swr/rast: FE/Binner - unify SIMD8/16 functions using simdlib types

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

5 files changed, 696 insertions, 1739 deletions

diff --git a/src/gallium/drivers/swr/rasterizer/core/binner.cpp b/src/gallium/drivers/swr/rasterizer/core/binner.cpp
index 9fe1b01711..e09ff7a399 100644
--- a/src/gallium/drivers/swr/rasterizer/core/binner.cpp
+++ b/src/gallium/drivers/swr/rasterizer/core/binner.cpp
@@ -36,13 +36,26 @@
 #include "tilemgr.h"
 
 // Function Prototype
-void BinPostSetupLines(DRAW_CONTEXT *pDC, PA_STATE& pa, uint32_t workerId, simdvector prims[3], simdscalar vRecipW[2], uint32_t primMask, simdscalari const &primID, simdscalari const &viewportIdx);
-void BinPostSetupPoints(DRAW_CONTEXT *pDC, PA_STATE& pa, uint32_t workerId, simdvector prims[], uint32_t primMask, simdscalari const &primID, simdscalari const &viewportIdx);
+template <typename SIMD_T, uint32_t SIMD_WIDTH>
+void BinPostSetupLinesImpl(
+    DRAW_CONTEXT *pDC,
+    PA_STATE &pa,
+    uint32_t workerId,
+    typename SIMD_T::Vec4 prim[],
+    typename SIMD_T::Float recipW[],
+    uint32_t primMask,
+    typename SIMD_T::Integer const &primID,
+    typename SIMD_T::Integer const &viewportIdx);
 
-#if USE_SIMD16_FRONTEND
-void BinPostSetupLines_simd16(DRAW_CONTEXT *pDC, PA_STATE& pa, uint32_t workerId, simd16vector prims[3], simd16scalar vRecipW[2], uint32_t primMask, simd16scalari const &primID, simd16scalari const &viewportIdx);
-void BinPostSetupPoints_simd16(DRAW_CONTEXT *pDC, PA_STATE& pa, uint32_t workerId, simd16vector prims[], uint32_t primMask, simd16scalari const &primID, simd16scalari const &viewportIdx);
-#endif
+template <typename SIMD_T, uint32_t SIMD_WIDTH>
+void BinPostSetupPointsImpl(
+    DRAW_CONTEXT *pDC,
+    PA_STATE &pa,
+    uint32_t workerId,
+    typename SIMD_T::Vec4 prim[],
+    uint32_t primMask,
+    typename SIMD_T::Integer const &primID,
+    typename SIMD_T::Integer const &viewportIdx);
 
 //////////////////////////////////////////////////////////////////////////
 /// @brief Processes attributes for the backend based on linkage mask and
@@ -209,148 +222,123 @@ INLINE void ProcessAttributes(
 /// @param scisYmax - output vector of per-prmitive scissor rect Ymax data.
 //
 /// @todo:  Look at speeding this up -- weigh against corresponding costs in rasterizer.
-template<size_t SimdWidth>
-struct GatherScissors
+static void GatherScissors(const SWR_RECT *pScissorsInFixedPoint, const uint32_t *pViewportIndex,
+    simdscalari &scisXmin, simdscalari &scisYmin, simdscalari &scisXmax, simdscalari &scisYmax)
 {
-    static void Gather(const SWR_RECT* pScissorsInFixedPoint, const uint32_t* pViewportIndex,
-        simdscalari &scisXmin, simdscalari &scisYmin,
-        simdscalari &scisXmax, simdscalari &scisYmax)
-    {
-        SWR_INVALID("Unhandled Simd Width in Scissor Rect Gather");
-    }
-};
-
-template<>
-struct GatherScissors<8>
-{
-    static void Gather(const SWR_RECT* pScissorsInFixedPoint, const uint32_t* pViewportIndex,
-        simdscalari &scisXmin, simdscalari &scisYmin,
-        simdscalari &scisXmax, simdscalari &scisYmax)
-    {
-        scisXmin = _simd_set_epi32(pScissorsInFixedPoint[pViewportIndex[0]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[1]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[2]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[3]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[4]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[5]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[6]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[7]].xmin);
-        scisYmin = _simd_set_epi32(pScissorsInFixedPoint[pViewportIndex[0]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[1]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[2]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[3]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[4]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[5]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[6]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[7]].ymin);
-        scisXmax = _simd_set_epi32(pScissorsInFixedPoint[pViewportIndex[0]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[1]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[2]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[3]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[4]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[5]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[6]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[7]].xmax);
-        scisYmax = _simd_set_epi32(pScissorsInFixedPoint[pViewportIndex[0]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[1]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[2]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[3]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[4]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[5]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[6]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[7]].ymax);
-    }
-};
-
-#if USE_SIMD16_FRONTEND
-template<size_t SimdWidth>
-struct GatherScissors_simd16
-{
-    static void Gather(const SWR_RECT* pScissorsInFixedPoint, const uint32_t* pViewportIndex,
-        simd16scalari &scisXmin, simd16scalari &scisYmin,
-        simd16scalari &scisXmax, simd16scalari &scisYmax)
-    {
-        SWR_INVALID("Unhandled Simd Width in Scissor Rect Gather");
-    }
-};
+    scisXmin = _simd_set_epi32(
+        pScissorsInFixedPoint[pViewportIndex[0]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[1]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[2]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[3]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[4]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[5]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[6]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[7]].xmin);
+    scisYmin = _simd_set_epi32(
+        pScissorsInFixedPoint[pViewportIndex[0]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[1]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[2]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[3]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[4]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[5]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[6]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[7]].ymin);
+    scisXmax = _simd_set_epi32(
+        pScissorsInFixedPoint[pViewportIndex[0]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[1]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[2]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[3]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[4]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[5]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[6]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[7]].xmax);
+    scisYmax = _simd_set_epi32(
+        pScissorsInFixedPoint[pViewportIndex[0]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[1]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[2]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[3]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[4]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[5]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[6]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[7]].ymax);
+}
 
-template<>
-struct GatherScissors_simd16<16>
+static void GatherScissors(const SWR_RECT *pScissorsInFixedPoint, const uint32_t *pViewportIndex,
+    simd16scalari &scisXmin, simd16scalari &scisYmin, simd16scalari &scisXmax, simd16scalari &scisYmax)
 {
-    static void Gather(const SWR_RECT* pScissorsInFixedPoint, const uint32_t* pViewportIndex,
-        simd16scalari &scisXmin, simd16scalari &scisYmin,
-        simd16scalari &scisXmax, simd16scalari &scisYmax) {
-        scisXmin = _simd16_set_epi32(pScissorsInFixedPoint[pViewportIndex[0]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[1]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[2]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[3]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[4]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[5]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[6]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[7]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[8]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[9]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[10]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[11]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[12]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[13]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[14]].xmin,
-            pScissorsInFixedPoint[pViewportIndex[15]].xmin);
-
-        scisYmin = _simd16_set_epi32(pScissorsInFixedPoint[pViewportIndex[0]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[1]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[2]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[3]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[4]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[5]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[6]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[7]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[8]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[9]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[10]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[11]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[12]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[13]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[14]].ymin,
-            pScissorsInFixedPoint[pViewportIndex[15]].ymin);
-
-        scisXmax = _simd16_set_epi32(pScissorsInFixedPoint[pViewportIndex[0]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[1]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[2]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[3]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[4]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[5]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[6]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[7]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[8]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[9]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[10]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[11]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[12]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[13]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[14]].xmax,
-            pScissorsInFixedPoint[pViewportIndex[15]].xmax);
-
-        scisYmax = _simd16_set_epi32(pScissorsInFixedPoint[pViewportIndex[0]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[1]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[2]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[3]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[4]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[5]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[6]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[7]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[8]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[9]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[10]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[11]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[12]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[13]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[14]].ymax,
-            pScissorsInFixedPoint[pViewportIndex[15]].ymax);
-    }
-};
+    scisXmin = _simd16_set_epi32(
+        pScissorsInFixedPoint[pViewportIndex[0]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[1]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[2]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[3]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[4]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[5]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[6]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[7]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[8]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[9]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[10]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[11]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[12]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[13]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[14]].xmin,
+        pScissorsInFixedPoint[pViewportIndex[15]].xmin);
+
+    scisYmin = _simd16_set_epi32(
+        pScissorsInFixedPoint[pViewportIndex[0]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[1]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[2]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[3]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[4]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[5]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[6]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[7]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[8]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[9]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[10]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[11]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[12]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[13]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[14]].ymin,
+        pScissorsInFixedPoint[pViewportIndex[15]].ymin);
+
+    scisXmax = _simd16_set_epi32(
+        pScissorsInFixedPoint[pViewportIndex[0]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[1]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[2]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[3]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[4]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[5]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[6]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[7]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[8]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[9]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[10]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[11]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[12]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[13]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[14]].xmax,
+        pScissorsInFixedPoint[pViewportIndex[15]].xmax);
+
+    scisYmax = _simd16_set_epi32(
+        pScissorsInFixedPoint[pViewportIndex[0]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[1]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[2]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[3]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[4]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[5]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[6]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[7]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[8]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[9]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[10]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[11]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[12]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[13]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[14]].ymax,
+        pScissorsInFixedPoint[pViewportIndex[15]].ymax);
+}
 
-#endif
 typedef void(*PFN_PROCESS_ATTRIBUTES)(DRAW_CONTEXT*, PA_STATE&, uint32_t, uint32_t, float*);
 
 struct ProcessAttributesChooser
@@ -416,6 +404,47 @@ void ProcessUserClipDist(PA_STATE& pa, uint32_t primIndex, uint8_t clipDistMask,
     }
 }
 
+// WA linux compiler issue with SIMDLIB and shift immediates
+#define SIMD_WA_SXXI_EPI32 1
+
+#if SIMD_WA_SXXI_EPI32
+template<int i>
+simdscalari simd_wa_slli_epi32(simdscalari a)
+{
+    return SIMD256::slli_epi32<i>(a);
+}
+
+template<int i>
+simd16scalari simd_wa_slli_epi32(simd16scalari a)
+{
+    return SIMD512::slli_epi32<i>(a);
+}
+
+template<int i>
+simdscalari simd_wa_srai_epi32(simdscalari a)
+{
+    return SIMD256::srai_epi32<i>(a);
+}
+
+template<int i>
+simd16scalari simd_wa_srai_epi32(simd16scalari a)
+{
+    return SIMD512::srai_epi32<i>(a);
+}
+
+#endif
+INLINE
+void TransposeVertices(simd4scalar(&dst)[8], const simdscalar &src0, const simdscalar &src1, const simdscalar &src2)
+{
+    vTranspose3x8(dst, src0, src1, src2);
+}
+
+INLINE
+void TransposeVertices(simd4scalar(&dst)[16], const simd16scalar &src0, const simd16scalar &src1, const simd16scalar &src2)
+{
+    vTranspose4x16(reinterpret_cast<simd16scalar(&)[4]>(dst), src0, src1, src2, _simd16_setzero_ps());
+}
+
 //////////////////////////////////////////////////////////////////////////
 /// @brief Bin triangle primitives to macro tiles. Performs setup, clipping
 ///        culling, viewport transform, etc.
@@ -426,14 +455,14 @@ void ProcessUserClipDist(PA_STATE& pa, uint32_t primIndex, uint8_t clipDistMask,
 /// @param primID - Primitive ID for each triangle.
 /// @param viewportIdx - viewport array index for each triangle.
 /// @tparam CT - ConservativeRastFETraits
-template <typename CT>
-void BinTriangles(
+template <typename SIMD_T, uint32_t SIMD_WIDTH, typename CT>
+void SIMDCALL BinTrianglesImpl(
     DRAW_CONTEXT *pDC,
-    PA_STATE& pa,
+    PA_STATE &pa,
     uint32_t workerId,
-    simdvector tri[3],
+    typename SIMD_T::Vec4 tri[3],
     uint32_t triMask,
-    simdscalari const &primID)
+    typename SIMD_T::Integer const &primID)
 {
     SWR_CONTEXT *pContext = pDC->pContext;
 
@@ -442,470 +471,26 @@ void BinTriangles(
     const API_STATE& state = GetApiState(pDC);
     const SWR_RASTSTATE& rastState = state.rastState;
     const SWR_FRONTEND_STATE& feState = state.frontendState;
-    MacroTileMgr *pTileMgr = pDC->pTileMgr;
-
-    simdscalar vRecipW0 = _simd_set1_ps(1.0f);
-    simdscalar vRecipW1 = _simd_set1_ps(1.0f);
-    simdscalar vRecipW2 = _simd_set1_ps(1.0f);
-
-    // Read viewport array index if needed
-    simdscalari viewportIdx = _simd_set1_epi32(0);
-    if (state.backendState.readViewportArrayIndex)
-    {
-        simdvector vpiAttrib[3];
-        pa.Assemble(VERTEX_SGV_SLOT, vpiAttrib);
-
-        // OOB indices => forced to zero.
-        simdscalari vpai = _simd_castps_si(vpiAttrib[0][VERTEX_SGV_VAI_COMP]);
-        vpai = _simd_max_epi32(_simd_setzero_si(), vpai);
-        simdscalari vNumViewports = _simd_set1_epi32(KNOB_NUM_VIEWPORTS_SCISSORS);
-        simdscalari vClearMask = _simd_cmplt_epi32(vpai, vNumViewports);
-        viewportIdx = _simd_and_si(vClearMask, vpai);
-    }
-
-    if (feState.vpTransformDisable)
-    {
-        // RHW is passed in directly when VP transform is disabled
-        vRecipW0 = tri[0].v[3];
-        vRecipW1 = tri[1].v[3];
-        vRecipW2 = tri[2].v[3];
-    }
-    else
-    {
-        // Perspective divide
-        vRecipW0 = _simd_div_ps(_simd_set1_ps(1.0f), tri[0].w);
-        vRecipW1 = _simd_div_ps(_simd_set1_ps(1.0f), tri[1].w);
-        vRecipW2 = _simd_div_ps(_simd_set1_ps(1.0f), tri[2].w);
-
-        tri[0].v[0] = _simd_mul_ps(tri[0].v[0], vRecipW0);
-        tri[1].v[0] = _simd_mul_ps(tri[1].v[0], vRecipW1);
-        tri[2].v[0] = _simd_mul_ps(tri[2].v[0], vRecipW2);
-
-        tri[0].v[1] = _simd_mul_ps(tri[0].v[1], vRecipW0);
-        tri[1].v[1] = _simd_mul_ps(tri[1].v[1], vRecipW1);
-        tri[2].v[1] = _simd_mul_ps(tri[2].v[1], vRecipW2);
-
-        tri[0].v[2] = _simd_mul_ps(tri[0].v[2], vRecipW0);
-        tri[1].v[2] = _simd_mul_ps(tri[1].v[2], vRecipW1);
-        tri[2].v[2] = _simd_mul_ps(tri[2].v[2], vRecipW2);
-
-        // Viewport transform to screen space coords
-        if (state.backendState.readViewportArrayIndex)
-        {
-            viewportTransform<3>(tri, state.vpMatrices, viewportIdx);
-        }
-        else
-        {
-            viewportTransform<3>(tri, state.vpMatrices);
-        }
-    }
-
-    // Adjust for pixel center location
-    simdscalar offset = g_pixelOffsets[rastState.pixelLocation];
-    tri[0].x = _simd_add_ps(tri[0].x, offset);
-    tri[0].y = _simd_add_ps(tri[0].y, offset);
-
-    tri[1].x = _simd_add_ps(tri[1].x, offset);
-    tri[1].y = _simd_add_ps(tri[1].y, offset);
-
-    tri[2].x = _simd_add_ps(tri[2].x, offset);
-    tri[2].y = _simd_add_ps(tri[2].y, offset);
-
-    simdscalari vXi[3], vYi[3];
-    // Set vXi, vYi to required fixed point precision
-    FPToFixedPoint(tri, vXi, vYi);
-
-    // triangle setup
-    simdscalari vAi[3], vBi[3];
-    triangleSetupABIntVertical(vXi, vYi, vAi, vBi);
-
-    // determinant
-    simdscalari vDet[2];
-    calcDeterminantIntVertical(vAi, vBi, vDet);
-
-    // cull zero area
-    int maskLo = _simd_movemask_pd(_simd_castsi_pd(_simd_cmpeq_epi64(vDet[0], _simd_setzero_si())));
-    int maskHi = _simd_movemask_pd(_simd_castsi_pd(_simd_cmpeq_epi64(vDet[1], _simd_setzero_si())));
-
-    int cullZeroAreaMask = maskLo | (maskHi << (KNOB_SIMD_WIDTH / 2));
-
-    uint32_t origTriMask = triMask;
-    // don't cull degenerate triangles if we're conservatively rasterizing
-    if (rastState.fillMode == SWR_FILLMODE_SOLID && !CT::IsConservativeT::value)
-    {
-        triMask &= ~cullZeroAreaMask;
-    }
-
-    // determine front winding tris
-    // CW  +det
-    // CCW det < 0;
-    // 0 area triangles are marked as backfacing regardless of winding order,
-    // which is required behavior for conservative rast and wireframe rendering
-    uint32_t frontWindingTris;
-    if (rastState.frontWinding == SWR_FRONTWINDING_CW)
-    {
-        maskLo = _simd_movemask_pd(_simd_castsi_pd(_simd_cmpgt_epi64(vDet[0], _simd_setzero_si())));
-        maskHi = _simd_movemask_pd(_simd_castsi_pd(_simd_cmpgt_epi64(vDet[1], _simd_setzero_si())));
-    }
-    else
-    {
-        maskLo = _simd_movemask_pd(_simd_castsi_pd(_simd_cmpgt_epi64(_simd_setzero_si(), vDet[0])));
-        maskHi = _simd_movemask_pd(_simd_castsi_pd(_simd_cmpgt_epi64(_simd_setzero_si(), vDet[1])));
-    }
-    frontWindingTris = maskLo | (maskHi << (KNOB_SIMD_WIDTH / 2));
-
-    // cull
-    uint32_t cullTris;
-    switch ((SWR_CULLMODE)rastState.cullMode)
-    {
-    case SWR_CULLMODE_BOTH:  cullTris = 0xffffffff; break;
-    case SWR_CULLMODE_NONE:  cullTris = 0x0; break;
-    case SWR_CULLMODE_FRONT: cullTris = frontWindingTris; break;
-        // 0 area triangles are marked as backfacing, which is required behavior for conservative rast
-    case SWR_CULLMODE_BACK:  cullTris = ~frontWindingTris; break;
-    default: SWR_INVALID("Invalid cull mode: %d", rastState.cullMode); cullTris = 0x0; break;
-    }
-
-    triMask &= ~cullTris;
-
-    if (origTriMask ^ triMask)
-    {
-        RDTSC_EVENT(FECullZeroAreaAndBackface, _mm_popcnt_u32(origTriMask ^ triMask), 0);
-    }
-
-    /// Note: these variable initializations must stay above any 'goto endBenTriangles'
-    // compute per tri backface
-    uint32_t frontFaceMask = frontWindingTris;
-    uint32_t *pPrimID = (uint32_t *)&primID;
-    const uint32_t *pViewportIndex = (uint32_t *)&viewportIdx;
-    DWORD triIndex = 0;
-    uint32_t edgeEnable;
-    PFN_WORK_FUNC pfnWork;
-    if (CT::IsConservativeT::value)
-    {
-        // determine which edges of the degenerate tri, if any, are valid to rasterize.
-        // used to call the appropriate templated rasterizer function
-        if (cullZeroAreaMask > 0)
-        {
-            // e0 = v1-v0
-            simdscalari x0x1Mask = _simd_cmpeq_epi32(vXi[0], vXi[1]);
-            simdscalari y0y1Mask = _simd_cmpeq_epi32(vYi[0], vYi[1]);
-            uint32_t e0Mask = _simd_movemask_ps(_simd_castsi_ps(_simd_and_si(x0x1Mask, y0y1Mask)));
-
-            // e1 = v2-v1
-            simdscalari x1x2Mask = _simd_cmpeq_epi32(vXi[1], vXi[2]);
-            simdscalari y1y2Mask = _simd_cmpeq_epi32(vYi[1], vYi[2]);
-            uint32_t e1Mask = _simd_movemask_ps(_simd_castsi_ps(_simd_and_si(x1x2Mask, y1y2Mask)));
-
-            // e2 = v0-v2
-            // if v0 == v1 & v1 == v2, v0 == v2
-            uint32_t e2Mask = e0Mask & e1Mask;
-            SWR_ASSERT(KNOB_SIMD_WIDTH == 8, "Need to update degenerate mask code for avx512");
-
-            // edge order: e0 = v0v1, e1 = v1v2, e2 = v0v2
-            // 32 bit binary: 0000 0000 0010 0100 1001 0010 0100 1001
-            e0Mask = pdep_u32(e0Mask, 0x00249249);
-            // 32 bit binary: 0000 0000 0100 1001 0010 0100 1001 0010
-            e1Mask = pdep_u32(e1Mask, 0x00492492);
-            // 32 bit binary: 0000 0000 1001 0010 0100 1001 0010 0100
-            e2Mask = pdep_u32(e2Mask, 0x00924924);
-
-            edgeEnable = (0x00FFFFFF & (~(e0Mask | e1Mask | e2Mask)));
-        }
-        else
-        {
-            edgeEnable = 0x00FFFFFF;
-        }
-    }
-    else
-    {
-        // degenerate triangles won't be sent to rasterizer; just enable all edges
-        pfnWork = GetRasterizerFunc(rastState.sampleCount, rastState.bIsCenterPattern, (rastState.conservativeRast > 0), 
-            (SWR_INPUT_COVERAGE)pDC->pState->state.psState.inputCoverage, EdgeValToEdgeState(ALL_EDGES_VALID), (state.scissorsTileAligned == false));
-    }
-
-    simdBBox bbox;
-
-    if (!triMask)
-    {
-        goto endBinTriangles;
-    }
-
-    // Calc bounding box of triangles
-    calcBoundingBoxIntVertical<CT>(tri, vXi, vYi, bbox);
-
-    // determine if triangle falls between pixel centers and discard
-    // only discard for non-MSAA case and when conservative rast is disabled
-    // (xmin + 127) & ~255
-    // (xmax + 128) & ~255
-    if((rastState.sampleCount == SWR_MULTISAMPLE_1X || rastState.bIsCenterPattern) &&
-        (!CT::IsConservativeT::value))
-    {
-        origTriMask = triMask;
-
-        int cullCenterMask;
-        {
-            simdscalari xmin = _simd_add_epi32(bbox.xmin, _simd_set1_epi32(127));
-            xmin = _simd_and_si(xmin, _simd_set1_epi32(~255));
-            simdscalari xmax = _simd_add_epi32(bbox.xmax, _simd_set1_epi32(128));
-            xmax = _simd_and_si(xmax, _simd_set1_epi32(~255));
-
-            simdscalari vMaskH = _simd_cmpeq_epi32(xmin, xmax);
-
-            simdscalari ymin = _simd_add_epi32(bbox.ymin, _simd_set1_epi32(127));
-            ymin = _simd_and_si(ymin, _simd_set1_epi32(~255));
-            simdscalari ymax = _simd_add_epi32(bbox.ymax, _simd_set1_epi32(128));
-            ymax = _simd_and_si(ymax, _simd_set1_epi32(~255));
-
-            simdscalari vMaskV = _simd_cmpeq_epi32(ymin, ymax);
-            vMaskV = _simd_or_si(vMaskH, vMaskV);
-            cullCenterMask = _simd_movemask_ps(_simd_castsi_ps(vMaskV));
-        }
-
-        triMask &= ~cullCenterMask;
-
-        if (origTriMask ^ triMask)
-        {
-            RDTSC_EVENT(FECullBetweenCenters, _mm_popcnt_u32(origTriMask ^ triMask), 0);
-        }
-    }
-
-    // Intersect with scissor/viewport. Subtract 1 ULP in x.8 fixed point since xmax/ymax edge is exclusive.
-    // Gather the AOS effective scissor rects based on the per-prim VP index.
-    /// @todo:  Look at speeding this up -- weigh against corresponding costs in rasterizer.
-    {
-        simdscalari scisXmin, scisYmin, scisXmax, scisYmax;
-        if (state.backendState.readViewportArrayIndex)
-        {
-            GatherScissors<KNOB_SIMD_WIDTH>::Gather(&state.scissorsInFixedPoint[0], pViewportIndex,
-                scisXmin, scisYmin, scisXmax, scisYmax);
-        }
-        else // broadcast fast path for non-VPAI case.
-        {
-            scisXmin = _simd_set1_epi32(state.scissorsInFixedPoint[0].xmin);
-            scisYmin = _simd_set1_epi32(state.scissorsInFixedPoint[0].ymin);
-            scisXmax = _simd_set1_epi32(state.scissorsInFixedPoint[0].xmax);
-            scisYmax = _simd_set1_epi32(state.scissorsInFixedPoint[0].ymax);
-        }
-
-        // Make triangle bbox inclusive
-        bbox.xmax = _simd_sub_epi32(bbox.xmax, _simd_set1_epi32(1));
-        bbox.ymax = _simd_sub_epi32(bbox.ymax, _simd_set1_epi32(1));
-
-        bbox.xmin = _simd_max_epi32(bbox.xmin, scisXmin);
-        bbox.ymin = _simd_max_epi32(bbox.ymin, scisYmin);
-        bbox.xmax = _simd_min_epi32(bbox.xmax, scisXmax);
-        bbox.ymax = _simd_min_epi32(bbox.ymax, scisYmax);
-    }
-
-    if (CT::IsConservativeT::value)
-    {
-        // in the case where a degenerate triangle is on a scissor edge, we need to make sure the primitive bbox has
-        // some area. Bump the xmax/ymax edges out 
-        simdscalari topEqualsBottom = _simd_cmpeq_epi32(bbox.ymin, bbox.ymax);
-        bbox.ymax = _simd_blendv_epi32(bbox.ymax, _simd_add_epi32(bbox.ymax, _simd_set1_epi32(1)), topEqualsBottom);
-        simdscalari leftEqualsRight = _simd_cmpeq_epi32(bbox.xmin, bbox.xmax);
-        bbox.xmax = _simd_blendv_epi32(bbox.xmax, _simd_add_epi32(bbox.xmax, _simd_set1_epi32(1)), leftEqualsRight);
-    }
-
-    // Cull tris completely outside scissor
-    {
-        simdscalari maskOutsideScissorX = _simd_cmpgt_epi32(bbox.xmin, bbox.xmax);
-        simdscalari maskOutsideScissorY = _simd_cmpgt_epi32(bbox.ymin, bbox.ymax);
-        simdscalari maskOutsideScissorXY = _simd_or_si(maskOutsideScissorX, maskOutsideScissorY);
-        uint32_t maskOutsideScissor = _simd_movemask_ps(_simd_castsi_ps(maskOutsideScissorXY));
-        triMask = triMask & ~maskOutsideScissor;
-    }
-
-endBinTriangles:
-
-    // Send surviving triangles to the line or point binner based on fill mode
-    if (rastState.fillMode == SWR_FILLMODE_WIREFRAME)
-    {
-        // Simple non-conformant wireframe mode, useful for debugging.
-        // Construct 3 SIMD lines out of the triangle and call the line binner for each SIMD
-        simdvector line[2];
-        simdscalar recipW[2];
-        line[0] = tri[0];
-        line[1] = tri[1];
-        recipW[0] = vRecipW0;
-        recipW[1] = vRecipW1;
-        BinPostSetupLines(pDC, pa, workerId, line, recipW, triMask, primID, viewportIdx);
-
-        line[0] = tri[1];
-        line[1] = tri[2];
-        recipW[0] = vRecipW1;
-        recipW[1] = vRecipW2;
-        BinPostSetupLines(pDC, pa, workerId, line, recipW, triMask, primID, viewportIdx);
-
-        line[0] = tri[2];
-        line[1] = tri[0];
-        recipW[0] = vRecipW2;
-        recipW[1] = vRecipW0;
-        BinPostSetupLines(pDC, pa, workerId, line, recipW, triMask, primID, viewportIdx);
-
-        AR_END(FEBinTriangles, 1);
-        return;
-    }
-    else if (rastState.fillMode == SWR_FILLMODE_POINT)
-    {
-        // Bin 3 points
-        BinPostSetupPoints(pDC, pa, workerId, &tri[0], triMask, primID, viewportIdx);
-        BinPostSetupPoints(pDC, pa, workerId, &tri[1], triMask, primID, viewportIdx);
-        BinPostSetupPoints(pDC, pa, workerId, &tri[2], triMask, primID, viewportIdx);
-        return;
-    }
-
-    // Convert triangle bbox to macrotile units.
-    bbox.xmin = _simd_srai_epi32(bbox.xmin, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-    bbox.ymin = _simd_srai_epi32(bbox.ymin, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-    bbox.xmax = _simd_srai_epi32(bbox.xmax, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-    bbox.ymax = _simd_srai_epi32(bbox.ymax, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-
-    OSALIGNSIMD(uint32_t) aMTLeft[KNOB_SIMD_WIDTH], aMTRight[KNOB_SIMD_WIDTH], aMTTop[KNOB_SIMD_WIDTH], aMTBottom[KNOB_SIMD_WIDTH];
-    _simd_store_si((simdscalari*)aMTLeft, bbox.xmin);
-    _simd_store_si((simdscalari*)aMTRight, bbox.xmax);
-    _simd_store_si((simdscalari*)aMTTop, bbox.ymin);
-    _simd_store_si((simdscalari*)aMTBottom, bbox.ymax);
-
-    // transpose verts needed for backend
-    /// @todo modify BE to take non-transformed verts
-    simd4scalar vHorizX[8], vHorizY[8], vHorizZ[8], vHorizW[8];
-    vTranspose3x8(vHorizX, tri[0].x, tri[1].x, tri[2].x);
-    vTranspose3x8(vHorizY, tri[0].y, tri[1].y, tri[2].y);
-    vTranspose3x8(vHorizZ, tri[0].z, tri[1].z, tri[2].z);
-    vTranspose3x8(vHorizW, vRecipW0, vRecipW1, vRecipW2);
-
-    // store render target array index
-    OSALIGNSIMD(uint32_t) aRTAI[KNOB_SIMD_WIDTH];
-    if (state.backendState.readRenderTargetArrayIndex)
-    {
-        simdvector vRtai[3];
-        pa.Assemble(VERTEX_SGV_SLOT, vRtai);
-        simdscalari vRtaii;
-        vRtaii = _simd_castps_si(vRtai[0][VERTEX_SGV_RTAI_COMP]);
-        _simd_store_si((simdscalari*)aRTAI, vRtaii);
-    }
-    else
-    {
-        _simd_store_si((simdscalari*)aRTAI, _simd_setzero_si());
-    }
-
-    // scan remaining valid triangles and bin each separately
-    while (_BitScanForward(&triIndex, triMask))
-    {
-        uint32_t linkageCount = state.backendState.numAttributes;
-        uint32_t numScalarAttribs = linkageCount * 4;
-
-        BE_WORK work;
-        work.type = DRAW;
-
-        bool isDegenerate;
-        if (CT::IsConservativeT::value)
-        {
-            // only rasterize valid edges if we have a degenerate primitive
-            int32_t triEdgeEnable = (edgeEnable >> (triIndex * 3)) & ALL_EDGES_VALID;
-            work.pfnWork = GetRasterizerFunc(rastState.sampleCount, rastState.bIsCenterPattern, (rastState.conservativeRast > 0), 
-                (SWR_INPUT_COVERAGE)pDC->pState->state.psState.inputCoverage, EdgeValToEdgeState(triEdgeEnable), (state.scissorsTileAligned == false));
-
-            // Degenerate triangles are required to be constant interpolated
-            isDegenerate = (triEdgeEnable != ALL_EDGES_VALID) ? true : false;
-        }
-        else
-        {
-            isDegenerate = false;
-            work.pfnWork = pfnWork;
-        }
-
-        // Select attribute processor
-        PFN_PROCESS_ATTRIBUTES pfnProcessAttribs = GetProcessAttributesFunc(3,
-            state.backendState.swizzleEnable, state.backendState.constantInterpolationMask, isDegenerate);
-
-        TRIANGLE_WORK_DESC &desc = work.desc.tri;
-
-        desc.triFlags.frontFacing = state.forceFront ? 1 : ((frontFaceMask >> triIndex) & 1);
-        desc.triFlags.renderTargetArrayIndex = aRTAI[triIndex];
-        desc.triFlags.viewportIndex = pViewportIndex[triIndex];
-
-        auto pArena = pDC->pArena;
-        SWR_ASSERT(pArena != nullptr);
-
-        // store active attribs
-        float *pAttribs = (float*)pArena->AllocAligned(numScalarAttribs * 3 * sizeof(float), 16);
-        desc.pAttribs = pAttribs;
-        desc.numAttribs = linkageCount;
-        pfnProcessAttribs(pDC, pa, triIndex, pPrimID[triIndex], desc.pAttribs);
-
-        // store triangle vertex data
-        desc.pTriBuffer = (float*)pArena->AllocAligned(4 * 4 * sizeof(float), 16);
-
-        SIMD128::store_ps(&desc.pTriBuffer[0], vHorizX[triIndex]);
-        SIMD128::store_ps(&desc.pTriBuffer[4], vHorizY[triIndex]);
-        SIMD128::store_ps(&desc.pTriBuffer[8], vHorizZ[triIndex]);
-        SIMD128::store_ps(&desc.pTriBuffer[12], vHorizW[triIndex]);
-
-        // store user clip distances
-        if (rastState.clipDistanceMask)
-        {
-            uint32_t numClipDist = _mm_popcnt_u32(rastState.clipDistanceMask);
-            desc.pUserClipBuffer = (float*)pArena->Alloc(numClipDist * 3 * sizeof(float));
-            ProcessUserClipDist<3>(pa, triIndex, rastState.clipDistanceMask, &desc.pTriBuffer[12], desc.pUserClipBuffer);
-        }
 
-        for (uint32_t y = aMTTop[triIndex]; y <= aMTBottom[triIndex]; ++y)
-        {
-            for (uint32_t x = aMTLeft[triIndex]; x <= aMTRight[triIndex]; ++x)
-            {
-#if KNOB_ENABLE_TOSS_POINTS
-                if (!KNOB_TOSS_SETUP_TRIS)
-#endif
-                {
-                    pTileMgr->enqueue(x, y, &work);
-                }
-            }
-        }
-                     triMask &= ~(1 << triIndex);
-    }
-
-    AR_END(FEBinTriangles, 1);
-}
-
-#if USE_SIMD16_FRONTEND
-template <typename CT>
-void SIMDCALL BinTriangles_simd16(
-    DRAW_CONTEXT *pDC,
-    PA_STATE& pa,
-    uint32_t workerId,
-    simd16vector tri[3],
-    uint32_t triMask,
-    simd16scalari const &primID)
-{
-    SWR_CONTEXT *pContext = pDC->pContext;
-
-    AR_BEGIN(FEBinTriangles, pDC->drawId);
+    MacroTileMgr *pTileMgr = pDC->pTileMgr;
 
-    const API_STATE& state = GetApiState(pDC);
-    const SWR_RASTSTATE& rastState = state.rastState;
-    const SWR_FRONTEND_STATE& feState = state.frontendState;
+    typename SIMD_T::Float vRecipW0 = SIMD_T::set1_ps(1.0f);
+    typename SIMD_T::Float vRecipW1 = SIMD_T::set1_ps(1.0f);
+    typename SIMD_T::Float vRecipW2 = SIMD_T::set1_ps(1.0f);
 
-    MacroTileMgr *pTileMgr = pDC->pTileMgr;
+    typename SIMD_T::Integer viewportIdx = SIMD_T::set1_epi32(0);
 
-    simd16scalar vRecipW0 = _simd16_set1_ps(1.0f);
-    simd16scalar vRecipW1 = _simd16_set1_ps(1.0f);
-    simd16scalar vRecipW2 = _simd16_set1_ps(1.0f);
-    
-    simd16scalari viewportIdx = _simd16_set1_epi32(0);
     if (state.backendState.readViewportArrayIndex)
     {
-        simd16vector vpiAttrib[3];
-        pa.Assemble_simd16(VERTEX_SGV_SLOT, vpiAttrib);
+        typename SIMD_T::Vec4 vpiAttrib[3];
+        pa.Assemble(VERTEX_SGV_SLOT, vpiAttrib);
 
         // OOB indices => forced to zero.
-        simd16scalari vpai = _simd16_castps_si(vpiAttrib[0][VERTEX_SGV_VAI_COMP]);
-        vpai = _simd16_max_epi32(_simd16_setzero_si(), vpai);
-        simd16scalari vNumViewports = _simd16_set1_epi32(KNOB_NUM_VIEWPORTS_SCISSORS);
-        simd16scalari vClearMask = _simd16_cmplt_epi32(vpai, vNumViewports);
-        viewportIdx = _simd16_and_si(vClearMask, vpai);
+        typename SIMD_T::Integer vpai = SIMD_T::castps_si(vpiAttrib[0][VERTEX_SGV_VAI_COMP]);
+        vpai = SIMD_T::max_epi32(SIMD_T::setzero_si(), vpai);
+        typename SIMD_T::Integer vNumViewports = SIMD_T::set1_epi32(KNOB_NUM_VIEWPORTS_SCISSORS);
+        typename SIMD_T::Integer vClearMask = SIMD_T::cmplt_epi32(vpai, vNumViewports);
+        viewportIdx = SIMD_T::and_si(vClearMask, vpai);
     }
 
     if (feState.vpTransformDisable)
@@ -918,21 +503,21 @@ void SIMDCALL BinTriangles_simd16(
     else
     {
         // Perspective divide
-        vRecipW0 = _simd16_div_ps(_simd16_set1_ps(1.0f), tri[0].w);
-        vRecipW1 = _simd16_div_ps(_simd16_set1_ps(1.0f), tri[1].w);
-        vRecipW2 = _simd16_div_ps(_simd16_set1_ps(1.0f), tri[2].w);
+        vRecipW0 = SIMD_T::div_ps(SIMD_T::set1_ps(1.0f), tri[0].w);
+        vRecipW1 = SIMD_T::div_ps(SIMD_T::set1_ps(1.0f), tri[1].w);
+        vRecipW2 = SIMD_T::div_ps(SIMD_T::set1_ps(1.0f), tri[2].w);
 
-        tri[0].v[0] = _simd16_mul_ps(tri[0].v[0], vRecipW0);
-        tri[1].v[0] = _simd16_mul_ps(tri[1].v[0], vRecipW1);
-        tri[2].v[0] = _simd16_mul_ps(tri[2].v[0], vRecipW2);
+        tri[0].v[0] = SIMD_T::mul_ps(tri[0].v[0], vRecipW0);
+        tri[1].v[0] = SIMD_T::mul_ps(tri[1].v[0], vRecipW1);
+        tri[2].v[0] = SIMD_T::mul_ps(tri[2].v[0], vRecipW2);
 
-        tri[0].v[1] = _simd16_mul_ps(tri[0].v[1], vRecipW0);
-        tri[1].v[1] = _simd16_mul_ps(tri[1].v[1], vRecipW1);
-        tri[2].v[1] = _simd16_mul_ps(tri[2].v[1], vRecipW2);
+        tri[0].v[1] = SIMD_T::mul_ps(tri[0].v[1], vRecipW0);
+        tri[1].v[1] = SIMD_T::mul_ps(tri[1].v[1], vRecipW1);
+        tri[2].v[1] = SIMD_T::mul_ps(tri[2].v[1], vRecipW2);
 
-        tri[0].v[2] = _simd16_mul_ps(tri[0].v[2], vRecipW0);
-        tri[1].v[2] = _simd16_mul_ps(tri[1].v[2], vRecipW1);
-        tri[2].v[2] = _simd16_mul_ps(tri[2].v[2], vRecipW2);
+        tri[0].v[2] = SIMD_T::mul_ps(tri[0].v[2], vRecipW0);
+        tri[1].v[2] = SIMD_T::mul_ps(tri[1].v[2], vRecipW1);
+        tri[2].v[2] = SIMD_T::mul_ps(tri[2].v[2], vRecipW2);
 
         // Viewport transform to screen space coords
         if (state.backendState.readViewportArrayIndex)
@@ -946,35 +531,34 @@ void SIMDCALL BinTriangles_simd16(
     }
 
     // Adjust for pixel center location
-    const simd16scalar offset = g_pixelOffsets_simd16[rastState.pixelLocation];
+    typename SIMD_T::Float offset = g_pixelOffsets<SIMD_T>[rastState.pixelLocation];
 
-    tri[0].x = _simd16_add_ps(tri[0].x, offset);
-    tri[0].y = _simd16_add_ps(tri[0].y, offset);
+    tri[0].x = SIMD_T::add_ps(tri[0].x, offset);
+    tri[0].y = SIMD_T::add_ps(tri[0].y, offset);
 
-    tri[1].x = _simd16_add_ps(tri[1].x, offset);
-    tri[1].y = _simd16_add_ps(tri[1].y, offset);
+    tri[1].x = SIMD_T::add_ps(tri[1].x, offset);
+    tri[1].y = SIMD_T::add_ps(tri[1].y, offset);
 
-    tri[2].x = _simd16_add_ps(tri[2].x, offset);
-    tri[2].y = _simd16_add_ps(tri[2].y, offset);
-
-    simd16scalari vXi[3], vYi[3];
+    tri[2].x = SIMD_T::add_ps(tri[2].x, offset);
+    tri[2].y = SIMD_T::add_ps(tri[2].y, offset);
 
     // Set vXi, vYi to required fixed point precision
-    FPToFixedPoint(tri, vXi, vYi);
+    typename SIMD_T::Integer vXi[3], vYi[3];
+    FPToFixedPoint<SIMD_T>(tri, vXi, vYi);
 
     // triangle setup
-    simd16scalari vAi[3], vBi[3];
+    typename SIMD_T::Integer vAi[3], vBi[3];
     triangleSetupABIntVertical(vXi, vYi, vAi, vBi);
 
     // determinant
-    simd16scalari vDet[2];
+    typename SIMD_T::Integer vDet[2];
     calcDeterminantIntVertical(vAi, vBi, vDet);
 
     // cull zero area
-    uint32_t maskLo = _simd16_movemask_pd(_simd16_castsi_pd(_simd16_cmpeq_epi64(vDet[0], _simd16_setzero_si())));
-    uint32_t maskHi = _simd16_movemask_pd(_simd16_castsi_pd(_simd16_cmpeq_epi64(vDet[1], _simd16_setzero_si())));
+    uint32_t maskLo = SIMD_T::movemask_pd(SIMD_T::castsi_pd(SIMD_T::cmpeq_epi64(vDet[0], SIMD_T::setzero_si())));
+    uint32_t maskHi = SIMD_T::movemask_pd(SIMD_T::castsi_pd(SIMD_T::cmpeq_epi64(vDet[1], SIMD_T::setzero_si())));
 
-    uint32_t cullZeroAreaMask = maskLo | (maskHi << (KNOB_SIMD16_WIDTH / 2));
+    uint32_t cullZeroAreaMask = maskLo | (maskHi << (SIMD_WIDTH / 2));
 
     // don't cull degenerate triangles if we're conservatively rasterizing
     uint32_t origTriMask = triMask;
@@ -991,15 +575,15 @@ void SIMDCALL BinTriangles_simd16(
     uint32_t frontWindingTris;
     if (rastState.frontWinding == SWR_FRONTWINDING_CW)
     {
-        maskLo = _simd16_movemask_pd(_simd16_castsi_pd(_simd16_cmpgt_epi64(vDet[0], _simd16_setzero_si())));
-        maskHi = _simd16_movemask_pd(_simd16_castsi_pd(_simd16_cmpgt_epi64(vDet[1], _simd16_setzero_si())));
+        maskLo = SIMD_T::movemask_pd(SIMD_T::castsi_pd(SIMD_T::cmpgt_epi64(vDet[0], SIMD_T::setzero_si())));
+        maskHi = SIMD_T::movemask_pd(SIMD_T::castsi_pd(SIMD_T::cmpgt_epi64(vDet[1], SIMD_T::setzero_si())));
     }
     else
     {
-        maskLo = _simd16_movemask_pd(_simd16_castsi_pd(_simd16_cmpgt_epi64(_simd16_setzero_si(), vDet[0])));
-        maskHi = _simd16_movemask_pd(_simd16_castsi_pd(_simd16_cmpgt_epi64(_simd16_setzero_si(), vDet[1])));
+        maskLo = SIMD_T::movemask_pd(SIMD_T::castsi_pd(SIMD_T::cmpgt_epi64(SIMD_T::setzero_si(), vDet[0])));
+        maskHi = SIMD_T::movemask_pd(SIMD_T::castsi_pd(SIMD_T::cmpgt_epi64(SIMD_T::setzero_si(), vDet[1])));
     }
-    frontWindingTris = maskLo | (maskHi << (KNOB_SIMD16_WIDTH / 2));
+    frontWindingTris = maskLo | (maskHi << (SIMD_WIDTH / 2));
 
     // cull
     uint32_t cullTris;
@@ -1036,16 +620,16 @@ void SIMDCALL BinTriangles_simd16(
         if (cullZeroAreaMask > 0)
         {
             // e0 = v1-v0
-            const simd16scalari x0x1Mask = _simd16_cmpeq_epi32(vXi[0], vXi[1]);
-            const simd16scalari y0y1Mask = _simd16_cmpeq_epi32(vYi[0], vYi[1]);
+            const typename SIMD_T::Integer x0x1Mask = SIMD_T::cmpeq_epi32(vXi[0], vXi[1]);
+            const typename SIMD_T::Integer y0y1Mask = SIMD_T::cmpeq_epi32(vYi[0], vYi[1]);
 
-            uint32_t e0Mask = _simd16_movemask_ps(_simd16_castsi_ps(_simd16_and_si(x0x1Mask, y0y1Mask)));
+            uint32_t e0Mask = SIMD_T::movemask_ps(SIMD_T::castsi_ps(SIMD_T::and_si(x0x1Mask, y0y1Mask)));
 
             // e1 = v2-v1
-            const simd16scalari x1x2Mask = _simd16_cmpeq_epi32(vXi[1], vXi[2]);
-            const simd16scalari y1y2Mask = _simd16_cmpeq_epi32(vYi[1], vYi[2]);
+            const typename SIMD_T::Integer x1x2Mask = SIMD_T::cmpeq_epi32(vXi[1], vXi[2]);
+            const typename SIMD_T::Integer y1y2Mask = SIMD_T::cmpeq_epi32(vYi[1], vYi[2]);
 
-            uint32_t e1Mask = _simd16_movemask_ps(_simd16_castsi_ps(_simd16_and_si(x1x2Mask, y1y2Mask)));
+            uint32_t e1Mask = SIMD_T::movemask_ps(SIMD_T::castsi_ps(SIMD_T::and_si(x1x2Mask, y1y2Mask)));
 
             // e2 = v0-v2
             // if v0 == v1 & v1 == v2, v0 == v2
@@ -1076,7 +660,7 @@ void SIMDCALL BinTriangles_simd16(
             (SWR_INPUT_COVERAGE)pDC->pState->state.psState.inputCoverage, EdgeValToEdgeState(ALL_EDGES_VALID), (state.scissorsTileAligned == false));
     }
 
-    simd16BBox bbox;
+    SIMDBBOX_T<SIMD_T> bbox;
 
     if (!triMask)
     {
@@ -1084,7 +668,7 @@ void SIMDCALL BinTriangles_simd16(
     }
 
     // Calc bounding box of triangles
-    calcBoundingBoxIntVertical<CT>(tri, vXi, vYi, bbox);
+    calcBoundingBoxIntVertical<SIMD_T, CT>(vXi, vYi, bbox);
 
     // determine if triangle falls between pixel centers and discard
     // only discard for non-MSAA case and when conservative rast is disabled
@@ -1098,22 +682,22 @@ void SIMDCALL BinTriangles_simd16(
         int cullCenterMask;
 
         {
-            simd16scalari xmin = _simd16_add_epi32(bbox.xmin, _simd16_set1_epi32(127));
-            xmin = _simd16_and_si(xmin, _simd16_set1_epi32(~255));
-            simd16scalari xmax = _simd16_add_epi32(bbox.xmax, _simd16_set1_epi32(128));
-            xmax = _simd16_and_si(xmax, _simd16_set1_epi32(~255));
+            typename SIMD_T::Integer xmin = SIMD_T::add_epi32(bbox.xmin, SIMD_T::set1_epi32(127));
+            xmin = SIMD_T::and_si(xmin, SIMD_T::set1_epi32(~255));
+            typename SIMD_T::Integer xmax = SIMD_T::add_epi32(bbox.xmax, SIMD_T::set1_epi32(128));
+            xmax = SIMD_T::and_si(xmax, SIMD_T::set1_epi32(~255));
 
-            simd16scalari vMaskH = _simd16_cmpeq_epi32(xmin, xmax);
+            typename SIMD_T::Integer vMaskH = SIMD_T::cmpeq_epi32(xmin, xmax);
 
-            simd16scalari ymin = _simd16_add_epi32(bbox.ymin, _simd16_set1_epi32(127));
-            ymin = _simd16_and_si(ymin, _simd16_set1_epi32(~255));
-            simd16scalari ymax = _simd16_add_epi32(bbox.ymax, _simd16_set1_epi32(128));
-            ymax = _simd16_and_si(ymax, _simd16_set1_epi32(~255));
+            typename SIMD_T::Integer ymin = SIMD_T::add_epi32(bbox.ymin, SIMD_T::set1_epi32(127));
+            ymin = SIMD_T::and_si(ymin, SIMD_T::set1_epi32(~255));
+            typename SIMD_T::Integer ymax = SIMD_T::add_epi32(bbox.ymax, SIMD_T::set1_epi32(128));
+            ymax = SIMD_T::and_si(ymax, SIMD_T::set1_epi32(~255));
 
-            simd16scalari vMaskV = _simd16_cmpeq_epi32(ymin, ymax);
+            typename SIMD_T::Integer vMaskV = SIMD_T::cmpeq_epi32(ymin, ymax);
 
-            vMaskV = _simd16_or_si(vMaskH, vMaskV);
-            cullCenterMask = _simd16_movemask_ps(_simd16_castsi_ps(vMaskV));
+            vMaskV = SIMD_T::or_si(vMaskH, vMaskV);
+            cullCenterMask = SIMD_T::movemask_ps(SIMD_T::castsi_ps(vMaskV));
         }
 
         triMask &= ~cullCenterMask;
@@ -1128,76 +712,82 @@ void SIMDCALL BinTriangles_simd16(
     // Gather the AOS effective scissor rects based on the per-prim VP index.
     /// @todo:  Look at speeding this up -- weigh against corresponding costs in rasterizer.
     {
-        simd16scalari scisXmin, scisYmin, scisXmax, scisYmax;
+        typename SIMD_T::Integer scisXmin, scisYmin, scisXmax, scisYmax;
 
         if (state.backendState.readViewportArrayIndex)
         {
-            GatherScissors_simd16<KNOB_SIMD16_WIDTH>::Gather(&state.scissorsInFixedPoint[0], pViewportIndex,
-                scisXmin, scisYmin, scisXmax, scisYmax);
+            GatherScissors(&state.scissorsInFixedPoint[0], pViewportIndex, scisXmin, scisYmin, scisXmax, scisYmax);
         }
         else // broadcast fast path for non-VPAI case.
         {
-            scisXmin = _simd16_set1_epi32(state.scissorsInFixedPoint[0].xmin);
-            scisYmin = _simd16_set1_epi32(state.scissorsInFixedPoint[0].ymin);
-            scisXmax = _simd16_set1_epi32(state.scissorsInFixedPoint[0].xmax);
-            scisYmax = _simd16_set1_epi32(state.scissorsInFixedPoint[0].ymax);
+            scisXmin = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].xmin);
+            scisYmin = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].ymin);
+            scisXmax = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].xmax);
+            scisYmax = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].ymax);
         }
 
         // Make triangle bbox inclusive
-        bbox.xmax = _simd16_sub_epi32(bbox.xmax, _simd16_set1_epi32(1));
-        bbox.ymax = _simd16_sub_epi32(bbox.ymax, _simd16_set1_epi32(1));
+        bbox.xmax = SIMD_T::sub_epi32(bbox.xmax, SIMD_T::set1_epi32(1));
+        bbox.ymax = SIMD_T::sub_epi32(bbox.ymax, SIMD_T::set1_epi32(1));
 
-        bbox.xmin = _simd16_max_epi32(bbox.xmin, scisXmin);
-        bbox.ymin = _simd16_max_epi32(bbox.ymin, scisYmin);
-        bbox.xmax = _simd16_min_epi32(bbox.xmax, scisXmax);
-        bbox.ymax = _simd16_min_epi32(bbox.ymax, scisYmax);
+        bbox.xmin = SIMD_T::max_epi32(bbox.xmin, scisXmin);
+        bbox.ymin = SIMD_T::max_epi32(bbox.ymin, scisYmin);
+        bbox.xmax = SIMD_T::min_epi32(bbox.xmax, scisXmax);
+        bbox.ymax = SIMD_T::min_epi32(bbox.ymax, scisYmax);
     }
 
     if (CT::IsConservativeT::value)
     {
         // in the case where a degenerate triangle is on a scissor edge, we need to make sure the primitive bbox has
         // some area. Bump the xmax/ymax edges out 
-        simd16scalari topEqualsBottom = _simd16_cmpeq_epi32(bbox.ymin, bbox.ymax);
-        bbox.ymax = _simd16_blendv_epi32(bbox.ymax, _simd16_add_epi32(bbox.ymax, _simd16_set1_epi32(1)), topEqualsBottom);
-        simd16scalari leftEqualsRight = _simd16_cmpeq_epi32(bbox.xmin, bbox.xmax);
-        bbox.xmax = _simd16_blendv_epi32(bbox.xmax, _simd16_add_epi32(bbox.xmax, _simd16_set1_epi32(1)), leftEqualsRight);
+
+        typename SIMD_T::Integer topEqualsBottom = SIMD_T::cmpeq_epi32(bbox.ymin, bbox.ymax);
+        bbox.ymax = SIMD_T::blendv_epi32(bbox.ymax, SIMD_T::add_epi32(bbox.ymax, SIMD_T::set1_epi32(1)), topEqualsBottom);
+
+        typename SIMD_T::Integer leftEqualsRight = SIMD_T::cmpeq_epi32(bbox.xmin, bbox.xmax);
+        bbox.xmax = SIMD_T::blendv_epi32(bbox.xmax, SIMD_T::add_epi32(bbox.xmax, SIMD_T::set1_epi32(1)), leftEqualsRight);
     }
 
     // Cull tris completely outside scissor
     {
-        simd16scalari maskOutsideScissorX = _simd16_cmpgt_epi32(bbox.xmin, bbox.xmax);
-        simd16scalari maskOutsideScissorY = _simd16_cmpgt_epi32(bbox.ymin, bbox.ymax);
-        simd16scalari maskOutsideScissorXY = _simd16_or_si(maskOutsideScissorX, maskOutsideScissorY);
-        uint32_t maskOutsideScissor = _simd16_movemask_ps(_simd16_castsi_ps(maskOutsideScissorXY));
+        typename SIMD_T::Integer maskOutsideScissorX = SIMD_T::cmpgt_epi32(bbox.xmin, bbox.xmax);
+        typename SIMD_T::Integer maskOutsideScissorY = SIMD_T::cmpgt_epi32(bbox.ymin, bbox.ymax);
+        typename SIMD_T::Integer maskOutsideScissorXY = SIMD_T::or_si(maskOutsideScissorX, maskOutsideScissorY);
+        uint32_t maskOutsideScissor = SIMD_T::movemask_ps(SIMD_T::castsi_ps(maskOutsideScissorXY));
         triMask = triMask & ~maskOutsideScissor;
     }
 
 endBinTriangles:
 
+
     // Send surviving triangles to the line or point binner based on fill mode
     if (rastState.fillMode == SWR_FILLMODE_WIREFRAME)
     {
         // Simple non-conformant wireframe mode, useful for debugging
         // construct 3 SIMD lines out of the triangle and call the line binner for each SIMD
-        simd16vector line[2];
-        simd16scalar recipW[2];
+        typename SIMD_T::Vec4 line[2];
+        typename SIMD_T::Float recipW[2];
+
         line[0] = tri[0];
         line[1] = tri[1];
         recipW[0] = vRecipW0;
         recipW[1] = vRecipW1;
-        BinPostSetupLines_simd16(pDC, pa, workerId, line, recipW, triMask, primID, viewportIdx);
+
+        BinPostSetupLinesImpl<SIMD_T, SIMD_WIDTH>(pDC, pa, workerId, line, recipW, triMask, primID, viewportIdx);
 
         line[0] = tri[1];
         line[1] = tri[2];
         recipW[0] = vRecipW1;
         recipW[1] = vRecipW2;
-        BinPostSetupLines_simd16(pDC, pa, workerId, line, recipW, triMask, primID, viewportIdx);
+
+        BinPostSetupLinesImpl<SIMD_T, SIMD_WIDTH>(pDC, pa, workerId, line, recipW, triMask, primID, viewportIdx);
 
         line[0] = tri[2];
         line[1] = tri[0];
         recipW[0] = vRecipW2;
         recipW[1] = vRecipW0;
-        BinPostSetupLines_simd16(pDC, pa, workerId, line, recipW, triMask, primID, viewportIdx);
+
+        BinPostSetupLinesImpl<SIMD_T, SIMD_WIDTH>(pDC, pa, workerId, line, recipW, triMask, primID, viewportIdx);
 
         AR_END(FEBinTriangles, 1);
         return;
@@ -1205,58 +795,61 @@ endBinTriangles:
     else if (rastState.fillMode == SWR_FILLMODE_POINT)
     {
         // Bin 3 points
-        BinPostSetupPoints_simd16(pDC, pa, workerId, &tri[0], triMask, primID, viewportIdx);
-        BinPostSetupPoints_simd16(pDC, pa, workerId, &tri[1], triMask, primID, viewportIdx);
-        BinPostSetupPoints_simd16(pDC, pa, workerId, &tri[2], triMask, primID, viewportIdx);
+        BinPostSetupPointsImpl<SIMD_T, SIMD_WIDTH>(pDC, pa, workerId, &tri[0], triMask, primID, viewportIdx);
+        BinPostSetupPointsImpl<SIMD_T, SIMD_WIDTH>(pDC, pa, workerId, &tri[1], triMask, primID, viewportIdx);
+        BinPostSetupPointsImpl<SIMD_T, SIMD_WIDTH>(pDC, pa, workerId, &tri[2], triMask, primID, viewportIdx);
+
+        AR_END(FEBinTriangles, 1);
         return;
     }
 
     // Convert triangle bbox to macrotile units.
-    bbox.xmin = _simd16_srai_epi32(bbox.xmin, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-    bbox.ymin = _simd16_srai_epi32(bbox.ymin, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-    bbox.xmax = _simd16_srai_epi32(bbox.xmax, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-    bbox.ymax = _simd16_srai_epi32(bbox.ymax, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
+#if SIMD_WA_SXXI_EPI32
+    bbox.xmin = simd_wa_srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmin);
+    bbox.ymin = simd_wa_srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymin);
+    bbox.xmax = simd_wa_srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmax);
+    bbox.ymax = simd_wa_srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymax);
+#else
+    bbox.xmin = SIMD_T::srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmin);
+    bbox.ymin = SIMD_T::srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymin);
+    bbox.xmax = SIMD_T::srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmax);
+    bbox.ymax = SIMD_T::srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymax);
+#endif
 
-    OSALIGNSIMD16(uint32_t) aMTLeft[KNOB_SIMD16_WIDTH], aMTRight[KNOB_SIMD16_WIDTH], aMTTop[KNOB_SIMD16_WIDTH], aMTBottom[KNOB_SIMD16_WIDTH];
+    OSALIGNSIMD16(uint32_t) aMTLeft[SIMD_WIDTH], aMTRight[SIMD_WIDTH], aMTTop[SIMD_WIDTH], aMTBottom[SIMD_WIDTH];
 
-    _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTLeft), bbox.xmin);
-    _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTRight), bbox.xmax);
-    _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTTop), bbox.ymin);
-    _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTBottom), bbox.ymax);
+    SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTLeft),   bbox.xmin);
+    SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTRight),  bbox.xmax);
+    SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTTop),    bbox.ymin);
+    SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTBottom), bbox.ymax);
 
     // transpose verts needed for backend
     /// @todo modify BE to take non-transformed verts
-    simd4scalar vHorizX[2][KNOB_SIMD_WIDTH]; // KNOB_SIMD16_WIDTH
-    simd4scalar vHorizY[2][KNOB_SIMD_WIDTH]; // KNOB_SIMD16_WIDTH
-    simd4scalar vHorizZ[2][KNOB_SIMD_WIDTH]; // KNOB_SIMD16_WIDTH
-    simd4scalar vHorizW[2][KNOB_SIMD_WIDTH]; // KNOB_SIMD16_WIDTH
-
-    vTranspose3x8(vHorizX[0], _simd16_extract_ps(tri[0].x, 0), _simd16_extract_ps(tri[1].x, 0), _simd16_extract_ps(tri[2].x, 0));
-    vTranspose3x8(vHorizY[0], _simd16_extract_ps(tri[0].y, 0), _simd16_extract_ps(tri[1].y, 0), _simd16_extract_ps(tri[2].y, 0));
-    vTranspose3x8(vHorizZ[0], _simd16_extract_ps(tri[0].z, 0), _simd16_extract_ps(tri[1].z, 0), _simd16_extract_ps(tri[2].z, 0));
-    vTranspose3x8(vHorizW[0], _simd16_extract_ps(vRecipW0, 0), _simd16_extract_ps(vRecipW1, 0), _simd16_extract_ps(vRecipW2, 0));
+    simd4scalar vHorizX[SIMD_WIDTH];
+    simd4scalar vHorizY[SIMD_WIDTH];
+    simd4scalar vHorizZ[SIMD_WIDTH];
+    simd4scalar vHorizW[SIMD_WIDTH];
 
-    vTranspose3x8(vHorizX[1], _simd16_extract_ps(tri[0].x, 1), _simd16_extract_ps(tri[1].x, 1), _simd16_extract_ps(tri[2].x, 1));
-    vTranspose3x8(vHorizY[1], _simd16_extract_ps(tri[0].y, 1), _simd16_extract_ps(tri[1].y, 1), _simd16_extract_ps(tri[2].y, 1));
-    vTranspose3x8(vHorizZ[1], _simd16_extract_ps(tri[0].z, 1), _simd16_extract_ps(tri[1].z, 1), _simd16_extract_ps(tri[2].z, 1));
-    vTranspose3x8(vHorizW[1], _simd16_extract_ps(vRecipW0, 1), _simd16_extract_ps(vRecipW1, 1), _simd16_extract_ps(vRecipW2, 1));
+    TransposeVertices(vHorizX, tri[0].x, tri[1].x, tri[2].x);
+    TransposeVertices(vHorizY, tri[0].y, tri[1].y, tri[2].y);
+    TransposeVertices(vHorizZ, tri[0].z, tri[1].z, tri[2].z);
+    TransposeVertices(vHorizW, vRecipW0, vRecipW1, vRecipW2);
 
     // store render target array index
-    OSALIGNSIMD16(uint32_t) aRTAI[KNOB_SIMD16_WIDTH];
+    OSALIGNSIMD16(uint32_t) aRTAI[SIMD_WIDTH];
     if (state.backendState.readRenderTargetArrayIndex)
     {
-        simd16vector vRtai[3];
-        pa.Assemble_simd16(VERTEX_SGV_SLOT, vRtai);
-        simd16scalari vRtaii;
-        vRtaii = _simd16_castps_si(vRtai[0][VERTEX_SGV_RTAI_COMP]);
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aRTAI), vRtaii);
+        typename SIMD_T::Vec4 vRtai[3];
+        pa.Assemble(VERTEX_SGV_SLOT, vRtai);
+        typename SIMD_T::Integer vRtaii;
+        vRtaii = SIMD_T::castps_si(vRtai[0][VERTEX_SGV_RTAI_COMP]);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aRTAI), vRtaii);
     }
     else
     {
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aRTAI), _simd16_setzero_si());
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aRTAI), SIMD_T::setzero_si());
     }
 
-
     // scan remaining valid triangles and bin each separately
     while (_BitScanForward(&triIndex, triMask))
     {
@@ -1305,15 +898,10 @@ endBinTriangles:
         // store triangle vertex data
         desc.pTriBuffer = (float*)pArena->AllocAligned(4 * 4 * sizeof(float), 16);
 
-        {
-            const uint32_t i = triIndex >> 3;   // triIndex / KNOB_SIMD_WIDTH
-            const uint32_t j = triIndex & 7;    // triIndex % KNOB_SIMD_WIDTH
-
-            _mm_store_ps(&desc.pTriBuffer[ 0], vHorizX[i][j]);
-            _mm_store_ps(&desc.pTriBuffer[ 4], vHorizY[i][j]);
-            _mm_store_ps(&desc.pTriBuffer[ 8], vHorizZ[i][j]);
-            _mm_store_ps(&desc.pTriBuffer[12], vHorizW[i][j]);
-        }
+        SIMD128::store_ps(&desc.pTriBuffer[0],  vHorizX[triIndex]);
+        SIMD128::store_ps(&desc.pTriBuffer[4],  vHorizY[triIndex]);
+        SIMD128::store_ps(&desc.pTriBuffer[8],  vHorizZ[triIndex]);
+        SIMD128::store_ps(&desc.pTriBuffer[12], vHorizW[triIndex]);
 
         // store user clip distances
         if (rastState.clipDistanceMask)
@@ -1336,12 +924,37 @@ endBinTriangles:
             }
         }
 
-        triMask &= ~(1 << triIndex);
+                     triMask &= ~(1 << triIndex);
     }
 
     AR_END(FEBinTriangles, 1);
 }
 
+template <typename CT>
+void BinTriangles(
+    DRAW_CONTEXT *pDC,
+    PA_STATE &pa,
+    uint32_t workerId,
+    simdvector tri[3],
+    uint32_t triMask,
+    simdscalari const &primID)
+{
+    BinTrianglesImpl<SIMD256, KNOB_SIMD_WIDTH, CT>(pDC, pa, workerId, tri, triMask, primID);
+}
+
+#if USE_SIMD16_FRONTEND
+template <typename CT>
+void SIMDCALL BinTriangles_simd16(
+    DRAW_CONTEXT *pDC,
+    PA_STATE &pa,
+    uint32_t workerId,
+    simd16vector tri[3],
+    uint32_t triMask,
+    simd16scalari const &primID)
+{
+    BinTrianglesImpl<SIMD512, KNOB_SIMD16_WIDTH, CT>(pDC, pa, workerId, tri, triMask, primID);
+}
+
 #endif
 struct FEBinTrianglesChooser
 {
@@ -1380,20 +993,21 @@ PFN_PROCESS_PRIMS_SIMD16 GetBinTrianglesFunc_simd16(bool IsConservative)
 
 #endif
 
-void BinPostSetupPoints(
+template <typename SIMD_T, uint32_t SIMD_WIDTH>
+void BinPostSetupPointsImpl(
     DRAW_CONTEXT *pDC,
-    PA_STATE& pa,
+    PA_STATE &pa,
     uint32_t workerId,
-    simdvector prim[],
+    typename SIMD_T::Vec4 prim[],
     uint32_t primMask,
-    simdscalari const &primID,
-    simdscalari const &viewportIdx)
+    typename SIMD_T::Integer const &primID,
+    typename SIMD_T::Integer const &viewportIdx)
 {
     SWR_CONTEXT *pContext = pDC->pContext;
 
     AR_BEGIN(FEBinPoints, pDC->drawId);
 
-    simdvector& primVerts = prim[0];
+    typename SIMD_T::Vec4 &primVerts = prim[0];
 
     const API_STATE& state = GetApiState(pDC);
     const SWR_RASTSTATE& rastState = state.rastState;
@@ -1404,64 +1018,88 @@ void BinPostSetupPoints(
         state.backendState.swizzleEnable, state.backendState.constantInterpolationMask);
 
     // convert to fixed point
-    simdscalari vXi, vYi;
-    vXi = fpToFixedPointVertical(primVerts.x);
-    vYi = fpToFixedPointVertical(primVerts.y);
+    typename SIMD_T::Integer vXi, vYi;
+
+    vXi = fpToFixedPointVertical<SIMD_T>(primVerts.x);
+    vYi = fpToFixedPointVertical<SIMD_T>(primVerts.y);
 
     if (CanUseSimplePoints(pDC))
     {
         // adjust for ymin-xmin rule
-        vXi = _simd_sub_epi32(vXi, _simd_set1_epi32(1));
-        vYi = _simd_sub_epi32(vYi, _simd_set1_epi32(1));
+        vXi = SIMD_T::sub_epi32(vXi, SIMD_T::set1_epi32(1));
+        vYi = SIMD_T::sub_epi32(vYi, SIMD_T::set1_epi32(1));
 
         // cull points off the ymin-xmin edge of the viewport
-        primMask &= ~_simd_movemask_ps(_simd_castsi_ps(vXi));
-        primMask &= ~_simd_movemask_ps(_simd_castsi_ps(vYi));
+        primMask &= ~SIMD_T::movemask_ps(SIMD_T::castsi_ps(vXi));
+        primMask &= ~SIMD_T::movemask_ps(SIMD_T::castsi_ps(vYi));
 
         // compute macro tile coordinates 
-        simdscalari macroX = _simd_srai_epi32(vXi, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-        simdscalari macroY = _simd_srai_epi32(vYi, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
+#if SIMD_WA_SXXI_EPI32
+        typename SIMD_T::Integer macroX = simd_wa_srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(vXi);
+        typename SIMD_T::Integer macroY = simd_wa_srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(vYi);
+#else
+        typename SIMD_T::Integer macroX = SIMD_T::srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(vXi);
+        typename SIMD_T::Integer macroY = SIMD_T::srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(vYi);
+#endif
+
+        OSALIGNSIMD16(uint32_t) aMacroX[SIMD_WIDTH], aMacroY[SIMD_WIDTH];
 
-        OSALIGNSIMD(uint32_t) aMacroX[KNOB_SIMD_WIDTH], aMacroY[KNOB_SIMD_WIDTH];
-        _simd_store_si((simdscalari*)aMacroX, macroX);
-        _simd_store_si((simdscalari*)aMacroY, macroY);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMacroX), macroX);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMacroY), macroY);
 
         // compute raster tile coordinates
-        simdscalari rasterX = _simd_srai_epi32(vXi, KNOB_TILE_X_DIM_SHIFT + FIXED_POINT_SHIFT);
-        simdscalari rasterY = _simd_srai_epi32(vYi, KNOB_TILE_Y_DIM_SHIFT + FIXED_POINT_SHIFT);
+#if SIMD_WA_SXXI_EPI32
+        typename SIMD_T::Integer rasterX = simd_wa_srai_epi32<KNOB_TILE_X_DIM_SHIFT + FIXED_POINT_SHIFT>(vXi);
+        typename SIMD_T::Integer rasterY = simd_wa_srai_epi32<KNOB_TILE_Y_DIM_SHIFT + FIXED_POINT_SHIFT>(vYi);
+#else
+        typename SIMD_T::Integer rasterX = SIMD_T::srai_epi32<KNOB_TILE_X_DIM_SHIFT + FIXED_POINT_SHIFT>(vXi);
+        typename SIMD_T::Integer rasterY = SIMD_T::srai_epi32<KNOB_TILE_Y_DIM_SHIFT + FIXED_POINT_SHIFT>(vYi);
+#endif
 
         // compute raster tile relative x,y for coverage mask
-        simdscalari tileAlignedX = _simd_slli_epi32(rasterX, KNOB_TILE_X_DIM_SHIFT);
-        simdscalari tileAlignedY = _simd_slli_epi32(rasterY, KNOB_TILE_Y_DIM_SHIFT);
+#if SIMD_WA_SXXI_EPI32
+        typename SIMD_T::Integer tileAlignedX = simd_wa_slli_epi32<KNOB_TILE_X_DIM_SHIFT>(rasterX);
+        typename SIMD_T::Integer tileAlignedY = simd_wa_slli_epi32<KNOB_TILE_Y_DIM_SHIFT>(rasterY);
+#else
+        typename SIMD_T::Integer tileAlignedX = SIMD_T::slli_epi32<KNOB_TILE_X_DIM_SHIFT>(rasterX);
+        typename SIMD_T::Integer tileAlignedY = SIMD_T::slli_epi32<KNOB_TILE_Y_DIM_SHIFT>(rasterY);
+#endif
+
+#if SIMD_WA_SXXI_EPI32
+        typename SIMD_T::Integer tileRelativeX = SIMD_T::sub_epi32(simd_wa_srai_epi32<FIXED_POINT_SHIFT>(vXi), tileAlignedX);
+        typename SIMD_T::Integer tileRelativeY = SIMD_T::sub_epi32(simd_wa_srai_epi32<FIXED_POINT_SHIFT>(vYi), tileAlignedY);
+#else
+        typename SIMD_T::Integer tileRelativeX = SIMD_T::sub_epi32(SIMD_T::srai_epi32<FIXED_POINT_SHIFT>(vXi), tileAlignedX);
+        typename SIMD_T::Integer tileRelativeY = SIMD_T::sub_epi32(SIMD_T::srai_epi32<FIXED_POINT_SHIFT>(vYi), tileAlignedY);
+#endif
+
+        OSALIGNSIMD16(uint32_t) aTileRelativeX[SIMD_WIDTH];
+        OSALIGNSIMD16(uint32_t) aTileRelativeY[SIMD_WIDTH];
 
-        simdscalari tileRelativeX = _simd_sub_epi32(_simd_srai_epi32(vXi, FIXED_POINT_SHIFT), tileAlignedX);
-        simdscalari tileRelativeY = _simd_sub_epi32(_simd_srai_epi32(vYi, FIXED_POINT_SHIFT), tileAlignedY);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aTileRelativeX), tileRelativeX);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aTileRelativeY), tileRelativeY);
 
-        OSALIGNSIMD(uint32_t) aTileRelativeX[KNOB_SIMD_WIDTH];
-        OSALIGNSIMD(uint32_t) aTileRelativeY[KNOB_SIMD_WIDTH];
-        _simd_store_si((simdscalari*)aTileRelativeX, tileRelativeX);
-        _simd_store_si((simdscalari*)aTileRelativeY, tileRelativeY);
+        OSALIGNSIMD16(uint32_t) aTileAlignedX[SIMD_WIDTH];
+        OSALIGNSIMD16(uint32_t) aTileAlignedY[SIMD_WIDTH];
 
-        OSALIGNSIMD(uint32_t) aTileAlignedX[KNOB_SIMD_WIDTH];
-        OSALIGNSIMD(uint32_t) aTileAlignedY[KNOB_SIMD_WIDTH];
-        _simd_store_si((simdscalari*)aTileAlignedX, tileAlignedX);
-        _simd_store_si((simdscalari*)aTileAlignedY, tileAlignedY);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aTileAlignedX), tileAlignedX);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aTileAlignedY), tileAlignedY);
 
-        OSALIGNSIMD(float) aZ[KNOB_SIMD_WIDTH];
-        _simd_store_ps((float*)aZ, primVerts.z);
+        OSALIGNSIMD16(float) aZ[SIMD_WIDTH];
+        SIMD_T::store_ps(reinterpret_cast<float *>(aZ), primVerts.z);
 
         // store render target array index
-        OSALIGNSIMD(uint32_t) aRTAI[KNOB_SIMD_WIDTH];
+        OSALIGNSIMD16(uint32_t) aRTAI[SIMD_WIDTH];
         if (state.backendState.readRenderTargetArrayIndex)
         {
-            simdvector vRtai;
+            typename SIMD_T::Vec4 vRtai;
             pa.Assemble(VERTEX_SGV_SLOT, &vRtai);
-            simdscalari vRtaii = _simd_castps_si(vRtai[VERTEX_SGV_RTAI_COMP]);
-            _simd_store_si((simdscalari*)aRTAI, vRtaii);
+            typename SIMD_T::Integer vRtaii = SIMD_T::castps_si(vRtai[VERTEX_SGV_RTAI_COMP]);
+            SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aRTAI), vRtaii);
         }
         else
         {
-            _simd_store_si((simdscalari*)aRTAI, _simd_setzero_si());
+            SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aRTAI), SIMD_T::setzero_si());
         }
 
         uint32_t *pPrimID = (uint32_t *)&primID;
@@ -1519,105 +1157,117 @@ void BinPostSetupPoints(
             {
                 pTileMgr->enqueue(aMacroX[primIndex], aMacroY[primIndex], &work);
             }
+
             primMask &= ~(1 << primIndex);
         }
     }
     else
     {
         // non simple points need to be potentially binned to multiple macro tiles
-        simdscalar vPointSize;
+        typename SIMD_T::Float vPointSize;
+
         if (rastState.pointParam)
         {
-            simdvector size[3];
+            typename SIMD_T::Vec4 size[3];
             pa.Assemble(VERTEX_SGV_SLOT, size);
             vPointSize = size[0][VERTEX_SGV_POINT_SIZE_COMP];
         }
         else
         {
-            vPointSize = _simd_set1_ps(rastState.pointSize);
+            vPointSize = SIMD_T::set1_ps(rastState.pointSize);
         }
 
         // bloat point to bbox
-        simdBBox bbox;
+        SIMDBBOX_T<SIMD_T> bbox;
+
         bbox.xmin = bbox.xmax = vXi;
         bbox.ymin = bbox.ymax = vYi;
 
-        simdscalar vHalfWidth = _simd_mul_ps(vPointSize, _simd_set1_ps(0.5f));
-        simdscalari vHalfWidthi = fpToFixedPointVertical(vHalfWidth);
-        bbox.xmin = _simd_sub_epi32(bbox.xmin, vHalfWidthi);
-        bbox.xmax = _simd_add_epi32(bbox.xmax, vHalfWidthi);
-        bbox.ymin = _simd_sub_epi32(bbox.ymin, vHalfWidthi);
-        bbox.ymax = _simd_add_epi32(bbox.ymax, vHalfWidthi);
+        typename SIMD_T::Float vHalfWidth = SIMD_T::mul_ps(vPointSize, SIMD_T::set1_ps(0.5f));
+        typename SIMD_T::Integer vHalfWidthi = fpToFixedPointVertical<SIMD_T>(vHalfWidth);
+
+        bbox.xmin = SIMD_T::sub_epi32(bbox.xmin, vHalfWidthi);
+        bbox.xmax = SIMD_T::add_epi32(bbox.xmax, vHalfWidthi);
+        bbox.ymin = SIMD_T::sub_epi32(bbox.ymin, vHalfWidthi);
+        bbox.ymax = SIMD_T::add_epi32(bbox.ymax, vHalfWidthi);
 
         // Intersect with scissor/viewport. Subtract 1 ULP in x.8 fixed point since xmax/ymax edge is exclusive.
         // Gather the AOS effective scissor rects based on the per-prim VP index.
         /// @todo:  Look at speeding this up -- weigh against corresponding costs in rasterizer.
         {
-            simdscalari scisXmin, scisYmin, scisXmax, scisYmax;
+            typename SIMD_T::Integer scisXmin, scisYmin, scisXmax, scisYmax;
+
             if (state.backendState.readViewportArrayIndex)
             {
-                GatherScissors<KNOB_SIMD_WIDTH>::Gather(&state.scissorsInFixedPoint[0], pViewportIndex,
-                    scisXmin, scisYmin, scisXmax, scisYmax);
+                GatherScissors(&state.scissorsInFixedPoint[0], pViewportIndex, scisXmin, scisYmin, scisXmax, scisYmax);
             }
             else // broadcast fast path for non-VPAI case.
             {
-                scisXmin = _simd_set1_epi32(state.scissorsInFixedPoint[0].xmin);
-                scisYmin = _simd_set1_epi32(state.scissorsInFixedPoint[0].ymin);
-                scisXmax = _simd_set1_epi32(state.scissorsInFixedPoint[0].xmax);
-                scisYmax = _simd_set1_epi32(state.scissorsInFixedPoint[0].ymax);
+                scisXmin = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].xmin);
+                scisYmin = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].ymin);
+                scisXmax = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].xmax);
+                scisYmax = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].ymax);
             }
 
-            bbox.xmin = _simd_max_epi32(bbox.xmin, scisXmin);
-            bbox.ymin = _simd_max_epi32(bbox.ymin, scisYmin);
-            bbox.xmax = _simd_min_epi32(_simd_sub_epi32(bbox.xmax, _simd_set1_epi32(1)), scisXmax);
-            bbox.ymax = _simd_min_epi32(_simd_sub_epi32(bbox.ymax, _simd_set1_epi32(1)), scisYmax);
+            bbox.xmin = SIMD_T::max_epi32(bbox.xmin, scisXmin);
+            bbox.ymin = SIMD_T::max_epi32(bbox.ymin, scisYmin);
+            bbox.xmax = SIMD_T::min_epi32(SIMD_T::sub_epi32(bbox.xmax, SIMD_T::set1_epi32(1)), scisXmax);
+            bbox.ymax = SIMD_T::min_epi32(SIMD_T::sub_epi32(bbox.ymax, SIMD_T::set1_epi32(1)), scisYmax);
         }
 
         // Cull bloated points completely outside scissor
-        simdscalari maskOutsideScissorX = _simd_cmpgt_epi32(bbox.xmin, bbox.xmax);
-        simdscalari maskOutsideScissorY = _simd_cmpgt_epi32(bbox.ymin, bbox.ymax);
-        simdscalari maskOutsideScissorXY = _simd_or_si(maskOutsideScissorX, maskOutsideScissorY);
-        uint32_t maskOutsideScissor = _simd_movemask_ps(_simd_castsi_ps(maskOutsideScissorXY));
+        typename SIMD_T::Integer maskOutsideScissorX = SIMD_T::cmpgt_epi32(bbox.xmin, bbox.xmax);
+        typename SIMD_T::Integer maskOutsideScissorY = SIMD_T::cmpgt_epi32(bbox.ymin, bbox.ymax);
+        typename SIMD_T::Integer maskOutsideScissorXY = SIMD_T::or_si(maskOutsideScissorX, maskOutsideScissorY);
+        uint32_t maskOutsideScissor = SIMD_T::movemask_ps(SIMD_T::castsi_ps(maskOutsideScissorXY));
         primMask = primMask & ~maskOutsideScissor;
 
         // Convert bbox to macrotile units.
-        bbox.xmin = _simd_srai_epi32(bbox.xmin, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-        bbox.ymin = _simd_srai_epi32(bbox.ymin, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-        bbox.xmax = _simd_srai_epi32(bbox.xmax, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-        bbox.ymax = _simd_srai_epi32(bbox.ymax, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
+#if SIMD_WA_SXXI_EPI32
+        bbox.xmin = simd_wa_srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmin);
+        bbox.ymin = simd_wa_srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymin);
+        bbox.xmax = simd_wa_srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmax);
+        bbox.ymax = simd_wa_srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymax);
+#else
+        bbox.xmin = SIMD_T::srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmin);
+        bbox.ymin = SIMD_T::srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymin);
+        bbox.xmax = SIMD_T::srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmax);
+        bbox.ymax = SIMD_T::srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymax);
+#endif
+
+        OSALIGNSIMD16(uint32_t) aMTLeft[SIMD_WIDTH], aMTRight[SIMD_WIDTH], aMTTop[SIMD_WIDTH], aMTBottom[SIMD_WIDTH];
 
-        OSALIGNSIMD(uint32_t) aMTLeft[KNOB_SIMD_WIDTH], aMTRight[KNOB_SIMD_WIDTH], aMTTop[KNOB_SIMD_WIDTH], aMTBottom[KNOB_SIMD_WIDTH];
-        _simd_store_si((simdscalari*)aMTLeft, bbox.xmin);
-        _simd_store_si((simdscalari*)aMTRight, bbox.xmax);
-        _simd_store_si((simdscalari*)aMTTop, bbox.ymin);
-        _simd_store_si((simdscalari*)aMTBottom, bbox.ymax);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTLeft),   bbox.xmin);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTRight),  bbox.xmax);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTTop),    bbox.ymin);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTBottom), bbox.ymax);
 
         // store render target array index
-        OSALIGNSIMD(uint32_t) aRTAI[KNOB_SIMD_WIDTH];
+        OSALIGNSIMD16(uint32_t) aRTAI[SIMD_WIDTH];
         if (state.backendState.readRenderTargetArrayIndex)
         {
-            simdvector vRtai[2];
+            typename SIMD_T::Vec4 vRtai[2];
             pa.Assemble(VERTEX_SGV_SLOT, vRtai);
-            simdscalari vRtaii = _simd_castps_si(vRtai[0][VERTEX_SGV_RTAI_COMP]);
-            _simd_store_si((simdscalari*)aRTAI, vRtaii);
+            typename SIMD_T::Integer vRtaii = SIMD_T::castps_si(vRtai[0][VERTEX_SGV_RTAI_COMP]);
+            SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aRTAI), vRtaii);
         }
         else
         {
-            _simd_store_si((simdscalari*)aRTAI, _simd_setzero_si());
+            SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aRTAI), SIMD_T::setzero_si());
         }
 
-        OSALIGNSIMD(float) aPointSize[KNOB_SIMD_WIDTH];
-        _simd_store_ps((float*)aPointSize, vPointSize);
+        OSALIGNSIMD16(float) aPointSize[SIMD_WIDTH];
+        _simd16_store_ps(reinterpret_cast<float *>(aPointSize), vPointSize);
 
         uint32_t *pPrimID = (uint32_t *)&primID;
 
-        OSALIGNSIMD(float) aPrimVertsX[KNOB_SIMD_WIDTH];
-        OSALIGNSIMD(float) aPrimVertsY[KNOB_SIMD_WIDTH];
-        OSALIGNSIMD(float) aPrimVertsZ[KNOB_SIMD_WIDTH];
+        OSALIGNSIMD16(float) aPrimVertsX[SIMD_WIDTH];
+        OSALIGNSIMD16(float) aPrimVertsY[SIMD_WIDTH];
+        OSALIGNSIMD16(float) aPrimVertsZ[SIMD_WIDTH];
 
-        _simd_store_ps((float*)aPrimVertsX, primVerts.x);
-        _simd_store_ps((float*)aPrimVertsY, primVerts.y);
-        _simd_store_ps((float*)aPrimVertsZ, primVerts.z);
+        SIMD_T::store_ps(reinterpret_cast<float *>(aPrimVertsX), primVerts.x);
+        SIMD_T::store_ps(reinterpret_cast<float *>(aPrimVertsY), primVerts.y);
+        SIMD_T::store_ps(reinterpret_cast<float *>(aPrimVertsZ), primVerts.z);
 
         // scan remaining valid prims and bin each separately
         const SWR_BACKEND_STATE& backendState = state.backendState;
@@ -1663,9 +1313,9 @@ void BinPostSetupPoints(
                 float one = 1.0f;
                 ProcessUserClipDist<1>(pa, primIndex, rastState.clipDistanceMask, &one, dists);
                 for (uint32_t i = 0; i < numClipDist; i++) {
-                    desc.pUserClipBuffer[3*i + 0] = 0.0f;
-                    desc.pUserClipBuffer[3*i + 1] = 0.0f;
-                    desc.pUserClipBuffer[3*i + 2] = dists[i];
+                    desc.pUserClipBuffer[3 * i + 0] = 0.0f;
+                    desc.pUserClipBuffer[3 * i + 1] = 0.0f;
+                    desc.pUserClipBuffer[3 * i + 2] = dists[i];
                 }
             }
 
@@ -1697,60 +1347,60 @@ void BinPostSetupPoints(
 /// @param workerId - thread's worker id. Even thread has a unique id.
 /// @param tri - Contains point position data for SIMDs worth of points.
 /// @param primID - Primitive ID for each point.
-void BinPoints(
+template <typename SIMD_T, uint32_t SIMD_WIDTH>
+void BinPointsImpl(
     DRAW_CONTEXT *pDC,
-    PA_STATE& pa,
+    PA_STATE &pa,
     uint32_t workerId,
-    simdvector prim[3],
+    typename SIMD_T::Vec4 prim[3],
     uint32_t primMask,
-    simdscalari const &primID)
+    typename SIMD_T::Integer const &primID)
 {
-    simdvector& primVerts = prim[0];
-
     const API_STATE& state = GetApiState(pDC);
     const SWR_FRONTEND_STATE& feState = state.frontendState;
     const SWR_RASTSTATE& rastState = state.rastState;
 
     // Read back viewport index if required
-    simdscalari viewportIdx = _simd_set1_epi32(0);
+    typename SIMD_T::Integer viewportIdx = SIMD_T::set1_epi32(0);
     if (state.backendState.readViewportArrayIndex)
     {
-        simdvector vpiAttrib[1];
+        typename SIMD_T::Vec4 vpiAttrib[1];
         pa.Assemble(VERTEX_SGV_SLOT, vpiAttrib);
-        simdscalari vpai = _simd_castps_si(vpiAttrib[0][VERTEX_SGV_VAI_COMP]);
 
         // OOB indices => forced to zero.
-        vpai = _simd_max_epi32(_simd_setzero_si(), vpai);
-        simdscalari vNumViewports = _simd_set1_epi32(KNOB_NUM_VIEWPORTS_SCISSORS);
-        simdscalari vClearMask = _simd_cmplt_epi32(vpai, vNumViewports);
-        viewportIdx = _simd_and_si(vClearMask, vpai);
+        typename SIMD_T::Integer vpai = SIMD_T::castps_si(vpiAttrib[0][VERTEX_SGV_VAI_COMP]);
+        vpai = SIMD_T::max_epi32(SIMD_T::setzero_si(), vpai);
+        typename SIMD_T::Integer vNumViewports = SIMD_T::set1_epi32(KNOB_NUM_VIEWPORTS_SCISSORS);
+        typename SIMD_T::Integer vClearMask = SIMD_T::cmplt_epi32(vpai, vNumViewports);
+        viewportIdx = SIMD_T::and_si(vClearMask, vpai);
     }
 
     if (!feState.vpTransformDisable)
     {
         // perspective divide
-        simdscalar vRecipW0 = _simd_div_ps(_simd_set1_ps(1.0f), primVerts.w);
-        primVerts.x = _simd_mul_ps(primVerts.x, vRecipW0);
-        primVerts.y = _simd_mul_ps(primVerts.y, vRecipW0);
-        primVerts.z = _simd_mul_ps(primVerts.z, vRecipW0);
+        typename SIMD_T::Float vRecipW0 = SIMD_T::div_ps(SIMD_T::set1_ps(1.0f), prim[0].w);
+
+        prim[0].x = SIMD_T::mul_ps(prim[0].x, vRecipW0);
+        prim[0].y = SIMD_T::mul_ps(prim[0].y, vRecipW0);
+        prim[0].z = SIMD_T::mul_ps(prim[0].z, vRecipW0);
 
         // viewport transform to screen coords
         if (state.backendState.readViewportArrayIndex)
         {
-            viewportTransform<1>(&primVerts, state.vpMatrices, viewportIdx);
+            viewportTransform<1>(prim, state.vpMatrices, viewportIdx);
         }
         else
         {
-            viewportTransform<1>(&primVerts, state.vpMatrices);
+            viewportTransform<1>(prim, state.vpMatrices);
         }
     }
 
-    // adjust for pixel center location
-    simdscalar offset = g_pixelOffsets[rastState.pixelLocation];
-    primVerts.x = _simd_add_ps(primVerts.x, offset);
-    primVerts.y = _simd_add_ps(primVerts.y, offset);
+    typename SIMD_T::Float offset = g_pixelOffsets<SIMD_T>[rastState.pixelLocation];
+
+    prim[0].x = SIMD_T::add_ps(prim[0].x, offset);
+    prim[0].y = SIMD_T::add_ps(prim[0].y, offset);
 
-    BinPostSetupPoints(
+    BinPostSetupPointsImpl<SIMD_T, SIMD_WIDTH>(
         pDC,
         pa,
         workerId,
@@ -1760,388 +1410,39 @@ void BinPoints(
         viewportIdx);
 }
 
-#if USE_SIMD16_FRONTEND
-void BinPostSetupPoints_simd16(
+void BinPoints(
     DRAW_CONTEXT *pDC,
-    PA_STATE& pa,
+    PA_STATE &pa,
     uint32_t workerId,
-    simd16vector prim[],
+    simdvector prim[3],
     uint32_t primMask,
-    simd16scalari const &primID,
-    simd16scalari const &viewportIdx)
+    simdscalari const &primID)
 {
-    SWR_CONTEXT *pContext = pDC->pContext;
-
-    AR_BEGIN(FEBinPoints, pDC->drawId);
-
-    simd16vector& primVerts = prim[0];
-
-    const API_STATE& state = GetApiState(pDC);
-    const SWR_RASTSTATE& rastState = state.rastState;
-    const uint32_t *pViewportIndex = (uint32_t *)&viewportIdx;
-
-    // Select attribute processor
-    PFN_PROCESS_ATTRIBUTES pfnProcessAttribs = GetProcessAttributesFunc(1,
-        state.backendState.swizzleEnable, state.backendState.constantInterpolationMask);
-
-    // convert to fixed point
-    simd16scalari vXi, vYi;
-
-    vXi = fpToFixedPointVertical(primVerts.x);
-    vYi = fpToFixedPointVertical(primVerts.y);
-
-    if (CanUseSimplePoints(pDC))
-    {
-        // adjust for ymin-xmin rule
-        vXi = _simd16_sub_epi32(vXi, _simd16_set1_epi32(1));
-        vYi = _simd16_sub_epi32(vYi, _simd16_set1_epi32(1));
-
-        // cull points off the ymin-xmin edge of the viewport
-        primMask &= ~_simd16_movemask_ps(_simd16_castsi_ps(vXi));
-        primMask &= ~_simd16_movemask_ps(_simd16_castsi_ps(vYi));
-
-        // compute macro tile coordinates 
-        simd16scalari macroX = _simd16_srai_epi32(vXi, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-        simd16scalari macroY = _simd16_srai_epi32(vYi, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-
-        OSALIGNSIMD16(uint32_t) aMacroX[KNOB_SIMD16_WIDTH], aMacroY[KNOB_SIMD16_WIDTH];
-
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aMacroX), macroX);
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aMacroY), macroY);
-
-        // compute raster tile coordinates
-        simd16scalari rasterX = _simd16_srai_epi32(vXi, KNOB_TILE_X_DIM_SHIFT + FIXED_POINT_SHIFT);
-        simd16scalari rasterY = _simd16_srai_epi32(vYi, KNOB_TILE_Y_DIM_SHIFT + FIXED_POINT_SHIFT);
-
-        // compute raster tile relative x,y for coverage mask
-        simd16scalari tileAlignedX = _simd16_slli_epi32(rasterX, KNOB_TILE_X_DIM_SHIFT);
-        simd16scalari tileAlignedY = _simd16_slli_epi32(rasterY, KNOB_TILE_Y_DIM_SHIFT);
-
-        simd16scalari tileRelativeX = _simd16_sub_epi32(_simd16_srai_epi32(vXi, FIXED_POINT_SHIFT), tileAlignedX);
-        simd16scalari tileRelativeY = _simd16_sub_epi32(_simd16_srai_epi32(vYi, FIXED_POINT_SHIFT), tileAlignedY);
-
-        OSALIGNSIMD16(uint32_t) aTileRelativeX[KNOB_SIMD16_WIDTH];
-        OSALIGNSIMD16(uint32_t) aTileRelativeY[KNOB_SIMD16_WIDTH];
-
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aTileRelativeX), tileRelativeX);
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aTileRelativeY), tileRelativeY);
-
-        OSALIGNSIMD16(uint32_t) aTileAlignedX[KNOB_SIMD16_WIDTH];
-        OSALIGNSIMD16(uint32_t) aTileAlignedY[KNOB_SIMD16_WIDTH];
-
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aTileAlignedX), tileAlignedX);
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aTileAlignedY), tileAlignedY);
-
-        OSALIGNSIMD16(float) aZ[KNOB_SIMD16_WIDTH];
-        _simd16_store_ps(reinterpret_cast<float *>(aZ), primVerts.z);
-
-        // store render target array index
-        OSALIGNSIMD16(uint32_t) aRTAI[KNOB_SIMD16_WIDTH];
-        if (state.backendState.readRenderTargetArrayIndex)
-        {
-            simd16vector vRtai;
-            pa.Assemble_simd16(VERTEX_SGV_SLOT, &vRtai);
-            simd16scalari vRtaii = _simd16_castps_si(vRtai[VERTEX_SGV_RTAI_COMP]);
-            _simd16_store_si(reinterpret_cast<simd16scalari *>(aRTAI), vRtaii);
-        }
-        else
-        {
-            _simd16_store_si(reinterpret_cast<simd16scalari *>(aRTAI), _simd16_setzero_si());
-        }
-
-        uint32_t *pPrimID = (uint32_t *)&primID;
-        DWORD primIndex = 0;
-
-        const SWR_BACKEND_STATE& backendState = pDC->pState->state.backendState;
-
-        // scan remaining valid triangles and bin each separately
-        while (_BitScanForward(&primIndex, primMask))
-        {
-            uint32_t linkageCount = backendState.numAttributes;
-            uint32_t numScalarAttribs = linkageCount * 4;
-
-            BE_WORK work;
-            work.type = DRAW;
-
-            TRIANGLE_WORK_DESC &desc = work.desc.tri;
-
-            // points are always front facing
-            desc.triFlags.frontFacing = 1;
-            desc.triFlags.renderTargetArrayIndex = aRTAI[primIndex];
-            desc.triFlags.viewportIndex = pViewportIndex[primIndex];
-
-            work.pfnWork = RasterizeSimplePoint;
-
-            auto pArena = pDC->pArena;
-            SWR_ASSERT(pArena != nullptr);
-
-            // store attributes
-            float *pAttribs = (float*)pArena->AllocAligned(3 * numScalarAttribs * sizeof(float), 16);
-            desc.pAttribs = pAttribs;
-            desc.numAttribs = linkageCount;
-
-            pfnProcessAttribs(pDC, pa, primIndex, pPrimID[primIndex], pAttribs);
-
-            // store raster tile aligned x, y, perspective correct z
-            float *pTriBuffer = (float*)pArena->AllocAligned(4 * sizeof(float), 16);
-            desc.pTriBuffer = pTriBuffer;
-            *(uint32_t*)pTriBuffer++ = aTileAlignedX[primIndex];
-            *(uint32_t*)pTriBuffer++ = aTileAlignedY[primIndex];
-            *pTriBuffer = aZ[primIndex];
-
-            uint32_t tX = aTileRelativeX[primIndex];
-            uint32_t tY = aTileRelativeY[primIndex];
-
-            // pack the relative x,y into the coverageMask, the rasterizer will
-            // generate the true coverage mask from it
-            work.desc.tri.triFlags.coverageMask = tX | (tY << 4);
-
-            // bin it
-            MacroTileMgr *pTileMgr = pDC->pTileMgr;
-#if KNOB_ENABLE_TOSS_POINTS
-            if (!KNOB_TOSS_SETUP_TRIS)
-#endif
-            {
-                pTileMgr->enqueue(aMacroX[primIndex], aMacroY[primIndex], &work);
-            }
-
-            primMask &= ~(1 << primIndex);
-        }
-    }
-    else
-    {
-        // non simple points need to be potentially binned to multiple macro tiles
-        simd16scalar vPointSize;
-
-        if (rastState.pointParam)
-        {
-            simd16vector size[3];
-            pa.Assemble_simd16(VERTEX_SGV_SLOT, size);
-            vPointSize = size[0][VERTEX_SGV_POINT_SIZE_COMP];
-        }
-        else
-        {
-            vPointSize = _simd16_set1_ps(rastState.pointSize);
-        }
-
-        // bloat point to bbox
-        simd16BBox bbox;
-
-        bbox.xmin = bbox.xmax = vXi;
-        bbox.ymin = bbox.ymax = vYi;
-
-        simd16scalar vHalfWidth = _simd16_mul_ps(vPointSize, _simd16_set1_ps(0.5f));
-        simd16scalari vHalfWidthi = fpToFixedPointVertical(vHalfWidth);
-
-        bbox.xmin = _simd16_sub_epi32(bbox.xmin, vHalfWidthi);
-        bbox.xmax = _simd16_add_epi32(bbox.xmax, vHalfWidthi);
-        bbox.ymin = _simd16_sub_epi32(bbox.ymin, vHalfWidthi);
-        bbox.ymax = _simd16_add_epi32(bbox.ymax, vHalfWidthi);
-
-        // Intersect with scissor/viewport. Subtract 1 ULP in x.8 fixed point since xmax/ymax edge is exclusive.
-        // Gather the AOS effective scissor rects based on the per-prim VP index.
-        /// @todo:  Look at speeding this up -- weigh against corresponding costs in rasterizer.
-        {
-            simd16scalari scisXmin, scisYmin, scisXmax, scisYmax;
-            if (state.backendState.readViewportArrayIndex)
-            {
-                GatherScissors_simd16<KNOB_SIMD16_WIDTH>::Gather(&state.scissorsInFixedPoint[0], pViewportIndex,
-                    scisXmin, scisYmin, scisXmax, scisYmax);
-            }
-            else // broadcast fast path for non-VPAI case.
-            {
-                scisXmin = _simd16_set1_epi32(state.scissorsInFixedPoint[0].xmin);
-                scisYmin = _simd16_set1_epi32(state.scissorsInFixedPoint[0].ymin);
-                scisXmax = _simd16_set1_epi32(state.scissorsInFixedPoint[0].xmax);
-                scisYmax = _simd16_set1_epi32(state.scissorsInFixedPoint[0].ymax);
-            }
-
-            bbox.xmin = _simd16_max_epi32(bbox.xmin, scisXmin);
-            bbox.ymin = _simd16_max_epi32(bbox.ymin, scisYmin);
-            bbox.xmax = _simd16_min_epi32(_simd16_sub_epi32(bbox.xmax, _simd16_set1_epi32(1)), scisXmax);
-            bbox.ymax = _simd16_min_epi32(_simd16_sub_epi32(bbox.ymax, _simd16_set1_epi32(1)), scisYmax);
-        }
-
-        // Cull bloated points completely outside scissor
-        simd16scalari maskOutsideScissorX = _simd16_cmpgt_epi32(bbox.xmin, bbox.xmax);
-        simd16scalari maskOutsideScissorY = _simd16_cmpgt_epi32(bbox.ymin, bbox.ymax);
-        simd16scalari maskOutsideScissorXY = _simd16_or_si(maskOutsideScissorX, maskOutsideScissorY);
-        uint32_t maskOutsideScissor = _simd16_movemask_ps(_simd16_castsi_ps(maskOutsideScissorXY));
-        primMask = primMask & ~maskOutsideScissor;
-
-        // Convert bbox to macrotile units.
-        bbox.xmin = _simd16_srai_epi32(bbox.xmin, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-        bbox.ymin = _simd16_srai_epi32(bbox.ymin, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-        bbox.xmax = _simd16_srai_epi32(bbox.xmax, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-        bbox.ymax = _simd16_srai_epi32(bbox.ymax, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-
-        OSALIGNSIMD16(uint32_t) aMTLeft[KNOB_SIMD16_WIDTH], aMTRight[KNOB_SIMD16_WIDTH], aMTTop[KNOB_SIMD16_WIDTH], aMTBottom[KNOB_SIMD16_WIDTH];
-
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTLeft),      bbox.xmin);
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTRight),     bbox.xmax);
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTTop),       bbox.ymin);
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTBottom),    bbox.ymax);
-
-        // store render target array index
-        OSALIGNSIMD16(uint32_t) aRTAI[KNOB_SIMD16_WIDTH];
-        if (state.backendState.readRenderTargetArrayIndex)
-        {
-            simd16vector vRtai[2];
-            pa.Assemble_simd16(VERTEX_SGV_SLOT, vRtai);
-            simd16scalari vRtaii = _simd16_castps_si(vRtai[0][VERTEX_SGV_RTAI_COMP]);
-            _simd16_store_si(reinterpret_cast<simd16scalari *>(aRTAI), vRtaii);
-        }
-        else
-        {
-            _simd16_store_si(reinterpret_cast<simd16scalari *>(aRTAI), _simd16_setzero_si());
-        }
-
-        OSALIGNSIMD16(float) aPointSize[KNOB_SIMD16_WIDTH];
-        _simd16_store_ps(reinterpret_cast<float *>(aPointSize), vPointSize);
-
-        uint32_t *pPrimID = (uint32_t *)&primID;
-
-        OSALIGNSIMD16(float) aPrimVertsX[KNOB_SIMD16_WIDTH];
-        OSALIGNSIMD16(float) aPrimVertsY[KNOB_SIMD16_WIDTH];
-        OSALIGNSIMD16(float) aPrimVertsZ[KNOB_SIMD16_WIDTH];
-
-        _simd16_store_ps(reinterpret_cast<float *>(aPrimVertsX), primVerts.x);
-        _simd16_store_ps(reinterpret_cast<float *>(aPrimVertsY), primVerts.y);
-        _simd16_store_ps(reinterpret_cast<float *>(aPrimVertsZ), primVerts.z);
-
-        // scan remaining valid prims and bin each separately
-        const SWR_BACKEND_STATE& backendState = state.backendState;
-        DWORD primIndex;
-        while (_BitScanForward(&primIndex, primMask))
-        {
-            uint32_t linkageCount = backendState.numAttributes;
-            uint32_t numScalarAttribs = linkageCount * 4;
-
-            BE_WORK work;
-            work.type = DRAW;
-
-            TRIANGLE_WORK_DESC &desc = work.desc.tri;
-
-            desc.triFlags.frontFacing = 1;
-            desc.triFlags.pointSize = aPointSize[primIndex];
-            desc.triFlags.renderTargetArrayIndex = aRTAI[primIndex];
-            desc.triFlags.viewportIndex = pViewportIndex[primIndex];
-
-            work.pfnWork = RasterizeTriPoint;
-
-            auto pArena = pDC->pArena;
-            SWR_ASSERT(pArena != nullptr);
-
-            // store active attribs
-            desc.pAttribs = (float*)pArena->AllocAligned(numScalarAttribs * 3 * sizeof(float), 16);
-            desc.numAttribs = linkageCount;
-            pfnProcessAttribs(pDC, pa, primIndex, pPrimID[primIndex], desc.pAttribs);
-
-            // store point vertex data
-            float *pTriBuffer = (float*)pArena->AllocAligned(4 * sizeof(float), 16);
-            desc.pTriBuffer = pTriBuffer;
-            *pTriBuffer++ = aPrimVertsX[primIndex];
-            *pTriBuffer++ = aPrimVertsY[primIndex];
-            *pTriBuffer = aPrimVertsZ[primIndex];
-
-            // store user clip distances
-            if (rastState.clipDistanceMask)
-            {
-                uint32_t numClipDist = _mm_popcnt_u32(rastState.clipDistanceMask);
-                desc.pUserClipBuffer = (float*)pArena->Alloc(numClipDist * 3 * sizeof(float));
-                float dists[8];
-                float one = 1.0f;
-                ProcessUserClipDist<1>(pa, primIndex, rastState.clipDistanceMask, &one, dists);
-                for (uint32_t i = 0; i < numClipDist; i++) {
-                    desc.pUserClipBuffer[3 * i + 0] = 0.0f;
-                    desc.pUserClipBuffer[3 * i + 1] = 0.0f;
-                    desc.pUserClipBuffer[3 * i + 2] = dists[i];
-                }
-            }
-
-            MacroTileMgr *pTileMgr = pDC->pTileMgr;
-            for (uint32_t y = aMTTop[primIndex]; y <= aMTBottom[primIndex]; ++y)
-            {
-                for (uint32_t x = aMTLeft[primIndex]; x <= aMTRight[primIndex]; ++x)
-                {
-#if KNOB_ENABLE_TOSS_POINTS
-                    if (!KNOB_TOSS_SETUP_TRIS)
-#endif
-                    {
-                        pTileMgr->enqueue(x, y, &work);
-                    }
-                }
-            }
-
-            primMask &= ~(1 << primIndex);
-        }
-    }
-
-    AR_END(FEBinPoints, 1);
+    BinPointsImpl<SIMD256, KNOB_SIMD_WIDTH>(
+        pDC,
+        pa,
+        workerId,
+        prim,
+        primMask,
+        primID);
 }
 
+#if USE_SIMD16_FRONTEND
 void SIMDCALL BinPoints_simd16(
     DRAW_CONTEXT *pDC,
-    PA_STATE& pa,
+    PA_STATE &pa,
     uint32_t workerId,
     simd16vector prim[3],
     uint32_t primMask,
     simd16scalari const &primID)
 {
-    simd16vector& primVerts = prim[0];
-
-    const API_STATE& state = GetApiState(pDC);
-    const SWR_FRONTEND_STATE& feState = state.frontendState;
-    const SWR_RASTSTATE& rastState = state.rastState;
-
-    // Read back viewport index if required
-    simd16scalari viewportIdx = _simd16_set1_epi32(0);
-    if (state.backendState.readViewportArrayIndex)
-    {
-        simd16vector vpiAttrib[1];
-        pa.Assemble_simd16(VERTEX_SGV_SLOT, vpiAttrib);
-
-        // OOB indices => forced to zero.
-        simd16scalari vpai = _simd16_castps_si(vpiAttrib[0][VERTEX_SGV_VAI_COMP]);
-        vpai = _simd16_max_epi32(_simd16_setzero_si(), vpai);
-        simd16scalari vNumViewports = _simd16_set1_epi32(KNOB_NUM_VIEWPORTS_SCISSORS);
-        simd16scalari vClearMask = _simd16_cmplt_epi32(vpai, vNumViewports);
-        viewportIdx = _simd16_and_si(vClearMask, vpai);
-    }
-
-    if (!feState.vpTransformDisable)
-    {
-        // perspective divide
-        simd16scalar vRecipW0 = _simd16_div_ps(_simd16_set1_ps(1.0f), primVerts.w);
-
-        primVerts.x = _simd16_mul_ps(primVerts.x, vRecipW0);
-        primVerts.y = _simd16_mul_ps(primVerts.y, vRecipW0);
-        primVerts.z = _simd16_mul_ps(primVerts.z, vRecipW0);
-
-        // viewport transform to screen coords
-        if (state.backendState.readViewportArrayIndex)
-        {
-            viewportTransform<1>(&primVerts, state.vpMatrices, viewportIdx);
-        }
-        else
-        {
-            viewportTransform<1>(&primVerts, state.vpMatrices);
-        }
-    }
-
-    const simd16scalar offset = g_pixelOffsets_simd16[rastState.pixelLocation];
-
-    primVerts.x = _simd16_add_ps(primVerts.x, offset);
-    primVerts.y = _simd16_add_ps(primVerts.y, offset);
-
-    BinPostSetupPoints_simd16(
+    BinPointsImpl<SIMD512, KNOB_SIMD16_WIDTH>(
         pDC,
         pa,
         workerId,
         prim,
         primMask,
-        primID,
-        viewportIdx);
+        primID);
 }
 
 #endif
@@ -2153,321 +1454,114 @@ void SIMDCALL BinPoints_simd16(
 /// @param tri - Contains line position data for SIMDs worth of points.
 /// @param primID - Primitive ID for each line.
 /// @param viewportIdx - Viewport Array Index for each line.
-void BinPostSetupLines(
+template <typename SIMD_T, uint32_t SIMD_WIDTH>
+void BinPostSetupLinesImpl(
     DRAW_CONTEXT *pDC,
-    PA_STATE& pa,
+    PA_STATE &pa,
     uint32_t workerId,
-    simdvector prim[],
-    simdscalar recipW[],
+    typename SIMD_T::Vec4 prim[],
+    typename SIMD_T::Float recipW[],
     uint32_t primMask,
-    simdscalari const &primID,
-    simdscalari const &viewportIdx)
+    typename SIMD_T::Integer const &primID,
+    typename SIMD_T::Integer const &viewportIdx)
 {
     SWR_CONTEXT *pContext = pDC->pContext;
 
     AR_BEGIN(FEBinLines, pDC->drawId);
 
-    const API_STATE& state = GetApiState(pDC);
-    const SWR_RASTSTATE& rastState = state.rastState;
+    const API_STATE &state = GetApiState(pDC);
+    const SWR_RASTSTATE &rastState = state.rastState;
 
     // Select attribute processor
     PFN_PROCESS_ATTRIBUTES pfnProcessAttribs = GetProcessAttributesFunc(2,
         state.backendState.swizzleEnable, state.backendState.constantInterpolationMask);
 
-    simdscalar& vRecipW0 = recipW[0];
-    simdscalar& vRecipW1 = recipW[1];
-
-    simd4scalar vHorizX[8], vHorizY[8], vHorizZ[8], vHorizW[8];
+    typename SIMD_T::Float &vRecipW0 = recipW[0];
+    typename SIMD_T::Float &vRecipW1 = recipW[1];
 
     // convert to fixed point
-    simdscalari vXi[2], vYi[2];
-    vXi[0] = fpToFixedPointVertical(prim[0].x);
-    vYi[0] = fpToFixedPointVertical(prim[0].y);
-    vXi[1] = fpToFixedPointVertical(prim[1].x);
-    vYi[1] = fpToFixedPointVertical(prim[1].y);
-
-    // compute x-major vs y-major mask
-    simdscalari xLength = _simd_abs_epi32(_simd_sub_epi32(vXi[0], vXi[1]));
-    simdscalari yLength = _simd_abs_epi32(_simd_sub_epi32(vYi[0], vYi[1]));
-    simdscalar vYmajorMask = _simd_castsi_ps(_simd_cmpgt_epi32(yLength, xLength));
-    uint32_t yMajorMask = _simd_movemask_ps(vYmajorMask);
-
-    // cull zero-length lines
-    simdscalari vZeroLengthMask = _simd_cmpeq_epi32(xLength, _simd_setzero_si());
-    vZeroLengthMask = _simd_and_si(vZeroLengthMask, _simd_cmpeq_epi32(yLength, _simd_setzero_si()));
-
-    primMask &= ~_simd_movemask_ps(_simd_castsi_ps(vZeroLengthMask));
-
-    uint32_t *pPrimID = (uint32_t *)&primID;
-    const uint32_t *pViewportIndex = (uint32_t *)&viewportIdx;
-
-    simdscalar vUnused = _simd_setzero_ps();
-
-    // Calc bounding box of lines
-    simdBBox bbox;
-    bbox.xmin = _simd_min_epi32(vXi[0], vXi[1]);
-    bbox.xmax = _simd_max_epi32(vXi[0], vXi[1]);
-    bbox.ymin = _simd_min_epi32(vYi[0], vYi[1]);
-    bbox.ymax = _simd_max_epi32(vYi[0], vYi[1]);
-
-    // bloat bbox by line width along minor axis
-    simdscalar vHalfWidth = _simd_set1_ps(rastState.lineWidth / 2.0f);
-    simdscalari vHalfWidthi = fpToFixedPointVertical(vHalfWidth);
-    simdBBox bloatBox;
-    bloatBox.xmin = _simd_sub_epi32(bbox.xmin, vHalfWidthi);
-    bloatBox.xmax = _simd_add_epi32(bbox.xmax, vHalfWidthi);
-    bloatBox.ymin = _simd_sub_epi32(bbox.ymin, vHalfWidthi);
-    bloatBox.ymax = _simd_add_epi32(bbox.ymax, vHalfWidthi);
-
-    bbox.xmin = _simd_blendv_epi32(bbox.xmin, bloatBox.xmin, vYmajorMask);
-    bbox.xmax = _simd_blendv_epi32(bbox.xmax, bloatBox.xmax, vYmajorMask);
-    bbox.ymin = _simd_blendv_epi32(bloatBox.ymin, bbox.ymin, vYmajorMask);
-    bbox.ymax = _simd_blendv_epi32(bloatBox.ymax, bbox.ymax, vYmajorMask);
-
-    // Intersect with scissor/viewport. Subtract 1 ULP in x.8 fixed point since xmax/ymax edge is exclusive.
-    {
-        simdscalari scisXmin, scisYmin, scisXmax, scisYmax;
-        if (state.backendState.readViewportArrayIndex)
-        {
-            GatherScissors<KNOB_SIMD_WIDTH>::Gather(&state.scissorsInFixedPoint[0], pViewportIndex,
-                scisXmin, scisYmin, scisXmax, scisYmax);
-        }
-        else // broadcast fast path for non-VPAI case.
-        {
-            scisXmin = _simd_set1_epi32(state.scissorsInFixedPoint[0].xmin);
-            scisYmin = _simd_set1_epi32(state.scissorsInFixedPoint[0].ymin);
-            scisXmax = _simd_set1_epi32(state.scissorsInFixedPoint[0].xmax);
-            scisYmax = _simd_set1_epi32(state.scissorsInFixedPoint[0].ymax);
-        }
+    typename SIMD_T::Integer vXi[2], vYi[2];
 
-        bbox.xmin = _simd_max_epi32(bbox.xmin, scisXmin);
-        bbox.ymin = _simd_max_epi32(bbox.ymin, scisYmin);
-        bbox.xmax = _simd_min_epi32(_simd_sub_epi32(bbox.xmax, _simd_set1_epi32(1)), scisXmax);
-        bbox.ymax = _simd_min_epi32(_simd_sub_epi32(bbox.ymax, _simd_set1_epi32(1)), scisYmax);
-    }
-
-    // Cull prims completely outside scissor
-    {
-        simdscalari maskOutsideScissorX = _simd_cmpgt_epi32(bbox.xmin, bbox.xmax);
-        simdscalari maskOutsideScissorY = _simd_cmpgt_epi32(bbox.ymin, bbox.ymax);
-        simdscalari maskOutsideScissorXY = _simd_or_si(maskOutsideScissorX, maskOutsideScissorY);
-        uint32_t maskOutsideScissor = _simd_movemask_ps(_simd_castsi_ps(maskOutsideScissorXY));
-        primMask = primMask & ~maskOutsideScissor;
-    }
-
-    if (!primMask)
-    {
-        goto endBinLines;
-    }
-
-    // Convert triangle bbox to macrotile units.
-    bbox.xmin = _simd_srai_epi32(bbox.xmin, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-    bbox.ymin = _simd_srai_epi32(bbox.ymin, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-    bbox.xmax = _simd_srai_epi32(bbox.xmax, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-    bbox.ymax = _simd_srai_epi32(bbox.ymax, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-
-    OSALIGNSIMD(uint32_t) aMTLeft[KNOB_SIMD_WIDTH], aMTRight[KNOB_SIMD_WIDTH], aMTTop[KNOB_SIMD_WIDTH], aMTBottom[KNOB_SIMD_WIDTH];
-    _simd_store_si((simdscalari*)aMTLeft, bbox.xmin);
-    _simd_store_si((simdscalari*)aMTRight, bbox.xmax);
-    _simd_store_si((simdscalari*)aMTTop, bbox.ymin);
-    _simd_store_si((simdscalari*)aMTBottom, bbox.ymax);
-
-    // transpose verts needed for backend
-    /// @todo modify BE to take non-transformed verts
-    vTranspose3x8(vHorizX, prim[0].x, prim[1].x, vUnused);
-    vTranspose3x8(vHorizY, prim[0].y, prim[1].y, vUnused);
-    vTranspose3x8(vHorizZ, prim[0].z, prim[1].z, vUnused);
-    vTranspose3x8(vHorizW, vRecipW0, vRecipW1, vUnused);
-
-    // store render target array index
-    OSALIGNSIMD(uint32_t) aRTAI[KNOB_SIMD_WIDTH];
-    if (state.backendState.readRenderTargetArrayIndex)
-    {
-        simdvector vRtai[2];
-        pa.Assemble(VERTEX_SGV_SLOT, vRtai);
-        simdscalari vRtaii = _simd_castps_si(vRtai[0][VERTEX_SGV_RTAI_COMP]);
-        _simd_store_si((simdscalari*)aRTAI, vRtaii);
-    }
-    else
-    {
-        _simd_store_si((simdscalari*)aRTAI, _simd_setzero_si());
-    }
-
-    // scan remaining valid prims and bin each separately
-    DWORD primIndex;
-    while (_BitScanForward(&primIndex, primMask))
-    {
-        uint32_t linkageCount = state.backendState.numAttributes;
-        uint32_t numScalarAttribs = linkageCount * 4;
-
-        BE_WORK work;
-        work.type = DRAW;
-
-        TRIANGLE_WORK_DESC &desc = work.desc.tri;
-
-        desc.triFlags.frontFacing = 1;
-        desc.triFlags.yMajor = (yMajorMask >> primIndex) & 1;
-        desc.triFlags.renderTargetArrayIndex = aRTAI[primIndex];
-        desc.triFlags.viewportIndex = pViewportIndex[primIndex];
-
-        work.pfnWork = RasterizeLine;
-
-        auto pArena = pDC->pArena;
-        SWR_ASSERT(pArena != nullptr);
-
-        // store active attribs
-        desc.pAttribs = (float*)pArena->AllocAligned(numScalarAttribs * 3 * sizeof(float), 16);
-        desc.numAttribs = linkageCount;
-        pfnProcessAttribs(pDC, pa, primIndex, pPrimID[primIndex], desc.pAttribs);
-
-        // store line vertex data
-        desc.pTriBuffer = (float*)pArena->AllocAligned(4 * 4 * sizeof(float), 16);
-        SIMD128::store_ps(&desc.pTriBuffer[0], vHorizX[primIndex]);
-        SIMD128::store_ps(&desc.pTriBuffer[4], vHorizY[primIndex]);
-        SIMD128::store_ps(&desc.pTriBuffer[8], vHorizZ[primIndex]);
-        SIMD128::store_ps(&desc.pTriBuffer[12], vHorizW[primIndex]);
-
-        // store user clip distances
-        if (rastState.clipDistanceMask)
-        {
-            uint32_t numClipDist = _mm_popcnt_u32(rastState.clipDistanceMask);
-            desc.pUserClipBuffer = (float*)pArena->Alloc(numClipDist * 2 * sizeof(float));
-            ProcessUserClipDist<2>(pa, primIndex, rastState.clipDistanceMask, &desc.pTriBuffer[12], desc.pUserClipBuffer);
-        }
-
-        MacroTileMgr *pTileMgr = pDC->pTileMgr;
-        for (uint32_t y = aMTTop[primIndex]; y <= aMTBottom[primIndex]; ++y)
-        {
-            for (uint32_t x = aMTLeft[primIndex]; x <= aMTRight[primIndex]; ++x)
-            {
-#if KNOB_ENABLE_TOSS_POINTS
-                if (!KNOB_TOSS_SETUP_TRIS)
-#endif
-                {
-                    pTileMgr->enqueue(x, y, &work);
-                }
-            }
-        }
-
-        primMask &= ~(1 << primIndex);
-    }
-
-endBinLines:
-
-    AR_END(FEBinLines, 1);
-}
-
-#if USE_SIMD16_FRONTEND
-void BinPostSetupLines_simd16(
-    DRAW_CONTEXT *pDC,
-    PA_STATE& pa,
-    uint32_t workerId,
-    simd16vector prim[],
-    simd16scalar recipW[],
-    uint32_t primMask,
-    simd16scalari const &primID,
-    simd16scalari const &viewportIdx)
-{
-    SWR_CONTEXT *pContext = pDC->pContext;
-
-    AR_BEGIN(FEBinLines, pDC->drawId);
-
-    const API_STATE& state = GetApiState(pDC);
-    const SWR_RASTSTATE& rastState = state.rastState;
-
-    // Select attribute processor
-    PFN_PROCESS_ATTRIBUTES pfnProcessAttribs = GetProcessAttributesFunc(2,
-        state.backendState.swizzleEnable, state.backendState.constantInterpolationMask);
-
-    simd16scalar& vRecipW0 = recipW[0];
-    simd16scalar& vRecipW1 = recipW[1];
-
-    // convert to fixed point
-    simd16scalari vXi[2], vYi[2];
-
-    vXi[0] = fpToFixedPointVertical(prim[0].x);
-    vYi[0] = fpToFixedPointVertical(prim[0].y);
-    vXi[1] = fpToFixedPointVertical(prim[1].x);
-    vYi[1] = fpToFixedPointVertical(prim[1].y);
+    vXi[0] = fpToFixedPointVertical<SIMD_T>(prim[0].x);
+    vYi[0] = fpToFixedPointVertical<SIMD_T>(prim[0].y);
+    vXi[1] = fpToFixedPointVertical<SIMD_T>(prim[1].x);
+    vYi[1] = fpToFixedPointVertical<SIMD_T>(prim[1].y);
 
     // compute x-major vs y-major mask
-    simd16scalari xLength = _simd16_abs_epi32(_simd16_sub_epi32(vXi[0], vXi[1]));
-    simd16scalari yLength = _simd16_abs_epi32(_simd16_sub_epi32(vYi[0], vYi[1]));
-    simd16scalar vYmajorMask = _simd16_castsi_ps(_simd16_cmpgt_epi32(yLength, xLength));
-    uint32_t yMajorMask = _simd16_movemask_ps(vYmajorMask);
+    typename SIMD_T::Integer xLength = SIMD_T::abs_epi32(SIMD_T::sub_epi32(vXi[0], vXi[1]));
+    typename SIMD_T::Integer yLength = SIMD_T::abs_epi32(SIMD_T::sub_epi32(vYi[0], vYi[1]));
+    typename SIMD_T::Float vYmajorMask = SIMD_T::castsi_ps(SIMD_T::cmpgt_epi32(yLength, xLength));
+    uint32_t yMajorMask = SIMD_T::movemask_ps(vYmajorMask);
 
     // cull zero-length lines
-    simd16scalari vZeroLengthMask = _simd16_cmpeq_epi32(xLength, _simd16_setzero_si());
-    vZeroLengthMask = _simd16_and_si(vZeroLengthMask, _simd16_cmpeq_epi32(yLength, _simd16_setzero_si()));
+    typename SIMD_T::Integer vZeroLengthMask = SIMD_T::cmpeq_epi32(xLength, SIMD_T::setzero_si());
+    vZeroLengthMask = SIMD_T::and_si(vZeroLengthMask, SIMD_T::cmpeq_epi32(yLength, SIMD_T::setzero_si()));
 
-    primMask &= ~_simd16_movemask_ps(_simd16_castsi_ps(vZeroLengthMask));
+    primMask &= ~SIMD_T::movemask_ps(SIMD_T::castsi_ps(vZeroLengthMask));
 
     uint32_t *pPrimID = (uint32_t *)&primID;
     const uint32_t *pViewportIndex = (uint32_t *)&viewportIdx;
 
     // Calc bounding box of lines
-    simd16BBox bbox;
-    bbox.xmin = _simd16_min_epi32(vXi[0], vXi[1]);
-    bbox.xmax = _simd16_max_epi32(vXi[0], vXi[1]);
-    bbox.ymin = _simd16_min_epi32(vYi[0], vYi[1]);
-    bbox.ymax = _simd16_max_epi32(vYi[0], vYi[1]);
+    SIMDBBOX_T<SIMD_T> bbox;
+    bbox.xmin = SIMD_T::min_epi32(vXi[0], vXi[1]);
+    bbox.xmax = SIMD_T::max_epi32(vXi[0], vXi[1]);
+    bbox.ymin = SIMD_T::min_epi32(vYi[0], vYi[1]);
+    bbox.ymax = SIMD_T::max_epi32(vYi[0], vYi[1]);
 
     // bloat bbox by line width along minor axis
-    simd16scalar vHalfWidth = _simd16_set1_ps(rastState.lineWidth / 2.0f);
-    simd16scalari vHalfWidthi = fpToFixedPointVertical(vHalfWidth);
+    typename SIMD_T::Float vHalfWidth = SIMD_T::set1_ps(rastState.lineWidth / 2.0f);
+    typename SIMD_T::Integer vHalfWidthi = fpToFixedPointVertical<SIMD_T>(vHalfWidth);
 
-    simd16BBox bloatBox;
+    SIMDBBOX_T<SIMD_T> bloatBox;
 
-    bloatBox.xmin = _simd16_sub_epi32(bbox.xmin, vHalfWidthi);
-    bloatBox.xmax = _simd16_add_epi32(bbox.xmax, vHalfWidthi);
-    bloatBox.ymin = _simd16_sub_epi32(bbox.ymin, vHalfWidthi);
-    bloatBox.ymax = _simd16_add_epi32(bbox.ymax, vHalfWidthi);
+    bloatBox.xmin = SIMD_T::sub_epi32(bbox.xmin, vHalfWidthi);
+    bloatBox.xmax = SIMD_T::add_epi32(bbox.xmax, vHalfWidthi);
+    bloatBox.ymin = SIMD_T::sub_epi32(bbox.ymin, vHalfWidthi);
+    bloatBox.ymax = SIMD_T::add_epi32(bbox.ymax, vHalfWidthi);
 
-    bbox.xmin = _simd16_blendv_epi32(bbox.xmin, bloatBox.xmin, vYmajorMask);
-    bbox.xmax = _simd16_blendv_epi32(bbox.xmax, bloatBox.xmax, vYmajorMask);
-    bbox.ymin = _simd16_blendv_epi32(bloatBox.ymin, bbox.ymin, vYmajorMask);
-    bbox.ymax = _simd16_blendv_epi32(bloatBox.ymax, bbox.ymax, vYmajorMask);
+    bbox.xmin = SIMD_T::blendv_epi32(bbox.xmin, bloatBox.xmin, vYmajorMask);
+    bbox.xmax = SIMD_T::blendv_epi32(bbox.xmax, bloatBox.xmax, vYmajorMask);
+    bbox.ymin = SIMD_T::blendv_epi32(bloatBox.ymin, bbox.ymin, vYmajorMask);
+    bbox.ymax = SIMD_T::blendv_epi32(bloatBox.ymax, bbox.ymax, vYmajorMask);
 
     // Intersect with scissor/viewport. Subtract 1 ULP in x.8 fixed point since xmax/ymax edge is exclusive.
     {
-        simd16scalari scisXmin, scisYmin, scisXmax, scisYmax;
+        typename SIMD_T::Integer scisXmin, scisYmin, scisXmax, scisYmax;
 
         if (state.backendState.readViewportArrayIndex)
         {
-            GatherScissors_simd16<KNOB_SIMD16_WIDTH>::Gather(&state.scissorsInFixedPoint[0], pViewportIndex,
-                scisXmin, scisYmin, scisXmax, scisYmax);
+            GatherScissors(&state.scissorsInFixedPoint[0], pViewportIndex, scisXmin, scisYmin, scisXmax, scisYmax);
         }
         else // broadcast fast path for non-VPAI case.
         {
-            scisXmin = _simd16_set1_epi32(state.scissorsInFixedPoint[0].xmin);
-            scisYmin = _simd16_set1_epi32(state.scissorsInFixedPoint[0].ymin);
-            scisXmax = _simd16_set1_epi32(state.scissorsInFixedPoint[0].xmax);
-            scisYmax = _simd16_set1_epi32(state.scissorsInFixedPoint[0].ymax);
+            scisXmin = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].xmin);
+            scisYmin = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].ymin);
+            scisXmax = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].xmax);
+            scisYmax = SIMD_T::set1_epi32(state.scissorsInFixedPoint[0].ymax);
         }
 
-        bbox.xmin = _simd16_max_epi32(bbox.xmin, scisXmin);
-        bbox.ymin = _simd16_max_epi32(bbox.ymin, scisYmin);
-        bbox.xmax = _simd16_min_epi32(_simd16_sub_epi32(bbox.xmax, _simd16_set1_epi32(1)), scisXmax);
-        bbox.ymax = _simd16_min_epi32(_simd16_sub_epi32(bbox.ymax, _simd16_set1_epi32(1)), scisYmax);
+        bbox.xmin = SIMD_T::max_epi32(bbox.xmin, scisXmin);
+        bbox.ymin = SIMD_T::max_epi32(bbox.ymin, scisYmin);
+        bbox.xmax = SIMD_T::min_epi32(SIMD_T::sub_epi32(bbox.xmax, SIMD_T::set1_epi32(1)), scisXmax);
+        bbox.ymax = SIMD_T::min_epi32(SIMD_T::sub_epi32(bbox.ymax, SIMD_T::set1_epi32(1)), scisYmax);
     }
 
     // Cull prims completely outside scissor
     {
-        simd16scalari maskOutsideScissorX = _simd16_cmpgt_epi32(bbox.xmin, bbox.xmax);
-        simd16scalari maskOutsideScissorY = _simd16_cmpgt_epi32(bbox.ymin, bbox.ymax);
-        simd16scalari maskOutsideScissorXY = _simd16_or_si(maskOutsideScissorX, maskOutsideScissorY);
-        uint32_t maskOutsideScissor = _simd16_movemask_ps(_simd16_castsi_ps(maskOutsideScissorXY));
+        typename SIMD_T::Integer maskOutsideScissorX = SIMD_T::cmpgt_epi32(bbox.xmin, bbox.xmax);
+        typename SIMD_T::Integer maskOutsideScissorY = SIMD_T::cmpgt_epi32(bbox.ymin, bbox.ymax);
+        typename SIMD_T::Integer maskOutsideScissorXY = SIMD_T::or_si(maskOutsideScissorX, maskOutsideScissorY);
+        uint32_t maskOutsideScissor = SIMD_T::movemask_ps(SIMD_T::castsi_ps(maskOutsideScissorXY));
         primMask = primMask & ~maskOutsideScissor;
     }
 
-    const simdscalar unused = _simd_setzero_ps();
-
     // transpose verts needed for backend
     /// @todo modify BE to take non-transformed verts
-    simd4scalar vHorizX[2][KNOB_SIMD_WIDTH]; // KNOB_SIMD16_WIDTH
-    simd4scalar vHorizY[2][KNOB_SIMD_WIDTH]; // KNOB_SIMD16_WIDTH
-    simd4scalar vHorizZ[2][KNOB_SIMD_WIDTH]; // KNOB_SIMD16_WIDTH
-    simd4scalar vHorizW[2][KNOB_SIMD_WIDTH]; // KNOB_SIMD16_WIDTH
+    simd4scalar vHorizX[SIMD_WIDTH];
+    simd4scalar vHorizY[SIMD_WIDTH];
+    simd4scalar vHorizZ[SIMD_WIDTH];
+    simd4scalar vHorizW[SIMD_WIDTH];
 
     if (!primMask)
     {
@@ -2475,40 +1569,42 @@ void BinPostSetupLines_simd16(
     }
 
     // Convert triangle bbox to macrotile units.
-    bbox.xmin = _simd16_srai_epi32(bbox.xmin, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-    bbox.ymin = _simd16_srai_epi32(bbox.ymin, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-    bbox.xmax = _simd16_srai_epi32(bbox.xmax, KNOB_MACROTILE_X_DIM_FIXED_SHIFT);
-    bbox.ymax = _simd16_srai_epi32(bbox.ymax, KNOB_MACROTILE_Y_DIM_FIXED_SHIFT);
-
-    OSALIGNSIMD16(uint32_t) aMTLeft[KNOB_SIMD16_WIDTH], aMTRight[KNOB_SIMD16_WIDTH], aMTTop[KNOB_SIMD16_WIDTH], aMTBottom[KNOB_SIMD16_WIDTH];
+#if SIMD_WA_SXXI_EPI32
+    bbox.xmin = simd_wa_srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmin);
+    bbox.ymin = simd_wa_srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymin);
+    bbox.xmax = simd_wa_srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmax);
+    bbox.ymax = simd_wa_srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymax);
+#else
+    bbox.xmin = SIMD_T::srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmin);
+    bbox.ymin = SIMD_T::srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymin);
+    bbox.xmax = SIMD_T::srai_epi32<KNOB_MACROTILE_X_DIM_FIXED_SHIFT>(bbox.xmax);
+    bbox.ymax = SIMD_T::srai_epi32<KNOB_MACROTILE_Y_DIM_FIXED_SHIFT>(bbox.ymax);
+#endif
 
-    _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTLeft),    bbox.xmin);
-    _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTRight),   bbox.xmax);
-    _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTTop),     bbox.ymin);
-    _simd16_store_si(reinterpret_cast<simd16scalari *>(aMTBottom),  bbox.ymax);
+    OSALIGNSIMD16(uint32_t) aMTLeft[SIMD_WIDTH], aMTRight[SIMD_WIDTH], aMTTop[SIMD_WIDTH], aMTBottom[SIMD_WIDTH];
 
-    vTranspose3x8(vHorizX[0], _simd16_extract_ps(prim[0].x, 0), _simd16_extract_ps(prim[1].x, 0), unused);
-    vTranspose3x8(vHorizY[0], _simd16_extract_ps(prim[0].y, 0), _simd16_extract_ps(prim[1].y, 0), unused);
-    vTranspose3x8(vHorizZ[0], _simd16_extract_ps(prim[0].z, 0), _simd16_extract_ps(prim[1].z, 0), unused);
-    vTranspose3x8(vHorizW[0], _simd16_extract_ps(vRecipW0, 0),  _simd16_extract_ps(vRecipW1, 0),  unused);
+    SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTLeft),   bbox.xmin);
+    SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTRight),  bbox.xmax);
+    SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTTop),    bbox.ymin);
+    SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aMTBottom), bbox.ymax);
 
-    vTranspose3x8(vHorizX[1], _simd16_extract_ps(prim[0].x, 1), _simd16_extract_ps(prim[1].x, 1), unused);
-    vTranspose3x8(vHorizY[1], _simd16_extract_ps(prim[0].y, 1), _simd16_extract_ps(prim[1].y, 1), unused);
-    vTranspose3x8(vHorizZ[1], _simd16_extract_ps(prim[0].z, 1), _simd16_extract_ps(prim[1].z, 1), unused);
-    vTranspose3x8(vHorizW[1], _simd16_extract_ps(vRecipW0, 1),  _simd16_extract_ps(vRecipW1, 1),  unused);
+    TransposeVertices(vHorizX, prim[0].x, prim[1].x, SIMD_T::setzero_ps());
+    TransposeVertices(vHorizY, prim[0].y, prim[1].y, SIMD_T::setzero_ps());
+    TransposeVertices(vHorizZ, prim[0].z, prim[1].z, SIMD_T::setzero_ps());
+    TransposeVertices(vHorizW, vRecipW0,  vRecipW1,  SIMD_T::setzero_ps());
 
     // store render target array index
-    OSALIGNSIMD16(uint32_t) aRTAI[KNOB_SIMD16_WIDTH];
+    OSALIGNSIMD16(uint32_t) aRTAI[SIMD_WIDTH];
     if (state.backendState.readRenderTargetArrayIndex)
     {
-        simd16vector vRtai[2];
-        pa.Assemble_simd16(VERTEX_SGV_SLOT, vRtai);
-        simd16scalari vRtaii = _simd16_castps_si(vRtai[0][VERTEX_SGV_RTAI_COMP]);
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aRTAI), vRtaii);
+        typename SIMD_T::Vec4 vRtai[2];
+        pa.Assemble(VERTEX_SGV_SLOT, vRtai);
+        typename SIMD_T::Integer vRtaii = SIMD_T::castps_si(vRtai[0][VERTEX_SGV_RTAI_COMP]);
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aRTAI), vRtaii);
     }
     else
     {
-        _simd16_store_si(reinterpret_cast<simd16scalari *>(aRTAI), _simd16_setzero_si());
+        SIMD_T::store_si(reinterpret_cast<typename SIMD_T::Integer *>(aRTAI), SIMD_T::setzero_si());
     }
 
     // scan remaining valid prims and bin each separately
@@ -2541,15 +1637,10 @@ void BinPostSetupLines_simd16(
         // store line vertex data
         desc.pTriBuffer = (float*)pArena->AllocAligned(4 * 4 * sizeof(float), 16);
 
-        {
-            const uint32_t i = primIndex >> 3;  // triIndex / KNOB_SIMD_WIDTH
-            const uint32_t j = primIndex & 7;   // triIndex % KNOB_SIMD_WIDTH
-
-            _mm_store_ps(&desc.pTriBuffer[ 0], vHorizX[i][j]);
-            _mm_store_ps(&desc.pTriBuffer[ 4], vHorizY[i][j]);
-            _mm_store_ps(&desc.pTriBuffer[ 8], vHorizZ[i][j]);
-            _mm_store_ps(&desc.pTriBuffer[12], vHorizW[i][j]);
-        }
+        _mm_store_ps(&desc.pTriBuffer[0],  vHorizX[primIndex]);
+        _mm_store_ps(&desc.pTriBuffer[4],  vHorizY[primIndex]);
+        _mm_store_ps(&desc.pTriBuffer[8],  vHorizZ[primIndex]);
+        _mm_store_ps(&desc.pTriBuffer[12], vHorizW[primIndex]);
 
         // store user clip distances
         if (rastState.clipDistanceMask)
@@ -2581,7 +1672,6 @@ endBinLines:
     AR_END(FEBinLines, 1);
 }
 
-#endif
 //////////////////////////////////////////////////////////////////////////
 /// @brief Bin SIMD lines to the backend.
 /// @param pDC - pointer to draw context.
@@ -2590,48 +1680,49 @@ endBinLines:
 /// @param tri - Contains line position data for SIMDs worth of points.
 /// @param primID - Primitive ID for each line.
 /// @param viewportIdx - Viewport Array Index for each line.
-void BinLines(
+template <typename SIMD_T, uint32_t SIMD_WIDTH>
+void SIMDCALL BinLinesImpl(
     DRAW_CONTEXT *pDC,
-    PA_STATE& pa,
+    PA_STATE &pa,
     uint32_t workerId,
-    simdvector prim[],
+    typename SIMD_T::Vec4 prim[3],
     uint32_t primMask,
-    simdscalari const &primID)
+    typename SIMD_T::Integer const &primID)
 {
     const API_STATE& state = GetApiState(pDC);
     const SWR_RASTSTATE& rastState = state.rastState;
     const SWR_FRONTEND_STATE& feState = state.frontendState;
 
-    simdscalar vRecipW[2] = { _simd_set1_ps(1.0f), _simd_set1_ps(1.0f) };
+    typename SIMD_T::Float vRecipW[2] = { SIMD_T::set1_ps(1.0f), SIMD_T::set1_ps(1.0f) };
 
-    simdscalari viewportIdx = _simd_set1_epi32(0);
+    typename SIMD_T::Integer viewportIdx = SIMD_T::set1_epi32(0);
     if (state.backendState.readViewportArrayIndex)
     {
-        simdvector vpiAttrib[2];
+        typename SIMD_T::Vec4 vpiAttrib[2];
         pa.Assemble(VERTEX_SGV_SLOT, vpiAttrib);
-        simdscalari vpai = _simd_castps_si(vpiAttrib[0][VERTEX_SGV_VAI_COMP]);
-        vpai = _simd_max_epi32(_simd_setzero_si(), vpai);
 
         // OOB indices => forced to zero.
-        simdscalari vNumViewports = _simd_set1_epi32(KNOB_NUM_VIEWPORTS_SCISSORS);
-        simdscalari vClearMask = _simd_cmplt_epi32(vpai, vNumViewports);
-        viewportIdx = _simd_and_si(vClearMask, vpai);
+        typename SIMD_T::Integer vpai = SIMD_T::castps_si(vpiAttrib[0][VERTEX_SGV_VAI_COMP]);
+        vpai = SIMD_T::max_epi32(SIMD_T::setzero_si(), vpai);
+        typename SIMD_T::Integer vNumViewports = SIMD_T::set1_epi32(KNOB_NUM_VIEWPORTS_SCISSORS);
+        typename SIMD_T::Integer vClearMask = SIMD_T::cmplt_epi32(vpai, vNumViewports);
+        viewportIdx = SIMD_T::and_si(vClearMask, vpai);
     }
 
     if (!feState.vpTransformDisable)
     {
         // perspective divide
-        vRecipW[0] = _simd_div_ps(_simd_set1_ps(1.0f), prim[0].w);
-        vRecipW[1] = _simd_div_ps(_simd_set1_ps(1.0f), prim[1].w);
+        vRecipW[0] = SIMD_T::div_ps(SIMD_T::set1_ps(1.0f), prim[0].w);
+        vRecipW[1] = SIMD_T::div_ps(SIMD_T::set1_ps(1.0f), prim[1].w);
 
-        prim[0].v[0] = _simd_mul_ps(prim[0].v[0], vRecipW[0]);
-        prim[1].v[0] = _simd_mul_ps(prim[1].v[0], vRecipW[1]);
+        prim[0].v[0] = SIMD_T::mul_ps(prim[0].v[0], vRecipW[0]);
+        prim[1].v[0] = SIMD_T::mul_ps(prim[1].v[0], vRecipW[1]);
 
-        prim[0].v[1] = _simd_mul_ps(prim[0].v[1], vRecipW[0]);
-        prim[1].v[1] = _simd_mul_ps(prim[1].v[1], vRecipW[1]);
+        prim[0].v[1] = SIMD_T::mul_ps(prim[0].v[1], vRecipW[0]);
+        prim[1].v[1] = SIMD_T::mul_ps(prim[1].v[1], vRecipW[1]);
 
-        prim[0].v[2] = _simd_mul_ps(prim[0].v[2], vRecipW[0]);
-        prim[1].v[2] = _simd_mul_ps(prim[1].v[2], vRecipW[1]);
+        prim[0].v[2] = SIMD_T::mul_ps(prim[0].v[2], vRecipW[0]);
+        prim[1].v[2] = SIMD_T::mul_ps(prim[1].v[2], vRecipW[1]);
 
         // viewport transform to screen coords
         if (state.backendState.readViewportArrayIndex)
@@ -2645,14 +1736,15 @@ void BinLines(
     }
 
     // adjust for pixel center location
-    simdscalar offset = g_pixelOffsets[rastState.pixelLocation];
-    prim[0].x = _simd_add_ps(prim[0].x, offset);
-    prim[0].y = _simd_add_ps(prim[0].y, offset);
+    typename SIMD_T::Float offset = g_pixelOffsets<SIMD_T>[rastState.pixelLocation];
+
+    prim[0].x = SIMD_T::add_ps(prim[0].x, offset);
+    prim[0].y = SIMD_T::add_ps(prim[0].y, offset);
 
-    prim[1].x = _simd_add_ps(prim[1].x, offset);
-    prim[1].y = _simd_add_ps(prim[1].y, offset);
+    prim[1].x = SIMD_T::add_ps(prim[1].x, offset);
+    prim[1].y = SIMD_T::add_ps(prim[1].y, offset);
 
-    BinPostSetupLines(
+    BinPostSetupLinesImpl<SIMD_T, SIMD_WIDTH>(
         pDC,
         pa,
         workerId,
@@ -2663,79 +1755,27 @@ void BinLines(
         viewportIdx);
 }
 
+void BinLines(
+    DRAW_CONTEXT *pDC,
+    PA_STATE &pa,
+    uint32_t workerId,
+    simdvector prim[],
+    uint32_t primMask,
+    simdscalari const &primID)
+{
+    BinLinesImpl<SIMD256, KNOB_SIMD_WIDTH>(pDC, pa, workerId, prim, primMask, primID);
+}
+
 #if USE_SIMD16_FRONTEND
 void SIMDCALL BinLines_simd16(
     DRAW_CONTEXT *pDC,
-    PA_STATE& pa,
+    PA_STATE &pa,
     uint32_t workerId,
     simd16vector prim[3],
     uint32_t primMask,
     simd16scalari const &primID)
 {
-    const API_STATE& state = GetApiState(pDC);
-    const SWR_RASTSTATE& rastState = state.rastState;
-    const SWR_FRONTEND_STATE& feState = state.frontendState;
-
-    simd16scalar vRecipW[2] = { _simd16_set1_ps(1.0f), _simd16_set1_ps(1.0f) };
-
-    simd16scalari viewportIdx = _simd16_set1_epi32(0);
-    if (state.backendState.readViewportArrayIndex)
-    {
-        simd16vector vpiAttrib[2];
-        pa.Assemble_simd16(VERTEX_SGV_SLOT, vpiAttrib);
-
-        // OOB indices => forced to zero.
-        simd16scalari vpai = _simd16_castps_si(vpiAttrib[0][VERTEX_SGV_VAI_COMP]);
-        vpai = _simd16_max_epi32(_simd16_setzero_si(), vpai);
-        simd16scalari vNumViewports = _simd16_set1_epi32(KNOB_NUM_VIEWPORTS_SCISSORS);
-        simd16scalari vClearMask = _simd16_cmplt_epi32(vpai, vNumViewports);
-        viewportIdx = _simd16_and_si(vClearMask, vpai);
-    }
-
-    if (!feState.vpTransformDisable)
-    {
-        // perspective divide
-        vRecipW[0] = _simd16_div_ps(_simd16_set1_ps(1.0f), prim[0].w);
-        vRecipW[1] = _simd16_div_ps(_simd16_set1_ps(1.0f), prim[1].w);
-
-        prim[0].v[0] = _simd16_mul_ps(prim[0].v[0], vRecipW[0]);
-        prim[1].v[0] = _simd16_mul_ps(prim[1].v[0], vRecipW[1]);
-
-        prim[0].v[1] = _simd16_mul_ps(prim[0].v[1], vRecipW[0]);
-        prim[1].v[1] = _simd16_mul_ps(prim[1].v[1], vRecipW[1]);
-
-        prim[0].v[2] = _simd16_mul_ps(prim[0].v[2], vRecipW[0]);
-        prim[1].v[2] = _simd16_mul_ps(prim[1].v[2], vRecipW[1]);
-
-        // viewport transform to screen coords
-        if (state.backendState.readViewportArrayIndex)
-        {
-            viewportTransform<2>(prim, state.vpMatrices, viewportIdx);
-        }
-        else
-        {
-            viewportTransform<2>(prim, state.vpMatrices);
-        }
-    }
-
-    // adjust for pixel center location
-    simd16scalar offset = g_pixelOffsets_simd16[rastState.pixelLocation];
-
-    prim[0].x = _simd16_add_ps(prim[0].x, offset);
-    prim[0].y = _simd16_add_ps(prim[0].y, offset);
-
-    prim[1].x = _simd16_add_ps(prim[1].x, offset);
-    prim[1].y = _simd16_add_ps(prim[1].y, offset);
-
-    BinPostSetupLines_simd16(
-        pDC,
-        pa,
-        workerId,
-        prim,
-        vRecipW,
-        primMask,
-        primID,
-        viewportIdx);
+    BinLinesImpl<SIMD512, KNOB_SIMD16_WIDTH>(pDC, pa, workerId, prim, primMask, primID);
 }
 
 #endif
diff --git a/src/gallium/drivers/swr/rasterizer/core/binner.h b/src/gallium/drivers/swr/rasterizer/core/binner.h
index 16161431d1..e842aa663b 100644
--- a/src/gallium/drivers/swr/rasterizer/core/binner.h
+++ b/src/gallium/drivers/swr/rasterizer/core/binner.h
@@ -31,67 +31,39 @@
 //////////////////////////////////////////////////////////////////////////
 /// @brief Offsets added to post-viewport vertex positions based on
 /// raster state.
-static const simdscalar g_pixelOffsets[SWR_PIXEL_LOCATION_UL + 1] =
+template <typename SIMD_T>
+static const typename SIMD_T::Float g_pixelOffsets[SWR_PIXEL_LOCATION_UL + 1] =
 {
-    _simd_set1_ps(0.0f),    // SWR_PIXEL_LOCATION_CENTER
-    _simd_set1_ps(0.5f),    // SWR_PIXEL_LOCATION_UL
+    SIMD_T::set1_ps(0.0f),  // SWR_PIXEL_LOCATION_CENTER
+    SIMD_T::set1_ps(0.5f),  // SWR_PIXEL_LOCATION_UL
 };
 
-#if USE_SIMD16_FRONTEND
-static const simd16scalar g_pixelOffsets_simd16[SWR_PIXEL_LOCATION_UL + 1] =
-{
-    _simd16_set1_ps(0.0f),  // SWR_PIXEL_LOCATION_CENTER
-    _simd16_set1_ps(0.5f),  // SWR_PIXEL_LOCATION_UL
-};
-
-#endif
 //////////////////////////////////////////////////////////////////////////
 /// @brief Convert the X,Y coords of a triangle to the requested Fixed 
 /// Point precision from FP32.
-template <typename PT = FixedPointTraits<Fixed_16_8>>
-INLINE simdscalari fpToFixedPointVertical(const simdscalar &vIn)
+template <typename SIMD_T, typename PT = FixedPointTraits<Fixed_16_8>>
+INLINE typename SIMD_T::Integer fpToFixedPointVertical(const typename SIMD_T::Float &vIn)
 {
-    simdscalar vFixed = _simd_mul_ps(vIn, _simd_set1_ps(PT::ScaleT::value));
-    return _simd_cvtps_epi32(vFixed);
+    return SIMD_T::cvtps_epi32(SIMD_T::mul_ps(vIn, SIMD_T::set1_ps(PT::ScaleT::value)));
 }
 
-#if USE_SIMD16_FRONTEND
-template <typename PT = FixedPointTraits<Fixed_16_8>>
-INLINE simd16scalari fpToFixedPointVertical(const simd16scalar &vIn)
-{
-    simd16scalar vFixed = _simd16_mul_ps(vIn, _simd16_set1_ps(PT::ScaleT::value));
-    return _simd16_cvtps_epi32(vFixed);
-}
-
-#endif
 //////////////////////////////////////////////////////////////////////////
 /// @brief Helper function to set the X,Y coords of a triangle to the 
 /// requested Fixed Point precision from FP32.
 /// @param tri: simdvector[3] of FP triangle verts
 /// @param vXi: fixed point X coords of tri verts
 /// @param vYi: fixed point Y coords of tri verts
-INLINE static void FPToFixedPoint(const simdvector * const tri, simdscalari(&vXi)[3], simdscalari(&vYi)[3])
-{
-    vXi[0] = fpToFixedPointVertical(tri[0].x);
-    vYi[0] = fpToFixedPointVertical(tri[0].y);
-    vXi[1] = fpToFixedPointVertical(tri[1].x);
-    vYi[1] = fpToFixedPointVertical(tri[1].y);
-    vXi[2] = fpToFixedPointVertical(tri[2].x);
-    vYi[2] = fpToFixedPointVertical(tri[2].y);
-}
-
-#if USE_SIMD16_FRONTEND
-INLINE static void FPToFixedPoint(const simd16vector * const tri, simd16scalari(&vXi)[3], simd16scalari(&vYi)[3])
+template <typename SIMD_T>
+INLINE static void FPToFixedPoint(const typename SIMD_T::Vec4 *const tri, typename SIMD_T::Integer(&vXi)[3], typename SIMD_T::Integer(&vYi)[3])
 {
-    vXi[0] = fpToFixedPointVertical(tri[0].x);
-    vYi[0] = fpToFixedPointVertical(tri[0].y);
-    vXi[1] = fpToFixedPointVertical(tri[1].x);
-    vYi[1] = fpToFixedPointVertical(tri[1].y);
-    vXi[2] = fpToFixedPointVertical(tri[2].x);
-    vYi[2] = fpToFixedPointVertical(tri[2].y);
+    vXi[0] = fpToFixedPointVertical<SIMD_T>(tri[0].x);
+    vYi[0] = fpToFixedPointVertical<SIMD_T>(tri[0].y);
+    vXi[1] = fpToFixedPointVertical<SIMD_T>(tri[1].x);
+    vYi[1] = fpToFixedPointVertical<SIMD_T>(tri[1].y);
+    vXi[2] = fpToFixedPointVertical<SIMD_T>(tri[2].x);
+    vYi[2] = fpToFixedPointVertical<SIMD_T>(tri[2].y);
 }
 
-#endif
 //////////////////////////////////////////////////////////////////////////
 /// @brief Calculate bounding box for current triangle
 /// @tparam CT: ConservativeRastFETraits type
@@ -100,124 +72,44 @@ INLINE static void FPToFixedPoint(const simd16vector * const tri, simd16scalari(
 /// @param bbox: fixed point bbox
 /// *Note*: expects vX, vY to be in the correct precision for the type 
 /// of rasterization. This avoids unnecessary FP->fixed conversions.
-template <typename CT>
-INLINE void calcBoundingBoxIntVertical(const simdvector * const tri, simdscalari(&vX)[3], simdscalari(&vY)[3], simdBBox &bbox)
+template <typename SIMD_T, typename CT>
+INLINE void calcBoundingBoxIntVertical(const typename SIMD_T::Integer(&vX)[3], const typename SIMD_T::Integer(&vY)[3], SIMDBBOX_T<SIMD_T> &bbox)
 {
-    simdscalari vMinX = vX[0];
-    vMinX = _simd_min_epi32(vMinX, vX[1]);
-    vMinX = _simd_min_epi32(vMinX, vX[2]);
+    typename SIMD_T::Integer vMinX = vX[0];
 
-    simdscalari vMaxX = vX[0];
-    vMaxX = _simd_max_epi32(vMaxX, vX[1]);
-    vMaxX = _simd_max_epi32(vMaxX, vX[2]);
+    vMinX = SIMD_T::min_epi32(vMinX, vX[1]);
+    vMinX = SIMD_T::min_epi32(vMinX, vX[2]);
 
-    simdscalari vMinY = vY[0];
-    vMinY = _simd_min_epi32(vMinY, vY[1]);
-    vMinY = _simd_min_epi32(vMinY, vY[2]);
+    typename SIMD_T::Integer vMaxX = vX[0];
 
-    simdscalari vMaxY = vY[0];
-    vMaxY = _simd_max_epi32(vMaxY, vY[1]);
-    vMaxY = _simd_max_epi32(vMaxY, vY[2]);
+    vMaxX = SIMD_T::max_epi32(vMaxX, vX[1]);
+    vMaxX = SIMD_T::max_epi32(vMaxX, vX[2]);
 
-    bbox.xmin = vMinX;
-    bbox.xmax = vMaxX;
-    bbox.ymin = vMinY;
-    bbox.ymax = vMaxY;
-}
+    typename SIMD_T::Integer vMinY = vY[0];
 
-#if USE_SIMD16_FRONTEND
-template <typename CT>
-INLINE void calcBoundingBoxIntVertical(const simd16vector * const tri, simd16scalari(&vX)[3], simd16scalari(&vY)[3], simd16BBox &bbox)
-{
-    simd16scalari vMinX = vX[0];
-
-    vMinX = _simd16_min_epi32(vMinX, vX[1]);
-    vMinX = _simd16_min_epi32(vMinX, vX[2]);
+    vMinY = SIMD_T::min_epi32(vMinY, vY[1]);
+    vMinY = SIMD_T::min_epi32(vMinY, vY[2]);
 
-    simd16scalari vMaxX = vX[0];
+    typename SIMD_T::Integer vMaxY = vY[0];
 
-    vMaxX = _simd16_max_epi32(vMaxX, vX[1]);
-    vMaxX = _simd16_max_epi32(vMaxX, vX[2]);
+    vMaxY = SIMD_T::max_epi32(vMaxY, vY[1]);
+    vMaxY = SIMD_T::max_epi32(vMaxY, vY[2]);
 
-    simd16scalari vMinY = vY[0];
+    if (CT::BoundingBoxOffsetT::value != 0)
+    {
+        /// Bounding box needs to be expanded by 1/512 before snapping to 16.8 for conservative rasterization
+        /// expand bbox by 1/256; coverage will be correctly handled in the rasterizer.
 
-    vMinY = _simd16_min_epi32(vMinY, vY[1]);
-    vMinY = _simd16_min_epi32(vMinY, vY[2]);
+        const typename SIMD_T::Integer value = SIMD_T::set1_epi32(CT::BoundingBoxOffsetT::value);
 
-    simd16scalari vMaxY = vY[0];
-
-    vMaxY = _simd16_max_epi32(vMaxY, vY[1]);
-    vMaxY = _simd16_max_epi32(vMaxY, vY[2]);
+        vMinX = SIMD_T::sub_epi32(vMinX, value);
+        vMaxX = SIMD_T::add_epi32(vMaxX, value);
+        vMinY = SIMD_T::sub_epi32(vMinY, value);
+        vMaxY = SIMD_T::add_epi32(vMaxY, value);
+    }
 
     bbox.xmin = vMinX;
     bbox.xmax = vMaxX;
     bbox.ymin = vMinY;
     bbox.ymax = vMaxY;
 }
-
-#endif
-//////////////////////////////////////////////////////////////////////////
-/// @brief FEConservativeRastT specialization of calcBoundingBoxIntVertical
-/// Offsets BBox for conservative rast
-template <>
-INLINE void calcBoundingBoxIntVertical<FEConservativeRastT>(const simdvector * const tri, simdscalari(&vX)[3], simdscalari(&vY)[3], simdBBox &bbox)
-{
-    // FE conservative rast traits
-    typedef FEConservativeRastT CT;
-
-    simdscalari vMinX = vX[0];
-    vMinX = _simd_min_epi32(vMinX, vX[1]);
-    vMinX = _simd_min_epi32(vMinX, vX[2]);
-
-    simdscalari vMaxX = vX[0];
-    vMaxX = _simd_max_epi32(vMaxX, vX[1]);
-    vMaxX = _simd_max_epi32(vMaxX, vX[2]);
-
-    simdscalari vMinY = vY[0];
-    vMinY = _simd_min_epi32(vMinY, vY[1]);
-    vMinY = _simd_min_epi32(vMinY, vY[2]);
-
-    simdscalari vMaxY = vY[0];
-    vMaxY = _simd_max_epi32(vMaxY, vY[1]);
-    vMaxY = _simd_max_epi32(vMaxY, vY[2]);
-
-    /// Bounding box needs to be expanded by 1/512 before snapping to 16.8 for conservative rasterization
-    /// expand bbox by 1/256; coverage will be correctly handled in the rasterizer.
-    bbox.xmin = _simd_sub_epi32(vMinX, _simd_set1_epi32(CT::BoundingBoxOffsetT::value));
-    bbox.xmax = _simd_add_epi32(vMaxX, _simd_set1_epi32(CT::BoundingBoxOffsetT::value));
-    bbox.ymin = _simd_sub_epi32(vMinY, _simd_set1_epi32(CT::BoundingBoxOffsetT::value));
-    bbox.ymax = _simd_add_epi32(vMaxY, _simd_set1_epi32(CT::BoundingBoxOffsetT::value));
-}
-
-#if USE_SIMD16_FRONTEND
-template <>
-INLINE void calcBoundingBoxIntVertical<FEConservativeRastT>(const simd16vector * const tri, simd16scalari(&vX)[3], simd16scalari(&vY)[3], simd16BBox &bbox)
-{
-    // FE conservative rast traits
-    typedef FEConservativeRastT CT;
-
-    simd16scalari vMinX = vX[0];
-    vMinX = _simd16_min_epi32(vMinX, vX[1]);
-    vMinX = _simd16_min_epi32(vMinX, vX[2]);
-
-    simd16scalari vMaxX = vX[0];
-    vMaxX = _simd16_max_epi32(vMaxX, vX[1]);
-    vMaxX = _simd16_max_epi32(vMaxX, vX[2]);
-
-    simd16scalari vMinY = vY[0];
-    vMinY = _simd16_min_epi32(vMinY, vY[1]);
-    vMinY = _simd16_min_epi32(vMinY, vY[2]);
-
-    simd16scalari vMaxY = vY[0];
-    vMaxY = _simd16_max_epi32(vMaxY, vY[1]);
-    vMaxY = _simd16_max_epi32(vMaxY, vY[2]);
-
-    /// Bounding box needs to be expanded by 1/512 before snapping to 16.8 for conservative rasterization
-    /// expand bbox by 1/256; coverage will be correctly handled in the rasterizer.
-    bbox.xmin = _simd16_sub_epi32(vMinX, _simd16_set1_epi32(CT::BoundingBoxOffsetT::value));
-    bbox.xmax = _simd16_add_epi32(vMaxX, _simd16_set1_epi32(CT::BoundingBoxOffsetT::value));
-    bbox.ymin = _simd16_sub_epi32(vMinY, _simd16_set1_epi32(CT::BoundingBoxOffsetT::value));
-    bbox.ymax = _simd16_add_epi32(vMaxY, _simd16_set1_epi32(CT::BoundingBoxOffsetT::value));
-}
-
-#endif
diff --git a/src/gallium/drivers/swr/rasterizer/core/conservativeRast.h b/src/gallium/drivers/swr/rasterizer/core/conservativeRast.h
index 1d8546959f..00c3a87c18 100644
--- a/src/gallium/drivers/swr/rasterizer/core/conservativeRast.h
+++ b/src/gallium/drivers/swr/rasterizer/core/conservativeRast.h
@@ -109,6 +109,7 @@ template <>
 struct ConservativeRastFETraits<StandardRastT>
 {
     typedef std::false_type IsConservativeT;
+    typedef std::integral_constant<uint32_t, 0> BoundingBoxOffsetT;
 };
 
 //////////////////////////////////////////////////////////////////////////
diff --git a/src/gallium/drivers/swr/rasterizer/core/pa.h b/src/gallium/drivers/swr/rasterizer/core/pa.h
index 87dba22bf8..b2ae8dafb9 100644
--- a/src/gallium/drivers/swr/rasterizer/core/pa.h
+++ b/src/gallium/drivers/swr/rasterizer/core/pa.h
@@ -91,6 +91,7 @@ struct PA_STATE
     virtual bool Assemble(uint32_t slot, simdvector verts[]) = 0;
 #if ENABLE_AVX512_SIMD16
     virtual bool Assemble_simd16(uint32_t slot, simd16vector verts[]) = 0;
+    virtual bool Assemble(uint32_t slot, simd16vector verts[]) = 0;
 #endif
     virtual void AssembleSingle(uint32_t slot, uint32_t primIndex, simd4scalar verts[]) = 0;
     virtual bool NextPrim() = 0;
@@ -203,6 +204,11 @@ struct PA_STATE_OPT : public PA_STATE
         return this->pfnPaFunc_simd16(*this, slot, verts);
     }
 
+    bool Assemble(uint32_t slot, simd16vector verts[])
+    {
+        return Assemble_simd16(slot, verts);
+    }
+
 #endif
     // Assembles 1 primitive. Each simdscalar is a vertex (xyzw).
     void AssembleSingle(uint32_t slot, uint32_t primIndex, simd4scalar verts[])
@@ -765,6 +771,11 @@ struct PA_STATE_CUT : public PA_STATE
         return true;
     }
 
+    bool Assemble(uint32_t slot, simd16vector verts[])
+    {
+        return Assemble_simd16(slot, verts);
+    }
+
 #endif
     void AssembleSingle(uint32_t slot, uint32_t triIndex, simd4scalar tri[3])
     {
@@ -1319,6 +1330,11 @@ struct PA_TESS : PA_STATE
         return true;
     }
 
+    bool Assemble(uint32_t slot, simd16vector verts[])
+    {
+        return Assemble_simd16(slot, verts);
+    }
+
 #endif
     void AssembleSingle(uint32_t slot, uint32_t primIndex, simd4scalar verts[])
     {
diff --git a/src/gallium/drivers/swr/rasterizer/core/utils.h b/src/gallium/drivers/swr/rasterizer/core/utils.h
index 392ee4ba94..b096d2120c 100644
--- a/src/gallium/drivers/swr/rasterizer/core/utils.h
+++ b/src/gallium/drivers/swr/rasterizer/core/utils.h
@@ -53,6 +53,14 @@ struct simd16BBox
 };
 #endif
 
+template<typename SIMD_T>
+struct SIMDBBOX_T
+{
+    typename SIMD_T::Integer            ymin;
+    typename SIMD_T::Integer            ymax;
+    typename SIMD_T::Integer            xmin;
+    typename SIMD_T::Integer            xmax;
+};
 
 // helper function to unroll loops
 template<int Begin, int End, int Step = 1>