[Mesa-dev] [PATCH 09/21] swr/rast: Use different handing for stream masks

[George Kyriazis]

---
 src/gallium/drivers/swr/rasterizer/common/os.h              | 3 +++
 src/gallium/drivers/swr/rasterizer/core/api.cpp             | 4 ++--
 src/gallium/drivers/swr/rasterizer/core/frontend.cpp        | 6 +++---
 src/gallium/drivers/swr/rasterizer/core/state.h             | 2 +-
 src/gallium/drivers/swr/rasterizer/jitter/streamout_jit.cpp | 2 ++
 5 files changed, 11 insertions(+), 6 deletions(-)

diff --git a/src/gallium/drivers/swr/rasterizer/common/os.h b/src/gallium/drivers/swr/rasterizer/common/os.h
index 5cfd12f..e779562 100644
--- a/src/gallium/drivers/swr/rasterizer/common/os.h
+++ b/src/gallium/drivers/swr/rasterizer/common/os.h
@@ -209,6 +209,9 @@ unsigned char _BitScanReverse(unsigned int *Index, unsigned int Mask)
     return (Mask != 0);
 }
 
+#define _BitScanForward64 _BitScanForward
+#define _BitScanReverse64 _BitScanReverse
+
 inline
 void *AlignedMalloc(size_t size, size_t alignment)
 {
diff --git a/src/gallium/drivers/swr/rasterizer/core/api.cpp b/src/gallium/drivers/swr/rasterizer/core/api.cpp
index e37e2e4..a2ee85d 100644
--- a/src/gallium/drivers/swr/rasterizer/core/api.cpp
+++ b/src/gallium/drivers/swr/rasterizer/core/api.cpp
@@ -976,14 +976,14 @@ void SetupPipeline(DRAW_CONTEXT *pDC)
 
     if (pState->state.soState.soEnable)
     {
-        uint32_t streamMasks = 0;
+        uint64_t streamMasks = 0;
         for (uint32_t i = 0; i < 4; ++i)
         {
             streamMasks |= pState->state.soState.streamMasks[i];
         }
 
         DWORD maxAttrib;
-        if (_BitScanReverse(&maxAttrib, streamMasks))
+        if (_BitScanReverse64(&maxAttrib, streamMasks))
         {
             pState->state.feNumAttributes = std::max(pState->state.feNumAttributes, (uint32_t)(maxAttrib + 1));
         }
diff --git a/src/gallium/drivers/swr/rasterizer/core/frontend.cpp b/src/gallium/drivers/swr/rasterizer/core/frontend.cpp
index 6e2bab3..1847c3e 100644
--- a/src/gallium/drivers/swr/rasterizer/core/frontend.cpp
+++ b/src/gallium/drivers/swr/rasterizer/core/frontend.cpp
@@ -528,10 +528,10 @@ static void StreamOut(
     for (uint32_t primIndex = 0; primIndex < numPrims; ++primIndex)
     {
         DWORD slot = 0;
-        uint32_t soMask = soState.streamMasks[streamIndex];
+        uint64_t soMask = soState.streamMasks[streamIndex];
 
         // Write all entries into primitive data buffer for SOS.
-        while (_BitScanForward(&slot, soMask))
+        while (_BitScanForward64(&slot, soMask))
         {
             simd4scalar attrib[MAX_NUM_VERTS_PER_PRIM];    // prim attribs (always 4 wide)
             uint32_t paSlot = slot + soState.vertexAttribOffset[streamIndex];
@@ -551,7 +551,7 @@ static void StreamOut(
                 _mm_store_ps((float*)pPrimDataAttrib, attrib[v]);
             }
 
-            soMask &= ~(1 << slot);
+            soMask &= ~(uint64_t(1) << slot);
         }
 
         // Update pPrimData pointer 
diff --git a/src/gallium/drivers/swr/rasterizer/core/state.h b/src/gallium/drivers/swr/rasterizer/core/state.h
index 217cf44..f160913 100644
--- a/src/gallium/drivers/swr/rasterizer/core/state.h
+++ b/src/gallium/drivers/swr/rasterizer/core/state.h
@@ -702,7 +702,7 @@ struct SWR_STREAMOUT_STATE
     // The stream masks specify which attributes are sent to which streams.
     // These masks help the FE to setup the pPrimData buffer that is passed
     // the Stream Output Shader (SOS) function.
-    uint32_t streamMasks[MAX_SO_STREAMS];
+    uint64_t streamMasks[MAX_SO_STREAMS];
 
     // Number of attributes, including position, per vertex that are streamed out.
     // This should match number of bits in stream mask.
diff --git a/src/gallium/drivers/swr/rasterizer/jitter/streamout_jit.cpp b/src/gallium/drivers/swr/rasterizer/jitter/streamout_jit.cpp
index 15a6bc4..f804900 100644
--- a/src/gallium/drivers/swr/rasterizer/jitter/streamout_jit.cpp
+++ b/src/gallium/drivers/swr/rasterizer/jitter/streamout_jit.cpp
@@ -313,6 +313,7 @@ struct StreamOutJit : public Builder
 
         JitManager::DumpToFile(soFunc, "SoFunc_optimized");
 
+
         return soFunc;
     }
 };
@@ -333,6 +334,7 @@ PFN_SO_FUNC JitStreamoutFunc(HANDLE hJitMgr, const HANDLE hFunc)
 
     pJitMgr->DumpAsm(func, "SoFunc_optimized");
 
+
     return pfnStreamOut;
 }
 
-- 
2.7.4