swr: [rasterizer core/memory] StoreTile: AVX512 progress

[mesa.git] / src / gallium / drivers / swr / rasterizer / core / utils.h

diff --git a/src/gallium/drivers/swr/rasterizer/core/utils.h b/src/gallium/drivers/swr/rasterizer/core/utils.h
index 8f968649a5733279dd1c1ea34890f23ac64f20ae..a2365757a7f904c891a1b6dd420ec78aee2a4fa6 100644 (file)
--- a/src/gallium/drivers/swr/rasterizer/core/utils.h
+++ b/src/gallium/drivers/swr/rasterizer/core/utils.h
@@ -421,8 +421,8 @@ struct Transpose32_32_32_32
 
         vTranspose4x8(vDst, _simd16_extract_ps(src0, 1), _simd16_extract_ps(src1, 1), _simd16_extract_ps(src2, 1), _simd16_extract_ps(src3, 1));
 
-        _simd16_store_ps(reinterpret_cast<float *>(pDst) + 32, reinterpret_cast<simd16scalar *>(vDst)[2]);
-        _simd16_store_ps(reinterpret_cast<float *>(pDst) + 48, reinterpret_cast<simd16scalar *>(vDst)[3]);
+        _simd16_store_ps(reinterpret_cast<float *>(pDst) + 32, reinterpret_cast<simd16scalar *>(vDst)[0]);
+        _simd16_store_ps(reinterpret_cast<float *>(pDst) + 48, reinterpret_cast<simd16scalar *>(vDst)[1]);
     }
 #endif
 };
@@ -474,8 +474,8 @@ struct Transpose32_32_32
 
         vTranspose3x8(vDst, _simd16_extract_ps(src0, 1), _simd16_extract_ps(src1, 1), _simd16_extract_ps(src2, 1));
 
-        _simd16_store_ps(reinterpret_cast<float *>(pDst) + 32, reinterpret_cast<simd16scalar *>(vDst)[2]);
-        _simd16_store_ps(reinterpret_cast<float *>(pDst) + 48, reinterpret_cast<simd16scalar *>(vDst)[3]);
+        _simd16_store_ps(reinterpret_cast<float *>(pDst) + 32, reinterpret_cast<simd16scalar *>(vDst)[0]);
+        _simd16_store_ps(reinterpret_cast<float *>(pDst) + 48, reinterpret_cast<simd16scalar *>(vDst)[1]);
     }
 #endif
 };