SWDEV-2 - Change OpenCL version number from 2786 to 2787.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2534 edit
SWDEV-165045 - Linux-Pro System Hard-Hang with Call Trace and VM Faults are observed while running ABAQUS-2018-OpenCL Sample on VEGA GPU
"-g" option caused shader compiler generated isa code which caused the VM fault. The root cause is still unknown. "-g" is not a feature for end user. So remove it.
Remove -g options from application. The affected API includes build, link and compile CL C kernel code.
Update ocltst compiler test to remove one test which uses "-g"
AMD_OCL_BUILD_OPTIONS_APPEND can be used to force "-g" option
Test:
Run the Abaqus as specfied in the JIRA twice. No VM fault.
In a simple OpenCL test application, driver can handle build option "-g -cl-mad-enable -g -D DDBUILD -D ok-g -g" correctly.
Teamcity test.
http://ocltc.amd.com:8111/viewModification.html?modId=112334&personal=true&buildTypeId=&tab=vcsModificationBuilds&show_all_builds=true
No new failure.
ReviewBoard: http://ocltc.amd.com/reviews/r/16255/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/platform/program.cpp#98 edit
... //depot/stg/opencl/drivers/opencl/tests/ocltst/module/compiler/OCLSeparateCompile.cpp#17 edit
[ROCm/clr commit: 463b3cd876]
SWDEV-165045 - Linux-Pro System Hard-Hang with Call Trace and VM Faults are observed while running ABAQUS-2018-OpenCL Sample on VEGA GPU
"-g" option caused shader compiler generated isa code which caused the VM fault. The root cause is still unknown. "-g" is not a feature for end user. So remove it.
Remove -g options from application. The affected API includes build, link and compile CL C kernel code.
Update ocltst compiler test to remove one test which uses "-g"
AMD_OCL_BUILD_OPTIONS_APPEND can be used to force "-g" option
Test:
Run the Abaqus as specfied in the JIRA twice. No VM fault.
In a simple OpenCL test application, driver can handle build option "-g -cl-mad-enable -g -D DDBUILD -D ok-g -g" correctly.
Teamcity test.
http://ocltc.amd.com:8111/viewModification.html?modId=112334&personal=true&buildTypeId=&tab=vcsModificationBuilds&show_all_builds=true
No new failure.
ReviewBoard: http://ocltc.amd.com/reviews/r/16255/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/platform/program.cpp#98 edit
... //depot/stg/opencl/drivers/opencl/tests/ocltst/module/compiler/OCLSeparateCompile.cpp#17 edit
SWDEV-79445 - OCL generic changes and code clean-up
- Modify CMakeLists.txt to switch to different library interface to match the COMgr requirement
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/CMakeLists.txt#16 edit
[ROCm/clr commit: 5b7a2bc281]
SWDEV-79445 - OCL generic changes and code clean-up
- Modify CMakeLists.txt to switch to different library interface to match the COMgr requirement
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/CMakeLists.txt#16 edit
+ cuSPARSE is supported up to CUDA 10.0
+ cuSPARSE_08 test (CUDA 10.0)
+ update CUSPARSE_API_supported_by_HIP.md
+ lit: add a rule for CUDA 10.0 tests excluding
[ROCm/clr commit: 424e4c8229]
+ cuSPARSE is supported up to CUDA 10.0
+ cuSPARSE_08 test (CUDA 10.0)
+ update CUSPARSE_API_supported_by_HIP.md
+ lit: add a rule for CUDA 10.0 tests excluding
+ cuSPARSE is supported up to CUDA 10.0
+ cuSPARSE_08 test (CUDA 10.0)
+ update CUSPARSE_API_supported_by_HIP.md
+ lit: add a rule for CUDA 10.0 tests excluding
[ROCm/hip commit: 116b9191f7]
+ cuSPARSE is supported up to CUDA 10.0
+ cuSPARSE_08 test (CUDA 10.0)
+ update CUSPARSE_API_supported_by_HIP.md
+ lit: add a rule for CUDA 10.0 tests excluding
This is to handle EINTR/EGAIN properly (issue #137), and also
make the code consistent with the rest.
Unfortunately posix_fallocate and mmap do not follow the classic
return code/errno pattern, so we need to write wrappers around those
functions.
[ROCm/rccl commit: eed8218e17]
This is to handle EINTR/EGAIN properly (issue #137), and also
make the code consistent with the rest.
Unfortunately posix_fallocate and mmap do not follow the classic
return code/errno pattern, so we need to write wrappers around those
functions.
SYSCHECKVAL was not retrying when a retry was needed. Since not all
calls are inside a loop, that means we could silently miss an
EINTR/EAGAIN return code.
Also rework the socket connection code and improve error reporting.
[ROCm/rccl commit: 302d538b73]
SYSCHECKVAL was not retrying when a retry was needed. Since not all
calls are inside a loop, that means we could silently miss an
EINTR/EAGAIN return code.
Also rework the socket connection code and improve error reporting.
foreman
Evgeny
Maneesh Gupta
Evgeny Mankov
Evgeny Mankov
David Addison
Jenkins
Sylvain Jeaugey
Siu Chi Chan