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

Fix wrong offset for vkCmdFillBuffer on VK_WHOLE_SIZE.

Browse Source 
A double offset would be applied, causing the word count to wrap around
and trigger an assertion and GPU lockup.
This commit is contained in:
Hans-Kristian Arntzen 2019-03-27 19:50:05 +01:00
parent ccdfa67a7b
commit bfd4d182cd
1 changed files with 4 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
MoltenVK/MoltenVK/Commands/MVKCmdTransfer.mm
View File

@ -1092,8 +1092,10 @@ void MVKCmdFillBuffer::setContent(VkBuffer dstBuffer,
void MVKCmdFillBuffer::encode(MVKCommandEncoder* cmdEncoder) {
id<MTLBuffer> dstMTLBuff = _dstBuffer->getMTLBuffer();
VkDeviceSize dstMTLBuffOffset = _dstBuffer->getMTLBufferOffset();
VkDeviceSize byteCnt = (_size == VK_WHOLE_SIZE) ? (_dstBuffer->getByteCount() - (dstMTLBuffOffset + _dstOffset)) : _size;
VkDeviceSize wordCnt = byteCnt >> 2;
VkDeviceSize byteCnt = (_size == VK_WHOLE_SIZE) ? (_dstBuffer->getByteCount() - _dstOffset) : _size;
// Round up in case of VK_WHOLE_SIZE on a buffer size which is not aligned to 4 bytes.
VkDeviceSize wordCnt = (byteCnt + 3) >> 2;
MVKAssert(mvkFits<uint32_t>(wordCnt),
"Buffer fill size must fit into a 32-bit unsigned integer.");