Develop stream 2025-01-20 (#515)

* Update copyright date python scripts * Fix warnings when generating sphinx docs * fix GenerateResourceSpec for gfx90a * fix relying on ROCPRIM_TARGET_ARCH * fix GenerateResourceSpec for gfx90a * Update documentation generation instructions * log more information during configure * CI changes to speedup development * fix usage of numeric_limits in tests * Fix rules overwrite for other then MR pipelines * Resolve "hook up additional rocThrust bitwise reproducibility tests into CI" * General changes CCCL 2.6 * Add missing copyright banners * Fix CMake warnings * Remove workaround for compiler issue for inclusive/exclusive scan with FixedVector items * Finalize changes for CCCL 2.6.0 * Resolve "Building rocRAND (used in benchmarks) takes significant time" * Remove extra dashes in the documentation its \param and \tparam * Documentation to ci upload * Adopt symbol visibility rules in rocThrust * Update copyright dates * Fix minor copyright year typo * Clang formatting on replacements of typedef * One more formatting issue in examples * Fix rebasing mistake in summary * cstdint.h is removed from thrust, replaced usage in rebased code * Add to CHANGELOG CCCL 2.6.0 changes --------- Co-authored-by: Sander Bos <[email protected]> Co-authored-by: Robin Voetter <[email protected]> Co-authored-by: Bence Parajdi <[email protected]> Co-authored-by: Nara Prasetya <[email protected]> Co-authored-by: Enrico Degregori <[email protected]> Co-authored-by: Beatriz Navidad Vilches <[email protected]> Co-authored-by: Anton Gorenko <[email protected]> Co-authored-by: Borys Petrov <[email protected]>
ROCm · Feb 28, 2025 · 0d5f02a · 0d5f02a
1 parent 0006584
commit 0d5f02a
Show file tree

Hide file tree

Showing 861 changed files with 11,517 additions and 12,822 deletions.
diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
@@ -1,5 +1,5 @@
 # ########################################################################
-# Copyright 2019-2024 Advanced Micro Devices, Inc.
+# Copyright 2019-2025 Advanced Micro Devices, Inc.
 # ########################################################################
 
 include:
@@ -23,6 +23,10 @@ stages:
   - test      # Tests if unit tests are passing (CTest)
   - benchmark # Runs the non-internal benchmarks (Google Benchmark)
 
+workflow:
+  rules:
+    - if: $CI_MERGE_REQUEST_LABELS !~ /CI Skip/
+
 variables:
   # Helper variables
   PACKAGE_DIR: $BUILD_DIR/package
@@ -74,7 +78,6 @@ copyright-date:
       -D CMAKE_CXX_COMPILER=hipcc
       -D CMAKE_BUILD_TYPE=Release
       -D BUILD_TEST=OFF
-      -D BUILD_HIPSTDPAR_TEST=OFF
       -D BUILD_EXAMPLE=OFF
       -D ROCM_DEP_ROCMCORE=OFF
       -D CMAKE_C_COMPILER_LAUNCHER=phc_sccache_c
@@ -116,7 +119,6 @@ copyright-date:
       -D CMAKE_BUILD_TYPE=$BUILD_TYPE
       -D BUILD_$BUILD_TARGET=ON
       -D GPU_TARGETS=$GPU_TARGETS
-      -D AMDGPU_TEST_TARGETS=$GPU_TARGETS
       -D RNG_SEED_COUNT=$rng_seed_count
       -D PRNG_SEEDS=$prng_seeds
       -D CMAKE_C_COMPILER_LAUNCHER=phc_sccache_c
@@ -137,7 +139,7 @@ copyright-date:
     - $CI_PROJECT_DIR/build/.ninja_log
     exclude:
     - $CI_PROJECT_DIR/build/**/*.o
-    expire_in: 2 weeks
+    expire_in: 1 day
 
 build:cmake-latest:
   stage: build
@@ -183,7 +185,7 @@ build:package:
     paths:
       - $PACKAGE_DIR/rocthrust*.deb
       - $PACKAGE_DIR/rocthrust*.zip
-    expire_in: 2 weeks
+    expire_in: 1 day
 
 build:windows:
   stage: build
@@ -231,7 +233,7 @@ build:windows:
     paths:
       - $CI_PROJECT_DIR/build/
       - $ROCPRIM_DIR/build/install
-    expire_in: 2 weeks
+    expire_in: 1 day
 
 test:package:
   stage: test
@@ -268,13 +270,18 @@ test:doc:
   extends:
     - .build:docs
     - .rules:test
+  artifacts:
+    paths:
+      - $DOCS_DIR/_build/html/
+    expire_in: 2 weeks
 
-test:
+.test:rocm:
   stage: test
+  tags:
+    - rocm
+    - $GPU
   extends:
     - .cmake-minimum
-    - .rules:test
-    - .gpus:rocm
   needs:
     - job: build:cmake-minimum
       parallel:
@@ -294,10 +301,76 @@ test:
     - HSA_ENABLE_SDMA=0 ctest
       --output-on-failure
       --repeat-until-fail 2
-      --tests-regex $GPU_TARGET
       --resource-spec-file ./resources.json
       --parallel $PARALLEL_JOBS
 
+test:rocm-any-gpu:
+  variables:
+    GPU: ""
+    PARALLEL_JOBS: 1
+  extends:
+    - .test:rocm
+  rules:
+    - if: $CI_MERGE_REQUEST_TITLE =~ /Draft:/ && $CI_MERGE_REQUEST_LABELS !~ /Arch::/
+
+test:rocm-label-arch:
+  extends:
+    - .gpus:rocm
+    - .test:rocm
+    - .rules:arch-labels
+
+test:rocm-all-gpus:
+  variables:
+    SHOULD_BE_UNDRAFTED: "true"
+  extends:
+    - .gpus:rocm
+    - .test:rocm
+    - .rules:test
+
+test-bitwise-repro-generate:
+  stage: test
+  extends:
+    - .cmake-minimum
+    - .rules:manual
+    - .gpus:rocm
+  needs:
+    - job: build:cmake-minimum
+      parallel:
+        matrix:
+          - BUILD_TYPE: Release
+            BUILD_TARGET: TEST
+            BUILD_VERSION: 14
+  cache:
+    key: database
+    paths:
+      - $CI_PROJECT_DIR/bitwise.db
+    policy: push
+  script:
+    - cd $CI_PROJECT_DIR/build/test/
+    - ROCTHRUST_BWR_PATH=$CI_PROJECT_DIR/bitwise.db ROCTHRUST_BWR_GENERATE=1 ./reproducibility.hip
+
+test-bitwise-repro:
+  stage: test
+  extends:
+    - .cmake-minimum
+    - .rules:test
+    - .gpus:rocm
+  needs:
+    - job: build:cmake-minimum
+      parallel:
+        matrix:
+          - BUILD_TYPE: Release
+            BUILD_TARGET: TEST
+            BUILD_VERSION: 14
+  cache:
+    key: database
+    paths:
+      - $CI_PROJECT_DIR/bitwise.db
+    policy: pull
+  script:
+    - cd $CI_PROJECT_DIR/build/test/
+    - ROCTHRUST_BWR_PATH=$CI_PROJECT_DIR/bitwise.db ./reproducibility.hip
+
 .rocm-windows:test:
   extends:
     - .gpus:rocm-windows
@@ -358,7 +431,7 @@ build:cuda-and-omp:
   tags:
     - build
   variables:
-    CCCL_GIT_BRANCH: v2.5.0
+    CCCL_GIT_BRANCH: v2.6.0
     CCCL_DIR: ${CI_PROJECT_DIR}/cccl
   needs: []
   script:
@@ -402,16 +475,17 @@ build:cuda-and-omp:
       - $CCCL_DIR/thrust/cmake/ThrustRunTest.cmake
       - $CCCL_DIR/thrust/cmake/ThrustRunExample.cmake
       - $CI_PROJECT_DIR/build/.ninja_log
-    expire_in: 1 week
+    expire_in: 1 day
 
-test:cuda-and-omp:
+.test:cuda-and-omp:
   stage: test
+  tags:
+    - nvcc
+    - $GPU
   needs:
     - build:cuda-and-omp
   extends:
     - .nvcc
-    - .gpus:nvcc
-    - .rules:test
   before_script:
     # This is only needed because of the legacy before_script in .gpus:nvcc would otherwise overwrite before_script
     - !reference [.nvcc, before_script]
@@ -420,6 +494,23 @@ test:cuda-and-omp:
     # These tests are executed on the build stage because they require sources
     - ctest --output-on-failure --exclude-regex "thrust.example.cmake.add_subdir|thrust.test.cmake.check_source_files"
 
+test:cuda-and-omp-any-gpu:
+  variables:
+    GPU: ""
+    PARALLEL_JOBS: 1
+  extends:
+    - .test:cuda-and-omp
+  rules:
+    - if: $CI_MERGE_REQUEST_TITLE =~ /Draft:/
+
+test:cuda-and-omp-all-gpus:
+  variables:
+    SHOULD_BE_UNDRAFTED: "true"
+  extends:
+    - .gpus:nvcc
+    - .test:cuda-and-omp
+    - .rules:test
+
 .benchmark-base:
   stage: benchmark
   extends:
@@ -430,7 +521,12 @@ test:cuda-and-omp:
 
 benchmark:
   needs:
-    - build:cmake-minimum
+    - job: build:cmake-minimum
+      parallel:
+        matrix:
+          - BUILD_TYPE: Release
+            BUILD_TARGET: BENCHMARKS
+            BUILD_VERSION: 14
   extends:
     - .cmake-minimum
     - .gpus:rocm
@@ -453,4 +549,4 @@ benchmark:
   artifacts:
     paths:
       - ${BENCHMARK_RESULT_DIR}
-    expire_in: 1 week
+    expire_in: 1 day
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -7,7 +7,8 @@ Documentation for rocThrust available at
 
 ### Added
 
-* gfx950 support
+* Added gfx950 support.
+* Merged changes from upstream CCCL/thrust 2.6.0
 
 ## rocThrust 3.3.0 for ROCm 6.4
 

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -1,5 +1,5 @@
 # ########################################################################
-# Copyright 2019-2024 Advanced Micro Devices, Inc.
+# Copyright 2019-2025 Advanced Micro Devices, Inc.
 # ########################################################################
 
 cmake_minimum_required(VERSION 3.10.2 FATAL_ERROR)
@@ -22,6 +22,12 @@ endif()
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 
+if (CMAKE_CXX_STANDARD EQUAL 14)
+  message(WARNING "C++14 will be deprecated in the next major release")
+elseif(NOT CMAKE_CXX_STANDARD EQUAL 17)
+  message(FATAL_ERROR "Only C++14 and C++17 are supported")
+endif()
+
 # Set HIP flags
 set(CMAKE_HIP_STANDARD 17)
 set(CMAKE_HIP_STANDARD_REQUIRED ON)
@@ -39,6 +45,7 @@ option(BUILD_HIPSTDPAR_TEST_WITH_TBB "Build hipstdpar tests with TBB" OFF)
 option(BUILD_EXAMPLES "Build examples" OFF)
 option(BUILD_BENCHMARKS "Build benchmarks" OFF)
 option(DOWNLOAD_ROCPRIM "Download rocPRIM and do not search for rocPRIM package" OFF)
+cmake_dependent_option(DOWNLOAD_ROCRAND "Download rocRAND and do not search for rocRAND package" OFF BUILD_BENCHMARKS OFF)
 option(DOWNLOAD_ROCRAND "Download rocRAND and do not search for rocRAND package" OFF)
 option(BUILD_ADDRESS_SANITIZER "Build with address sanitizer enabled" OFF)
 cmake_dependent_option(ENABLE_UPSTREAM_TESTS "Enable upstream (thrust) tests" ON BUILD_TEST OFF)
@@ -130,16 +137,9 @@ if (NOT THRUST_HOST_SYSTEM IN_LIST THRUST_HOST_SYSTEM_OPTIONS)
   )
 endif ()
 
-if(DISABLE_WERROR)
-  add_compile_options(-Wall -Wextra)
-else()
-  add_compile_options(-Wall -Wextra -Werror)
-endif()
-
-if (CMAKE_CXX_STANDARD EQUAL 14)
-  message(WARNING "C++14 will be deprecated in the next major release")
-elseif(NOT CMAKE_CXX_STANDARD EQUAL 17)
-  message(FATAL_ERROR "Only C++14 and C++17 are supported")
+set(COMPILE_OPTIONS -Wall -Wextra)
+if(NOT DISABLE_WERROR)
+  list(APPEND COMPILE_OPTIONS -Werror)
 endif()
 
 if (WIN32)
@@ -149,13 +149,13 @@ endif()
 
 # Address Sanitizer
 if(BUILD_ADDRESS_SANITIZER)
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsanitize=address -shared-libasan")
-  set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fsanitize=address -shared-libasan")
+  list(APPEND COMPILE_OPTIONS -fsanitize=address -shared-libasan)
   add_link_options(-fuse-ld=lld)
 endif()
 
 # Setup VERSION
 rocm_setup_version(VERSION "3.3.0")
+math(EXPR rocthrust_VERSION_NUMBER "${rocthrust_VERSION_MAJOR} * 100000 + ${rocthrust_VERSION_MINOR} * 100 + ${rocthrust_VERSION_PATCH}")
 
 # Print configuration summary
 include(cmake/Summary.cmake)
@@ -191,7 +191,6 @@ endif()
 # Benchmarks
 if(BUILD_BENCHMARKS)
   add_subdirectory(benchmarks)
-  add_subdirectory(internal/benchmark)
 endif()
 
 #Create header wrapper for backward compatibility