X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Fswrast%2Fs_texcombine.c;h=da4a013634ca39e2f6e3c3fde421883fbb28c214;hb=6749d77c690d2254b8ed3f16a653c41565ebab49;hp=27c5c15cf16bdee531fadf4f22be9d4c96b80e38;hpb=933f3b13c34c2ed9223755c0e7c7dc22f09d23e8;p=mesa.git diff --git a/src/mesa/swrast/s_texcombine.c b/src/mesa/swrast/s_texcombine.c index 27c5c15cf16..da4a013634c 100644 --- a/src/mesa/swrast/s_texcombine.c +++ b/src/mesa/swrast/s_texcombine.c @@ -1,6 +1,5 @@ /* * Mesa 3-D graphics library - * Version: 7.5 * * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * Copyright (C) 2009 VMware, Inc. All Rights Reserved. @@ -18,64 +17,105 @@ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. */ #include "main/glheader.h" #include "main/context.h" -#include "main/colormac.h" -#include "main/image.h" #include "main/imports.h" #include "main/macros.h" -#include "main/pixel.h" -#include "shader/prog_instruction.h" +#include "main/pixeltransfer.h" +#include "main/samplerobj.h" +#include "program/prog_instruction.h" #include "s_context.h" #include "s_texcombine.h" -#define MAX_COMBINER_TERMS 4 +/** + * Pointer to array of float[4] + * This type makes the code below more concise and avoids a lot of casting. + */ +typedef float (*float4_array)[4]; + + +/** + * Return array of texels for given unit. + */ +static inline float4_array +get_texel_array(SWcontext *swrast, GLuint unit) +{ +#ifdef _OPENMP + return (float4_array) (swrast->TexelBuffer + unit * SWRAST_MAX_WIDTH * 4 * omp_get_num_threads() + (SWRAST_MAX_WIDTH * 4 * omp_get_thread_num())); +#else + return (float4_array) (swrast->TexelBuffer + unit * SWRAST_MAX_WIDTH * 4); +#endif +} + /** - * Do texture application for GL_ARB/EXT_texture_env_combine. - * This function also supports GL_{EXT,ARB}_texture_env_dot3 and - * GL_ATI_texture_env_combine3. Since "classic" texture environments are - * implemented using GL_ARB_texture_env_combine-like state, this same function - * is used for classic texture environment application as well. + * Do texture application for: + * GL_EXT_texture_env_combine + * GL_ARB_texture_env_combine + * GL_EXT_texture_env_dot3 + * GL_ARB_texture_env_dot3 + * GL_ATI_texture_env_combine3 + * GL_NV_texture_env_combine4 + * conventional GL texture env modes * * \param ctx rendering context - * \param textureUnit the texture unit to apply - * \param n number of fragments to process (span width) + * \param unit the texture combiner unit * \param primary_rgba incoming fragment color array * \param texelBuffer pointer to texel colors for all texture units * - * \param rgba incoming colors, which get modified here + * \param span two fields are used in this function: + * span->end: number of fragments to process + * span->array->rgba: incoming/result fragment colors */ static void -texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, - CONST GLfloat (*primary_rgba)[4], - CONST GLfloat *texelBuffer, - GLchan (*rgbaChan)[4] ) +texture_combine( struct gl_context *ctx, GLuint unit, + const float4_array primary_rgba, + const GLfloat *texelBuffer, + SWspan *span ) { + SWcontext *swrast = SWRAST_CONTEXT(ctx); const struct gl_texture_unit *textureUnit = &(ctx->Texture.Unit[unit]); const struct gl_tex_env_combine_state *combine = textureUnit->_CurrentCombine; - const GLfloat (*argRGB [MAX_COMBINER_TERMS])[4]; - const GLfloat (*argA [MAX_COMBINER_TERMS])[4]; - const GLfloat RGBmult = (GLfloat) (1 << combine->ScaleShiftRGB); - const GLfloat Amult = (GLfloat) (1 << combine->ScaleShiftA); - const GLuint numColorArgs = combine->_NumArgsRGB; - const GLuint numAlphaArgs = combine->_NumArgsA; - GLfloat ccolor[MAX_COMBINER_TERMS][MAX_WIDTH][4]; /* temp color buffers */ - GLfloat rgba[MAX_WIDTH][4]; + float4_array argRGB[MAX_COMBINER_TERMS]; + float4_array argA[MAX_COMBINER_TERMS]; + const GLfloat scaleRGB = (GLfloat) (1 << combine->ScaleShiftRGB); + const GLfloat scaleA = (GLfloat) (1 << combine->ScaleShiftA); + const GLuint numArgsRGB = combine->_NumArgsRGB; + const GLuint numArgsA = combine->_NumArgsA; + float4_array ccolor[4], rgba; GLuint i, term; + GLuint n = span->end; + GLchan (*rgbaChan)[4] = span->array->rgba; + + /* alloc temp pixel buffers */ + rgba = malloc(4 * n * sizeof(GLfloat)); + if (!rgba) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_combine"); + return; + } - ASSERT(ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine); - ASSERT(CONST_SWRAST_CONTEXT(ctx)->_AnyTextureCombine); + for (i = 0; i < numArgsRGB || i < numArgsA; i++) { + ccolor[i] = malloc(4 * n * sizeof(GLfloat)); + if (!ccolor[i]) { + while (i) { + free(ccolor[i]); + i--; + } + _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_combine"); + free(rgba); + return; + } + } for (i = 0; i < n; i++) { rgba[i][RCOMP] = CHAN_TO_FLOAT(rgbaChan[i][RCOMP]); @@ -97,24 +137,23 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, /* * Do operand setup for up to 4 operands. Loop over the terms. */ - for (term = 0; term < numColorArgs; term++) { + for (term = 0; term < numArgsRGB; term++) { const GLenum srcRGB = combine->SourceRGB[term]; const GLenum operandRGB = combine->OperandRGB[term]; switch (srcRGB) { case GL_TEXTURE: - argRGB[term] = (const GLfloat (*)[4]) - (texelBuffer + unit * (n * 4 * sizeof(GLfloat))); + argRGB[term] = get_texel_array(swrast, unit); break; case GL_PRIMARY_COLOR: argRGB[term] = primary_rgba; break; case GL_PREVIOUS: - argRGB[term] = (const GLfloat (*)[4]) rgba; + argRGB[term] = rgba; break; case GL_CONSTANT: { - GLfloat (*c)[4] = ccolor[term]; + float4_array c = ccolor[term]; GLfloat red = textureUnit->EnvColor[0]; GLfloat green = textureUnit->EnvColor[1]; GLfloat blue = textureUnit->EnvColor[2]; @@ -122,47 +161,46 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, for (i = 0; i < n; i++) { ASSIGN_4V(c[i], red, green, blue, alpha); } - argRGB[term] = (const GLfloat (*)[4]) ccolor[term]; + argRGB[term] = ccolor[term]; } break; /* GL_ATI_texture_env_combine3 allows GL_ZERO & GL_ONE as sources. */ case GL_ZERO: { - GLfloat (*c)[4] = ccolor[term]; + float4_array c = ccolor[term]; for (i = 0; i < n; i++) { ASSIGN_4V(c[i], 0.0F, 0.0F, 0.0F, 0.0F); } - argRGB[term] = (const GLfloat (*)[4]) ccolor[term]; + argRGB[term] = ccolor[term]; } break; case GL_ONE: { - GLfloat (*c)[4] = ccolor[term]; + float4_array c = ccolor[term]; for (i = 0; i < n; i++) { ASSIGN_4V(c[i], 1.0F, 1.0F, 1.0F, 1.0F); } - argRGB[term] = (const GLfloat (*)[4]) ccolor[term]; + argRGB[term] = ccolor[term]; } break; default: /* ARB_texture_env_crossbar source */ { const GLuint srcUnit = srcRGB - GL_TEXTURE0; - ASSERT(srcUnit < ctx->Const.MaxTextureUnits); - if (!ctx->Texture.Unit[srcUnit]._ReallyEnabled) - return; - argRGB[term] = (const GLfloat (*)[4]) - (texelBuffer + srcUnit * (n * 4 * sizeof(GLfloat))); + assert(srcUnit < ctx->Const.MaxTextureUnits); + if (!ctx->Texture.Unit[srcUnit]._Current) + goto end; + argRGB[term] = get_texel_array(swrast, srcUnit); } } if (operandRGB != GL_SRC_COLOR) { - const GLfloat (*src)[4] = argRGB[term]; - GLfloat (*dst)[4] = ccolor[term]; + float4_array src = argRGB[term]; + float4_array dst = ccolor[term]; /* point to new arg[term] storage */ - argRGB[term] = (const GLfloat (*)[4]) ccolor[term]; + argRGB[term] = ccolor[term]; switch (operandRGB) { case GL_ONE_MINUS_SRC_COLOR: @@ -174,15 +212,15 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, break; case GL_SRC_ALPHA: for (i = 0; i < n; i++) { - dst[i][RCOMP] = src[i][ACOMP]; - dst[i][GCOMP] = src[i][ACOMP]; + dst[i][RCOMP] = + dst[i][GCOMP] = dst[i][BCOMP] = src[i][ACOMP]; } break; case GL_ONE_MINUS_SRC_ALPHA: for (i = 0; i < n; i++) { - dst[i][RCOMP] = 1.0F - src[i][ACOMP]; - dst[i][GCOMP] = 1.0F - src[i][ACOMP]; + dst[i][RCOMP] = + dst[i][GCOMP] = dst[i][BCOMP] = 1.0F - src[i][ACOMP]; } break; @@ -193,66 +231,64 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, } /* - * Set up the argA[i] pointers + * Set up the argA[term] pointers */ - for (term = 0; term < numAlphaArgs; term++) { + for (term = 0; term < numArgsA; term++) { const GLenum srcA = combine->SourceA[term]; const GLenum operandA = combine->OperandA[term]; switch (srcA) { case GL_TEXTURE: - argA[term] = (const GLfloat (*)[4]) - (texelBuffer + unit * (n * 4 * sizeof(GLfloat))); + argA[term] = get_texel_array(swrast, unit); break; case GL_PRIMARY_COLOR: argA[term] = primary_rgba; break; case GL_PREVIOUS: - argA[term] = (const GLfloat (*)[4]) rgba; + argA[term] = rgba; break; case GL_CONSTANT: { - GLfloat (*c)[4] = ccolor[term]; + float4_array c = ccolor[term]; GLfloat alpha = textureUnit->EnvColor[3]; for (i = 0; i < n; i++) c[i][ACOMP] = alpha; - argA[term] = (const GLfloat (*)[4]) ccolor[term]; + argA[term] = ccolor[term]; } break; /* GL_ATI_texture_env_combine3 allows GL_ZERO & GL_ONE as sources. */ case GL_ZERO: { - GLfloat (*c)[4] = ccolor[term]; + float4_array c = ccolor[term]; for (i = 0; i < n; i++) c[i][ACOMP] = 0.0F; - argA[term] = (const GLfloat (*)[4]) ccolor[term]; + argA[term] = ccolor[term]; } break; case GL_ONE: { - GLfloat (*c)[4] = ccolor[term]; + float4_array c = ccolor[term]; for (i = 0; i < n; i++) c[i][ACOMP] = 1.0F; - argA[term] = (const GLfloat (*)[4]) ccolor[term]; + argA[term] = ccolor[term]; } break; default: /* ARB_texture_env_crossbar source */ { const GLuint srcUnit = srcA - GL_TEXTURE0; - ASSERT(srcUnit < ctx->Const.MaxTextureUnits); - if (!ctx->Texture.Unit[srcUnit]._ReallyEnabled) - return; - argA[term] = (const GLfloat (*)[4]) - (texelBuffer + srcUnit * (n * 4 * sizeof(GLfloat))); + assert(srcUnit < ctx->Const.MaxTextureUnits); + if (!ctx->Texture.Unit[srcUnit]._Current) + goto end; + argA[term] = get_texel_array(swrast, srcUnit); } } if (operandA == GL_ONE_MINUS_SRC_ALPHA) { - const GLfloat (*src)[4] = argA[term]; - GLfloat (*dst)[4] = ccolor[term]; - argA[term] = (const GLfloat (*)[4]) ccolor[term]; + float4_array src = argA[term]; + float4_array dst = ccolor[term]; + argA[term] = ccolor[term]; for (i = 0; i < n; i++) { dst[i][ACOMP] = 1.0F - src[i][ACOMP]; } @@ -261,24 +297,24 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, /* RGB channel combine */ { - const GLfloat (*arg0)[4] = (const GLfloat (*)[4]) argRGB[0]; - const GLfloat (*arg1)[4] = (const GLfloat (*)[4]) argRGB[1]; - const GLfloat (*arg2)[4] = (const GLfloat (*)[4]) argRGB[2]; - const GLfloat (*arg3)[4] = (const GLfloat (*)[4]) argRGB[3]; + float4_array arg0 = argRGB[0]; + float4_array arg1 = argRGB[1]; + float4_array arg2 = argRGB[2]; + float4_array arg3 = argRGB[3]; switch (combine->ModeRGB) { case GL_REPLACE: for (i = 0; i < n; i++) { - rgba[i][RCOMP] = arg0[i][RCOMP] * RGBmult; - rgba[i][GCOMP] = arg0[i][GCOMP] * RGBmult; - rgba[i][BCOMP] = arg0[i][BCOMP] * RGBmult; + rgba[i][RCOMP] = arg0[i][RCOMP] * scaleRGB; + rgba[i][GCOMP] = arg0[i][GCOMP] * scaleRGB; + rgba[i][BCOMP] = arg0[i][BCOMP] * scaleRGB; } break; case GL_MODULATE: for (i = 0; i < n; i++) { - rgba[i][RCOMP] = arg0[i][RCOMP] * arg1[i][RCOMP] * RGBmult; - rgba[i][GCOMP] = arg0[i][GCOMP] * arg1[i][GCOMP] * RGBmult; - rgba[i][BCOMP] = arg0[i][BCOMP] * arg1[i][BCOMP] * RGBmult; + rgba[i][RCOMP] = arg0[i][RCOMP] * arg1[i][RCOMP] * scaleRGB; + rgba[i][GCOMP] = arg0[i][GCOMP] * arg1[i][GCOMP] * scaleRGB; + rgba[i][BCOMP] = arg0[i][BCOMP] * arg1[i][BCOMP] * scaleRGB; } break; case GL_ADD: @@ -286,19 +322,19 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, /* (a * b) + (c * d) */ for (i = 0; i < n; i++) { rgba[i][RCOMP] = (arg0[i][RCOMP] * arg1[i][RCOMP] + - arg2[i][RCOMP] * arg3[i][RCOMP]) * RGBmult; + arg2[i][RCOMP] * arg3[i][RCOMP]) * scaleRGB; rgba[i][GCOMP] = (arg0[i][GCOMP] * arg1[i][GCOMP] + - arg2[i][GCOMP] * arg3[i][GCOMP]) * RGBmult; + arg2[i][GCOMP] * arg3[i][GCOMP]) * scaleRGB; rgba[i][BCOMP] = (arg0[i][BCOMP] * arg1[i][BCOMP] + - arg2[i][BCOMP] * arg3[i][BCOMP]) * RGBmult; + arg2[i][BCOMP] * arg3[i][BCOMP]) * scaleRGB; } } else { /* 2-term addition */ for (i = 0; i < n; i++) { - rgba[i][RCOMP] = (arg0[i][RCOMP] + arg1[i][RCOMP]) * RGBmult; - rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP]) * RGBmult; - rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP]) * RGBmult; + rgba[i][RCOMP] = (arg0[i][RCOMP] + arg1[i][RCOMP]) * scaleRGB; + rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP]) * scaleRGB; + rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP]) * scaleRGB; } } break; @@ -306,112 +342,93 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, if (textureUnit->EnvMode == GL_COMBINE4_NV) { /* (a * b) + (c * d) - 0.5 */ for (i = 0; i < n; i++) { - rgba[i][RCOMP] = (arg0[i][RCOMP] + arg1[i][RCOMP] * - arg2[i][RCOMP] + arg3[i][RCOMP] - 0.5) * RGBmult; - rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP] * - arg2[i][GCOMP] + arg3[i][GCOMP] - 0.5) * RGBmult; - rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP] * - arg2[i][BCOMP] + arg3[i][BCOMP] - 0.5) * RGBmult; + rgba[i][RCOMP] = (arg0[i][RCOMP] * arg1[i][RCOMP] + + arg2[i][RCOMP] * arg3[i][RCOMP] - 0.5F) * scaleRGB; + rgba[i][GCOMP] = (arg0[i][GCOMP] * arg1[i][GCOMP] + + arg2[i][GCOMP] * arg3[i][GCOMP] - 0.5F) * scaleRGB; + rgba[i][BCOMP] = (arg0[i][BCOMP] * arg1[i][BCOMP] + + arg2[i][BCOMP] * arg3[i][BCOMP] - 0.5F) * scaleRGB; } } else { for (i = 0; i < n; i++) { - rgba[i][RCOMP] = (arg0[i][RCOMP] + arg1[i][RCOMP] - 0.5) * RGBmult; - rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP] - 0.5) * RGBmult; - rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP] - 0.5) * RGBmult; + rgba[i][RCOMP] = (arg0[i][RCOMP] + arg1[i][RCOMP] - 0.5F) * scaleRGB; + rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP] - 0.5F) * scaleRGB; + rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP] - 0.5F) * scaleRGB; } } break; case GL_INTERPOLATE: for (i = 0; i < n; i++) { rgba[i][RCOMP] = (arg0[i][RCOMP] * arg2[i][RCOMP] + - arg1[i][RCOMP] * (1.0F - arg2[i][RCOMP])) * RGBmult; + arg1[i][RCOMP] * (1.0F - arg2[i][RCOMP])) * scaleRGB; rgba[i][GCOMP] = (arg0[i][GCOMP] * arg2[i][GCOMP] + - arg1[i][GCOMP] * (1.0F - arg2[i][GCOMP])) * RGBmult; + arg1[i][GCOMP] * (1.0F - arg2[i][GCOMP])) * scaleRGB; rgba[i][BCOMP] = (arg0[i][BCOMP] * arg2[i][BCOMP] + - arg1[i][BCOMP] * (1.0F - arg2[i][BCOMP])) * RGBmult; + arg1[i][BCOMP] * (1.0F - arg2[i][BCOMP])) * scaleRGB; } break; case GL_SUBTRACT: for (i = 0; i < n; i++) { - rgba[i][RCOMP] = (arg0[i][RCOMP] - arg1[i][RCOMP]) * RGBmult; - rgba[i][GCOMP] = (arg0[i][GCOMP] - arg1[i][GCOMP]) * RGBmult; - rgba[i][BCOMP] = (arg0[i][BCOMP] - arg1[i][BCOMP]) * RGBmult; + rgba[i][RCOMP] = (arg0[i][RCOMP] - arg1[i][RCOMP]) * scaleRGB; + rgba[i][GCOMP] = (arg0[i][GCOMP] - arg1[i][GCOMP]) * scaleRGB; + rgba[i][BCOMP] = (arg0[i][BCOMP] - arg1[i][BCOMP]) * scaleRGB; } break; case GL_DOT3_RGB_EXT: case GL_DOT3_RGBA_EXT: /* Do not scale the result by 1 2 or 4 */ for (i = 0; i < n; i++) { - GLfloat dot = ((arg0[i][RCOMP]-0.5F) * (arg1[i][RCOMP]-0.5F) + - (arg0[i][GCOMP]-0.5F) * (arg1[i][GCOMP]-0.5F) + - (arg0[i][BCOMP]-0.5F) * (arg1[i][BCOMP]-0.5F)) + GLfloat dot = ((arg0[i][RCOMP] - 0.5F) * (arg1[i][RCOMP] - 0.5F) + + (arg0[i][GCOMP] - 0.5F) * (arg1[i][GCOMP] - 0.5F) + + (arg0[i][BCOMP] - 0.5F) * (arg1[i][BCOMP] - 0.5F)) * 4.0F; dot = CLAMP(dot, 0.0F, 1.0F); - rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = (GLfloat) dot; + rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = dot; } break; case GL_DOT3_RGB: case GL_DOT3_RGBA: /* DO scale the result by 1 2 or 4 */ for (i = 0; i < n; i++) { - GLfloat dot = ((arg0[i][RCOMP]-0.5F) * (arg1[i][RCOMP]-0.5F) + - (arg0[i][GCOMP]-0.5F) * (arg1[i][GCOMP]-0.5F) + - (arg0[i][BCOMP]-0.5F) * (arg1[i][BCOMP]-0.5F)) - * 4.0F * RGBmult; - dot = CLAMP(dot, 0.0, 1.0F); - rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = (GLfloat) dot; + GLfloat dot = ((arg0[i][RCOMP] - 0.5F) * (arg1[i][RCOMP] - 0.5F) + + (arg0[i][GCOMP] - 0.5F) * (arg1[i][GCOMP] - 0.5F) + + (arg0[i][BCOMP] - 0.5F) * (arg1[i][BCOMP] - 0.5F)) + * 4.0F * scaleRGB; + dot = CLAMP(dot, 0.0F, 1.0F); + rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = dot; } break; case GL_MODULATE_ADD_ATI: for (i = 0; i < n; i++) { rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) + - arg1[i][RCOMP]) * RGBmult; + arg1[i][RCOMP]) * scaleRGB; rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) + - arg1[i][GCOMP]) * RGBmult; + arg1[i][GCOMP]) * scaleRGB; rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) + - arg1[i][BCOMP]) * RGBmult; + arg1[i][BCOMP]) * scaleRGB; } break; case GL_MODULATE_SIGNED_ADD_ATI: for (i = 0; i < n; i++) { rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) + - arg1[i][RCOMP] - 0.5) * RGBmult; + arg1[i][RCOMP] - 0.5F) * scaleRGB; rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) + - arg1[i][GCOMP] - 0.5) * RGBmult; + arg1[i][GCOMP] - 0.5F) * scaleRGB; rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) + - arg1[i][BCOMP] - 0.5) * RGBmult; + arg1[i][BCOMP] - 0.5F) * scaleRGB; } break; case GL_MODULATE_SUBTRACT_ATI: for (i = 0; i < n; i++) { rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) - - arg1[i][RCOMP]) * RGBmult; + arg1[i][RCOMP]) * scaleRGB; rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) - - arg1[i][GCOMP]) * RGBmult; + arg1[i][GCOMP]) * scaleRGB; rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) - - arg1[i][BCOMP]) * RGBmult; + arg1[i][BCOMP]) * scaleRGB; } break; - case GL_BUMP_ENVMAP_ATI: - { - /* this produces a fixed rgba color, and the coord calc is done elsewhere */ - for (i = 0; i < n; i++) { - /* rgba result is 0,0,0,1 */ -#if CHAN_TYPE == GL_FLOAT - rgba[i][RCOMP] = 0.0; - rgba[i][GCOMP] = 0.0; - rgba[i][BCOMP] = 0.0; - rgba[i][ACOMP] = 1.0; -#else - rgba[i][RCOMP] = 0; - rgba[i][GCOMP] = 0; - rgba[i][BCOMP] = 0; - rgba[i][ACOMP] = CHAN_MAX; -#endif - } - } - return; /* no alpha processing */ default: _mesa_problem(ctx, "invalid combine mode"); } @@ -419,21 +436,20 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, /* Alpha channel combine */ { - const GLfloat (*arg0)[4] = (const GLfloat (*)[4]) argA[0]; - const GLfloat (*arg1)[4] = (const GLfloat (*)[4]) argA[1]; - const GLfloat (*arg2)[4] = (const GLfloat (*)[4]) argA[2]; - const GLfloat (*arg3)[4] = (const GLfloat (*)[4]) argA[3]; + float4_array arg0 = argA[0]; + float4_array arg1 = argA[1]; + float4_array arg2 = argA[2]; + float4_array arg3 = argA[3]; switch (combine->ModeA) { case GL_REPLACE: for (i = 0; i < n; i++) { - GLfloat a = arg0[i][ACOMP] * Amult; - rgba[i][ACOMP] = (GLfloat) MIN2(a, 1.0F); + rgba[i][ACOMP] = arg0[i][ACOMP] * scaleA; } break; case GL_MODULATE: for (i = 0; i < n; i++) { - rgba[i][ACOMP] = arg0[i][ACOMP] * arg1[i][ACOMP] * Amult; + rgba[i][ACOMP] = arg0[i][ACOMP] * arg1[i][ACOMP] * scaleA; } break; case GL_ADD: @@ -441,13 +457,13 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, /* (a * b) + (c * d) */ for (i = 0; i < n; i++) { rgba[i][ACOMP] = (arg0[i][ACOMP] * arg1[i][ACOMP] + - arg2[i][ACOMP] * arg3[i][ACOMP]) * Amult; + arg2[i][ACOMP] * arg3[i][ACOMP]) * scaleA; } } else { /* two-term add */ for (i = 0; i < n; i++) { - rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP]) * Amult; + rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP]) * scaleA; } } break; @@ -457,44 +473,44 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, for (i = 0; i < n; i++) { rgba[i][ACOMP] = (arg0[i][ACOMP] * arg1[i][ACOMP] + arg2[i][ACOMP] * arg3[i][ACOMP] - - 0.5) * Amult; + 0.5F) * scaleA; } } else { /* a + b - 0.5 */ for (i = 0; i < n; i++) { - rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP] - 0.5F) * Amult; + rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP] - 0.5F) * scaleA; } } break; case GL_INTERPOLATE: - for (i=0; iarray->rgba values are of CHAN type so set + * span->array->ChanType field accordingly. + */ + span->array->ChanType = CHAN_TYPE; +end: + for (i = 0; i < numArgsRGB || i < numArgsA; i++) { + free(ccolor[i]); + } + free(rgba); +} /** @@ -529,7 +554,7 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, * See GL_EXT_texture_swizzle. */ static void -swizzle_texels(GLuint swizzle, GLuint count, GLfloat (*texels)[4]) +swizzle_texels(GLuint swizzle, GLuint count, float4_array texels) { const GLuint swzR = GET_SWZ(swizzle, 0); const GLuint swzG = GET_SWZ(swizzle, 1); @@ -555,401 +580,60 @@ swizzle_texels(GLuint swizzle, GLuint count, GLfloat (*texels)[4]) /** - * Apply a conventional OpenGL texture env mode (REPLACE, ADD, BLEND, - * MODULATE, or DECAL) to an array of fragments. - * Input: textureUnit - pointer to texture unit to apply - * format - base internal texture format - * n - number of fragments - * primary_rgba - primary colors (may alias rgba for single texture) - * texels - array of texel colors - * InOut: rgba - incoming fragment colors modified by texel colors - * according to the texture environment mode. + * Apply texture mapping to a span of fragments. */ -static void -texture_apply( const GLcontext *ctx, - const struct gl_texture_unit *texUnit, - GLuint n, - CONST GLfloat primary_rgba[][4], CONST GLfloat texel[][4], - GLchan rgbaChan[][4] ) +void +_swrast_texture_span( struct gl_context *ctx, SWspan *span ) { - GLint baseLevel; - GLuint i; - GLfloat Rc, Gc, Bc, Ac; - GLenum format; - GLfloat rgba[MAX_WIDTH][4]; - - (void) primary_rgba; - - ASSERT(texUnit); - ASSERT(texUnit->_Current); - - baseLevel = texUnit->_Current->BaseLevel; - ASSERT(texUnit->_Current->Image[0][baseLevel]); + SWcontext *swrast = SWRAST_CONTEXT(ctx); + float4_array primary_rgba; + GLuint unit; - format = texUnit->_Current->Image[0][baseLevel]->_BaseFormat; + if (!swrast->TexelBuffer) { +#ifdef _OPENMP + const GLint maxThreads = omp_get_max_threads(); + + /* TexelBuffer memory allocation needs to be done in a critical section + * as this code runs in a parallel loop. + * When entering the section, first check if TexelBuffer has been + * initialized already by another thread while this thread was waiting. + */ + #pragma omp critical + if (!swrast->TexelBuffer) { +#else + const GLint maxThreads = 1; +#endif - if (format == GL_COLOR_INDEX || format == GL_YCBCR_MESA) { - format = GL_RGBA; /* a bit of a hack */ - } - else if (format == GL_DEPTH_COMPONENT || format == GL_DEPTH_STENCIL_EXT) { - format = texUnit->_Current->DepthMode; - } + /* TexelBuffer is also global and normally shared by all SWspan + * instances; when running with multiple threads, create one per + * thread. + */ + swrast->TexelBuffer = + malloc(ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxTextureImageUnits * maxThreads * + SWRAST_MAX_WIDTH * 4 * sizeof(GLfloat)); +#ifdef _OPENMP + } /* critical section */ +#endif - if (texUnit->EnvMode != GL_REPLACE) { - /* convert GLchan colors to GLfloat */ - for (i = 0; i < n; i++) { - rgba[i][RCOMP] = CHAN_TO_FLOAT(rgbaChan[i][RCOMP]); - rgba[i][GCOMP] = CHAN_TO_FLOAT(rgbaChan[i][GCOMP]); - rgba[i][BCOMP] = CHAN_TO_FLOAT(rgbaChan[i][BCOMP]); - rgba[i][ACOMP] = CHAN_TO_FLOAT(rgbaChan[i][ACOMP]); + if (!swrast->TexelBuffer) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_combine"); + return; } } - switch (texUnit->EnvMode) { - case GL_REPLACE: - switch (format) { - case GL_ALPHA: - for (i=0;iEnvColor[0]; - Gc = texUnit->EnvColor[1]; - Bc = texUnit->EnvColor[2]; - Ac = texUnit->EnvColor[3]; - switch (format) { - case GL_ALPHA: - for (i=0;iend * 4 * sizeof(GLfloat)); - /* convert GLfloat colors to GLchan */ - for (i = 0; i < n; i++) { - CLAMPED_FLOAT_TO_CHAN(rgbaChan[i][RCOMP], rgba[i][RCOMP]); - CLAMPED_FLOAT_TO_CHAN(rgbaChan[i][GCOMP], rgba[i][GCOMP]); - CLAMPED_FLOAT_TO_CHAN(rgbaChan[i][BCOMP], rgba[i][BCOMP]); - CLAMPED_FLOAT_TO_CHAN(rgbaChan[i][ACOMP], rgba[i][ACOMP]); + if (!primary_rgba) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_span"); + return; } -} - - -/** - * Apply texture mapping to a span of fragments. - */ -void -_swrast_texture_span( GLcontext *ctx, SWspan *span ) -{ - SWcontext *swrast = SWRAST_CONTEXT(ctx); - GLfloat primary_rgba[MAX_WIDTH][4]; - GLuint unit; - - ASSERT(span->end < MAX_WIDTH); + assert(span->end <= SWRAST_MAX_WIDTH); /* * Save copy of the incoming fragment colors (the GL_PRIMARY_COLOR) */ - if (swrast->_AnyTextureCombine) { + if (swrast->_TextureCombinePrimary) { GLuint i; for (i = 0; i < span->end; i++) { primary_rgba[i][RCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][RCOMP]); @@ -959,100 +643,25 @@ _swrast_texture_span( GLcontext *ctx, SWspan *span ) } } - /* First must sample all bump maps */ - for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) { - if (ctx->Texture.Unit[unit]._ReallyEnabled && - ctx->Texture.Unit[unit]._CurrentCombine->ModeRGB == GL_BUMP_ENVMAP_ATI) { - const GLfloat (*texcoords)[4] - = (const GLfloat (*)[4]) - span->array->attribs[FRAG_ATTRIB_TEX0 + unit]; - GLfloat (*targetcoords)[4] - = (GLfloat (*)[4]) - span->array->attribs[FRAG_ATTRIB_TEX0 + - ctx->Texture.Unit[unit].BumpTarget - GL_TEXTURE0]; - - const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - const struct gl_texture_object *curObj = texUnit->_Current; - GLfloat *lambda = span->array->lambda[unit]; - GLchan (*texels)[4] = (GLchan (*)[4]) - (swrast->TexelBuffer + unit * (span->end * 4 * sizeof(GLchan))); - GLuint i; - GLfloat rotMatrix00 = ctx->Texture.Unit[unit].RotMatrix[0]; - GLfloat rotMatrix01 = ctx->Texture.Unit[unit].RotMatrix[1]; - GLfloat rotMatrix10 = ctx->Texture.Unit[unit].RotMatrix[2]; - GLfloat rotMatrix11 = ctx->Texture.Unit[unit].RotMatrix[3]; - - /* adjust texture lod (lambda) */ - if (span->arrayMask & SPAN_LAMBDA) { - if (texUnit->LodBias + curObj->LodBias != 0.0F) { - /* apply LOD bias, but don't clamp yet */ - const GLfloat bias = CLAMP(texUnit->LodBias + curObj->LodBias, - -ctx->Const.MaxTextureLodBias, - ctx->Const.MaxTextureLodBias); - GLuint i; - for (i = 0; i < span->end; i++) { - lambda[i] += bias; - } - } - - if (curObj->MinLod != -1000.0 || curObj->MaxLod != 1000.0) { - /* apply LOD clamping to lambda */ - const GLfloat min = curObj->MinLod; - const GLfloat max = curObj->MaxLod; - GLuint i; - for (i = 0; i < span->end; i++) { - GLfloat l = lambda[i]; - lambda[i] = CLAMP(l, min, max); - } - } - } - - /* Sample the texture (span->end = number of fragments) */ - swrast->TextureSample[unit]( ctx, texUnit->_Current, span->end, - texcoords, lambda, texels ); - - /* manipulate the span values of the bump target - not sure this can work correctly even ignoring - the problem that channel is unsigned */ - for (i = 0; i < span->end; i++) { -#if CHAN_TYPE == GL_FLOAT - targetcoords[i][0] += (texels[i][0] * rotMatrix00 + texels[i][1] * - rotMatrix01) / targetcoords[i][3]; - targetcoords[i][1] += (texels[i][0] * rotMatrix10 + texels[i][1] * - rotMatrix11) / targetcoords[i][3]; -#else - targetcoords[i][0] += (CHAN_TO_FLOAT(texels[i][1]) * rotMatrix00 + - CHAN_TO_FLOAT(texels[i][1]) * rotMatrix01) / - targetcoords[i][3]; - targetcoords[i][1] += (CHAN_TO_FLOAT(texels[i][0]) * rotMatrix10 + - CHAN_TO_FLOAT(texels[i][1]) * rotMatrix11) / - targetcoords[i][3]; -#endif - } - } - } - /* * Must do all texture sampling before combining in order to - * accomodate GL_ARB_texture_env_crossbar. + * accommodate GL_ARB_texture_env_crossbar. */ for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) { - if (ctx->Texture.Unit[unit]._ReallyEnabled && - ctx->Texture.Unit[unit]._CurrentCombine->ModeRGB != GL_BUMP_ENVMAP_ATI) { - const GLfloat (*texcoords)[4] - = (const GLfloat (*)[4]) - span->array->attribs[FRAG_ATTRIB_TEX0 + unit]; - const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; + const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; + if (texUnit->_Current) { + const GLfloat (*texcoords)[4] = (const GLfloat (*)[4]) + span->array->attribs[VARYING_SLOT_TEX0 + unit]; const struct gl_texture_object *curObj = texUnit->_Current; + const struct gl_sampler_object *samp = _mesa_get_samplerobj(ctx, unit); GLfloat *lambda = span->array->lambda[unit]; - GLfloat (*texels)[4] = (GLfloat (*)[4]) - (swrast->TexelBuffer + unit * (span->end * 4 * sizeof(GLfloat))); + float4_array texels = get_texel_array(swrast, unit); /* adjust texture lod (lambda) */ if (span->arrayMask & SPAN_LAMBDA) { - if (texUnit->LodBias + curObj->LodBias != 0.0F) { + if (texUnit->LodBias + samp->LodBias != 0.0F) { /* apply LOD bias, but don't clamp yet */ - const GLfloat bias = CLAMP(texUnit->LodBias + curObj->LodBias, + const GLfloat bias = CLAMP(texUnit->LodBias + samp->LodBias, -ctx->Const.MaxTextureLodBias, ctx->Const.MaxTextureLodBias); GLuint i; @@ -1061,10 +670,11 @@ _swrast_texture_span( GLcontext *ctx, SWspan *span ) } } - if (curObj->MinLod != -1000.0 || curObj->MaxLod != 1000.0) { + if (samp->MinLod != -1000.0F || + samp->MaxLod != 1000.0F) { /* apply LOD clamping to lambda */ - const GLfloat min = curObj->MinLod; - const GLfloat max = curObj->MaxLod; + const GLfloat min = samp->MinLod; + const GLfloat max = samp->MaxLod; GLuint i; for (i = 0; i < span->end; i++) { GLfloat l = lambda[i]; @@ -1072,15 +682,23 @@ _swrast_texture_span( GLcontext *ctx, SWspan *span ) } } } + else if (samp->MaxAnisotropy > 1.0F && + samp->MinFilter == GL_LINEAR_MIPMAP_LINEAR) { + /* sample_lambda_2d_aniso is beeing used as texture_sample_func, + * it requires the current SWspan *span as an additional parameter. + * In order to keep the same function signature, the unused lambda + * parameter will be modified to actually contain the SWspan pointer. + * This is a Hack. To make it right, the texture_sample_func + * signature and all implementing functions need to be modified. + */ + /* "hide" SWspan struct; cast to (GLfloat *) to suppress warning */ + lambda = (GLfloat *)span; + } /* Sample the texture (span->end = number of fragments) */ - swrast->TextureSample[unit]( ctx, texUnit->_Current, span->end, - texcoords, lambda, texels ); - - /* GL_SGI_texture_color_table */ - if (texUnit->ColorTableEnabled) { - _mesa_lookup_rgba_float(&texUnit->ColorTable, span->end, texels); - } + swrast->TextureSample[unit]( ctx, samp, + ctx->Texture.Unit[unit]._Current, + span->end, texcoords, lambda, texels ); /* GL_EXT_texture_swizzle */ if (curObj->_Swizzle != SWIZZLE_NOOP) { @@ -1089,31 +707,14 @@ _swrast_texture_span( GLcontext *ctx, SWspan *span ) } } - /* * OK, now apply the texture (aka texture combine/blend). * We modify the span->color.rgba values. */ for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) { - if (ctx->Texture.Unit[unit]._ReallyEnabled) { - const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - if (texUnit->_CurrentCombine != &texUnit->_EnvMode ) { - texture_combine( ctx, unit, span->end, - (CONST GLfloat (*)[4]) primary_rgba, - swrast->TexelBuffer, - span->array->rgba ); - } - else { - /* conventional texture blend */ - const GLfloat (*texels)[4] = (const GLfloat (*)[4]) - (swrast->TexelBuffer + unit * - (span->end * 4 * sizeof(GLfloat))); - - - texture_apply( ctx, texUnit, span->end, - (CONST GLfloat (*)[4]) primary_rgba, texels, - span->array->rgba ); - } - } + if (ctx->Texture.Unit[unit]._Current) + texture_combine(ctx, unit, primary_rgba, swrast->TexelBuffer, span); } + + free(primary_rgba); }