Граф коммитов

170 Коммитов

Автор SHA1 Сообщение Дата
German Andryeyev fd8ec70847 Avoid null stream allocation in hipFree
- Add option to skip stream allocation on access.
- Avoid null stream allocation in ihipFree, so an inactive device
won't be initialized

Change-Id: Id24426640df59a5e7a08b2dd9dcd4d67758b84bf
2020-05-04 17:17:36 -04:00
Christophe Paquot 79834f3d0f Merge "SWDEV-216213 - Delete kernel function_ in hip::Function destructor" into amd-master-next 2020-05-04 11:36:52 -04:00
kjayapra-amd 6bad7ef790 SWDEV-216213 - Delete kernel function_ in hip::Function destructor
Change-Id: I0f6c44927f453ac63b5b84552a5fba07f04a8a8e
2020-05-03 23:08:58 -04:00
Vlad Sytchenko dcd466773b Fix invalid check in hipTexRefGetAddressMode()
This resolves https://github.com/ROCm-Developer-Tools/HIP/issues/2048.

Change-Id: Iba3653b011434a450a5c3ce10cfd4aa39d318527
2020-05-03 14:48:32 -04:00
kjayapra-amd 1ed3af2789 SWDEV-231874 - Make hipMemset patten size 8byte aligned when possible
Change-Id: Ida98bd89212af9b00f3c9c7c5d22ae81f3b5396a
2020-05-01 23:04:04 -04:00
Alex Xie d890d77da4 SWDEV-221166 - Detect support for large bar access through HIP runtime API
Change-Id: Iaa9756c1b5e40c1ab5afb38e44a6699fa5f6c13f
2020-05-01 20:39:52 -04:00
Matt Arsenault 253962c9c4 Don't directly link pthreads
This was already searching for the right threads library, but ignoring
the result.

Change-Id: I10d898245696135c1ef928c7715efce8ec6b939f
2020-05-01 08:28:36 -04:00
Matt Arsenault b69e33038b Don't emit generated hip_prof_str.h to source directory
The build should never touch the source directory, and only write
generated files to the output directory.

Also track the dependency with a custom command and target, as is the
normal method for dependencies on generated files.

Change-Id: I9d835256c643aeef241d26ca05ab390ebba65111
2020-05-01 08:24:56 -04:00
Matt Arsenault 1f5cc41d64 Find python to generate prof API headers.
Don't rely on shell commands. On Ubuntu 20.04, /usr/bin/python no
longer exists, so I'm seeing some failures from somewhere assuming the
path.

The top level CMakeLists also repeats exactly this, which should be
fixed.

Change-Id: I56b26742920f0dc40b363b409892bd41cfa485ef
2020-04-30 18:21:03 -04:00
Michael Hong Bin Liao 61932e1785 Merge "[vdi] Skip null texture object in hipDestroyTextureObject." into amd-master-next 2020-04-30 16:07:58 -04:00
Payam Ghafari b24e5fea52 Merge "renaming vdi to rocclr, part 1" into amd-master-next 2020-04-29 19:42:20 -04:00
Laurent Morichetti f21fd3e22e Don't make a copy of the embedded device binaries
Embedded device binaries should not be cloned if we want the ROCR's
code object URI for this binary to point to the container (host
executable or shared library) instead of a memory location.

Depends-On: I7973bb0243f5a2d1b639b8a88445cfe6af919dd7
Change-Id: Id3afc981e027e5371114d011f2caaa6414f5fc58
2020-04-29 18:40:18 -04:00
Payam 4c3f8221e9 renaming vdi to rocclr, part 1
Change-Id: Idec9be2b6174217abcebaa1776e59168670740b1
2020-04-29 17:54:37 -04:00
Saleel Kudchadker 34b32e35c9 Fix error code when hipEventRecord is not called
Change-Id: I6672d4320280fd87fd2786bc54214d40944da845
2020-04-29 13:13:52 -04:00
Michael LIAO cbe9023639 [vdi] Skip null texture object in hipDestroyTextureObject.
- To match both CUDA and HCC runtime behavior.

Change-Id: I072b006dd554e17f8341f391d33bf6224a125a7e
2020-04-29 12:20:42 -04:00
German Andryeyev deb2c399c9 Clean-up the list of blocking streams
- Insert the stream into the list on the host queue creation,
instead of stream creation

Change-Id: Ib25053019f7df97e5bc786922a6587b9514852d3
2020-04-29 11:37:43 -04:00
Vlad Sytchenko ded92a5d2a Fix order initialization warning
Change-Id: I77ddf1dbec2d11c273163efae6526d3deca92dbf
2020-04-28 20:07:51 -04:00
Vlad Sytchenko cf52b5fca4 Handle copy cases where the uses pases a bad ptr
Change-Id: I4490b8519e4d0dde25b845f9ca7e81c1d80d4f28
2020-04-28 11:04:29 -04:00
Freddy Paul b75b602601 hip-on-vdi: Enable RUNPATH and versioning
1. Enable versioning for HIP libraries
2. Enable RUNPATH on HIP libraries to find ROCm dependencies
3. Control HIP package install using CPACK_INSTALL_PREFIX

Change-Id: I9096da56592e9c78cdba329011f6e1bae99adaa2
2020-04-27 19:48:44 -04:00
kjayapra-amd 1f3789d5a2 SWDEV-232008 - Handle cases where Memcpy size is less than Symbol Size.
Change-Id: I046bc70445dfd4f6fdfd06415f53d1fc06f425b4
2020-04-27 10:47:28 -04:00
root 2689246de6 Merge master into amd-master-next
Change-Id: I3fc1dc0c860d627053537581e75561e8a7efe327
2020-04-26 22:19:37 +00:00
Michael LIAO a0acf7bdaa [vdi] Revise the symbol management.
- As different modules may have symbols with the same name, each symbol
  needs identifying with a pair of the module handle and the symbol
  name.

Change-Id: I85650a787d9a424545154cc40ebd59e706fa358f
2020-04-24 22:28:31 -04:00
German Andryeyev f7f7337bae SWDEV-232918
hipEventRecord is much slower in hipclang/vdi

- Make sure default streams don't sync each other.
- Add null stream into the list of default streams.
- Code clean-up to simplify queue look-up.

Change-Id: I36e1fc8d86a600e3dce806694d95d146ed8afd03
2020-04-24 14:40:43 -04:00
Michael Hong Bin Liao 7519ba64af Merge "[vdi] Fix hang due to recursive locking." into amd-master-next 2020-04-24 10:30:26 -04:00
Michael LIAO 04794c6359 [vdi] Fix hang due to recursive locking.
- Declare recursive lock for global function map.

Change-Id: I3e792e21688c980343c4fd1c61dcad97f7a890cd
2020-04-23 23:32:56 -04:00
kjayapra-amd b443172d12 SWDEV - 231874 - Do Dword aligned memset if the total size aligns.
Change-Id: Id05db4cfd9c43b2cffa3cec8b02f1cd07f340dd6
2020-04-23 21:08:50 -04:00
Maneesh Gupta a0b5dfd625 Merge in the rocclr based hip runtime (#2032)
* Merge master-next changes in master (include vdi development in master branch)
2020-04-23 09:12:06 -07:00
Evgeny Shcherbakov 7d5cd08796 Merge "cleanup" into amd-master-next 2020-04-22 23:28:22 -04:00
Tao Sang 793dbf5bd5 Fix hip_get_devices failure in lammps
Support hipDeviceAttributeIntegrated in hipDeviceGetAttribute() with hip-vdi rt

Change-Id: Ie5ba81222af3554a843c184ae75af7f369a3c24b
2020-04-22 15:16:59 -04:00
Evgeny 6823232b3a cleanup
Change-Id: Ia0ee0e4cab2ee8eaa9931024681d0db5b2802594
2020-04-22 13:42:45 -05:00
Michael LIAO 16d9fe5e37 [vdi] Refactor texture/surface reference support.
Change-Id: I8014d82aae7139ef5f95e4b50c4fc6da200dbc9d
2020-04-21 11:56:48 -04:00
German Andryeyev db70fc66b7 SWDEV-231579 - [hipclang-vdi-rocm][perf]
- HIPPerfDispatchSpeed disparity between HIP/HCC vs HIP/VDI
Insert a wait marker command in the default stream only when
HIP has pending operations on other async streams

Change-Id: I68660a54867fab7571ba57eb1df5feb1bca1c61a
2020-04-21 10:05:26 -04:00
kjayapra-amd 2eb8cc8e90 SWDEV-229480 - Improve error messages in HIP Layer.
Change-Id: I054b979d3aa6cf6ed4ca14a9393bdcba757772ff
2020-04-20 21:23:52 -04:00
Tao Sang da27fd2b09 Solve issues with hip-vdi runtime static lib
1.Combine libamdhip64_static_base.a and libamdvdi_static.a into libamdhip64_static.a.
2.Let hipcc use -use-staticlib to link libamdhip64_static.a.
3.Add some samples for static lib.
4.Fix compiling failure of code object.

Change-Id: Ic8c95228eb139058da8b5d66ba8439486154ca6f
2020-04-20 15:37:02 -04:00
Tao Sang f246761dee Revert "Solve issues with hip-vdi runtime static lib"
This reverts commit 4c2ab3f41e.

Reason for revert: It is causing dkms-no-npi-hipclang broken.
It is top priority to maintain dkms-no-npi-hipclang build, otherwise we lose track of regression analysis.
So revert the change for now and recommit it after fixing it.

 

Change-Id: Ia5136e888baecb6148c6c18eedbf37066fcb1eaa
2020-04-17 10:13:08 -05:00
Tao Sang 4c2ab3f41e Solve issues with hip-vdi runtime static lib
1.Combine libamdhip64_static_base.a and libamdvdi_static.a into libamdhip64_static.a.
2.Let hipcc use -use-staticlib to link libamdhip64_static.a.
3.Add some samples for static lib.
4.Fix compiling failure of code object.
Change-Id: Ia2333622a8d05639b90974c4c5d3d85654ba0138
2020-04-15 13:46:36 -04:00
Karthik Jayaprakash 85fdbb85a2 Merge "SWDEV-227602 - Adding support for hipFuncGetAttribute" into amd-master-next 2020-04-15 12:18:32 -04:00
Vlad Sytchenko f95cdb09c2 Correctly calculate size of the copy region
Since we adjust we adjust the start of the region, amd::BufferRect::end_ is no longer the size, just the offset as to where the region ends.

The actual size of the region is (amd::BufferRect::end_ - amd::BufferRect::start_).

Change-Id: I8425d8bdfb20f485740863813e762e8923d9ee94
2020-04-14 12:25:17 -04:00
Yaxun (Sam) Liu 88304c15e6 Fix MIOpen build failure
This is charrypick of 9ead991784
and https://github.com/ROCm-Developer-Tools/HIP/pull/2009

Fix cmake config file

Removed cmake target files under packaging directory.

Merged cmake config .in files for HIP-Clang and HCC as one.

Use cmake generated target files in both install and packaging.

This makes cmake config file consistent for make install and
make package.

Let device side malloc/free return nullptr and trap

Change-Id: I448f3ea2d4934648089bad371debc203f895cba6
2020-04-13 23:01:31 -04:00
kjayapra-amd a87f517873 SWDEV-227602 - Adding support for hipFuncGetAttribute
Change-Id: I16511274653c8c5521447eb2ed0fc5331dae8cba
2020-04-13 21:23:04 -04:00
Vlad Sytchenko 5d8d752da5 Default stream needs to be initialized
Change-Id: I5cbf9076bffb51c0e99601393645f443c1658e13
2020-04-13 15:32:35 -04:00
Evgeny Shcherbakov 58c805fbfe Merge "fxing C compatibility (amd-master-next)" into amd-master-next 2020-04-10 13:41:47 -04:00
Evgeny b8b5bdf44b fxing C compatibility (amd-master-next)
Change-Id: Ib95b953bb49e0edbe044789b6ff81aaccb87f85f
2020-04-10 00:08:09 -05:00
Vladislav Sytchenko a81e7d63c6 Correctly check max 1D image buffer size
VDI reports the limits in pixels, but user provides the size in bytes.

Make sure both values are in pixels before doing comparisons.

Change-Id: I082c7175c9fa4383e0b0ee38ff8c047c26ff20b4
2020-04-09 21:37:43 -04:00
Vladislav Sytchenko 7592fd18d7 Fix Windows build
Change-Id: I8e219f8200875e3c46c1f54348317ba7ad8ae8ba
2020-04-09 20:00:29 -04:00
Christophe Paquot 6584470377 Merge "Remove a map lookup whenever we were getting the default stream" into amd-master-next 2020-04-09 18:35:46 -04:00
Vladislav Sytchenko 2613b22e8a Fix warnings when compiling with clang
The following warnings are addressed:

comparison of different enumeration types in switch statement

Change-Id: I6cb3948aeab7287851c57ecc1d4b3a439ab14ec6
2020-04-09 17:07:48 -04:00
Christophe Paquot 4d9c540f4f Remove a map lookup whenever we were getting the default stream
Change-Id: I64b6d1deea41d81e94a58a83de287e78923656b3
2020-04-09 12:44:21 -07:00
Evgeny 1e4e4cb229 enabling hipPrintString (to master-next)
Change-Id: I28859f3dbe5b867a858ca1d76c93e6fab6a68d1f
2020-04-09 09:57:27 -05:00
Saleel Kudchadker f99db12c82 Merge "Revert "Wake up commandQueue before returning"" into amd-master-next 2020-04-06 18:52:29 -04:00