Merge branch 'release/rocm-rel-4.3' into xnack_on_hmm

.gitlab-ci.yml

@@ -14,7 +14,7 @@ stages:
 
 variables:
   # Tested CMake versions
-  CMAKE_MINIMUM_VERSION: 3.5.2
+  CMAKE_MINIMUM_VERSION: 3.10.2
   CMAKE_LATEST_VERSION: 3.17.0
   # Cluster config
 
@@ -23,11 +23,11 @@ variables:
   # Helper variables
   DEPS_DIR: $CI_PROJECT_DIR/deps
   BUILD_DIR: $CI_PROJECT_DIR/build
-  CMAKE_MINIMUM_URL: "https://cmake.org/files/v3.5/cmake-3.5.1-Linux-x86_64.tar.gz"
-  CMAKE_MINIMUM_PATH: "$CI_PROJECT_DIR/deps/cmake-3.5.2"
+  CMAKE_MINIMUM_URL: "https://cmake.org/files/v3.10/cmake-3.10.2-Linux-x86_64.tar.gz"
+  CMAKE_MINIMUM_PATH: "$CI_PROJECT_DIR/deps/cmake-3.10.2"
   CMAKE_LATEST_URL: "https://cmake.org/files/v3.16/cmake-3.16.0-Linux-x86_64.tar.gz"
   CMAKE_LATEST_PATH: "$CI_PROJECT_DIR/deps/cmake-3.17.0"
-  BUILD_MINIMUM_DIR: $CI_PROJECT_DIR/build/cmake-3.5.2
+  BUILD_MINIMUM_DIR: $CI_PROJECT_DIR/build/cmake-3.10.2
   BUILD_LATEST_DIR: $CI_PROJECT_DIR/build/cmake-3.17.0
   PACKAGE_DIR: $CI_PROJECT_DIR/build/package
 
@@ -189,8 +189,8 @@ deploy:deb:
       -G "DEB;ZIP"
       -B $PACKAGE_DIR
     # Install rocPRIM and rocThrust
-    - $SUDO_CMD dpkg -i $PACKAGE_DIR/rocprim-*.deb
-    - $SUDO_CMD dpkg -i $PACKAGE_DIR/rocthrust-*.deb
+    - $SUDO_CMD dpkg -i $PACKAGE_DIR/rocprim*.deb
+    - $SUDO_CMD dpkg -i $PACKAGE_DIR/rocthrust*.deb
     # Test install
     - mkdir -p package_test
     - cd package_test

.jenkins/common.groovy

@@ -8,6 +8,8 @@ def runCompileCommand(platform, project, jobName, boolean debug=false, boolean s
     String buildTypeArg = debug ? '-DCMAKE_BUILD_TYPE=Debug' : '-DCMAKE_BUILD_TYPE=Release'
     String buildTypeDir = debug ? 'debug' : 'release'
     String cmake = platform.jenkinsLabel.contains('centos') ? 'cmake3' : 'cmake'
+    //Set CI node's gfx arch as target if PR, otherwise use default targets of the library
+    String amdgpuTargets = env.BRANCH_NAME.startsWith('PR-') ? '-DAMDGPU_TARGETS=\$gfx_arch' : ''
 
     def getRocPRIM = auxiliary.getLibrary('rocPRIM', platform.jenkinsLabel, 'develop', sameOrg)
 
@@ -16,14 +18,14 @@ def runCompileCommand(platform, project, jobName, boolean debug=false, boolean s
                 ${getRocPRIM}
                 cd ${project.paths.project_build_prefix}
                 mkdir -p build/${buildTypeDir} && cd build/${buildTypeDir}
-                ${cmake} -DCMAKE_CXX_COMPILER=/opt/rocm/bin/hipcc ${buildTypeArg} -DBUILD_TEST=ON -DBUILD_BENCHMARK=ON ../..
+                ${auxiliary.gfxTargetParser()}
+                ${cmake} -DCMAKE_CXX_COMPILER=/opt/rocm/bin/hipcc ${buildTypeArg} ${amdgpuTargets} -DBUILD_TEST=ON -DBUILD_BENCHMARK=ON ../..
                 make -j\$(nproc)
                 """
 
     platform.runCommand(this, command)
 }
 
-
 def runTestCommand (platform, project)
 {
     String sudo = auxiliary.sudo(platform.jenkinsLabel)

.jenkins/precheckin.groovy

@@ -15,6 +15,8 @@ def runCI =
 
     def prj = new rocProject('rocThrust', 'precheckin')
 
+    prj.timeout.compile = 420
+
     // Define test architectures, optional rocm version argument is available
     def nodes = new dockerNodes(nodeDetails, jobName, prj)
 

.jenkins/staticanalysis.groovy

@@ -0,0 +1,68 @@
+#!/usr/bin/env groovy
+// This shared library is available at https://github.com/ROCmSoftwarePlatform/rocJENKINS/
+@Library('rocJenkins@pong') _
+
+// This is file for internal AMD use.
+// If you are interested in running your own Jenkins, please raise a github issue for assistance.
+
+import com.amd.project.*
+import com.amd.docker.*
+import java.nio.file.Path
+
+def runCompileCommand(platform, project, jobName, boolean debug=false)
+{
+    project.paths.construct_build_prefix()
+
+    def command = """#!/usr/bin/env bash
+            set -x
+            ${project.paths.project_build_prefix}/docs/run_doc.sh
+            """
+
+    try
+    {
+        platform.runCommand(this, command)
+    }
+    catch(e)
+    {
+        throw e
+    }
+
+    publishHTML([allowMissing: false,
+                alwaysLinkToLastBuild: false,
+                keepAll: false,
+                reportDir: "${project.paths.project_build_prefix}/docs/_build/html",
+                reportFiles: "index.html",
+                reportName: "Documentation",
+                reportTitles: "Documentation"])
+}
+
+def runCI = 
+{
+    nodeDetails, jobName->
+
+    def prj  = new rocProject('rocThrust', 'StaticAnalysis')
+
+    // Define test architectures, optional rocm version argument is available
+    def nodes = new dockerNodes(nodeDetails, jobName, prj)
+
+    boolean formatCheck = false
+    boolean staticAnalysis = true
+
+    def compileCommand =
+    {
+        platform, project->
+
+        runCompileCommand(platform, project, jobName, false)
+    }
+
+    buildProject(prj , formatCheck, nodes.dockerArray, compileCommand, null, null, staticAnalysis)
+}
+
+ci: { 
+    String urlJobName = auxiliary.getTopJobName(env.BUILD_URL)
+
+    properties(auxiliary.addCommonProperties([pipelineTriggers([cron('0 1 * * 6')])]))
+    stage(urlJobName) {
+        runCI([ubuntu18:['any']], urlJobName)
+    }
+}

.jenkins/staticlibrary.groovy

@@ -10,6 +10,8 @@ def runCI =
 
     def prj = new rocProject('rocThrust', 'Static Library PreCheckin')
 
+    prj.timeout.compile = 420
+
     def nodes = new dockerNodes(nodeDetails, jobName, prj)
 
     def commonGroovy

CHANGELOG.md

@@ -2,7 +2,26 @@
 
 Full documentation for rocThrust is available at [https://rocthrust.readthedocs.io/en/latest/](https://rocthrust.readthedocs.io/en/latest/)
 
-## [Unreleased rocThrust-2.10.7 for ROCm 4.0.0]
+## [rocThrust-2.11.0 for ROCm 4.3.0]
+### Added
+- Updated to match upstream Thrust 1.11
+- gfx90a support added
+- gfx803 support re-enabled
+
+## [rocThrust-2.10.9 for ROCm 4.2.0]
+### Added
+- Updated to match upstream Thrust 1.10
+### Changed
+- Minimum cmake version required for building rocThrust is now 3.10.2
+### Fixed
+- Size zero inputs are now properly handled with newer ROCm builds that no longer allow zero-size kernel grid/block dimensions
+- Warning of unused results fixed.
+
+## [rocThrust-2.10.8 for ROCm 4.1.0]
+### Added
+- No new features
+
+## [rocThrust-2.10.7 for ROCm 4.0.0]
 ### Added
 - Updated to upstream Thrust 1.10.0
 - Implemented runtime error for unsupported algorithms and disabled respective tests.

CMakeLists.txt

@@ -2,7 +2,7 @@
 # Copyright 2019-2020 Advanced Micro Devices, Inc.
 # ########################################################################
 
-cmake_minimum_required(VERSION 3.5.1 FATAL_ERROR)
+cmake_minimum_required(VERSION 3.10.2 FATAL_ERROR)
 
 # Install prefix
 if(WIN32)
@@ -32,7 +32,18 @@ endif()
 
 set(CMAKE_INSTALL_RPATH_USE_LINK_PATH TRUE CACHE BOOL "Add paths to linker search and installed rpath")
 
+# rocm-cmake contains common cmake code for rocm projects to help
+# setup and install
+find_package( ROCM CONFIG )
+include( ROCMSetupVersion )
+include( ROCMCreatePackage )
+include( ROCMInstallTargets )
+include( ROCMPackageConfigHelpers )
+include( ROCMInstallSymlinks )
+include( ROCMCheckTargetIds OPTIONAL )
+
 # Detect compiler support for target ID
+# This section is deprecated. Please use rocm_check_target_ids for future use.
 if( CMAKE_CXX_COMPILER MATCHES ".*/hipcc$" )
   execute_process(COMMAND ${CMAKE_CXX_COMPILER} "--help"
     OUTPUT_VARIABLE CXX_OUTPUT
@@ -41,12 +52,20 @@ if( CMAKE_CXX_COMPILER MATCHES ".*/hipcc$" )
   string(REGEX MATCH ".mcode\-object\-version" TARGET_ID_SUPPORT ${CXX_OUTPUT})
 endif()
 
-# Use target ID syntax if supported for AMDGPU_TARGETS
-if(TARGET_ID_SUPPORT)
-  set(AMDGPU_TARGETS gfx900:xnack-;gfx906:xnack-;gfx908:xnack-;gfx908:xnack+ CACHE STRING "List of specific machine types for library to target")
+#Set the AMDGPU_TARGETS with backward compatiblity
+if(COMMAND rocm_check_target_ids)
+   rocm_check_target_ids(DEFAULT_AMDGPU_TARGETS
+       TARGETS "gfx803;gfx900:xnack-;gfx906:xnack-;gfx908:xnack-;gfx90a:xnack-;gfx90a:xnack+;gfx1030"
+   )
 else()
-  set(AMDGPU_TARGETS gfx900;gfx906;gfx908 CACHE STRING "List of specific machine types for library to target")
+    # Use target ID syntax if supported for AMDGPU_TARGETS
+    if(TARGET_ID_SUPPORT)
+        set(DEFAULT_AMDGPU_TARGETS "gfx803;gfx900:xnack-;gfx906:xnack-;gfx908:xnack-;gfx1030")
+    else()
+        set(DEFAULT_AMDGPU_TARGETS "gfx803;gfx900;gfx906;gfx908")
+    endif()
 endif()
+set(AMDGPU_TARGETS "${DEFAULT_AMDGPU_TARGETS}" CACHE STRING "List of specific machine types for library to target")
 
 # Verify that hcc or hipcc compiler is used on ROCM platform
 include(cmake/VerifyCompiler.cmake)
@@ -80,11 +99,17 @@ set(CMAKE_CXX_STANDARD 14)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 
+if(DISABLE_WERROR)
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wextra")
+else()
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wextra -Werror")
+endif()
+
 # Get dependencies
 include(cmake/Dependencies.cmake)
 
 # Setup VERSION
-rocm_setup_version(VERSION "2.10.7")
+rocm_setup_version(VERSION "2.10.9")
 
 # Print configuration summary
 include(cmake/Summary.cmake)

cmake/Dependencies.cmake

@@ -60,38 +60,3 @@ if(BUILD_TEST)
   )
   find_package(GTest REQUIRED)
 endif()
-
-# Find or download/install rocm-cmake project
-find_package(ROCM QUIET CONFIG PATHS /opt/rocm)
-if(NOT ROCM_FOUND)
-  set(rocm_cmake_tag "master" CACHE STRING "rocm-cmake tag to download")
-  file(
-    DOWNLOAD https://github.com/RadeonOpenCompute/rocm-cmake/archive/${rocm_cmake_tag}.zip
-    ${CMAKE_CURRENT_BINARY_DIR}/rocm-cmake-${rocm_cmake_tag}.zip
-    STATUS rocm_cmake_download_status LOG rocm_cmake_download_log
-  )
-  list(GET rocm_cmake_download_status 0 rocm_cmake_download_error_code)
-  if(rocm_cmake_download_error_code)
-    message(FATAL_ERROR "Error: downloading "
-      "https://github.com/RadeonOpenCompute/rocm-cmake/archive/${rocm_cmake_tag}.zip failed "
-      "error_code: ${rocm_cmake_download_error_code} "
-      "log: ${rocm_cmake_download_log} "
-    )
-  endif()
-
-  execute_process(
-    COMMAND ${CMAKE_COMMAND} -E tar xzf ${CMAKE_CURRENT_BINARY_DIR}/rocm-cmake-${rocm_cmake_tag}.zip
-    WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
-    RESULT_VARIABLE rocm_cmake_unpack_error_code
-  )
-  if(rocm_cmake_unpack_error_code)
-    message(FATAL_ERROR "Error: unpacking  ${CMAKE_CURRENT_BINARY_DIR}/rocm-cmake-${rocm_cmake_tag}.zip failed")
-  endif()
-  find_package(ROCM REQUIRED CONFIG PATHS ${CMAKE_CURRENT_BINARY_DIR}/rocm-cmake-${rocm_cmake_tag})
-endif()
-
-include(ROCMSetupVersion)
-include(ROCMCreatePackage)
-include(ROCMInstallTargets)
-include(ROCMPackageConfigHelpers)
-include(ROCMInstallSymlinks)

docs/Doxyfile

@@ -32,7 +32,7 @@ DOXYFILE_ENCODING      = UTF-8
 # title of most generated pages and in a few other places.
 # The default value is: My Project.
 
-PROJECT_NAME           = "RCCL"
+PROJECT_NAME           = thrust
 
 # The PROJECT_NUMBER tag can be used to enter a project or revision number. This
 # could be handy for archiving the generated documentation or if some version
@@ -162,7 +162,7 @@ FULL_PATH_NAMES        = YES
 # will be relative from the directory where doxygen is started.
 # This tag requires that the tag FULL_PATH_NAMES is set to YES.
 
-STRIP_FROM_PATH        = 
+STRIP_FROM_PATH        = ..
 
 # The STRIP_FROM_INC_PATH tag can be used to strip a user-defined part of the
 # path mentioned in the documentation of a class, which tells the reader which
@@ -2029,7 +2029,7 @@ ENABLE_PREPROCESSING   = YES
 # The default value is: NO.
 # This tag requires that the tag ENABLE_PREPROCESSING is set to YES.
 
-MACRO_EXPANSION        = NO
+MACRO_EXPANSION        = YES
 
 # If the EXPAND_ONLY_PREDEF and MACRO_EXPANSION tags are both set to YES then
 # the macro expansion is limited to the macros specified with the PREDEFINED and
@@ -2044,7 +2044,7 @@ EXPAND_ONLY_PREDEF     = NO
 # The default value is: YES.
 # This tag requires that the tag ENABLE_PREPROCESSING is set to YES.
 
-SEARCH_INCLUDES        = YES
+SEARCH_INCLUDES        = NO
 
 # The INCLUDE_PATH tag can be used to specify one or more directories that
 # contain include files that are not input files but should be processed by the
@@ -2069,7 +2069,20 @@ INCLUDE_FILE_PATTERNS  =
 # recursively expanded use the := operator instead of the = operator.
 # This tag requires that the tag ENABLE_PREPROCESSING is set to YES.
 
-PREDEFINED             = 
+PREDEFINED             = THRUST_NOEXCEPT=noexcept \
+                         "THRUST_INLINE_CONSTANT=inline constexpr" \
+                         "THRUST_DEFAULT={}" \
+                         "THRUST_NODISCARD=[[nodiscard]]" \
+                         "THRUST_MR_DEFAULT_ALIGNMENT=alignof(max_align_t)" \
+                         "THRUST_FINAL=final" \
+                         "THRUST_OVERRIDE=" \
+                         "THRUST_BEGIN_NS=namespace thrust {" \
+                         "THRUST_END_NS=}" \
+                         "cuda_cub=system::cuda" \
+                         THRUST_PREVENT_MACRO_SUBSTITUTION= \
+                         __thrust_exec_check_disable__= \
+                         DOXYGEN_DOCUMENTATION_BUILD= \
+                         DOXYGEN_SHOULD_SKIP_THIS=
 
 # If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then this
 # tag can be used to specify a list of macro names that should be expanded. The
@@ -2088,7 +2101,7 @@ EXPAND_AS_DEFINED      =
 # The default value is: YES.
 # This tag requires that the tag ENABLE_PREPROCESSING is set to YES.
 
-SKIP_FUNCTION_MACROS   = YES
+SKIP_FUNCTION_MACROS   = NO
 
 #---------------------------------------------------------------------------
 # Configuration options related to external references

docs/Makefile

@@ -12,9 +12,13 @@ BUILDDIR      = _build
 help:
 	@$(SPHINXBUILD) -M help "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
 
-.PHONY: help Makefile
+.PHONY: help Makefile clean
 
 # Catch-all target: route all unknown targets to Sphinx using the new
 # "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).
 %: Makefile
 	@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+
+clean:
+	rm -rf docBin/ api/
+	@$(SPHINXBUILD) -M clean "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)