StreamHPC 2023-11-21 (DeviceMemcpy::Batched) (#314)

* ci: use build instead rocm-build and nvcc-build tags

This allows the build job to be performed by any runner configured
for building, instead of the ROCm-specialized builder. As the
target architectures are specified ahead of time, the GPU is not
needed during the build process, and may be performed by any builder.

* feat: Add interface for batched memcpy from rocPRIM and CUB

* style(device_memcpy): improve formatting

---------

Co-authored-by: Robin Voetter <[email protected]>
Co-authored-by: Gergely Mészáros <[email protected]>

.clang-format

@@ -59,8 +59,6 @@ BraceWrapping:
     AfterStruct: true
     AfterUnion: true
     AfterExternBlock: false
-    BeforeCatch: true
-    BeforeElse: true
     BeforeLambdaBody: true
     BeforeWhile: true
     IndentBraces: false

.gitlab-ci.yml

@@ -45,7 +45,7 @@ clang-format:
   stage: lint
   needs: []
   tags:
-    - rocm-build
+    - build
   variables:
     CLANG_FORMAT: "/opt/rocm/llvm/bin/clang-format"
     GIT_CLANG_FORMAT: "/opt/rocm/llvm/bin/git-clang-format"
@@ -62,7 +62,7 @@ copyright-date:
   stage: lint
   needs: []
   tags:
-    - rocm-build
+    - build
   rules:
     - if: '$CI_PIPELINE_SOURCE == "merge_request_event"'
   script:
@@ -107,7 +107,7 @@ build:rocm:
     - .rules:build
   stage: build
   tags:
-    - rocm-build
+    - build
   needs: []
   script:
     - cmake
@@ -145,7 +145,7 @@ build:rocm-benchmark:
     - .rules:build
   stage: build
   tags:
-    - rocm-build
+    - build
   needs: []
   script:
     - cmake
@@ -304,7 +304,7 @@ build:nvcc:
     - .nvcc
     - .rules:build
   tags:
-    - nvcc-build
+    - build
   needs: []
   script:
     - cmake
@@ -339,7 +339,7 @@ build:nvcc-benchmark:
     - .nvcc
     - .rules:build
   tags:
-    - nvcc-build
+    - build
   needs: []
   script:
     - cmake

CHANGELOG.md

@@ -10,6 +10,9 @@ See README.md on how to build the hipCUB documentation using Doxygen.
 ### Fixed
 - Fixed build issues with `rmake.py` on Windows when using VS 2017 15.8 or later due to a breaking fix with extended aligned storage.
 
+### Added
+- Added interface `DeviceMemcpy::Batched` for batched memcpy from rocPRIM and CUB.
+
 ## (Unreleased) hipCUB-2.13.1 for ROCm 5.7.0
 ### Changed
 - CUB backend references CUB and Thrust version 2.0.1.

benchmark/CMakeLists.txt

@@ -1,6 +1,6 @@
 # MIT License
 #
-# Copyright (c) 2020 Advanced Micro Devices, Inc. All rights reserved.
+# Copyright (c) 2020-2023 Advanced Micro Devices, Inc. All rights reserved.
 #
 # Permission is hereby granted, free of charge, to any person obtaining a copy
 # of this software and associated documentation files (the "Software"), to deal
@@ -80,6 +80,7 @@ add_hipcub_benchmark(benchmark_block_run_length_decode.cpp)
 add_hipcub_benchmark(benchmark_block_scan.cpp)
 add_hipcub_benchmark(benchmark_block_shuffle.cpp)
 add_hipcub_benchmark(benchmark_device_adjacent_difference.cpp)
+add_hipcub_benchmark(benchmark_device_batch_memcpy.cpp)
 add_hipcub_benchmark(benchmark_device_histogram.cpp)
 add_hipcub_benchmark(benchmark_device_memory.cpp)
 add_hipcub_benchmark(benchmark_device_merge_sort.cpp)