
-    lanespeed = 2.5f * speed * 0.8f;
-  else if (speed <= 5)
-    lanespeed = 8.0f * (1<<(speed-3)) * 128/130;
-  else
-    lanespeed = 8.0f * (1<<(speed-3)) * 242/256; /* assume Gen8 will be 256 GT/s and so on */
-
-  /* linkspeed in GB/s */
-  *linkspeed = lanespeed * width / 8;
   return 0;
 }
 

src/3rdparty/hwloc/src/shmem.c

@@ -23,6 +23,7 @@ struct hwloc_shmem_header {
   uint32_t header_length; /* where the actual topology starts in the file/mapping */
   uint64_t mmap_address; /* virtual address to pass to mmap */
   uint64_t mmap_length; /* length to pass to mmap (includes the header) */
+  /* we will pad the end to a multiple of pointer size so that the topology is well aligned */
 };
 
 #define HWLOC_SHMEM_MALLOC_ALIGN 8UL
@@ -85,6 +86,7 @@ hwloc_shmem_topology_write(hwloc_topology_t topology,
   hwloc_topology_t new;
   struct hwloc_tma tma;
   struct hwloc_shmem_header header;
+  uint32_t header_length = (sizeof(header) + sizeof(void*) - 1) & ~(sizeof(void*) - 1); /* pad to a multiple of pointer size */
   void *mmap_res;
   int err;
 
@@ -100,7 +102,7 @@ hwloc_shmem_topology_write(hwloc_topology_t topology,
   hwloc_internal_memattrs_refresh(topology);
 
   header.header_version = HWLOC_SHMEM_HEADER_VERSION;
-  header.header_length = sizeof(header);
+  header.header_length = header_length;
   header.mmap_address = (uintptr_t) mmap_address;
   header.mmap_length = length;
 
@@ -127,7 +129,7 @@ hwloc_shmem_topology_write(hwloc_topology_t topology,
 
   tma.malloc = tma_shmem_malloc;
   tma.dontfree = 1;
-  tma.data = (char *)mmap_res + sizeof(header);
+  tma.data = (char *)mmap_res + header_length;
   err = hwloc__topology_dup(&new, topology, &tma);
   if (err < 0)
     return err;
@@ -154,6 +156,7 @@ hwloc_shmem_topology_adopt(hwloc_topology_t *topologyp,
 {
   hwloc_topology_t new, old;
   struct hwloc_shmem_header header;
+  uint32_t header_length = (sizeof(header) + sizeof(void*) - 1) & ~(sizeof(void*) - 1); /* pad to a multiple of pointer size */
   void *mmap_res;
   int err;
 
@@ -171,7 +174,7 @@ hwloc_shmem_topology_adopt(hwloc_topology_t *topologyp,
     return -1;
 
   if (header.header_version != HWLOC_SHMEM_HEADER_VERSION
-      || header.header_length != sizeof(header)
+      || header.header_length != header_length
       || header.mmap_address != (uintptr_t) mmap_address
       || header.mmap_length != length) {
     errno = EINVAL;
@@ -186,7 +189,7 @@ hwloc_shmem_topology_adopt(hwloc_topology_t *topologyp,
     goto out_with_mmap;
   }
 
-  old = (hwloc_topology_t)((char*)mmap_address + sizeof(header));
+  old = (hwloc_topology_t)((char*)mmap_address + header_length);
   if (hwloc_topology_abi_check(old) < 0) {
     errno = EINVAL;
     goto out_with_mmap;

src/3rdparty/hwloc/src/topology-synthetic.c

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2022 Inria.  All rights reserved.
+ * Copyright © 2009-2023 Inria.  All rights reserved.
  * Copyright © 2009-2010 Université Bordeaux
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -23,6 +23,7 @@ struct hwloc_synthetic_attr_s {
   unsigned depth; /* For caches/groups */
   hwloc_obj_cache_type_t cachetype; /* For caches */
   hwloc_uint64_t memorysize; /* For caches/memory */
+  hwloc_uint64_t memorysidecachesize; /* Single level of memory-side-cache in-front of a NUMA node */
 };
 
 struct hwloc_synthetic_indexes_s {
@@ -380,6 +381,9 @@ hwloc_synthetic_parse_attrs(const char *attrs, const char **next_posp,
     } else if (!iscache && !strncmp("memory=", attrs, 7)) {
       memorysize = hwloc_synthetic_parse_memory_attr(attrs+7, &attrs);
 
+    } else if (!strncmp("memorysidecachesize=", attrs, 20)) {
+      sattr->memorysidecachesize = hwloc_synthetic_parse_memory_attr(attrs+20, &attrs);
+
     } else if (!strncmp("indexes=", attrs, 8)) {
       index_string = attrs+8;
       attrs += 8;
@@ -387,10 +391,9 @@ hwloc_synthetic_parse_attrs(const char *attrs, const char **next_posp,
       attrs += index_string_length;
 
     } else {
-      if (verbose)
+      size_t length = strcspn(attrs, " )");
-	fprintf(stderr, "Unknown attribute at '%s'\n", attrs);
+      fprintf(stderr, "hwloc/synthetic: Ignoring unknown attribute at '%s'\n", attrs);
-      errno = EINVAL;
+      attrs += length;
-      return -1;
     }
 
     if (' ' == *attrs)
@@ -416,6 +419,32 @@ hwloc_synthetic_parse_attrs(const char *attrs, const char **next_posp,
   return 0;
 }
 
+static void
+hwloc_synthetic_set_default_attrs(struct hwloc_synthetic_attr_s *sattr,
+                                  int *type_count)
+{
+  hwloc_obj_type_t type = sattr->type;
+
+  if (type == HWLOC_OBJ_GROUP) {
+    if (sattr->depth == (unsigned)-1)
+      sattr->depth = type_count[HWLOC_OBJ_GROUP]--;
+
+  } else if (hwloc__obj_type_is_cache(type)) {
+    if (!sattr->memorysize) {
+      if (1 == sattr->depth)
+        /* 32KiB in L1 */
+        sattr->memorysize = 32*1024;
+      else
+        /* *4 at each level, starting from 1MiB for L2, unified */
+        sattr->memorysize = 256ULL*1024 << (2*sattr->depth);
+    }
+
+  } else if (type == HWLOC_OBJ_NUMANODE && !sattr->memorysize) {
+    /* 1GiB in memory nodes. */
+    sattr->memorysize = 1024*1024*1024;
+  }
+}
+
 /* frees level until arity = 0 */
 static void
 hwloc_synthetic_free_levels(struct hwloc_synthetic_backend_data_s *data)
@@ -465,6 +494,7 @@ hwloc_backend_synthetic_init(struct hwloc_synthetic_backend_data_s *data,
   data->level[0].indexes.string = NULL;
   data->level[0].indexes.array = NULL;
   data->level[0].attr.memorysize = 0;
+  data->level[0].attr.memorysidecachesize = 0;
   data->level[0].attached = NULL;
   type_count[HWLOC_OBJ_MACHINE] = 1;
   if (*description == '(') {
@@ -514,6 +544,7 @@ hwloc_backend_synthetic_init(struct hwloc_synthetic_backend_data_s *data,
       if (attached) {
 	attached->attr.type = type;
 	attached->attr.memorysize = 0;
+	attached->attr.memorysidecachesize = 0;
 	/* attached->attr.depth and .cachetype unused */
 	attached->next = NULL;
 	pprev = &data->level[count-1].attached;
@@ -601,7 +632,7 @@ hwloc_backend_synthetic_init(struct hwloc_synthetic_backend_data_s *data,
     }
     if (!item) {
       if (verbose)
-	fprintf(stderr,"Synthetic string with disallow 0 number of objects at '%s'\n", pos);
+	fprintf(stderr,"Synthetic string with disallowed 0 number of objects at '%s'\n", pos);
       errno = EINVAL;
       goto error;
     }
@@ -611,6 +642,7 @@ hwloc_backend_synthetic_init(struct hwloc_synthetic_backend_data_s *data,
     data->level[count].indexes.string = NULL;
     data->level[count].indexes.array = NULL;
     data->level[count].attr.memorysize = 0;
+    data->level[count].attr.memorysidecachesize = 0;
     if (*next_pos == '(') {
       err = hwloc_synthetic_parse_attrs(next_pos+1, &next_pos, &data->level[count].attr, &data->level[count].indexes, verbose);
       if (err < 0)
@@ -796,6 +828,7 @@ hwloc_backend_synthetic_init(struct hwloc_synthetic_backend_data_s *data,
     data->level[1].indexes.string = NULL;
     data->level[1].indexes.array = NULL;
     data->level[1].attr.memorysize = 0;
+    data->level[1].attr.memorysidecachesize = 0;
     data->level[1].totalwidth = data->level[0].totalwidth;
     /* update arity to insert a single NUMA node per parent */
     data->level[1].arity = data->level[0].arity;
@@ -803,30 +836,14 @@ hwloc_backend_synthetic_init(struct hwloc_synthetic_backend_data_s *data,
     count++;
   }
 
+  /* set default attributes that depend on the depth/hierarchy of levels */
   for (i=0; i<count; i++) {
+    struct hwloc_synthetic_attached_s *attached;
     struct hwloc_synthetic_level_data_s *curlevel = &data->level[i];
-    hwloc_obj_type_t type = curlevel->attr.type;
+    hwloc_synthetic_set_default_attrs(&curlevel->attr, type_count);
-
+    for(attached = curlevel->attached; attached != NULL; attached = attached->next)
-    if (type == HWLOC_OBJ_GROUP) {
+      hwloc_synthetic_set_default_attrs(&attached->attr, type_count);
-      if (curlevel->attr.depth == (unsigned)-1)
+    hwloc_synthetic_process_indexes(data, &curlevel->indexes, curlevel->totalwidth, verbose);
-	curlevel->attr.depth = type_count[HWLOC_OBJ_GROUP]--;
-
-    } else if (hwloc__obj_type_is_cache(type)) {
-      if (!curlevel->attr.memorysize) {
-	if (1 == curlevel->attr.depth)
-	  /* 32KiB in L1 */
-	  curlevel->attr.memorysize = 32*1024;
-	else
-	  /* *4 at each level, starting from 1MiB for L2, unified */
-	  curlevel->attr.memorysize = 256ULL*1024 << (2*curlevel->attr.depth);
-      }
-
-    } else if (type == HWLOC_OBJ_NUMANODE && !curlevel->attr.memorysize) {
-      /* 1GiB in memory nodes. */
-      curlevel->attr.memorysize = 1024*1024*1024;
-    }
-
-    hwloc_synthetic_process_indexes(data, &data->level[i].indexes, data->level[i].totalwidth, verbose);
   }
 
   hwloc_synthetic_process_indexes(data, &data->numa_attached_indexes, data->numa_attached_nr, verbose);
@@ -859,6 +876,12 @@ hwloc_synthetic_set_attr(struct hwloc_synthetic_attr_s *sattr,
     obj->attr->numanode.page_types[0].size = 4096;
     obj->attr->numanode.page_types[0].count = sattr->memorysize / 4096;
     break;
+  case HWLOC_OBJ_MEMCACHE:
+    obj->attr->cache.depth = 1;
+    obj->attr->cache.linesize = 64;
+    obj->attr->cache.type = HWLOC_OBJ_CACHE_UNIFIED;
+    obj->attr->cache.size = sattr->memorysidecachesize;
+    break;
   case HWLOC_OBJ_PACKAGE:
   case HWLOC_OBJ_DIE:
     break;
@@ -926,6 +949,14 @@ hwloc_synthetic_insert_attached(struct hwloc_topology *topology,
 
   hwloc__insert_object_by_cpuset(topology, NULL, child, "synthetic:attached");
 
+  if (attached->attr.memorysidecachesize) {
+    hwloc_obj_t mscachechild = hwloc_alloc_setup_object(topology, HWLOC_OBJ_MEMCACHE, HWLOC_UNKNOWN_INDEX);
+    mscachechild->cpuset = hwloc_bitmap_dup(set);
+    mscachechild->nodeset = hwloc_bitmap_dup(child->nodeset);
+    hwloc_synthetic_set_attr(&attached->attr, mscachechild);
+    hwloc__insert_object_by_cpuset(topology, NULL, mscachechild, "synthetic:attached:mscache");
+  }
+
   hwloc_synthetic_insert_attached(topology, data, attached->next, set);
 }
 
@@ -977,6 +1008,14 @@ hwloc__look_synthetic(struct hwloc_topology *topology,
     hwloc_synthetic_set_attr(&curlevel->attr, obj);
 
     hwloc__insert_object_by_cpuset(topology, NULL, obj, "synthetic");
+
+    if (type == HWLOC_OBJ_NUMANODE && curlevel->attr.memorysidecachesize) {
+      hwloc_obj_t mscachechild = hwloc_alloc_setup_object(topology, HWLOC_OBJ_MEMCACHE, HWLOC_UNKNOWN_INDEX);
+      mscachechild->cpuset = hwloc_bitmap_dup(set);
+      mscachechild->nodeset = hwloc_bitmap_dup(obj->nodeset);
+      hwloc_synthetic_set_attr(&curlevel->attr, mscachechild);
+      hwloc__insert_object_by_cpuset(topology, NULL, mscachechild, "synthetic:mscache");
+    }
   }
 
   hwloc_synthetic_insert_attached(topology, data, curlevel->attached, set);
@@ -1217,6 +1256,7 @@ hwloc__export_synthetic_indexes(hwloc_obj_t *level, unsigned total,
 
 static int
 hwloc__export_synthetic_obj_attr(struct hwloc_topology * topology,
+                                 unsigned long flags,
 				 hwloc_obj_t obj,
 				 char *buffer, size_t buflen)
 {
@@ -1224,6 +1264,7 @@ hwloc__export_synthetic_obj_attr(struct hwloc_topology * topology,
   const char * prefix = "(";
   char cachesize[64] = "";
   char memsize[64] = "";
+  char memorysidecachesize[64] = "";
   int needindexes = 0;
 
   if (hwloc__obj_type_is_cache(obj->type) && obj->attr->cache.size) {
@@ -1236,6 +1277,19 @@ hwloc__export_synthetic_obj_attr(struct hwloc_topology * topology,
 	     prefix, (unsigned long long) obj->attr->numanode.local_memory);
     prefix = separator;
   }
+  if (obj->type == HWLOC_OBJ_NUMANODE && !(flags & HWLOC_TOPOLOGY_EXPORT_SYNTHETIC_FLAG_V1)) {
+    hwloc_obj_t memorysidecache = obj->parent;
+    hwloc_uint64_t size = 0;
+    while (memorysidecache && memorysidecache->type == HWLOC_OBJ_MEMCACHE) {
+      size += memorysidecache->attr->cache.size;
+      memorysidecache = memorysidecache->parent;
+    }
+    if (size) {
+      snprintf(memorysidecachesize, sizeof(memorysidecachesize), "%smemorysidecachesize=%llu",
+               prefix, (unsigned long long) size);
+      prefix = separator;
+    }
+  }
   if (!obj->logical_index /* only display indexes once per level (not for non-first NUMA children, etc.) */
       && (obj->type == HWLOC_OBJ_PU || obj->type == HWLOC_OBJ_NUMANODE)) {
     hwloc_obj_t cur = obj;
@@ -1247,12 +1301,12 @@ hwloc__export_synthetic_obj_attr(struct hwloc_topology * topology,
       cur = cur->next_cousin;
     }
   }
-  if (*cachesize || *memsize || needindexes) {
+  if (*cachesize || *memsize || *memorysidecachesize || needindexes) {
     ssize_t tmplen = buflen;
     char *tmp = buffer;
     int res, ret = 0;
 
-    res = hwloc_snprintf(tmp, tmplen, "%s%s%s", cachesize, memsize, needindexes ? "" : ")");
+    res = hwloc_snprintf(tmp, tmplen, "%s%s%s%s", cachesize, memsize, memorysidecachesize, needindexes ? "" : ")");
     if (hwloc__export_synthetic_update_status(&ret, &tmp, &tmplen, res) < 0)
       return -1;
 
@@ -1326,7 +1380,7 @@ hwloc__export_synthetic_obj(struct hwloc_topology * topology, unsigned long flag
 
   if (!(flags & HWLOC_TOPOLOGY_EXPORT_SYNTHETIC_FLAG_NO_ATTRS)) {
     /* obj attributes */
-    res = hwloc__export_synthetic_obj_attr(topology, obj, tmp, tmplen);
+    res = hwloc__export_synthetic_obj_attr(topology, flags, obj, tmp, tmplen);
     if (hwloc__export_synthetic_update_status(&ret, &tmp, &tmplen, res) < 0)
       return -1;
   }
@@ -1351,7 +1405,7 @@ hwloc__export_synthetic_memory_children(struct hwloc_topology * topology, unsign
 
   if (flags & HWLOC_TOPOLOGY_EXPORT_SYNTHETIC_FLAG_V1) {
     /* v1: export a single NUMA child */
-    if (parent->memory_arity > 1 || mchild->type != HWLOC_OBJ_NUMANODE) {
+    if (parent->memory_arity > 1) {
       /* not supported */
       if (verbose)
 	fprintf(stderr, "Cannot export to synthetic v1 if multiple memory children are attached to the same location.\n");
@@ -1362,6 +1416,9 @@ hwloc__export_synthetic_memory_children(struct hwloc_topology * topology, unsign
     if (needprefix)
       hwloc__export_synthetic_add_char(&ret, &tmp, &tmplen, ' ');
 
+    /* ignore memcaches and export the NUMA node */
+    while (mchild->type != HWLOC_OBJ_NUMANODE)
+      mchild = mchild->memory_first_child;
     res = hwloc__export_synthetic_obj(topology, flags, mchild, 1, tmp, tmplen);
     if (hwloc__export_synthetic_update_status(&ret, &tmp, &tmplen, res) < 0)
       return -1;
@@ -1369,16 +1426,25 @@ hwloc__export_synthetic_memory_children(struct hwloc_topology * topology, unsign
   }
 
   while (mchild) {
-    /* FIXME: really recurse to export memcaches and numanode,
+    /* The core doesn't support shared memcache for now (because ACPI and Linux don't).
+     * So, for each mchild here, recurse only in the first children at each level.
+     *
+     * FIXME: whenever supported by the core, really recurse to export memcaches and numanode,
      * but it requires clever parsing of [ memcache [numa] [numa] ] during import,
      * better attaching of things to describe the hierarchy.
      */
     hwloc_obj_t numanode = mchild;
-    /* only export the first NUMA node leaf of each memory child
+    /* Only export the first NUMA node leaf of each memory child.
-     * FIXME: This assumes mscache aren't shared between nodes, that's true in current platforms
+     * Memcaches are ignored here, they will be summed and exported as a single attribute
+     * of the NUMA node in hwloc__export_synthetic_obj().
      */
     while (numanode && numanode->type != HWLOC_OBJ_NUMANODE) {
-      assert(numanode->arity == 1);
+      if (verbose && numanode->memory_arity > 1) {
+        static int warned = 0;
+        if (!warned)
+          fprintf(stderr, "Ignoring non-first memory children at non-first level of memory hierarchy.\n");
+        warned = 1;
+      }
       numanode = numanode->memory_first_child;
     }
     assert(numanode); /* there's always a numanode at the bottom of the memory tree */
@@ -1511,17 +1577,21 @@ hwloc_topology_export_synthetic(struct hwloc_topology * topology,
 
   if (flags & HWLOC_TOPOLOGY_EXPORT_SYNTHETIC_FLAG_V1) {
     /* v1 requires all NUMA at the same level */
-    hwloc_obj_t node;
+    hwloc_obj_t node, parent;
     signed pdepth;
 
     node = hwloc_get_obj_by_type(topology, HWLOC_OBJ_NUMANODE, 0);
     assert(node);
-    assert(hwloc__obj_type_is_normal(node->parent->type)); /* only depth-1 memory children for now */
+    parent = node->parent;
-    pdepth = node->parent->depth;
+    while (!hwloc__obj_type_is_normal(parent->type))
+      parent = parent->parent;
+    pdepth = parent->depth;
 
     while ((node = node->next_cousin) != NULL) {
-      assert(hwloc__obj_type_is_normal(node->parent->type)); /* only depth-1 memory children for now */
+      parent = node->parent;
-      if (node->parent->depth != pdepth) {
+      while (!hwloc__obj_type_is_normal(parent->type))
+        parent = parent->parent;
+      if (parent->depth != pdepth) {
 	if (verbose)
 	  fprintf(stderr, "Cannot export to synthetic v1 if memory is attached to parents at different depths.\n");
 	errno = EINVAL;
@@ -1534,7 +1604,7 @@ hwloc_topology_export_synthetic(struct hwloc_topology * topology,
 
   if (!(flags & HWLOC_TOPOLOGY_EXPORT_SYNTHETIC_FLAG_NO_ATTRS)) {
     /* obj attributes */
-    res = hwloc__export_synthetic_obj_attr(topology, obj, tmp, tmplen);
+    res = hwloc__export_synthetic_obj_attr(topology, flags, obj, tmp, tmplen);
     if (res > 0)
       needprefix = 1;
     if (hwloc__export_synthetic_update_status(&ret, &tmp, &tmplen, res) < 0)

src/3rdparty/hwloc/src/topology-windows.c

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2022 Inria.  All rights reserved.
+ * Copyright © 2009-2025 Inria.  All rights reserved.
  * Copyright © 2009-2012, 2020 Université Bordeaux
  * Copyright © 2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -56,6 +56,9 @@ typedef enum _LOGICAL_PROCESSOR_RELATIONSHIP {
   RelationCache,
   RelationProcessorPackage,
   RelationGroup,
+  RelationProcessorDie,
+  RelationNumaNodeEx, /* only used to *request* extended numa info only, but included in RelationAll, never returned on output */
+  RelationProcessorModule,
   RelationAll = 0xffff
 } LOGICAL_PROCESSOR_RELATIONSHIP;
 #else /* HAVE_LOGICAL_PROCESSOR_RELATIONSHIP */
@@ -64,6 +67,11 @@ typedef enum _LOGICAL_PROCESSOR_RELATIONSHIP {
 #    define RelationGroup 4
 #    define RelationAll 0xffff
 #  endif /* HAVE_RELATIONPROCESSORPACKAGE */
+#  ifndef HAVE_RELATIONPROCESSORDIE
+#    define RelationProcessorDie 5
+#    define RelationNumaNodeEx 6
+#    define RelationProcessorModule 7
+#  endif
 #endif /* HAVE_LOGICAL_PROCESSOR_RELATIONSHIP */
 
 #ifndef HAVE_GROUP_AFFINITY
@@ -220,7 +228,7 @@ static void hwloc_win_get_function_ptrs(void)
 #pragma GCC diagnostic ignored "-Wcast-function-type"
 #endif
 
-    kernel32 = LoadLibrary("kernel32.dll");
+    kernel32 = LoadLibrary(TEXT("kernel32.dll"));
     if (kernel32) {
       GetActiveProcessorGroupCountProc =
 	(PFN_GETACTIVEPROCESSORGROUPCOUNT) GetProcAddress(kernel32, "GetActiveProcessorGroupCount");
@@ -249,12 +257,12 @@ static void hwloc_win_get_function_ptrs(void)
     }
 
     if (!QueryWorkingSetExProc) {
-      HMODULE psapi = LoadLibrary("psapi.dll");
+      HMODULE psapi = LoadLibrary(TEXT("psapi.dll"));
       if (psapi)
         QueryWorkingSetExProc = (PFN_QUERYWORKINGSETEX) GetProcAddress(psapi, "QueryWorkingSetEx");
     }
 
-    ntdll = GetModuleHandle("ntdll");
+    ntdll = GetModuleHandle(TEXT("ntdll"));
     RtlGetVersionProc = (PFN_RTLGETVERSION) GetProcAddress(ntdll, "RtlGetVersion");
 
 #if HWLOC_HAVE_GCC_W_CAST_FUNCTION_TYPE
@@ -366,8 +374,8 @@ hwloc_win_get_processor_groups(void)
   hwloc_debug("found %lu windows processor groups\n", nr_processor_groups);
 
   if (nr_processor_groups > 1 && SIZEOF_VOID_P == 4) {
-    if (HWLOC_SHOW_ALL_ERRORS())
+    if (HWLOC_SHOW_CRITICAL_ERRORS())
-      fprintf(stderr, "hwloc: multiple processor groups found on 32bits Windows, topology may be invalid/incomplete.\n");
+      fprintf(stderr, "hwloc/windows: multiple processor groups found on 32bits Windows, topology may be invalid/incomplete.\n");
   }
 
   length = 0;
@@ -987,7 +995,11 @@ hwloc_look_windows(struct hwloc_backend *backend, struct hwloc_disc_status *dsta
   OSVERSIONINFOEX osvi;
   char versionstr[20];
   char hostname[122] = "";
-  unsigned hostname_size = sizeof(hostname);
+#if !defined(__CYGWIN__)
+  DWORD hostname_size = sizeof(hostname);
+#else
+  size_t hostname_size = sizeof(hostname);
+#endif
   int has_efficiencyclass = 0;
   struct hwloc_win_efficiency_classes eclasses;
   char *env = getenv("HWLOC_WINDOWS_PROCESSOR_GROUP_OBJS");
@@ -1051,15 +1063,20 @@ hwloc_look_windows(struct hwloc_backend *backend, struct hwloc_disc_status *dsta
         unsigned efficiency_class = 0;
         GROUP_AFFINITY *GroupMask;
 
-        /* Ignore unknown caches */
+	if (procInfo->Relationship == RelationCache) {
-	if (procInfo->Relationship == RelationCache
+          if (!topology->want_some_cpu_caches)
-		&& procInfo->Cache.Type != CacheUnified
+            /* TODO: check if RelationAll&~RelationCache works? */
+            continue;
+          if (procInfo->Cache.Type != CacheUnified
               && procInfo->Cache.Type != CacheData
               && procInfo->Cache.Type != CacheInstruction)
+            /* Ignore unknown caches */
             continue;
+        }
 
 	id = HWLOC_UNKNOWN_INDEX;
 	switch (procInfo->Relationship) {
+          case RelationNumaNodeEx: /* only used on input anyway */
 	  case RelationNumaNode:
 	    type = HWLOC_OBJ_NUMANODE;
             /* Starting with Windows 11 and Server 2022, the GroupCount field is valid and >=1
@@ -1082,6 +1099,16 @@ hwloc_look_windows(struct hwloc_backend *backend, struct hwloc_disc_status *dsta
 	    num = procInfo->Processor.GroupCount;
 	    GroupMask = procInfo->Processor.GroupMask;
 	    break;
+	  case RelationProcessorDie:
+            type = HWLOC_OBJ_DIE;
+            num = procInfo->Processor.GroupCount;
+            GroupMask = procInfo->Processor.GroupMask;
+            break;
+	  case RelationProcessorModule:
+            type = HWLOC_OBJ_GROUP;
+            num = procInfo->Processor.GroupCount;
+            GroupMask = procInfo->Processor.GroupMask;
+            break;
 	  case RelationCache:
 	    type = (procInfo->Cache.Type == CacheInstruction ? HWLOC_OBJ_L1ICACHE : HWLOC_OBJ_L1CACHE) + procInfo->Cache.Level - 1;
             /* GroupCount added approximately with NumaNode.GroupCount above */
@@ -1203,6 +1230,19 @@ hwloc_look_windows(struct hwloc_backend *backend, struct hwloc_disc_status *dsta
 		continue;
 	    }
 	    break;
+          case HWLOC_OBJ_GROUP:
+            switch (procInfo->Relationship) {
+            case RelationGroup:
+              obj->attr->group.kind = HWLOC_GROUP_KIND_WINDOWS_PROCESSOR_GROUP;
+              break;
+            case RelationProcessorModule:
+              obj->attr->group.kind = HWLOC_GROUP_KIND_INTEL_MODULE;
+              obj->subtype = strdup("Module");
+              break;
+            default:
+              obj->attr->group.kind = HWLOC_GROUP_KIND_WINDOWS_RELATIONSHIP_UNKNOWN;
+            }
+            break;
 	  default:
 	    break;
 	}

src/3rdparty/hwloc/src/topology-x86.c

@@ -1,11 +1,11 @@
 /*
- * Copyright © 2010-2022 Inria.  All rights reserved.
+ * Copyright © 2010-2025 Inria.  All rights reserved.
  * Copyright © 2010-2013 Université Bordeaux
  * Copyright © 2010-2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
  *
  *
- * This backend is only used when the operating system does not export
+ * This backend is mostly used when the operating system does not export
  * the necessary hardware topology information to user-space applications.
  * Currently, FreeBSD and NetBSD only add PUs and then fallback to this
  * backend for CPU/Cache discovery.
@@ -15,6 +15,7 @@
  * on various architectures, without having to use this x86-specific code.
  * But this backend is still used after them to annotate some objects with
  * additional details (CPU info in Package, Inclusiveness in Caches).
+ * It may also be enabled manually to work-around bugs in native OS discovery.
  */
 
 #include "private/autogen/config.h"
@@ -38,6 +39,12 @@ struct hwloc_x86_backend_data_s {
   int apicid_unique;
   char *src_cpuiddump_path;
   int is_knl;
+  int is_hybrid;
+  int found_die_ids;
+  int found_complex_ids;
+  int found_unit_ids;
+  int found_module_ids;
+  int found_tile_ids;
 };
 
 /************************************
@@ -80,7 +87,7 @@ cpuiddump_read(const char *dirpath, unsigned idx)
 
   cpuiddump = malloc(sizeof(*cpuiddump));
   if (!cpuiddump) {
-    fprintf(stderr, "Failed to allocate cpuiddump for PU #%u, ignoring cpuiddump.\n", idx);
+    fprintf(stderr, "hwloc/x86: Failed to allocate cpuiddump for PU #%u, ignoring cpuiddump.\n", idx);
     goto out;
   }
 
@@ -91,7 +98,7 @@ cpuiddump_read(const char *dirpath, unsigned idx)
   snprintf(filename, filenamelen, "%s/pu%u", dirpath, idx);
   file = fopen(filename, "r");
   if (!file) {
-    fprintf(stderr, "Could not read dumped cpuid file %s, ignoring cpuiddump.\n", filename);
+    fprintf(stderr, "hwloc/x86: Could not read dumped cpuid file %s, ignoring cpuiddump.\n", filename);
     goto out_with_filename;
   }
 
@@ -100,7 +107,7 @@ cpuiddump_read(const char *dirpath, unsigned idx)
     nr++;
   cpuiddump->entries = malloc(nr * sizeof(struct cpuiddump_entry));
   if (!cpuiddump->entries) {
-    fprintf(stderr, "Failed to allocate %u cpuiddump entries for PU #%u, ignoring cpuiddump.\n", nr, idx);
+    fprintf(stderr, "hwloc/x86: Failed to allocate %u cpuiddump entries for PU #%u, ignoring cpuiddump.\n", nr, idx);
     goto out_with_file;
   }
 
@@ -156,7 +163,7 @@ cpuiddump_find_by_input(unsigned *eax, unsigned *ebx, unsigned *ecx, unsigned *e
     return;
   }
 
-  fprintf(stderr, "Couldn't find %x,%x,%x,%x in dumped cpuid, returning 0s.\n",
+  fprintf(stderr, "hwloc/x86: Couldn't find %x,%x,%x,%x in dumped cpuid, returning 0s.\n",
 	  *eax, *ebx, *ecx, *edx);
   *eax = 0;
   *ebx = 0;
@@ -210,7 +217,8 @@ struct procinfo {
 #define TILE 4
 #define MODULE 5
 #define DIE 6
-#define HWLOC_X86_PROCINFO_ID_NR 7
+#define COMPLEX 7
+#define HWLOC_X86_PROCINFO_ID_NR 8
   unsigned ids[HWLOC_X86_PROCINFO_ID_NR];
   unsigned *otherids;
   unsigned levels;
@@ -314,7 +322,7 @@ static void read_amd_caches_topoext(struct procinfo *infos, struct cpuiddump *sr
   /* the code below doesn't want any other cache yet */
   assert(!infos->numcaches);
 
-  for (cachenum = 0; ; cachenum++) {
+  for (cachenum = 0; cachenum<16 /* guard */; cachenum++) {
     eax = 0x8000001d;
     ecx = cachenum;
     cpuid_or_from_dump(&eax, &ebx, &ecx, &edx, src_cpuiddump);
@@ -325,7 +333,7 @@ static void read_amd_caches_topoext(struct procinfo *infos, struct cpuiddump *sr
 
   cache = infos->cache = malloc(infos->numcaches * sizeof(*infos->cache));
   if (cache) {
-    for (cachenum = 0; ; cachenum++) {
+    for (cachenum = 0; cachenum<16 /* guard */; cachenum++) {
       unsigned long linesize, linepart, ways, sets;
       eax = 0x8000001d;
       ecx = cachenum;
@@ -378,7 +386,7 @@ static void read_intel_caches(struct hwloc_x86_backend_data_s *data, struct proc
   unsigned cachenum;
   struct cacheinfo *cache;
 
-  for (cachenum = 0; ; cachenum++) {
+  for (cachenum = 0; cachenum<16 /* guard */; cachenum++) {
     eax = 0x04;
     ecx = cachenum;
     cpuid_or_from_dump(&eax, &ebx, &ecx, &edx, src_cpuiddump);
@@ -400,7 +408,7 @@ static void read_intel_caches(struct hwloc_x86_backend_data_s *data, struct proc
     infos->cache = tmpcaches;
     cache = &infos->cache[oldnumcaches];
 
-    for (cachenum = 0; ; cachenum++) {
+    for (cachenum = 0; cachenum<16 /* guard */; cachenum++) {
       unsigned long linesize, linepart, ways, sets;
       eax = 0x04;
       ecx = cachenum;
@@ -480,7 +488,7 @@ static void read_amd_cores_legacy(struct procinfo *infos, struct cpuiddump *src_
 }
 
 /* AMD unit/node from CPUID 0x8000001e leaf (topoext) */
-static void read_amd_cores_topoext(struct procinfo *infos, unsigned long flags, struct cpuiddump *src_cpuiddump)
+static void read_amd_cores_topoext(struct hwloc_x86_backend_data_s *data, struct procinfo *infos, unsigned long flags __hwloc_attribute_unused, struct cpuiddump *src_cpuiddump)
 {
   unsigned apic_id, nodes_per_proc = 0;
   unsigned eax, ebx, ecx, edx;
@@ -489,7 +497,6 @@ static void read_amd_cores_topoext(struct procinfo *infos, unsigned long flags,
   cpuid_or_from_dump(&eax, &ebx, &ecx, &edx, src_cpuiddump);
   infos->apicid = apic_id = eax;
 
-  if (flags & HWLOC_X86_DISC_FLAG_TOPOEXT_NUMANODES) {
     if (infos->cpufamilynumber == 0x16) {
       /* ecx is reserved */
       infos->ids[NODE] = 0;
@@ -504,12 +511,12 @@ static void read_amd_cores_topoext(struct procinfo *infos, unsigned long flags,
         || (infos->cpufamilynumber == 0x19 && nodes_per_proc > 1)) {
       hwloc_debug("warning: undefined nodes_per_proc value %u, assuming it means %u\n", nodes_per_proc, nodes_per_proc);
     }
-  }
 
   if (infos->cpufamilynumber <= 0x16) { /* topoext appeared in 0x15 and compute-units were only used in 0x15 and 0x16 */
     unsigned cores_per_unit;
     /* coreid was obtained from read_amd_cores_legacy() earlier */
     infos->ids[UNIT] = ebx & 0xff;
+    data->found_unit_ids = 1;
     cores_per_unit = ((ebx >> 8) & 0xff) + 1;
     hwloc_debug("topoext %08x, %u nodes, node %u, %u cores in unit %u\n", apic_id, nodes_per_proc, infos->ids[NODE], cores_per_unit, infos->ids[UNIT]);
     /* coreid and unitid are package-wide (core 0-15 and unit 0-7 on 16-core 2-NUMAnode processor).
@@ -524,19 +531,29 @@ static void read_amd_cores_topoext(struct procinfo *infos, unsigned long flags,
   }
 }
 
-/* Intel core/thread or even die/module/tile from CPUID 0x0b or 0x1f leaves (v1 and v2 extended topology enumeration) */
+/* Intel core/thread or even die/module/tile from CPUID 0x0b or 0x1f leaves (v1 and v2 extended topology enumeration)
-static void read_intel_cores_exttopoenum(struct procinfo *infos, unsigned leaf, struct cpuiddump *src_cpuiddump)
+ * or AMD core/thread or even complex/ccd from CPUID 0x0b or 0x80000026 (extended CPU topology)
+ */
+static void read_extended_topo(struct hwloc_x86_backend_data_s *data, struct procinfo *infos, unsigned leaf, enum cpuid_type cpuid_type __hwloc_attribute_unused, struct cpuiddump *src_cpuiddump)
 {
-  unsigned level, apic_nextshift, apic_number, apic_type, apic_id = 0, apic_shift = 0, id;
+  unsigned level, apic_nextshift, apic_type, apic_id = 0, apic_shift = 0, id;
   unsigned threadid __hwloc_attribute_unused = 0; /* shut-up compiler */
   unsigned eax, ebx, ecx = 0, edx;
   int apic_packageshift = 0;
 
-  for (level = 0; ; level++) {
+  for (level = 0; level<32 /* guard */; level++) {
     ecx = level;
     eax = leaf;
     cpuid_or_from_dump(&eax, &ebx, &ecx, &edx, src_cpuiddump);
-    if (!eax && !ebx)
+    /* Intel specifies that the 0x0b/0x1f loop should stop when we get "invalid domain" (0 in ecx[8:15])
+     * (if so, we also get 0 in eax/ebx for invalid subleaves). Zhaoxin implements this too.
+     * However AMD rather says that the 0x80000026/0x0b loop should stop when we get "no thread at this level" (0 in ebx[0:15]).
+     *
+     * Linux kernel <= 6.8 used "invalid domain" for both Intel and AMD (in detect_extended_topology())
+     * but x86 discovery revamp in 6.9 now properly checks both Intel and AMD conditions (in topo_subleaf()).
+     * So let's assume we are allowed to break-out once one of the Intel+AMD conditions is met.
+     */
+    if (!(ebx & 0xffff) || !(ecx & 0xff00))
       break;
     apic_packageshift = eax & 0x1f;
   }
@@ -545,42 +562,63 @@ static void read_intel_cores_exttopoenum(struct procinfo *infos, unsigned leaf,
     infos->otherids = malloc(level * sizeof(*infos->otherids));
     if (infos->otherids) {
       infos->levels = level;
-      for (level = 0; ; level++) {
+      for (level = 0; level<32 /* guard */; level++) {
 	ecx = level;
 	eax = leaf;
 	cpuid_or_from_dump(&eax, &ebx, &ecx, &edx, src_cpuiddump);
-	if (!eax && !ebx)
+        if (!(ebx & 0xffff) || !(ecx & 0xff00))
           break;
 	apic_nextshift = eax & 0x1f;
-	apic_number = ebx & 0xffff;
 	apic_type = (ecx & 0xff00) >> 8;
 	apic_id = edx;
 	id = (apic_id >> apic_shift) & ((1 << (apic_packageshift - apic_shift)) - 1);
-	hwloc_debug("x2APIC %08x %u: nextshift %u num %2u type %u id %2u\n", apic_id, level, apic_nextshift, apic_number, apic_type, id);
+	hwloc_debug("x2APIC %08x %u: nextshift %u nextnumber %2u type %u id %2u\n",
+                    apic_id,
+                    level,
+                    apic_nextshift,
+                    ebx & 0xffff /* number of threads in next level */,
+                    apic_type,
+                    id);
 	infos->apicid = apic_id;
 	infos->otherids[level] = UINT_MAX;
         switch (apic_type) {
         case 1:
           threadid = id;
-	  /* apic_number is the actual number of threads per core */
           break;
         case 2:
           infos->ids[CORE] = id;
-	  /* apic_number is the actual number of threads per die */
           break;
         case 3:
+          if (leaf == 0x80000026) {
+            data->found_complex_ids = 1;
+            infos->ids[COMPLEX] = id;
+          } else {
+            data->found_module_ids = 1;
             infos->ids[MODULE] = id;
-	  /* apic_number is the actual number of threads per tile */
+          }
           break;
         case 4:
+          if (leaf == 0x80000026) {
+            data->found_die_ids = 1;
+            infos->ids[DIE] = id;
+          } else {
+            data->found_tile_ids = 1;
             infos->ids[TILE] = id;
-	  /* apic_number is the actual number of threads per die */
+          }
           break;
         case 5:
+          if (leaf == 0x80000026) {
+            goto unknown_type;
+          } else {
+            data->found_die_ids = 1;
             infos->ids[DIE] = id;
-	  /* apic_number is the actual number of threads per package */
+          }
           break;
+        case 6:
+          /* TODO: "DieGrp" on Intel */
+          /* fallthrough */
         default:
+        unknown_type:
           hwloc_debug("x2APIC %u: unknown type %u\n", level, apic_type);
           infos->otherids[level] = apic_id >> apic_shift;
           break;
@@ -615,7 +653,13 @@ static void look_proc(struct hwloc_backend *backend, struct procinfo *infos, uns
   cpuid_or_from_dump(&eax, &ebx, &ecx, &edx, src_cpuiddump);
   infos->apicid = ebx >> 24;
   if (edx & (1 << 28)) {
-    legacy_max_log_proc = 1 << hwloc_flsl(((ebx >> 16) & 0xff) - 1);
+    unsigned ebx_16_23 = (ebx >> 16) & 0xff;
+    if (ebx_16_23) {
+      legacy_max_log_proc = 1 << hwloc_flsl(ebx_16_23 - 1);
+    } else {
+      hwloc_debug("HTT bit set in CPUID 0x01.edx, but legacy_max_proc = 0 in ebx, assuming legacy_max_log_proc = 1\n");
+      legacy_max_log_proc = 1;
+    }
   } else {
     hwloc_debug("HTT bit not set in CPUID 0x01.edx, assuming legacy_max_log_proc = 1\n");
     legacy_max_log_proc = 1;
@@ -704,12 +748,13 @@ static void look_proc(struct hwloc_backend *backend, struct procinfo *infos, uns
   }
 
   if (highest_cpuid >= 0x1a && has_hybrid(features)) {
-    /* Get hybrid cpu information from cpuid 0x1a */
+    /* Get hybrid cpu information from cpuid 0x1a on Intel */
     eax = 0x1a;
     ecx = 0;
     cpuid_or_from_dump(&eax, &ebx, &ecx, &edx, src_cpuiddump);
     infos->hybridcoretype = eax >> 24;
     infos->hybridnativemodel = eax & 0xffffff;
+    data->is_hybrid = 1;
   }
 
   /*********************************************************************************
@@ -731,23 +776,30 @@ static void look_proc(struct hwloc_backend *backend, struct procinfo *infos, uns
      *
      * Only needed when x2apic supported if NUMA nodes are needed.
      */
-    read_amd_cores_topoext(infos, flags, src_cpuiddump);
+    read_amd_cores_topoext(data, infos, flags, src_cpuiddump);
   }
 
-  if ((cpuid_type == intel) && highest_cpuid >= 0x1f) {
+  if ((cpuid_type == amd) && highest_ext_cpuid >= 0x80000026) {
+    /* Get socket/die/complex/core/thread information from cpuid 0x80000026
+     * (AMD Extended CPU Topology)
+     */
+    read_extended_topo(data, infos, 0x80000026, cpuid_type, src_cpuiddump);
+
+  } else if ((cpuid_type == intel || cpuid_type == zhaoxin) && highest_cpuid >= 0x1f) {
     /* Get package/die/module/tile/core/thread information from cpuid 0x1f
      * (Intel v2 Extended Topology Enumeration)
      */
-    read_intel_cores_exttopoenum(infos, 0x1f, src_cpuiddump);
+    read_extended_topo(data, infos, 0x1f, cpuid_type, src_cpuiddump);
 
   } else if ((cpuid_type == intel || cpuid_type == amd || cpuid_type == zhaoxin)
 	     && highest_cpuid >= 0x0b && has_x2apic(features)) {
     /* Get package/core/thread information from cpuid 0x0b
      * (Intel v1 Extended Topology Enumeration)
      */
-    read_intel_cores_exttopoenum(infos, 0x0b, src_cpuiddump);
+    read_extended_topo(data, infos, 0x0b, cpuid_type, src_cpuiddump);
   }
 
+ if (backend->topology->want_some_cpu_caches) {
   /**************************************
    * Get caches from CPU-specific leaves
    */
@@ -845,6 +897,7 @@ static void look_proc(struct hwloc_backend *backend, struct procinfo *infos, uns
       }
     }
   }
+ }
 
   if (hwloc_bitmap_isset(data->apicid_set, infos->apicid))
     data->apicid_unique = 0;
@@ -1046,21 +1099,34 @@ static void summarize(struct hwloc_backend *backend, struct procinfo *infos, uns
 
   if (hwloc_filter_check_keep_object_type(topology, HWLOC_OBJ_GROUP)) {
     if (fulldiscovery) {
+      if (data->found_unit_ids) {
+        /* Look for AMD Complex inside packages */
+        hwloc_bitmap_copy(remaining_cpuset, complete_cpuset);
+        hwloc_x86_add_groups(topology, infos, nbprocs, remaining_cpuset,
+                             COMPLEX, "Complex",
+                             HWLOC_GROUP_KIND_AMD_COMPLEX, 0);
+      }
+      if (data->found_unit_ids) {
         /* Look for AMD Compute units inside packages */
         hwloc_bitmap_copy(remaining_cpuset, complete_cpuset);
         hwloc_x86_add_groups(topology, infos, nbprocs, remaining_cpuset,
                              UNIT, "Compute Unit",
                              HWLOC_GROUP_KIND_AMD_COMPUTE_UNIT, 0);
+      }
+      if (data->found_module_ids) {
         /* Look for Intel Modules inside packages */
         hwloc_bitmap_copy(remaining_cpuset, complete_cpuset);
         hwloc_x86_add_groups(topology, infos, nbprocs, remaining_cpuset,
                              MODULE, "Module",
                              HWLOC_GROUP_KIND_INTEL_MODULE, 0);
+      }
+      if (data->found_tile_ids) {
         /* Look for Intel Tiles inside packages */
         hwloc_bitmap_copy(remaining_cpuset, complete_cpuset);
         hwloc_x86_add_groups(topology, infos, nbprocs, remaining_cpuset,
                              TILE, "Tile",
                              HWLOC_GROUP_KIND_INTEL_TILE, 0);
+      }
 
       /* Look for unknown objects */
       if (infos[one].otherids) {
@@ -1094,7 +1160,8 @@ static void summarize(struct hwloc_backend *backend, struct procinfo *infos, uns
     }
   }
 
-  if (hwloc_filter_check_keep_object_type(topology, HWLOC_OBJ_DIE)) {
+  if (data->found_die_ids
+      && hwloc_filter_check_keep_object_type(topology, HWLOC_OBJ_DIE)) {
     /* Look for Intel Dies inside packages */
     if (fulldiscovery) {
       hwloc_bitmap_t die_cpuset;
@@ -1349,8 +1416,11 @@ look_procs(struct hwloc_backend *backend, struct procinfo *infos, unsigned long
   if (data->apicid_unique) {
     summarize(backend, infos, flags);
 
-    if (has_hybrid(features) && !(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_CPUKINDS)) {
+    if (data->is_hybrid
+        && !(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_CPUKINDS)) {
       /* use hybrid info for cpukinds */
+      if (cpuid_type == intel) {
+        /* Hybrid Intel */
         hwloc_bitmap_t atomset = hwloc_bitmap_alloc();
         hwloc_bitmap_t coreset = hwloc_bitmap_alloc();
         for(i=0; i<nbprocs; i++) {
@@ -1381,8 +1451,10 @@ look_procs(struct hwloc_backend *backend, struct procinfo *infos, unsigned long
         }
       }
     }
-  /* if !data->apicid_unique, do nothing and return success, so that the caller does nothing either */
+  } else {
-
+    hwloc_debug("x86 APIC IDs aren't unique, x86 discovery ignored.\n");
+    /* do nothing and return success, so that the caller does nothing either */
+  }
   return 0;
 }
 
@@ -1459,7 +1531,15 @@ int hwloc_look_x86(struct hwloc_backend *backend, unsigned long flags)
   unsigned i;
   unsigned highest_cpuid;
   unsigned highest_ext_cpuid;
-  /* This stores cpuid features with the same indexing as Linux */
+  /* This stores cpuid features with the same indexing as Linux:
+   * [0] = 0x1 edx
+   * [1] = 0x80000001 edx
+   * [4] = 0x1 ecx
+   * [6] = 0x80000001 ecx
+   * [9] = 0x7/0 ebx
+   * [16] = 0x7/0 ecx
+   * [18] = 0x7/0 edx
+   */
   unsigned features[19] = { 0 };
   struct procinfo *infos = NULL;
   enum cpuid_type cpuid_type = unknown;
@@ -1579,6 +1659,7 @@ int hwloc_look_x86(struct hwloc_backend *backend, unsigned long flags)
     ecx = 0;
     cpuid_or_from_dump(&eax, &ebx, &ecx, &edx, src_cpuiddump);
     features[9] = ebx;
+    features[16] = ecx;
     features[18] = edx;
   }
 
@@ -1667,7 +1748,7 @@ hwloc_x86_discover(struct hwloc_backend *backend, struct hwloc_disc_status *dsta
 
   if (topology->levels[0][0]->cpuset) {
     /* somebody else discovered things, reconnect levels so that we can look at them */
-    hwloc_topology_reconnect(topology, 0);
+    hwloc__reconnect(topology, 0);
     if (topology->nb_levels == 2 && topology->level_nbobjects[1] == data->nbprocs) {
       /* only PUs were discovered, as much as we would, complete the topology with everything else */
       alreadypus = 1;
@@ -1730,17 +1811,17 @@ hwloc_x86_check_cpuiddump_input(const char *src_cpuiddump_path, hwloc_bitmap_t s
   sprintf(path, "%s/hwloc-cpuid-info", src_cpuiddump_path);
   file = fopen(path, "r");
   if (!file) {
-    fprintf(stderr, "Couldn't open dumped cpuid summary %s\n", path);
+    fprintf(stderr, "hwloc/x86: Couldn't open dumped cpuid summary %s\n", path);
     goto out_with_path;
   }
   if (!fgets(line, sizeof(line), file)) {
-    fprintf(stderr, "Found read dumped cpuid summary in %s\n", path);
+    fprintf(stderr, "hwloc/x86: Found read dumped cpuid summary in %s\n", path);
     fclose(file);
     goto out_with_path;
   }
   fclose(file);
-  if (strcmp(line, "Architecture: x86\n")) {
+  if (strncmp(line, "Architecture: x86", 17)) {
-    fprintf(stderr, "Found non-x86 dumped cpuid summary in %s: %s\n", path, line);
+    fprintf(stderr, "hwloc/x86: Found non-x86 dumped cpuid summary in %s: %s\n", path, line);
     goto out_with_path;
   }
   free(path);
@@ -1752,19 +1833,19 @@ hwloc_x86_check_cpuiddump_input(const char *src_cpuiddump_path, hwloc_bitmap_t s
       if (!*end)
 	hwloc_bitmap_set(set, idx);
       else
-	fprintf(stderr, "Ignoring invalid dirent `%s' in dumped cpuid directory `%s'\n",
+	fprintf(stderr, "hwloc/x86: Ignoring invalid dirent `%s' in dumped cpuid directory `%s'\n",
 		dirent->d_name, src_cpuiddump_path);
     }
   }
   closedir(dir);
 
   if (hwloc_bitmap_iszero(set)) {
-    fprintf(stderr, "Did not find any valid pu%%u entry in dumped cpuid directory `%s'\n",
+    fprintf(stderr, "hwloc/x86: Did not find any valid pu%%u entry in dumped cpuid directory `%s'\n",
 	    src_cpuiddump_path);
     return -1;
   } else if (hwloc_bitmap_last(set) != hwloc_bitmap_weight(set) - 1) {
     /* The x86 backends enforces contigous set of PUs starting at 0 so far */
-    fprintf(stderr, "Found non-contigous pu%%u range in dumped cpuid directory `%s'\n",
+    fprintf(stderr, "hwloc/x86: Found non-contigous pu%%u range in dumped cpuid directory `%s'\n",
 	    src_cpuiddump_path);
     return -1;
   }
@@ -1816,9 +1897,15 @@ hwloc_x86_component_instantiate(struct hwloc_topology *topology,
 
   /* default values */
   data->is_knl = 0;
+  data->is_hybrid = 0;
   data->apicid_set = hwloc_bitmap_alloc();
   data->apicid_unique = 1;
   data->src_cpuiddump_path = NULL;
+  data->found_die_ids = 0;
+  data->found_complex_ids = 0;
+  data->found_unit_ids = 0;
+  data->found_module_ids = 0;
+  data->found_tile_ids = 0;
 
   src_cpuiddump_path = getenv("HWLOC_CPUID_PATH");
   if (src_cpuiddump_path) {
@@ -1829,7 +1916,7 @@ hwloc_x86_component_instantiate(struct hwloc_topology *topology,
       assert(!hwloc_bitmap_iszero(set)); /* enforced by hwloc_x86_check_cpuiddump_input() */
       data->nbprocs = hwloc_bitmap_weight(set);
     } else {
-      fprintf(stderr, "Ignoring dumped cpuid directory.\n");
+      fprintf(stderr, "hwloc/x86: Ignoring dumped cpuid directory.\n");
     }
     hwloc_bitmap_free(set);
   }

src/3rdparty/hwloc/src/topology-xml-nolibxml.c

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2020 Inria.  All rights reserved.
+ * Copyright © 2009-2024 Inria.  All rights reserved.
  * Copyright © 2009-2011 Université Bordeaux
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -41,7 +41,7 @@ typedef struct hwloc__nolibxml_import_state_data_s {
 static char *
 hwloc__nolibxml_import_ignore_spaces(char *buffer)
 {
-  return buffer + strspn(buffer, " \t\n");
+  return buffer + strspn(buffer, " \t\n\r");
 }
 
 static int
@@ -411,12 +411,12 @@ hwloc_nolibxml_backend_init(struct hwloc_xml_backend_data_s *bdata,
   bdata->data = nbdata;
 
   if (xmlbuffer) {
-    nbdata->buffer = malloc(xmlbuflen+1);
+    nbdata->buffer = malloc(xmlbuflen);
     if (!nbdata->buffer)
       goto out_with_nbdata;
-    nbdata->buflen = xmlbuflen+1;
+    nbdata->buflen = xmlbuflen;
     memcpy(nbdata->buffer, xmlbuffer, xmlbuflen);
-    nbdata->buffer[xmlbuflen] = '\0';
+    nbdata->buffer[xmlbuflen-1] = '\0'; /* make sure it's there as requested in the API */
 
   } else {
     int err = hwloc_nolibxml_read_file(xmlpath, &nbdata->buffer, &nbdata->buflen);
@@ -453,8 +453,9 @@ hwloc_nolibxml_import_diff(struct hwloc__xml_import_state_s *state,
     buffer = malloc(xmlbuflen);
     if (!buffer)
       goto out;
-    memcpy(buffer, xmlbuffer, xmlbuflen);
     buflen = xmlbuflen;
+    memcpy(buffer, xmlbuffer, xmlbuflen);
+    buffer[xmlbuflen-1] = '\0'; /* make sure it's there as requested in the API */
 
   } else {
     ret = hwloc_nolibxml_read_file(xmlpath, &buffer, &buflen);

src/3rdparty/hwloc/src/topology-xml.c

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2022 Inria.  All rights reserved.
+ * Copyright © 2009-2025 Inria.  All rights reserved.
  * Copyright © 2009-2011, 2020 Université Bordeaux
  * Copyright © 2009-2018 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -415,6 +415,20 @@ hwloc__xml_import_object_attr(struct hwloc_topology *topology,
     }
   }
 
+  else if (!strcmp(name, "numanode_type")) {
+    switch (obj->type) {
+    case HWLOC_OBJ_NUMANODE: {
+      /* ignored for now, here for possible forward compat */
+      break;
+    }
+    default:
+      if (hwloc__xml_verbose())
+	fprintf(stderr, "%s: ignoring numanode_type attribute for non-NUMA object\n",
+		state->global->msgprefix);
+      break;
+    }
+  }
+
   else if (data->version_major < 2) {
     /************************
      * deprecated from 1.x
@@ -562,7 +576,13 @@ hwloc__xml_import_pagetype(hwloc_topology_t topology __hwloc_attribute_unused, s
     char *attrname, *attrvalue;
     if (state->global->next_attr(state, &attrname, &attrvalue) < 0)
       break;
-    if (!strcmp(attrname, "size"))
+    if (!strcmp(attrname, "info")) {
+      char *infoname, *infovalue;
+      int ret = hwloc___xml_import_info(&infoname, &infovalue, state);
+      if (ret < 0)
+        return -1;
+      /* ignored */
+    } else if (!strcmp(attrname, "size"))
       size = strtoull(attrvalue, NULL, 10);
     else if (!strcmp(attrname, "count"))
       count = strtoull(attrvalue, NULL, 10);
@@ -866,14 +886,23 @@ hwloc__xml_import_object(hwloc_topology_t topology,
 	  /* deal with possible future type */
 	  obj->type = HWLOC_OBJ_GROUP;
 	  obj->attr->group.kind = HWLOC_GROUP_KIND_INTEL_MODULE;
-	} else if (!strcasecmp(attrvalue, "MemCache")) {
+	} else if (!strcasecmp(attrvalue, "Cluster")) {
+	  /* deal with possible future type */
+	  obj->type = HWLOC_OBJ_GROUP;
+	  obj->attr->group.kind = HWLOC_GROUP_KIND_LINUX_CLUSTER;
+	}
+#if 0
+        /* reenable if there's ever a future type that should be ignored without being an error */
+        else if (!strcasecmp(attrvalue, "MemCache")) {
 	  /* ignore possible future type */
 	  obj->type = _HWLOC_OBJ_FUTURE;
 	  ignored = 1;
 	  if (hwloc__xml_verbose())
 	    fprintf(stderr, "%s: %s object not-supported, will be ignored\n",
 		    state->global->msgprefix, attrvalue);
-	} else {
+	}
+#endif
+        else {
 	  if (hwloc__xml_verbose())
 	    fprintf(stderr, "%s: unrecognized object type string %s\n",
 		    state->global->msgprefix, attrvalue);
@@ -948,22 +977,22 @@ hwloc__xml_import_object(hwloc_topology_t topology,
     if (hwloc__obj_type_is_normal(obj->type)) {
       if (!hwloc__obj_type_is_normal(parent->type)) {
 	if (hwloc__xml_verbose())
-	  fprintf(stderr, "normal object %s cannot be child of non-normal parent %s\n",
+	  fprintf(stderr, "%s: normal object %s cannot be child of non-normal parent %s\n",
-		  hwloc_obj_type_string(obj->type), hwloc_obj_type_string(parent->type));
+		  state->global->msgprefix, hwloc_obj_type_string(obj->type), hwloc_obj_type_string(parent->type));
 	goto error_with_object;
       }
     } else if (hwloc__obj_type_is_memory(obj->type)) {
       if (hwloc__obj_type_is_io(parent->type) || HWLOC_OBJ_MISC == parent->type) {
 	if (hwloc__xml_verbose())
-	  fprintf(stderr, "Memory object %s cannot be child of non-normal-or-memory parent %s\n",
+	  fprintf(stderr, "%s: Memory object %s cannot be child of non-normal-or-memory parent %s\n",
-		  hwloc_obj_type_string(obj->type), hwloc_obj_type_string(parent->type));
+		  state->global->msgprefix, hwloc_obj_type_string(obj->type), hwloc_obj_type_string(parent->type));
 	goto error_with_object;
       }
     } else if (hwloc__obj_type_is_io(obj->type)) {
       if (hwloc__obj_type_is_memory(parent->type) || HWLOC_OBJ_MISC == parent->type) {
 	if (hwloc__xml_verbose())
-	  fprintf(stderr, "I/O object %s cannot be child of non-normal-or-I/O parent %s\n",
+	  fprintf(stderr, "%s: I/O object %s cannot be child of non-normal-or-I/O parent %s\n",
-		  hwloc_obj_type_string(obj->type), hwloc_obj_type_string(parent->type));
+		  state->global->msgprefix, hwloc_obj_type_string(obj->type), hwloc_obj_type_string(parent->type));
 	goto error_with_object;
       }
     }
@@ -1160,6 +1189,48 @@ hwloc__xml_import_object(hwloc_topology_t topology,
     data->last_numanode = obj;
   }
 
+  /* 3.0 forward compatibility */
+  if (data->version_major >= 3 && obj->type == HWLOC_OBJ_OS_DEVICE) {
+    /* osdev.type changed into bitmak in 3.0 */
+    if (obj->attr->osdev.type & 3 /* STORAGE|MEMORY for BLOCK */) {
+      obj->attr->osdev.type = HWLOC_OBJ_OSDEV_BLOCK;
+    } else if (obj->attr->osdev.type & 8 /* COPROC for COPROC and rsmi/nvml GPUs */) {
+      if (obj->subtype && (!strcmp(obj->subtype, "RSMI") || !strcmp(obj->subtype, "NVML")))
+        obj->attr->osdev.type = HWLOC_OBJ_OSDEV_GPU;
+      else
+        obj->attr->osdev.type = HWLOC_OBJ_OSDEV_COPROC;
+    } else if (obj->attr->osdev.type & 4 /* GPU for non-COPROC GPUs */) {
+      obj->attr->osdev.type = HWLOC_OBJ_OSDEV_GPU;
+    } else if (obj->attr->osdev.type & 32 /* OFED */) {
+      obj->attr->osdev.type = HWLOC_OBJ_OSDEV_OPENFABRICS;
+    } else if (obj->attr->osdev.type & 16 /* NET for NET and BXI v2-fake-OFED */) {
+      if (obj->subtype && !strcmp(obj->subtype, "BXI"))
+        obj->attr->osdev.type = HWLOC_OBJ_OSDEV_OPENFABRICS;
+      else
+        obj->attr->osdev.type = HWLOC_OBJ_OSDEV_NETWORK;
+    } else if (obj->attr->osdev.type & 64 /* DMA */) {
+      obj->attr->osdev.type = HWLOC_OBJ_OSDEV_DMA;
+    } else { /* none or unknown */
+      obj->attr->osdev.type = (hwloc_obj_osdev_type_t)  -1;
+    }
+    /* Backend info only in root */
+    if (obj->subtype && !hwloc_obj_get_info_by_name(obj, "Backend")) {
+      if (!strcmp(obj->subtype, "CUDA")) {
+        hwloc_obj_add_info(obj, "Backend", "CUDA");
+      } else if  (!strcmp(obj->subtype, "NVML")) {
+        hwloc_obj_add_info(obj, "Backend", "NVML");
+      } else if  (!strcmp(obj->subtype, "OpenCL")) {
+        hwloc_obj_add_info(obj, "Backend", "OpenCL");
+      } else if  (!strcmp(obj->subtype, "RSMI")) {
+        hwloc_obj_add_info(obj, "Backend", "RSMI");
+      } else if  (!strcmp(obj->subtype, "LevelZero")) {
+        hwloc_obj_add_info(obj, "Backend", "LevelZero");
+      } else if  (!strcmp(obj->subtype, "Display")) {
+        hwloc_obj_add_info(obj, "Backend", "GL");
+      }
+    }
+  }
+
   if (!hwloc_filter_check_keep_object(topology, obj)) {
     /* Ignore this object instead of inserting it.
      *
@@ -1296,7 +1367,7 @@ hwloc__xml_v2import_support(hwloc_topology_t topology,
     HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_support) == 4*sizeof(void*));
     HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_discovery_support) == 6);
     HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_cpubind_support) == 11);
-    HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_membind_support) == 15);
+    HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_membind_support) == 16);
     HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_misc_support) == 1);
 #endif
 
@@ -1330,6 +1401,7 @@ hwloc__xml_v2import_support(hwloc_topology_t topology,
     else DO(membind,firsttouch_membind);
     else DO(membind,bind_membind);
     else DO(membind,interleave_membind);
+    else DO(membind,weighted_interleave_membind);
     else DO(membind,nexttouch_membind);
     else DO(membind,migrate_membind);
     else DO(membind,get_area_memlocation);
@@ -1388,6 +1460,10 @@ hwloc__xml_v2import_distances(hwloc_topology_t topology,
     }
     else if (!strcmp(attrname, "kind")) {
       kind = strtoul(attrvalue, NULL, 10);
+      /* forward compat with "HOPS" kind in v3 */
+      if (kind & (1UL<<5))
+        /* hops becomes latency */
+        kind = (kind & ~(1UL<<5)) | HWLOC_DISTANCES_KIND_MEANS_LATENCY;
     }
     else if (!strcmp(attrname, "name")) {
       name = attrvalue;
@@ -1433,7 +1509,14 @@ hwloc__xml_v2import_distances(hwloc_topology_t topology,
     if (ret <= 0)
       break;
 
-    if (!strcmp(tag, "indexes"))
+    if (!strcmp(tag, "info")) {
+      char *infoname, *infovalue;
+      ret = hwloc___xml_import_info(&infoname, &infovalue, state);
+      if (ret < 0)
+        goto out_with_arrays;
+      /* ignored */
+      continue;
+    } else if (!strcmp(tag, "indexes"))
       is_index = 1;
     else if (!strcmp(tag, "u64values"))
       is_u64values = 1;
@@ -1766,6 +1849,10 @@ hwloc__xml_import_memattr(hwloc_topology_t topology,
 
     if (!strcmp(tag, "memattr_value")) {
       ret = hwloc__xml_import_memattr_value(topology, id, flags, &childstate);
+    } else if (!strcmp(tag, "info")) {
+      char *infoname, *infovalue;
+      ret = hwloc___xml_import_info(&infoname, &infovalue, &childstate);
+      /* ignored */
     } else {
       if (hwloc__xml_verbose())
         fprintf(stderr, "%s: memattr with unrecognized child %s\n",
@@ -2094,9 +2181,10 @@ hwloc_look_xml(struct hwloc_backend *backend, struct hwloc_disc_status *dstatus)
   if (ret < 0)
     goto failed;
 
-  if (data->version_major > 2) {
+  if (data->version_major > 3
+      || (data->version_major == 3 && data->version_minor > 0)) {
     if (hwloc__xml_verbose())
-      fprintf(stderr, "%s: cannot import XML version %u.%u > 2\n",
+      fprintf(stderr, "%s: cannot import XML version %u.%u > 3.0\n",
 	      data->msgprefix, data->version_major, data->version_minor);
     goto err;
   }
@@ -2144,6 +2232,13 @@ hwloc_look_xml(struct hwloc_backend *backend, struct hwloc_disc_status *dstatus)
         ret = hwloc__xml_import_cpukind(topology, &childstate);
         if (ret < 0)
           goto failed;
+      } else if (!strcmp(tag, "info")) {
+        char *infoname, *infovalue;
+        ret = hwloc___xml_import_info(&infoname, &infovalue, &childstate);
+        if (ret < 0)
+          goto failed;
+        /* move 3.x topology info back to the root object */
+        hwloc_obj_add_info(topology->levels[0][0], infoname, infovalue);
       } else {
 	if (hwloc__xml_verbose())
 	  fprintf(stderr, "%s: ignoring unknown tag `%s' after root object.\n",
@@ -3020,7 +3115,7 @@ hwloc__xml_v2export_support(hwloc__xml_export_state_t parentstate, hwloc_topolog
   HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_support) == 4*sizeof(void*));
   HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_discovery_support) == 6);
   HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_cpubind_support) == 11);
-  HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_membind_support) == 15);
+  HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_membind_support) == 16);
   HWLOC_BUILD_ASSERT(sizeof(struct hwloc_topology_misc_support) == 1);
 #endif
 
@@ -3065,6 +3160,7 @@ hwloc__xml_v2export_support(hwloc__xml_export_state_t parentstate, hwloc_topolog
   DO(membind,firsttouch_membind);
   DO(membind,bind_membind);
   DO(membind,interleave_membind);
+  DO(membind,weighted_interleave_membind);
   DO(membind,nexttouch_membind);
   DO(membind,migrate_membind);
   DO(membind,get_area_memlocation);

src/3rdparty/hwloc/src/topology.c

@@ -1,6 +1,6 @@
 /*
  * Copyright © 2009 CNRS
- * Copyright © 2009-2022 Inria.  All rights reserved.
+ * Copyright © 2009-2025 Inria.  All rights reserved.
  * Copyright © 2009-2012, 2020 Université Bordeaux
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
  * Copyright © 2022 IBM Corporation.  All rights reserved.
@@ -54,56 +54,6 @@
 #endif
 
 
-#ifdef HWLOC_HAVE_LEVELZERO
-/*
- * Define ZES_ENABLE_SYSMAN=1 early so that the LevelZero backend gets Sysman enabled.
- *
- * Only if the levelzero was enabled in this build so that we don't enable sysman
- * for external levelzero users when hwloc doesn't need it. If somebody ever loads
- * an external levelzero plugin in a hwloc library built without levelzero (unlikely),
- * he may have to manually set ZES_ENABLE_SYSMAN=1.
- *
- * Use the constructor if supported and/or the Windows DllMain callback.
- * Do it in the main hwloc library instead of the levelzero component because
- * the latter could be loaded later as a plugin.
- *
- * L0 seems to be using getenv() to check this variable on Windows
- * (at least in the Intel Compute-Runtime of March 2021),
- * but setenv() doesn't seem to exist on Windows, hence use putenv() to set the variable.
- *
- * For the record, Get/SetEnvironmentVariable() is not exactly the same as getenv/putenv():
- * - getenv() doesn't see what was set with SetEnvironmentVariable()
- * - GetEnvironmentVariable() doesn't see putenv() in cygwin (while it does in MSVC and MinGW).
- * Hence, if L0 ever switches from getenv() to GetEnvironmentVariable(),
- * it will break in cygwin, we'll have to use both putenv() and SetEnvironmentVariable().
- * Hopefully L0 will provide a way to enable Sysman without env vars before it happens.
- */
-#if HWLOC_HAVE_ATTRIBUTE_CONSTRUCTOR
-static void hwloc_constructor(void) __attribute__((constructor));
-static void hwloc_constructor(void)
-{
-  if (!getenv("ZES_ENABLE_SYSMAN"))
-#ifdef HWLOC_WIN_SYS
-    putenv("ZES_ENABLE_SYSMAN=1");
-#else
-    setenv("ZES_ENABLE_SYSMAN", "1", 1);
-#endif
-}
-#endif
-#ifdef HWLOC_WIN_SYS
-BOOL WINAPI DllMain(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpReserved)
-{
-  if (fdwReason == DLL_PROCESS_ATTACH) {
-    if (!getenv("ZES_ENABLE_SYSMAN"))
-      /* Windows does not have a setenv, so use putenv. */
-      putenv((char *) "ZES_ENABLE_SYSMAN=1");
-  }
-  return TRUE;
-}
-#endif
-#endif /* HWLOC_HAVE_LEVELZERO */
-
-
 unsigned hwloc_get_api_version(void)
 {
   return HWLOC_API_VERSION;
@@ -146,21 +96,24 @@ report_insert_error_format_obj(char *buf, size_t buflen, hwloc_obj_t obj)
   char typestr[64];
   char *cpusetstr;
   char *nodesetstr = NULL;
+  char indexstr[64] = "";
+  char groupstr[64] = "";
 
   hwloc_obj_type_snprintf(typestr, sizeof(typestr), obj, 0);
   hwloc_bitmap_asprintf(&cpusetstr, obj->cpuset);
+  if (obj->os_index != HWLOC_UNKNOWN_INDEX)
+    snprintf(indexstr, sizeof(indexstr), "P#%u ", obj->os_index);
+  if (obj->type == HWLOC_OBJ_GROUP)
+    snprintf(groupstr, sizeof(groupstr), "groupkind %u-%u ", obj->attr->group.kind, obj->attr->group.subkind);
   if (obj->nodeset) /* may be missing during insert */
     hwloc_bitmap_asprintf(&nodesetstr, obj->nodeset);
-  if (obj->os_index != HWLOC_UNKNOWN_INDEX)
+  snprintf(buf, buflen, "%s (%s%s%s%s%scpuset %s%s%s)",
-    snprintf(buf, buflen, "%s (P#%u cpuset %s%s%s)",
+           typestr,
-             typestr, obj->os_index, cpusetstr,
+           indexstr,
-             nodesetstr ? " nodeset " : "",
+           obj->subtype ? "subtype " : "", obj->subtype ? obj->subtype : "", obj->subtype ? " " : "",
-             nodesetstr ? nodesetstr : "");
+           groupstr,
-  else
+           cpusetstr,
-    snprintf(buf, buflen, "%s (cpuset %s%s%s)",
+           nodesetstr ? " nodeset " : "", nodesetstr ? nodesetstr : "");
-             typestr, cpusetstr,
-             nodesetstr ? " nodeset " : "",
-             nodesetstr ? nodesetstr : "");
   free(cpusetstr);
   free(nodesetstr);
 }
@@ -176,10 +129,11 @@ static void report_insert_error(hwloc_obj_t new, hwloc_obj_t old, const char *ms
     report_insert_error_format_obj(oldstr, sizeof(oldstr), old);
 
     fprintf(stderr, "****************************************************************************\n");
-    fprintf(stderr, "* hwloc %s received invalid information from the operating system.\n", HWLOC_VERSION);
+    fprintf(stderr, "* hwloc %s received invalid information.\n", HWLOC_VERSION);
     fprintf(stderr, "*\n");
-    fprintf(stderr, "* Failed with: %s\n", msg);
+    fprintf(stderr, "* Failed with error: %s\n", msg);
-    fprintf(stderr, "* while inserting %s at %s\n", newstr, oldstr);
+    fprintf(stderr, "* while inserting %s\n", newstr);
+    fprintf(stderr, "* at %s\n", oldstr);
     fprintf(stderr, "* coming from: %s\n", reason);
     fprintf(stderr, "*\n");
     fprintf(stderr, "* The following FAQ entry in the hwloc documentation may help:\n");
@@ -461,6 +415,20 @@ hwloc_debug_print_objects(int indent __hwloc_attribute_unused, hwloc_obj_t obj)
 #define hwloc_debug_print_objects(indent, obj) do { /* nothing */ } while (0)
 #endif /* !HWLOC_DEBUG */
 
+int hwloc_obj_set_subtype(hwloc_topology_t topology __hwloc_attribute_unused, hwloc_obj_t obj, const char *subtype)
+{
+  char *new = NULL;
+  if (subtype) {
+    new = strdup(subtype);
+    if (!new)
+      return -1;
+  }
+  if (obj->subtype)
+    free(obj->subtype);
+  obj->subtype = new;
+  return 0;
+}
+
 void hwloc__free_infos(struct hwloc_info_s *infos, unsigned count)
 {
   unsigned i;
@@ -679,6 +647,7 @@ unlink_and_free_object_and_children(hwloc_obj_t *pobj)
 void
 hwloc_free_object_and_children(hwloc_obj_t obj)
 {
+  if (obj)
     unlink_and_free_object_and_children(&obj);
 }
 
@@ -1925,17 +1894,78 @@ hwloc_topology_alloc_group_object(struct hwloc_topology *topology)
   return hwloc_alloc_setup_object(topology, HWLOC_OBJ_GROUP, HWLOC_UNKNOWN_INDEX);
 }
 
+int
+hwloc_topology_free_group_object(struct hwloc_topology *topology, hwloc_obj_t obj)
+{
+  if (!topology->is_loaded) {
+    /* this could actually work, see insert() below */
+    errno = EINVAL;
+    return -1;
+  }
+  if (topology->adopted_shmem_addr) {
+    errno = EPERM;
+    return -1;
+  }
+  hwloc_free_unlinked_object(obj);
+  return 0;
+}
+
 static void hwloc_propagate_symmetric_subtree(hwloc_topology_t topology, hwloc_obj_t root);
 static void propagate_total_memory(hwloc_obj_t obj);
 static void hwloc_set_group_depth(hwloc_topology_t topology);
 static void hwloc_connect_children(hwloc_obj_t parent);
 static int hwloc_connect_levels(hwloc_topology_t topology);
 static int hwloc_connect_special_levels(hwloc_topology_t topology);
+static int hwloc_filter_levels_keep_structure(hwloc_topology_t topology);
+
+/* reconnect children and levels,
+ * and optionnally merged identical levels while keeping structure.
+ */
+int
+hwloc__reconnect(struct hwloc_topology *topology, unsigned long flags)
+{
+  int merged_levels = 0;
+
+  if (topology->modified) {
+    hwloc_connect_children(topology->levels[0][0]);
+
+    if (hwloc_connect_levels(topology) < 0)
+      return -1;
+  }
+
+  if (flags & _HWLOC_RECONNECT_FLAG_KEEPSTRUCTURE) {
+    merged_levels = hwloc_filter_levels_keep_structure(topology);
+    /* If > 0, we merged some levels,
+     * some child+parent special children list may have been merged,
+     * hence specials level might need reordering,
+     * So reconnect special levels only here at the end.
+     */
+  }
+
+  if (topology->modified || merged_levels) {
+    if (hwloc_connect_special_levels(topology) < 0)
+      return -1;
+  }
+
+  topology->modified = 0;
+  return 0;
+}
+
+int
+hwloc_topology_reconnect(struct hwloc_topology *topology, unsigned long flags)
+{
+  if (flags) {
+    errno = EINVAL;
+    return -1;
+  }
+
+  return hwloc__reconnect(topology, 0);
+}
 
 hwloc_obj_t
 hwloc_topology_insert_group_object(struct hwloc_topology *topology, hwloc_obj_t obj)
 {
-  hwloc_obj_t res, root;
+  hwloc_obj_t res, root, child;
   int cmp;
 
   if (!topology->is_loaded) {
@@ -1945,6 +1975,7 @@ hwloc_topology_insert_group_object(struct hwloc_topology *topology, hwloc_obj_t
     return NULL;
   }
   if (topology->adopted_shmem_addr) {
+    hwloc_free_unlinked_object(obj);
     errno = EPERM;
     return NULL;
   }
@@ -1998,6 +2029,7 @@ hwloc_topology_insert_group_object(struct hwloc_topology *topology, hwloc_obj_t
     res = hwloc__insert_object_by_cpuset(topology, NULL, obj, NULL /* do not show errors on stdout */);
   } else {
     /* just merge root */
+    hwloc_free_unlinked_object(obj);
     res = root;
   }
 
@@ -2021,9 +2053,19 @@ hwloc_topology_insert_group_object(struct hwloc_topology *topology, hwloc_obj_t
 
   /* properly inserted */
   hwloc_obj_add_children_sets(res);
-  if (hwloc_topology_reconnect(topology, 0) < 0)
+  /* reconnect levels.
+   * no need to filter levels keep_structure because groups are either auto-merged
+   * or have the dont_merge attribute */
+  if (hwloc__reconnect(topology, 0) < 0)
     return NULL;
 
+  /* Compute group total_memory. */
+  res->total_memory = 0;
+  for_each_child(child, res)
+    res->total_memory += child->total_memory;
+  for_each_memory_child(child, res)
+    res->total_memory += child->total_memory;
+
   hwloc_propagate_symmetric_subtree(topology, topology->levels[0][0]);
   hwloc_set_group_depth(topology);
 
@@ -2258,7 +2300,9 @@ hwloc_obj_add_other_obj_sets(hwloc_obj_t dst, hwloc_obj_t src)
   if ((_src)->_set) {								\
     if (!(_dst)->_set)								\
       (_dst)->_set = hwloc_bitmap_alloc();					\
-    hwloc_bitmap_or((_dst)->_set, (_dst)->_set, (_src)->_set);	\
+    if (!(_dst)->_set								\
+        || hwloc_bitmap_or((_dst)->_set, (_dst)->_set, (_src)->_set) < 0)	\
+      return -1;								\
   }
   ADD_OTHER_OBJ_SET(dst, src, cpuset);
   ADD_OTHER_OBJ_SET(dst, src, complete_cpuset);
@@ -2504,26 +2548,13 @@ hwloc_compare_levels_structure(hwloc_topology_t topology, unsigned i)
   return 0;
 }
 
-/* return > 0 if any level was removed.
+/* return > 0 if any level was removed. */
- * performs its own reconnect internally if needed
- */
 static int
 hwloc_filter_levels_keep_structure(hwloc_topology_t topology)
 {
   unsigned i, j;
   int res = 0;
 
-  if (topology->modified) {
-    /* WARNING: hwloc_topology_reconnect() is duplicated partially here
-     * and at the end of this function:
-     * - we need normal levels before merging.
-     * - and we'll need to update special levels after merging.
-     */
-    hwloc_connect_children(topology->levels[0][0]);
-    if (hwloc_connect_levels(topology) < 0)
-      return -1;
-  }
-
   /* start from the bottom since we'll remove intermediate levels */
   for(i=topology->nb_levels-1; i>0; i--) {
     int replacechild = 0, replaceparent = 0;
@@ -2545,9 +2576,15 @@ hwloc_filter_levels_keep_structure(hwloc_topology_t topology)
       if (type1 == HWLOC_OBJ_GROUP && hwloc_dont_merge_group_level(topology, i))
 	replacechild = 0;
     }
-    if (!replacechild && !replaceparent)
+    if (!replacechild && !replaceparent) {
+      /* always merge Die into Package when levels are identical */
+      if (type1 == HWLOC_OBJ_PACKAGE && type2 == HWLOC_OBJ_DIE)
+        replacechild = 1;
+    }
+    if (!replacechild && !replaceparent) {
       /* no ignoring */
       continue;
+    }
     /* Decide which one to actually replace */
     if (replaceparent && replacechild) {
       /* If both may be replaced, look at obj_type_priority */
@@ -2690,20 +2727,6 @@ hwloc_filter_levels_keep_structure(hwloc_topology_t topology)
     }
   }
 
-
-  if (res > 0 || topology-> modified) {
-    /* WARNING: hwloc_topology_reconnect() is duplicated partially here
-     * and at the beginning of this function.
-     * If we merged some levels, some child+parent special children lisst
-     * may have been merged, hence specials level might need reordering,
-     * So reconnect special levels only here at the end
-     * (it's not needed at the beginning of this function).
-     */
-    if (hwloc_connect_special_levels(topology) < 0)
-      return -1;
-    topology->modified = 0;
-  }
-
   return 0;
 }
 
@@ -3232,33 +3255,6 @@ hwloc_connect_levels(hwloc_topology_t topology)
   return 0;
 }
 
-int
-hwloc_topology_reconnect(struct hwloc_topology *topology, unsigned long flags)
-{
-  /* WARNING: when updating this function, the replicated code must
-   * also be updated inside hwloc_filter_levels_keep_structure()
-   */
-
-  if (flags) {
-    errno = EINVAL;
-    return -1;
-  }
-  if (!topology->modified)
-    return 0;
-
-  hwloc_connect_children(topology->levels[0][0]);
-
-  if (hwloc_connect_levels(topology) < 0)
-    return -1;
-
-  if (hwloc_connect_special_levels(topology) < 0)
-    return -1;
-
-  topology->modified = 0;
-
-  return 0;
-}
-
 /* for regression testing, make sure the order of io devices
  * doesn't change with the dentry order in the filesystem
  *
@@ -3515,32 +3511,13 @@ hwloc_discover(struct hwloc_topology *topology,
     hwloc_debug_print_objects(0, topology->levels[0][0]);
   }
 
-  /* see if we should ignore the root now that we know how many children it has */
-  if (!hwloc_filter_check_keep_object(topology, topology->levels[0][0])
-      && topology->levels[0][0]->first_child && !topology->levels[0][0]->first_child->next_sibling) {
-    hwloc_obj_t oldroot = topology->levels[0][0];
-    hwloc_obj_t newroot = oldroot->first_child;
-    /* switch to the new root */
-    newroot->parent = NULL;
-    topology->levels[0][0] = newroot;
-    /* move oldroot memory/io/misc children before newroot children */
-    if (oldroot->memory_first_child)
-      prepend_siblings_list(&newroot->memory_first_child, oldroot->memory_first_child, newroot);
-    if (oldroot->io_first_child)
-      prepend_siblings_list(&newroot->io_first_child, oldroot->io_first_child, newroot);
-    if (oldroot->misc_first_child)
-      prepend_siblings_list(&newroot->misc_first_child, oldroot->misc_first_child, newroot);
-    /* destroy oldroot and use the new one */
-    hwloc_free_unlinked_object(oldroot);
-  }
-
   /*
    * All object cpusets and nodesets are properly set now.
    */
 
   /* Now connect handy pointers to make remaining discovery easier. */
   hwloc_debug("%s", "\nOk, finished tweaking, now connect\n");
-  if (hwloc_topology_reconnect(topology, 0) < 0)
+  if (hwloc__reconnect(topology, 0) < 0)
     return -1;
   hwloc_debug_print_objects(0, topology->levels[0][0]);
 
@@ -3596,12 +3573,12 @@ hwloc_discover(struct hwloc_topology *topology,
   }
   hwloc_debug_print_objects(0, topology->levels[0][0]);
 
+  /* reconnect all (new groups might have appears, IO added, etc),
+   * and (now that everything was added) remove identical levels while keeping structure
+   */
   hwloc_debug("%s", "\nRemoving levels with HWLOC_TYPE_FILTER_KEEP_STRUCTURE\n");
-  if (hwloc_filter_levels_keep_structure(topology) < 0)
+  if (hwloc__reconnect(topology, _HWLOC_RECONNECT_FLAG_KEEPSTRUCTURE) < 0)
     return -1;
-  /* takes care of reconnecting children/levels internally,
-   * because it needs normal levels.
-   * and it's often needed below because of Groups inserted for I/Os anyway */
   hwloc_debug_print_objects(0, topology->levels[0][0]);
 
   /* accumulate children memory in total_memory fields (only once parent is set) */
@@ -3730,6 +3707,7 @@ hwloc__topology_init (struct hwloc_topology **topologyp,
 
   hwloc__topology_filter_init(topology);
 
+  /* always initialize since we don't know flags to disable those yet */
   hwloc_internal_distances_init(topology);
   hwloc_internal_memattrs_init(topology);
   hwloc_internal_cpukinds_init(topology);
@@ -3942,8 +3920,12 @@ int
 hwloc_topology_set_cache_types_filter(hwloc_topology_t topology, enum hwloc_type_filter_e filter)
 {
   unsigned i;
-  for(i=HWLOC_OBJ_L1CACHE; i<HWLOC_OBJ_L3ICACHE; i++)
+  if (topology->is_loaded) {
-    hwloc_topology_set_type_filter(topology, (hwloc_obj_type_t) i, filter);
+    errno = EBUSY;
+    return -1;
+  }
+  for(i=HWLOC_OBJ_L1CACHE; i<=HWLOC_OBJ_L3ICACHE; i++)
+    hwloc__topology_set_type_filter(topology, (hwloc_obj_type_t) i, filter);
   return 0;
 }
 
@@ -3951,17 +3933,25 @@ int
 hwloc_topology_set_icache_types_filter(hwloc_topology_t topology, enum hwloc_type_filter_e filter)
 {
   unsigned i;
-  for(i=HWLOC_OBJ_L1ICACHE; i<HWLOC_OBJ_L3ICACHE; i++)
+  if (topology->is_loaded) {
-    hwloc_topology_set_type_filter(topology, (hwloc_obj_type_t) i, filter);
+    errno = EBUSY;
+    return -1;
+  }
+  for(i=HWLOC_OBJ_L1ICACHE; i<=HWLOC_OBJ_L3ICACHE; i++)
+    hwloc__topology_set_type_filter(topology, (hwloc_obj_type_t) i, filter);
   return 0;
 }
 
 int
 hwloc_topology_set_io_types_filter(hwloc_topology_t topology, enum hwloc_type_filter_e filter)
 {
-  hwloc_topology_set_type_filter(topology, HWLOC_OBJ_BRIDGE, filter);
+  if (topology->is_loaded) {
-  hwloc_topology_set_type_filter(topology, HWLOC_OBJ_PCI_DEVICE, filter);
+    errno = EBUSY;
-  hwloc_topology_set_type_filter(topology, HWLOC_OBJ_OS_DEVICE, filter);
+    return -1;
+  }
+  hwloc__topology_set_type_filter(topology, HWLOC_OBJ_BRIDGE, filter);
+  hwloc__topology_set_type_filter(topology, HWLOC_OBJ_PCI_DEVICE, filter);
+  hwloc__topology_set_type_filter(topology, HWLOC_OBJ_OS_DEVICE, filter);
   return 0;
 }
 
@@ -3982,9 +3972,12 @@ hwloc_topology_clear (struct hwloc_topology *topology)
 {
   /* no need to set to NULL after free() since callers will call setup_defaults() or just destroy the rest of the topology */
   unsigned l;
+
+  /* always destroy cpukinds/distances/memattrs since there are always initialized during init() */
   hwloc_internal_cpukinds_destroy(topology);
   hwloc_internal_distances_destroy(topology);
   hwloc_internal_memattrs_destroy(topology);
+
   hwloc_free_object_and_children(topology->levels[0][0]);
   hwloc_bitmap_free(topology->allowed_cpuset);
   hwloc_bitmap_free(topology->allowed_nodeset);
@@ -4024,6 +4017,7 @@ hwloc_topology_load (struct hwloc_topology *topology)
 {
   struct hwloc_disc_status dstatus;
   const char *env;
+  unsigned i;
   int err;
 
   if (topology->is_loaded) {
@@ -4032,9 +4026,19 @@ hwloc_topology_load (struct hwloc_topology *topology)
   }
 
   /* initialize envvar-related things */
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_DISTANCES))
     hwloc_internal_distances_prepare(topology);
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_MEMATTRS))
     hwloc_internal_memattrs_prepare(topology);
 
+  /* check if any cpu cache filter is not NONE */
+  topology->want_some_cpu_caches = 0;
+  for(i=HWLOC_OBJ_L1CACHE; i<=HWLOC_OBJ_L3ICACHE; i++)
+    if (topology->type_filter[i] != HWLOC_TYPE_FILTER_KEEP_NONE) {
+      topology->want_some_cpu_caches = 1;
+      break;
+    }
+
   if (getenv("HWLOC_XML_USERDATA_NOT_DECODED"))
     topology->userdata_not_decoded = 1;
 
@@ -4110,9 +4114,12 @@ hwloc_topology_load (struct hwloc_topology *topology)
 #endif
     hwloc_topology_check(topology);
 
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_CPUKINDS)) {
     /* Rank cpukinds */
     hwloc_internal_cpukinds_rank(topology);
+  }
 
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_DISTANCES)) {
     /* Mark distances objs arrays as invalid since we may have removed objects
      * from the topology after adding the distances (remove_empty, etc).
      * It would be hard to actually verify whether it's needed.
@@ -4122,11 +4129,17 @@ hwloc_topology_load (struct hwloc_topology *topology)
      * don't refresh() concurrently (disallowed).
      */
     hwloc_internal_distances_refresh(topology);
+  }
 
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_MEMATTRS)) {
+    int force_memtiers = (getenv("HWLOC_MEMTIERS_REFRESH") != NULL);
     /* Same for memattrs */
     hwloc_internal_memattrs_need_refresh(topology);
     hwloc_internal_memattrs_refresh(topology);
-  hwloc_internal_memattrs_guess_memory_tiers(topology);
+    /* update memtiers unless XML */
+    if (force_memtiers || strcmp(topology->backends->component->name, "xml"))
+      hwloc_internal_memattrs_guess_memory_tiers(topology, force_memtiers);
+  }
 
   topology->is_loaded = 1;
 
@@ -4185,20 +4198,11 @@ restrict_object_by_cpuset(hwloc_topology_t topology, unsigned long flags, hwloc_
     hwloc_bitmap_andnot(obj->cpuset, obj->cpuset, droppedcpuset);
     hwloc_bitmap_andnot(obj->complete_cpuset, obj->complete_cpuset, droppedcpuset);
     modified = 1;
-  } else {
-    if ((flags & HWLOC_RESTRICT_FLAG_REMOVE_CPULESS)
-	&& hwloc_bitmap_iszero(obj->complete_cpuset)) {
-      /* we're empty, there's a NUMAnode below us, it'll be removed this time */
-      modified = 1;
   }
-    /* nodeset cannot intersect unless cpuset intersects or is empty */
+  if (droppednodeset && hwloc_bitmap_intersects(obj->complete_nodeset, droppednodeset)) {
-    if (droppednodeset)
-      assert(!hwloc_bitmap_intersects(obj->complete_nodeset, droppednodeset)
-	     || hwloc_bitmap_iszero(obj->complete_cpuset));
-  }
-  if (droppednodeset) {
     hwloc_bitmap_andnot(obj->nodeset, obj->nodeset, droppednodeset);
     hwloc_bitmap_andnot(obj->complete_nodeset, obj->complete_nodeset, droppednodeset);
+    modified = 1;
   }
 
   if (modified) {
@@ -4251,20 +4255,11 @@ restrict_object_by_nodeset(hwloc_topology_t topology, unsigned long flags, hwloc
     hwloc_bitmap_andnot(obj->nodeset, obj->nodeset, droppednodeset);
     hwloc_bitmap_andnot(obj->complete_nodeset, obj->complete_nodeset, droppednodeset);
     modified = 1;
-  } else {
-    if ((flags & HWLOC_RESTRICT_FLAG_REMOVE_MEMLESS)
-	&& hwloc_bitmap_iszero(obj->complete_nodeset)) {
-      /* we're empty, there's a PU below us, it'll be removed this time */
-      modified = 1;
   }
-    /* cpuset cannot intersect unless nodeset intersects or is empty */
+  if (droppedcpuset && hwloc_bitmap_intersects(obj->complete_cpuset, droppedcpuset)) {
-    if (droppedcpuset)
-      assert(!hwloc_bitmap_intersects(obj->complete_cpuset, droppedcpuset)
-	     || hwloc_bitmap_iszero(obj->complete_nodeset));
-  }
-  if (droppedcpuset) {
     hwloc_bitmap_andnot(obj->cpuset, obj->cpuset, droppedcpuset);
     hwloc_bitmap_andnot(obj->complete_cpuset, obj->complete_cpuset, droppedcpuset);
+    modified = 1;
   }
 
   if (modified) {
@@ -4430,16 +4425,21 @@ hwloc_topology_restrict(struct hwloc_topology *topology, hwloc_const_bitmap_t se
   hwloc_bitmap_free(droppedcpuset);
   hwloc_bitmap_free(droppednodeset);
 
-  if (hwloc_filter_levels_keep_structure(topology) < 0) /* takes care of reconnecting internally */
+  if (hwloc__reconnect(topology, _HWLOC_RECONNECT_FLAG_KEEPSTRUCTURE) < 0)
     goto out;
 
-  /* some objects may have disappeared, we need to update distances objs arrays */
+  /* some objects may have disappeared and sets were modified,
+   * we need to update distances, etc */
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_DISTANCES))
     hwloc_internal_distances_invalidate_cached_objs(topology);
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_MEMATTRS))
     hwloc_internal_memattrs_need_refresh(topology);
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_CPUKINDS))
+    hwloc_internal_cpukinds_restrict(topology);
+
 
   hwloc_propagate_symmetric_subtree(topology, topology->levels[0][0]);
   propagate_total_memory(topology->levels[0][0]);
-  hwloc_internal_cpukinds_restrict(topology);
 
 #ifndef HWLOC_DEBUG
   if (getenv("HWLOC_DEBUG_CHECK"))
@@ -4527,8 +4527,11 @@ hwloc_topology_allow(struct hwloc_topology *topology,
 int
 hwloc_topology_refresh(struct hwloc_topology *topology)
 {
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_CPUKINDS))
     hwloc_internal_cpukinds_rank(topology);
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_DISTANCES))
     hwloc_internal_distances_refresh(topology);
+  if (!(topology->flags & HWLOC_TOPOLOGY_FLAG_NO_MEMATTRS))
     hwloc_internal_memattrs_refresh(topology);
   return 0;
 }
@@ -5044,6 +5047,8 @@ hwloc_topology_check(struct hwloc_topology *topology)
   unsigned i;
   int j, depth;
 
+  assert(!topology->modified);
+
   /* make sure we can use ranges to check types */
 
   /* hwloc__obj_type_is_{,d,i}cache() want cache types to be ordered like this */
@@ -5081,6 +5086,9 @@ hwloc_topology_check(struct hwloc_topology *topology)
   for(i=HWLOC_OBJ_TYPE_MIN; i<HWLOC_OBJ_TYPE_MAX; i++)
     assert(obj_type_order[obj_order_type[i]] == i);
 
+  if (!topology->is_loaded)
+    return;
+
   depth = hwloc_topology_get_depth(topology);
 
   assert(!topology->modified);

src/3rdparty/libethash/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.1)
+cmake_minimum_required(VERSION 3.10)
 project (ethash C)
 
 set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} -Os")

src/3rdparty/libethash/endian.h

@@ -31,7 +31,7 @@
 #include <libkern/OSByteOrder.h>
 #define ethash_swap_u32(input_) OSSwapInt32(input_)
 #define ethash_swap_u64(input_) OSSwapInt64(input_)
-#elif defined(__FreeBSD__) || defined(__DragonFly__) || defined(__NetBSD__)
+#elif defined(__FreeBSD__) || defined(__DragonFly__) || defined(__NetBSD__) || defined(__HAIKU__)
 #define ethash_swap_u32(input_) bswap32(input_)
 #define ethash_swap_u64(input_) bswap64(input_)
 #elif defined(__OpenBSD__)

src/3rdparty/llhttp/api.c

@@ -4,7 +4,7 @@
 
 #include "llhttp.h"
 
-#define CALLBACK_MAYBE(PARSER, NAME, ...)                                     \
+#define CALLBACK_MAYBE(PARSER, NAME)                                          \
   do {                                                                        \
     const llhttp_settings_t* settings;                                        \
     settings = (const llhttp_settings_t*) (PARSER)->settings;                 \
@@ -12,7 +12,22 @@
       err = 0;                                                                \
       break;                                                                  \
     }                                                                         \
-    err = settings->NAME(__VA_ARGS__);                                        \
+    err = settings->NAME((PARSER));                                           \
+  } while (0)
+
+#define SPAN_CALLBACK_MAYBE(PARSER, NAME, START, LEN)                         \
+  do {                                                                        \
+    const llhttp_settings_t* settings;                                        \
+    settings = (const llhttp_settings_t*) (PARSER)->settings;                 \
+    if (settings == NULL || settings->NAME == NULL) {                         \
+      err = 0;                                                                \
+      break;                                                                  \
+    }                                                                         \
+    err = settings->NAME((PARSER), (START), (LEN));                           \
+    if (err == -1) {                                                          \
+      err = HPE_USER;                                                         \
+      llhttp_set_error_reason((PARSER), "Span callback error in " #NAME);     \
+    }                                                                         \
   } while (0)
 
 void llhttp_init(llhttp_t* parser, llhttp_type_t type,
@@ -31,21 +46,25 @@ extern int wasm_on_url(llhttp_t* p, const char* at, size_t length);
 extern int wasm_on_status(llhttp_t* p, const char* at, size_t length);
 extern int wasm_on_header_field(llhttp_t* p, const char* at, size_t length);
 extern int wasm_on_header_value(llhttp_t* p, const char* at, size_t length);
-extern int wasm_on_headers_complete(llhttp_t * p);
+extern int wasm_on_headers_complete(llhttp_t * p, int status_code,
+                                    uint8_t upgrade, int should_keep_alive);
 extern int wasm_on_body(llhttp_t* p, const char* at, size_t length);
 extern int wasm_on_message_complete(llhttp_t * p);
 
+static int wasm_on_headers_complete_wrap(llhttp_t* p) {
+  return wasm_on_headers_complete(p, p->status_code, p->upgrade,
+                                  llhttp_should_keep_alive(p));
+}
+
 const llhttp_settings_t wasm_settings = {
-  wasm_on_message_begin,
+  .on_message_begin = wasm_on_message_begin,
-  wasm_on_url,
+  .on_url = wasm_on_url,
-  wasm_on_status,
+  .on_status = wasm_on_status,
-  wasm_on_header_field,
+  .on_header_field = wasm_on_header_field,
-  wasm_on_header_value,
+  .on_header_value = wasm_on_header_value,
-  wasm_on_headers_complete,
+  .on_headers_complete = wasm_on_headers_complete_wrap,
-  wasm_on_body,
+  .on_body = wasm_on_body,
-  wasm_on_message_complete,
+  .on_message_complete = wasm_on_message_complete,
-  NULL,
-  NULL,
 };
 
 
@@ -59,6 +78,8 @@ void llhttp_free(llhttp_t* parser) {
   free(parser);
 }
 
+#endif  // defined(__wasm__)
+
 /* Some getters required to get stuff from the parser */
 
 uint8_t llhttp_get_type(llhttp_t* parser) {
@@ -85,14 +106,12 @@ uint8_t llhttp_get_upgrade(llhttp_t* parser) {
   return parser->upgrade;
 }
 
-#endif  // defined(__wasm__)
-
 
 void llhttp_reset(llhttp_t* parser) {
   llhttp_type_t type = parser->type;
   const llhttp_settings_t* settings = parser->settings;
   void* data = parser->data;
-  uint8_t lenient_flags = parser->lenient_flags;
+  uint16_t lenient_flags = parser->lenient_flags;
 
   llhttp__internal_init(parser);
 
@@ -123,7 +142,7 @@ llhttp_errno_t llhttp_finish(llhttp_t* parser) {
 
   switch (parser->finish) {
     case HTTP_FINISH_SAFE_WITH_CB:
-      CALLBACK_MAYBE(parser, on_message_complete, parser);
+      CALLBACK_MAYBE(parser, on_message_complete);
       if (err != HPE_OK) return err;
 
     /* FALLTHROUGH */
@@ -199,12 +218,21 @@ const char* llhttp_errno_name(llhttp_errno_t err) {
 const char* llhttp_method_name(llhttp_method_t method) {
 #define HTTP_METHOD_GEN(NUM, NAME, STRING) case HTTP_##NAME: return #STRING;
   switch (method) {
-    HTTP_METHOD_MAP(HTTP_METHOD_GEN)
+    HTTP_ALL_METHOD_MAP(HTTP_METHOD_GEN)
     default: abort();
   }
 #undef HTTP_METHOD_GEN
 }
 
+const char* llhttp_status_name(llhttp_status_t status) {
+#define HTTP_STATUS_GEN(NUM, NAME, STRING) case HTTP_STATUS_##NAME: return #STRING;
+  switch (status) {
+    HTTP_STATUS_MAP(HTTP_STATUS_GEN)
+    default: abort();
+  }
+#undef HTTP_STATUS_GEN
+}
+
 
 void llhttp_set_lenient_headers(llhttp_t* parser, int enabled) {
   if (enabled) {
@@ -232,103 +260,236 @@ void llhttp_set_lenient_keep_alive(llhttp_t* parser, int enabled) {
   }
 }
 
+void llhttp_set_lenient_transfer_encoding(llhttp_t* parser, int enabled) {
+  if (enabled) {
+    parser->lenient_flags |= LENIENT_TRANSFER_ENCODING;
+  } else {
+    parser->lenient_flags &= ~LENIENT_TRANSFER_ENCODING;
+  }
+}
+
+void llhttp_set_lenient_version(llhttp_t* parser, int enabled) {
+  if (enabled) {
+    parser->lenient_flags |= LENIENT_VERSION;
+  } else {
+    parser->lenient_flags &= ~LENIENT_VERSION;
+  }
+}
+
+void llhttp_set_lenient_data_after_close(llhttp_t* parser, int enabled) {
+  if (enabled) {
+    parser->lenient_flags |= LENIENT_DATA_AFTER_CLOSE;
+  } else {
+    parser->lenient_flags &= ~LENIENT_DATA_AFTER_CLOSE;
+  }
+}
+
+void llhttp_set_lenient_optional_lf_after_cr(llhttp_t* parser, int enabled) {
+  if (enabled) {
+    parser->lenient_flags |= LENIENT_OPTIONAL_LF_AFTER_CR;
+  } else {
+    parser->lenient_flags &= ~LENIENT_OPTIONAL_LF_AFTER_CR;
+  }
+}
+
+void llhttp_set_lenient_optional_crlf_after_chunk(llhttp_t* parser, int enabled) {
+  if (enabled) {
+    parser->lenient_flags |= LENIENT_OPTIONAL_CRLF_AFTER_CHUNK;
+  } else {
+    parser->lenient_flags &= ~LENIENT_OPTIONAL_CRLF_AFTER_CHUNK;
+  }
+}
+
+void llhttp_set_lenient_optional_cr_before_lf(llhttp_t* parser, int enabled) {
+  if (enabled) {
+    parser->lenient_flags |= LENIENT_OPTIONAL_CR_BEFORE_LF;
+  } else {
+    parser->lenient_flags &= ~LENIENT_OPTIONAL_CR_BEFORE_LF;
+  }
+}
+
+void llhttp_set_lenient_spaces_after_chunk_size(llhttp_t* parser, int enabled) {
+  if (enabled) {
+    parser->lenient_flags |= LENIENT_SPACES_AFTER_CHUNK_SIZE;
+  } else {
+    parser->lenient_flags &= ~LENIENT_SPACES_AFTER_CHUNK_SIZE;
+  }
+}
+
 /* Callbacks */
 
 
 int llhttp__on_message_begin(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_message_begin, s);
+  CALLBACK_MAYBE(s, on_message_begin);
+  return err;
+}
+
+
+int llhttp__on_protocol(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  SPAN_CALLBACK_MAYBE(s, on_protocol, p, endp - p);
+  return err;
+}
+
+
+int llhttp__on_protocol_complete(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  CALLBACK_MAYBE(s, on_protocol_complete);
   return err;
 }
 
 
 int llhttp__on_url(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_url, s, p, endp - p);
+  SPAN_CALLBACK_MAYBE(s, on_url, p, endp - p);
   return err;
 }
 
 
 int llhttp__on_url_complete(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_url_complete, s);
+  CALLBACK_MAYBE(s, on_url_complete);
   return err;
 }
 
 
 int llhttp__on_status(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_status, s, p, endp - p);
+  SPAN_CALLBACK_MAYBE(s, on_status, p, endp - p);
   return err;
 }
 
 
 int llhttp__on_status_complete(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_status_complete, s);
+  CALLBACK_MAYBE(s, on_status_complete);
+  return err;
+}
+
+
+int llhttp__on_method(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  SPAN_CALLBACK_MAYBE(s, on_method, p, endp - p);
+  return err;
+}
+
+
+int llhttp__on_method_complete(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  CALLBACK_MAYBE(s, on_method_complete);
+  return err;
+}
+
+
+int llhttp__on_version(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  SPAN_CALLBACK_MAYBE(s, on_version, p, endp - p);
+  return err;
+}
+
+
+int llhttp__on_version_complete(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  CALLBACK_MAYBE(s, on_version_complete);
   return err;
 }
 
 
 int llhttp__on_header_field(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_header_field, s, p, endp - p);
+  SPAN_CALLBACK_MAYBE(s, on_header_field, p, endp - p);
   return err;
 }
 
 
 int llhttp__on_header_field_complete(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_header_field_complete, s);
+  CALLBACK_MAYBE(s, on_header_field_complete);
   return err;
 }
 
 
 int llhttp__on_header_value(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_header_value, s, p, endp - p);
+  SPAN_CALLBACK_MAYBE(s, on_header_value, p, endp - p);
   return err;
 }
 
 
 int llhttp__on_header_value_complete(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_header_value_complete, s);
+  CALLBACK_MAYBE(s, on_header_value_complete);
   return err;
 }
 
 
 int llhttp__on_headers_complete(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_headers_complete, s);
+  CALLBACK_MAYBE(s, on_headers_complete);
   return err;
 }
 
 
 int llhttp__on_message_complete(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_message_complete, s);
+  CALLBACK_MAYBE(s, on_message_complete);
   return err;
 }
 
 
 int llhttp__on_body(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_body, s, p, endp - p);
+  SPAN_CALLBACK_MAYBE(s, on_body, p, endp - p);
   return err;
 }
 
 
 int llhttp__on_chunk_header(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_chunk_header, s);
+  CALLBACK_MAYBE(s, on_chunk_header);
+  return err;
+}
+
+
+int llhttp__on_chunk_extension_name(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  SPAN_CALLBACK_MAYBE(s, on_chunk_extension_name, p, endp - p);
+  return err;
+}
+
+
+int llhttp__on_chunk_extension_name_complete(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  CALLBACK_MAYBE(s, on_chunk_extension_name_complete);
+  return err;
+}
+
+
+int llhttp__on_chunk_extension_value(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  SPAN_CALLBACK_MAYBE(s, on_chunk_extension_value, p, endp - p);
+  return err;
+}
+
+
+int llhttp__on_chunk_extension_value_complete(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  CALLBACK_MAYBE(s, on_chunk_extension_value_complete);
   return err;
 }
 
 
 int llhttp__on_chunk_complete(llhttp_t* s, const char* p, const char* endp) {
   int err;
-  CALLBACK_MAYBE(s, on_chunk_complete, s);
+  CALLBACK_MAYBE(s, on_chunk_complete);
+  return err;
+}
+
+
+int llhttp__on_reset(llhttp_t* s, const char* p, const char* endp) {
+  int err;
+  CALLBACK_MAYBE(s, on_reset);
   return err;
 }
 

src/3rdparty/llhttp/api.h

@@ -1,253 +0,0 @@
-#ifndef INCLUDE_LLHTTP_API_H_
-#define INCLUDE_LLHTTP_API_H_
-#ifdef __cplusplus
-extern "C" {
-#endif
-#include <stddef.h>
-
-#if defined(__wasm__)
-#define LLHTTP_EXPORT __attribute__((visibility("default")))
-#else
-#define LLHTTP_EXPORT
-#endif
-
-typedef llhttp__internal_t llhttp_t;
-typedef struct llhttp_settings_s llhttp_settings_t;
-
-typedef int (*llhttp_data_cb)(llhttp_t*, const char *at, size_t length);
-typedef int (*llhttp_cb)(llhttp_t*);
-
-struct llhttp_settings_s {
-  /* Possible return values 0, -1, `HPE_PAUSED` */
-  llhttp_cb      on_message_begin;
-
-  llhttp_data_cb on_url;
-  llhttp_data_cb on_status;
-  llhttp_data_cb on_header_field;
-  llhttp_data_cb on_header_value;
-
-  /* Possible return values:
-   * 0  - Proceed normally
-   * 1  - Assume that request/response has no body, and proceed to parsing the
-   *      next message
-   * 2  - Assume absence of body (as above) and make `llhttp_execute()` return
-   *      `HPE_PAUSED_UPGRADE`
-   * -1 - Error
-   * `HPE_PAUSED`
-   */
-  llhttp_cb      on_headers_complete;
-
-  llhttp_data_cb on_body;
-
-  /* Possible return values 0, -1, `HPE_PAUSED` */
-  llhttp_cb      on_message_complete;
-
-  /* When on_chunk_header is called, the current chunk length is stored
-   * in parser->content_length.
-   * Possible return values 0, -1, `HPE_PAUSED`
-   */
-  llhttp_cb      on_chunk_header;
-  llhttp_cb      on_chunk_complete;
-
-  llhttp_cb      on_url_complete;
-  llhttp_cb      on_status_complete;
-  llhttp_cb      on_header_field_complete;
-  llhttp_cb      on_header_value_complete;
-};
-
-/* Initialize the parser with specific type and user settings.
- *
- * NOTE: lifetime of `settings` has to be at least the same as the lifetime of
- * the `parser` here. In practice, `settings` has to be either a static
- * variable or be allocated with `malloc`, `new`, etc.
- */
-LLHTTP_EXPORT
-void llhttp_init(llhttp_t* parser, llhttp_type_t type,
-                 const llhttp_settings_t* settings);
-
-#if defined(__wasm__)
-
-LLHTTP_EXPORT
-llhttp_t* llhttp_alloc(llhttp_type_t type);
-
-LLHTTP_EXPORT
-void llhttp_free(llhttp_t* parser);
-
-LLHTTP_EXPORT
-uint8_t llhttp_get_type(llhttp_t* parser);
-
-LLHTTP_EXPORT
-uint8_t llhttp_get_http_major(llhttp_t* parser);
-
-LLHTTP_EXPORT
-uint8_t llhttp_get_http_minor(llhttp_t* parser);
-
-LLHTTP_EXPORT
-uint8_t llhttp_get_method(llhttp_t* parser);
-
-LLHTTP_EXPORT
-int llhttp_get_status_code(llhttp_t* parser);
-
-LLHTTP_EXPORT
-uint8_t llhttp_get_upgrade(llhttp_t* parser);
-
-#endif  // defined(__wasm__)
-
-/* Reset an already initialized parser back to the start state, preserving the
- * existing parser type, callback settings, user data, and lenient flags.
- */
-LLHTTP_EXPORT
-void llhttp_reset(llhttp_t* parser);
-
-/* Initialize the settings object */
-LLHTTP_EXPORT
-void llhttp_settings_init(llhttp_settings_t* settings);
-
-/* Parse full or partial request/response, invoking user callbacks along the
- * way.
- *
- * If any of `llhttp_data_cb` returns errno not equal to `HPE_OK` - the parsing
- * interrupts, and such errno is returned from `llhttp_execute()`. If
- * `HPE_PAUSED` was used as a errno, the execution can be resumed with
- * `llhttp_resume()` call.
- *
- * In a special case of CONNECT/Upgrade request/response `HPE_PAUSED_UPGRADE`
- * is returned after fully parsing the request/response. If the user wishes to
- * continue parsing, they need to invoke `llhttp_resume_after_upgrade()`.
- *
- * NOTE: if this function ever returns a non-pause type error, it will continue
- * to return the same error upon each successive call up until `llhttp_init()`
- * is called.
- */
-LLHTTP_EXPORT
-llhttp_errno_t llhttp_execute(llhttp_t* parser, const char* data, size_t len);
-
-/* This method should be called when the other side has no further bytes to
- * send (e.g. shutdown of readable side of the TCP connection.)
- *
- * Requests without `Content-Length` and other messages might require treating
- * all incoming bytes as the part of the body, up to the last byte of the
- * connection. This method will invoke `on_message_complete()` callback if the
- * request was terminated safely. Otherwise a error code would be returned.
- */
-LLHTTP_EXPORT
-llhttp_errno_t llhttp_finish(llhttp_t* parser);
-
-/* Returns `1` if the incoming message is parsed until the last byte, and has
- * to be completed by calling `llhttp_finish()` on EOF
- */
-LLHTTP_EXPORT
-int llhttp_message_needs_eof(const llhttp_t* parser);
-
-/* Returns `1` if there might be any other messages following the last that was
- * successfully parsed.
- */
-LLHTTP_EXPORT
-int llhttp_should_keep_alive(const llhttp_t* parser);
-
-/* Make further calls of `llhttp_execute()` return `HPE_PAUSED` and set
- * appropriate error reason.
- *
- * Important: do not call this from user callbacks! User callbacks must return
- * `HPE_PAUSED` if pausing is required.
- */
-LLHTTP_EXPORT
-void llhttp_pause(llhttp_t* parser);
-
-/* Might be called to resume the execution after the pause in user's callback.
- * See `llhttp_execute()` above for details.
- *
- * Call this only if `llhttp_execute()` returns `HPE_PAUSED`.
- */
-LLHTTP_EXPORT
-void llhttp_resume(llhttp_t* parser);
-
-/* Might be called to resume the execution after the pause in user's callback.
- * See `llhttp_execute()` above for details.
- *
- * Call this only if `llhttp_execute()` returns `HPE_PAUSED_UPGRADE`
- */
-LLHTTP_EXPORT
-void llhttp_resume_after_upgrade(llhttp_t* parser);
-
-/* Returns the latest return error */
-LLHTTP_EXPORT
-llhttp_errno_t llhttp_get_errno(const llhttp_t* parser);
-
-/* Returns the verbal explanation of the latest returned error.
- *
- * Note: User callback should set error reason when returning the error. See
- * `llhttp_set_error_reason()` for details.
- */
-LLHTTP_EXPORT
-const char* llhttp_get_error_reason(const llhttp_t* parser);
-
-/* Assign verbal description to the returned error. Must be called in user
- * callbacks right before returning the errno.
- *
- * Note: `HPE_USER` error code might be useful in user callbacks.
- */
-LLHTTP_EXPORT
-void llhttp_set_error_reason(llhttp_t* parser, const char* reason);
-
-/* Returns the pointer to the last parsed byte before the returned error. The
- * pointer is relative to the `data` argument of `llhttp_execute()`.
- *
- * Note: this method might be useful for counting the number of parsed bytes.
- */
-LLHTTP_EXPORT
-const char* llhttp_get_error_pos(const llhttp_t* parser);
-
-/* Returns textual name of error code */
-LLHTTP_EXPORT
-const char* llhttp_errno_name(llhttp_errno_t err);
-
-/* Returns textual name of HTTP method */
-LLHTTP_EXPORT
-const char* llhttp_method_name(llhttp_method_t method);
-
-
-/* Enables/disables lenient header value parsing (disabled by default).
- *
- * Lenient parsing disables header value token checks, extending llhttp's
- * protocol support to highly non-compliant clients/server. No
- * `HPE_INVALID_HEADER_TOKEN` will be raised for incorrect header values when
- * lenient parsing is "on".
- *
- * **(USE AT YOUR OWN RISK)**
- */
-LLHTTP_EXPORT
-void llhttp_set_lenient_headers(llhttp_t* parser, int enabled);
-
-
-/* Enables/disables lenient handling of conflicting `Transfer-Encoding` and
- * `Content-Length` headers (disabled by default).
- *
- * Normally `llhttp` would error when `Transfer-Encoding` is present in
- * conjunction with `Content-Length`. This error is important to prevent HTTP
- * request smuggling, but may be less desirable for small number of cases
- * involving legacy servers.
- *
- * **(USE AT YOUR OWN RISK)**
- */
-LLHTTP_EXPORT
-void llhttp_set_lenient_chunked_length(llhttp_t* parser, int enabled);
-
-
-/* Enables/disables lenient handling of `Connection: close` and HTTP/1.0
- * requests responses.
- *
- * Normally `llhttp` would error on (in strict mode) or discard (in loose mode)
- * the HTTP request/response after the request/response with `Connection: close`
- * and `Content-Length`. This is important to prevent cache poisoning attacks,
- * but might interact badly with outdated and insecure clients. With this flag
- * the extra request/response will be parsed normally.
- *
- * **(USE AT YOUR OWN RISK)**
- */
-void llhttp_set_lenient_keep_alive(llhttp_t* parser, int enabled);
-
-#ifdef __cplusplus
-}  /* extern "C" */
-#endif
-#endif  /* INCLUDE_LLHTTP_API_H_ */

src/3rdparty/llhttp/http.c

@@ -39,20 +39,41 @@ int llhttp__after_headers_complete(llhttp_t* parser, const char* p,
   int hasBody;
 
   hasBody = parser->flags & F_CHUNKED || parser->content_length > 0;
-  if (parser->upgrade && (parser->method == HTTP_CONNECT ||
+  if (
-                          (parser->flags & F_SKIPBODY) || !hasBody)) {
+      (parser->upgrade && (parser->method == HTTP_CONNECT ||
+                          (parser->flags & F_SKIPBODY) || !hasBody)) ||
+      /* See RFC 2616 section 4.4 - 1xx e.g. Continue */
+      (parser->type == HTTP_RESPONSE && parser->status_code == 101)
+  ) {
     /* Exit, the rest of the message is in a different protocol. */
     return 1;
   }
 
-  if (parser->flags & F_SKIPBODY) {
+  if (parser->type == HTTP_RESPONSE && parser->status_code == 100) {
+    /* No body, restart as the message is complete */
+    return 0;
+  }
+
+  /* See RFC 2616 section 4.4 */
+  if (
+    parser->flags & F_SKIPBODY ||         /* response to a HEAD request */
+    (
+      parser->type == HTTP_RESPONSE && (
+        parser->status_code == 102 ||     /* Processing */
+        parser->status_code == 103 ||     /* Early Hints */
+        parser->status_code == 204 ||     /* No Content */
+        parser->status_code == 304        /* Not Modified */
+      )
+    )
+  ) {
     return 0;
   } else if (parser->flags & F_CHUNKED) {
     /* chunked encoding - ignore Content-Length header, prepare for a chunk */
     return 2;
   } else if (parser->flags & F_TRANSFER_ENCODING) {
     if (parser->type == HTTP_REQUEST &&
-        (parser->lenient_flags & LENIENT_CHUNKED_LENGTH) == 0) {
+        (parser->lenient_flags & LENIENT_CHUNKED_LENGTH) == 0 &&
+        (parser->lenient_flags & LENIENT_TRANSFER_ENCODING) == 0) {
       /* RFC 7230 3.3.3 */
 
       /* If a Transfer-Encoding header field

src/3rdparty/llhttp/llhttp.h

@@ -1,14 +1,11 @@
+
 #ifndef INCLUDE_LLHTTP_H_
 #define INCLUDE_LLHTTP_H_
 
-#define LLHTTP_VERSION_MAJOR 5
+#define LLHTTP_VERSION_MAJOR 9
-#define LLHTTP_VERSION_MINOR 1
+#define LLHTTP_VERSION_MINOR 3
 #define LLHTTP_VERSION_PATCH 0
 
-#ifndef LLHTTP_STRICT_MODE
-# define LLHTTP_STRICT_MODE 0
-#endif
-
 #ifndef INCLUDE_LLHTTP_ITSELF_H_
 #define INCLUDE_LLHTTP_ITSELF_H_
 #ifdef __cplusplus
@@ -33,11 +30,12 @@ struct llhttp__internal_s {
   uint8_t http_major;
   uint8_t http_minor;
   uint8_t header_state;
-  uint8_t lenient_flags;
+  uint16_t lenient_flags;
   uint8_t upgrade;
   uint8_t finish;
   uint16_t flags;
   uint16_t status_code;
+  uint8_t initial_message_completed;
   void* settings;
 };
 
@@ -49,6 +47,7 @@ int llhttp__internal_execute(llhttp__internal_t* s, const char* p, const char* e
 #endif
 #endif  /* INCLUDE_LLHTTP_ITSELF_H_ */
 
+
 #ifndef LLLLHTTP_C_HEADERS_
 #define LLLLHTTP_C_HEADERS_
 #ifdef __cplusplus
@@ -59,8 +58,10 @@ enum llhttp_errno {
   HPE_OK = 0,
   HPE_INTERNAL = 1,
   HPE_STRICT = 2,
+  HPE_CR_EXPECTED = 25,
   HPE_LF_EXPECTED = 3,
   HPE_UNEXPECTED_CONTENT_LENGTH = 4,
+  HPE_UNEXPECTED_SPACE = 30,
   HPE_CLOSED_CONNECTION = 5,
   HPE_INVALID_METHOD = 6,
   HPE_INVALID_URL = 7,
@@ -80,7 +81,17 @@ enum llhttp_errno {
   HPE_PAUSED = 21,
   HPE_PAUSED_UPGRADE = 22,
   HPE_PAUSED_H2_UPGRADE = 23,
-  HPE_USER = 24
+  HPE_USER = 24,
+  HPE_CB_URL_COMPLETE = 26,
+  HPE_CB_STATUS_COMPLETE = 27,
+  HPE_CB_METHOD_COMPLETE = 32,
+  HPE_CB_VERSION_COMPLETE = 33,
+  HPE_CB_HEADER_FIELD_COMPLETE = 28,
+  HPE_CB_HEADER_VALUE_COMPLETE = 29,
+  HPE_CB_CHUNK_EXTENSION_NAME_COMPLETE = 34,
+  HPE_CB_CHUNK_EXTENSION_VALUE_COMPLETE = 35,
+  HPE_CB_RESET = 31,
+  HPE_CB_PROTOCOL_COMPLETE = 38
 };
 typedef enum llhttp_errno llhttp_errno_t;
 
@@ -100,7 +111,14 @@ typedef enum llhttp_flags llhttp_flags_t;
 enum llhttp_lenient_flags {
   LENIENT_HEADERS = 0x1,
   LENIENT_CHUNKED_LENGTH = 0x2,
-  LENIENT_KEEP_ALIVE = 0x4
+  LENIENT_KEEP_ALIVE = 0x4,
+  LENIENT_TRANSFER_ENCODING = 0x8,
+  LENIENT_VERSION = 0x10,
+  LENIENT_DATA_AFTER_CLOSE = 0x20,
+  LENIENT_OPTIONAL_LF_AFTER_CR = 0x40,
+  LENIENT_OPTIONAL_CRLF_AFTER_CHUNK = 0x80,
+  LENIENT_OPTIONAL_CR_BEFORE_LF = 0x100,
+  LENIENT_SPACES_AFTER_CHUNK_SIZE = 0x200
 };
 typedef enum llhttp_lenient_flags llhttp_lenient_flags_t;
 
@@ -164,16 +182,122 @@ enum llhttp_method {
   HTTP_SET_PARAMETER = 42,
   HTTP_REDIRECT = 43,
   HTTP_RECORD = 44,
-  HTTP_FLUSH = 45
+  HTTP_FLUSH = 45,
+  HTTP_QUERY = 46
 };
 typedef enum llhttp_method llhttp_method_t;
 
+enum llhttp_status {
+  HTTP_STATUS_CONTINUE = 100,
+  HTTP_STATUS_SWITCHING_PROTOCOLS = 101,
+  HTTP_STATUS_PROCESSING = 102,
+  HTTP_STATUS_EARLY_HINTS = 103,
+  HTTP_STATUS_RESPONSE_IS_STALE = 110,
+  HTTP_STATUS_REVALIDATION_FAILED = 111,
+  HTTP_STATUS_DISCONNECTED_OPERATION = 112,
+  HTTP_STATUS_HEURISTIC_EXPIRATION = 113,
+  HTTP_STATUS_MISCELLANEOUS_WARNING = 199,
+  HTTP_STATUS_OK = 200,
+  HTTP_STATUS_CREATED = 201,
+  HTTP_STATUS_ACCEPTED = 202,
+  HTTP_STATUS_NON_AUTHORITATIVE_INFORMATION = 203,
+  HTTP_STATUS_NO_CONTENT = 204,
+  HTTP_STATUS_RESET_CONTENT = 205,
+  HTTP_STATUS_PARTIAL_CONTENT = 206,
+  HTTP_STATUS_MULTI_STATUS = 207,
+  HTTP_STATUS_ALREADY_REPORTED = 208,
+  HTTP_STATUS_TRANSFORMATION_APPLIED = 214,
+  HTTP_STATUS_IM_USED = 226,
+  HTTP_STATUS_MISCELLANEOUS_PERSISTENT_WARNING = 299,
+  HTTP_STATUS_MULTIPLE_CHOICES = 300,
+  HTTP_STATUS_MOVED_PERMANENTLY = 301,
+  HTTP_STATUS_FOUND = 302,
+  HTTP_STATUS_SEE_OTHER = 303,
+  HTTP_STATUS_NOT_MODIFIED = 304,
+  HTTP_STATUS_USE_PROXY = 305,
+  HTTP_STATUS_SWITCH_PROXY = 306,
+  HTTP_STATUS_TEMPORARY_REDIRECT = 307,
+  HTTP_STATUS_PERMANENT_REDIRECT = 308,
+  HTTP_STATUS_BAD_REQUEST = 400,
+  HTTP_STATUS_UNAUTHORIZED = 401,
+  HTTP_STATUS_PAYMENT_REQUIRED = 402,
+  HTTP_STATUS_FORBIDDEN = 403,
+  HTTP_STATUS_NOT_FOUND = 404,
+  HTTP_STATUS_METHOD_NOT_ALLOWED = 405,
+  HTTP_STATUS_NOT_ACCEPTABLE = 406,
+  HTTP_STATUS_PROXY_AUTHENTICATION_REQUIRED = 407,
+  HTTP_STATUS_REQUEST_TIMEOUT = 408,
+  HTTP_STATUS_CONFLICT = 409,
+  HTTP_STATUS_GONE = 410,
+  HTTP_STATUS_LENGTH_REQUIRED = 411,
+  HTTP_STATUS_PRECONDITION_FAILED = 412,
+  HTTP_STATUS_PAYLOAD_TOO_LARGE = 413,
+  HTTP_STATUS_URI_TOO_LONG = 414,
+  HTTP_STATUS_UNSUPPORTED_MEDIA_TYPE = 415,
+  HTTP_STATUS_RANGE_NOT_SATISFIABLE = 416,
+  HTTP_STATUS_EXPECTATION_FAILED = 417,
+  HTTP_STATUS_IM_A_TEAPOT = 418,
+  HTTP_STATUS_PAGE_EXPIRED = 419,
+  HTTP_STATUS_ENHANCE_YOUR_CALM = 420,
+  HTTP_STATUS_MISDIRECTED_REQUEST = 421,
+  HTTP_STATUS_UNPROCESSABLE_ENTITY = 422,
+  HTTP_STATUS_LOCKED = 423,
+  HTTP_STATUS_FAILED_DEPENDENCY = 424,
+  HTTP_STATUS_TOO_EARLY = 425,
+  HTTP_STATUS_UPGRADE_REQUIRED = 426,
+  HTTP_STATUS_PRECONDITION_REQUIRED = 428,
+  HTTP_STATUS_TOO_MANY_REQUESTS = 429,
+  HTTP_STATUS_REQUEST_HEADER_FIELDS_TOO_LARGE_UNOFFICIAL = 430,
+  HTTP_STATUS_REQUEST_HEADER_FIELDS_TOO_LARGE = 431,
+  HTTP_STATUS_LOGIN_TIMEOUT = 440,
+  HTTP_STATUS_NO_RESPONSE = 444,
+  HTTP_STATUS_RETRY_WITH = 449,
+  HTTP_STATUS_BLOCKED_BY_PARENTAL_CONTROL = 450,
+  HTTP_STATUS_UNAVAILABLE_FOR_LEGAL_REASONS = 451,
+  HTTP_STATUS_CLIENT_CLOSED_LOAD_BALANCED_REQUEST = 460,
+  HTTP_STATUS_INVALID_X_FORWARDED_FOR = 463,
+  HTTP_STATUS_REQUEST_HEADER_TOO_LARGE = 494,
+  HTTP_STATUS_SSL_CERTIFICATE_ERROR = 495,
+  HTTP_STATUS_SSL_CERTIFICATE_REQUIRED = 496,
+  HTTP_STATUS_HTTP_REQUEST_SENT_TO_HTTPS_PORT = 497,
+  HTTP_STATUS_INVALID_TOKEN = 498,
+  HTTP_STATUS_CLIENT_CLOSED_REQUEST = 499,
+  HTTP_STATUS_INTERNAL_SERVER_ERROR = 500,
+  HTTP_STATUS_NOT_IMPLEMENTED = 501,
+  HTTP_STATUS_BAD_GATEWAY = 502,
+  HTTP_STATUS_SERVICE_UNAVAILABLE = 503,
+  HTTP_STATUS_GATEWAY_TIMEOUT = 504,
+  HTTP_STATUS_HTTP_VERSION_NOT_SUPPORTED = 505,
+  HTTP_STATUS_VARIANT_ALSO_NEGOTIATES = 506,
+  HTTP_STATUS_INSUFFICIENT_STORAGE = 507,
+  HTTP_STATUS_LOOP_DETECTED = 508,
+  HTTP_STATUS_BANDWIDTH_LIMIT_EXCEEDED = 509,
+  HTTP_STATUS_NOT_EXTENDED = 510,
+  HTTP_STATUS_NETWORK_AUTHENTICATION_REQUIRED = 511,
+  HTTP_STATUS_WEB_SERVER_UNKNOWN_ERROR = 520,
+  HTTP_STATUS_WEB_SERVER_IS_DOWN = 521,
+  HTTP_STATUS_CONNECTION_TIMEOUT = 522,
+  HTTP_STATUS_ORIGIN_IS_UNREACHABLE = 523,
+  HTTP_STATUS_TIMEOUT_OCCURED = 524,
+  HTTP_STATUS_SSL_HANDSHAKE_FAILED = 525,
+  HTTP_STATUS_INVALID_SSL_CERTIFICATE = 526,
+  HTTP_STATUS_RAILGUN_ERROR = 527,
+  HTTP_STATUS_SITE_IS_OVERLOADED = 529,
+  HTTP_STATUS_SITE_IS_FROZEN = 530,
+  HTTP_STATUS_IDENTITY_PROVIDER_AUTHENTICATION_ERROR = 561,
+  HTTP_STATUS_NETWORK_READ_TIMEOUT = 598,
+  HTTP_STATUS_NETWORK_CONNECT_TIMEOUT = 599
+};
+typedef enum llhttp_status llhttp_status_t;
+
 #define HTTP_ERRNO_MAP(XX) \
   XX(0, OK, OK) \
   XX(1, INTERNAL, INTERNAL) \
   XX(2, STRICT, STRICT) \
+  XX(25, CR_EXPECTED, CR_EXPECTED) \
   XX(3, LF_EXPECTED, LF_EXPECTED) \
   XX(4, UNEXPECTED_CONTENT_LENGTH, UNEXPECTED_CONTENT_LENGTH) \
+  XX(30, UNEXPECTED_SPACE, UNEXPECTED_SPACE) \
   XX(5, CLOSED_CONNECTION, CLOSED_CONNECTION) \
   XX(6, INVALID_METHOD, INVALID_METHOD) \
   XX(7, INVALID_URL, INVALID_URL) \
@@ -194,9 +318,74 @@ typedef enum llhttp_method llhttp_method_t;
   XX(22, PAUSED_UPGRADE, PAUSED_UPGRADE) \
   XX(23, PAUSED_H2_UPGRADE, PAUSED_H2_UPGRADE) \
   XX(24, USER, USER) \
+  XX(26, CB_URL_COMPLETE, CB_URL_COMPLETE) \
+  XX(27, CB_STATUS_COMPLETE, CB_STATUS_COMPLETE) \
+  XX(32, CB_METHOD_COMPLETE, CB_METHOD_COMPLETE) \
+  XX(33, CB_VERSION_COMPLETE, CB_VERSION_COMPLETE) \
+  XX(28, CB_HEADER_FIELD_COMPLETE, CB_HEADER_FIELD_COMPLETE) \
+  XX(29, CB_HEADER_VALUE_COMPLETE, CB_HEADER_VALUE_COMPLETE) \
+  XX(34, CB_CHUNK_EXTENSION_NAME_COMPLETE, CB_CHUNK_EXTENSION_NAME_COMPLETE) \
+  XX(35, CB_CHUNK_EXTENSION_VALUE_COMPLETE, CB_CHUNK_EXTENSION_VALUE_COMPLETE) \
+  XX(31, CB_RESET, CB_RESET) \
+  XX(38, CB_PROTOCOL_COMPLETE, CB_PROTOCOL_COMPLETE) \
 
 
 #define HTTP_METHOD_MAP(XX) \
+  XX(0, DELETE, DELETE) \
+  XX(1, GET, GET) \
+  XX(2, HEAD, HEAD) \
+  XX(3, POST, POST) \
+  XX(4, PUT, PUT) \
+  XX(5, CONNECT, CONNECT) \
+  XX(6, OPTIONS, OPTIONS) \
+  XX(7, TRACE, TRACE) \
+  XX(8, COPY, COPY) \
+  XX(9, LOCK, LOCK) \
+  XX(10, MKCOL, MKCOL) \
+  XX(11, MOVE, MOVE) \
+  XX(12, PROPFIND, PROPFIND) \
+  XX(13, PROPPATCH, PROPPATCH) \
+  XX(14, SEARCH, SEARCH) \
+  XX(15, UNLOCK, UNLOCK) \
+  XX(16, BIND, BIND) \
+  XX(17, REBIND, REBIND) \
+  XX(18, UNBIND, UNBIND) \
+  XX(19, ACL, ACL) \
+  XX(20, REPORT, REPORT) \
+  XX(21, MKACTIVITY, MKACTIVITY) \
+  XX(22, CHECKOUT, CHECKOUT) \
+  XX(23, MERGE, MERGE) \
+  XX(24, MSEARCH, M-SEARCH) \
+  XX(25, NOTIFY, NOTIFY) \
+  XX(26, SUBSCRIBE, SUBSCRIBE) \
+  XX(27, UNSUBSCRIBE, UNSUBSCRIBE) \
+  XX(28, PATCH, PATCH) \
+  XX(29, PURGE, PURGE) \
+  XX(30, MKCALENDAR, MKCALENDAR) \
+  XX(31, LINK, LINK) \
+  XX(32, UNLINK, UNLINK) \
+  XX(33, SOURCE, SOURCE) \
+  XX(46, QUERY, QUERY) \
+
+
+#define RTSP_METHOD_MAP(XX) \
+  XX(1, GET, GET) \
+  XX(3, POST, POST) \
+  XX(6, OPTIONS, OPTIONS) \
+  XX(35, DESCRIBE, DESCRIBE) \
+  XX(36, ANNOUNCE, ANNOUNCE) \
+  XX(37, SETUP, SETUP) \
+  XX(38, PLAY, PLAY) \
+  XX(39, PAUSE, PAUSE) \
+  XX(40, TEARDOWN, TEARDOWN) \
+  XX(41, GET_PARAMETER, GET_PARAMETER) \
+  XX(42, SET_PARAMETER, SET_PARAMETER) \
+  XX(43, REDIRECT, REDIRECT) \
+  XX(44, RECORD, RECORD) \
+  XX(45, FLUSH, FLUSH) \
+
+
+#define HTTP_ALL_METHOD_MAP(XX) \
   XX(0, DELETE, DELETE) \
   XX(1, GET, GET) \
   XX(2, HEAD, HEAD) \
@@ -243,14 +432,117 @@ typedef enum llhttp_method llhttp_method_t;
   XX(43, REDIRECT, REDIRECT) \
   XX(44, RECORD, RECORD) \
   XX(45, FLUSH, FLUSH) \
+  XX(46, QUERY, QUERY) \
 
 
+#define HTTP_STATUS_MAP(XX) \
+  XX(100, CONTINUE, CONTINUE) \
+  XX(101, SWITCHING_PROTOCOLS, SWITCHING_PROTOCOLS) \
+  XX(102, PROCESSING, PROCESSING) \
+  XX(103, EARLY_HINTS, EARLY_HINTS) \
+  XX(110, RESPONSE_IS_STALE, RESPONSE_IS_STALE) \
+  XX(111, REVALIDATION_FAILED, REVALIDATION_FAILED) \
+  XX(112, DISCONNECTED_OPERATION, DISCONNECTED_OPERATION) \
+  XX(113, HEURISTIC_EXPIRATION, HEURISTIC_EXPIRATION) \
+  XX(199, MISCELLANEOUS_WARNING, MISCELLANEOUS_WARNING) \
+  XX(200, OK, OK) \
+  XX(201, CREATED, CREATED) \
+  XX(202, ACCEPTED, ACCEPTED) \
+  XX(203, NON_AUTHORITATIVE_INFORMATION, NON_AUTHORITATIVE_INFORMATION) \
+  XX(204, NO_CONTENT, NO_CONTENT) \
+  XX(205, RESET_CONTENT, RESET_CONTENT) \
+  XX(206, PARTIAL_CONTENT, PARTIAL_CONTENT) \
+  XX(207, MULTI_STATUS, MULTI_STATUS) \
+  XX(208, ALREADY_REPORTED, ALREADY_REPORTED) \
+  XX(214, TRANSFORMATION_APPLIED, TRANSFORMATION_APPLIED) \
+  XX(226, IM_USED, IM_USED) \
+  XX(299, MISCELLANEOUS_PERSISTENT_WARNING, MISCELLANEOUS_PERSISTENT_WARNING) \
+  XX(300, MULTIPLE_CHOICES, MULTIPLE_CHOICES) \
+  XX(301, MOVED_PERMANENTLY, MOVED_PERMANENTLY) \
+  XX(302, FOUND, FOUND) \
+  XX(303, SEE_OTHER, SEE_OTHER) \
+  XX(304, NOT_MODIFIED, NOT_MODIFIED) \
+  XX(305, USE_PROXY, USE_PROXY) \
+  XX(306, SWITCH_PROXY, SWITCH_PROXY) \
+  XX(307, TEMPORARY_REDIRECT, TEMPORARY_REDIRECT) \
+  XX(308, PERMANENT_REDIRECT, PERMANENT_REDIRECT) \
+  XX(400, BAD_REQUEST, BAD_REQUEST) \
+  XX(401, UNAUTHORIZED, UNAUTHORIZED) \
+  XX(402, PAYMENT_REQUIRED, PAYMENT_REQUIRED) \
+  XX(403, FORBIDDEN, FORBIDDEN) \
+  XX(404, NOT_FOUND, NOT_FOUND) \
+  XX(405, METHOD_NOT_ALLOWED, METHOD_NOT_ALLOWED) \
+  XX(406, NOT_ACCEPTABLE, NOT_ACCEPTABLE) \
+  XX(407, PROXY_AUTHENTICATION_REQUIRED, PROXY_AUTHENTICATION_REQUIRED) \
+  XX(408, REQUEST_TIMEOUT, REQUEST_TIMEOUT) \
+  XX(409, CONFLICT, CONFLICT) \
+  XX(410, GONE, GONE) \
+  XX(411, LENGTH_REQUIRED, LENGTH_REQUIRED) \
+  XX(412, PRECONDITION_FAILED, PRECONDITION_FAILED) \
+  XX(413, PAYLOAD_TOO_LARGE, PAYLOAD_TOO_LARGE) \
+  XX(414, URI_TOO_LONG, URI_TOO_LONG) \
+  XX(415, UNSUPPORTED_MEDIA_TYPE, UNSUPPORTED_MEDIA_TYPE) \
+  XX(416, RANGE_NOT_SATISFIABLE, RANGE_NOT_SATISFIABLE) \
+  XX(417, EXPECTATION_FAILED, EXPECTATION_FAILED) \
+  XX(418, IM_A_TEAPOT, IM_A_TEAPOT) \
+  XX(419, PAGE_EXPIRED, PAGE_EXPIRED) \
+  XX(420, ENHANCE_YOUR_CALM, ENHANCE_YOUR_CALM) \
+  XX(421, MISDIRECTED_REQUEST, MISDIRECTED_REQUEST) \
+  XX(422, UNPROCESSABLE_ENTITY, UNPROCESSABLE_ENTITY) \
+  XX(423, LOCKED, LOCKED) \
+  XX(424, FAILED_DEPENDENCY, FAILED_DEPENDENCY) \
+  XX(425, TOO_EARLY, TOO_EARLY) \
+  XX(426, UPGRADE_REQUIRED, UPGRADE_REQUIRED) \
+  XX(428, PRECONDITION_REQUIRED, PRECONDITION_REQUIRED) \
+  XX(429, TOO_MANY_REQUESTS, TOO_MANY_REQUESTS) \
+  XX(430, REQUEST_HEADER_FIELDS_TOO_LARGE_UNOFFICIAL, REQUEST_HEADER_FIELDS_TOO_LARGE_UNOFFICIAL) \
+  XX(431, REQUEST_HEADER_FIELDS_TOO_LARGE, REQUEST_HEADER_FIELDS_TOO_LARGE) \
+  XX(440, LOGIN_TIMEOUT, LOGIN_TIMEOUT) \
+  XX(444, NO_RESPONSE, NO_RESPONSE) \
+  XX(449, RETRY_WITH, RETRY_WITH) \
+  XX(450, BLOCKED_BY_PARENTAL_CONTROL, BLOCKED_BY_PARENTAL_CONTROL) \
+  XX(451, UNAVAILABLE_FOR_LEGAL_REASONS, UNAVAILABLE_FOR_LEGAL_REASONS) \
+  XX(460, CLIENT_CLOSED_LOAD_BALANCED_REQUEST, CLIENT_CLOSED_LOAD_BALANCED_REQUEST) \
+  XX(463, INVALID_X_FORWARDED_FOR, INVALID_X_FORWARDED_FOR) \
+  XX(494, REQUEST_HEADER_TOO_LARGE, REQUEST_HEADER_TOO_LARGE) \
+  XX(495, SSL_CERTIFICATE_ERROR, SSL_CERTIFICATE_ERROR) \
+  XX(496, SSL_CERTIFICATE_REQUIRED, SSL_CERTIFICATE_REQUIRED) \
+  XX(497, HTTP_REQUEST_SENT_TO_HTTPS_PORT, HTTP_REQUEST_SENT_TO_HTTPS_PORT) \
+  XX(498, INVALID_TOKEN, INVALID_TOKEN) \
+  XX(499, CLIENT_CLOSED_REQUEST, CLIENT_CLOSED_REQUEST) \
+  XX(500, INTERNAL_SERVER_ERROR, INTERNAL_SERVER_ERROR) \
+  XX(501, NOT_IMPLEMENTED, NOT_IMPLEMENTED) \
+  XX(502, BAD_GATEWAY, BAD_GATEWAY) \
+  XX(503, SERVICE_UNAVAILABLE, SERVICE_UNAVAILABLE) \
+  XX(504, GATEWAY_TIMEOUT, GATEWAY_TIMEOUT) \
+  XX(505, HTTP_VERSION_NOT_SUPPORTED, HTTP_VERSION_NOT_SUPPORTED) \
+  XX(506, VARIANT_ALSO_NEGOTIATES, VARIANT_ALSO_NEGOTIATES) \
+  XX(507, INSUFFICIENT_STORAGE, INSUFFICIENT_STORAGE) \
+  XX(508, LOOP_DETECTED, LOOP_DETECTED) \
+  XX(509, BANDWIDTH_LIMIT_EXCEEDED, BANDWIDTH_LIMIT_EXCEEDED) \
+  XX(510, NOT_EXTENDED, NOT_EXTENDED) \
+  XX(511, NETWORK_AUTHENTICATION_REQUIRED, NETWORK_AUTHENTICATION_REQUIRED) \
+  XX(520, WEB_SERVER_UNKNOWN_ERROR, WEB_SERVER_UNKNOWN_ERROR) \
+  XX(521, WEB_SERVER_IS_DOWN, WEB_SERVER_IS_DOWN) \
+  XX(522, CONNECTION_TIMEOUT, CONNECTION_TIMEOUT) \
+  XX(523, ORIGIN_IS_UNREACHABLE, ORIGIN_IS_UNREACHABLE) \
+  XX(524, TIMEOUT_OCCURED, TIMEOUT_OCCURED) \
+  XX(525, SSL_HANDSHAKE_FAILED, SSL_HANDSHAKE_FAILED) \
+  XX(526, INVALID_SSL_CERTIFICATE, INVALID_SSL_CERTIFICATE) \
+  XX(527, RAILGUN_ERROR, RAILGUN_ERROR) \
+  XX(529, SITE_IS_OVERLOADED, SITE_IS_OVERLOADED) \
+  XX(530, SITE_IS_FROZEN, SITE_IS_FROZEN) \
+  XX(561, IDENTITY_PROVIDER_AUTHENTICATION_ERROR, IDENTITY_PROVIDER_AUTHENTICATION_ERROR) \
+  XX(598, NETWORK_READ_TIMEOUT, NETWORK_READ_TIMEOUT) \
+  XX(599, NETWORK_CONNECT_TIMEOUT, NETWORK_CONNECT_TIMEOUT) \
+
 
 #ifdef __cplusplus
 }  /* extern "C" */
 #endif
 #endif  /* LLLLHTTP_C_HEADERS_ */
 
+
 #ifndef INCLUDE_LLHTTP_API_H_
 #define INCLUDE_LLHTTP_API_H_
 #ifdef __cplusplus
@@ -274,10 +566,16 @@ struct llhttp_settings_s {
   /* Possible return values 0, -1, `HPE_PAUSED` */
   llhttp_cb      on_message_begin;
 
+  /* Possible return values 0, -1, HPE_USER */
+  llhttp_data_cb on_protocol;
   llhttp_data_cb on_url;
   llhttp_data_cb on_status;
+  llhttp_data_cb on_method;
+  llhttp_data_cb on_version;
   llhttp_data_cb on_header_field;
   llhttp_data_cb on_header_value;
+  llhttp_data_cb      on_chunk_extension_name;
+  llhttp_data_cb      on_chunk_extension_value;
 
   /* Possible return values:
    * 0  - Proceed normally
@@ -290,10 +588,20 @@ struct llhttp_settings_s {
    */
   llhttp_cb      on_headers_complete;
 
+  /* Possible return values 0, -1, HPE_USER */
   llhttp_data_cb on_body;
 
   /* Possible return values 0, -1, `HPE_PAUSED` */
   llhttp_cb      on_message_complete;
+  llhttp_cb      on_protocol_complete;
+  llhttp_cb      on_url_complete;
+  llhttp_cb      on_status_complete;
+  llhttp_cb      on_method_complete;
+  llhttp_cb      on_version_complete;
+  llhttp_cb      on_header_field_complete;
+  llhttp_cb      on_header_value_complete;
+  llhttp_cb      on_chunk_extension_name_complete;
+  llhttp_cb      on_chunk_extension_value_complete;
 
   /* When on_chunk_header is called, the current chunk length is stored
    * in parser->content_length.
@@ -301,11 +609,7 @@ struct llhttp_settings_s {
    */
   llhttp_cb      on_chunk_header;
   llhttp_cb      on_chunk_complete;
-
+  llhttp_cb      on_reset;
-  llhttp_cb      on_url_complete;
-  llhttp_cb      on_status_complete;
-  llhttp_cb      on_header_field_complete;
-  llhttp_cb      on_header_value_complete;
 };
 
 /* Initialize the parser with specific type and user settings.
@@ -318,8 +622,6 @@ LLHTTP_EXPORT
 void llhttp_init(llhttp_t* parser, llhttp_type_t type,
                  const llhttp_settings_t* settings);
 
-#if defined(__wasm__)
-
 LLHTTP_EXPORT
 llhttp_t* llhttp_alloc(llhttp_type_t type);
 
@@ -344,8 +646,6 @@ int llhttp_get_status_code(llhttp_t* parser);
 LLHTTP_EXPORT
 uint8_t llhttp_get_upgrade(llhttp_t* parser);
 
-#endif  // defined(__wasm__)
-
 /* Reset an already initialized parser back to the start state, preserving the
  * existing parser type, callback settings, user data, and lenient flags.
  */
@@ -459,6 +759,9 @@ const char* llhttp_errno_name(llhttp_errno_t err);
 LLHTTP_EXPORT
 const char* llhttp_method_name(llhttp_method_t method);
 
+/* Returns textual name of HTTP status */
+LLHTTP_EXPORT
+const char* llhttp_status_name(llhttp_status_t status);
 
 /* Enables/disables lenient header value parsing (disabled by default).
  *
@@ -467,7 +770,8 @@ const char* llhttp_method_name(llhttp_method_t method);
  * `HPE_INVALID_HEADER_TOKEN` will be raised for incorrect header values when
  * lenient parsing is "on".
  *
- * **(USE AT YOUR OWN RISK)**
+ * **Enabling this flag can pose a security issue since you will be exposed to
+ * request smuggling attacks. USE WITH CAUTION!**
  */
 LLHTTP_EXPORT
 void llhttp_set_lenient_headers(llhttp_t* parser, int enabled);
@@ -481,7 +785,8 @@ void llhttp_set_lenient_headers(llhttp_t* parser, int enabled);
  * request smuggling, but may be less desirable for small number of cases
  * involving legacy servers.
  *
- * **(USE AT YOUR OWN RISK)**
+ * **Enabling this flag can pose a security issue since you will be exposed to
+ * request smuggling attacks. USE WITH CAUTION!**
  */
 LLHTTP_EXPORT
 void llhttp_set_lenient_chunked_length(llhttp_t* parser, int enabled);
@@ -496,13 +801,105 @@ void llhttp_set_lenient_chunked_length(llhttp_t* parser, int enabled);
  * but might interact badly with outdated and insecure clients. With this flag
  * the extra request/response will be parsed normally.
  *
- * **(USE AT YOUR OWN RISK)**
+ * **Enabling this flag can pose a security issue since you will be exposed to
+ * poisoning attacks. USE WITH CAUTION!**
  */
+LLHTTP_EXPORT
 void llhttp_set_lenient_keep_alive(llhttp_t* parser, int enabled);
 
+/* Enables/disables lenient handling of `Transfer-Encoding` header.
+ *
+ * Normally `llhttp` would error when a `Transfer-Encoding` has `chunked` value
+ * and another value after it (either in a single header or in multiple
+ * headers whose value are internally joined using `, `).
+ * This is mandated by the spec to reliably determine request body size and thus
+ * avoid request smuggling.
+ * With this flag the extra value will be parsed normally.
+ *
+ * **Enabling this flag can pose a security issue since you will be exposed to
+ * request smuggling attacks. USE WITH CAUTION!**
+ */
+LLHTTP_EXPORT
+void llhttp_set_lenient_transfer_encoding(llhttp_t* parser, int enabled);
+
+/* Enables/disables lenient handling of HTTP version.
+ *
+ * Normally `llhttp` would error when the HTTP version in the request or status line
+ * is not `0.9`, `1.0`, `1.1` or `2.0`.
+ * With this flag the invalid value will be parsed normally.
+ *
+ * **Enabling this flag can pose a security issue since you will allow unsupported
+ * HTTP versions. USE WITH CAUTION!**
+ */
+LLHTTP_EXPORT
+void llhttp_set_lenient_version(llhttp_t* parser, int enabled);
+
+/* Enables/disables lenient handling of additional data received after a message ends
+ * and keep-alive is disabled.
+ *
+ * Normally `llhttp` would error when additional unexpected data is received if the message
+ * contains the `Connection` header with `close` value.
+ * With this flag the extra data will discarded without throwing an error.
+ *
+ * **Enabling this flag can pose a security issue since you will be exposed to
+ * poisoning attacks. USE WITH CAUTION!**
+ */
+LLHTTP_EXPORT
+void llhttp_set_lenient_data_after_close(llhttp_t* parser, int enabled);
+
+/* Enables/disables lenient handling of incomplete CRLF sequences.
+ *
+ * Normally `llhttp` would error when a CR is not followed by LF when terminating the
+ * request line, the status line, the headers or a chunk header.
+ * With this flag only a CR is required to terminate such sections.
+ *
+ * **Enabling this flag can pose a security issue since you will be exposed to
+ * request smuggling attacks. USE WITH CAUTION!**
+ */
+LLHTTP_EXPORT
+void llhttp_set_lenient_optional_lf_after_cr(llhttp_t* parser, int enabled);
+
+/*
+ * Enables/disables lenient handling of line separators.
+ *
+ * Normally `llhttp` would error when a LF is not preceded by CR when terminating the
+ * request line, the status line, the headers, a chunk header or a chunk data.
+ * With this flag only a LF is required to terminate such sections.
+ *
+ * **Enabling this flag can pose a security issue since you will be exposed to
+ * request smuggling attacks. USE WITH CAUTION!**
+ */
+LLHTTP_EXPORT
+void llhttp_set_lenient_optional_cr_before_lf(llhttp_t* parser, int enabled);
+
+/* Enables/disables lenient handling of chunks not separated via CRLF.
+ *
+ * Normally `llhttp` would error when after a chunk data a CRLF is missing before
+ * starting a new chunk.
+ * With this flag the new chunk can start immediately after the previous one.
+ *
+ * **Enabling this flag can pose a security issue since you will be exposed to
+ * request smuggling attacks. USE WITH CAUTION!**
+ */
+LLHTTP_EXPORT
+void llhttp_set_lenient_optional_crlf_after_chunk(llhttp_t* parser, int enabled);
+
+/* Enables/disables lenient handling of spaces after chunk size.
+ *
+ * Normally `llhttp` would error when after a chunk size is followed by one or more
+ * spaces are present instead of a CRLF or `;`.
+ * With this flag this check is disabled.
+ *
+ * **Enabling this flag can pose a security issue since you will be exposed to
+ * request smuggling attacks. USE WITH CAUTION!**
+ */
+LLHTTP_EXPORT
+void llhttp_set_lenient_spaces_after_chunk_size(llhttp_t* parser, int enabled);
+
 #ifdef __cplusplus
 }  /* extern "C" */
 #endif
 #endif  /* INCLUDE_LLHTTP_API_H_ */
 
+
 #endif  /* INCLUDE_LLHTTP_H_ */

src/3rdparty/rapidjson/allocators.h

@@ -19,6 +19,7 @@
 #include "internal/meta.h"
 
 #include <memory>
+#include <limits>
 
 #if RAPIDJSON_HAS_CXX11
 #include <type_traits>
@@ -433,7 +434,7 @@ namespace internal {
 template<typename T, typename A>
 inline T* Realloc(A& a, T* old_p, size_t old_n, size_t new_n)
 {
-    RAPIDJSON_NOEXCEPT_ASSERT(old_n <= SIZE_MAX / sizeof(T) && new_n <= SIZE_MAX / sizeof(T));
+    RAPIDJSON_NOEXCEPT_ASSERT(old_n <= (std::numeric_limits<size_t>::max)() / sizeof(T) && new_n <= (std::numeric_limits<size_t>::max)() / sizeof(T));
     return static_cast<T*>(a.Realloc(old_p, old_n * sizeof(T), new_n * sizeof(T)));
 }
 
@@ -496,9 +497,9 @@ public:
 #endif
 
     /* implicit */
-    StdAllocator(const BaseAllocator& allocator) RAPIDJSON_NOEXCEPT :
+    StdAllocator(const BaseAllocator& baseAllocator) RAPIDJSON_NOEXCEPT :
         allocator_type(),
-        baseAllocator_(allocator)
+        baseAllocator_(baseAllocator)
     { }
 
     ~StdAllocator() RAPIDJSON_NOEXCEPT

src/3rdparty/rapidjson/document.h

@@ -75,7 +75,7 @@ class GenericDocument;
     User can define this to use CrtAllocator or MemoryPoolAllocator.
 */
 #ifndef RAPIDJSON_DEFAULT_ALLOCATOR
-#define RAPIDJSON_DEFAULT_ALLOCATOR MemoryPoolAllocator<CrtAllocator>
+#define RAPIDJSON_DEFAULT_ALLOCATOR ::RAPIDJSON_NAMESPACE::MemoryPoolAllocator<::RAPIDJSON_NAMESPACE::CrtAllocator>
 #endif
 
 /*! \def RAPIDJSON_DEFAULT_STACK_ALLOCATOR
@@ -85,7 +85,7 @@ class GenericDocument;
     User can define this to use CrtAllocator or MemoryPoolAllocator.
 */
 #ifndef RAPIDJSON_DEFAULT_STACK_ALLOCATOR
-#define RAPIDJSON_DEFAULT_STACK_ALLOCATOR CrtAllocator
+#define RAPIDJSON_DEFAULT_STACK_ALLOCATOR ::RAPIDJSON_NAMESPACE::CrtAllocator
 #endif
 
 /*! \def RAPIDJSON_VALUE_DEFAULT_OBJECT_CAPACITY
@@ -1033,7 +1033,7 @@ public:
                 return false;
             for (ConstMemberIterator lhsMemberItr = MemberBegin(); lhsMemberItr != MemberEnd(); ++lhsMemberItr) {
                 typename RhsType::ConstMemberIterator rhsMemberItr = rhs.FindMember(lhsMemberItr->name);
-                if (rhsMemberItr == rhs.MemberEnd() || lhsMemberItr->value != rhsMemberItr->value)
+                if (rhsMemberItr == rhs.MemberEnd() || (!(lhsMemberItr->value == rhsMemberItr->value)))
                     return false;
             }
             return true;
@@ -1042,7 +1042,7 @@ public:
             if (data_.a.size != rhs.data_.a.size)
                 return false;
             for (SizeType i = 0; i < data_.a.size; i++)
-                if ((*this)[i] != rhs[i])
+                if (!((*this)[i] == rhs[i]))
                     return false;
             return true;
 
@@ -1078,6 +1078,7 @@ public:
     */
     template <typename T> RAPIDJSON_DISABLEIF_RETURN((internal::OrExpr<internal::IsPointer<T>,internal::IsGenericValue<T> >), (bool)) operator==(const T& rhs) const { return *this == GenericValue(rhs); }
 
+#ifndef __cpp_impl_three_way_comparison
     //! Not-equal-to operator
     /*! \return !(*this == rhs)
      */
@@ -1092,7 +1093,6 @@ public:
      */
     template <typename T> RAPIDJSON_DISABLEIF_RETURN((internal::IsGenericValue<T>), (bool)) operator!=(const T& rhs) const { return !(*this == rhs); }
 
-#ifndef __cpp_lib_three_way_comparison
     //! Equal-to operator with arbitrary types (symmetric version)
     /*! \return (rhs == lhs)
      */
@@ -1230,13 +1230,28 @@ public:
         else {
             RAPIDJSON_ASSERT(false);    // see above note
 
-            // This will generate -Wexit-time-destructors in clang
+#if RAPIDJSON_HAS_CXX11
-            // static GenericValue NullValue;
+            // Use thread-local storage to prevent races between threads.
-            // return NullValue;
+            // Use static buffer and placement-new to prevent destruction, with
-
+            // alignas() to ensure proper alignment.
-            // Use static buffer and placement-new to prevent destruction
+            alignas(GenericValue) thread_local static char buffer[sizeof(GenericValue)];
-            static char buffer[sizeof(GenericValue)];
             return *new (buffer) GenericValue();
+#elif defined(_MSC_VER) && _MSC_VER < 1900
+            // There's no way to solve both thread locality and proper alignment
+            // simultaneously.
+            __declspec(thread) static char buffer[sizeof(GenericValue)];
+            return *new (buffer) GenericValue();
+#elif defined(__GNUC__) || defined(__clang__)
+            // This will generate -Wexit-time-destructors in clang, but that's
+            // better than having under-alignment.
+            __thread static GenericValue buffer;
+            return buffer;
+#else
+            // Don't know what compiler this is, so don't know how to ensure
+            // thread-locality.
+            static GenericValue buffer;
+            return buffer;
+#endif
         }
     }
     template <typename SourceAllocator>
@@ -2430,13 +2445,14 @@ private:
             data_.f.flags = kShortStringFlag;
             data_.ss.SetLength(s.length);
             str = data_.ss.str;
+            std::memmove(str, s, s.length * sizeof(Ch));
         } else {
             data_.f.flags = kCopyStringFlag;
             data_.s.length = s.length;
             str = static_cast<Ch *>(allocator.Malloc((s.length + 1) * sizeof(Ch)));
             SetStringPointer(str);
-        }
             std::memcpy(str, s, s.length * sizeof(Ch));
+        }
         str[s.length] = '\0';
     }
 
@@ -2486,6 +2502,7 @@ public:
     typedef typename Encoding::Ch Ch;                       //!< Character type derived from Encoding.
     typedef GenericValue<Encoding, Allocator> ValueType;    //!< Value type of the document.
     typedef Allocator AllocatorType;                        //!< Allocator type from template parameter.
+    typedef StackAllocator StackAllocatorType;              //!< StackAllocator type from template parameter.
 
     //! Constructor
     /*! Creates an empty document of specified type.

src/3rdparty/rapidjson/encodings.h

@@ -177,10 +177,10 @@ struct UTF8 {
 
     template <typename InputStream, typename OutputStream>
     static bool Validate(InputStream& is, OutputStream& os) {
-#define RAPIDJSON_COPY() os.Put(c = is.Take())
+#define RAPIDJSON_COPY() if (c != '\0') os.Put(c = is.Take())
 #define RAPIDJSON_TRANS(mask) result &= ((GetRange(static_cast<unsigned char>(c)) & mask) != 0)
 #define RAPIDJSON_TAIL() RAPIDJSON_COPY(); RAPIDJSON_TRANS(0x70)
-        Ch c;
+        Ch c = static_cast<Ch>(-1);
         RAPIDJSON_COPY();
         if (!(c & 0x80))
             return true;

src/3rdparty/rapidjson/error/en.h

@@ -104,11 +104,65 @@ inline const RAPIDJSON_ERROR_CHARTYPE* GetValidateError_En(ValidateErrorCode val
         case kValidateErrorType:                        return RAPIDJSON_ERROR_STRING("Property has a type '%actual' that is not in the following list: '%expected'.");
 
         case kValidateErrorOneOf:                       return RAPIDJSON_ERROR_STRING("Property did not match any of the sub-schemas specified by 'oneOf', refer to following errors.");
-        case kValidateErrorOneOfMatch:                  return RAPIDJSON_ERROR_STRING("Property matched more than one of the sub-schemas specified by 'oneOf'.");
+        case kValidateErrorOneOfMatch:                  return RAPIDJSON_ERROR_STRING("Property matched more than one of the sub-schemas specified by 'oneOf', indices '%matches'.");
         case kValidateErrorAllOf:                       return RAPIDJSON_ERROR_STRING("Property did not match all of the sub-schemas specified by 'allOf', refer to following errors.");
         case kValidateErrorAnyOf:                       return RAPIDJSON_ERROR_STRING("Property did not match any of the sub-schemas specified by 'anyOf', refer to following errors.");
         case kValidateErrorNot:                         return RAPIDJSON_ERROR_STRING("Property matched the sub-schema specified by 'not'.");
 
+        case kValidateErrorReadOnly:                    return RAPIDJSON_ERROR_STRING("Property is read-only but has been provided when validation is for writing.");
+        case kValidateErrorWriteOnly:                   return RAPIDJSON_ERROR_STRING("Property is write-only but has been provided when validation is for reading.");
+
+        default:                                        return RAPIDJSON_ERROR_STRING("Unknown error.");
+    }
+}
+
+//! Maps error code of schema document compilation into error message.
+/*!
+    \ingroup RAPIDJSON_ERRORS
+    \param schemaErrorCode Error code obtained from compiling the schema document.
+    \return the error message.
+    \note User can make a copy of this function for localization.
+        Using switch-case is safer for future modification of error codes.
+*/
+  inline const RAPIDJSON_ERROR_CHARTYPE* GetSchemaError_En(SchemaErrorCode schemaErrorCode) {
+      switch (schemaErrorCode) {
+          case kSchemaErrorNone:                        return RAPIDJSON_ERROR_STRING("No error.");
+
+          case kSchemaErrorStartUnknown:                return RAPIDJSON_ERROR_STRING("Pointer '%value' to start of schema does not resolve to a location in the document.");
+          case kSchemaErrorRefPlainName:                return RAPIDJSON_ERROR_STRING("$ref fragment '%value' must be a JSON pointer.");
+          case kSchemaErrorRefInvalid:                  return RAPIDJSON_ERROR_STRING("$ref must not be an empty string.");
+          case kSchemaErrorRefPointerInvalid:           return RAPIDJSON_ERROR_STRING("$ref fragment '%value' is not a valid JSON pointer at offset '%offset'.");
+          case kSchemaErrorRefUnknown:                  return RAPIDJSON_ERROR_STRING("$ref '%value' does not resolve to a location in the target document.");
+          case kSchemaErrorRefCyclical:                 return RAPIDJSON_ERROR_STRING("$ref '%value' is cyclical.");
+          case kSchemaErrorRefNoRemoteProvider:         return RAPIDJSON_ERROR_STRING("$ref is remote but there is no remote provider.");
+          case kSchemaErrorRefNoRemoteSchema:           return RAPIDJSON_ERROR_STRING("$ref '%value' is remote but the remote provider did not return a schema.");
+          case kSchemaErrorRegexInvalid:                return RAPIDJSON_ERROR_STRING("Invalid regular expression '%value' in 'pattern' or 'patternProperties'.");
+          case kSchemaErrorSpecUnknown:                 return RAPIDJSON_ERROR_STRING("JSON schema draft or OpenAPI version is not recognized.");
+          case kSchemaErrorSpecUnsupported:             return RAPIDJSON_ERROR_STRING("JSON schema draft or OpenAPI version is not supported.");
+          case kSchemaErrorSpecIllegal:                 return RAPIDJSON_ERROR_STRING("Both JSON schema draft and OpenAPI version found in document.");
+          case kSchemaErrorReadOnlyAndWriteOnly:        return RAPIDJSON_ERROR_STRING("Property must not be both 'readOnly' and 'writeOnly'.");
+
+          default:                                      return RAPIDJSON_ERROR_STRING("Unknown error.");
+    }
+  }
+
+//! Maps error code of pointer parse into error message.
+/*!
+    \ingroup RAPIDJSON_ERRORS
+    \param pointerParseErrorCode Error code obtained from pointer parse.
+    \return the error message.
+    \note User can make a copy of this function for localization.
+        Using switch-case is safer for future modification of error codes.
+*/
+inline const RAPIDJSON_ERROR_CHARTYPE* GetPointerParseError_En(PointerParseErrorCode pointerParseErrorCode) {
+    switch (pointerParseErrorCode) {
+        case kPointerParseErrorNone:                       return RAPIDJSON_ERROR_STRING("No error.");
+
+        case kPointerParseErrorTokenMustBeginWithSolidus:  return RAPIDJSON_ERROR_STRING("A token must begin with a '/'.");
+        case kPointerParseErrorInvalidEscape:              return RAPIDJSON_ERROR_STRING("Invalid escape.");
+        case kPointerParseErrorInvalidPercentEncoding:     return RAPIDJSON_ERROR_STRING("Invalid percent encoding in URI fragment.");
+        case kPointerParseErrorCharacterMustPercentEncode: return RAPIDJSON_ERROR_STRING("A character must be percent encoded in a URI fragment.");
+
         default:                                           return RAPIDJSON_ERROR_STRING("Unknown error.");
     }
 }

src/3rdparty/rapidjson/error/error.h

@@ -42,7 +42,7 @@ RAPIDJSON_DIAG_OFF(padded)
 ///////////////////////////////////////////////////////////////////////////////
 // RAPIDJSON_ERROR_STRING
 
-//! Macro for converting string literial to \ref RAPIDJSON_ERROR_CHARTYPE[].
+//! Macro for converting string literal to \ref RAPIDJSON_ERROR_CHARTYPE[].
 /*! \ingroup RAPIDJSON_ERRORS
     By default this conversion macro does nothing.
     On Windows, user can define this macro as \c _T(x) for supporting both
@@ -185,14 +185,17 @@ enum ValidateErrorCode {
     kValidateErrorPatternProperties,           //!< See other errors.
     kValidateErrorDependencies,                //!< Object has missing property or schema dependencies.
 
-    kValidateErrorEnum,                        //!< Property has a value that is not one of its allowed enumerated values
+    kValidateErrorEnum,                        //!< Property has a value that is not one of its allowed enumerated values.
-    kValidateErrorType,                        //!< Property has a type that is not allowed by the schema..
+    kValidateErrorType,                        //!< Property has a type that is not allowed by the schema.
 
     kValidateErrorOneOf,                       //!< Property did not match any of the sub-schemas specified by 'oneOf'.
     kValidateErrorOneOfMatch,                  //!< Property matched more than one of the sub-schemas specified by 'oneOf'.
     kValidateErrorAllOf,                       //!< Property did not match all of the sub-schemas specified by 'allOf'.
     kValidateErrorAnyOf,                       //!< Property did not match any of the sub-schemas specified by 'anyOf'.
-    kValidateErrorNot                          //!< Property matched the sub-schema specified by 'not'.
+    kValidateErrorNot,                         //!< Property matched the sub-schema specified by 'not'.
+
+    kValidateErrorReadOnly,                    //!< Property is read-only but has been provided when validation is for writing
+    kValidateErrorWriteOnly                    //!< Property is write-only but has been provided when validation is for reading
 };
 
 //! Function pointer type of GetValidateError().
@@ -207,6 +210,72 @@ enum ValidateErrorCode {
 */
 typedef const RAPIDJSON_ERROR_CHARTYPE* (*GetValidateErrorFunc)(ValidateErrorCode);
 
+///////////////////////////////////////////////////////////////////////////////
+// SchemaErrorCode
+
+//! Error codes when validating.
+/*! \ingroup RAPIDJSON_ERRORS
+    \see GenericSchemaValidator
+*/
+enum SchemaErrorCode {
+    kSchemaErrorNone = 0,                      //!< No error.
+
+    kSchemaErrorStartUnknown,                  //!< Pointer to start of schema does not resolve to a location in the document
+    kSchemaErrorRefPlainName,                  //!< $ref fragment must be a JSON pointer
+    kSchemaErrorRefInvalid,                    //!< $ref must not be an empty string
+    kSchemaErrorRefPointerInvalid,             //!< $ref fragment is not a valid JSON pointer at offset
+    kSchemaErrorRefUnknown,                    //!< $ref does not resolve to a location in the target document
+    kSchemaErrorRefCyclical,                   //!< $ref is cyclical
+    kSchemaErrorRefNoRemoteProvider,           //!< $ref is remote but there is no remote provider
+    kSchemaErrorRefNoRemoteSchema,             //!< $ref is remote but the remote provider did not return a schema
+    kSchemaErrorRegexInvalid,                  //!< Invalid regular expression in 'pattern' or 'patternProperties'
+    kSchemaErrorSpecUnknown,                   //!< JSON schema draft or OpenAPI version is not recognized
+    kSchemaErrorSpecUnsupported,               //!< JSON schema draft or OpenAPI version is not supported
+    kSchemaErrorSpecIllegal,                   //!< Both JSON schema draft and OpenAPI version found in document
+    kSchemaErrorReadOnlyAndWriteOnly           //!< Property must not be both 'readOnly' and 'writeOnly'
+};
+
+//! Function pointer type of GetSchemaError().
+/*! \ingroup RAPIDJSON_ERRORS
+
+    This is the prototype for \c GetSchemaError_X(), where \c X is a locale.
+    User can dynamically change locale in runtime, e.g.:
+\code
+    GetSchemaErrorFunc GetSchemaError = GetSchemaError_En; // or whatever
+    const RAPIDJSON_ERROR_CHARTYPE* s = GetSchemaError(validator.GetInvalidSchemaCode());
+\endcode
+*/
+typedef const RAPIDJSON_ERROR_CHARTYPE* (*GetSchemaErrorFunc)(SchemaErrorCode);
+
+///////////////////////////////////////////////////////////////////////////////
+// PointerParseErrorCode
+
+//! Error code of JSON pointer parsing.
+/*! \ingroup RAPIDJSON_ERRORS
+    \see GenericPointer::GenericPointer, GenericPointer::GetParseErrorCode
+*/
+enum PointerParseErrorCode {
+    kPointerParseErrorNone = 0,                     //!< The parse is successful
+
+    kPointerParseErrorTokenMustBeginWithSolidus,    //!< A token must begin with a '/'
+    kPointerParseErrorInvalidEscape,                //!< Invalid escape
+    kPointerParseErrorInvalidPercentEncoding,       //!< Invalid percent encoding in URI fragment
+    kPointerParseErrorCharacterMustPercentEncode    //!< A character must percent encoded in URI fragment
+};
+
+//! Function pointer type of GetPointerParseError().
+/*! \ingroup RAPIDJSON_ERRORS
+
+    This is the prototype for \c GetPointerParseError_X(), where \c X is a locale.
+    User can dynamically change locale in runtime, e.g.:
+\code
+    GetPointerParseErrorFunc GetPointerParseError = GetPointerParseError_En; // or whatever
+    const RAPIDJSON_ERROR_CHARTYPE* s = GetPointerParseError(pointer.GetParseErrorCode());
+\endcode
+*/
+typedef const RAPIDJSON_ERROR_CHARTYPE* (*GetPointerParseErrorFunc)(PointerParseErrorCode);
+
+
 RAPIDJSON_NAMESPACE_END
 
 #ifdef __clang__

src/3rdparty/rapidjson/internal/biginteger.h

@@ -19,7 +19,11 @@
 
 #if defined(_MSC_VER) && !defined(__INTEL_COMPILER) && defined(_M_AMD64)
 #include <intrin.h> // for _umul128
+#if !defined(_ARM64EC_)
 #pragma intrinsic(_umul128)
+#else
+#pragma comment(lib,"softintrin")
+#endif
 #endif
 
 RAPIDJSON_NAMESPACE_BEGIN
@@ -255,7 +259,7 @@ private:
         if (low < k)
             (*outHigh)++;
         return low;
-#elif (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) && defined(__x86_64__)
+#elif defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) && defined(__x86_64__)
         __extension__ typedef unsigned __int128 uint128;
         uint128 p = static_cast<uint128>(a) * static_cast<uint128>(b);
         p += k;

src/3rdparty/rapidjson/internal/diyfp.h

@@ -25,7 +25,11 @@
 
 #if defined(_MSC_VER) && defined(_M_AMD64) && !defined(__INTEL_COMPILER)
 #include <intrin.h>
+#if !defined(_ARM64EC_)
 #pragma intrinsic(_umul128)
+#else
+#pragma comment(lib,"softintrin")
+#endif
 #endif
 
 RAPIDJSON_NAMESPACE_BEGIN
@@ -75,7 +79,7 @@ struct DiyFp {
         if (l & (uint64_t(1) << 63)) // rounding
             h++;
         return DiyFp(h, e + rhs.e + 64);
-#elif (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) && defined(__x86_64__)
+#elif defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) && defined(__x86_64__)
         __extension__ typedef unsigned __int128 uint128;
         uint128 p = static_cast<uint128>(f) * static_cast<uint128>(rhs.f);
         uint64_t h = static_cast<uint64_t>(p >> 64);

src/3rdparty/rapidjson/internal/dtoa.h

@@ -58,11 +58,11 @@ inline int CountDecimalDigit32(uint32_t n) {
 }
 
 inline void DigitGen(const DiyFp& W, const DiyFp& Mp, uint64_t delta, char* buffer, int* len, int* K) {
-    static const uint64_t kPow10[] = { 1U, 10U, 100U, 1000U, 10000U, 100000U, 1000000U, 10000000U, 100000000U,
+    static const uint64_t kPow10[] = { 1ULL, 10ULL, 100ULL, 1000ULL, 10000ULL, 100000ULL, 1000000ULL, 10000000ULL, 100000000ULL,
-                                       1000000000U, 10000000000U, 100000000000U, 1000000000000U,
+                                       1000000000ULL, 10000000000ULL, 100000000000ULL, 1000000000000ULL,
-                                       10000000000000U, 100000000000000U, 1000000000000000U,
+                                       10000000000000ULL, 100000000000000ULL, 1000000000000000ULL,
-                                       10000000000000000U, 100000000000000000U, 1000000000000000000U,
+                                       10000000000000000ULL, 100000000000000000ULL, 1000000000000000000ULL,
-                                       10000000000000000000U };
+                                       10000000000000000000ULL };
     const DiyFp one(uint64_t(1) << -Mp.e, Mp.e);
     const DiyFp wp_w = Mp - W;
     uint32_t p1 = static_cast<uint32_t>(Mp.f >> -one.e);

src/3rdparty/rapidjson/internal/regex.h

@@ -615,7 +615,7 @@ public:
         RAPIDJSON_ASSERT(regex_.IsValid());
         if (!allocator_)
             ownAllocator_ = allocator_ = RAPIDJSON_NEW(Allocator)();
-        stateSet_ = static_cast<unsigned*>(allocator_->Malloc(GetStateSetSize()));
+        stateSet_ = static_cast<uint32_t*>(allocator_->Malloc(GetStateSetSize()));
         state0_.template Reserve<SizeType>(regex_.stateCount_);
         state1_.template Reserve<SizeType>(regex_.stateCount_);
     }

src/3rdparty/rapidjson/internal/strtod.h

@@ -134,7 +134,7 @@ inline bool StrtodDiyFp(const Ch* decimals, int dLen, int dExp, double* result)
     int i = 0;   // 2^64 - 1 = 18446744073709551615, 1844674407370955161 = 0x1999999999999999
     for (; i < dLen; i++) {
         if (significand  >  RAPIDJSON_UINT64_C2(0x19999999, 0x99999999) ||
-            (significand == RAPIDJSON_UINT64_C2(0x19999999, 0x99999999) && decimals[i] > Ch('5')))
+            (significand == RAPIDJSON_UINT64_C2(0x19999999, 0x99999999) && decimals[i] >= Ch('5')))
             break;
         significand = significand * 10u + static_cast<unsigned>(decimals[i] - Ch('0'));
     }

src/3rdparty/rapidjson/pointer.h

@@ -18,6 +18,7 @@
 #include "document.h"
 #include "uri.h"
 #include "internal/itoa.h"
+#include "error/error.h" // PointerParseErrorCode
 
 #ifdef __clang__
 RAPIDJSON_DIAG_PUSH
@@ -27,23 +28,16 @@ RAPIDJSON_DIAG_PUSH
 RAPIDJSON_DIAG_OFF(4512) // assignment operator could not be generated
 #endif
 
+#if defined(RAPIDJSON_CPLUSPLUS) && RAPIDJSON_CPLUSPLUS >= 201703L
+#define RAPIDJSON_IF_CONSTEXPR if constexpr
+#else
+#define RAPIDJSON_IF_CONSTEXPR if
+#endif
+
 RAPIDJSON_NAMESPACE_BEGIN
 
 static const SizeType kPointerInvalidIndex = ~SizeType(0);  //!< Represents an invalid index in GenericPointer::Token
 
-//! Error code of parsing.
-/*! \ingroup RAPIDJSON_ERRORS
-    \see GenericPointer::GenericPointer, GenericPointer::GetParseErrorCode
-*/
-enum PointerParseErrorCode {
-    kPointerParseErrorNone = 0,                     //!< The parse is successful
-
-    kPointerParseErrorTokenMustBeginWithSolidus,    //!< A token must begin with a '/'
-    kPointerParseErrorInvalidEscape,                //!< Invalid escape
-    kPointerParseErrorInvalidPercentEncoding,       //!< Invalid percent encoding in URI fragment
-    kPointerParseErrorCharacterMustPercentEncode    //!< A character must percent encoded in URI fragment
-};
-
 ///////////////////////////////////////////////////////////////////////////////
 // GenericPointer
 
@@ -303,7 +297,7 @@ public:
         SizeType length = static_cast<SizeType>(end - buffer);
         buffer[length] = '\0';
 
-        if (sizeof(Ch) == 1) {
+        RAPIDJSON_IF_CONSTEXPR (sizeof(Ch) == 1) {
             Token token = { reinterpret_cast<Ch*>(buffer), length, index };
             return Append(token, allocator);
         }
@@ -902,10 +896,16 @@ private:
             std::memcpy(nameBuffer_, rhs.nameBuffer_, nameBufferSize * sizeof(Ch));
         }
 
-        // Adjust pointers to name buffer
+        // The names of each token point to a string in the nameBuffer_. The
-        std::ptrdiff_t diff = nameBuffer_ - rhs.nameBuffer_;
+        // previous memcpy copied over string pointers into the rhs.nameBuffer_,
-        for (Token *t = tokens_; t != tokens_ + rhs.tokenCount_; ++t)
+        // but they should point to the strings in the new nameBuffer_.
-            t->name += diff;
+        for (size_t i = 0; i < rhs.tokenCount_; ++i) {
+          // The offset between the string address and the name buffer should
+          // still be constant, so we can just get this offset and set each new
+          // token name according the new buffer start + the known offset.
+          std::ptrdiff_t name_offset = rhs.tokens_[i].name - rhs.nameBuffer_;
+          tokens_[i].name = nameBuffer_ + name_offset;
+        }
 
         return nameBuffer_ + nameBufferSize;
     }

src/3rdparty/rapidjson/rapidjson.h

@@ -268,7 +268,7 @@
 #  elif defined(_BIG_ENDIAN) && !defined(_LITTLE_ENDIAN)
 #    define RAPIDJSON_ENDIAN RAPIDJSON_BIGENDIAN
 // Detect with architecture macros
-#  elif defined(__sparc) || defined(__sparc__) || defined(_POWER) || defined(__powerpc__) || defined(__ppc__) || defined(__hpux) || defined(__hppa) || defined(_MIPSEB) || defined(_POWER) || defined(__s390__)
+#  elif defined(__sparc) || defined(__sparc__) || defined(_POWER) || defined(__powerpc__) || defined(__ppc__) || defined(__ppc64__) || defined(__hpux) || defined(__hppa) || defined(_MIPSEB) || defined(_POWER) || defined(__s390__)
 #    define RAPIDJSON_ENDIAN RAPIDJSON_BIGENDIAN
 #  elif defined(__i386__) || defined(__alpha__) || defined(__ia64) || defined(__ia64__) || defined(_M_IX86) || defined(_M_IA64) || defined(_M_ALPHA) || defined(__amd64) || defined(__amd64__) || defined(_M_AMD64) || defined(__x86_64) || defined(__x86_64__) || defined(_M_X64) || defined(__bfin__)
 #    define RAPIDJSON_ENDIAN RAPIDJSON_LITTLEENDIAN

src/3rdparty/rapidjson/reader.h

@@ -1433,7 +1433,7 @@ private:
     class NumberStream<InputStream, StackCharacter, true, false> : public NumberStream<InputStream, StackCharacter, false, false> {
         typedef NumberStream<InputStream, StackCharacter, false, false> Base;
     public:
-        NumberStream(GenericReader& reader, InputStream& is) : Base(reader, is), stackStream(reader.stack_) {}
+        NumberStream(GenericReader& reader, InputStream& s) : Base(reader, s), stackStream(reader.stack_) {}
 
         RAPIDJSON_FORCEINLINE Ch TakePush() {
             stackStream.Put(static_cast<StackCharacter>(Base::is.Peek()));
@@ -1459,7 +1459,7 @@ private:
     class NumberStream<InputStream, StackCharacter, true, true> : public NumberStream<InputStream, StackCharacter, true, false> {
         typedef NumberStream<InputStream, StackCharacter, true, false> Base;
     public:
-        NumberStream(GenericReader& reader, InputStream& is) : Base(reader, is) {}
+        NumberStream(GenericReader& reader, InputStream& s) : Base(reader, s) {}
 
         RAPIDJSON_FORCEINLINE Ch Take() { return Base::TakePush(); }
     };
@@ -1584,7 +1584,7 @@ private:
         // Parse frac = decimal-point 1*DIGIT
         int expFrac = 0;
         size_t decimalPosition;
-        if (Consume(s, '.')) {
+        if (!useNanOrInf && Consume(s, '.')) {
             decimalPosition = s.Length();
 
             if (RAPIDJSON_UNLIKELY(!(s.Peek() >= '0' && s.Peek() <= '9')))
@@ -1631,7 +1631,7 @@ private:
 
         // Parse exp = e [ minus / plus ] 1*DIGIT
         int exp = 0;
-        if (Consume(s, 'e') || Consume(s, 'E')) {
+        if (!useNanOrInf && (Consume(s, 'e') || Consume(s, 'E'))) {
             if (!useDouble) {
                 d = static_cast<double>(use64bit ? i64 : i);
                 useDouble = true;
@@ -1694,7 +1694,7 @@ private:
             }
             else {
                 SizeType numCharsToCopy = static_cast<SizeType>(s.Length());
-                GenericStringStream<UTF8<NumberCharacter>> srcStream(s.Pop());
+                GenericStringStream<UTF8<NumberCharacter> > srcStream(s.Pop());
                 StackStream<typename TargetEncoding::Ch> dstStream(stack_);
                 while (numCharsToCopy--) {
                     Transcoder<UTF8<typename TargetEncoding::Ch>, TargetEncoding>::Transcode(srcStream, dstStream);

src/3rdparty/rapidjson/uri.h

@@ -238,20 +238,27 @@ private:
 
         // Allocate one block containing each part of the URI (5) plus base plus full URI, all null terminated.
         // Order: scheme, auth, path, query, frag, base, uri
+        // Note need to set, increment, assign in 3 stages to avoid compiler warning bug.
         size_t total = (3 * len + 7) * sizeof(Ch);
         scheme_ = static_cast<Ch*>(allocator_->Malloc(total));
         *scheme_ = '\0';
-        auth_ = scheme_ + 1;
+        auth_ = scheme_;
+        auth_++;
         *auth_ = '\0';
-        path_ = auth_ + 1;
+        path_ = auth_;
+        path_++;
         *path_ = '\0';
-        query_ = path_ + 1;
+        query_ = path_;
+        query_++;
         *query_ = '\0';
-        frag_ = query_ + 1;
+        frag_ = query_;
+        frag_++;
         *frag_ = '\0';
-        base_ = frag_ + 1;
+        base_ = frag_;
+        base_++;
         *base_ = '\0';
-        uri_ = base_ + 1;
+        uri_ = base_;
+        uri_++;
         *uri_ = '\0';
         return total;
     }
@@ -293,7 +300,9 @@ private:
             }
         }
         // Look for auth (//([^/?#]*))?
-        auth_ = scheme_ + GetSchemeStringLength() + 1;
+        // Note need to set, increment, assign in 3 stages to avoid compiler warning bug.
+        auth_ = scheme_ + GetSchemeStringLength();
+        auth_++;
         *auth_ = '\0';
         if (start < len - 1 && uri[start] == '/' && uri[start + 1] == '/') {
             pos2 = start + 2;
@@ -308,7 +317,9 @@ private:
             start = pos2;
         }
         // Look for path ([^?#]*)
-        path_ = auth_ + GetAuthStringLength() + 1;
+        // Note need to set, increment, assign in 3 stages to avoid compiler warning bug.
+        path_ = auth_ + GetAuthStringLength();
+        path_++;
         *path_ = '\0';
         if (start < len) {
             pos2 = start;
@@ -326,7 +337,9 @@ private:
             }
         }
         // Look for query (\?([^#]*))?
-        query_ = path_ + GetPathStringLength() + 1;
+        // Note need to set, increment, assign in 3 stages to avoid compiler warning bug.
+        query_ = path_ + GetPathStringLength();
+        query_++;
         *query_ = '\0';
         if (start < len && uri[start] == '?') {
             pos2 = start + 1;
@@ -341,7 +354,9 @@ private:
             }
         }
         // Look for fragment (#(.*))?
-        frag_ = query_ + GetQueryStringLength() + 1;
+        // Note need to set, increment, assign in 3 stages to avoid compiler warning bug.
+        frag_ = query_ + GetQueryStringLength();
+        frag_++;
         *frag_ = '\0';
         if (start < len && uri[start] == '#') {
             std::memcpy(frag_, &uri[start], (len - start) * sizeof(Ch));

src/3rdparty/rapidjson/writer.h

@@ -67,6 +67,7 @@ enum WriteFlag {
     kWriteNoFlags = 0,              //!< No flags are set.
     kWriteValidateEncodingFlag = 1, //!< Validate encoding of JSON strings.
     kWriteNanAndInfFlag = 2,        //!< Allow writing of Infinity, -Infinity and NaN.
+    kWriteNanAndInfNullFlag = 4,    //!< Allow writing of Infinity, -Infinity and NaN as null.
     kWriteDefaultFlags = RAPIDJSON_WRITE_DEFAULT_FLAGS  //!< Default write flags. Can be customized by defining RAPIDJSON_WRITE_DEFAULT_FLAGS
 };
 
@@ -349,8 +350,13 @@ protected:
 
     bool WriteDouble(double d) {
         if (internal::Double(d).IsNanOrInf()) {
-            if (!(writeFlags & kWriteNanAndInfFlag))
+            if (!(writeFlags & kWriteNanAndInfFlag) && !(writeFlags & kWriteNanAndInfNullFlag))
                 return false;
+            if (writeFlags & kWriteNanAndInfNullFlag) {
+                PutReserve(*os_, 4);
+                PutUnsafe(*os_, 'n'); PutUnsafe(*os_, 'u'); PutUnsafe(*os_, 'l'); PutUnsafe(*os_, 'l');
+                return true;
+            }
             if (internal::Double(d).IsNan()) {
                 PutReserve(*os_, 3);
                 PutUnsafe(*os_, 'N'); PutUnsafe(*os_, 'a'); PutUnsafe(*os_, 'N');
@@ -549,6 +555,11 @@ inline bool Writer<StringBuffer>::WriteDouble(double d) {
         // Note: This code path can only be reached if (RAPIDJSON_WRITE_DEFAULT_FLAGS & kWriteNanAndInfFlag).
         if (!(kWriteDefaultFlags & kWriteNanAndInfFlag))
             return false;
+        if (kWriteDefaultFlags & kWriteNanAndInfNullFlag) {
+            PutReserve(*os_, 4);
+            PutUnsafe(*os_, 'n'); PutUnsafe(*os_, 'u'); PutUnsafe(*os_, 'l'); PutUnsafe(*os_, 'l');
+            return true;
+        }
         if (internal::Double(d).IsNan()) {
             PutReserve(*os_, 3);
             PutUnsafe(*os_, 'N'); PutUnsafe(*os_, 'a'); PutUnsafe(*os_, 'N');

src/App.cpp

@@ -6,8 +6,8 @@
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
  * Copyright 2018      Lee Clagett <https://github.com/vtnerd>
- * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2024 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2024 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -23,7 +23,6 @@
  *   along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-
 #include <cstdlib>
 #include <uv.h>
 
@@ -61,13 +60,13 @@ int xmrig::App::exec()
         return 2;
     }
 
-    m_signals = std::make_shared<Signals>(this);
-
     int rc = 0;
     if (background(rc)) {
         return rc;
     }
 
+    m_signals = std::make_shared<Signals>(this);
+
     rc = m_controller->init();
     if (rc != 0) {
         return rc;

src/App_unix.cpp

@@ -5,8 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2024 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2024 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -22,7 +22,6 @@
  *   along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-
 #include <cstdlib>
 #include <csignal>
 #include <cerrno>
@@ -53,16 +52,9 @@ bool xmrig::App::background(int &rc)
         return true;
     }
 
-    i = setsid();
+    if (setsid() < 0) {
-
-    if (i < 0) {
         LOG_ERR("setsid() failed (errno = %d)", errno);
     }
 
-    i = chdir("/");
-    if (i < 0) {
-        LOG_ERR("chdir() failed (errno = %d)", errno);
-    }
-
     return false;
 }

src/backend/common/Hashrate.cpp

@@ -30,10 +30,10 @@
 #include "base/tools/Handle.h"
 
 
-inline static const char *format(double h, char *buf, size_t size)
+inline static const char *format(std::pair<bool, double> h, char *buf, size_t size)
 {
-    if (std::isnormal(h)) {
+    if (h.first) {
-        snprintf(buf, size, (h < 100.0) ? "%04.2f" : "%03.1f", h);
+        snprintf(buf, size, (h.second < 100.0) ? "%04.2f" : "%03.1f", h.second);
         return buf;
     }
 
@@ -80,15 +80,16 @@ double xmrig::Hashrate::average() const
 }
 
 
-const char *xmrig::Hashrate::format(double h, char *buf, size_t size)
+const char *xmrig::Hashrate::format(std::pair<bool, double> h, char *buf, size_t size)
 {
     return ::format(h, buf, size);
 }
 
 
-rapidjson::Value xmrig::Hashrate::normalize(double d)
+rapidjson::Value xmrig::Hashrate::normalize(std::pair<bool, double> d)
 {
-    return Json::normalize(d, false);
+    using namespace rapidjson;
+    return d.first ? Value(floor(d.second * 100.0) / 100.0) : Value(kNullType);
 }
 
 
@@ -122,11 +123,11 @@ rapidjson::Value xmrig::Hashrate::toJSON(size_t threadId, rapidjson::Document &d
 #endif
 
 
-double xmrig::Hashrate::hashrate(size_t index, size_t ms) const
+std::pair<bool, double> xmrig::Hashrate::hashrate(size_t index, size_t ms) const
 {
     assert(index < m_threads);
     if (index >= m_threads) {
-        return nan("");
+        return { false, 0.0 };
     }
 
     uint64_t earliestHashCount = 0;
@@ -157,17 +158,27 @@ double xmrig::Hashrate::hashrate(size_t index, size_t ms) const
     } while (idx != idx_start);
 
     if (!haveFullSet || earliestStamp == 0 || lastestStamp == 0) {
-        return nan("");
+        return { false, 0.0 };
     }
 
-    if (lastestStamp - earliestStamp == 0) {
+    if (lastestHashCnt == earliestHashCount) {
-        return nan("");
+        return { true, 0.0 };
+    }
+
+    if (lastestStamp == earliestStamp) {
+        return { false, 0.0 };
     }
 
     const auto hashes = static_cast<double>(lastestHashCnt - earliestHashCount);
-    const auto time   = static_cast<double>(lastestStamp - earliestStamp) / 1000.0;
+    const auto time   = static_cast<double>(lastestStamp - earliestStamp);
 
-    return hashes / time;
+    const auto hr = hashes * 1000.0 / time;
+
+    if (!std::isnormal(hr)) {
+        return { false, 0.0 };
+    }
+
+    return { true, hr };
 }
 
 

src/backend/common/Hashrate.h

@@ -47,16 +47,16 @@ public:
     Hashrate(size_t threads);
     ~Hashrate();
 
-    inline double calc(size_t ms) const                                     { const double data = hashrate(0U, ms); return std::isnormal(data) ? data : 0.0; }
+    inline std::pair<bool, double> calc(size_t ms) const                    { return hashrate(0U, ms); }
-    inline double calc(size_t threadId, size_t ms) const                    { return hashrate(threadId + 1, ms); }
+    inline std::pair<bool, double> calc(size_t threadId, size_t ms) const   { return hashrate(threadId + 1, ms); }
     inline size_t threads() const                                           { return m_threads > 0U ? m_threads - 1U : 0U; }
     inline void add(size_t threadId, uint64_t count, uint64_t timestamp)    { addData(threadId + 1U, count, timestamp); }
     inline void add(uint64_t count, uint64_t timestamp)                     { addData(0U, count, timestamp); }
 
     double average() const;
 
-    static const char *format(double h, char *buf, size_t size);
+    static const char *format(std::pair<bool, double> h, char *buf, size_t size);
-    static rapidjson::Value normalize(double d);
+    static rapidjson::Value normalize(std::pair<bool, double> d);
 
 #   ifdef XMRIG_FEATURE_API
     rapidjson::Value toJSON(rapidjson::Document &doc) const;
@@ -64,7 +64,7 @@ public:
 #   endif
 
 private:
-    double hashrate(size_t index, size_t ms) const;
+    std::pair<bool, double> hashrate(size_t index, size_t ms) const;
     void addData(size_t index, uint64_t count, uint64_t timestamp);
 
     constexpr static size_t kBucketSize = 2 << 11;

src/backend/common/misc/PciTopology.h

@@ -1,6 +1,6 @@
 /* XMRig
- * Copyright (c) 2018-2021 SChernykh   <https://github.com/SChernykh>
+ * Copyright (c) 2018-2024 SChernykh   <https://github.com/SChernykh>
- * Copyright (c) 2016-2021 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright (c) 2016-2024 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -19,10 +19,8 @@
 #ifndef XMRIG_PCITOPOLOGY_H
 #define XMRIG_PCITOPOLOGY_H
 
-
 #include <cstdio>
 
-
 #include "base/tools/String.h"
 
 
@@ -33,7 +31,14 @@ class PciTopology
 {
 public:
     PciTopology() = default;
-    PciTopology(uint32_t bus, uint32_t device, uint32_t function) : m_valid(true), m_bus(bus), m_device(device), m_function(function) {}
+
+    template<typename T>
+    inline PciTopology(T bus, T device, T function)
+        : m_valid(true),
+          m_bus(static_cast<uint8_t>(bus)),
+          m_device(static_cast<uint8_t>(device)),
+          m_function(static_cast<uint8_t>(function))
+    {}
 
     inline bool isEqual(const PciTopology &other) const     { return m_valid == other.m_valid && toUint32() == other.toUint32(); }
     inline bool isValid() const                             { return m_valid; }
@@ -70,4 +75,4 @@ private:
 } // namespace xmrig
 
 
-#endif /* XMRIG_PCITOPOLOGY_H */
+#endif // XMRIG_PCITOPOLOGY_H

src/backend/cpu/CpuConfig_gen.h

@@ -1,6 +1,6 @@
 /* XMRig
- * Copyright (c) 2018-2021 SChernykh   <https://github.com/SChernykh>
+ * Copyright (c) 2018-2024 SChernykh   <https://github.com/SChernykh>
- * Copyright (c) 2016-2021 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright (c) 2016-2024 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
  *   it under the terms of the GNU General Public License as published by
@@ -122,17 +122,6 @@ size_t inline generate<Algorithm::RANDOM_X>(Threads<CpuThreads> &threads, uint32
         }
     }
 
-    if (!threads.isExist(Algorithm::RX_KEVA)) {
-        auto keva = cpuInfo->threads(Algorithm::RX_KEVA, limit);
-        if (keva == wow) {
-            threads.setAlias(Algorithm::RX_KEVA, Algorithm::kRX_WOW);
-            ++count;
-        }
-        else {
-            count += threads.move(Algorithm::kRX_KEVA, std::move(keva));
-        }
-    }
-
     if (!threads.isExist(Algorithm::RX_WOW)) {
         count += threads.move(Algorithm::kRX_WOW, std::move(wow));
     }