SWDEV-189500 - [HIP] Have to force async=false for host to device case as well
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_memory.cpp#61 edit
[Reason] To be compatible with CUDA [#1133]
Update HIP code, hipify-clang, tests and docs
[TODO] Add support of the corresponding functions on nvcc fallback path
SWDEV-79445 - HIP generic changes and code clean-up
- Correct elapsed time calculation. Use event start and end.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_event.cpp#13 edit
+ Add option -print-stats-csv to dump statistics to CSV file
+ If -o-dir is specified, CSV file will be dumped there
+ Generate 1 summary file sum_stat.csv in case of multiple sources
Typo introduced here:
commit 67abac1365
Author: Alex Voicu <alexandru.voicu@amd.com>
Date: Mon Jun 24 20:02:09 2019 -0500
Put 3-wide vector types on a ketogenic diet. (#1180)
* Remove flags parameter from hipOccupancyMaxPotentialBlockSize
This commit makes the hipOccupancyMaxPotentialBlockSize method consistent with hcc path and the CUDA API.
SWDEV-194872 - [HIP] CUDA and HCC sync after a DeviceToHost async copy.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_memory.cpp#60 edit
module_api_global relies on a HCC only feature which allows host code
to write to device variables. This feature does not exist in CUDA
or hip-clang, which causes the sample not working in CUDA or hip-clang.
This patch fixes the sample by using standard features of CUDA and
hip-clang. The fixed sample works in HCC, CUDA and hip-clang.
module_api_global relies on a HCC only feature which allows host code
to write to device variables. This feature does not exist in CUDA
or hip-clang, which causes the sample not working in CUDA or hip-clang.
This patch fixes the sample by using standard features of CUDA and
hip-clang. The fixed sample works in HCC, CUDA and hip-clang.
SWDEV-193938 - [HIP] RCCL test fails
Set default stream to null-stream
In hipStreamWaitEvent if event_ is null, get the last queued command instead.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_event.cpp#12 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_event.hpp#4 edit
SWDEV-145570 - Fix device name mismatch.
Not only gfx906 can have device name with +xnack etc.
Other devices e.g. gfx900 could have that too.
Make the previous fix more generic.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_platform.cpp#33 edit
This fixes a bug where GCC++ on Ubuntu 18.04 creates failing executables compared to GCC++ on 16.04 and clang++. While creating function names on Ubuntu 18.04, dl_phdr_info seems to provide a non-zero value for dlpi_addr on initial iteration, and an empty string in dlpi_name. This is causing failure when linking with g++, since the empty string prevents the kernel function from being loaded. Clang++ and GCC on UB16 provide a zero value for dlpi_addr. To fix this, we need to verify both addr and name exists, so that /proc/self/exe can be properly loaded.
* Put 3-wide vector types on a ketogenic diet.
* Remove needless include.
* Do not be narrow-minded.
* Do not be narrow-minded.
* Put the C people on a diet too.
SWDEV-184710 - Support hipLaunchCooperativeKernelMultiDevice()
- Clean-up the loop for the launch on each device
- Add hipExtLaunchMultiKernelMultiDevice()
http://ocltc.amd.com/reviews/r/17573/
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_module.cpp#29 edit
Evgeny Mankov
Jenkins
kpyzhov
Yaxun (Sam) Liu
foreman
mhbliao
Jatin Chaudhary
Maneesh Gupta
Aryan Salmanpour
darkbuck
searlmc1
Jannik
eshcherb
Anusha Godavarthy Surya
wkwchau
Aaron Enye Shi
Alex Voicu