From cd6c200223f7c6f5bac6bd2f2991bccf363fa7d9 Mon Sep 17 00:00:00 2001 From: Tim Rowley Date: Sun, 26 Mar 2017 15:46:42 -0500 Subject: [PATCH] swr: [rasterizer core] SIMD16 Frontend WIP Implement widened VS output for SIMD16 Reviewed-by: Bruce Cherniak --- .../drivers/swr/rasterizer/core/frontend.cpp | 42 +++++----------------- src/gallium/drivers/swr/rasterizer/core/state.h | 9 +++-- 2 files changed, 14 insertions(+), 37 deletions(-) diff --git a/src/gallium/drivers/swr/rasterizer/core/frontend.cpp b/src/gallium/drivers/swr/rasterizer/core/frontend.cpp index 9df7eeadc10..8cf234cd67a 100644 --- a/src/gallium/drivers/swr/rasterizer/core/frontend.cpp +++ b/src/gallium/drivers/swr/rasterizer/core/frontend.cpp @@ -1543,6 +1543,8 @@ void ProcessDraw( vsContext_lo.pVin = &vin_lo; vsContext_hi.pVin = &vin_hi; + vsContext_lo.AlternateOffset = 0; + vsContext_hi.AlternateOffset = 1; SWR_FETCH_CONTEXT fetchInfo_lo = { 0 }; @@ -1612,20 +1614,18 @@ void ProcessDraw( pvCutIndices_hi = &reinterpret_cast(&pa.GetNextVsIndices())[1]; } - simdvertex vout_lo; - simdvertex vout_hi; - - vsContext_lo.pVout = &vout_lo; - vsContext_hi.pVout = &vout_hi; - simd16vertex &vout = pa.GetNextVsOutput(); + vsContext_lo.pVout = reinterpret_cast(&vout); + vsContext_hi.pVout = reinterpret_cast(&vout); + if (i < endVertex) { // 1. Execute FS/VS for a single SIMD. AR_BEGIN(FEFetchShader, pDC->drawId); state.pfnFetchFunc(fetchInfo_lo, vin_lo); - if ((i + KNOB_SIMD_WIDTH) < endVertex) + + if ((i + KNOB_SIMD_WIDTH) < endVertex) // 1/2 of KNOB_SIMD16_WIDTH { state.pfnFetchFunc(fetchInfo_hi, vin_hi); } @@ -1655,35 +1655,9 @@ void ProcessDraw( AR_BEGIN(FEVertexShader, pDC->drawId); state.pfnVertexFunc(GetPrivateState(pDC), &vsContext_lo); - // copy SIMD vout_lo to lo part of SIMD16 vout - { - const uint32_t attribCount = sizeof(vout.attrib) / sizeof(vout.attrib[0]); - - for (uint32_t i = 0; i < attribCount; i += 1) - { - for (uint32_t j = 0; j < 4; j += 1) - { - vout.attrib[i][j] = _simd16_insert_ps(_simd16_setzero_ps(), vout_lo.attrib[i][j], 0); - } - } - } - - if ((i + KNOB_SIMD_WIDTH) < endVertex) + if ((i + KNOB_SIMD_WIDTH) < endVertex) // 1/2 of KNOB_SIMD16_WIDTH { state.pfnVertexFunc(GetPrivateState(pDC), &vsContext_hi); - - // copy SIMD vout_hi to hi part of SIMD16 vout - { - const uint32_t attribCount = sizeof(vout.attrib) / sizeof(vout.attrib[0]); - - for (uint32_t i = 0; i < attribCount; i += 1) - { - for (uint32_t j = 0; j < 4; j += 1) - { - vout.attrib[i][j] = _simd16_insert_ps(vout.attrib[i][j], vout_hi.attrib[i][j], 1); - } - } - } } AR_END(FEVertexShader, 0); diff --git a/src/gallium/drivers/swr/rasterizer/core/state.h b/src/gallium/drivers/swr/rasterizer/core/state.h index bb1336c429f..623e70a1519 100644 --- a/src/gallium/drivers/swr/rasterizer/core/state.h +++ b/src/gallium/drivers/swr/rasterizer/core/state.h @@ -201,9 +201,12 @@ struct SWR_VS_CONTEXT simdvertex* pVin; // IN: SIMD input vertex data store simdvertex* pVout; // OUT: SIMD output vertex data store - uint32_t InstanceID; // IN: Instance ID, constant across all verts of the SIMD - simdscalari VertexID; // IN: Vertex ID - simdscalari mask; // IN: Active mask for shader + uint32_t InstanceID; // IN: Instance ID, constant across all verts of the SIMD + simdscalari VertexID; // IN: Vertex ID + simdscalari mask; // IN: Active mask for shader +#if USE_SIMD16_FRONTEND + uint32_t AlternateOffset; // IN: amount to offset for interleaving even/odd simd8 in simd16vertex output +#endif }; ///////////////////////////////////////////////////////////////////////// -- 2.11.0