GitLab

c9d0659f CMakeLists.txt: configurable ROCm-DS.cmake repo <> Thu Apr 24 20:17:45 2025 +0000 <> domcharrier <docharri@amd.com>
Make the repository from which to obtain the ROCm-DS.cmake
script configurable via the environment variable
`ROCm-DS_CMAKE_SCRIPT_REPO`.

96ca34f2 explicitly call out libhipcxx dependency <> Wed Apr 16 02:20:16 2025 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

deb37164 revise package creation: do not disable test builds when creating package <> Tue Apr 15 05:33:33 2025 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

c317769a CMakeLists.txt: adopt ROCm-DS_CMAKE_MODULE_PATH <> Thu Apr 3 16:05:06 2025 +0000 <> domcharrier <docharri@amd.com>
Adopt a new CMake filepath parameter ROCm-DS_CMAKE_MODULE_PATH
and get rid of boolean parameter ROCm-DS_CMAKE_FROM_MODULE_PATH
(and specification of CMAKE_MODULE_PATH).

The new parameter ROCm-DS_CMAKE_MODULE_PATH can be set also
from a shell via `export ROCm-DS_CMAKE_MODULE_PATH=...`.

55b11daa fix(CPack): write version_config.hpp <> Tue Apr 1 06:39:54 2025 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

df6b17bc hotfix: add hash_bench to CU_SOURCES <> Mon Mar 31 09:37:46 2025 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

60287a98 fix: use CMake package version 0.2.0 <> Mon Mar 31 01:03:27 2025 -0700 <> Philipp Samfass <philipp.samfass@amd.com>
Details:
When creating packages for different hipCo versions,
the package names need to be disjunct and use a different
version (internal requirement).

a9cf2e7f feat(WIP): add option CUCO_ONLY_CREATE_PACKAGE to only create a package <> Fri Mar 28 05:07:28 2025 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

dd8d666a feat: allow hipCo to use a local rocm-ds-cmake installation <> Fri Mar 28 04:18:13 2025 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

1c5f8ab4 doc(README.md): add further requirements and build instructions <> Thu Feb 27 15:27:16 2025 +0100 <> Michael Obersteiner <Michael.Obersteiner@amd.com>
With this pull request, we add additional documentation and update the requirements:

- Update requirements (including necessary ROCm packages and git)
- Added note for installation of potentially missing ROCm components
- Updated build instructions for tests
4f28a0d0 updated requirements and build information in README <> Thu Feb 27 01:32:58 2025 -0800 <> Michael Obersteiner <Michael.Obersteiner@amd.com>

58fc4afc fix: remove CUDART_VERSION definition <> Thu Feb 27 11:20:11 2025 +0100 <> Michael Obersteiner <Michael.Obersteiner@amd.com>

434ac049 added date to note <> Thu Feb 27 01:39:34 2025 -0800 <> Michael Obersteiner <Michael.Obersteiner@amd.com>

fb175cb4 added note for unsupported barrier <> Thu Feb 27 01:35:05 2025 -0800 <> Michael Obersteiner <Michael.Obersteiner@amd.com>

4939f93a removed CUDART_VERSION definition <> Tue Feb 25 07:43:09 2025 -0800 <> Michael Obersteiner <Michael.Obersteiner@amd.com>

d8566f12 doc(README.md): review README wrt wavefront size 32 support <> Thu Feb 20 04:36:30 2025 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

84a46572 fix/feat: improve robustness of build when compiling for wave size 32 <> Thu Feb 20 04:23:22 2025 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

26e4860b fix(CMakeLists.txt): Fix default build behavior for tests/benchmarks Only build tests and benchmarks per default for standalone builds, not if hipCo is consumed by another project. <> Mon Jan 20 10:45:38 2025 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

b39c1eb7 Fix(open_addressing_ref_impl.cuh): Add __host__ decorator for extract_payload <> Thu Jan 16 23:41:37 2025 -0800 <> Philipp Samfass <philipp.samfass@amd.com>
Details:
The hipGraph build requires extract_payload to
be a `__host__` function, too. On ROCm side,
we do not have an equivalent of `expt-relaxed-constexpr`,
so we need to add the `__host__` decorator explicitly.

88d96259 fix (pair.inl): Use correct libhipcxx namespace. <> Fri Jan 17 01:38:42 2025 -0800 <> Philipp Samfass <philipp.samfass@amd.com>
Details:

Older libhipcxx versions use the "hip" namespace name
while more recent versions use "cuda". We therefore now use
the macros from libhipcxx to put the tuple_helpers.inl symbols
into the correct libhipcxx namespace. This ensures that hipCo builds
fine with libhipcxx v2.2.0.

2cc52bf4 fix: remove license from readme <> Thu Nov 21 18:33:02 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

d2b97428 refactor: unify todo comments <> Wed Nov 20 12:15:23 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

3eabfd2a refactor: use reverse hipification for cuda APIs to minimize code changes <> Wed Nov 20 12:01:55 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

22f28e63 update minimum ROCm requirement to 6.3 <> Wed Nov 20 10:33:45 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

5168122f reverse hipify: hipcub -> cub <> Tue Nov 19 19:05:48 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

bb67177d fix(shared_memory_example.cu): fix/war compiler warning <> Tue Nov 19 17:46:27 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

7c3a0c0b fix(hash_test.cu): fix compiler warnings about large integer literals <> Tue Nov 19 17:31:30 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

ff666598 fix/war(functors.cuh): Fixes corrupted key/value tuples arising with aow_storage (window size 2) in static_map unit tests (retrieve_all API). <> Tue Nov 19 16:00:40 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

a5aecbef fix/war(open_addressing_ref_impl.cuh): Add missing switch case to fix compiler warnings <> Tue Nov 19 15:23:51 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

1bc88ddd fix(static_set): fix compiler warnings for unused variables/aliases <> Tue Nov 19 15:22:54 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

3eddc3f8 fix/war: set default CG size for static set to 4 consistent with the upstream. <> Tue Nov 19 15:19:35 2024 +0100 <> Philipp Samfass <psamfass@amd.com>
Failing unit tests related to rehash have been observed in the original code with CG size 1 (static set)
and they happen with the hipCo port, too. These issues are going to get addressed in a future upgrade.

d100731f refactor: remove no longer required code for path without cg in static_map <> Mon Nov 18 12:00:35 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

d295116f fix(benchmarks): fix/war compiler error due to constexpr <> Mon Nov 18 10:55:44 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

c3e7bec1 doc(README.md): added clarifications regarding support of wavesize 32/64 into README <> Wed Oct 16 04:02:55 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

ed3ce671 fix(CMakeLists.txt): per default, do not build for both gfx1100 and gfx9* archs as they use different wavefront sizes which are incompatible <> Wed Oct 16 03:17:04 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

f7164aea fix(*/static_multimap.cuh): force user at compile time to decide whether warp/wavesize 32 should be used. This is necessary as HIP does not accept to create cooperative groups larger than tile_size 32 if compiling for warp/wavesize 32. <> Fri Oct 11 14:59:37 2024 +0200 <> Philipp Samfass <psamfass@amd.com>

addc5c64 remove no longer needed contains/count implementations of static_multimap <> Sun Nov 17 08:39:49 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

b66e978a fix(static_set/kernels.cuh): fix UB <> Sun Nov 17 08:29:32 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

bb8c5388 remove no longer needed insert_if variant without CG (static_map) <> Sun Nov 17 08:24:50 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

f107f65a removed no longer needed retrieve_outer variant <> Sun Nov 17 08:11:35 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

ad6ddd30 fix/feat: hipification of static_map_ref.inl <> Fri Nov 15 07:31:38 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

df30a4e1 fix(static_set_ref.inl): use original signature/type for cooperative group <> Fri Aug 9 01:48:20 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

fbd60898 refactor: use cuda namespace alias for libhipcxx to reduce code changes <> Fri Nov 15 05:07:06 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

601b1edf feat/fix(remove cooperative group extensions workaround): ROCm 6.3 ships the coop group APIs that are needed for hipCollections. <> Thu Aug 8 05:05:26 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>
In this commit, the workarounds taken earlier with a custom
implementation of the missing CG APIs ourselves is removed and
 we now employ the new ROCm CG API implementations instead.

18350ecf fix(open_addressing_ref_impl.cuh): fix compiler warning about unused alias <> Fri Nov 15 03:59:54 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

f28d789d fix: do not use this in constexpr functions (compiler error!) <> Fri Nov 15 03:53:10 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

025f8cf6 fix/war: fix for "error: __device__ function 'with' cannot overload __host__ __device__ function 'with'" <> Fri Nov 15 03:45:53 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

72937d08 fix ambiguous calls error for catch2 <> Thu Nov 14 08:25:40 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

a4f35ff8 fix(tests/CMakeLists.txt): interpret *.cu as HIP files <> Thu Nov 14 07:16:32 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

20c48511 fix(static_multimap): remove outdated count operations without CG <> Wed Nov 13 06:16:22 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

945543d3 fix: fix compiler errors due to missing template/typename keywords <> Wed Nov 13 03:43:33 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

ed9cbb2e fix/war(__config): comment unsupported CUDA checks <> Wed Nov 13 03:27:26 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

8304dcdf fix(non_match_test.hip): fix compiler warning for unused alias <> Wed Nov 13 02:17:21 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

566d0260 fix(host_bulk_example.hip): fix compiler warning for unused variable <> Wed Nov 13 02:16:16 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

fcf01f8d cleanup: remove no longer supported benchmarks (CG sizes, block sizes) <> Wed Nov 13 01:26:10 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

a0dfa82d doc(README.md): document support for gfx1100 <> Wed Oct 2 12:22:43 2024 +0200 <> Philipp Samfass <psamfass@amd.com>

4cd9ffbf feat(CMakeLists.txt): add gfx1100 as supported architecture to hipCo <> Wed Oct 2 10:30:54 2024 +0200 <> Philipp Samfass <psamfass@amd.com>

c4f572ca feat(wavefront size portability): Use __POPCC and __FFS as well as lane_mask to support wavefront sizes 32 and 64. <> Wed Oct 2 10:37:49 2024 +0200 <> Philipp Samfass <psamfass@amd.com>

75f39ecc feat(include/hipco/detail/utils.cuh): add portable wavefront-size dependent __POPC and __FFS device functions <> Wed Oct 2 10:33:31 2024 +0200 <> Philipp Samfass <psamfass@amd.com>

f30a5933 hotfix(CMakeLists.txt): ignore compiler warning about narrowing const references that arise with recent clang versions: <> Wed Oct 2 00:11:07 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>
"error: non-constant-expression cannot be narrowed from type
'std::size_t' (aka 'unsigned long') to 'int' in initializer list
[-Wc++11-narrowing-const-reference]"

Note: this is a breaking change that breaks backwards compatibility with
older ROCm releases than 6.2

f3cd5759 fix: use publicly available hipBench port <> Mon Sep 30 15:13:06 2024 +0200 <> Philipp Samfass <psamfass@amd.com>

7ecba482 feat/fix(CMakeLists.txt): Add xnack+ archs to supported archs towards enabling ASAN builds. <> Tue Jul 30 04:32:17 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

0864dcd5 fix/feat: enable inlining WAR depending on HIP_VERSION <> Tue Aug 27 02:25:25 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

e37a5b3e doc: increase minimum required ROCm version to 6.2 <> Mon Aug 26 06:27:17 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

d8e5677b Revert "tmpfix/war: workaround for invalid modulo results arising in hipDF, see SWDEV-436805" <> Mon Aug 26 05:36:45 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

84a49e24 cleanup: delete notebooks which should not be part of publicly released branch (might be restored internally on separate branch later) <> Tue Aug 27 01:01:26 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>

38eda9f4 Add license file <> Tue May 7 16:35:27 2024 +0000 <> monoatamd <mnorouzi@amd.com>

10e9957d fix issues with contains and find benchmarks in STATIC_MAP_BLOCK_BENCH <> Mon May 6 15:59:21 2024 +0000 <> monoatamd <mnorouzi@amd.com>

646d4d39 Updated licenses <> Mon May 6 10:56:32 2024 +0000 <> monoatamd <mnorouzi@amd.com>

ed8563a6 cleanup <> Fri May 3 16:12:50 2024 +0000 <> monoatamd <mnorouzi@amd.com>

168210d4 tmpfix(CMake): don't install hip extension headers into include root directory to not create conflicts with other packages that use/install these headers and that may have a different version. <> Wed Jul 17 23:53:07 2024 -0700 <> Philipp Samfass <philipp.samfass@amd.com>
Todo: maybe we need to look for an even cleaner way, e.g., by making the
hip extensions and independent package.

10e08ca1 Fix some issues related to find and contains benchmarks that raise with STATIC_MAP_BLOCK_BENCH <> Mon May 6 16:33:54 2024 +0000 <> monoatamd <mnorouzi@amd.com>

e2459adc Update README.md <> Mon Mar 25 08:57:08 2024 +0100 <> Philipp Samfass <philipp.samfass@amd.com>

3ed44dd9 feat/fix: enable more arch strings <> Tue Mar 19 07:45:46 2024 -0500 <> Philipp Samfass <psamfass@amd.com>

1c9735cc feat/maint: add support for MI300 gfx archs <> Tue Mar 12 05:18:34 2024 -0500 <> Philipp Samfass <psamfass@amd.com>

35d24bb5 tmpfix/war: workaround for invalid modulo results arising in hipDF, see SWDEV-436805 <> Wed Feb 14 07:24:36 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

1e132004 Removed more unnecessary cuda files <> Thu Jan 11 08:41:39 2024 +0000 <> monoatamd <mnorouzi@amd.com>

ff3409ce removed unnecessary cuda files. Enable an example in static_set <> Wed Jan 10 16:43:18 2024 +0000 <> monoatamd <mnorouzi@amd.com>

aac8edbf Disable benchmark per default <> Tue Nov 14 15:52:46 2023 +0000 <> Mohammad Norouzi <mnorouzi@amd.com>

c1e34e9a Remove probe_benchmark <> Mon Nov 13 18:09:30 2023 +0000 <> Mohammad Norouzi <mnorouzi@amd.com>

b4b65287 Enable static_map tile and block benchmakrs <> Mon Nov 13 12:51:22 2023 +0000 <> Mohammad Norouzi <mnorouzi@amd.com>

548e8afb enable static_set_cg_bench <> Mon Nov 13 12:43:55 2023 +0000 <> Mohammad Norouzi <mnorouzi@amd.com>

f4a3e2cd reenable STATIC_SET_BENCH <> Fri Nov 10 03:34:11 2023 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

32289533 improved naming of externally visible compiler defines <> Thu Nov 9 09:15:21 2023 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

781b8f82 Delete SECURITY.md <> Tue Nov 7 14:59:50 2023 +0100 <> Philipp Samfass <philipp.samfass@amd.com>

9c733647 removed LICENSE file that does not apply <> Tue Nov 7 05:40:34 2023 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

2e2fc3e2 fix README: minimum required ROCm version is 5.7.0 (a necessary fix for thrust has been added with 5.7) <> Tue Nov 7 05:03:50 2023 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

9810fd37 Enable static_map/rehash_test <> Tue Nov 7 10:23:07 2023 +0000 <> monoatamd <mnorouzi@amd.com>

e104576d Enable key_sentinel_test in static_map <> Mon Nov 6 17:07:36 2023 +0000 <> monoatamd <mnorouzi@amd.com>

8734e963 Enable all tests in static_set, static_multimap and static_map (except two tests) <> Mon Nov 6 14:51:45 2023 +0000 <> monoatamd <mnorouzi@amd.com>

2323d0f2 fix(cuda.cuh): fix compiler warning about missing inline keyword <> Tue Nov 19 15:33:23 2024 +0100 <> Philipp Samfass <psamfass@amd.com>

d578631c Added typename to eliminate warnings <> Tue Oct 31 09:17:31 2023 +0000 <> Mohammad Norouzi <mnorouzi@amd.com>

7bbd0620 addressed more merge conflicts + started hipification (WIP) <> Thu Oct 12 14:19:26 2023 +0000 <> Philipp Samfass <philipp.samfass@amd.com>

7edf5d3f fix: added checking for new architectures that are now supported in ROCm 5.7+ <> Wed Oct 11 07:57:35 2023 +0000 <> Philipp Samfass <philipp.samfass@amd.com>

24f37de8 fix: adapt README to reflect recent changes to build <> Wed Oct 11 07:51:56 2023 +0000 <> Philipp Samfass <philipp.samfass@amd.com>

37aad3d9 fix: select only supported architectures per default and check in CMake build for unsupported architectures (and return an error) <> Wed Oct 11 07:48:12 2023 +0000 <> Philipp Samfass <philipp.samfass@amd.com>

311ba4e0 fix: fix formatting of CMakeLists.txt <> Wed Oct 11 06:41:13 2023 +0000 <> Philipp Samfass <philipp.samfass@amd.com>

40bbe161 Fix: install CG groups extensions when running make install. <> Fri Oct 6 13:28:49 2023 +0000 <> Philipp Samfass <philipp.samfass@amd.com>

1396c2a5 Update README.md <> Thu Oct 5 18:30:37 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

af2bde6c Update README.md <> Thu Oct 5 18:29:15 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

be89bab4 update static_map_tile and _block benchmarks to use hipco instead of cuco <> Wed Jul 12 09:43:31 2023 +0000 <> mnorouzi <mnorouzi@amd.com>

0a7c082f update implementation of static_map with user-defined tile and block sizes <> Wed Jul 12 08:55:38 2023 +0000 <> mnorouzi <mnorouzi@amd.com>

bc8e6b9e replace CuCollections -> hipCollections in comments <> Wed Jul 12 10:29:47 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

e8ced6df made set_mask protected <> Wed Jul 12 09:54:47 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

275e701e fixed header include paths <> Tue Jul 11 19:43:23 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

e4738a78 reflected namespace structure in folder structure <> Tue Jul 11 18:43:14 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

33dc32fc wrapped hip extensions into a separate namespace <> Tue Jul 11 18:15:22 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

6dc1d68d Make branch mergeable with dev <> Tue Jul 11 12:48:13 2023 +0000 <> monoatamd <mnorouzi@amd.com>

0d6e1982 moved cooperative group extensions into separate folder + simplified code <> Tue Jul 11 13:48:47 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

d554d61f renamed header file ending to make it consistent <> Tue Jul 11 11:39:55 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

19b77563 added more documentation and excluded *impl symbols in Doxyfile for ROCm-docs-core documentation <> Tue Jul 11 08:50:27 2023 +0200 <> Philipp Samfass <psamfass@amd.com>

4dd734d4 added scripts and configuration files to generate documentation in style of rocm-docs-core <> Mon Jul 10 19:29:29 2023 +0200 <> Philipp Samfass <psamfass@amd.com>

644fddde fix: fixed compiler errors <> Fri Jul 7 09:49:40 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

b5fb5198 feat: added doxygen documentation for cooperative groups extension <> Fri Jul 7 09:32:11 2023 +0200 <> Philipp Samfass <psamfass@amd.com>

5dc41a0a feat: Add script to run probe_bench <> Tue Jul 4 14:50:46 2023 +0000 <> monoatamd <mnorouzi@amd.com>

6c1b3c4c fixed compiler error (pass key instead of key-value pair) <> Tue Jul 4 16:35:55 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

0a51471c fixed CI checking scripts for style and documentation <> Tue Jul 4 14:33:42 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

d2dea5a2 feat: made cg sizes used for static_set configurable <> Tue Jul 4 13:22:52 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

f486e088 fixed example, which did not use cooperative group extension <> Tue Jul 4 12:30:06 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

c9368444 bug: Use CG workaournd for static_map <> Thu Jun 29 11:25:08 2023 +0200 <> monoatamd <mnorouzi@amd.com>

e0e9a207 added baseline to performance evaluation notebook <> Tue Jun 27 13:45:28 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

bcf3ea5b feat: Add charts for different block sizes <> Tue Jun 27 07:17:54 2023 +0000 <> monoatamd <mnorouzi@amd.com>

06bedd21 update charts for static_map with different tile sizes <> Mon Jun 26 19:59:53 2023 +0200 <> monoatamd <mnorouzi@amd.com>

3f59bcc1 bug: Fix tile size issue <> Mon Jun 26 16:35:27 2023 +0000 <> monoatamd <mnorouzi@amd.com>

36f4ac40 added analysis notebook for static set (different CG tile sizes) <> Mon Jun 26 18:26:53 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

7ec21b6e feat: Remove optnone from warp primitives <> Mon Jun 26 15:01:58 2023 +0000 <> monoatamd <mnorouzi@amd.com>

c7008046 added benchmark for testing different probing cg sizes with static_set <> Mon Jun 26 15:42:13 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

4d7a15ab bug: Fix benchmark path <> Mon Jun 26 11:48:43 2023 +0000 <> monoatamd <mnorouzi@amd.com>

16ca4996 feat: Add notebook for static_map with different tile sizes <> Mon Jun 26 08:59:19 2023 +0000 <> monoatamd <mnorouzi@amd.com>

00c2ed8b added plotting of basline variant to CG performance analysis notebook <> Mon Jun 26 09:10:34 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

e0800615 feat: add tile size as param for benchmarking <> Fri Jun 23 14:48:29 2023 +0000 <> monoatamd <mnorouzi@amd.com>

a8caff60 feat: added Juptyer notebook to evaluate results of STATIC_MULTIMAP_CG_BENCHMARK <> Thu Jun 22 12:17:57 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

32010513 feat: removed APIs implementations that are no longer needed after re-design of CG extensions <> Thu Jun 22 09:50:54 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

05fdd490 feat: added benchmark that tests different probing cg sizes for STATIC_MULTIMAP with hipBench <> Wed Jun 21 16:40:54 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

14ab482b initial porting of static_set (WIP) <> Wed Jun 21 14:03:42 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

8246af35 feat: Fix problem with 64 CG sizes <> Wed Jun 21 08:34:13 2023 +0000 <> monoatamd <mnorouzi@amd.com>

d3f6670a fix: set warp size to 64 or HIP <> Tue Jun 20 10:27:36 2023 +0000 <> Philipp Samfass <philipp.samfass@amd.com>

ceb6e013 fix: size of input mask for count_least_significant_bits <> Tue Jun 20 10:05:30 2023 +0000 <> Philipp Samfass <philipp.samfass@amd.com>

708977fe disabled dynamic map tests and benchmarks as they are not supported yet <> Mon Jun 19 15:00:04 2023 +0000 <> Philipp Samfass <philipp.samfass@amd.com>

76b2d355 bug: Remove eps files <> Fri Jun 16 11:45:56 2023 +0000 <> monoatamd <mnorouzi@amd.com>

1384a43a feat: add charts for static_map benchmarks <> Fri Jun 16 11:41:58 2023 +0000 <> monoatamd <mnorouzi@amd.com>

6e69ed8a feat: re-design impl. of CG workaround for static_map <> Thu Jun 15 15:31:48 2023 +0000 <> monoatamd <mnorouzi@amd.com>

d61f206d feature: added contains() support for static_multimap that does not use parallel probing based on cooperative groups <> Thu Jun 15 15:35:17 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

fba52cb8 feat: re-design CG workaround <> Thu Jun 15 12:45:38 2023 +0000 <> monoatamd <mnorouzi@amd.com>

f13934bd feat: added IPython analysis notebook for static_multimap benchmarks with HIP port <> Tue Jun 13 13:38:04 2023 +0000 <> Philipp Samfass <philipp.samfass@amd.com>

6719cb8d feat: Add macro to select tile_size for cooperative groups <> Mon Jun 12 11:34:05 2023 +0000 <> monoatamd <mnorouzi@amd.com>

be32fc28 feat: Implement CG workaround for static_map <> Fri Jun 9 15:13:18 2023 +0000 <> mnorouzi <mnorouzi@amd.com>

4ed44b16 fixed function names and enabled flushing variant when using cg size 1 for parallel probing, as it performs better in benchmarks <> Wed Jun 7 17:30:13 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

abcde22b feat: Refactor of CG workaround (WIP) <> Wed Jun 7 16:30:00 2023 +0200 <> monoatamd <mnorouzi@amd.com>

7c6085ce addressed open todo and made ballot implementation more robust <> Wed Jun 7 15:14:06 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

0b465e4a change from tiled active flushing cg to original active flushing cg + fixed some warnings <> Wed Jun 7 11:21:13 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

e571ee00 fix for thread_rank and size in cooperative groups extension <> Mon Jun 5 17:54:46 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

36ea6ee2 fix: use 64 bit variant of popc in multiple places which fixes several issues. Todo: fix multimap test for sizes >= 1 000 000 key-value pairs <> Mon Jun 5 12:11:50 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

24127a04 fixed some corner cases for thread_rank and shuffle in cooperative groups extensions <> Fri Jun 2 14:06:57 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

468ea910 - added implementations of retrieve in static_multimap with and without outputbuffer flushing/parallel probing based on cooperative groups - added implementation of count_outer - added todos for missing APIs <> Fri Jun 2 08:10:16 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

74f46904 feat: Enable benchmarks and integrate hipBench <> Wed May 31 11:42:22 2023 +0000 <> monoatamd <mnorouzi@amd.com>

174f7cf3 feat: Enable benchmarks and  Integrate hipBench <> Wed May 31 11:35:57 2023 +0000 <> monoatamd <mnorouzi@amd.com>

f8bbbff7 feat: impl. workaround for binary_partition <> Mon May 29 10:21:10 2023 +0000 <> monoatamd <mnorouzi@amd.com>

202ce0d2 feat: Impl. binary_partition in static_multimap <> Wed May 24 10:43:05 2023 +0000 <> monoatamd <mnorouzi@amd.com>

21832baa feat: Encapsulate CG workaround. Added HIP_CG_SIZE macro to select the impl. <> Wed May 24 08:22:53 2023 +0000 <> monoatamd <mnorouzi@amd.com>

301ff0a5 WIP: implementation of static_multimap for HIP without cooperative groups 1) set default probing cg size to 1 for double hashing 2) implemented insert, insert_if and contains without cg 3) deactivated unsupported tests for static multimap <> Tue May 23 12:02:24 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

48831b76 feature: select non-cg path for HIP at compile time for static_map <> Mon May 22 17:56:58 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

ba2d4e32 added insert_if kernel variant that does not require cooperative groups <> Mon May 22 17:55:12 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

cf830780 added more comments on various issues <> Wed May 17 11:02:22 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

94d8396c added todos regarding CMake + some clean-up <> Wed May 17 08:42:46 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

537df001 compile for native GPU architecture per default only when using HIP <> Tue May 16 16:29:24 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

cdd7b7fe feat: Enable test compilation + Disable warnings <> Tue May 9 11:41:27 2023 +0000 <> monoatamd <mnorouzi@amd.com>

44147d61 fixed several compiler warnings for examplesx <> Mon May 8 12:13:46 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

d061f840 feat: Fix ambiguous call to get_wrapper in catch2 <> Mon May 8 10:02:05 2023 +0000 <> monoatamd <mnorouzi@amd.com>

8aefb4d2 fixed custom_type_test and added comments <> Fri May 5 18:40:23 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

46d5e7ed feat: Enable 2 more tests <> Fri May 5 08:11:12 2023 +0000 <> monoatamd <mnorouzi@amd.com>

de1cc57b feat: Enable more tests <> Fri May 5 07:40:58 2023 +0000 <> monoatamd <mnorouzi@amd.com>

a26b9a33 fix: Ignre g.ballot and g.any. Dsble failing tsts <> Fri May 5 06:46:58 2023 +0000 <> monoatamd <mnorouzi@amd.com>

2556bb4d fixed compiler errors related to dependent templates for HIP,  examples are building now (but not running...)! <> Thu May 4 16:32:46 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

5434b123 fix/war: introduce WAR for missing/unused proclaim_return_type <> Sun Nov 17 06:38:31 2024 -0800 <> Philipp Samfass <philipp.samfass@amd.com>

c17d246a fixed more compiler errors + more hipification <> Wed May 3 18:15:16 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

95d77f0d various fixes for compilation + fixed hipcub integration into CMake + disabled unsupported cg functions ballot and any for which we will need a workaround <> Wed May 3 13:34:16 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

98dcd1a1 fixed further template-related errors <> Wed May 3 09:50:39 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

f569c42d fixed cub header dependencies for HIP + workaround for non-existing atomicCAS_block (todo!) <> Wed May 3 09:34:18 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

af605f17 further hipification <> Wed May 3 08:41:28 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

58265c1e feat: enable tests and hipify tests and benchmarks <> Tue May 2 13:41:01 2023 +0000 <> monoatamd <mnorouzi@amd.com>

570e44d3 adapting CMake config: change to rapids-cmake HIP port <> Tue May 2 09:51:03 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>

839b0299 further adaptions to CMake config for enabling HIP, renamed examples from <example name>.cu to <example name>.hip, added CMake functions to configure rocthrust and libhipcxx <> Tue May 2 11:23:51 2023 +0200 <> Philipp Samfass <philipp.samfass@amd.com>
Co-authored-by: psamfass <psamfass@amd.com>
Co-authored-by: monoatamd <mnorouzi@amd.com>
Co-authored-by: domcharrier <docharri@amd.com>
Co-authored-by: reger-men <adil.lashab@amd.com>
Co-authored-by: obersteiner <moberste@amd.com>