ECR #333756 - HSA Finalizer: Make sure size of kernarg segment, alignment of kernarg, private and group segments are multiple of 16. Update ORCA runtime assert. [ OpenCL integration of CL 1151953]
Change by Nikolay Haustov
Testing: http://ocltc:8111/viewModification.html?modId=51851&personal=true&init=1&tab=vcsModificationBuilds
Also fix uncovered problem in test.
Testing: pre-checkin
Reviewed by: German
Affected files ...
... //depot/stg/opencl/drivers/opencl/compiler/lib/backends/gpu/sc/HSAIL/hsail-fin/HSAILFinalizer.cpp#16 integrate
... //depot/stg/opencl/drivers/opencl/compiler/lib/backends/gpu/sc/HSAIL/tests/src/finalizer/features/structural_analysis/short_circuit/short_circuit06.hsail#4 integrate
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.cpp#284 edit
EPR #403782 - IOMMU2/SVM
- Disable DX interop on SVM. This is a feature for SVM and may need more work.
ReviewBoardURL = http://ocltc.amd.com/reviews/r/7555/diff/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.cpp#310 edit
ECR #304775 - Mipmaps support
- Following CL#1151650. Change the comparison condition to 1.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuresource.cpp#216 edit
ECR #304775 - Mipmaps support
- Enable miplevel flag even for the first mip level when runtime creates a view. Otherwise GSL may change the pitch alignment for the created view.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuresource.cpp#215 edit
EPR #397491 - Disable generic address space only on 32 bit Linux
Reviewed by: German
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gslbe/src/rt/GSLDevice.cpp#116 edit
ECR #333753 - Compiler Lib/RT: libutils.h usage removal due to non-API interface
Utils are to be used only by Compiler Lib itself.
Testing: pre checkin
Reviewers: German Andryeyev, Brian Sumner, Yaxun Liu
Affected files ...
... //depot/stg/opencl/drivers/opencl/compiler/lib/utils/v0_8/libUtils.cpp#8 edit
... //depot/stg/opencl/drivers/opencl/compiler/lib/utils/v0_8/libUtils.h#17 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/device.cpp#178 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuprogram.cpp#193 edit
EPR #419351 - clEnqueueNDRange crash if one doesn't create a device queue and use device enqueue in the kernel
- add a check for defQueue is NULL in case the app didn't create one.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#510 edit
EPR #419065 - [CQE OCL][ISV][QR][G] FAHBenchmark application is crashing.
Two issues:
1. Remove clearing of profileEnabled_ since it may cause incorrect kernel execution time measurement.
2. Blit kernels causes assertion in getWavesPerSH since they do not have wave limiters. Remove the assert. If a kernel has no wave limiter, returns 0 in getWavesPerSH.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#361 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuwavelimiter.cpp#5 edit
ECR #304775 - Wave limiter: Fix bug in adaptation.
Dumped waves/simd value is incorrect.
Should exit adptation only after the changed waves/simd value is applied.
Added wave limiter manager to handle situation that one kernel is enqueued to more than one queues. Create wave limiter for each virtual device.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/device.hpp#245 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.cpp#283 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.hpp#109 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#360 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuwavelimiter.cpp#4 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuwavelimiter.hpp#3 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/command.cpp#70 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/command.hpp#76 edit
EPR #410821 - Reduced the maxMemAlloc and global memory size on APUs to 75% of uncachedRemoteRAM.
http://ocltc.amd.com/reviews/r/7425/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#509 edit
ECR #304775 - Wave limiter: Fix crash in CompuBenchCL video composition due to profiling data not collected correctly.
Gpuvirtual.cpp only collects profiling data when all events have profiling enabled. Fixed it by adding a member to indicate at least one event has profiling enabled and collect profiling data.
Improved adptation by changing waves/simd only when the last change has been enforced. Also detecting discontinuities in measured data and discard them.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#359 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.hpp#129 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuwavelimiter.cpp#3 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuwavelimiter.hpp#2 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.hpp#230 edit
ECR #304775 - Fix offline compilation for Hawaii with -cl-fp32-correctly-rounded-divide-sqrt flag
- check for cl-fp32-correctly-rounded-divide-sqrt support uses device info, but device info was never fully updated for offline devices. This change will update device info structure for offline devices as well.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#508 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.hpp#143 edit
EPR #403782 - IOMMU2/SVM on CZ Win10, GL Interop changes for SVM.
- Pass flag to GL to disable forceRemoteAllocation during GLDissociate
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gslbe/src/rt/GSLDeviceGL.cpp#23 edit
ECR #304775 - Wave limiter: fix bug about some variables not initialized before being used.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuwavelimiter.cpp#2 edit
EPR #416903 - reduce maximum work load time to avoid BSOD/TDR when running the FlopsCL benchmarks on Kaveri
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.cpp#309 edit
ECR #399840 - OpenCL Runtime HW Debug support development - resolve the TDR issues on Kaveri.
1. update the resource descriptors in the runtime trap handler to match those in the HSA HW debug implementation
2. force to use SDMA for device memory map function, which is called when using clHwDbgSetGlobalMemoryAMD() and clEnqueueMapImage() functions, for HW debug
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuresource.cpp#214 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gputrap.hpp#3 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#358 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gslbe/src/rt/GSLDevice.cpp#115 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gslbe/src/rt/GSLDevice.h#47 edit
EPR #412821 - Disable ThreadTrace for SI for now.
There's hang on some SI (Verde). So only enable ThreadTrace for CI+ for now.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.cpp#308 edit
ECR #333753 - ORCA RT: aclCompilerInit usage fix
Passing pointers on CRT memory management functions to dynamically loaded library is unsafe and may lead to memory corruption. But actually the pointers aren't actually passed.
They are cut off by the struct size: sizeof(aclCompilerOptions_0_8) and hence never being called.
P.S.
If it was aclCompilerOptions_0_8_1 the pointers would be taken into account and a heap corruption would occur. So the change intended to liquidate possible misunderstanding of malloc & free functions usage, and doesn't change the behaviour of aclCompilerInit function at all.
Testing: pre check-in
Reviewers: German Andreev, Stanislav Mekhanoshin
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/cpu/cpudevice.cpp#271 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#507 edit
EPR #403782 - IOMMU2/SVM
- Handle case of only one DMA engine available, for example with SVM.
ReviewBoardURL = http://ocltc.amd.com/reviews/r/7284/diff/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#506 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.hpp#142 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#357 edit
EPR #403782 - IOMMU2/SVM
- Enable SCOption_R1200_ENABLE_XNACK whenever IOMMUv2 is supported.
- Add "-sc-xnack-iommu" option for compile and link and pass this to SCWrapper in the options string.
ReviewBoardURL = http://ocltc.amd.com/reviews/r/7266/diff/
Affected files ...
... //depot/stg/opencl/drivers/opencl/compiler/lib/backends/gpu/scwrapper/SI/scStateSI.cpp#30 edit
... //depot/stg/opencl/drivers/opencl/compiler/lib/utils/OPTIONS.def#122 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.cpp#282 edit
EPR #394115 - [OpenCL][Arctic Islands] adding aiPlus_ variable for AI and post AI features
ReviewBoardURL = http://ocltc.amd.com/reviews/r/7167/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.cpp#307 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.hpp#90 edit
ECR #399840 - OpenCL Runtime HW Debug support development - add support to the VI asics & support the use case of debug registeration in a pre-dispatch callback function
** Cross branch check-in with CL1131894
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_debugger_amd.cpp#7 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudebugmanager.cpp#9 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#501 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.hpp#139 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gputrap.hpp#2 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#355 edit
ECR #304775 - Mipmaps support
- Create views for the specified mip level in the transfer operations
- OCL requires just one mip level transfer. Thus we could keep the original blit kernels and just create a view for the specified mip level.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_memobj.cpp#75 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#500 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/memory.cpp#119 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/memory.hpp#90 edit
EPR #394115 - Adding the environment variable "GPU_SELECT_COMPUTE_RINGS_ID" to select a specific compute queue for OCL submission. This EV was requested from KMD team for testing the CWSR demo on CZ.
ReviewBoardURL = http://ocltc.amd.com/reviews/r/7082/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#354 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gslbe/src/rt/GSLDevice.cpp#111 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gslbe/src/rt/GSLDevice.h#44 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.hpp#226 edit
EPR #414133 - HSA HLC: fixed device enqueue code for no-capture case
A dead store to a context structure may be generated by clang, while our code was only
expecting a bitcast.
Affected files ...
... //depot/stg/opencl/drivers/opencl/compiler/llvm36/lib/Transforms/Scalar/AMDLowerEnqueueKernel.cpp#6 edit
EPR #415638 - Improve APU performance
- Force remote allocation of local and persistent memory to Remote from RemoteUSWC:
- Use gpu copy for remote/pinned image/buffer.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpublit.cpp#114 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuresource.cpp#211 edit
EPR #413960 - Platform Atomics on CZ.
1. Since CMMQS can support 4G atomics range on CZ, the atomics range on KV is reverted to 4G.
2. On KV, the platform atomics is implemented by using ATC/IOMMU. But on CZ, it should be implemented without using ATC/IOMMU due to HW changes. So, _svmAtomicsUsingIOMMU is introduced to differentiate the two implementations.
3. _priSupport was renamed _svmAtomics due to different implementation on KV vs CZ.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.cpp#304 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gslbe/src/include/cal/cal.h#33 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gslbe/src/rt/GSLDevice.cpp#109 edit
ECR #304775 - In addition to CL#1126564
- Keep just tile index field from the parent
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuresource.cpp#209 edit
ECR #304775 - Use the tiling index from the parent object instead of forcing it to 0.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuresource.cpp#208 edit
ECR #304775 - Enable TC compatible htile mode by default.
- The feature is available on VI+, but interop just requires different tiling mode.
- In the future we may need another way to control this feature for interop.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gslbe/src/rt/GSLDevice.cpp#108 edit
EPR #405824 - Back out changelist 1125022 because of failures in Linux.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#497 edit
foreman