SWDEV-193423 - HIP/VDI - Support for lazy hsa queue creation
- Add queue pool support for HSA HW queues. GPU_MAX_HW_QUEUES controls the pool size. The current default value is 4 (the number of active pipes on GPU).
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.cpp#132 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.hpp#38 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.cpp#81 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.hpp#24 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.hpp#314 edit
SWDEV-79445 - OCL generic changes and code clean-up
- Make sure PAL_DISABLE_SDMA is fully functional. CP DMA is used for buffer transfers currently and kernels for images and buffer rect copies.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palblit.cpp#32 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#150 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#92 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.hpp#25 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#142 edit
SWDEV-144570 - Use Comgr feature flag around COMGR related API usage.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/devprogram.cpp#52 edit
SWDEV-189140 - Add P2P support in PAL path
- Don't add host mem direct access flag for P2P buffers
- Avoid sync cocherency logic for P2P buffers
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palmemory.cpp#27 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palmemory.hpp#13 edit
SWDEV-189787 - [NWNIT] WX9100 crashing under heavy load in Agisoft
- Don't report SPIR extension, since HSAIL doesn't support it
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#90 edit
SWDEV-196199 - [Navi10] Corruption is observed when running Premier Pro Benchmarks using Adobe Premier Pro 2019
- In addition to CL#1968527. Add an extra transfer in the case the app reads image without kernel execution
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#139 edit
SWDEV-174198 - Put back the fix of segfault running integer_ops.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.cpp#80 edit
SWDEV-174198 - Back out changelist 1968648.
Since it causes soft-hang in multiple_device_context test.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.cpp#79 edit
SWDEV-195023 - [CQE OCL][Navi10][RESOLVE] corruption seen in thumbnail for mxf clip after enabling temporal denoiser in Davinci resolve app
- Add a workaround for missing custom pitch in gfx10 HW. It can be disabled with GPU_IMAGE_BUFFER_WAR=0. Workaround implements double copy with an image without pitch.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palmemory.cpp#26 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palmemory.hpp#12 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#89 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.hpp#24 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#138 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.hpp#62 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.hpp#313 edit
SWDEV-192353 - 12% Performance drop observed while running ROC_OCL_Perf_CompubenchCL_GPU_W64 only on Vega10 with Win7
1. Reduce CmdAllocator sub-allocation size to 4KB.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#137 edit
SWDEV-161424 - Fix broken option handling in Comgr path
Introduces another potential bug if any options in Options::llvmOptions contain spaces, but this existed before the switch to Comgr.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/devprogram.cpp#50 edit
SWDEV-193973 - [OpenCL][NV21] Add support for Navi21
1. Enable OCLOfflineCompilation for gfx1010/1011/1012.
2. Skip gfx1030 as there is no support for this target in the device libraries for now.
ReviewBoardURL = http://ocltc.amd.com/reviews/r/17611/diff/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/devkernel.cpp#24 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palkernel.cpp#80 edit
... //depot/stg/opencl/drivers/opencl/tests/ocltst/module/runtime/OCLOfflineCompilation.cpp#21 edit
SWDEV-190204 - [CQE OCL][NAVI 14] "Unsupported device(gfx1012) for the test! " for OCLPerfCounters which is subtest of oclruntime.
Fix the ID converstion array.
Tests:
1. run the test in Navi 10.
2. http://ocltc.amd.com:8111/viewModification.html?modId=122114&personal=true&tab=vcsModificationBuilds
3. The CQE has confirmed the idea fixed the issue in Navi14
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palcounters.cpp#22 edit
... //depot/stg/opencl/drivers/opencl/tests/ocltst/module/runtime/OCLPerfCounters.cpp#48 edit
SWDEV-144570 - Adding entries on to P2P Access devices in RocM, to create deviceMemories_ for P2P devices too.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.cpp#129 edit
SWDEV-79445 - OCL generic changes and code clean-up
Align scratch buffer size to match the HW requirement. HSAIL path has an alignment in SC, but LC doesn't have any
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palkernel.cpp#79 edit
SWDEV-191674 - Handling cases where p2p memcpy is initiated from device 1. (No Large Bar/P2P staging).
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#136 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.cpp#77 edit
SWDEV-180872 - Runtime support changes for Cooperative Group Features
- Keep this feature for Linux only. Windows doesn't enable GWS by default
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#85 edit
SWDEV-180872 - Runtime support changes for Cooperative Group Features
- Initial implementation of the core functionality. Disabled by default. Use GPU_ENABLE_COOP_GROUPS=1 to enable the feature.
- Runtime uses device queue for cooperative executions with a synchronization on the launched queue.
- The current implementation is pure runtime change and it can work if only one app uses this feature. No ROCr/KFD support was added or tested
- Only inline assembler was tested
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_device.cpp#20 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_device_runtime.cpp#15 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_hcc.def.in#15 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_hcc.map.in#17 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_module.cpp#28 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_platform.cpp#32 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/device.hpp#338 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#606 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.hpp#171 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palblit.cpp#31 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palblit.hpp#9 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#142 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.hpp#39 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palschedcl.cpp#6 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#135 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.hpp#61 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocblit.cpp#32 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocblit.hpp#12 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.cpp#127 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.hpp#37 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocschedcl.cpp#3 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.cpp#75 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.hpp#23 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/command.cpp#94 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/command.hpp#92 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.hpp#311 edit
SWDEV-187399 - Linux-Pro Pyro_Explosion.hip Image difference is observed specific to W9000 ASIC in OCL mode
The issue disappeared when we invalidate L1 cache after each dispatch.
This change also fix ocltst runtime OCLMultiQueue test in Tahiti Brahma driver.
Tests:
http://ocltc.amd.com:8111/viewModification.html?modId=121519&personal=true&tab=vcsModificationBuilds
Test in the Houdini app. The symptom disappeared.
This change also fix ocltst runtime OCLMultiQueue test in Tahiti Brahma driver.
ReviewBoard: http://ocltc.amd.com/reviews/r/17509/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.cpp#337 edit
SWDEV-79445 - OCL generic changes and code clean-up
- Force LC for HIP, since it doesn't support HSAIL path
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#83 edit
SWDEV-79445 - OCL generic changes and code clean-up
Optimize scratch buffer calculation in the preparation for coop group launch, since the current limit affects max waves calculation:
- Switch to 32 waves per CU as the max possible limit
- Use vgprs count for the waves limit calculation to avoid unconditional possible max
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#141 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.hpp#38 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#82 edit
SWDEV-79445 - OCL generic changes and code clean-up
- Use PAL shader core properties instead of the local device info
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldefs.hpp#53 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#140 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#134 edit
SWDEV-132899 - [OCL][GFX10] increase the numScratchWavesPerCu in Wave32 mode and use the actual num of CUs not the total num of WGPs when calculating the scratch buffer size
ReviewBoardURL = http://ocltc.amd.com/reviews/r/17474/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#139 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#81 edit
foreman