fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-05 16:08:04 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

swr: [rasterizer core] Affinitize thread scratch space to numa node of worker

Browse source 
Acked-by: Brian Paul <brianp@vmware.com>
This commit is contained in:
Tim Rowley 2016-03-24 16:20:02 -06:00
parent f89f6d562a
commit c25244f2f7
3 changed files with 16 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

16
src/gallium/drivers/swr/rasterizer/core/api.cpp
View file

@ -93,8 +93,16 @@ HANDLE SwrCreateContext(
///@note We could lazily allocate this but its rather small amount of memory.
for (uint32_t i = 0; i < pContext->NumWorkerThreads; ++i)
{
///@todo Use numa API for allocations using numa information from thread data (if exists).
pContext->pScratch[i] = (uint8_t*)_aligned_malloc((32 * 1024), KNOB_SIMD_WIDTH * 4);
#if defined(_WIN32)
uint32_t numaNode = pContext->threadPool.pThreadData ?
pContext->threadPool.pThreadData[i].numaId : 0;
pContext->pScratch[i] = (uint8_t*)VirtualAllocExNuma(
GetCurrentProcess(), nullptr, 32 * sizeof(KILOBYTE),
MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE,
numaNode);
#else
pContext->pScratch[i] = (uint8_t*)_aligned_malloc(32 * sizeof(KILOBYTE), KNOB_SIMD_WIDTH * 4);
#endif
}
// State setup AFTER context is fully initialized
@ -138,7 +146,11 @@ void SwrDestroyContext(HANDLE hContext)
// Free scratch space.
for (uint32_t i = 0; i < pContext->NumWorkerThreads; ++i)
{
#if defined(_WIN32)
VirtualFree(pContext->pScratch[i], 0, MEM_RELEASE);
#else
_aligned_free(pContext->pScratch[i]);
#endif
}
delete(pContext->pHotTileMgr);

2
src/gallium/drivers/swr/rasterizer/core/arena.h
View file

@ -209,7 +209,7 @@ struct CachingAllocatorT : DefaultAllocator
};
typedef CachingAllocatorT<> CachingAllocator;
template<typename T = DefaultAllocator, size_t BlockSizeT = (128 * 1024)>
template<typename T = DefaultAllocator, size_t BlockSizeT = 128 * sizeof(KILOBYTE)>
class TArena
{
public:

2
src/gallium/drivers/swr/rasterizer/core/backend.cpp
View file

@ -83,7 +83,7 @@ void ProcessComputeBE(DRAW_CONTEXT* pDC, uint32_t workerId, uint32_t threadGroup
if (pDC->pSpillFill[workerId] == nullptr)
{
///@todo Add state which indicates the spill fill size.
pDC->pSpillFill[workerId] = (uint8_t*)pDC->pArena->AllocAlignedSync(4096 * 1024, sizeof(float) * 8);
pDC->pSpillFill[workerId] = (uint8_t*)pDC->pArena->AllocAlignedSync(4 * sizeof(MEGABYTE), sizeof(float) * 8);
}
const API_STATE& state = GetApiState(pDC);