Ryujinx/moltenvk
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

vkCmdFillBuffer: Use the older method to dispatch the compute kernel.

Browse Source 
We're not even running more than one thread, let alone taking advantage
of the new method's automatic threadgroup sizing.
This commit is contained in:
Chip Davis 2018-09-05 13:51:05 -05:00
parent 7fb064fdee
commit eb06c9dd5f
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
MoltenVK/MoltenVK/Commands/MVKCmdTransfer.mm
View File

@ -1012,7 +1012,7 @@ void MVKCmdFillBuffer::encode(MVKCommandEncoder* cmdEncoder) {
[mtlComputeEnc setComputePipelineState: cmdEncoder->getCommandEncodingPool()->getCmdFillBufferMTLComputePipelineState()];
[mtlComputeEnc setBuffer: dstMTLBuff offset: dstMTLBuffOffset atIndex: 0];
[mtlComputeEnc setBytes: &fillInfo length: sizeof(fillInfo) atIndex: 1];
[mtlComputeEnc dispatchThreads: MTLSizeMake(1, 1, 1) threadsPerThreadgroup: MTLSizeMake(1, 1, 1)];
[mtlComputeEnc dispatchThreadgroups: MTLSizeMake(1, 1, 1) threadsPerThreadgroup: MTLSizeMake(1, 1, 1)];
[mtlComputeEnc popDebugGroup];
}