SWDEV-151739 - [CQE OCL][DTB][Perf][QR][DTB-BLOCKER][VEGA10] Upto 18% performance drop observed while running Video Composition test sub test of Compubench due to faulty CL#1544622
- Implement customized TS tracking for managed buffers. The common TS tracking mechanism saves the event of the last command, assuming SDMA and compute operations occur in order, but for managed buffers it's not the case. Also managed buffer doesn't have to validate TS for the parent resource.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palblit.cpp#21 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palconstbuf.cpp#11 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palconstbuf.hpp#9 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palmemory.hpp#6 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palresource.hpp#22 edit
[ROCm/clr commit: 79ba5904dc]
SWDEV-151739 - [CQE OCL][DTB][Perf][QR][DTB-BLOCKER][VEGA10] Upto 18% performance drop observed while running Video Composition test sub test of Compubench due to faulty CL#1544622
- Implement customized TS tracking for managed buffers. The common TS tracking mechanism saves the event of the last command, assuming SDMA and compute operations occur in order, but for managed buffers it's not the case. Also managed buffer doesn't have to validate TS for the parent resource.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palblit.cpp#21 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palconstbuf.cpp#11 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palconstbuf.hpp#9 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palmemory.hpp#6 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palresource.hpp#22 edit
SWDEV-2 - Change OpenCL version number from 2639 to 2640.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2387 edit
[ROCm/clr commit: d6618edf9f]
SWDEV-2 - Change OpenCL version number from 2639 to 2640.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2387 edit
Use async. signal handler to satisfy dependencies for SDMA blits.
Change-Id: Ifa8d3ee6810509f400a568ca2387ac6ab3ab7c36
[ROCm/ROCR-Runtime commit: 7cd6e366ed]
SWDEV-151852 - [CQE OCL][DTB-BLOCKER][VEGA10][WIN][LNX][QR] Observed crash while running Multiple device context test due to FAULTY CL#1544622
- Add a check for nullptr in case if virtual device object was destroyed before memory allocation
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palconstbuf.cpp#10 edit
[ROCm/clr commit: 3416ad385d]
SWDEV-151852 - [CQE OCL][DTB-BLOCKER][VEGA10][WIN][LNX][QR] Observed crash while running Multiple device context test due to FAULTY CL#1544622
- Add a check for nullptr in case if virtual device object was destroyed before memory allocation
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palconstbuf.cpp#10 edit
SWDEV-2 - Change OpenCL version number from 2638 to 2639.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2386 edit
[ROCm/clr commit: b785abe728]
SWDEV-2 - Change OpenCL version number from 2638 to 2639.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2386 edit
SWDEV-79445 - OCL generic changes and code clean-up
- Report an error if persistent memory allocation failed. The original logic was switching to local memory, but SSG can't work with it. The logic should work properly under Linux, but in Windows OS never fails persistent allocation.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#86 edit
[ROCm/clr commit: 1c91644f1b]
SWDEV-79445 - OCL generic changes and code clean-up
- Report an error if persistent memory allocation failed. The original logic was switching to local memory, but SSG can't work with it. The logic should work properly under Linux, but in Windows OS never fails persistent allocation.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#86 edit
SWDEV-145570 - [HIP] Get hipCtx_simple to pass
Implemented hipCtxGetDevice
hipCtxCreate must push the created context onto the context stack
hipCtxDestroy must check if the top of the stack is the context being destroy
and not just pop the top of the stack w/o checking.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_context.cpp#8 edit
[ROCm/clr commit: 894f01c206]
SWDEV-145570 - [HIP] Get hipCtx_simple to pass
Implemented hipCtxGetDevice
hipCtxCreate must push the created context onto the context stack
hipCtxDestroy must check if the top of the stack is the context being destroy
and not just pop the top of the stack w/o checking.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_context.cpp#8 edit
SWDEV-145570 - [HIP] Get hipCtx_simple to pass
Implemented hipCtxGetDevice
hipCtxCreate must push the created context onto the context stack
hipCtxDestroy must check if the top of the stack is the context being destroy
and not just pop the top of the stack w/o checking.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_context.cpp#8 edit
This change fixes the following bugs that were discovered while debuggnig TF unit test failures (cwise_ops_test)
1. __hisinf and __hisnan routines
Both had incorrect implementations.
2. abs
A "long long" (64bit int) version was missing, resulting in the 32bit version being used for 64bit ints (which resulted in incorrect results, when the value passed in was outside the 32bit int range)
3. lgamma
We seemed to have a custom version for the 'double' datatype (which was giving incorrect results). Replaced it with a call to the 'double' version of the underlying 'hc::precision_math::lgamma'
[ROCm/clr commit: 65a90c55e7]
This change fixes the following bugs that were discovered while debuggnig TF unit test failures (cwise_ops_test)
1. __hisinf and __hisnan routines
Both had incorrect implementations.
2. abs
A "long long" (64bit int) version was missing, resulting in the 32bit version being used for 64bit ints (which resulted in incorrect results, when the value passed in was outside the 32bit int range)
3. lgamma
We seemed to have a custom version for the 'double' datatype (which was giving incorrect results). Replaced it with a call to the 'double' version of the underlying 'hc::precision_math::lgamma'
This change fixes the following bugs that were discovered while debuggnig TF unit test failures (cwise_ops_test)
1. __hisinf and __hisnan routines
Both had incorrect implementations.
2. abs
A "long long" (64bit int) version was missing, resulting in the 32bit version being used for 64bit ints (which resulted in incorrect results, when the value passed in was outside the 32bit int range)
3. lgamma
We seemed to have a custom version for the 'double' datatype (which was giving incorrect results). Replaced it with a call to the 'double' version of the underlying 'hc::precision_math::lgamma'
SWDEV-2 - Change OpenCL version number from 2637 to 2638.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2385 edit
[ROCm/clr commit: 03b72c3007]
SWDEV-2 - Change OpenCL version number from 2637 to 2638.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2385 edit
SWDEV-79445 - OCL generic changes and code clean-up
- Switch scheduler parameters to the managed buffer allocation
- Use 8 bytes for the mem clear pattern in the program loader
- Use local memory for the hsa queue data
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palprogram.cpp#63 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#93 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.hpp#49 edit
[ROCm/clr commit: 30088df072]
SWDEV-79445 - OCL generic changes and code clean-up
- Switch scheduler parameters to the managed buffer allocation
- Use 8 bytes for the mem clear pattern in the program loader
- Use local memory for the hsa queue data
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palprogram.cpp#63 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#93 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.hpp#49 edit
foreman
Sean Keely
Evgeny
Deven Desai
Rahul Garg
Paul
Evgeny Mankov
srinivas Charupally
Aaron Enye Shi