X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fgallium%2Fdrivers%2Fswr%2Frasterizer%2Fcore%2Fapi.cpp;h=3922606b3a69d04605de00424aa48d11d4ed91f2;hb=29e1c4a8a9f26ce41aa53dc9bf39852a8530adc6;hp=bc36cfb7727cb003404cd78d2270d906a25e32dd;hpb=7cf187d08ae6a64c959de1cdf9004f5fb2fd097a;p=mesa.git

diff --git a/src/gallium/drivers/swr/rasterizer/core/api.cpp b/src/gallium/drivers/swr/rasterizer/core/api.cpp
index bc36cfb7727..3922606b3a6 100644
--- a/src/gallium/drivers/swr/rasterizer/core/api.cpp
+++ b/src/gallium/drivers/swr/rasterizer/core/api.cpp
@@ -75,6 +75,17 @@ HANDLE SwrCreateContext(
     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->dcRing[dc].pArena = new CachingArena(pContext->cachingArenaAllocator);
@@ -84,7 +95,7 @@ HANDLE SwrCreateContext(
         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));
@@ -95,9 +106,8 @@ HANDLE SwrCreateContext(
     }
 
     // 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;
@@ -207,13 +217,18 @@ 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();
@@ -431,6 +446,20 @@ void SwrWaitForIdle(HANDLE hContext)
     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,