Граф коммитов

3159 Коммитов

Автор SHA1 Сообщение Дата
Evgeny Mankov a2b6b1e7e9 [HIPIFY][hipify-perl] Formatting
[ROCm/hip commit: defc6f8155]
2019-04-23 17:55:47 +03:00
David Salinas d4c72ad357 Revert "append the ELF flags for sram-ecc and xnack to the target triple per code object"
This reverts commit ee887dd6a2.


[ROCm/hip commit: 5843530a06]
2019-04-18 11:49:40 -04:00
Evgeny Mankov 43d7714fe4 [HIPIFY][SPARSE] cuSPARSE 10.1 support
[ROCm/hip commit: e1c87d8cae]
2019-04-16 14:59:44 +03:00
Evgeny Mankov fec9f62e0f [HIPIFY][BLAS] cuBLAS 10.1 support
[ROCm/hip commit: 032c3bf5b8]
2019-04-16 12:52:58 +03:00
Evgeny Mankov 6268243c74 Merge pull request #1023 from emankov/master
[HIPIFY][cuDNN] Add partial cudnnRNNBiasMode_t support

[ROCm/hip commit: ea389cd7b8]
2019-04-16 11:03:22 +03:00
Evgeny Mankov a7e36937ca [HIPIFY][cuDNN] Add partial cudnnRNNBiasMode_t support
[ROCm/hip commit: 1b36987c5f]
2019-04-16 11:01:01 +03:00
Maneesh Gupta f529a6af13 Merge pull request #995 from david-salinas/add_sram-ecc_and_xnack_flags_to_triple
Append the ELF flags for sram-ecc and xnack to the target triple per code object

[ROCm/hip commit: 8309632e2d]
2019-04-16 09:10:04 +05:30
Maneesh Gupta 8b7cd90f0c Merge pull request #1019 from scchan/lazy_binding
minor workaround for lazy binding

[ROCm/hip commit: d789aef46e]
2019-04-16 08:36:10 +05:30
Jeff Daily a0172ca884 In hipFree, synchronize owner of memory (#1018)
* In hipFree, if memory is associated with a device, synchronize that device's streams.

This changes the behavior from synchronizing the currently set TLS device.

* All devices sync in hipFree for _appId=-1 case.

* Revert "All devices sync in hipFree for _appId=-1 case."

This reverts commit 1efb34d6a8426661e45bc5f763422a1147aeac10.

* add HIP_SYNC_FREE env var


[ROCm/hip commit: 2b3037a6ea]
2019-04-16 08:35:55 +05:30
Mr-LiuSw ba23b5c5db add little changes in hip_runtime_api.h to work with c language (#1017)
* Update hip_runtime_api.h

when i try to use mpicc or gcc to compile a c language code which call some hip runtime api , error occured as
> /path/to/hcc_detail/hip_runtime_api.h:2268:33: error: unknown type name ‘hipFuncAttributes’; 
> hipFuncGetAttributes(hipFuncAttributes* attr, const void* func);
 
add ' struct ' for the first parameter of hipFuncGetAttributes will get ride of this problem.


[ROCm/hip commit: dc07b1b06c]
2019-04-16 08:35:36 +05:30
Aaron Enye Shi 882006555b Use COMgr to read Kernel Args Metadata (#1006)
* Add CMAKE dep to amd_comgr

* Use COMGR for read_kernarg_metadata in COV2

* Do not assume kernargs exist

* Add proper metadata destroy cleanup

* Use a process function for easier destroy

* Remove old read_kernarg_metadata

* Clean up HCC, prints, names

* Use COMGR in CMAKE by default

* Move metadata lookup for keyword values into helper

* Remove C string usage for lookup_keyword_value

* Guard COMGR for non-NVCC path

* Add hip_hcc dependency on comgr package

* Add lifetime to metadata nodes

* Find COMGR config file for amd_comgr target

* Move set_active data earlier


[ROCm/hip commit: 8a548bf40b]
2019-04-16 08:34:39 +05:30
Evgeny Mankov e73bfab77c [HIPIFY] cuDNN 7.5.0.56 support
[ROCm/hip commit: d3a810ab11]
2019-04-15 15:46:46 +03:00
Maneesh Gupta ab22256587 [ci] Enable tests on ROCm 2.3
Change-Id: Id344ef600b0868f36f2e7ac08d5664234d88835b


[ROCm/hip commit: 5a0e14fbb1]
2019-04-15 12:38:01 +05:30
Yaxun (Sam) Liu 9899d0523e hip-clang: Add __align__
CUDA has __align__. Define eqivalent for hip-clang.


[ROCm/hip commit: b9f8f977fb]
2019-04-10 14:17:18 -04:00
Evgeny Mankov 9087c06f8f [HIPIFY] CUDA 10.1 Runtime API support
[ROCm/hip commit: d2d2889bfe]
2019-04-10 18:41:36 +03:00
Evgeny Mankov ac64c184fb [HIPIFY] CUDA 10.1 Driver API support
[ROCm/hip commit: b0879cd959]
2019-04-10 15:03:34 +03:00
Maneesh Gupta ae6c61575e Merge pull request #1013 from yxsamliu/config
Fix hip-config.cmake for hip-clang

[ROCm/hip commit: fa6ba987a5]
2019-04-10 07:53:22 +00:00
Yaxun (Sam) Liu cc4acbdd08 hip-clang: fix hipcc options
Use -O3 for default optimization level.
Fix --hip-device-lib-path for warnings about unused option.


[ROCm/hip commit: a31036ab74]
2019-04-09 22:31:13 -04:00
Yaxun (Sam) Liu a38f40a151 Fix hip-config.cmake for hip-clang
Define HIP_COMPILER and HIP_RUNTIME in hip-config.cmake to
facilitate other packages identify hip-clang and HIP/VDI
runtime.


[ROCm/hip commit: cce99bfa24]
2019-04-09 16:41:01 -04:00
Yaxun (Sam) Liu 0f62034e00 hip-clang: add include path for hsa headers
[ROCm/hip commit: c0e0ea9a12]
2019-04-09 15:01:50 -04:00
Maneesh Gupta a863c0dbee Merge pull request #1010 from ROCm-Developer-Tools/mangupta-patch-1
[ci] Disable hipLaunchParm.tst for now

[ROCm/hip commit: 035a3e5aa2]
2019-04-09 03:19:56 +00:00
Evgeny Mankov cb90b40344 [HIPIFY][cmake][doc] Update CMakeLists and Readme cause CUDA 10.1 and clang 8.0.0 are released
+ Fix the error in cmake: clang 8.0.0 supports CUDA 10.0, not 10.1; only trunk clang 9.0.0 supports CUDA 10.1.
+ Update Readme:
  1. clang's bug 36384 is gone, so latest stable release for Linux is clang 8.0.0, which supports CUDA up to 10.0;
  2. update dependencies, testing and building notes.


[ROCm/hip commit: 53091e5378]
2019-04-08 20:05:21 +03:00
Maneesh Gupta eabfd4c6f4 [ci] Disable hipLaunchParm.tst for now
Due to a HCC bug in rocm-head, the test fails to compile. Disabling the test until the issue is resolved.

Change-Id: Ib4e7baf0b9c2cb5f5dbe38e6dd2bab894d28886a


[ROCm/hip commit: 0a52c6661d]
2019-04-08 16:14:25 +05:30
Yaxun (Sam) Liu cf4bdb8b55 Fix regression on multi-gpu due to PR#997
[ROCm/hip commit: 271fdc4e4d]
2019-04-05 22:54:41 -04:00
Evgeny Mankov be808fd3ca Merge pull request #1003 from acowley/patch-1
hipify: translate cudaErrorInvalidSymbol

[ROCm/hip commit: 25bb74a2bb]
2019-04-05 20:21:27 +03:00
David Salinas ee887dd6a2 append the ELF flags for sram-ecc and xnack to the target triple per code object
[ROCm/hip commit: c61f265657]
2019-04-05 13:17:11 -04:00
Evgeny Mankov 93433d9fe1 [HIPIFY][doc] Fix typos and minor text inaccuracies in HIP FAQ
[ROCm/hip commit: 6e70b8494b]
2019-04-05 19:19:38 +03:00
Siu Chi Chan 23df3b5879 minor workaround for lazy binding
[ROCm/hip commit: e126f7254c]
2019-04-02 17:28:06 -04:00
Anthony Cowley b00a98f27a hipify: translate cudaErrorInvalidSymbol
[ROCm/hip commit: e9b1881c60]
2019-04-02 08:54:37 -04:00
Maneesh Gupta 8a93d8e9aa Merge pull request #998 from yxsamliu/doc
hip-clang: update installation guide.

[ROCm/hip commit: c19874e0f1]
2019-04-02 05:08:18 +00:00
Maneesh Gupta 1fbff012b4 Merge pull request #997 from yxsamliu/mgpu
hip-clang: fix kernel not found on multi-gpu

[ROCm/hip commit: 5b1e63ee2d]
2019-04-02 05:07:31 +00:00
Evgeny Mankov 7bbf825c74 [HIPIFY][tests] Fix typo in test for CUDA 10.x
[ROCm/hip commit: 8bca9f1d7a]
2019-04-01 19:52:08 +03:00
Evgeny Mankov 9e32584753 [HIPIFY][cmake] Update CMakeLists and Readme cause CUDA 10.1 and clang 8.0.0 are released
[ROCm/hip commit: 7c221450fa]
2019-04-01 19:44:52 +03:00
Yaxun Sam Liu 5072c98f32 hip-clang: fix kernel not found on multi-gpu
__hipRegisterFunction is called during by .init functions during program initialization.
It calls hipModuleGetFunction to locate kernel symbol in code objects. hipModuleGetFunction
assumes current device when locating kernel symbols. This works for HCC but not for hip-clang,
since hip-clang needs to locate kernel symbols for different devices without switching
between devices.

This patch introduces a new hsa agent parameter to ihipModuleGetFunction, which allows
__hipRegisterFunction to choose the correct hsa agent when locating kernel symbols. By
default it uses this_agent(), therefore this patch has no impact on HCC.


[ROCm/hip commit: 98b9e92908]
2019-03-31 10:08:20 -04:00
Yaxun (Sam) Liu 595d95cfe8 Update INSTALL.md
[ROCm/hip commit: ed2fbb27e4]
2019-03-30 08:29:08 -04:00
Yaxun Sam Liu c0682dda96 hip-clang: update installation guide.
[ROCm/hip commit: 409aa49218]
2019-03-30 08:24:49 -04:00
Wen-Heng (Jack) Chung 1cc94f9369 Make hipModuleGetGlobal be in HIP runtime so it can be discovered at runtime (#981)
* Make hipModuleGetGlobal be in HIP runtime so it can be discovered at runtime

In HIP PR #929, quite a few HIP public APIs were made as inline functions with
hidden visibility. It was necessary to support applications with shared
libraries with GPU kernels launched via hipLaunchKernelGGL(), after HIP runtime
is initialized.

In empirical tests, the implementation has been proved to be a bit too
excessive, especially for hipModuleGetGlobal(). The function is used by another
type of client applications which relies on the existence of this function
within HIP runtime so global symbols from HSA code objects loaded dynamically
at runtime can be retrieved programmtically.

This commit moves hipModuleGetGlobal() back to src/hip_module.cpp, and makes it
visible and not inline, to fulfill requirements for applications
aforementioned. It does not change the behavior of applications depending on
hipLaunchKernelGGL().

* Add HIP_INIT_API into the implementation of hipModuleGetGlobal

Address review comments.

* Fix failing HIP unit tests


[ROCm/hip commit: 4b7177ac42]
2019-03-29 03:45:04 +00:00
Maneesh Gupta f63c4be10e Merge pull request #992 from gargrahul/handle_d2d_memcpy2d
Handle D2D in memcpy2D

[ROCm/hip commit: d0e5fbeb72]
2019-03-28 04:41:36 +00:00
Jeff Daily fb2beb0c88 improve program state commentary
Disambiguate calling many varibles "agent".
More detail in exception message.
Create and discard map placeholders; no need to call std::vector::clear() on map value.


[ROCm/hip commit: 7aada87cbd]
2019-03-27 21:40:27 +00:00
Rahul Garg 0eaa29ad06 Handle D2D in memcpy2D
[ROCm/hip commit: 0c55db8552]
2019-03-28 02:21:45 +05:30
Jeff Daily 21f5e142f5 load program state once per agent
[ROCm/hip commit: c9117de8eb]
2019-03-27 18:19:10 +00:00
Maneesh Gupta 87772a479e Merge pull request #987 from gargrahul/fix_hostmalloc_double_device_map
Avoid double mapping of devices to hostMalloc buffer

[ROCm/hip commit: 08e12e730e]
2019-03-27 05:23:47 +00:00
Maneesh Gupta 464c023bb5 Merge pull request #990 from mhbliao/hliao/master/sw
SWDEV-184380 Fix hcc compilation

[ROCm/hip commit: e66dfa6ea3]
2019-03-27 05:23:26 +00:00
Michael LIAO a9a397bc4b SWDEV-184380 Fix hcc compilation
- `hcc` has no builtin. Need to invoke LLVM intrinsic directly.


[ROCm/hip commit: 939d153c1a]
2019-03-26 15:20:17 -04:00
Rahul Garg d98d5ca12a Let hipHostMalloc always share/map pinned host ptr
[ROCm/hip commit: f0af073793]
2019-03-26 10:19:13 +05:30
Rahul Garg c6ef785464 Avoid double mapping of devices to hostMalloc buffer
[ROCm/hip commit: 5e917d70f3]
2019-03-25 23:07:05 +05:30
Michael LIAO 94b7856a1a [hip] Fix typo in macro hipLaunchKernel
[ROCm/hip commit: b3497824ec]
2019-03-25 12:06:46 -04:00
Maneesh Gupta 82fd86e63f Merge pull request #970 from mangupta/swdev-172995
hipExtMallocWithFlags implementation

[ROCm/hip commit: 30b5c02ec4]
2019-03-25 07:46:53 +00:00
Maneesh Gupta 8c6b5bf266 Merge pull request #962 from gargrahul/add_2d_copy_fallback
Add 2D fallback to use copy kernel

[ROCm/hip commit: eb03d50de9]
2019-03-25 07:46:43 +00:00
Rahul Garg 63e926cec7 2D Fallback needs hcc workweek 19101 or higher
[ROCm/hip commit: 9bbfbceb64]
2019-03-25 12:07:28 +05:30