From 2621c460e39479113e8ca6192f78091481bcd458 Mon Sep 17 00:00:00 2001 From: Christoph Oelckers Date: Sun, 23 Apr 2017 10:39:07 +0200 Subject: [PATCH] Revert "- add CPU.bSSE2 branching support to softpoly" This reverts commit 545241aa06ef73320cf84c3ae57950186a01ca83. These commits are WIP stuff that should not be in the master branch. A work branch has been created to finalize this code. --- src/polyrenderer/drawers/poly_drawer32.h | 782 ------------------ src/polyrenderer/drawers/poly_drawer32_avx2.h | 739 ----------------- src/polyrenderer/drawers/poly_drawer32_sse2.h | 68 +- src/polyrenderer/drawers/poly_triangle.cpp | 201 ++--- src/polyrenderer/drawers/poly_triangle.h | 6 +- src/polyrenderer/drawers/screen_triangle.cpp | 277 ++----- src/polyrenderer/drawers/screen_triangle.h | 5 - src/polyrenderer/math/tri_matrix.cpp | 49 +- 8 files changed, 189 insertions(+), 1938 deletions(-) delete mode 100644 src/polyrenderer/drawers/poly_drawer32.h delete mode 100644 src/polyrenderer/drawers/poly_drawer32_avx2.h diff --git a/src/polyrenderer/drawers/poly_drawer32.h b/src/polyrenderer/drawers/poly_drawer32.h deleted file mode 100644 index 1d83eb1ba..000000000 --- a/src/polyrenderer/drawers/poly_drawer32.h +++ /dev/null @@ -1,782 +0,0 @@ -/* -** Projected triangle drawer -** Copyright (c) 2016 Magnus Norddahl -** -** This software is provided 'as-is', without any express or implied -** warranty. In no event will the authors be held liable for any damages -** arising from the use of this software. -** -** Permission is granted to anyone to use this software for any purpose, -** including commercial applications, and to alter it and redistribute it -** freely, subject to the following restrictions: -** -** 1. The origin of this software must not be misrepresented; you must not -** claim that you wrote the original software. If you use this software -** in a product, an acknowledgment in the product documentation would be -** appreciated but is not required. -** 2. Altered source versions must be plainly marked as such, and must not be -** misrepresented as being the original software. -** 3. This notice may not be removed or altered from any source distribution. -** -*/ - -#pragma once - -#include "screen_triangle.h" - -namespace TriScreenDrawerModes -{ - namespace - { - struct BgraColor - { - uint32_t b, g, r, a; - BgraColor() { } - BgraColor(uint32_t c) : b(BPART(c)), g(GPART(c)), r(RPART(c)), a(APART(c)) { } - BgraColor &operator=(uint32_t c) { b = BPART(c); g = GPART(c); r = RPART(c); a = APART(c); return *this; } - operator uint32_t() const { return MAKEARGB(a, r, g, b); } - }; - } - - template - FORCEINLINE unsigned int Sample32(int32_t u, int32_t v, const uint32_t *texPixels, int texWidth, int texHeight, uint32_t oneU, uint32_t oneV, uint32_t color, const uint32_t *translation) - { - uint32_t texel; - if (SamplerT::Mode == (int)Samplers::Shaded || SamplerT::Mode == (int)Samplers::Stencil || SamplerT::Mode == (int)Samplers::Fill || SamplerT::Mode == (int)Samplers::Fuzz) - { - return color; - } - else if (SamplerT::Mode == (int)Samplers::Translated) - { - const uint8_t *texpal = (const uint8_t *)texPixels; - uint32_t texelX = ((((uint32_t)u << 8) >> 16) * texWidth) >> 16; - uint32_t texelY = ((((uint32_t)v << 8) >> 16) * texHeight) >> 16; - return translation[texpal[texelX * texHeight + texelY]]; - } - else if (FilterModeT::Mode == (int)FilterModes::Nearest) - { - uint32_t texelX = ((((uint32_t)u << 8) >> 16) * texWidth) >> 16; - uint32_t texelY = ((((uint32_t)v << 8) >> 16) * texHeight) >> 16; - texel = texPixels[texelX * texHeight + texelY]; - } - else - { - u -= oneU >> 1; - v -= oneV >> 1; - - unsigned int frac_x0 = (((uint32_t)u << 8) >> FRACBITS) * texWidth; - unsigned int frac_x1 = ((((uint32_t)u << 8) + oneU) >> FRACBITS) * texWidth; - unsigned int frac_y0 = (((uint32_t)v << 8) >> FRACBITS) * texHeight; - unsigned int frac_y1 = ((((uint32_t)v << 8) + oneV) >> FRACBITS) * texHeight; - unsigned int x0 = frac_x0 >> FRACBITS; - unsigned int x1 = frac_x1 >> FRACBITS; - unsigned int y0 = frac_y0 >> FRACBITS; - unsigned int y1 = frac_y1 >> FRACBITS; - - unsigned int p00 = texPixels[x0 * texHeight + y0]; - unsigned int p01 = texPixels[x0 * texHeight + y1]; - unsigned int p10 = texPixels[x1 * texHeight + y0]; - unsigned int p11 = texPixels[x1 * texHeight + y1]; - - unsigned int inv_a = (frac_x1 >> (FRACBITS - 4)) & 15; - unsigned int inv_b = (frac_y1 >> (FRACBITS - 4)) & 15; - unsigned int a = 16 - inv_a; - unsigned int b = 16 - inv_b; - - unsigned int sred = (RPART(p00) * (a * b) + RPART(p01) * (inv_a * b) + RPART(p10) * (a * inv_b) + RPART(p11) * (inv_a * inv_b) + 127) >> 8; - unsigned int sgreen = (GPART(p00) * (a * b) + GPART(p01) * (inv_a * b) + GPART(p10) * (a * inv_b) + GPART(p11) * (inv_a * inv_b) + 127) >> 8; - unsigned int sblue = (BPART(p00) * (a * b) + BPART(p01) * (inv_a * b) + BPART(p10) * (a * inv_b) + BPART(p11) * (inv_a * inv_b) + 127) >> 8; - unsigned int salpha = (APART(p00) * (a * b) + APART(p01) * (inv_a * b) + APART(p10) * (a * inv_b) + APART(p11) * (inv_a * inv_b) + 127) >> 8; - - texel = (salpha << 24) | (sred << 16) | (sgreen << 8) | sblue; - } - - if (SamplerT::Mode == (int)Samplers::Skycap) - { - int start_fade = 2; // How fast it should fade out - - int alpha_top = clamp(v >> (16 - start_fade), 0, 256); - int alpha_bottom = clamp(((2 << 24) - v) >> (16 - start_fade), 0, 256); - int a = MIN(alpha_top, alpha_bottom); - int inv_a = 256 - a; - - uint32_t r = RPART(texel); - uint32_t g = GPART(texel); - uint32_t b = BPART(texel); - uint32_t fg_a = APART(texel); - uint32_t bg_red = RPART(color); - uint32_t bg_green = GPART(color); - uint32_t bg_blue = BPART(color); - r = (r * a + bg_red * inv_a + 127) >> 8; - g = (g * a + bg_green * inv_a + 127) >> 8; - b = (b * a + bg_blue * inv_a + 127) >> 8; - return MAKEARGB(fg_a, r, g, b); - } - else - { - return texel; - } - } - - template - FORCEINLINE unsigned int SampleShade32(int32_t u, int32_t v, const uint32_t *texPixels, int texWidth, int texHeight, int &fuzzpos) - { - if (SamplerT::Mode == (int)Samplers::Shaded) - { - const uint8_t *texpal = (const uint8_t *)texPixels; - uint32_t texelX = ((((uint32_t)u << 8) >> 16) * texWidth) >> 16; - uint32_t texelY = ((((uint32_t)v << 8) >> 16) * texHeight) >> 16; - unsigned int sampleshadeout = texpal[texelX * texHeight + texelY]; - sampleshadeout += sampleshadeout >> 7; // 255 -> 256 - return sampleshadeout; - } - else if (SamplerT::Mode == (int)Samplers::Stencil) - { - uint32_t texelX = ((((uint32_t)u << 8) >> 16) * texWidth) >> 16; - uint32_t texelY = ((((uint32_t)v << 8) >> 16) * texHeight) >> 16; - unsigned int sampleshadeout = APART(texPixels[texelX * texHeight + texelY]); - sampleshadeout += sampleshadeout >> 7; // 255 -> 256 - return sampleshadeout; - } - else if (SamplerT::Mode == (int)Samplers::Fuzz) - { - uint32_t texelX = ((((uint32_t)u << 8) >> 16) * texWidth) >> 16; - uint32_t texelY = ((((uint32_t)v << 8) >> 16) * texHeight) >> 16; - unsigned int sampleshadeout = APART(texPixels[texelX * texHeight + texelY]); - sampleshadeout += sampleshadeout >> 7; // 255 -> 256 - sampleshadeout = (sampleshadeout * fuzzcolormap[fuzzpos++]) >> 5; - if (fuzzpos >= FUZZTABLE) fuzzpos = 0; - return sampleshadeout; - } - else - { - return 0; - } - } - - template - FORCEINLINE BgraColor Shade32(BgraColor fgcolor, BgraColor mlight, uint32_t desaturate, uint32_t inv_desaturate, BgraColor shade_fade, BgraColor shade_light) - { - if (ShadeModeT::Mode == (int)ShadeMode::Simple) - { - fgcolor.r = (fgcolor.r * mlight.r) >> 8; - fgcolor.g = (fgcolor.g * mlight.g) >> 8; - fgcolor.b = (fgcolor.b * mlight.b) >> 8; - } - else if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - uint32_t intensity = ((fgcolor.r * 77 + fgcolor.g * 143 + fgcolor.b * 37) >> 8) * desaturate; - fgcolor.r = (((shade_fade.r + ((fgcolor.r * inv_desaturate + intensity) >> 8) * mlight.r) >> 8) * shade_light.r) >> 8; - fgcolor.g = (((shade_fade.g + ((fgcolor.g * inv_desaturate + intensity) >> 8) * mlight.g) >> 8) * shade_light.g) >> 8; - fgcolor.b = (((shade_fade.b + ((fgcolor.b * inv_desaturate + intensity) >> 8) * mlight.b) >> 8) * shade_light.b) >> 8; - } - return fgcolor; - } - - template - FORCEINLINE BgraColor Blend32(BgraColor fgcolor, BgraColor bgcolor, uint32_t ifgcolor, uint32_t ifgshade, uint32_t srcalpha, uint32_t destalpha) - { - if (BlendT::Mode == (int)BlendModes::Opaque) - { - return fgcolor; - } - else if (BlendT::Mode == (int)BlendModes::Masked) - { - return (ifgcolor == 0) ? bgcolor : fgcolor; - } - else if (BlendT::Mode == (int)BlendModes::AddSrcColorOneMinusSrcColor) - { - uint32_t srcred = fgcolor.r + (fgcolor.r >> 7); - uint32_t srcgreen = fgcolor.g + (fgcolor.g >> 7); - uint32_t srcblue = fgcolor.b + (fgcolor.b >> 7); - uint32_t inv_srcred = 256 - srcred; - uint32_t inv_srcgreen = 256 - srcgreen; - uint32_t inv_srcblue = 256 - srcblue; - - BgraColor outcolor; - outcolor.r = (fgcolor.r * srcred + bgcolor.r * inv_srcred) >> 8; - outcolor.g = (fgcolor.g * srcgreen + bgcolor.g * inv_srcgreen) >> 8; - outcolor.b = (fgcolor.b * srcblue + bgcolor.b * inv_srcblue) >> 8; - outcolor.a = 255; - return outcolor; - } - else if (BlendT::Mode == (int)BlendModes::Shaded) - { - uint32_t alpha = ifgshade; - uint32_t inv_alpha = 256 - alpha; - - BgraColor outcolor; - outcolor.r = (fgcolor.r * alpha + bgcolor.r * inv_alpha) >> 8; - outcolor.g = (fgcolor.g * alpha + bgcolor.g * inv_alpha) >> 8; - outcolor.b = (fgcolor.b * alpha + bgcolor.b * inv_alpha) >> 8; - outcolor.a = 255; - return outcolor; - } - else if (BlendT::Mode == (int)BlendModes::AddClampShaded) - { - uint32_t alpha = ifgshade; - BgraColor outcolor; - outcolor.r = ((fgcolor.r * alpha) >> 8) + bgcolor.r; - outcolor.g = ((fgcolor.g * alpha) >> 8) + bgcolor.g; - outcolor.b = ((fgcolor.b * alpha) >> 8) + bgcolor.b; - outcolor.a = 255; - return outcolor; - } - else - { - uint32_t alpha = APART(ifgcolor); - alpha += alpha >> 7; // 255->256 - uint32_t inv_alpha = 256 - alpha; - - uint32_t bgalpha = (destalpha * alpha + (inv_alpha << 8) + 128) >> 8; - uint32_t fgalpha = (srcalpha * alpha + 128) >> 8; - - fgcolor.r *= fgalpha; - fgcolor.g *= fgalpha; - fgcolor.b *= fgalpha; - bgcolor.r *= bgalpha; - bgcolor.g *= bgalpha; - bgcolor.b *= bgalpha; - - BgraColor outcolor; - if (BlendT::Mode == (int)BlendModes::AddClamp) - { - outcolor.r = MIN((fgcolor.r + bgcolor.r) >> 8, 255); - outcolor.g = MIN((fgcolor.g + bgcolor.g) >> 8, 255); - outcolor.b = MIN((fgcolor.b + bgcolor.b) >> 8, 255); - } - else if (BlendT::Mode == (int)BlendModes::SubClamp) - { - outcolor.r = MAX(int32_t(fgcolor.r - bgcolor.r) >> 8, 0); - outcolor.g = MAX(int32_t(fgcolor.g - bgcolor.g) >> 8, 0); - outcolor.b = MAX(int32_t(fgcolor.b - bgcolor.b) >> 8, 0); - } - else if (BlendT::Mode == (int)BlendModes::RevSubClamp) - { - outcolor.r = MAX(int32_t(bgcolor.r - fgcolor.r) >> 8, 0); - outcolor.g = MAX(int32_t(bgcolor.g - fgcolor.g) >> 8, 0); - outcolor.b = MAX(int32_t(bgcolor.b - fgcolor.b) >> 8, 0); - } - outcolor.a = 255; - return outcolor; - } - } -} - -template -class TriScreenDrawer32 -{ -public: - static void Execute(int x, int y, uint32_t mask0, uint32_t mask1, const TriDrawTriangleArgs *args) - { - using namespace TriScreenDrawerModes; - - bool is_simple_shade = args->uniforms->SimpleShade(); - - if (SamplerT::Mode == (int)Samplers::Texture) - { - bool is_nearest_filter = args->uniforms->NearestFilter(); - - if (is_simple_shade) - { - if (is_nearest_filter) - DrawBlock(x, y, mask0, mask1, args); - else - DrawBlock(x, y, mask0, mask1, args); - } - else - { - if (is_nearest_filter) - DrawBlock(x, y, mask0, mask1, args); - else - DrawBlock(x, y, mask0, mask1, args); - } - } - else if (SamplerT::Mode == (int)Samplers::Fuzz) - { - DrawBlock(x, y, mask0, mask1, args); - } - else // no linear filtering for translated, shaded, stencil, fill or skycap - { - if (is_simple_shade) - { - DrawBlock(x, y, mask0, mask1, args); - } - else - { - DrawBlock(x, y, mask0, mask1, args); - } - } - } - -private: - template - FORCEINLINE static void DrawBlock(int destX, int destY, uint32_t mask0, uint32_t mask1, const TriDrawTriangleArgs *args) - { - using namespace TriScreenDrawerModes; - - bool is_fixed_light = args->uniforms->FixedLight(); - uint32_t lightmask = is_fixed_light ? 0 : 0xffffffff; - uint32_t srcalpha = args->uniforms->SrcAlpha(); - uint32_t destalpha = args->uniforms->DestAlpha(); - - int fuzzpos = (ScreenTriangle::FuzzStart + destX * 123 + destY) % FUZZTABLE; - - // Calculate gradients - const TriVertex &v1 = *args->v1; - ScreenTriangleStepVariables gradientX = args->gradientX; - ScreenTriangleStepVariables gradientY = args->gradientY; - ScreenTriangleStepVariables blockPosY; - blockPosY.W = v1.w + gradientX.W * (destX - v1.x) + gradientY.W * (destY - v1.y); - blockPosY.U = v1.u * v1.w + gradientX.U * (destX - v1.x) + gradientY.U * (destY - v1.y); - blockPosY.V = v1.v * v1.w + gradientX.V * (destX - v1.x) + gradientY.V * (destY - v1.y); - gradientX.W *= 8.0f; - gradientX.U *= 8.0f; - gradientX.V *= 8.0f; - - // Output - uint32_t * RESTRICT destOrg = (uint32_t*)args->dest; - int pitch = args->pitch; - uint32_t *dest = destOrg + destX + destY * pitch; - - // Light - uint32_t light = args->uniforms->Light(); - float shade = 2.0f - (light + 12.0f) / 128.0f; - float globVis = args->uniforms->GlobVis() * (1.0f / 32.0f); - light += (light >> 7); // 255 -> 256 - - // Sampling stuff - uint32_t color = args->uniforms->Color(); - const uint32_t * RESTRICT translation = (const uint32_t *)args->uniforms->Translation(); - const uint32_t * RESTRICT texPixels = (const uint32_t *)args->uniforms->TexturePixels(); - uint32_t texWidth = args->uniforms->TextureWidth(); - uint32_t texHeight = args->uniforms->TextureHeight(); - uint32_t oneU, oneV; - if (SamplerT::Mode != (int)Samplers::Fill) - { - oneU = ((0x800000 + texWidth - 1) / texWidth) * 2 + 1; - oneV = ((0x800000 + texHeight - 1) / texHeight) * 2 + 1; - } - else - { - oneU = 0; - oneV = 0; - } - - // Shade constants - int inv_desaturate; - BgraColor shade_fade, shade_light; - int desaturate; - if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - shade_fade.r = args->uniforms->ShadeFadeRed(); - shade_fade.g = args->uniforms->ShadeFadeGreen(); - shade_fade.b = args->uniforms->ShadeFadeBlue(); - shade_light.r = args->uniforms->ShadeLightRed(); - shade_light.g = args->uniforms->ShadeLightGreen(); - shade_light.b = args->uniforms->ShadeLightBlue(); - desaturate = args->uniforms->ShadeDesaturate(); - inv_desaturate = 256 - desaturate; - } - else - { - inv_desaturate = 0; - shade_fade.r = 0; - shade_fade.g = 0; - shade_fade.b = 0; - shade_light.r = 0; - shade_light.g = 0; - shade_light.b = 0; - desaturate = 0; - } - - if (mask0 == 0xffffffff && mask1 == 0xffffffff) - { - for (int y = 0; y < 8; y++) - { - float rcpW = 0x01000000 / blockPosY.W; - int32_t posU = (int32_t)(blockPosY.U * rcpW); - int32_t posV = (int32_t)(blockPosY.V * rcpW); - - fixed_t lightpos = FRACUNIT - (int)(clamp(shade - MIN(24.0f / 32.0f, globVis * blockPosY.W), 0.0f, 31.0f / 32.0f) * (float)FRACUNIT); - lightpos = (lightpos & lightmask) | ((light << 8) & ~lightmask); - - ScreenTriangleStepVariables blockPosX = blockPosY; - blockPosX.W += gradientX.W; - blockPosX.U += gradientX.U; - blockPosX.V += gradientX.V; - - rcpW = 0x01000000 / blockPosX.W; - int32_t nextU = (int32_t)(blockPosX.U * rcpW); - int32_t nextV = (int32_t)(blockPosX.V * rcpW); - int32_t stepU = (nextU - posU) / 8; - int32_t stepV = (nextV - posV) / 8; - - fixed_t lightnext = FRACUNIT - (fixed_t)(clamp(shade - MIN(24.0f / 32.0f, globVis * blockPosX.W), 0.0f, 31.0f / 32.0f) * (float)FRACUNIT); - fixed_t lightstep = (lightnext - lightpos) / 8; - lightstep = lightstep & lightmask; - - for (int ix = 0; ix < 8; ix++) - { - // Load bgcolor - BgraColor bgcolor; - if (BlendT::Mode != (int)BlendModes::Opaque) - bgcolor = dest[ix]; - else - bgcolor = 0; - - // Sample fgcolor - unsigned int ifgcolor = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - // Setup light - int lightpos0 = lightpos >> 8; - lightpos += lightstep; - BgraColor mlight; - mlight.r = lightpos0; - mlight.g = lightpos0; - mlight.b = lightpos0; - - BgraColor shade_fade_lit; - if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - uint32_t inv_light = 256 - lightpos0; - shade_fade_lit.r = shade_fade.r * inv_light; - shade_fade_lit.g = shade_fade.g * inv_light; - shade_fade_lit.b = shade_fade.b * inv_light; - } - else - { - shade_fade_lit.r = 0; - shade_fade_lit.g = 0; - shade_fade_lit.b = 0; - } - - // Shade and blend - BgraColor fgcolor = Shade32(ifgcolor, mlight, desaturate, inv_desaturate, shade_fade_lit, shade_light); - BgraColor outcolor = Blend32(fgcolor, bgcolor, ifgcolor, ifgshade, srcalpha, destalpha); - - // Store result - dest[ix] = outcolor; - } - - blockPosY.W += gradientY.W; - blockPosY.U += gradientY.U; - blockPosY.V += gradientY.V; - - dest += pitch; - } - } - else - { - // mask0 loop: - for (int y = 0; y < 4; y++) - { - float rcpW = 0x01000000 / blockPosY.W; - int32_t posU = (int32_t)(blockPosY.U * rcpW); - int32_t posV = (int32_t)(blockPosY.V * rcpW); - - fixed_t lightpos = FRACUNIT - (fixed_t)(clamp(shade - MIN(24.0f / 32.0f, globVis * blockPosY.W), 0.0f, 31.0f / 32.0f) * (float)FRACUNIT); - lightpos = (lightpos & lightmask) | ((light << 8) & ~lightmask); - - ScreenTriangleStepVariables blockPosX = blockPosY; - blockPosX.W += gradientX.W; - blockPosX.U += gradientX.U; - blockPosX.V += gradientX.V; - - rcpW = 0x01000000 / blockPosX.W; - int32_t nextU = (int32_t)(blockPosX.U * rcpW); - int32_t nextV = (int32_t)(blockPosX.V * rcpW); - int32_t stepU = (nextU - posU) / 8; - int32_t stepV = (nextV - posV) / 8; - - fixed_t lightnext = FRACUNIT - (fixed_t)(clamp(shade - MIN(24.0f / 32.0f, globVis * blockPosX.W), 0.0f, 31.0f / 32.0f) * (float)FRACUNIT); - fixed_t lightstep = (lightnext - lightpos) / 8; - lightstep = lightstep & lightmask; - - for (int x = 0; x < 8; x++) - { - // Load bgcolor - BgraColor bgcolor; - if (BlendT::Mode != (int)BlendModes::Opaque) - { - if (mask0 & (1 << 31)) bgcolor = dest[x]; - } - else - bgcolor = 0; - - // Sample fgcolor - unsigned int ifgcolor = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - // Setup light - int lightpos0 = lightpos >> 8; - lightpos += lightstep; - BgraColor mlight; - mlight.r = lightpos0; - mlight.g = lightpos0; - mlight.b = lightpos0; - - BgraColor shade_fade_lit; - if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - uint32_t inv_light = 256 - lightpos0; - shade_fade_lit.r = shade_fade.r * inv_light; - shade_fade_lit.g = shade_fade.g * inv_light; - shade_fade_lit.b = shade_fade.b * inv_light; - } - else - { - shade_fade_lit.r = 0; - shade_fade_lit.g = 0; - shade_fade_lit.b = 0; - } - - // Shade and blend - BgraColor fgcolor = Shade32(ifgcolor, mlight, desaturate, inv_desaturate, shade_fade_lit, shade_light); - BgraColor outcolor = Blend32(fgcolor, bgcolor, ifgcolor, ifgshade, srcalpha, destalpha); - - // Store result - if (mask0 & (1 << 31)) dest[x] = outcolor; - - mask0 <<= 1; - } - - blockPosY.W += gradientY.W; - blockPosY.U += gradientY.U; - blockPosY.V += gradientY.V; - - dest += pitch; - } - - // mask1 loop: - for (int y = 0; y < 4; y++) - { - float rcpW = 0x01000000 / blockPosY.W; - int32_t posU = (int32_t)(blockPosY.U * rcpW); - int32_t posV = (int32_t)(blockPosY.V * rcpW); - - fixed_t lightpos = FRACUNIT - (fixed_t)(clamp(shade - MIN(24.0f / 32.0f, globVis * blockPosY.W), 0.0f, 31.0f / 32.0f) * (float)FRACUNIT); - lightpos = (lightpos & lightmask) | ((light << 8) & ~lightmask); - - ScreenTriangleStepVariables blockPosX = blockPosY; - blockPosX.W += gradientX.W; - blockPosX.U += gradientX.U; - blockPosX.V += gradientX.V; - - rcpW = 0x01000000 / blockPosX.W; - int32_t nextU = (int32_t)(blockPosX.U * rcpW); - int32_t nextV = (int32_t)(blockPosX.V * rcpW); - int32_t stepU = (nextU - posU) / 8; - int32_t stepV = (nextV - posV) / 8; - - fixed_t lightnext = FRACUNIT - (fixed_t)(clamp(shade - MIN(24.0f / 32.0f, globVis * blockPosX.W), 0.0f, 31.0f / 32.0f) * (float)FRACUNIT); - fixed_t lightstep = (lightnext - lightpos) / 8; - lightstep = lightstep & lightmask; - - for (int x = 0; x < 8; x++) - { - // Load bgcolor - BgraColor bgcolor; - if (BlendT::Mode != (int)BlendModes::Opaque) - { - if (mask1 & (1 << 31)) bgcolor = dest[x]; - } - else - bgcolor = 0; - - // Sample fgcolor - unsigned int ifgcolor = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - // Setup light - int lightpos0 = lightpos >> 8; - lightpos += lightstep; - BgraColor mlight; - mlight.r = lightpos0; - mlight.g = lightpos0; - mlight.b = lightpos0; - - BgraColor shade_fade_lit; - if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - uint32_t inv_light = 256 - lightpos0; - shade_fade_lit.r = shade_fade.r * inv_light; - shade_fade_lit.g = shade_fade.g * inv_light; - shade_fade_lit.b = shade_fade.b * inv_light; - } - else - { - shade_fade_lit.r = 0; - shade_fade_lit.g = 0; - shade_fade_lit.b = 0; - } - - // Shade and blend - BgraColor fgcolor = Shade32(ifgcolor, mlight, desaturate, inv_desaturate, shade_fade_lit, shade_light); - BgraColor outcolor = Blend32(fgcolor, bgcolor, ifgcolor, ifgshade, srcalpha, destalpha); - - // Store result - if (mask1 & (1 << 31)) dest[x] = outcolor; - - mask1 <<= 1; - } - - blockPosY.W += gradientY.W; - blockPosY.U += gradientY.U; - blockPosY.V += gradientY.V; - - dest += pitch; - } - } - } -}; - -template -class RectScreenDrawer32 -{ -public: - static void Execute(const void *destOrg, int destWidth, int destHeight, int destPitch, const RectDrawArgs *args, WorkerThreadData *thread) - { - using namespace TriScreenDrawerModes; - - if (args->SimpleShade()) - { - Loop(destOrg, destWidth, destHeight, destPitch, args, thread); - } - else - { - Loop(destOrg, destWidth, destHeight, destPitch, args, thread); - } - } - -private: - template - FORCEINLINE static void Loop(const void *destOrg, int destWidth, int destHeight, int destPitch, const RectDrawArgs *args, WorkerThreadData *thread) - { - using namespace TriScreenDrawerModes; - - int x0 = clamp((int)(args->X0() + 0.5f), 0, destWidth); - int x1 = clamp((int)(args->X1() + 0.5f), 0, destWidth); - int y0 = clamp((int)(args->Y0() + 0.5f), 0, destHeight); - int y1 = clamp((int)(args->Y1() + 0.5f), 0, destHeight); - - if (x1 <= x0 || y1 <= y0) - return; - - uint32_t srcalpha = args->SrcAlpha(); - uint32_t destalpha = args->DestAlpha(); - - // Setup step variables - float fstepU = (args->U1() - args->U0()) / (args->X1() - args->X0()); - float fstepV = (args->V1() - args->V0()) / (args->Y1() - args->Y0()); - uint32_t startU = (int32_t)((args->U0() + (x0 + 0.5f - args->X0()) * fstepU) * 0x1000000); - uint32_t startV = (int32_t)((args->V0() + (y0 + 0.5f - args->Y0()) * fstepV) * 0x1000000); - uint32_t stepU = (int32_t)(fstepU * 0x1000000); - uint32_t stepV = (int32_t)(fstepV * 0x1000000); - - // Sampling stuff - uint32_t color = args->Color(); - const uint32_t * RESTRICT translation = (const uint32_t *)args->Translation(); - const uint32_t * RESTRICT texPixels = (const uint32_t *)args->TexturePixels(); - uint32_t texWidth = args->TextureWidth(); - uint32_t texHeight = args->TextureHeight(); - uint32_t oneU, oneV; - if (SamplerT::Mode != (int)Samplers::Fill) - { - oneU = ((0x800000 + texWidth - 1) / texWidth) * 2 + 1; - oneV = ((0x800000 + texHeight - 1) / texHeight) * 2 + 1; - } - else - { - oneU = 0; - oneV = 0; - } - - // Setup light - uint32_t lightpos = args->Light(); - lightpos += lightpos >> 7; // 255 -> 256 - BgraColor mlight; - - // Shade constants - int inv_desaturate; - BgraColor shade_fade_lit, shade_light; - int desaturate; - if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - uint32_t inv_light = 256 - lightpos; - shade_fade_lit.r = args->ShadeFadeRed() * inv_light; - shade_fade_lit.g = args->ShadeFadeGreen() * inv_light; - shade_fade_lit.b = args->ShadeFadeBlue() * inv_light; - shade_light.r = args->ShadeLightRed(); - shade_light.g = args->ShadeLightGreen(); - shade_light.b = args->ShadeLightBlue(); - desaturate = args->ShadeDesaturate(); - inv_desaturate = 256 - desaturate; - mlight.r = lightpos; - mlight.g = lightpos; - mlight.b = lightpos; - } - else - { - inv_desaturate = 0; - shade_fade_lit.r = 0; - shade_fade_lit.g = 0; - shade_fade_lit.b = 0; - shade_light.r = 0; - shade_light.g = 0; - shade_light.b = 0; - desaturate = 0; - mlight.r = lightpos; - mlight.g = lightpos; - mlight.b = lightpos; - } - - int count = x1 - x0; - - int fuzzpos = (ScreenTriangle::FuzzStart + x0 * 123 + y0) % FUZZTABLE; - - uint32_t posV = startV; - for (int y = y0; y < y1; y++, posV += stepV) - { - int coreBlock = y / 8; - if (coreBlock % thread->num_cores != thread->core) - { - fuzzpos = (fuzzpos + count) % FUZZTABLE; - continue; - } - - uint32_t *dest = ((uint32_t*)destOrg) + y * destPitch + x0; - - uint32_t posU = startU; - for (int i = 0; i < count; i++) - { - // Load bgcolor - BgraColor bgcolor; - if (BlendT::Mode != (int)BlendModes::Opaque) - bgcolor = *dest; - else - bgcolor = 0; - - // Sample fgcolor - unsigned int ifgcolor = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - - // Shade and blend - BgraColor fgcolor = Shade32(ifgcolor, mlight, desaturate, inv_desaturate, shade_fade_lit, shade_light); - BgraColor outcolor = Blend32(fgcolor, bgcolor, ifgcolor, ifgshade, srcalpha, destalpha); - - // Store result - *dest = outcolor; - dest++; - } - } - } -}; diff --git a/src/polyrenderer/drawers/poly_drawer32_avx2.h b/src/polyrenderer/drawers/poly_drawer32_avx2.h deleted file mode 100644 index 9091ae21a..000000000 --- a/src/polyrenderer/drawers/poly_drawer32_avx2.h +++ /dev/null @@ -1,739 +0,0 @@ -/* -** Projected triangle drawer -** Copyright (c) 2016 Magnus Norddahl -** -** This software is provided 'as-is', without any express or implied -** warranty. In no event will the authors be held liable for any damages -** arising from the use of this software. -** -** Permission is granted to anyone to use this software for any purpose, -** including commercial applications, and to alter it and redistribute it -** freely, subject to the following restrictions: -** -** 1. The origin of this software must not be misrepresented; you must not -** claim that you wrote the original software. If you use this software -** in a product, an acknowledgment in the product documentation would be -** appreciated but is not required. -** 2. Altered source versions must be plainly marked as such, and must not be -** misrepresented as being the original software. -** 3. This notice may not be removed or altered from any source distribution. -** -*/ - -#pragma once - -#include "screen_triangle.h" - -#ifdef _MSC_VER -#pragma warning(disable: 4752) // warning C4752 : found Intel(R) Advanced Vector Extensions; consider using /arch:AVX -#endif - -namespace TriScreenDrawerModes -{ - template - FORCEINLINE unsigned int VECTORCALL Sample32_AVX2(int32_t u, int32_t v, const uint32_t *texPixels, int texWidth, int texHeight, uint32_t oneU, uint32_t oneV, uint32_t color, const uint32_t *translation) - { - uint32_t texel; - if (SamplerT::Mode == (int)Samplers::Shaded || SamplerT::Mode == (int)Samplers::Stencil || SamplerT::Mode == (int)Samplers::Fill || SamplerT::Mode == (int)Samplers::Fuzz) - { - return color; - } - else if (SamplerT::Mode == (int)Samplers::Translated) - { - const uint8_t *texpal = (const uint8_t *)texPixels; - uint32_t texelX = ((((uint32_t)u << 8) >> 16) * texWidth) >> 16; - uint32_t texelY = ((((uint32_t)v << 8) >> 16) * texHeight) >> 16; - return translation[texpal[texelX * texHeight + texelY]]; - } - else if (FilterModeT::Mode == (int)FilterModes::Nearest) - { - uint32_t texelX = ((((uint32_t)u << 8) >> 16) * texWidth) >> 16; - uint32_t texelY = ((((uint32_t)v << 8) >> 16) * texHeight) >> 16; - texel = texPixels[texelX * texHeight + texelY]; - } - else - { - u -= oneU >> 1; - v -= oneV >> 1; - - unsigned int frac_x0 = (((uint32_t)u << 8) >> FRACBITS) * texWidth; - unsigned int frac_x1 = ((((uint32_t)u << 8) + oneU) >> FRACBITS) * texWidth; - unsigned int frac_y0 = (((uint32_t)v << 8) >> FRACBITS) * texHeight; - unsigned int frac_y1 = ((((uint32_t)v << 8) + oneV) >> FRACBITS) * texHeight; - unsigned int x0 = frac_x0 >> FRACBITS; - unsigned int x1 = frac_x1 >> FRACBITS; - unsigned int y0 = frac_y0 >> FRACBITS; - unsigned int y1 = frac_y1 >> FRACBITS; - - unsigned int p00 = texPixels[x0 * texHeight + y0]; - unsigned int p01 = texPixels[x0 * texHeight + y1]; - unsigned int p10 = texPixels[x1 * texHeight + y0]; - unsigned int p11 = texPixels[x1 * texHeight + y1]; - - unsigned int inv_a = (frac_x1 >> (FRACBITS - 4)) & 15; - unsigned int inv_b = (frac_y1 >> (FRACBITS - 4)) & 15; - unsigned int a = 16 - inv_a; - unsigned int b = 16 - inv_b; - - unsigned int sred = (RPART(p00) * (a * b) + RPART(p01) * (inv_a * b) + RPART(p10) * (a * inv_b) + RPART(p11) * (inv_a * inv_b) + 127) >> 8; - unsigned int sgreen = (GPART(p00) * (a * b) + GPART(p01) * (inv_a * b) + GPART(p10) * (a * inv_b) + GPART(p11) * (inv_a * inv_b) + 127) >> 8; - unsigned int sblue = (BPART(p00) * (a * b) + BPART(p01) * (inv_a * b) + BPART(p10) * (a * inv_b) + BPART(p11) * (inv_a * inv_b) + 127) >> 8; - unsigned int salpha = (APART(p00) * (a * b) + APART(p01) * (inv_a * b) + APART(p10) * (a * inv_b) + APART(p11) * (inv_a * inv_b) + 127) >> 8; - - texel = (salpha << 24) | (sred << 16) | (sgreen << 8) | sblue; - } - - if (SamplerT::Mode == (int)Samplers::Skycap) - { - int start_fade = 2; // How fast it should fade out - - int alpha_top = clamp(v >> (16 - start_fade), 0, 256); - int alpha_bottom = clamp(((2 << 24) - v) >> (16 - start_fade), 0, 256); - int a = MIN(alpha_top, alpha_bottom); - int inv_a = 256 - a; - - uint32_t r = RPART(texel); - uint32_t g = GPART(texel); - uint32_t b = BPART(texel); - uint32_t fg_a = APART(texel); - uint32_t bg_red = RPART(color); - uint32_t bg_green = GPART(color); - uint32_t bg_blue = BPART(color); - r = (r * a + bg_red * inv_a + 127) >> 8; - g = (g * a + bg_green * inv_a + 127) >> 8; - b = (b * a + bg_blue * inv_a + 127) >> 8; - return MAKEARGB(fg_a, r, g, b); - } - else - { - return texel; - } - } - - template - FORCEINLINE unsigned int VECTORCALL SampleShade32_AVX2(int32_t u, int32_t v, const uint32_t *texPixels, int texWidth, int texHeight, int &fuzzpos) - { - if (SamplerT::Mode == (int)Samplers::Shaded) - { - const uint8_t *texpal = (const uint8_t *)texPixels; - uint32_t texelX = ((((uint32_t)u << 8) >> 16) * texWidth) >> 16; - uint32_t texelY = ((((uint32_t)v << 8) >> 16) * texHeight) >> 16; - unsigned int sampleshadeout = texpal[texelX * texHeight + texelY]; - sampleshadeout += sampleshadeout >> 7; // 255 -> 256 - return sampleshadeout; - } - else if (SamplerT::Mode == (int)Samplers::Stencil) - { - uint32_t texelX = ((((uint32_t)u << 8) >> 16) * texWidth) >> 16; - uint32_t texelY = ((((uint32_t)v << 8) >> 16) * texHeight) >> 16; - unsigned int sampleshadeout = APART(texPixels[texelX * texHeight + texelY]); - sampleshadeout += sampleshadeout >> 7; // 255 -> 256 - return sampleshadeout; - } - else if (SamplerT::Mode == (int)Samplers::Fuzz) - { - uint32_t texelX = ((((uint32_t)u << 8) >> 16) * texWidth) >> 16; - uint32_t texelY = ((((uint32_t)v << 8) >> 16) * texHeight) >> 16; - unsigned int sampleshadeout = APART(texPixels[texelX * texHeight + texelY]); - sampleshadeout += sampleshadeout >> 7; // 255 -> 256 - sampleshadeout = (sampleshadeout * fuzzcolormap[fuzzpos++]) >> 5; - if (fuzzpos >= FUZZTABLE) fuzzpos = 0; - return sampleshadeout; - } - else - { - return 0; - } - } - - template - FORCEINLINE __m256i VECTORCALL Shade32_AVX2(__m256i fgcolor, __m256i mlight, __m256i desaturate, __m256i inv_desaturate, __m256i shade_fade, __m256i shade_light) - { - if (ShadeModeT::Mode == (int)ShadeMode::Simple) - { - fgcolor = _mm256_srli_epi16(_mm256_mullo_epi16(fgcolor, mlight), 8); - } - else if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - __m256i intensity = _mm256_mullo_epi16(fgcolor, _mm256_set_epi16(0, 77, 143, 37, 0, 77, 143, 37, 0, 77, 143, 37, 0, 77, 143, 37)); - intensity = _mm256_add_epi16(intensity, _mm256_srli_epi64(intensity, 32)); - intensity = _mm256_add_epi16(intensity, _mm256_srli_epi64(intensity, 16)); - intensity = _mm256_srli_epi16(intensity, 8); - intensity = _mm256_mullo_epi16(intensity, desaturate); - intensity = _mm256_shufflehi_epi16(_mm256_shufflelo_epi16(intensity, _MM_SHUFFLE(3, 0, 0, 0)), _MM_SHUFFLE(3, 0, 0, 0)); - - fgcolor = _mm256_srli_epi16(_mm256_add_epi16(_mm256_mullo_epi16(fgcolor, inv_desaturate), intensity), 8); - fgcolor = _mm256_mullo_epi16(fgcolor, mlight); - fgcolor = _mm256_srli_epi16(_mm256_add_epi16(shade_fade, fgcolor), 8); - fgcolor = _mm256_srli_epi16(_mm256_mullo_epi16(fgcolor, shade_light), 8); - } - return fgcolor; - } - - template - FORCEINLINE __m256i VECTORCALL Blend32_AVX2(__m256i fgcolor, __m256i bgcolor, __m256i ifgcolor, __m256i ifgshade, __m256i srcalpha, __m256i destalpha) - { - if (BlendT::Mode == (int)BlendModes::Opaque) - { - __m256i outcolor = fgcolor; - outcolor = _mm256_packus_epi16(outcolor, _mm256_setzero_si256()); - return outcolor; - } - else if (BlendT::Mode == (int)BlendModes::Masked) - { - __m256i mask = _mm256_cmpeq_epi32(_mm256_packus_epi16(fgcolor, _mm256_setzero_si256()), _mm256_setzero_si256()); - mask = _mm256_unpacklo_epi8(mask, _mm256_setzero_si256()); - __m256i outcolor = _mm256_or_si256(_mm256_and_si256(mask, bgcolor), _mm256_andnot_si256(mask, fgcolor)); - outcolor = _mm256_packus_epi16(outcolor, _mm256_setzero_si256()); - outcolor = _mm256_or_si256(outcolor, _mm256_set1_epi32(0xff000000)); - return outcolor; - } - else if (BlendT::Mode == (int)BlendModes::AddSrcColorOneMinusSrcColor) - { - __m256i inv_srccolor = _mm256_sub_epi16(_mm256_set1_epi16(256), _mm256_add_epi16(fgcolor, _mm256_srli_epi16(fgcolor, 7))); - __m256i outcolor = _mm256_add_epi16(fgcolor, _mm256_srli_epi16(_mm256_mullo_epi16(bgcolor, inv_srccolor), 8)); - outcolor = _mm256_packus_epi16(outcolor, _mm256_setzero_si256()); - return outcolor; - } - else if (BlendT::Mode == (int)BlendModes::Shaded) - { - ifgshade = _mm256_srli_epi32(_mm256_add_epi32(_mm256_mul_epu32(ifgshade, srcalpha), _mm256_set1_epi32(128)), 8); - __m256i alpha = _mm256_shufflehi_epi16(_mm256_shufflelo_epi16(ifgshade, _MM_SHUFFLE(0, 0, 0, 0)), _MM_SHUFFLE(0, 0, 0, 0)); - __m256i inv_alpha = _mm256_sub_epi16(_mm256_set1_epi16(256), alpha); - - fgcolor = _mm256_mullo_epi16(fgcolor, alpha); - bgcolor = _mm256_mullo_epi16(bgcolor, inv_alpha); - __m256i outcolor = _mm256_srli_epi16(_mm256_add_epi16(fgcolor, bgcolor), 8); - outcolor = _mm256_packus_epi16(outcolor, _mm256_setzero_si256()); - outcolor = _mm256_or_si256(outcolor, _mm256_set1_epi32(0xff000000)); - return outcolor; - } - else if (BlendT::Mode == (int)BlendModes::AddClampShaded) - { - ifgshade = _mm256_srli_epi32(_mm256_add_epi32(_mm256_mul_epu32(ifgshade, srcalpha), _mm256_set1_epi32(128)), 8); - __m256i alpha = _mm256_shufflehi_epi16(_mm256_shufflelo_epi16(ifgshade, _MM_SHUFFLE(0, 0, 0, 0)), _MM_SHUFFLE(0, 0, 0, 0)); - __m256i inv_alpha = _mm256_sub_epi16(_mm256_set1_epi16(256), alpha); - - fgcolor = _mm256_srli_epi16(_mm256_mullo_epi16(fgcolor, alpha), 8); - __m256i outcolor = _mm256_add_epi16(fgcolor, bgcolor); - outcolor = _mm256_packus_epi16(outcolor, _mm256_setzero_si256()); - outcolor = _mm256_or_si256(outcolor, _mm256_set1_epi32(0xff000000)); - return outcolor; - } - else - { - __m256i alpha = _mm256_shufflehi_epi16(_mm256_shufflelo_epi16(ifgcolor, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3)); - alpha = _mm256_srli_epi16(_mm256_add_epi16(alpha, _mm256_srli_epi16(alpha, 7)), 1); // 255->128 - __m256i inv_alpha = _mm256_sub_epi16(_mm256_set1_epi16(128), alpha); - - __m256i bgalpha = _mm256_srli_epi16(_mm256_add_epi16(_mm256_add_epi16(_mm256_mullo_epi16(destalpha, alpha), _mm256_slli_epi16(inv_alpha, 8)), _mm256_set1_epi32(64)), 7); - __m256i fgalpha = _mm256_srli_epi16(_mm256_add_epi16(_mm256_mullo_epi16(srcalpha, alpha), _mm256_set1_epi32(64)), 7); - - fgcolor = _mm256_mullo_epi16(fgcolor, fgalpha); - bgcolor = _mm256_mullo_epi16(bgcolor, bgalpha); - - __m256i fg_lo = _mm256_unpacklo_epi16(fgcolor, _mm256_setzero_si256()); - __m256i bg_lo = _mm256_unpacklo_epi16(bgcolor, _mm256_setzero_si256()); - __m256i fg_hi = _mm256_unpackhi_epi16(fgcolor, _mm256_setzero_si256()); - __m256i bg_hi = _mm256_unpackhi_epi16(bgcolor, _mm256_setzero_si256()); - - __m256i out_lo, out_hi; - if (BlendT::Mode == (int)BlendModes::AddClamp) - { - out_lo = _mm256_add_epi32(fg_lo, bg_lo); - out_hi = _mm256_add_epi32(fg_hi, bg_hi); - } - else if (BlendT::Mode == (int)BlendModes::SubClamp) - { - out_lo = _mm256_sub_epi32(fg_lo, bg_lo); - out_hi = _mm256_sub_epi32(fg_hi, bg_hi); - } - else if (BlendT::Mode == (int)BlendModes::RevSubClamp) - { - out_lo = _mm256_sub_epi32(bg_lo, fg_lo); - out_hi = _mm256_sub_epi32(bg_hi, fg_hi); - } - - out_lo = _mm256_srai_epi32(out_lo, 8); - out_hi = _mm256_srai_epi32(out_hi, 8); - __m256i outcolor = _mm256_packs_epi32(out_lo, out_hi); - outcolor = _mm256_packus_epi16(outcolor, _mm256_setzero_si256()); - outcolor = _mm256_or_si256(outcolor, _mm256_set1_epi32(0xff000000)); - return outcolor; - } - } -} - -template -class TriScreenDrawer32_AVX2 -{ -public: - static void Execute(int x, int y, uint32_t mask0, uint32_t mask1, const TriDrawTriangleArgs *args) - { - using namespace TriScreenDrawerModes; - - bool is_simple_shade = args->uniforms->SimpleShade(); - - if (SamplerT::Mode == (int)Samplers::Texture) - { - bool is_nearest_filter = args->uniforms->NearestFilter(); - - if (is_simple_shade) - { - if (is_nearest_filter) - DrawBlock(x, y, mask0, mask1, args); - else - DrawBlock(x, y, mask0, mask1, args); - } - else - { - if (is_nearest_filter) - DrawBlock(x, y, mask0, mask1, args); - else - DrawBlock(x, y, mask0, mask1, args); - } - } - else if (SamplerT::Mode == (int)Samplers::Fuzz) - { - DrawBlock(x, y, mask0, mask1, args); - } - else // no linear filtering for translated, shaded, stencil, fill or skycap - { - if (is_simple_shade) - { - DrawBlock(x, y, mask0, mask1, args); - } - else - { - DrawBlock(x, y, mask0, mask1, args); - } - } - } - -private: - template - FORCEINLINE static void VECTORCALL DrawBlock(int destX, int destY, uint32_t mask0, uint32_t mask1, const TriDrawTriangleArgs *args) - { - using namespace TriScreenDrawerModes; - - bool is_fixed_light = args->uniforms->FixedLight(); - __m128i lightmask = _mm_set1_epi32(is_fixed_light ? 0 : 0xffffffff); - __m256i srcalpha = _mm256_set1_epi16(args->uniforms->SrcAlpha()); - __m256i destalpha = _mm256_set1_epi16(args->uniforms->DestAlpha()); - - int fuzzpos = (ScreenTriangle::FuzzStart + destX * 123 + destY) % FUZZTABLE; - - // Light - uint32_t light = args->uniforms->Light(); - float shade = MIN(2.0f - (light + 12.0f) / 128.0f, 31.0f / 32.0f); - float globVis = args->uniforms->GlobVis() * (1.0f / 32.0f); - light += (light >> 7); // 255 -> 256 - light <<= 8; - __m128i fixedlight = _mm_set1_epi32(light); - - // Calculate gradients - const TriVertex &v1 = *args->v1; - __m128 gradientX = _mm_setr_ps(args->gradientX.W, args->gradientX.U, args->gradientX.V, 0.0f); - __m128 gradientY = _mm_setr_ps(args->gradientY.W, args->gradientY.U, args->gradientY.V, 0.0f); - __m128 blockPosY = _mm_add_ps(_mm_add_ps( - _mm_setr_ps(v1.w, v1.u * v1.w, v1.v * v1.w, globVis), - _mm_mul_ps(gradientX, _mm_set1_ps(destX - v1.x))), - _mm_mul_ps(gradientY, _mm_set1_ps(destY - v1.y))); - gradientX = _mm_mul_ps(gradientX, _mm_set1_ps(8.0f)); - - // Output - uint32_t * RESTRICT destOrg = (uint32_t*)args->dest; - int pitch = args->pitch; - uint32_t *dest = destOrg + destX + destY * pitch; - int offset_next_line = pitch - 8; - - // Sampling stuff - uint32_t color = args->uniforms->Color(); - const uint32_t * RESTRICT translation = (const uint32_t *)args->uniforms->Translation(); - const uint32_t * RESTRICT texPixels = (const uint32_t *)args->uniforms->TexturePixels(); - uint32_t texWidth = args->uniforms->TextureWidth(); - uint32_t texHeight = args->uniforms->TextureHeight(); - uint32_t oneU, oneV; - if (SamplerT::Mode != (int)Samplers::Fill) - { - oneU = ((0x800000 + texWidth - 1) / texWidth) * 2 + 1; - oneV = ((0x800000 + texHeight - 1) / texHeight) * 2 + 1; - } - else - { - oneU = 0; - oneV = 0; - } - - // Shade constants - __m256i inv_desaturate, shade_fade, shade_light; - __m256i desaturate; - if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - inv_desaturate = _mm256_setr_epi16( - 256, 256 - args->uniforms->ShadeDesaturate(), 256 - args->uniforms->ShadeDesaturate(), 256 - args->uniforms->ShadeDesaturate(), - 256, 256 - args->uniforms->ShadeDesaturate(), 256 - args->uniforms->ShadeDesaturate(), 256 - args->uniforms->ShadeDesaturate(), - 256, 256 - args->uniforms->ShadeDesaturate(), 256 - args->uniforms->ShadeDesaturate(), 256 - args->uniforms->ShadeDesaturate(), - 256, 256 - args->uniforms->ShadeDesaturate(), 256 - args->uniforms->ShadeDesaturate(), 256 - args->uniforms->ShadeDesaturate()); - shade_fade = _mm256_set_epi16( - args->uniforms->ShadeFadeAlpha(), args->uniforms->ShadeFadeRed(), args->uniforms->ShadeFadeGreen(), args->uniforms->ShadeFadeBlue(), - args->uniforms->ShadeFadeAlpha(), args->uniforms->ShadeFadeRed(), args->uniforms->ShadeFadeGreen(), args->uniforms->ShadeFadeBlue(), - args->uniforms->ShadeFadeAlpha(), args->uniforms->ShadeFadeRed(), args->uniforms->ShadeFadeGreen(), args->uniforms->ShadeFadeBlue(), - args->uniforms->ShadeFadeAlpha(), args->uniforms->ShadeFadeRed(), args->uniforms->ShadeFadeGreen(), args->uniforms->ShadeFadeBlue()); - shade_light = _mm256_set_epi16( - args->uniforms->ShadeLightAlpha(), args->uniforms->ShadeLightRed(), args->uniforms->ShadeLightGreen(), args->uniforms->ShadeLightBlue(), - args->uniforms->ShadeLightAlpha(), args->uniforms->ShadeLightRed(), args->uniforms->ShadeLightGreen(), args->uniforms->ShadeLightBlue(), - args->uniforms->ShadeLightAlpha(), args->uniforms->ShadeLightRed(), args->uniforms->ShadeLightGreen(), args->uniforms->ShadeLightBlue(), - args->uniforms->ShadeLightAlpha(), args->uniforms->ShadeLightRed(), args->uniforms->ShadeLightGreen(), args->uniforms->ShadeLightBlue()); - desaturate = _mm256_sub_epi16(_mm256_set1_epi16(256), inv_desaturate); - } - else - { - inv_desaturate = _mm256_setzero_si256(); - shade_fade = _mm256_setzero_si256(); - shade_fade = _mm256_setzero_si256(); - shade_light = _mm256_setzero_si256(); - desaturate = _mm256_setzero_si256(); - } - - if (mask0 == 0xffffffff && mask1 == 0xffffffff) - { - for (int y = 0; y < 8; y++) - { - __m128 blockPosX = _mm_add_ps(blockPosY, gradientX); - __m128 W = _mm_shuffle_ps(blockPosY, blockPosX, _MM_SHUFFLE(0, 0, 0, 0)); - __m128 rcpW = _mm_div_ps(_mm_set1_ps((float)0x01000000), W); - __m128i posUV = _mm_cvtps_epi32(_mm_mul_ps(_mm_shuffle_ps(blockPosY, blockPosX, _MM_SHUFFLE(2, 1, 2, 1)), rcpW)); - - __m128 vis = _mm_mul_ps(_mm_shuffle_ps(blockPosY, blockPosX, _MM_SHUFFLE(3, 3, 3, 3)), W); - __m128i lightpospair = _mm_sub_epi32( - _mm_set1_epi32(FRACUNIT), - _mm_cvtps_epi32(_mm_mul_ps( - _mm_max_ps(_mm_sub_ps(_mm_set1_ps(shade), _mm_min_ps(_mm_set1_ps(24.0f / 32.0f), vis)), _mm_setzero_ps()), - _mm_set1_ps((float)FRACUNIT)))); - lightpospair = _mm_or_si128(_mm_and_si128(lightmask, lightpospair), _mm_andnot_si128(lightmask, fixedlight)); - - int32_t posU = _mm_cvtsi128_si32(posUV); - int32_t posV = _mm_cvtsi128_si32(_mm_srli_si128(posUV, 4)); - int32_t nextU = _mm_cvtsi128_si32(_mm_srli_si128(posUV, 8)); - int32_t nextV = _mm_cvtsi128_si32(_mm_srli_si128(posUV, 12)); - int32_t lightpos = _mm_cvtsi128_si32(lightpospair); - int32_t lightnext = _mm_cvtsi128_si32(_mm_srli_si128(lightpospair, 8)); - int32_t stepU = (nextU - posU) >> 3; - int32_t stepV = (nextV - posV) >> 3; - fixed_t lightstep = (lightnext - lightpos) >> 3; - - for (int ix = 0; ix < 2; ix++) - { - // Load bgcolor - __m256i bgcolor; - if (BlendT::Mode != (int)BlendModes::Opaque) - { - __m128i bgpacked = _mm_loadu_si128((__m128i*)dest); - bgcolor = _mm256_set_m128i(_mm_unpackhi_epi8(bgpacked, _mm_setzero_si128()), _mm_unpacklo_epi8(bgpacked, _mm_setzero_si128())); - } - else - bgcolor = _mm256_setzero_si256(); - - // Sample fgcolor - unsigned int ifgcolor0 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade0 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - unsigned int ifgcolor1 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade1 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - unsigned int ifgcolor2 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade2 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - unsigned int ifgcolor3 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade3 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - // Setup light - int lightpos0 = lightpos >> 8; - lightpos += lightstep; - int lightpos1 = lightpos >> 8; - lightpos += lightstep; - int lightpos2 = lightpos >> 8; - lightpos += lightstep; - int lightpos3 = lightpos >> 8; - lightpos += lightstep; - __m256i mlight = _mm256_set_epi16( - 256, lightpos3, lightpos3, lightpos3, - 256, lightpos2, lightpos2, lightpos2, - 256, lightpos1, lightpos1, lightpos1, - 256, lightpos0, lightpos0, lightpos0); - - __m256i shade_fade_lit; - if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - __m256i inv_light = _mm256_sub_epi16(_mm256_set_epi16(0, 256, 256, 256, 0, 256, 256, 256, 0, 256, 256, 256, 0, 256, 256, 256), mlight); - shade_fade_lit = _mm256_mullo_epi16(shade_fade, inv_light); - } - else - { - shade_fade_lit = _mm256_setzero_si256(); - } - - // Shade and blend - __m128i fgpacked = _mm_set_epi32(ifgcolor3, ifgcolor2, ifgcolor1, ifgcolor0); - __m128i shadepacked = _mm_set_epi32(ifgshade3, ifgshade2, ifgshade1, ifgshade0); - __m256i mifgcolor = _mm256_set_m128i(_mm_unpackhi_epi8(fgpacked, _mm_setzero_si128()), _mm_unpacklo_epi8(fgpacked, _mm_setzero_si128())); - __m256i mifgshade = _mm256_set_m128i(_mm_unpackhi_epi32(shadepacked, shadepacked), _mm_unpacklo_epi32(shadepacked, shadepacked)); - __m256i fgcolor = mifgcolor; - fgcolor = Shade32_AVX2(fgcolor, mlight, desaturate, inv_desaturate, shade_fade_lit, shade_light); - __m256i outcolor = Blend32_AVX2(fgcolor, bgcolor, mifgcolor, mifgshade, srcalpha, destalpha); - - // Store result - _mm_storeu_si128((__m128i*)dest, _mm_or_si128(_mm256_extracti128_si256(outcolor, 0), _mm_slli_si128(_mm256_extracti128_si256(outcolor, 1), 8))); - dest += 4; - } - - blockPosY = _mm_add_ps(blockPosY, gradientY); - - dest += offset_next_line; - } - } - else - { - // mask0 loop: - for (int y = 0; y < 4; y++) - { - __m128 blockPosX = _mm_add_ps(blockPosY, gradientX); - __m128 W = _mm_shuffle_ps(blockPosY, blockPosX, _MM_SHUFFLE(0, 0, 0, 0)); - __m128 rcpW = _mm_div_ps(_mm_set1_ps((float)0x01000000), W); - __m128i posUV = _mm_cvtps_epi32(_mm_mul_ps(_mm_shuffle_ps(blockPosY, blockPosX, _MM_SHUFFLE(2, 1, 2, 1)), rcpW)); - - __m128 vis = _mm_mul_ps(_mm_shuffle_ps(blockPosY, blockPosX, _MM_SHUFFLE(3, 3, 3, 3)), W); - __m128i lightpospair = _mm_sub_epi32( - _mm_set1_epi32(FRACUNIT), - _mm_cvtps_epi32(_mm_mul_ps( - _mm_max_ps(_mm_sub_ps(_mm_set1_ps(shade), _mm_min_ps(_mm_set1_ps(24.0f / 32.0f), vis)), _mm_setzero_ps()), - _mm_set1_ps((float)FRACUNIT)))); - lightpospair = _mm_or_si128(_mm_and_si128(lightmask, lightpospair), _mm_andnot_si128(lightmask, fixedlight)); - - int32_t posU = _mm_cvtsi128_si32(posUV); - int32_t posV = _mm_cvtsi128_si32(_mm_srli_si128(posUV, 4)); - int32_t nextU = _mm_cvtsi128_si32(_mm_srli_si128(posUV, 8)); - int32_t nextV = _mm_cvtsi128_si32(_mm_srli_si128(posUV, 12)); - int32_t lightpos = _mm_cvtsi128_si32(lightpospair); - int32_t lightnext = _mm_cvtsi128_si32(_mm_srli_si128(lightpospair, 8)); - int32_t stepU = (nextU - posU) >> 3; - int32_t stepV = (nextV - posV) >> 3; - fixed_t lightstep = (lightnext - lightpos) >> 3; - - for (int x = 0; x < 2; x++) - { - // Load bgcolor - uint32_t desttmp[4]; - __m256i bgcolor; - if (BlendT::Mode != (int)BlendModes::Opaque) - { - if (mask0 & (1 << 31)) desttmp[0] = dest[0]; - if (mask0 & (1 << 30)) desttmp[1] = dest[1]; - if (mask0 & (1 << 29)) desttmp[2] = dest[2]; - if (mask0 & (1 << 28)) desttmp[3] = dest[3]; - - __m128i bgpacked = _mm_loadu_si128((__m128i*)(desttmp)); - bgcolor = _mm256_set_m128i(_mm_unpackhi_epi8(bgpacked, _mm_setzero_si128()), _mm_unpacklo_epi8(bgpacked, _mm_setzero_si128())); - } - else - bgcolor = _mm256_setzero_si256(); - - // Sample fgcolor - unsigned int ifgcolor0 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade0 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - unsigned int ifgcolor1 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade1 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - unsigned int ifgcolor2 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade2 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - unsigned int ifgcolor3 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade3 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - // Setup light - int lightpos0 = lightpos >> 8; - lightpos += lightstep; - int lightpos1 = lightpos >> 8; - lightpos += lightstep; - int lightpos2 = lightpos >> 8; - lightpos += lightstep; - int lightpos3 = lightpos >> 8; - lightpos += lightstep; - __m256i mlight = _mm256_set_epi16( - 256, lightpos3, lightpos3, lightpos3, - 256, lightpos2, lightpos2, lightpos2, - 256, lightpos1, lightpos1, lightpos1, - 256, lightpos0, lightpos0, lightpos0); - - __m256i shade_fade_lit; - if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - __m256i inv_light = _mm256_sub_epi16(_mm256_set_epi16(0, 256, 256, 256, 0, 256, 256, 256, 0, 256, 256, 256, 0, 256, 256, 256), mlight); - shade_fade_lit = _mm256_mullo_epi16(shade_fade, inv_light); - } - else - { - shade_fade_lit = _mm256_setzero_si256(); - } - - // Shade and blend - __m128i fgpacked = _mm_set_epi32(ifgcolor3, ifgcolor2, ifgcolor1, ifgcolor0); - __m128i shadepacked = _mm_set_epi32(ifgshade3, ifgshade2, ifgshade1, ifgshade0); - __m256i mifgcolor = _mm256_set_m128i(_mm_unpackhi_epi8(fgpacked, _mm_setzero_si128()), _mm_unpacklo_epi8(fgpacked, _mm_setzero_si128())); - __m256i mifgshade = _mm256_set_m128i(_mm_unpackhi_epi32(shadepacked, shadepacked), _mm_unpacklo_epi32(shadepacked, shadepacked)); - __m256i fgcolor = mifgcolor; - fgcolor = Shade32_AVX2(fgcolor, mlight, desaturate, inv_desaturate, shade_fade_lit, shade_light); - __m256i outcolor = Blend32_AVX2(fgcolor, bgcolor, mifgcolor, mifgshade, srcalpha, destalpha); - - // Store result - _mm_storeu_si128((__m128i*)desttmp, _mm_or_si128(_mm256_extracti128_si256(outcolor, 0), _mm_slli_si128(_mm256_extracti128_si256(outcolor, 1), 8))); - if (mask0 & (1 << 31)) dest[0] = desttmp[0]; - if (mask0 & (1 << 30)) dest[1] = desttmp[1]; - if (mask0 & (1 << 29)) dest[2] = desttmp[2]; - if (mask0 & (1 << 28)) dest[3] = desttmp[3]; - - mask0 <<= 4; - dest += 4; - } - - blockPosY = _mm_add_ps(blockPosY, gradientY); - - dest += offset_next_line; - } - - // mask1 loop: - for (int y = 0; y < 4; y++) - { - __m128 blockPosX = _mm_add_ps(blockPosY, gradientX); - __m128 W = _mm_shuffle_ps(blockPosY, blockPosX, _MM_SHUFFLE(0, 0, 0, 0)); - __m128 rcpW = _mm_div_ps(_mm_set1_ps((float)0x01000000), W); - __m128i posUV = _mm_cvtps_epi32(_mm_mul_ps(_mm_shuffle_ps(blockPosY, blockPosX, _MM_SHUFFLE(2, 1, 2, 1)), rcpW)); - - __m128 vis = _mm_mul_ps(_mm_shuffle_ps(blockPosY, blockPosX, _MM_SHUFFLE(3, 3, 3, 3)), W); - __m128i lightpospair = _mm_sub_epi32( - _mm_set1_epi32(FRACUNIT), - _mm_cvtps_epi32(_mm_mul_ps( - _mm_max_ps(_mm_sub_ps(_mm_set1_ps(shade), _mm_min_ps(_mm_set1_ps(24.0f / 32.0f), vis)), _mm_setzero_ps()), - _mm_set1_ps((float)FRACUNIT)))); - lightpospair = _mm_or_si128(_mm_and_si128(lightmask, lightpospair), _mm_andnot_si128(lightmask, fixedlight)); - - int32_t posU = _mm_cvtsi128_si32(posUV); - int32_t posV = _mm_cvtsi128_si32(_mm_srli_si128(posUV, 4)); - int32_t nextU = _mm_cvtsi128_si32(_mm_srli_si128(posUV, 8)); - int32_t nextV = _mm_cvtsi128_si32(_mm_srli_si128(posUV, 12)); - int32_t lightpos = _mm_cvtsi128_si32(lightpospair); - int32_t lightnext = _mm_cvtsi128_si32(_mm_srli_si128(lightpospair, 8)); - int32_t stepU = (nextU - posU) >> 3; - int32_t stepV = (nextV - posV) >> 3; - fixed_t lightstep = (lightnext - lightpos) >> 3; - - for (int x = 0; x < 2; x++) - { - // Load bgcolor - uint32_t desttmp[4]; - __m256i bgcolor; - if (BlendT::Mode != (int)BlendModes::Opaque) - { - if (mask1 & (1 << 31)) desttmp[0] = dest[0]; - if (mask1 & (1 << 30)) desttmp[1] = dest[1]; - if (mask1 & (1 << 29)) desttmp[2] = dest[2]; - if (mask1 & (1 << 28)) desttmp[3] = dest[3]; - - __m128i bgpacked = _mm_loadu_si128((__m128i*)(desttmp)); - bgcolor = _mm256_set_m128i(_mm_unpackhi_epi8(bgpacked, _mm_setzero_si128()), _mm_unpacklo_epi8(bgpacked, _mm_setzero_si128())); - } - else - bgcolor = _mm256_setzero_si256(); - - // Sample fgcolor - unsigned int ifgcolor0 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade0 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - unsigned int ifgcolor1 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade1 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - unsigned int ifgcolor2 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade2 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - unsigned int ifgcolor3 = Sample32_AVX2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - unsigned int ifgshade3 = SampleShade32_AVX2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); - posU += stepU; - posV += stepV; - - // Setup light - int lightpos0 = lightpos >> 8; - lightpos += lightstep; - int lightpos1 = lightpos >> 8; - lightpos += lightstep; - int lightpos2 = lightpos >> 8; - lightpos += lightstep; - int lightpos3 = lightpos >> 8; - lightpos += lightstep; - __m256i mlight = _mm256_set_epi16( - 256, lightpos3, lightpos3, lightpos3, - 256, lightpos2, lightpos2, lightpos2, - 256, lightpos1, lightpos1, lightpos1, - 256, lightpos0, lightpos0, lightpos0); - - __m256i shade_fade_lit; - if (ShadeModeT::Mode == (int)ShadeMode::Advanced) - { - __m256i inv_light = _mm256_sub_epi16(_mm256_set_epi16(0, 256, 256, 256, 0, 256, 256, 256, 0, 256, 256, 256, 0, 256, 256, 256), mlight); - shade_fade_lit = _mm256_mullo_epi16(shade_fade, inv_light); - } - else - { - shade_fade_lit = _mm256_setzero_si256(); - } - - // Shade and blend - __m128i fgpacked = _mm_set_epi32(ifgcolor3, ifgcolor2, ifgcolor1, ifgcolor0); - __m128i shadepacked = _mm_set_epi32(ifgshade3, ifgshade2, ifgshade1, ifgshade0); - __m256i mifgcolor = _mm256_set_m128i(_mm_unpackhi_epi8(fgpacked, _mm_setzero_si128()), _mm_unpacklo_epi8(fgpacked, _mm_setzero_si128())); - __m256i mifgshade = _mm256_set_m128i(_mm_unpackhi_epi32(shadepacked, shadepacked), _mm_unpacklo_epi32(shadepacked, shadepacked)); - __m256i fgcolor = mifgcolor; - fgcolor = Shade32_AVX2(fgcolor, mlight, desaturate, inv_desaturate, shade_fade_lit, shade_light); - __m256i outcolor = Blend32_AVX2(fgcolor, bgcolor, mifgcolor, mifgshade, srcalpha, destalpha); - - // Store result - _mm_storeu_si128((__m128i*)desttmp, _mm_or_si128(_mm256_extracti128_si256(outcolor, 0), _mm_slli_si128(_mm256_extracti128_si256(outcolor, 1), 8))); - if (mask1 & (1 << 31)) dest[0] = desttmp[0]; - if (mask1 & (1 << 30)) dest[1] = desttmp[1]; - if (mask1 & (1 << 29)) dest[2] = desttmp[2]; - if (mask1 & (1 << 28)) dest[3] = desttmp[3]; - - mask1 <<= 4; - dest += 4; - } - - blockPosY = _mm_add_ps(blockPosY, gradientY); - - dest += offset_next_line; - } - } - } -}; diff --git a/src/polyrenderer/drawers/poly_drawer32_sse2.h b/src/polyrenderer/drawers/poly_drawer32_sse2.h index 2f690f7e8..5125c93c7 100644 --- a/src/polyrenderer/drawers/poly_drawer32_sse2.h +++ b/src/polyrenderer/drawers/poly_drawer32_sse2.h @@ -27,7 +27,7 @@ namespace TriScreenDrawerModes { template - FORCEINLINE unsigned int VECTORCALL Sample32_SSE2(int32_t u, int32_t v, const uint32_t *texPixels, int texWidth, int texHeight, uint32_t oneU, uint32_t oneV, uint32_t color, const uint32_t *translation) + FORCEINLINE unsigned int VECTORCALL Sample32(int32_t u, int32_t v, const uint32_t *texPixels, int texWidth, int texHeight, uint32_t oneU, uint32_t oneV, uint32_t color, const uint32_t *translation) { uint32_t texel; if (SamplerT::Mode == (int)Samplers::Shaded || SamplerT::Mode == (int)Samplers::Stencil || SamplerT::Mode == (int)Samplers::Fill || SamplerT::Mode == (int)Samplers::Fuzz) @@ -107,7 +107,7 @@ namespace TriScreenDrawerModes } template - FORCEINLINE unsigned int VECTORCALL SampleShade32_SSE2(int32_t u, int32_t v, const uint32_t *texPixels, int texWidth, int texHeight, int &fuzzpos) + FORCEINLINE unsigned int VECTORCALL SampleShade32(int32_t u, int32_t v, const uint32_t *texPixels, int texWidth, int texHeight, int &fuzzpos) { if (SamplerT::Mode == (int)Samplers::Shaded) { @@ -143,7 +143,7 @@ namespace TriScreenDrawerModes } template - FORCEINLINE __m128i VECTORCALL Shade32_SSE2(__m128i fgcolor, __m128i mlight, unsigned int ifgcolor0, unsigned int ifgcolor1, int desaturate, __m128i inv_desaturate, __m128i shade_fade, __m128i shade_light) + FORCEINLINE __m128i VECTORCALL Shade32(__m128i fgcolor, __m128i mlight, unsigned int ifgcolor0, unsigned int ifgcolor1, int desaturate, __m128i inv_desaturate, __m128i shade_fade, __m128i shade_light) { if (ShadeModeT::Mode == (int)ShadeMode::Simple) { @@ -172,7 +172,7 @@ namespace TriScreenDrawerModes } template - FORCEINLINE __m128i VECTORCALL Blend32_SSE2(__m128i fgcolor, __m128i bgcolor, unsigned int ifgcolor0, unsigned int ifgcolor1, unsigned int ifgshade0, unsigned int ifgshade1, uint32_t srcalpha, uint32_t destalpha) + FORCEINLINE __m128i VECTORCALL Blend32(__m128i fgcolor, __m128i bgcolor, unsigned int ifgcolor0, unsigned int ifgcolor1, unsigned int ifgshade0, unsigned int ifgshade1, uint32_t srcalpha, uint32_t destalpha) { if (BlendT::Mode == (int)BlendModes::Opaque) { @@ -275,7 +275,7 @@ namespace TriScreenDrawerModes } template -class TriScreenDrawer32_SSE2 +class TriScreenDrawer32 { public: static void Execute(int x, int y, uint32_t mask0, uint32_t mask1, const TriDrawTriangleArgs *args) @@ -430,13 +430,13 @@ private: // Sample fgcolor unsigned int ifgcolor[2], ifgshade[2]; - ifgcolor[0] = Sample32_SSE2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - ifgshade[0] = SampleShade32_SSE2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); + ifgcolor[0] = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); + ifgshade[0] = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); posU += stepU; posV += stepV; - ifgcolor[1] = Sample32_SSE2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - ifgshade[1] = SampleShade32_SSE2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); + ifgcolor[1] = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); + ifgshade[1] = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); posU += stepU; posV += stepV; @@ -460,8 +460,8 @@ private: // Shade and blend __m128i fgcolor = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)ifgcolor), _mm_setzero_si128()); - fgcolor = Shade32_SSE2(fgcolor, mlight, ifgcolor[0], ifgcolor[1], desaturate, inv_desaturate, shade_fade_lit, shade_light); - __m128i outcolor = Blend32_SSE2(fgcolor, bgcolor, ifgcolor[0], ifgcolor[1], ifgshade[0], ifgshade[1], srcalpha, destalpha); + fgcolor = Shade32(fgcolor, mlight, ifgcolor[0], ifgcolor[1], desaturate, inv_desaturate, shade_fade_lit, shade_light); + __m128i outcolor = Blend32(fgcolor, bgcolor, ifgcolor[0], ifgcolor[1], ifgshade[0], ifgshade[1], srcalpha, destalpha); // Store result _mm_storel_epi64((__m128i*)(dest + ix * 2), outcolor); @@ -517,13 +517,13 @@ private: // Sample fgcolor unsigned int ifgcolor[2], ifgshade[2]; - ifgcolor[0] = Sample32_SSE2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - ifgshade[0] = SampleShade32_SSE2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); + ifgcolor[0] = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); + ifgshade[0] = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); posU += stepU; posV += stepV; - ifgcolor[1] = Sample32_SSE2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - ifgshade[1] = SampleShade32_SSE2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); + ifgcolor[1] = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); + ifgshade[1] = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); posU += stepU; posV += stepV; @@ -547,8 +547,8 @@ private: // Shade and blend __m128i fgcolor = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)ifgcolor), _mm_setzero_si128()); - fgcolor = Shade32_SSE2(fgcolor, mlight, ifgcolor[0], ifgcolor[1], desaturate, inv_desaturate, shade_fade_lit, shade_light); - __m128i outcolor = Blend32_SSE2(fgcolor, bgcolor, ifgcolor[0], ifgcolor[1], ifgshade[0], ifgshade[1], srcalpha, destalpha); + fgcolor = Shade32(fgcolor, mlight, ifgcolor[0], ifgcolor[1], desaturate, inv_desaturate, shade_fade_lit, shade_light); + __m128i outcolor = Blend32(fgcolor, bgcolor, ifgcolor[0], ifgcolor[1], ifgshade[0], ifgshade[1], srcalpha, destalpha); // Store result _mm_storel_epi64((__m128i*)desttmp, outcolor); @@ -606,13 +606,13 @@ private: // Sample fgcolor unsigned int ifgcolor[2], ifgshade[2]; - ifgcolor[0] = Sample32_SSE2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - ifgshade[0] = SampleShade32_SSE2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); + ifgcolor[0] = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); + ifgshade[0] = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); posU += stepU; posV += stepV; - ifgcolor[1] = Sample32_SSE2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - ifgshade[1] = SampleShade32_SSE2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); + ifgcolor[1] = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); + ifgshade[1] = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); posU += stepU; posV += stepV; @@ -636,8 +636,8 @@ private: // Shade and blend __m128i fgcolor = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)ifgcolor), _mm_setzero_si128()); - fgcolor = Shade32_SSE2(fgcolor, mlight, ifgcolor[0], ifgcolor[1], desaturate, inv_desaturate, shade_fade_lit, shade_light); - __m128i outcolor = Blend32_SSE2(fgcolor, bgcolor, ifgcolor[0], ifgcolor[1], ifgshade[0], ifgshade[1], srcalpha, destalpha); + fgcolor = Shade32(fgcolor, mlight, ifgcolor[0], ifgcolor[1], desaturate, inv_desaturate, shade_fade_lit, shade_light); + __m128i outcolor = Blend32(fgcolor, bgcolor, ifgcolor[0], ifgcolor[1], ifgshade[0], ifgshade[1], srcalpha, destalpha); // Store result _mm_storel_epi64((__m128i*)desttmp, outcolor); @@ -658,7 +658,7 @@ private: }; template -class RectScreenDrawer32_SSE2 +class RectScreenDrawer32 { public: static void Execute(const void *destOrg, int destWidth, int destHeight, int destPitch, const RectDrawArgs *args, WorkerThreadData *thread) @@ -780,18 +780,18 @@ private: // Sample fgcolor unsigned int ifgcolor[2], ifgshade[2]; - ifgcolor[0] = Sample32_SSE2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - ifgshade[0] = SampleShade32_SSE2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); + ifgcolor[0] = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); + ifgshade[0] = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); posU += stepU; - ifgcolor[1] = Sample32_SSE2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - ifgshade[1] = SampleShade32_SSE2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); + ifgcolor[1] = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); + ifgshade[1] = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); posU += stepU; // Shade and blend __m128i fgcolor = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)ifgcolor), _mm_setzero_si128()); - fgcolor = Shade32_SSE2(fgcolor, mlight, ifgcolor[0], ifgcolor[1], desaturate, inv_desaturate, shade_fade_lit, shade_light); - __m128i outcolor = Blend32_SSE2(fgcolor, bgcolor, ifgcolor[0], ifgcolor[1], ifgshade[0], ifgshade[1], srcalpha, destalpha); + fgcolor = Shade32(fgcolor, mlight, ifgcolor[0], ifgcolor[1], desaturate, inv_desaturate, shade_fade_lit, shade_light); + __m128i outcolor = Blend32(fgcolor, bgcolor, ifgcolor[0], ifgcolor[1], ifgshade[0], ifgshade[1], srcalpha, destalpha); // Store result _mm_storel_epi64((__m128i*)dest, outcolor); @@ -809,16 +809,16 @@ private: // Sample fgcolor unsigned int ifgcolor[2], ifgshade[2]; - ifgcolor[0] = Sample32_SSE2(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); - ifgshade[0] = SampleShade32_SSE2(posU, posV, texPixels, texWidth, texHeight, fuzzpos); + ifgcolor[0] = Sample32(posU, posV, texPixels, texWidth, texHeight, oneU, oneV, color, translation); + ifgshade[0] = SampleShade32(posU, posV, texPixels, texWidth, texHeight, fuzzpos); ifgcolor[1] = 0; ifgshade[1] = 0; posU += stepU; // Shade and blend __m128i fgcolor = _mm_unpacklo_epi8(_mm_loadl_epi64((__m128i*)ifgcolor), _mm_setzero_si128()); - fgcolor = Shade32_SSE2(fgcolor, mlight, ifgcolor[0], ifgcolor[1], desaturate, inv_desaturate, shade_fade_lit, shade_light); - __m128i outcolor = Blend32_SSE2(fgcolor, bgcolor, ifgcolor[0], ifgcolor[1], ifgshade[0], ifgshade[1], srcalpha, destalpha); + fgcolor = Shade32(fgcolor, mlight, ifgcolor[0], ifgcolor[1], desaturate, inv_desaturate, shade_fade_lit, shade_light); + __m128i outcolor = Blend32(fgcolor, bgcolor, ifgcolor[0], ifgcolor[1], ifgshade[0], ifgshade[1], srcalpha, destalpha); // Store result *dest = _mm_cvtsi128_si32(outcolor); diff --git a/src/polyrenderer/drawers/poly_triangle.cpp b/src/polyrenderer/drawers/poly_triangle.cpp index e4d91a65e..84dd13897 100644 --- a/src/polyrenderer/drawers/poly_triangle.cpp +++ b/src/polyrenderer/drawers/poly_triangle.cpp @@ -152,8 +152,14 @@ ShadedTriVertex PolyTriangleDrawer::shade_vertex(const TriMatrix &objectToClip, return sv; } -void PolyTriangleDrawer::clip_to_viewport(TriVertex *clippedvert, int numclipvert) +void PolyTriangleDrawer::draw_shaded_triangle(const ShadedTriVertex *vert, bool ccw, TriDrawTriangleArgs *args, WorkerThreadData *thread) { + // Cull, clip and generate additional vertices as needed + TriVertex clippedvert[max_additional_vertices]; + int numclipvert = clipedge(vert, clippedvert); + +#ifdef NO_SSE + // Map to 2D viewport: for (int j = 0; j < numclipvert; j++) { auto &v = clippedvert[j]; @@ -168,11 +174,8 @@ void PolyTriangleDrawer::clip_to_viewport(TriVertex *clippedvert, int numclipver v.x = viewport_x + viewport_width * (1.0f + v.x) * 0.5f; v.y = viewport_y + viewport_height * (1.0f - v.y) * 0.5f; } -} - -#ifndef NO_SSE -void PolyTriangleDrawer::clip_to_viewport_sse2(TriVertex *clippedvert, int numclipvert) -{ +#else + // Map to 2D viewport: __m128 mviewport_x = _mm_set1_ps((float)viewport_x); __m128 mviewport_y = _mm_set1_ps((float)viewport_y); __m128 mviewport_halfwidth = _mm_set1_ps(viewport_width * 0.5f); @@ -203,21 +206,8 @@ void PolyTriangleDrawer::clip_to_viewport_sse2(TriVertex *clippedvert, int numcl _mm_storeu_ps(&clippedvert[j + 2].x, vz); _mm_storeu_ps(&clippedvert[j + 3].x, vw); } -} #endif -void PolyTriangleDrawer::draw_shaded_triangle(const ShadedTriVertex *vert, bool ccw, TriDrawTriangleArgs *args, WorkerThreadData *thread) -{ - // Cull, clip and generate additional vertices as needed - TriVertex clippedvert[max_additional_vertices]; - int numclipvert = CPU.bSSE2 ? clipedge_sse2(vert, clippedvert) : clipedge(vert, clippedvert); - - // Map to 2D viewport: - if (CPU.bSSE2) - clip_to_viewport_sse2(clippedvert, numclipvert); - else - clip_to_viewport(clippedvert, numclipvert); - // Keep varyings in -128 to 128 range if possible if (numclipvert > 0) { @@ -266,6 +256,7 @@ int PolyTriangleDrawer::clipedge(const ShadedTriVertex *verts, TriVertex *clippe // halfspace clip distances static const int numclipdistances = 7; +#ifdef NO_SSE float clipdistance[numclipdistances * 3]; bool needsclipping = false; float *clipd = clipdistance; @@ -292,6 +283,43 @@ int PolyTriangleDrawer::clipedge(const ShadedTriVertex *verts, TriVertex *clippe } return 3; } +#else + __m128 mx = _mm_loadu_ps(&verts[0].x); + __m128 my = _mm_loadu_ps(&verts[1].x); + __m128 mz = _mm_loadu_ps(&verts[2].x); + __m128 mw = _mm_setzero_ps(); + _MM_TRANSPOSE4_PS(mx, my, mz, mw); + __m128 clipd0 = _mm_add_ps(mx, mw); + __m128 clipd1 = _mm_sub_ps(mw, mx); + __m128 clipd2 = _mm_add_ps(my, mw); + __m128 clipd3 = _mm_sub_ps(mw, my); + __m128 clipd4 = _mm_add_ps(mz, mw); + __m128 clipd5 = _mm_sub_ps(mw, mz); + __m128 clipd6 = _mm_setr_ps(verts[0].clipDistance0, verts[1].clipDistance0, verts[2].clipDistance0, 0.0f); + __m128 mneedsclipping = _mm_cmplt_ps(clipd0, _mm_setzero_ps()); + mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd1, _mm_setzero_ps())); + mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd2, _mm_setzero_ps())); + mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd3, _mm_setzero_ps())); + mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd4, _mm_setzero_ps())); + mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd5, _mm_setzero_ps())); + mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd6, _mm_setzero_ps())); + if (_mm_movemask_ps(mneedsclipping) == 0) + { + for (int i = 0; i < 3; i++) + { + memcpy(clippedvert + i, verts + i, sizeof(TriVertex)); + } + return 3; + } + float clipdistance[numclipdistances * 4]; + _mm_storeu_ps(clipdistance, clipd0); + _mm_storeu_ps(clipdistance + 4, clipd1); + _mm_storeu_ps(clipdistance + 8, clipd2); + _mm_storeu_ps(clipdistance + 12, clipd3); + _mm_storeu_ps(clipdistance + 16, clipd4); + _mm_storeu_ps(clipdistance + 20, clipd5); + _mm_storeu_ps(clipdistance + 24, clipd6); +#endif // use barycentric weights while clipping vertices float weights[max_additional_vertices * 3 * 2]; @@ -314,6 +342,7 @@ int PolyTriangleDrawer::clipedge(const ShadedTriVertex *verts, TriVertex *clippe for (int i = 0; i < inputverts; i++) { int j = (i + 1) % inputverts; +#ifdef NO_SSE float clipdistance1 = clipdistance[0 * numclipdistances + p] * input[i * 3 + 0] + clipdistance[1 * numclipdistances + p] * input[i * 3 + 1] + @@ -323,6 +352,17 @@ int PolyTriangleDrawer::clipedge(const ShadedTriVertex *verts, TriVertex *clippe clipdistance[0 * numclipdistances + p] * input[j * 3 + 0] + clipdistance[1 * numclipdistances + p] * input[j * 3 + 1] + clipdistance[2 * numclipdistances + p] * input[j * 3 + 2]; +#else + float clipdistance1 = + clipdistance[0 + p * 4] * input[i * 3 + 0] + + clipdistance[1 + p * 4] * input[i * 3 + 1] + + clipdistance[2 + p * 4] * input[i * 3 + 2]; + + float clipdistance2 = + clipdistance[0 + p * 4] * input[j * 3 + 0] + + clipdistance[1 + p * 4] * input[j * 3 + 1] + + clipdistance[2 + p * 4] * input[j * 3 + 2]; +#endif // Clip halfspace if ((clipdistance1 >= 0.0f || clipdistance2 >= 0.0f) && outputverts + 1 < max_additional_vertices) @@ -369,129 +409,6 @@ int PolyTriangleDrawer::clipedge(const ShadedTriVertex *verts, TriVertex *clippe return inputverts; } -#ifndef NO_SSE -int PolyTriangleDrawer::clipedge_sse2(const ShadedTriVertex *verts, TriVertex *clippedvert) -{ - // Clip and cull so that the following is true for all vertices: - // -v.w <= v.x <= v.w - // -v.w <= v.y <= v.w - // -v.w <= v.z <= v.w - - // halfspace clip distances - static const int numclipdistances = 7; - __m128 mx = _mm_loadu_ps(&verts[0].x); - __m128 my = _mm_loadu_ps(&verts[1].x); - __m128 mz = _mm_loadu_ps(&verts[2].x); - __m128 mw = _mm_setzero_ps(); - _MM_TRANSPOSE4_PS(mx, my, mz, mw); - __m128 clipd0 = _mm_add_ps(mx, mw); - __m128 clipd1 = _mm_sub_ps(mw, mx); - __m128 clipd2 = _mm_add_ps(my, mw); - __m128 clipd3 = _mm_sub_ps(mw, my); - __m128 clipd4 = _mm_add_ps(mz, mw); - __m128 clipd5 = _mm_sub_ps(mw, mz); - __m128 clipd6 = _mm_setr_ps(verts[0].clipDistance0, verts[1].clipDistance0, verts[2].clipDistance0, 0.0f); - __m128 mneedsclipping = _mm_cmplt_ps(clipd0, _mm_setzero_ps()); - mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd1, _mm_setzero_ps())); - mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd2, _mm_setzero_ps())); - mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd3, _mm_setzero_ps())); - mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd4, _mm_setzero_ps())); - mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd5, _mm_setzero_ps())); - mneedsclipping = _mm_or_ps(mneedsclipping, _mm_cmplt_ps(clipd6, _mm_setzero_ps())); - if (_mm_movemask_ps(mneedsclipping) == 0) - { - for (int i = 0; i < 3; i++) - { - memcpy(clippedvert + i, verts + i, sizeof(TriVertex)); - } - return 3; - } - float clipdistance[numclipdistances * 4]; - _mm_storeu_ps(clipdistance, clipd0); - _mm_storeu_ps(clipdistance + 4, clipd1); - _mm_storeu_ps(clipdistance + 8, clipd2); - _mm_storeu_ps(clipdistance + 12, clipd3); - _mm_storeu_ps(clipdistance + 16, clipd4); - _mm_storeu_ps(clipdistance + 20, clipd5); - _mm_storeu_ps(clipdistance + 24, clipd6); - - // use barycentric weights while clipping vertices - float weights[max_additional_vertices * 3 * 2]; - for (int i = 0; i < 3; i++) - { - weights[i * 3 + 0] = 0.0f; - weights[i * 3 + 1] = 0.0f; - weights[i * 3 + 2] = 0.0f; - weights[i * 3 + i] = 1.0f; - } - - // Clip against each halfspace - float *input = weights; - float *output = weights + max_additional_vertices * 3; - int inputverts = 3; - for (int p = 0; p < numclipdistances; p++) - { - // Clip each edge - int outputverts = 0; - for (int i = 0; i < inputverts; i++) - { - int j = (i + 1) % inputverts; - float clipdistance1 = - clipdistance[0 + p * 4] * input[i * 3 + 0] + - clipdistance[1 + p * 4] * input[i * 3 + 1] + - clipdistance[2 + p * 4] * input[i * 3 + 2]; - - float clipdistance2 = - clipdistance[0 + p * 4] * input[j * 3 + 0] + - clipdistance[1 + p * 4] * input[j * 3 + 1] + - clipdistance[2 + p * 4] * input[j * 3 + 2]; - - // Clip halfspace - if ((clipdistance1 >= 0.0f || clipdistance2 >= 0.0f) && outputverts + 1 < max_additional_vertices) - { - float t1 = (clipdistance1 < 0.0f) ? MAX(-clipdistance1 / (clipdistance2 - clipdistance1), 0.0f) : 0.0f; - float t2 = (clipdistance2 < 0.0f) ? MIN(1.0f + clipdistance2 / (clipdistance1 - clipdistance2), 1.0f) : 1.0f; - - // add t1 vertex - for (int k = 0; k < 3; k++) - output[outputverts * 3 + k] = input[i * 3 + k] * (1.0f - t1) + input[j * 3 + k] * t1; - outputverts++; - - if (t2 != 1.0f && t2 > t1) - { - // add t2 vertex - for (int k = 0; k < 3; k++) - output[outputverts * 3 + k] = input[i * 3 + k] * (1.0f - t2) + input[j * 3 + k] * t2; - outputverts++; - } - } - } - std::swap(input, output); - inputverts = outputverts; - if (inputverts == 0) - break; - } - - // Convert barycentric weights to actual vertices - for (int i = 0; i < inputverts; i++) - { - auto &v = clippedvert[i]; - memset(&v, 0, sizeof(TriVertex)); - for (int w = 0; w < 3; w++) - { - float weight = input[i * 3 + w]; - v.x += verts[w].x * weight; - v.y += verts[w].y * weight; - v.z += verts[w].z * weight; - v.w += verts[w].w * weight; - v.u += verts[w].u * weight; - v.v += verts[w].v * weight; - } - } - return inputverts; -} -#endif - ///////////////////////////////////////////////////////////////////////////// DrawPolyTrianglesCommand::DrawPolyTrianglesCommand(const PolyDrawArgs &args, bool mirror) diff --git a/src/polyrenderer/drawers/poly_triangle.h b/src/polyrenderer/drawers/poly_triangle.h index ca92ac735..c939149d3 100644 --- a/src/polyrenderer/drawers/poly_triangle.h +++ b/src/polyrenderer/drawers/poly_triangle.h @@ -47,12 +47,8 @@ private: static ShadedTriVertex shade_vertex(const TriMatrix &objectToClip, const float *clipPlane, const TriVertex &v); static void draw_arrays(const PolyDrawArgs &args, WorkerThreadData *thread); static void draw_shaded_triangle(const ShadedTriVertex *vertices, bool ccw, TriDrawTriangleArgs *args, WorkerThreadData *thread); - static void clip_to_viewport(TriVertex *clippedvert, int numclipvert); + static int clipedge(const ShadedTriVertex *verts, TriVertex *clippedvert); -#ifndef NO_SSE - static void clip_to_viewport_sse2(TriVertex *clippedvert, int numclipvert); - static int clipedge_sse2(const ShadedTriVertex *verts, TriVertex *clippedvert); -#endif static int viewport_x, viewport_y, viewport_width, viewport_height, dest_pitch, dest_width, dest_height; static bool dest_bgra; diff --git a/src/polyrenderer/drawers/screen_triangle.cpp b/src/polyrenderer/drawers/screen_triangle.cpp index bbfb0361f..682ede81d 100644 --- a/src/polyrenderer/drawers/screen_triangle.cpp +++ b/src/polyrenderer/drawers/screen_triangle.cpp @@ -36,20 +36,12 @@ #include "poly_triangle.h" #include "swrenderer/drawers/r_draw_rgba.h" #include "screen_triangle.h" -#include "poly_drawer32.h" -#include "poly_drawer8.h" #ifndef NO_SSE #include "poly_drawer32_sse2.h" #endif +#include "poly_drawer8.h" #include "x86.h" -namespace -{ - class SSE2CPU { public: static const int HasSSE2 = 1; }; - class GenericCPU { public: static const int HasSSE2 = 0; }; -} - -template class TriangleBlock { public: @@ -123,17 +115,9 @@ private: void ClipTest(); void StencilWrite(); void SubsectorWrite(); - -#ifndef NO_SSE - void CoverageTestSSE2(); - void StencilEqualTestSSE2(); - void SubsectorTestSSE2(); - void SubsectorWriteSSE2(); -#endif }; -template -TriangleBlock::TriangleBlock(const TriDrawTriangleArgs *args) +TriangleBlock::TriangleBlock(const TriDrawTriangleArgs *args) { const TriVertex &v1 = *args->v1; const TriVertex &v2 = *args->v2; @@ -162,32 +146,19 @@ TriangleBlock::TriangleBlock(const TriDrawTriangleArgs *args) const int X2 = (int)round(16.0f * v2.x); const int X3 = (int)round(16.0f * v3.x); #else - int Y1, Y2, Y3, X1, X2, X3; - if (CPUType::HasSSE2 == 1) - { - int tempround[4 * 3]; - __m128 m16 = _mm_set1_ps(16.0f); - __m128 mhalf = _mm_set1_ps(65536.5f); - __m128i m65536 = _mm_set1_epi32(65536); - _mm_storeu_si128((__m128i*)tempround, _mm_sub_epi32(_mm_cvtps_epi32(_mm_add_ps(_mm_mul_ps(_mm_loadu_ps((const float*)&v1), m16), mhalf)), m65536)); - _mm_storeu_si128((__m128i*)(tempround + 4), _mm_sub_epi32(_mm_cvtps_epi32(_mm_add_ps(_mm_mul_ps(_mm_loadu_ps((const float*)&v2), m16), mhalf)), m65536)); - _mm_storeu_si128((__m128i*)(tempround + 8), _mm_sub_epi32(_mm_cvtps_epi32(_mm_add_ps(_mm_mul_ps(_mm_loadu_ps((const float*)&v3), m16), mhalf)), m65536)); - X1 = tempround[0]; - X2 = tempround[4]; - X3 = tempround[8]; - Y1 = tempround[1]; - Y2 = tempround[5]; - Y3 = tempround[9]; - } - else - { - Y1 = (int)round(16.0f * v1.y); - Y2 = (int)round(16.0f * v2.y); - Y3 = (int)round(16.0f * v3.y); - X1 = (int)round(16.0f * v1.x); - X2 = (int)round(16.0f * v2.x); - X3 = (int)round(16.0f * v3.x); - } + int tempround[4 * 3]; + __m128 m16 = _mm_set1_ps(16.0f); + __m128 mhalf = _mm_set1_ps(65536.5f); + __m128i m65536 = _mm_set1_epi32(65536); + _mm_storeu_si128((__m128i*)tempround, _mm_sub_epi32(_mm_cvtps_epi32(_mm_add_ps(_mm_mul_ps(_mm_loadu_ps((const float*)&v1), m16), mhalf)), m65536)); + _mm_storeu_si128((__m128i*)(tempround + 4), _mm_sub_epi32(_mm_cvtps_epi32(_mm_add_ps(_mm_mul_ps(_mm_loadu_ps((const float*)&v2), m16), mhalf)), m65536)); + _mm_storeu_si128((__m128i*)(tempround + 8), _mm_sub_epi32(_mm_cvtps_epi32(_mm_add_ps(_mm_mul_ps(_mm_loadu_ps((const float*)&v3), m16), mhalf)), m65536)); + const int X1 = tempround[0]; + const int X2 = tempround[4]; + const int X3 = tempround[8]; + const int Y1 = tempround[1]; + const int Y2 = tempround[5]; + const int Y3 = tempround[9]; #endif // Deltas @@ -233,32 +204,28 @@ TriangleBlock::TriangleBlock(const TriDrawTriangleArgs *args) if (DY31 < 0 || (DY31 == 0 && DX31 > 0)) C3++; #ifndef NO_SSE - if (CPUType::HasSSE2 == 1) - { - mFDY12Offset = _mm_setr_epi32(0, FDY12, FDY12 * 2, FDY12 * 3); - mFDY23Offset = _mm_setr_epi32(0, FDY23, FDY23 * 2, FDY23 * 3); - mFDY31Offset = _mm_setr_epi32(0, FDY31, FDY31 * 2, FDY31 * 3); - mFDY12x4 = _mm_set1_epi32(FDY12 * 4); - mFDY23x4 = _mm_set1_epi32(FDY23 * 4); - mFDY31x4 = _mm_set1_epi32(FDY31 * 4); - mFDX12 = _mm_set1_epi32(FDX12); - mFDX23 = _mm_set1_epi32(FDX23); - mFDX31 = _mm_set1_epi32(FDX31); - mC1 = _mm_set1_epi32(C1); - mC2 = _mm_set1_epi32(C2); - mC3 = _mm_set1_epi32(C3); - mDX12 = _mm_set1_epi32(DX12); - mDY12 = _mm_set1_epi32(DY12); - mDX23 = _mm_set1_epi32(DX23); - mDY23 = _mm_set1_epi32(DY23); - mDX31 = _mm_set1_epi32(DX31); - mDY31 = _mm_set1_epi32(DY31); - } + mFDY12Offset = _mm_setr_epi32(0, FDY12, FDY12 * 2, FDY12 * 3); + mFDY23Offset = _mm_setr_epi32(0, FDY23, FDY23 * 2, FDY23 * 3); + mFDY31Offset = _mm_setr_epi32(0, FDY31, FDY31 * 2, FDY31 * 3); + mFDY12x4 = _mm_set1_epi32(FDY12 * 4); + mFDY23x4 = _mm_set1_epi32(FDY23 * 4); + mFDY31x4 = _mm_set1_epi32(FDY31 * 4); + mFDX12 = _mm_set1_epi32(FDX12); + mFDX23 = _mm_set1_epi32(FDX23); + mFDX31 = _mm_set1_epi32(FDX31); + mC1 = _mm_set1_epi32(C1); + mC2 = _mm_set1_epi32(C2); + mC3 = _mm_set1_epi32(C3); + mDX12 = _mm_set1_epi32(DX12); + mDY12 = _mm_set1_epi32(DY12); + mDX23 = _mm_set1_epi32(DX23); + mDY23 = _mm_set1_epi32(DY23); + mDX31 = _mm_set1_epi32(DX31); + mDY31 = _mm_set1_epi32(DY31); #endif } -template -void TriangleBlock::Loop(const TriDrawTriangleArgs *args, WorkerThreadData *thread) +void TriangleBlock::Loop(const TriDrawTriangleArgs *args, WorkerThreadData *thread) { // First block line for this thread int core = thread->core; @@ -270,18 +237,9 @@ void TriangleBlock::Loop(const TriDrawTriangleArgs *args, WorkerThreadD bool writeColor = args->uniforms->WriteColor(); bool writeStencil = args->uniforms->WriteStencil(); bool writeSubsector = args->uniforms->WriteSubsector(); - int bmode = (int)args->uniforms->BlendMode(); - // Find the drawer function for the given blend mode -#ifndef NO_SSE - void(*drawFunc)(int, int, uint32_t, uint32_t, const TriDrawTriangleArgs *); - if (CPUType::HasSSE2 == 1) - drawFunc = args->destBgra ? ScreenTriangle::TriDrawers32_SSE2[bmode] : ScreenTriangle::TriDrawers8[bmode]; - else - drawFunc = args->destBgra ? ScreenTriangle::TriDrawers32[bmode] : ScreenTriangle::TriDrawers8[bmode]; -#else + int bmode = (int)args->uniforms->BlendMode(); auto drawFunc = args->destBgra ? ScreenTriangle::TriDrawers32[bmode] : ScreenTriangle::TriDrawers8[bmode]; -#endif // Loop through blocks for (int y = start_miny; y < maxy; y += q * num_cores) @@ -291,11 +249,7 @@ void TriangleBlock::Loop(const TriDrawTriangleArgs *args, WorkerThreadD X = x; Y = y; - if (CPUType::HasSSE2 == 1) - CoverageTestSSE2(); - else - CoverageTest(); - + CoverageTest(); if (Mask0 == 0 && Mask1 == 0) continue; @@ -306,11 +260,7 @@ void TriangleBlock::Loop(const TriDrawTriangleArgs *args, WorkerThreadD // To do: make the stencil test use its own flag for comparison mode instead of abusing the subsector test.. if (!subsectorTest) { - if (CPUType::HasSSE2 == 1) - StencilEqualTestSSE2(); - else - StencilEqualTest(); - + StencilEqualTest(); if (Mask0 == 0 && Mask1 == 0) continue; } @@ -320,11 +270,7 @@ void TriangleBlock::Loop(const TriDrawTriangleArgs *args, WorkerThreadD if (Mask0 == 0 && Mask1 == 0) continue; - if (CPUType::HasSSE2 == 1) - SubsectorTestSSE2(); - else - SubsectorTest(); - + SubsectorTest(); if (Mask0 == 0 && Mask1 == 0) continue; } @@ -334,18 +280,14 @@ void TriangleBlock::Loop(const TriDrawTriangleArgs *args, WorkerThreadD if (writeStencil) StencilWrite(); if (writeSubsector) - { - if (CPUType::HasSSE2 == 1) - SubsectorWriteSSE2(); - else - SubsectorWrite(); - } + SubsectorWrite(); } } } -template -void TriangleBlock::SubsectorTest() +#ifdef NO_SSE + +void TriangleBlock::SubsectorTest() { int block = (X >> 3) + (Y >> 3) * subsectorPitch; uint32_t *subsector = subsectorGBuffer + block * 64; @@ -371,10 +313,9 @@ void TriangleBlock::SubsectorTest() Mask1 = Mask1 & mask1; } -#ifndef NO_SSE +#else -template -void TriangleBlock::SubsectorTestSSE2() +void TriangleBlock::SubsectorTest() { int block = (X >> 3) + (Y >> 3) * subsectorPitch; uint32_t *subsector = subsectorGBuffer + block * 64; @@ -402,8 +343,7 @@ void TriangleBlock::SubsectorTestSSE2() #endif -template -void TriangleBlock::ClipTest() +void TriangleBlock::ClipTest() { static const uint32_t clipxmask[8] = { @@ -437,8 +377,9 @@ void TriangleBlock::ClipTest() Mask1 = Mask1 & xmask & ymask1; } -template -void TriangleBlock::StencilEqualTest() +#ifdef NO_SSE + +void TriangleBlock::StencilEqualTest() { // Stencil test the whole block, if possible int block = (X >> 3) + (Y >> 3) * stencilPitch; @@ -481,10 +422,9 @@ void TriangleBlock::StencilEqualTest() } } -#ifndef NO_SSE +#else -template -void TriangleBlock::StencilEqualTestSSE2() +void TriangleBlock::StencilEqualTest() { // Stencil test the whole block, if possible int block = (X >> 3) + (Y >> 3) * stencilPitch; @@ -550,8 +490,7 @@ void TriangleBlock::StencilEqualTestSSE2() #endif -template -void TriangleBlock::StencilGreaterEqualTest() +void TriangleBlock::StencilGreaterEqualTest() { // Stencil test the whole block, if possible int block = (X >> 3) + (Y >> 3) * stencilPitch; @@ -594,8 +533,9 @@ void TriangleBlock::StencilGreaterEqualTest() } } -template -void TriangleBlock::CoverageTest() +#ifdef NO_SSE + +void TriangleBlock::CoverageTest() { // Corners of block int x0 = X << 4; @@ -692,10 +632,9 @@ void TriangleBlock::CoverageTest() } } -#ifndef NO_SSE +#else -template -void TriangleBlock::CoverageTestSSE2() +void TriangleBlock::CoverageTest() { // Corners of block int x0 = X << 4; @@ -805,8 +744,7 @@ void TriangleBlock::CoverageTestSSE2() #endif -template -void TriangleBlock::StencilWrite() +void TriangleBlock::StencilWrite() { int block = (X >> 3) + (Y >> 3) * stencilPitch; uint8_t *stencilBlock = &stencilValues[block * 64]; @@ -856,8 +794,9 @@ void TriangleBlock::StencilWrite() } } -template -void TriangleBlock::SubsectorWrite() +#ifdef NO_SSE + +void TriangleBlock::SubsectorWrite() { int block = (X >> 3) + (Y >> 3) * subsectorPitch; uint32_t *subsector = subsectorGBuffer + block * 64; @@ -890,10 +829,9 @@ void TriangleBlock::SubsectorWrite() } } -#ifndef NO_SSE +#else -template -void TriangleBlock::SubsectorWriteSSE2() +void TriangleBlock::SubsectorWrite() { int block = (X >> 3) + (Y >> 3) * subsectorPitch; uint32_t *subsector = subsectorGBuffer + block * 64; @@ -950,21 +888,8 @@ void TriangleBlock::SubsectorWriteSSE2() void ScreenTriangle::Draw(const TriDrawTriangleArgs *args, WorkerThreadData *thread) { -#ifdef NO_SSE - TriangleBlock block(args); + TriangleBlock block(args); block.Loop(args, thread); -#else - if (CPU.bSSE2) - { - TriangleBlock block(args); - block.Loop(args, thread); - } - else - { - TriangleBlock block(args); - block.Loop(args, thread); - } -#endif } void(*ScreenTriangle::TriDrawers8[])(int, int, uint32_t, uint32_t, const TriDrawTriangleArgs *) = @@ -994,6 +919,15 @@ void(*ScreenTriangle::TriDrawers8[])(int, int, uint32_t, uint32_t, const TriDraw &TriScreenDrawer8::Execute // Fuzz }; +#ifdef NO_SSE + +void(*ScreenTriangle::TriDrawers32[])(int, int, uint32_t, uint32_t, const TriDrawTriangleArgs *) = +{ + nullptr +}; + +#else + void(*ScreenTriangle::TriDrawers32[])(int, int, uint32_t, uint32_t, const TriDrawTriangleArgs *) = { &TriScreenDrawer32::Execute, // TextureOpaque @@ -1021,35 +955,6 @@ void(*ScreenTriangle::TriDrawers32[])(int, int, uint32_t, uint32_t, const TriDra &TriScreenDrawer32::Execute // Fuzz }; -#ifndef NO_SSE - -void(*ScreenTriangle::TriDrawers32_SSE2[])(int, int, uint32_t, uint32_t, const TriDrawTriangleArgs *) = -{ - &TriScreenDrawer32_SSE2::Execute, // TextureOpaque - &TriScreenDrawer32_SSE2::Execute, // TextureMasked - &TriScreenDrawer32_SSE2::Execute, // TextureAdd - &TriScreenDrawer32_SSE2::Execute, // TextureSub - &TriScreenDrawer32_SSE2::Execute, // TextureRevSub - &TriScreenDrawer32_SSE2::Execute, // TextureAddSrcColor - &TriScreenDrawer32_SSE2::Execute, // TranslatedOpaque - &TriScreenDrawer32_SSE2::Execute, // TranslatedMasked - &TriScreenDrawer32_SSE2::Execute, // TranslatedAdd - &TriScreenDrawer32_SSE2::Execute, // TranslatedSub - &TriScreenDrawer32_SSE2::Execute, // TranslatedRevSub - &TriScreenDrawer32_SSE2::Execute, // TranslatedAddSrcColor - &TriScreenDrawer32_SSE2::Execute, // Shaded - &TriScreenDrawer32_SSE2::Execute, // AddShaded - &TriScreenDrawer32_SSE2::Execute, // Stencil - &TriScreenDrawer32_SSE2::Execute, // AddStencil - &TriScreenDrawer32_SSE2::Execute, // FillOpaque - &TriScreenDrawer32_SSE2::Execute, // FillAdd - &TriScreenDrawer32_SSE2::Execute, // FillSub - &TriScreenDrawer32_SSE2::Execute, // FillRevSub - &TriScreenDrawer32_SSE2::Execute, // FillAddSrcColor - &TriScreenDrawer32_SSE2::Execute, // Skycap - &TriScreenDrawer32_SSE2::Execute // Fuzz -}; - #endif void(*ScreenTriangle::RectDrawers8[])(const void *, int, int, int, const RectDrawArgs *, WorkerThreadData *) = @@ -1079,6 +984,15 @@ void(*ScreenTriangle::RectDrawers8[])(const void *, int, int, int, const RectDra &RectScreenDrawer8::Execute // Fuzz }; +#ifdef NO_SSE + +void(*ScreenTriangle::RectDrawers32[])(const void *, int, int, int, const RectDrawArgs *, WorkerThreadData *) = +{ + nullptr +}; + +#else + void(*ScreenTriangle::RectDrawers32[])(const void *, int, int, int, const RectDrawArgs *, WorkerThreadData *) = { &RectScreenDrawer32::Execute, // TextureOpaque @@ -1106,35 +1020,6 @@ void(*ScreenTriangle::RectDrawers32[])(const void *, int, int, int, const RectDr &RectScreenDrawer32::Execute // Fuzz }; -#ifndef NO_SSE - -void(*ScreenTriangle::RectDrawers32_SSE2[])(const void *, int, int, int, const RectDrawArgs *, WorkerThreadData *) = -{ - &RectScreenDrawer32_SSE2::Execute, // TextureOpaque - &RectScreenDrawer32_SSE2::Execute, // TextureMasked - &RectScreenDrawer32_SSE2::Execute, // TextureAdd - &RectScreenDrawer32_SSE2::Execute, // TextureSub - &RectScreenDrawer32_SSE2::Execute, // TextureRevSub - &RectScreenDrawer32_SSE2::Execute, // TextureAddSrcColor - &RectScreenDrawer32_SSE2::Execute, // TranslatedOpaque - &RectScreenDrawer32_SSE2::Execute, // TranslatedMasked - &RectScreenDrawer32_SSE2::Execute, // TranslatedAdd - &RectScreenDrawer32_SSE2::Execute, // TranslatedSub - &RectScreenDrawer32_SSE2::Execute, // TranslatedRevSub - &RectScreenDrawer32_SSE2::Execute, // TranslatedAddSrcColor - &RectScreenDrawer32_SSE2::Execute, // Shaded - &RectScreenDrawer32_SSE2::Execute, // AddShaded - &RectScreenDrawer32_SSE2::Execute, // Stencil - &RectScreenDrawer32_SSE2::Execute, // AddStencil - &RectScreenDrawer32_SSE2::Execute, // FillOpaque - &RectScreenDrawer32_SSE2::Execute, // FillAdd - &RectScreenDrawer32_SSE2::Execute, // FillSub - &RectScreenDrawer32_SSE2::Execute, // FillRevSub - &RectScreenDrawer32_SSE2::Execute, // FillAddSrcColor - &RectScreenDrawer32_SSE2::Execute, // Skycap - &RectScreenDrawer32_SSE2::Execute // Fuzz -}; - #endif int ScreenTriangle::FuzzStart = 0; diff --git a/src/polyrenderer/drawers/screen_triangle.h b/src/polyrenderer/drawers/screen_triangle.h index 615a0c631..3dd4c24eb 100644 --- a/src/polyrenderer/drawers/screen_triangle.h +++ b/src/polyrenderer/drawers/screen_triangle.h @@ -131,11 +131,6 @@ public: static void(*RectDrawers8[])(const void *, int, int, int, const RectDrawArgs *, WorkerThreadData *); static void(*RectDrawers32[])(const void *, int, int, int, const RectDrawArgs *, WorkerThreadData *); -#ifndef NO_SSE - static void(*TriDrawers32_SSE2[])(int, int, uint32_t, uint32_t, const TriDrawTriangleArgs *); - static void(*RectDrawers32_SSE2[])(const void *, int, int, int, const RectDrawArgs *, WorkerThreadData *); -#endif - static int FuzzStart; }; diff --git a/src/polyrenderer/math/tri_matrix.cpp b/src/polyrenderer/math/tri_matrix.cpp index 5c88a96ec..f6bb03a5e 100644 --- a/src/polyrenderer/math/tri_matrix.cpp +++ b/src/polyrenderer/math/tri_matrix.cpp @@ -185,42 +185,21 @@ ShadedTriVertex TriMatrix::operator*(TriVertex v) const sv.y = vy; sv.z = vz; sv.w = vw; +#else + __m128 m0 = _mm_loadu_ps(matrix); + __m128 m1 = _mm_loadu_ps(matrix + 4); + __m128 m2 = _mm_loadu_ps(matrix + 8); + __m128 m3 = _mm_loadu_ps(matrix + 12); + __m128 mv = _mm_loadu_ps(&v.x); + m0 = _mm_mul_ps(m0, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(0, 0, 0, 0))); + m1 = _mm_mul_ps(m1, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(1, 1, 1, 1))); + m2 = _mm_mul_ps(m2, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(2, 2, 2, 2))); + m3 = _mm_mul_ps(m3, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(3, 3, 3, 3))); + mv = _mm_add_ps(_mm_add_ps(_mm_add_ps(m0, m1), m2), m3); + ShadedTriVertex sv; + _mm_storeu_ps(&sv.x, mv); +#endif sv.u = v.u; sv.v = v.v; return sv; -#else - if (CPU.bSSE2) - { - __m128 m0 = _mm_loadu_ps(matrix); - __m128 m1 = _mm_loadu_ps(matrix + 4); - __m128 m2 = _mm_loadu_ps(matrix + 8); - __m128 m3 = _mm_loadu_ps(matrix + 12); - __m128 mv = _mm_loadu_ps(&v.x); - m0 = _mm_mul_ps(m0, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(0, 0, 0, 0))); - m1 = _mm_mul_ps(m1, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(1, 1, 1, 1))); - m2 = _mm_mul_ps(m2, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(2, 2, 2, 2))); - m3 = _mm_mul_ps(m3, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(3, 3, 3, 3))); - mv = _mm_add_ps(_mm_add_ps(_mm_add_ps(m0, m1), m2), m3); - ShadedTriVertex sv; - _mm_storeu_ps(&sv.x, mv); - sv.u = v.u; - sv.v = v.v; - return sv; - } - else - { - float vx = matrix[0 * 4 + 0] * v.x + matrix[1 * 4 + 0] * v.y + matrix[2 * 4 + 0] * v.z + matrix[3 * 4 + 0] * v.w; - float vy = matrix[0 * 4 + 1] * v.x + matrix[1 * 4 + 1] * v.y + matrix[2 * 4 + 1] * v.z + matrix[3 * 4 + 1] * v.w; - float vz = matrix[0 * 4 + 2] * v.x + matrix[1 * 4 + 2] * v.y + matrix[2 * 4 + 2] * v.z + matrix[3 * 4 + 2] * v.w; - float vw = matrix[0 * 4 + 3] * v.x + matrix[1 * 4 + 3] * v.y + matrix[2 * 4 + 3] * v.z + matrix[3 * 4 + 3] * v.w; - ShadedTriVertex sv; - sv.x = vx; - sv.y = vy; - sv.z = vz; - sv.w = vw; - sv.u = v.u; - sv.v = v.v; - return sv; - } -#endif }