P4 to Git Change 1584092 by zukhan@zukhan-lc-win10 on 2018/07/23 15:24:19
SWDEV-155860 - [OCL][LC] 62 Failures running on OCL/LC Windows drivers
- Conformance test basic->vload_private was failing due to alignment assert check on akc->workitem_private_segment_byte_size.
- Determined needed to add alignment, so added alignment into the header file, and removed assert check (as alignment was done afterwards).
ReviewBoardURL = http://ocltc.amd.com/reviews/r/15451/diff/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palkernel.cpp#57 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palkernel.hpp#18 edit
[ROCm/clr commit: 12a063779d]
此提交包含在:
1 父節點
44dca0ab3b
當前提交
6a30645259
2 file changed
+1
-2
No files matched your search
@@ -420,7 +420,6 @@ bool HSAILKernel::aqlCreateHWInfo(amd::hsa::loader::Symbol* sym) {
|
|||||||
return false;
|
return false;
|
||||||
}
|
}
|
||||||
|
|
||||||
assert((akc->workitem_private_segment_byte_size & 3) == 0 && "Scratch must be DWORD aligned");
|
|
||||||
workGroupInfo_.scratchRegs_ =
|
workGroupInfo_.scratchRegs_ =
|
||||||
amd::alignUp(akc->workitem_private_segment_byte_size, 16) / sizeof(uint);
|
amd::alignUp(akc->workitem_private_segment_byte_size, 16) / sizeof(uint);
|
||||||
workGroupInfo_.privateMemSize_ = akc->workitem_private_segment_byte_size;
|
workGroupInfo_.privateMemSize_ = akc->workitem_private_segment_byte_size;
|
||||||
|
|||||||
@@ -161,7 +161,7 @@ class HSAILKernel : public device::Kernel {
|
|||||||
size_t argsBufferSize() const { return cpuAqlCode_->kernarg_segment_byte_size; }
|
size_t argsBufferSize() const { return cpuAqlCode_->kernarg_segment_byte_size; }
|
||||||
|
|
||||||
//! Returns spill reg size per workitem
|
//! Returns spill reg size per workitem
|
||||||
int spillSegSize() const { return cpuAqlCode_->workitem_private_segment_byte_size; }
|
int spillSegSize() const { return amd::alignUp(cpuAqlCode_->workitem_private_segment_byte_size, sizeof(uint32_t)); }
|
||||||
|
|
||||||
//! Returns TRUE if kernel uses dynamic parallelism
|
//! Returns TRUE if kernel uses dynamic parallelism
|
||||||
bool dynamicParallelism() const { return (flags_.dynamicParallelism_) ? true : false; }
|
bool dynamicParallelism() const { return (flags_.dynamicParallelism_) ? true : false; }
|
||||||
|
|||||||
Reference in new issue
封鎖使用者