/****************************************************************************
-* Copyright (C) 2014-2015 Intel Corporation. All Rights Reserved.
+* Copyright (C) 2014-2016 Intel Corporation. All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
#include "core/threads.h"
#include "core/tilemgr.h"
#include "core/clip.h"
+#include "core/utils.h"
#include "common/simdintrin.h"
#include "common/os.h"
void SetupDefaultState(SWR_CONTEXT *pContext);
+static INLINE SWR_CONTEXT* GetContext(HANDLE hContext)
+{
+ return (SWR_CONTEXT*)hContext;
+}
+
//////////////////////////////////////////////////////////////////////////
/// @brief Create SWR Context.
/// @param pCreateInfo - pointer to creation info.
RDTSC_RESET();
RDTSC_INIT(0);
- void* pContextMem = _aligned_malloc(sizeof(SWR_CONTEXT), KNOB_SIMD_WIDTH * 4);
+ void* pContextMem = AlignedMalloc(sizeof(SWR_CONTEXT), KNOB_SIMD_WIDTH * 4);
memset(pContextMem, 0, sizeof(SWR_CONTEXT));
SWR_CONTEXT *pContext = new (pContextMem) SWR_CONTEXT();
pContext->dcRing.Init(KNOB_MAX_DRAWS_IN_FLIGHT);
pContext->dsRing.Init(KNOB_MAX_DRAWS_IN_FLIGHT);
- pContext->pMacroTileManagerArray = (MacroTileMgr*)_aligned_malloc(sizeof(MacroTileMgr) * KNOB_MAX_DRAWS_IN_FLIGHT, 64);
- pContext->pDispatchQueueArray = (DispatchQueue*)_aligned_malloc(sizeof(DispatchQueue) * KNOB_MAX_DRAWS_IN_FLIGHT, 64);
+ pContext->pMacroTileManagerArray = (MacroTileMgr*)AlignedMalloc(sizeof(MacroTileMgr) * KNOB_MAX_DRAWS_IN_FLIGHT, 64);
+ pContext->pDispatchQueueArray = (DispatchQueue*)AlignedMalloc(sizeof(DispatchQueue) * KNOB_MAX_DRAWS_IN_FLIGHT, 64);
+
+ pContext->threadInfo.MAX_WORKER_THREADS = KNOB_MAX_WORKER_THREADS;
+ pContext->threadInfo.MAX_NUMA_NODES = KNOB_MAX_NUMA_NODES;
+ pContext->threadInfo.MAX_CORES_PER_NUMA_NODE = KNOB_MAX_CORES_PER_NUMA_NODE;
+ pContext->threadInfo.MAX_THREADS_PER_CORE = KNOB_MAX_THREADS_PER_CORE;
+ pContext->threadInfo.SINGLE_THREADED = KNOB_SINGLE_THREADED;
+
+ if (pCreateInfo->pThreadInfo)
+ {
+ pContext->threadInfo = *pCreateInfo->pThreadInfo;
+ }
for (uint32_t dc = 0; dc < KNOB_MAX_DRAWS_IN_FLIGHT; ++dc)
{
pContext->dsRing[dc].pArena = new CachingArena(pContext->cachingArenaAllocator);
}
- if (!KNOB_SINGLE_THREADED)
+ if (!pContext->threadInfo.SINGLE_THREADED)
{
memset(&pContext->WaitLock, 0, sizeof(pContext->WaitLock));
memset(&pContext->FifosNotEmpty, 0, sizeof(pContext->FifosNotEmpty));
}
// Calling createThreadPool() above can set SINGLE_THREADED
- if (KNOB_SINGLE_THREADED)
+ if (pContext->threadInfo.SINGLE_THREADED)
{
- SET_KNOB(HYPERTHREADED_FE, false);
pContext->NumWorkerThreads = 1;
pContext->NumFEThreads = 1;
pContext->NumBEThreads = 1;
MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE,
numaNode);
#else
- pContext->pScratch[i] = (uint8_t*)_aligned_malloc(32 * sizeof(KILOBYTE), KNOB_SIMD_WIDTH * 4);
+ pContext->pScratch[i] = (uint8_t*)AlignedMalloc(32 * sizeof(KILOBYTE), KNOB_SIMD_WIDTH * 4);
#endif
}
// initialize function pointer tables
InitClearTilesTable();
- // initialize store tiles function
+ // initialize callback functions
pContext->pfnLoadTile = pCreateInfo->pfnLoadTile;
pContext->pfnStoreTile = pCreateInfo->pfnStoreTile;
pContext->pfnClearTile = pCreateInfo->pfnClearTile;
+ pContext->pfnUpdateSoWriteOffset = pCreateInfo->pfnUpdateSoWriteOffset;
// pass pointer to bucket manager back to caller
#ifdef KNOB_ENABLE_RDTSC
void SwrDestroyContext(HANDLE hContext)
{
- SWR_CONTEXT *pContext = (SWR_CONTEXT*)hContext;
+ SWR_CONTEXT *pContext = GetContext(hContext);
DestroyThreadPool(pContext, &pContext->threadPool);
// free the fifos
pContext->pDispatchQueueArray[i].~DispatchQueue();
}
- _aligned_free(pContext->pDispatchQueueArray);
- _aligned_free(pContext->pMacroTileManagerArray);
+ AlignedFree(pContext->pDispatchQueueArray);
+ AlignedFree(pContext->pMacroTileManagerArray);
// 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]);
+ AlignedFree(pContext->pScratch[i]);
#endif
}
delete(pContext->pHotTileMgr);
pContext->~SWR_CONTEXT();
- _aligned_free((SWR_CONTEXT*)hContext);
+ AlignedFree(GetContext(hContext));
}
void CopyState(DRAW_STATE& dst, const DRAW_STATE& src)
pContext->FifosNotEmpty.notify_all();
}
+static TileSet gSingleThreadLockedTiles;
+
template<bool IsDraw>
void QueueWork(SWR_CONTEXT *pContext)
{
// then moved on if all work is done.)
pContext->pCurDrawContext->threadsDone = pContext->NumFEThreads + pContext->NumBEThreads;
+ if (IsDraw)
+ {
+ InterlockedIncrement((volatile LONG*)&pContext->drawsOutstandingFE);
+ }
+
_ReadWriteBarrier();
{
std::unique_lock<std::mutex> lock(pContext->WaitLock);
pContext->dcRing.Enqueue();
}
- if (KNOB_SINGLE_THREADED)
+ if (pContext->threadInfo.SINGLE_THREADED)
{
// flush denormals to 0
uint32_t mxcsr = _mm_getcsr();
if (IsDraw)
{
- static TileSet lockedTiles;
- uint64_t curDraw[2] = { pContext->pCurDrawContext->drawId, pContext->pCurDrawContext->drawId };
+ uint32_t curDraw[2] = { pContext->pCurDrawContext->drawId, pContext->pCurDrawContext->drawId };
WorkOnFifoFE(pContext, 0, curDraw[0]);
- WorkOnFifoBE(pContext, 0, curDraw[1], lockedTiles, 0, 0);
+ WorkOnFifoBE(pContext, 0, curDraw[1], gSingleThreadLockedTiles, 0, 0);
}
else
{
- uint64_t curDispatch = pContext->pCurDrawContext->drawId;
+ uint32_t curDispatch = pContext->pCurDrawContext->drawId;
WorkOnCompute(pContext, 0, curDispatch);
}
SWR_ASSERT(pCurDrawContext->pArena->IsEmpty() == true);
- pCurDrawContext->dependency = 0;
+ pCurDrawContext->dependent = false;
pCurDrawContext->pContext = pContext;
pCurDrawContext->isCompute = false; // Dispatch has to set this to true.
pCurDrawContext->doneFE = false;
pCurDrawContext->FeLock = 0;
pCurDrawContext->threadsDone = 0;
+ pCurDrawContext->retireCallback.pfnCallbackFunc = nullptr;
+
+ memset(&pCurDrawContext->dynState, 0, sizeof(pCurDrawContext->dynState));
// Assign unique drawId for this DC
pCurDrawContext->drawId = pContext->dcRing.GetHead();
void* pOutputStateBlock,
size_t memSize)
{
- SWR_CONTEXT *pContext = (SWR_CONTEXT*)hContext;
+ SWR_CONTEXT *pContext = GetContext(hContext);
auto pSrc = GetDrawState(pContext);
SWR_ASSERT(pOutputStateBlock && memSize >= sizeof(*pSrc));
const void* pStateBlock,
size_t memSize)
{
- SWR_CONTEXT *pContext = (SWR_CONTEXT*)hContext;
+ SWR_CONTEXT *pContext = GetContext(hContext);
auto pDst = GetDrawState(pContext);
SWR_ASSERT(pStateBlock && memSize >= sizeof(*pDst));
pState->rastState.frontWinding = SWR_FRONTWINDING_CCW;
}
-static INLINE SWR_CONTEXT* GetContext(HANDLE hContext)
-{
- return (SWR_CONTEXT*)hContext;
-}
-
void SwrSync(HANDLE hContext, PFN_CALLBACK_FUNC pfnFunc, uint64_t userData, uint64_t userData2, uint64_t userData3)
{
RDTSC_START(APISync);
pDC->FeWork.type = SYNC;
pDC->FeWork.pfnWork = ProcessSync;
- pDC->FeWork.desc.sync.pfnCallbackFunc = pfnFunc;
- pDC->FeWork.desc.sync.userData = userData;
- pDC->FeWork.desc.sync.userData2 = userData2;
- pDC->FeWork.desc.sync.userData3 = userData3;
- // cannot execute until all previous draws have completed
- pDC->dependency = pDC->drawId - 1;
+ // Setup callback function
+ pDC->retireCallback.pfnCallbackFunc = pfnFunc;
+ pDC->retireCallback.userData = userData;
+ pDC->retireCallback.userData2 = userData2;
+ pDC->retireCallback.userData3 = userData3;
//enqueue
QueueDraw(pContext);
RDTSC_STOP(APIWaitForIdle, 1, 0);
}
+void SwrWaitForIdleFE(HANDLE hContext)
+{
+ SWR_CONTEXT *pContext = GetContext(hContext);
+
+ RDTSC_START(APIWaitForIdle);
+
+ while (pContext->drawsOutstandingFE > 0)
+ {
+ _mm_pause();
+ }
+
+ RDTSC_STOP(APIWaitForIdle, 1, 0);
+}
+
void SwrSetVertexBuffers(
HANDLE hContext,
uint32_t numBuffers,
void SwrSetCsFunc(
HANDLE hContext,
PFN_CS_FUNC pfnCsFunc,
- uint32_t totalThreadsInGroup)
+ uint32_t totalThreadsInGroup,
+ uint32_t totalSpillFillSize)
{
API_STATE* pState = GetDrawState(GetContext(hContext));
pState->pfnCsFunc = pfnCsFunc;
pState->totalThreadsInGroup = totalThreadsInGroup;
+ pState->totalSpillFillSize = totalSpillFillSize;
}
void SwrSetTsState(
pState->pfnBlendFunc[renderTarget] = pfnBlendFunc;
}
-void SwrSetLinkage(
- HANDLE hContext,
- uint32_t mask,
- const uint8_t* pMap)
-{
- API_STATE* pState = GetDrawState(GetContext(hContext));
-
- static const uint8_t IDENTITY_MAP[] =
- {
- 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
- 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
- };
- static_assert(sizeof(IDENTITY_MAP) == sizeof(pState->linkageMap),
- "Update for new value of MAX_ATTRIBUTES");
-
- pState->linkageMask = mask;
- pState->linkageCount = _mm_popcnt_u32(mask);
-
- if (!pMap)
- {
- pMap = IDENTITY_MAP;
- }
- memcpy(pState->linkageMap, pMap, pState->linkageCount);
-}
-
// update guardband multipliers for the viewport
void updateGuardband(API_STATE *pState)
{
}
else
{
+ // the vp width and height must be added to origin un-rounded then the result round to -inf.
+ // The cast to int works for rounding assuming all [left, right, top, bottom] are positive.
left = (int32_t)pState->vp[0].x;
- right = (int32_t)pState->vp[0].x + (int32_t)pState->vp[0].width;
+ right = (int32_t)(pState->vp[0].x + pState->vp[0].width);
top = (int32_t)pState->vp[0].y;
- bottom = (int32_t)pState->vp[0].y + (int32_t)pState->vp[0].height;
+ bottom = (int32_t)(pState->vp[0].y + pState->vp[0].height);
}
right = std::min<uint32_t>(right, KNOB_MAX_SCISSOR_X);
pState->scissorInFixedPoint.bottom = bottom * FIXED_POINT_SCALE - 1;
}
}
+
// templated backend function tables
-extern PFN_BACKEND_FUNC gBackendNullPs[SWR_MULTISAMPLE_TYPE_MAX];
-extern PFN_BACKEND_FUNC gBackendSingleSample[2][2];
-extern PFN_BACKEND_FUNC gBackendPixelRateTable[SWR_MULTISAMPLE_TYPE_MAX][SWR_MSAA_SAMPLE_PATTERN_MAX][SWR_INPUT_COVERAGE_MAX][2][2];
-extern PFN_BACKEND_FUNC gBackendSampleRateTable[SWR_MULTISAMPLE_TYPE_MAX][SWR_INPUT_COVERAGE_MAX][2];
-extern PFN_OUTPUT_MERGER gBackendOutputMergerTable[SWR_NUM_RENDERTARGETS + 1][SWR_MULTISAMPLE_TYPE_MAX];
-extern PFN_CALC_PIXEL_BARYCENTRICS gPixelBarycentricTable[2];
-extern PFN_CALC_SAMPLE_BARYCENTRICS gSampleBarycentricTable[2];
-extern PFN_CALC_CENTROID_BARYCENTRICS gCentroidBarycentricTable[SWR_MULTISAMPLE_TYPE_MAX][2][2][2];
+extern PFN_BACKEND_FUNC gBackendNullPs[SWR_MULTISAMPLE_TYPE_COUNT];
+extern PFN_BACKEND_FUNC gBackendSingleSample[SWR_INPUT_COVERAGE_COUNT][2][2];
+extern PFN_BACKEND_FUNC gBackendPixelRateTable[SWR_MULTISAMPLE_TYPE_COUNT][SWR_MSAA_SAMPLE_PATTERN_COUNT][SWR_INPUT_COVERAGE_COUNT][2][2][2];
+extern PFN_BACKEND_FUNC gBackendSampleRateTable[SWR_MULTISAMPLE_TYPE_COUNT][SWR_INPUT_COVERAGE_COUNT][2][2];
void SetupPipeline(DRAW_CONTEXT *pDC)
{
DRAW_STATE* pState = pDC->pState;
const SWR_RASTSTATE &rastState = pState->state.rastState;
const SWR_PS_STATE &psState = pState->state.psState;
BACKEND_FUNCS& backendFuncs = pState->backendFuncs;
- const uint32_t forcedSampleCount = (rastState.bForcedSampleCount) ? 1 : 0;
+ const uint32_t forcedSampleCount = (rastState.forcedSampleCount) ? 1 : 0;
// setup backend
if (psState.pfnPixelShader == nullptr)
{
backendFuncs.pfnBackend = gBackendNullPs[pState->state.rastState.sampleCount];
- // always need to generate I & J per sample for Z interpolation
- backendFuncs.pfnCalcSampleBarycentrics = gSampleBarycentricTable[1];
}
else
{
- const bool bMultisampleEnable = ((rastState.sampleCount > SWR_MULTISAMPLE_1X) || rastState.bForcedSampleCount) ? 1 : 0;
+ const bool bMultisampleEnable = ((rastState.sampleCount > SWR_MULTISAMPLE_1X) || rastState.forcedSampleCount) ? 1 : 0;
const uint32_t centroid = ((psState.barycentricsMask & SWR_BARYCENTRIC_CENTROID_MASK) > 0) ? 1 : 0;
-
- // currently only support 'normal' input coverage
- SWR_ASSERT(psState.inputCoverage == SWR_INPUT_COVERAGE_NORMAL ||
- psState.inputCoverage == SWR_INPUT_COVERAGE_NONE);
+ const uint32_t canEarlyZ = (psState.forceEarlyZ || (!psState.writesODepth && !psState.usesSourceDepth && !psState.usesUAV)) ? 1 : 0;
SWR_BARYCENTRICS_MASK barycentricsMask = (SWR_BARYCENTRICS_MASK)psState.barycentricsMask;
{
// always need to generate I & J per sample for Z interpolation
barycentricsMask = (SWR_BARYCENTRICS_MASK)(barycentricsMask | SWR_BARYCENTRIC_PER_SAMPLE_MASK);
- backendFuncs.pfnBackend = gBackendPixelRateTable[rastState.sampleCount][rastState.samplePattern][psState.inputCoverage][centroid][forcedSampleCount];
- backendFuncs.pfnOutputMerger = gBackendOutputMergerTable[psState.numRenderTargets][pState->state.blendState.sampleCount];
+ backendFuncs.pfnBackend = gBackendPixelRateTable[rastState.sampleCount][rastState.samplePattern][psState.inputCoverage][centroid][forcedSampleCount][canEarlyZ];
}
else
{
// always need to generate I & J per pixel for Z interpolation
barycentricsMask = (SWR_BARYCENTRICS_MASK)(barycentricsMask | SWR_BARYCENTRIC_PER_PIXEL_MASK);
- backendFuncs.pfnBackend = gBackendSingleSample[psState.inputCoverage][centroid];
- backendFuncs.pfnOutputMerger = gBackendOutputMergerTable[psState.numRenderTargets][SWR_MULTISAMPLE_1X];
+ backendFuncs.pfnBackend = gBackendSingleSample[psState.inputCoverage][centroid][canEarlyZ];
}
break;
case SWR_SHADING_RATE_SAMPLE:
SWR_ASSERT(rastState.samplePattern == SWR_MSAA_STANDARD_PATTERN);
// always need to generate I & J per sample for Z interpolation
barycentricsMask = (SWR_BARYCENTRICS_MASK)(barycentricsMask | SWR_BARYCENTRIC_PER_SAMPLE_MASK);
- backendFuncs.pfnBackend = gBackendSampleRateTable[rastState.sampleCount][psState.inputCoverage][centroid];
- backendFuncs.pfnOutputMerger = gBackendOutputMergerTable[psState.numRenderTargets][pState->state.blendState.sampleCount];
+ backendFuncs.pfnBackend = gBackendSampleRateTable[rastState.sampleCount][psState.inputCoverage][centroid][canEarlyZ];
break;
default:
SWR_ASSERT(0 && "Invalid shading rate");
break;
}
-
- // setup pointer to function that generates necessary barycentrics required by the PS
- bool bBarycentrics = (barycentricsMask & SWR_BARYCENTRIC_PER_PIXEL_MASK) > 0 ? 1 : 0;
- backendFuncs.pfnCalcPixelBarycentrics = gPixelBarycentricTable[bBarycentrics];
-
- bBarycentrics = (barycentricsMask & SWR_BARYCENTRIC_PER_SAMPLE_MASK) > 0 ? 1 : 0;
- backendFuncs.pfnCalcSampleBarycentrics = gSampleBarycentricTable[bBarycentrics];
-
- bBarycentrics = (barycentricsMask & SWR_BARYCENTRIC_CENTROID_MASK) > 0 ? 1 : 0;
- backendFuncs.pfnCalcCentroidBarycentrics = gCentroidBarycentricTable[rastState.sampleCount][bBarycentrics][rastState.samplePattern][forcedSampleCount];
}
PFN_PROCESS_PRIMS pfnBinner;
break;
default:
pState->pfnProcessPrims = ClipTriangles;
- pfnBinner = BinTriangles;
+ pfnBinner = GetBinTrianglesFunc((rastState.conservativeRast > 0));
break;
};
(pState->state.depthStencilState.depthWriteEnable == FALSE) &&
(pState->state.depthStencilState.stencilTestEnable == FALSE) &&
(pState->state.depthStencilState.stencilWriteEnable == FALSE) &&
- (pState->state.linkageCount == 0))
+ (pState->state.backendState.numAttributes == 0))
{
pState->pfnProcessPrims = nullptr;
- pState->state.linkageMask = 0;
}
if (pState->state.soState.rasterizerDisable == true)
{
pState->pfnProcessPrims = nullptr;
- pState->state.linkageMask = 0;
}
- // set up the frontend attrib mask
- pState->state.feAttribMask = pState->state.linkageMask;
+ // set up the frontend attribute count
+ pState->state.feNumAttributes = 0;
+ const SWR_BACKEND_STATE& backendState = pState->state.backendState;
+ if (backendState.swizzleEnable)
+ {
+ // attribute swizzling is enabled, iterate over the map and record the max attribute used
+ for (uint32_t i = 0; i < backendState.numAttributes; ++i)
+ {
+ pState->state.feNumAttributes = std::max(pState->state.feNumAttributes, (uint32_t)backendState.swizzleMap[i].sourceAttrib + 1);
+ }
+ }
+ else
+ {
+ pState->state.feNumAttributes = pState->state.backendState.numAttributes;
+ }
+
if (pState->state.soState.soEnable)
{
+ uint32_t streamMasks = 0;
for (uint32_t i = 0; i < 4; ++i)
{
- pState->state.feAttribMask |= pState->state.soState.streamMasks[i];
+ streamMasks |= pState->state.soState.streamMasks[i];
+ }
+
+ DWORD maxAttrib;
+ if (_BitScanReverse(&maxAttrib, streamMasks))
+ {
+ pState->state.feNumAttributes = std::max(pState->state.feNumAttributes, (uint32_t)(maxAttrib + 1));
}
}
return vertsPerDraw;
}
-// Recursive template used to auto-nest conditionals. Converts dynamic boolean function
-// arguments to static template arguments.
-template <bool... ArgsB>
-struct FEDrawChooser
-{
- // Last Arg Terminator
- static PFN_FE_WORK_FUNC GetFunc(bool bArg)
- {
- if (bArg)
- {
- return ProcessDraw<ArgsB..., true>;
- }
-
- return ProcessDraw<ArgsB..., false>;
- }
-
- // Recursively parse args
- template <typename... TArgsT>
- static PFN_FE_WORK_FUNC GetFunc(bool bArg, TArgsT... remainingArgs)
- {
- if (bArg)
- {
- return FEDrawChooser<ArgsB..., true>::GetFunc(remainingArgs...);
- }
-
- return FEDrawChooser<ArgsB..., false>::GetFunc(remainingArgs...);
- }
-};
-
-// Selector for correct templated Draw front-end function
-INLINE
-static PFN_FE_WORK_FUNC GetFEDrawFunc(bool IsIndexed, bool HasTessellation, bool HasGeometryShader, bool HasStreamOut, bool RasterizerEnabled)
-{
- return FEDrawChooser<>::GetFunc(IsIndexed, HasTessellation, HasGeometryShader, HasStreamOut, RasterizerEnabled);
-}
-
//////////////////////////////////////////////////////////////////////////
/// @brief DrawInstanced
SWR_CONTEXT *pContext = GetContext(hContext);
DRAW_CONTEXT* pDC = GetDrawContext(pContext);
- int32_t maxVertsPerDraw = MaxVertsPerDraw(pDC, numVertices, topology);
+ uint32_t maxVertsPerDraw = MaxVertsPerDraw(pDC, numVertices, topology);
uint32_t primsPerDraw = GetNumPrims(topology, maxVertsPerDraw);
- int32_t remainingVerts = numVertices;
+ uint32_t remainingVerts = numVertices;
API_STATE *pState = &pDC->pState->state;
pState->topology = topology;
InitDraw(pDC, isSplitDraw);
pDC->FeWork.type = DRAW;
- pDC->FeWork.pfnWork = GetFEDrawFunc(
+ pDC->FeWork.pfnWork = GetProcessDrawFunc(
false, // IsIndexed
+ false, // bEnableCutIndex
pState->tsState.tsEnable,
pState->gsState.gsEnable,
pState->soState.soEnable,
DRAW_CONTEXT* pDC = GetDrawContext(pContext);
API_STATE* pState = &pDC->pState->state;
- int32_t maxIndicesPerDraw = MaxVertsPerDraw(pDC, numIndices, topology);
+ uint32_t maxIndicesPerDraw = MaxVertsPerDraw(pDC, numIndices, topology);
uint32_t primsPerDraw = GetNumPrims(topology, maxIndicesPerDraw);
- int32_t remainingIndices = numIndices;
+ uint32_t remainingIndices = numIndices;
uint32_t indexSize = 0;
switch (pState->indexBuffer.format)
InitDraw(pDC, isSplitDraw);
pDC->FeWork.type = DRAW;
- pDC->FeWork.pfnWork = GetFEDrawFunc(
+ pDC->FeWork.pfnWork = GetProcessDrawFunc(
true, // IsIndexed
+ pState->frontendState.bEnableCutIndex,
pState->tsState.tsEnable,
pState->gsState.gsEnable,
pState->soState.soEnable,
HANDLE hContext,
uint32_t attachmentMask)
{
- SWR_CONTEXT *pContext = (SWR_CONTEXT*)hContext;
+ if (KNOB_TOSS_DRAW)
+ {
+ return;
+ }
+
+ SWR_CONTEXT *pContext = GetContext(hContext);
DRAW_CONTEXT* pDC = GetDrawContext(pContext);
pDC->FeWork.type = DISCARDINVALIDATETILES;
uint32_t attachmentMask,
SWR_RECT rect)
{
- SWR_CONTEXT *pContext = (SWR_CONTEXT*)hContext;
+ if (KNOB_TOSS_DRAW)
+ {
+ return;
+ }
+
+ SWR_CONTEXT *pContext = GetContext(hContext);
DRAW_CONTEXT* pDC = GetDrawContext(pContext);
// Queue a load to the hottile
}
RDTSC_START(APIDispatch);
- SWR_CONTEXT *pContext = (SWR_CONTEXT*)hContext;
+ SWR_CONTEXT *pContext = GetContext(hContext);
DRAW_CONTEXT* pDC = GetDrawContext(pContext);
pDC->isCompute = true; // This is a compute context.
SWR_RENDERTARGET_ATTACHMENT attachment,
SWR_TILE_STATE postStoreTileState)
{
+ if (KNOB_TOSS_DRAW)
+ {
+ return;
+ }
+
RDTSC_START(APIStoreTiles);
- SWR_CONTEXT *pContext = (SWR_CONTEXT*)hContext;
+ SWR_CONTEXT *pContext = GetContext(hContext);
DRAW_CONTEXT* pDC = GetDrawContext(pContext);
SetupMacroTileScissors(pDC);
float z,
uint8_t stencil)
{
+ if (KNOB_TOSS_DRAW)
+ {
+ return;
+ }
+
RDTSC_START(APIClearRenderTarget);
- SWR_CONTEXT *pContext = (SWR_CONTEXT*)hContext;
+ SWR_CONTEXT *pContext = GetContext(hContext);
DRAW_CONTEXT* pDC = GetDrawContext(pContext);
pDC->FeWork.desc.queryStats.pStats = pStats;
// cannot execute until all previous draws have completed
- pDC->dependency = pDC->drawId - 1;
+ pDC->dependent = true;
//enqueue
QueueDraw(pContext);