X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Fswrast%2Fs_texcombine.c;h=1ee906effd981393895b5024a6e8cd0ed7d53d96;hb=a79ca675f3d61d22853657831d8b1c3a6b27e9fa;hp=38b5633887f4fde7d3f8e4f130d4852f7344692b;hpb=e97681c7f551a2a2a6bd5eff0f4192a870c816c0;p=mesa.git diff --git a/src/mesa/swrast/s_texcombine.c b/src/mesa/swrast/s_texcombine.c index 38b5633887f..1ee906effd9 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. @@ -18,787 +17,506 @@ * 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 PROD(A,B) ( (GLuint)(A) * ((GLuint)(B)+1) ) -#define S_PROD(A,B) ( (GLint)(A) * ((GLint)(B)+1) ) -#if CHAN_BITS == 32 -typedef GLfloat ChanTemp; +/** + * 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 -typedef GLuint ChanTemp; + 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 GLchan (*primary_rgba)[4], - CONST GLchan *texelBuffer, - GLchan (*rgba)[4] ) +texture_combine( struct gl_context *ctx, GLuint unit, + const float4_array primary_rgba, + const GLfloat *texelBuffer, + SWspan *span ) { - const struct gl_texture_unit *textureUnit = &(ctx->Texture.Unit[unit]); - const GLchan (*argRGB [4])[4]; - const GLchan (*argA [4])[4]; - const GLint RGBshift = textureUnit->_CurrentCombine->ScaleShiftRGB; - const GLuint Ashift = textureUnit->_CurrentCombine->ScaleShiftA; -#if CHAN_TYPE == GL_FLOAT - const GLchan RGBmult = (GLfloat) (1 << RGBshift); - const GLchan Amult = (GLfloat) (1 << Ashift); -#else - const GLint half = (CHAN_MAX + 1) / 2; -#endif - static const GLchan one[4] = { CHAN_MAX, CHAN_MAX, CHAN_MAX, CHAN_MAX }; - static const GLchan zero[4] = { 0, 0, 0, 0 }; - const GLuint numColorArgs = textureUnit->_CurrentCombine->_NumArgsRGB; - const GLuint numAlphaArgs = textureUnit->_CurrentCombine->_NumArgsA; - GLchan ccolor[4][MAX_WIDTH][4]; - GLuint i, j; + SWcontext *swrast = SWRAST_CONTEXT(ctx); + const struct gl_fixedfunc_texture_unit *textureUnit = + &ctx->Texture.FixedFuncUnit[unit]; + const struct gl_tex_env_combine_state *combine = textureUnit->_CurrentCombine; + 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]); + 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]); + } /* printf("modeRGB 0x%x modeA 0x%x srcRGB1 0x%x srcA1 0x%x srcRGB2 0x%x srcA2 0x%x\n", - textureUnit->_CurrentCombine->ModeRGB, - textureUnit->_CurrentCombine->ModeA, - textureUnit->_CurrentCombine->SourceRGB[0], - textureUnit->_CurrentCombine->SourceA[0], - textureUnit->_CurrentCombine->SourceRGB[1], - textureUnit->_CurrentCombine->SourceA[1]); + combine->ModeRGB, + combine->ModeA, + combine->SourceRGB[0], + combine->SourceA[0], + combine->SourceRGB[1], + combine->SourceA[1]); */ /* * Do operand setup for up to 4 operands. Loop over the terms. */ - for (j = 0; j < numColorArgs; j++) { - const GLenum srcRGB = textureUnit->_CurrentCombine->SourceRGB[j]; + for (term = 0; term < numArgsRGB; term++) { + const GLenum srcRGB = combine->SourceRGB[term]; + const GLenum operandRGB = combine->OperandRGB[term]; switch (srcRGB) { case GL_TEXTURE: - argRGB[j] = (const GLchan (*)[4]) - (texelBuffer + unit * (n * 4 * sizeof(GLchan))); + argRGB[term] = get_texel_array(swrast, unit); break; case GL_PRIMARY_COLOR: - argRGB[j] = primary_rgba; + argRGB[term] = primary_rgba; break; case GL_PREVIOUS: - argRGB[j] = (const GLchan (*)[4]) rgba; + argRGB[term] = rgba; break; case GL_CONSTANT: { - GLchan (*c)[4] = ccolor[j]; - GLchan red, green, blue, alpha; - UNCLAMPED_FLOAT_TO_CHAN(red, textureUnit->EnvColor[0]); - UNCLAMPED_FLOAT_TO_CHAN(green, textureUnit->EnvColor[1]); - UNCLAMPED_FLOAT_TO_CHAN(blue, textureUnit->EnvColor[2]); - UNCLAMPED_FLOAT_TO_CHAN(alpha, textureUnit->EnvColor[3]); + float4_array c = ccolor[term]; + GLfloat red = textureUnit->EnvColor[0]; + GLfloat green = textureUnit->EnvColor[1]; + GLfloat blue = textureUnit->EnvColor[2]; + GLfloat alpha = textureUnit->EnvColor[3]; for (i = 0; i < n; i++) { - c[i][RCOMP] = red; - c[i][GCOMP] = green; - c[i][BCOMP] = blue; - c[i][ACOMP] = alpha; + ASSIGN_4V(c[i], red, green, blue, alpha); } - argRGB[j] = (const GLchan (*)[4]) ccolor[j]; + argRGB[term] = ccolor[term]; } break; /* GL_ATI_texture_env_combine3 allows GL_ZERO & GL_ONE as sources. */ case GL_ZERO: - argRGB[j] = & zero; + { + 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] = ccolor[term]; + } break; case GL_ONE: - argRGB[j] = & one; + { + 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] = 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[j] = (const GLchan (*)[4]) - (texelBuffer + srcUnit * (n * 4 * sizeof(GLchan))); + assert(srcUnit < ctx->Const.MaxTextureUnits); + if (!ctx->Texture.Unit[srcUnit]._Current) + goto end; + argRGB[term] = get_texel_array(swrast, srcUnit); } } - if (textureUnit->_CurrentCombine->OperandRGB[j] != GL_SRC_COLOR) { - const GLchan (*src)[4] = argRGB[j]; - GLchan (*dst)[4] = ccolor[j]; + if (operandRGB != GL_SRC_COLOR) { + float4_array src = argRGB[term]; + float4_array dst = ccolor[term]; - /* point to new arg[j] storage */ - argRGB[j] = (const GLchan (*)[4]) ccolor[j]; + /* point to new arg[term] storage */ + argRGB[term] = ccolor[term]; - if (textureUnit->_CurrentCombine->OperandRGB[j] == GL_ONE_MINUS_SRC_COLOR) { + switch (operandRGB) { + case GL_ONE_MINUS_SRC_COLOR: for (i = 0; i < n; i++) { - dst[i][RCOMP] = CHAN_MAX - src[i][RCOMP]; - dst[i][GCOMP] = CHAN_MAX - src[i][GCOMP]; - dst[i][BCOMP] = CHAN_MAX - src[i][BCOMP]; + dst[i][RCOMP] = 1.0F - src[i][RCOMP]; + dst[i][GCOMP] = 1.0F - src[i][GCOMP]; + dst[i][BCOMP] = 1.0F - src[i][BCOMP]; } - } - else if (textureUnit->_CurrentCombine->OperandRGB[j] == GL_SRC_ALPHA) { + 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]; } - } - else { - ASSERT(textureUnit->_CurrentCombine->OperandRGB[j] ==GL_ONE_MINUS_SRC_ALPHA); + break; + case GL_ONE_MINUS_SRC_ALPHA: for (i = 0; i < n; i++) { - dst[i][RCOMP] = CHAN_MAX - src[i][ACOMP]; - dst[i][GCOMP] = CHAN_MAX - src[i][ACOMP]; - dst[i][BCOMP] = CHAN_MAX - src[i][ACOMP]; + dst[i][RCOMP] = + dst[i][GCOMP] = + dst[i][BCOMP] = 1.0F - src[i][ACOMP]; } + break; + default: + _mesa_problem(ctx, "Bad operandRGB"); } } } /* - * Set up the argA[i] pointers + * Set up the argA[term] pointers */ - for (j = 0; j < numAlphaArgs; j++) { - const GLenum srcA = textureUnit->_CurrentCombine->SourceA[j]; + for (term = 0; term < numArgsA; term++) { + const GLenum srcA = combine->SourceA[term]; + const GLenum operandA = combine->OperandA[term]; switch (srcA) { case GL_TEXTURE: - argA[j] = (const GLchan (*)[4]) - (texelBuffer + unit * (n * 4 * sizeof(GLchan))); + argA[term] = get_texel_array(swrast, unit); break; case GL_PRIMARY_COLOR: - argA[j] = primary_rgba; + argA[term] = primary_rgba; break; case GL_PREVIOUS: - argA[j] = (const GLchan (*)[4]) rgba; + argA[term] = rgba; break; case GL_CONSTANT: { - GLchan alpha, (*c)[4] = ccolor[j]; - UNCLAMPED_FLOAT_TO_CHAN(alpha, textureUnit->EnvColor[3]); + float4_array c = ccolor[term]; + GLfloat alpha = textureUnit->EnvColor[3]; for (i = 0; i < n; i++) c[i][ACOMP] = alpha; - argA[j] = (const GLchan (*)[4]) ccolor[j]; + argA[term] = ccolor[term]; } break; /* GL_ATI_texture_env_combine3 allows GL_ZERO & GL_ONE as sources. */ case GL_ZERO: - argA[j] = & zero; + { + float4_array c = ccolor[term]; + for (i = 0; i < n; i++) + c[i][ACOMP] = 0.0F; + argA[term] = ccolor[term]; + } break; case GL_ONE: - argA[j] = & one; + { + float4_array c = ccolor[term]; + for (i = 0; i < n; i++) + c[i][ACOMP] = 1.0F; + 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[j] = (const GLchan (*)[4]) - (texelBuffer + srcUnit * (n * 4 * sizeof(GLchan))); + assert(srcUnit < ctx->Const.MaxTextureUnits); + if (!ctx->Texture.Unit[srcUnit]._Current) + goto end; + argA[term] = get_texel_array(swrast, srcUnit); } } - if (textureUnit->_CurrentCombine->OperandA[j] == GL_ONE_MINUS_SRC_ALPHA) { - const GLchan (*src)[4] = argA[j]; - GLchan (*dst)[4] = ccolor[j]; - argA[j] = (const GLchan (*)[4]) ccolor[j]; + if (operandA == GL_ONE_MINUS_SRC_ALPHA) { + float4_array src = argA[term]; + float4_array dst = ccolor[term]; + argA[term] = ccolor[term]; for (i = 0; i < n; i++) { - dst[i][ACOMP] = CHAN_MAX - src[i][ACOMP]; + dst[i][ACOMP] = 1.0F - src[i][ACOMP]; } } } - /* - * Do the texture combine. - */ - switch (textureUnit->_CurrentCombine->ModeRGB) { + /* RGB channel combine */ + { + 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: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - if (RGBshift) { - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][RCOMP] = arg0[i][RCOMP] * RGBmult; - rgba[i][GCOMP] = arg0[i][GCOMP] * RGBmult; - rgba[i][BCOMP] = arg0[i][BCOMP] * RGBmult; -#else - GLuint r = (GLuint) arg0[i][RCOMP] << RGBshift; - GLuint g = (GLuint) arg0[i][GCOMP] << RGBshift; - GLuint b = (GLuint) arg0[i][BCOMP] << RGBshift; - rgba[i][RCOMP] = MIN2(r, CHAN_MAX); - rgba[i][GCOMP] = MIN2(g, CHAN_MAX); - rgba[i][BCOMP] = MIN2(b, CHAN_MAX); -#endif - } - } - else { - for (i = 0; i < n; i++) { - rgba[i][RCOMP] = arg0[i][RCOMP]; - rgba[i][GCOMP] = arg0[i][GCOMP]; - rgba[i][BCOMP] = arg0[i][BCOMP]; - } - } + for (i = 0; i < n; i++) { + 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: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; -#if CHAN_TYPE != GL_FLOAT - const GLint shift = CHAN_BITS - RGBshift; -#endif - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - 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; -#else - GLuint r = PROD(arg0[i][RCOMP], arg1[i][RCOMP]) >> shift; - GLuint g = PROD(arg0[i][GCOMP], arg1[i][GCOMP]) >> shift; - GLuint b = PROD(arg0[i][BCOMP], arg1[i][BCOMP]) >> shift; - rgba[i][RCOMP] = (GLchan) MIN2(r, CHAN_MAX); - rgba[i][GCOMP] = (GLchan) MIN2(g, CHAN_MAX); - rgba[i][BCOMP] = (GLchan) MIN2(b, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + 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: if (textureUnit->EnvMode == GL_COMBINE4_NV) { /* (a * b) + (c * d) */ - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argRGB[2]; - const GLchan (*arg3)[4] = (const GLchan (*)[4]) argRGB[3]; for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT 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; -#else - const GLint shift = CHAN_BITS - RGBshift; - GLint r = (PROD(arg0[i][RCOMP], arg1[i][RCOMP]) >> shift) + - (PROD(arg2[i][RCOMP], arg3[i][RCOMP]) >> shift); - GLint g = (PROD(arg0[i][GCOMP], arg1[i][GCOMP]) >> shift) + - (PROD(arg2[i][GCOMP], arg3[i][GCOMP]) >> shift); - GLint b = (PROD(arg0[i][BCOMP], arg1[i][BCOMP]) >> shift) + - (PROD(arg2[i][BCOMP], arg3[i][BCOMP]) >> shift); - rgba[i][RCOMP] = (GLchan) MIN2(r, CHAN_MAX); - rgba[i][GCOMP] = (GLchan) MIN2(g, CHAN_MAX); - rgba[i][BCOMP] = (GLchan) MIN2(b, CHAN_MAX); -#endif + arg2[i][BCOMP] * arg3[i][BCOMP]) * scaleRGB; } } else { /* 2-term addition */ - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - 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; -#else - GLint r = ((GLint) arg0[i][RCOMP] + (GLint) arg1[i][RCOMP]) << RGBshift; - GLint g = ((GLint) arg0[i][GCOMP] + (GLint) arg1[i][GCOMP]) << RGBshift; - GLint b = ((GLint) arg0[i][BCOMP] + (GLint) arg1[i][BCOMP]) << RGBshift; - rgba[i][RCOMP] = (GLchan) MIN2(r, CHAN_MAX); - rgba[i][GCOMP] = (GLchan) MIN2(g, CHAN_MAX); - rgba[i][BCOMP] = (GLchan) MIN2(b, CHAN_MAX); -#endif + 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_SIGNED: if (textureUnit->EnvMode == GL_COMBINE4_NV) { /* (a * b) + (c * d) - 0.5 */ - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argRGB[2]; - const GLchan (*arg3)[4] = (const GLchan (*)[4]) argRGB[3]; for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - 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; -#else - GLint r = (((PROD(arg0[i][RCOMP], arg1[i][RCOMP]) + - PROD(arg2[i][RCOMP], arg3[i][RCOMP])) >> CHAN_BITS) - half) - << RGBshift; - GLint g = (((PROD(arg0[i][GCOMP], arg1[i][GCOMP]) + - PROD(arg2[i][GCOMP], arg3[i][GCOMP])) >> CHAN_BITS) - half) - << RGBshift; - GLint b = (((PROD(arg0[i][BCOMP], arg1[i][BCOMP]) + - PROD(arg2[i][BCOMP], arg3[i][BCOMP])) >> CHAN_BITS) - half) - << RGBshift; - rgba[i][RCOMP] = (GLchan) CLAMP(r, 0, CHAN_MAX); - rgba[i][GCOMP] = (GLchan) CLAMP(g, 0, CHAN_MAX); - rgba[i][BCOMP] = (GLchan) CLAMP(b, 0, CHAN_MAX); -#endif + 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 { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - 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; -#else - GLint r = (GLint) arg0[i][RCOMP] + (GLint) arg1[i][RCOMP] - half; - GLint g = (GLint) arg0[i][GCOMP] + (GLint) arg1[i][GCOMP] - half; - GLint b = (GLint) arg0[i][BCOMP] + (GLint) arg1[i][BCOMP] - half; - r = (r < 0) ? 0 : r << RGBshift; - g = (g < 0) ? 0 : g << RGBshift; - b = (b < 0) ? 0 : b << RGBshift; - rgba[i][RCOMP] = (GLchan) MIN2(r, CHAN_MAX); - rgba[i][GCOMP] = (GLchan) MIN2(g, CHAN_MAX); - rgba[i][BCOMP] = (GLchan) MIN2(b, CHAN_MAX); -#endif + 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: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argRGB[2]; -#if CHAN_TYPE != GL_FLOAT - const GLint shift = CHAN_BITS - RGBshift; -#endif - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][RCOMP] = (arg0[i][RCOMP] * arg2[i][RCOMP] + - arg1[i][RCOMP] * (CHAN_MAXF - arg2[i][RCOMP])) * RGBmult; - rgba[i][GCOMP] = (arg0[i][GCOMP] * arg2[i][GCOMP] + - arg1[i][GCOMP] * (CHAN_MAXF - arg2[i][GCOMP])) * RGBmult; - rgba[i][BCOMP] = (arg0[i][BCOMP] * arg2[i][BCOMP] + - arg1[i][BCOMP] * (CHAN_MAXF - arg2[i][BCOMP])) * RGBmult; -#else - GLuint r = (PROD(arg0[i][RCOMP], arg2[i][RCOMP]) - + PROD(arg1[i][RCOMP], CHAN_MAX - arg2[i][RCOMP])) - >> shift; - GLuint g = (PROD(arg0[i][GCOMP], arg2[i][GCOMP]) - + PROD(arg1[i][GCOMP], CHAN_MAX - arg2[i][GCOMP])) - >> shift; - GLuint b = (PROD(arg0[i][BCOMP], arg2[i][BCOMP]) - + PROD(arg1[i][BCOMP], CHAN_MAX - arg2[i][BCOMP])) - >> shift; - rgba[i][RCOMP] = (GLchan) MIN2(r, CHAN_MAX); - rgba[i][GCOMP] = (GLchan) MIN2(g, CHAN_MAX); - rgba[i][BCOMP] = (GLchan) MIN2(b, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + rgba[i][RCOMP] = (arg0[i][RCOMP] * arg2[i][RCOMP] + + 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])) * scaleRGB; + rgba[i][BCOMP] = (arg0[i][BCOMP] * arg2[i][BCOMP] + + arg1[i][BCOMP] * (1.0F - arg2[i][BCOMP])) * scaleRGB; } break; case GL_SUBTRACT: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - 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; -#else - GLint r = ((GLint) arg0[i][RCOMP] - (GLint) arg1[i][RCOMP]) << RGBshift; - GLint g = ((GLint) arg0[i][GCOMP] - (GLint) arg1[i][GCOMP]) << RGBshift; - GLint b = ((GLint) arg0[i][BCOMP] - (GLint) arg1[i][BCOMP]) << RGBshift; - rgba[i][RCOMP] = (GLchan) CLAMP(r, 0, CHAN_MAX); - rgba[i][GCOMP] = (GLchan) CLAMP(g, 0, CHAN_MAX); - rgba[i][BCOMP] = (GLchan) CLAMP(b, 0, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + 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 */ - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - GLchan 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, CHAN_MAXF); -#else - GLint dot = (S_PROD((GLint)arg0[i][RCOMP] - half, - (GLint)arg1[i][RCOMP] - half) + - S_PROD((GLint)arg0[i][GCOMP] - half, - (GLint)arg1[i][GCOMP] - half) + - S_PROD((GLint)arg0[i][BCOMP] - half, - (GLint)arg1[i][BCOMP] - half)) >> 6; - dot = CLAMP(dot, 0, CHAN_MAX); -#endif - rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = (GLchan) dot; - } + /* 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)) + * 4.0F; + dot = CLAMP(dot, 0.0F, 1.0F); + 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 */ - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - GLchan 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, CHAN_MAXF); -#else - GLint dot = (S_PROD((GLint)arg0[i][RCOMP] - half, - (GLint)arg1[i][RCOMP] - half) + - S_PROD((GLint)arg0[i][GCOMP] - half, - (GLint)arg1[i][GCOMP] - half) + - S_PROD((GLint)arg0[i][BCOMP] - half, - (GLint)arg1[i][BCOMP] - half)) >> 6; - dot <<= RGBshift; - dot = CLAMP(dot, 0, CHAN_MAX); -#endif - rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = (GLchan) dot; - } + /* 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 * scaleRGB; + dot = CLAMP(dot, 0.0F, 1.0F); + rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = dot; } break; case GL_MODULATE_ADD_ATI: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argRGB[2]; -#if CHAN_TYPE != GL_FLOAT - const GLint shift = CHAN_BITS - RGBshift; -#endif - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) + arg1[i][RCOMP]) * RGBmult; - rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) + arg1[i][GCOMP]) * RGBmult; - rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) + arg1[i][BCOMP]) * RGBmult; -#else - GLuint r = (PROD(arg0[i][RCOMP], arg2[i][RCOMP]) - + ((GLuint) arg1[i][RCOMP] << CHAN_BITS)) >> shift; - GLuint g = (PROD(arg0[i][GCOMP], arg2[i][GCOMP]) - + ((GLuint) arg1[i][GCOMP] << CHAN_BITS)) >> shift; - GLuint b = (PROD(arg0[i][BCOMP], arg2[i][BCOMP]) - + ((GLuint) arg1[i][BCOMP] << CHAN_BITS)) >> shift; - rgba[i][RCOMP] = (GLchan) MIN2(r, CHAN_MAX); - rgba[i][GCOMP] = (GLchan) MIN2(g, CHAN_MAX); - rgba[i][BCOMP] = (GLchan) MIN2(b, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) + + arg1[i][RCOMP]) * scaleRGB; + rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) + + arg1[i][GCOMP]) * scaleRGB; + rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) + + arg1[i][BCOMP]) * scaleRGB; } break; case GL_MODULATE_SIGNED_ADD_ATI: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argRGB[2]; -#if CHAN_TYPE != GL_FLOAT - const GLint shift = CHAN_BITS - RGBshift; -#endif - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) + arg1[i][RCOMP] - 0.5) * RGBmult; - rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) + arg1[i][GCOMP] - 0.5) * RGBmult; - rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) + arg1[i][BCOMP] - 0.5) * RGBmult; -#else - GLint r = (S_PROD(arg0[i][RCOMP], arg2[i][RCOMP]) - + (((GLint) arg1[i][RCOMP] - half) << CHAN_BITS)) - >> shift; - GLint g = (S_PROD(arg0[i][GCOMP], arg2[i][GCOMP]) - + (((GLint) arg1[i][GCOMP] - half) << CHAN_BITS)) - >> shift; - GLint b = (S_PROD(arg0[i][BCOMP], arg2[i][BCOMP]) - + (((GLint) arg1[i][BCOMP] - half) << CHAN_BITS)) - >> shift; - rgba[i][RCOMP] = (GLchan) CLAMP(r, 0, CHAN_MAX); - rgba[i][GCOMP] = (GLchan) CLAMP(g, 0, CHAN_MAX); - rgba[i][BCOMP] = (GLchan) CLAMP(b, 0, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) + + arg1[i][RCOMP] - 0.5F) * scaleRGB; + rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) + + arg1[i][GCOMP] - 0.5F) * scaleRGB; + rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) + + arg1[i][BCOMP] - 0.5F) * scaleRGB; } break; case GL_MODULATE_SUBTRACT_ATI: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argRGB[2]; -#if CHAN_TYPE != GL_FLOAT - const GLint shift = CHAN_BITS - RGBshift; -#endif - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) - arg1[i][RCOMP]) * RGBmult; - rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) - arg1[i][GCOMP]) * RGBmult; - rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) - arg1[i][BCOMP]) * RGBmult; -#else - GLint r = (S_PROD(arg0[i][RCOMP], arg2[i][RCOMP]) - - ((GLint) arg1[i][RCOMP] << CHAN_BITS)) - >> shift; - GLint g = (S_PROD(arg0[i][GCOMP], arg2[i][GCOMP]) - - ((GLint) arg1[i][GCOMP] << CHAN_BITS)) - >> shift; - GLint b = (S_PROD(arg0[i][BCOMP], arg2[i][BCOMP]) - - ((GLint) arg1[i][BCOMP] << CHAN_BITS)) - >> shift; - rgba[i][RCOMP] = (GLchan) CLAMP(r, 0, CHAN_MAX); - rgba[i][GCOMP] = (GLchan) CLAMP(g, 0, CHAN_MAX); - rgba[i][BCOMP] = (GLchan) CLAMP(b, 0, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) - + arg1[i][RCOMP]) * scaleRGB; + rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) - + arg1[i][GCOMP]) * scaleRGB; + rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) - + arg1[i][BCOMP]) * scaleRGB; } break; default: _mesa_problem(ctx, "invalid combine mode"); + } } - switch (textureUnit->_CurrentCombine->ModeA) { + /* Alpha channel combine */ + { + 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: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - if (Ashift) { - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - GLchan a = arg0[i][ACOMP] * Amult; -#else - GLuint a = (GLuint) arg0[i][ACOMP] << Ashift; -#endif - rgba[i][ACOMP] = (GLchan) MIN2(a, CHAN_MAX); - } - } - else { - for (i = 0; i < n; i++) { - rgba[i][ACOMP] = arg0[i][ACOMP]; - } - } + for (i = 0; i < n; i++) { + rgba[i][ACOMP] = arg0[i][ACOMP] * scaleA; } break; case GL_MODULATE: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1]; -#if CHAN_TYPE != GL_FLOAT - const GLint shift = CHAN_BITS - Ashift; -#endif - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][ACOMP] = arg0[i][ACOMP] * arg1[i][ACOMP] * Amult; -#else - GLuint a = (PROD(arg0[i][ACOMP], arg1[i][ACOMP]) >> shift); - rgba[i][ACOMP] = (GLchan) MIN2(a, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + rgba[i][ACOMP] = arg0[i][ACOMP] * arg1[i][ACOMP] * scaleA; } break; case GL_ADD: if (textureUnit->EnvMode == GL_COMBINE4_NV) { /* (a * b) + (c * d) */ - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argA[2]; - const GLchan (*arg3)[4] = (const GLchan (*)[4]) argA[3]; for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT rgba[i][ACOMP] = (arg0[i][ACOMP] * arg1[i][ACOMP] + - arg2[i][ACOMP] * arg3[i][ACOMP]) * Amult; -#else - const GLint shift = CHAN_BITS - Ashift; - GLint a = (PROD(arg0[i][ACOMP], arg1[i][ACOMP]) >> shift) + - (PROD(arg2[i][ACOMP], arg3[i][ACOMP]) >> shift); - rgba[i][ACOMP] = (GLchan) MIN2(a, CHAN_MAX); -#endif + arg2[i][ACOMP] * arg3[i][ACOMP]) * scaleA; } } else { /* two-term add */ - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1]; for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP]) * Amult; -#else - GLint a = ((GLint) arg0[i][ACOMP] + arg1[i][ACOMP]) << Ashift; - rgba[i][ACOMP] = (GLchan) MIN2(a, CHAN_MAX); -#endif + rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP]) * scaleA; } } break; case GL_ADD_SIGNED: if (textureUnit->EnvMode == GL_COMBINE4_NV) { /* (a * b) + (c * d) - 0.5 */ - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argA[2]; - const GLchan (*arg3)[4] = (const GLchan (*)[4]) argA[3]; for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT rgba[i][ACOMP] = (arg0[i][ACOMP] * arg1[i][ACOMP] + arg2[i][ACOMP] * arg3[i][ACOMP] - - 0.5) * Amult; -#else - GLint a = (((PROD(arg0[i][ACOMP], arg1[i][ACOMP]) + - PROD(arg2[i][ACOMP], arg3[i][ACOMP])) >> CHAN_BITS) - half) - << Ashift; - rgba[i][ACOMP] = (GLchan) CLAMP(a, 0, CHAN_MAX); -#endif + 0.5F) * scaleA; } } else { /* a + b - 0.5 */ - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1]; for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP] - 0.5F) * Amult; -#else - GLint a = (GLint) arg0[i][ACOMP] + (GLint) arg1[i][ACOMP] -half; - a = (a < 0) ? 0 : a << Ashift; - rgba[i][ACOMP] = (GLchan) CLAMP(a, 0, CHAN_MAX); -#endif + rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP] - 0.5F) * scaleA; } } break; case GL_INTERPOLATE: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argA[2]; -#if CHAN_TYPE != GL_FLOAT - const GLint shift = CHAN_BITS - Ashift; -#endif - for (i=0; i> shift; - rgba[i][ACOMP] = (GLchan) MIN2(a, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + rgba[i][ACOMP] = (arg0[i][ACOMP] * arg2[i][ACOMP] + + arg1[i][ACOMP] * (1.0F - arg2[i][ACOMP])) + * scaleA; } break; case GL_SUBTRACT: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1]; - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][ACOMP] = (arg0[i][ACOMP] - arg1[i][ACOMP]) * Amult; -#else - GLint a = ((GLint) arg0[i][ACOMP] - (GLint) arg1[i][ACOMP]) << Ashift; - rgba[i][ACOMP] = (GLchan) CLAMP(a, 0, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + rgba[i][ACOMP] = (arg0[i][ACOMP] - arg1[i][ACOMP]) * scaleA; } break; case GL_MODULATE_ADD_ATI: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argA[2]; -#if CHAN_TYPE != GL_FLOAT - const GLint shift = CHAN_BITS - Ashift; -#endif - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP]) + arg1[i][ACOMP]) * Amult; -#else - GLint a = (PROD(arg0[i][ACOMP], arg2[i][ACOMP]) - + ((GLuint) arg1[i][ACOMP] << CHAN_BITS)) - >> shift; - rgba[i][ACOMP] = (GLchan) CLAMP(a, 0, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP]) + + arg1[i][ACOMP]) * scaleA; } break; case GL_MODULATE_SIGNED_ADD_ATI: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argA[2]; -#if CHAN_TYPE != GL_FLOAT - const GLint shift = CHAN_BITS - Ashift; -#endif - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP]) + arg1[i][ACOMP] - 0.5F) * Amult; -#else - GLint a = (S_PROD(arg0[i][ACOMP], arg2[i][ACOMP]) - + (((GLint) arg1[i][ACOMP] - half) << CHAN_BITS)) - >> shift; - rgba[i][ACOMP] = (GLchan) CLAMP(a, 0, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP]) + + arg1[i][ACOMP] - 0.5F) * scaleA; } break; case GL_MODULATE_SUBTRACT_ATI: - { - const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0]; - const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1]; - const GLchan (*arg2)[4] = (const GLchan (*)[4]) argA[2]; -#if CHAN_TYPE != GL_FLOAT - const GLint shift = CHAN_BITS - Ashift; -#endif - for (i = 0; i < n; i++) { -#if CHAN_TYPE == GL_FLOAT - rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP]) - arg1[i][ACOMP]) * Amult; -#else - GLint a = (S_PROD(arg0[i][ACOMP], arg2[i][ACOMP]) - - ((GLint) arg1[i][ACOMP] << CHAN_BITS)) - >> shift; - rgba[i][ACOMP] = (GLchan) CLAMP(a, 0, CHAN_MAX); -#endif - } + for (i = 0; i < n; i++) { + rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP]) + - arg1[i][ACOMP]) * scaleA; } break; default: _mesa_problem(ctx, "invalid combine mode"); + } } /* Fix the alpha component for GL_DOT3_RGBA_EXT/ARB combining. @@ -806,14 +524,30 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, * were written such that the GL_COMBINE_ALPHA value could be set to * GL_DOT3. */ - if (textureUnit->_CurrentCombine->ModeRGB == GL_DOT3_RGBA_EXT || - textureUnit->_CurrentCombine->ModeRGB == GL_DOT3_RGBA) { + if (combine->ModeRGB == GL_DOT3_RGBA_EXT || + combine->ModeRGB == GL_DOT3_RGBA) { for (i = 0; i < n; i++) { rgba[i][ACOMP] = rgba[i][RCOMP]; } } + + for (i = 0; i < n; i++) { + UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][RCOMP], rgba[i][RCOMP]); + UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][GCOMP], rgba[i][GCOMP]); + UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][BCOMP], rgba[i][BCOMP]); + UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][ACOMP], rgba[i][ACOMP]); + } + /* The span->array->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); } -#undef PROD /** @@ -821,17 +555,17 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n, * See GL_EXT_texture_swizzle. */ static void -swizzle_texels(GLuint swizzle, GLuint count, GLchan (*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); const GLuint swzB = GET_SWZ(swizzle, 2); const GLuint swzA = GET_SWZ(swizzle, 3); - GLchan vector[6]; + GLfloat vector[6]; GLuint i; vector[SWIZZLE_ZERO] = 0; - vector[SWIZZLE_ONE] = CHAN_MAX; + vector[SWIZZLE_ONE] = 1.0F; for (i = 0; i < count; i++) { vector[SWIZZLE_X] = texels[i][0]; @@ -847,397 +581,88 @@ swizzle_texels(GLuint swizzle, GLuint count, GLchan (*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 GLchan primary_rgba[][4], CONST GLchan texel[][4], - GLchan rgba[][4] ) +void +_swrast_texture_span( struct gl_context *ctx, SWspan *span ) { - GLint baseLevel; - GLuint i; - GLchan Rc, Gc, Bc, Ac; - GLenum format; - (void) primary_rgba; - - ASSERT(texUnit); - ASSERT(texUnit->_Current); + SWcontext *swrast = SWRAST_CONTEXT(ctx); + float4_array primary_rgba; + GLuint unit; - baseLevel = texUnit->_Current->BaseLevel; - ASSERT(texUnit->_Current->Image[0][baseLevel]); + 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 - format = texUnit->_Current->Image[0][baseLevel]->_BaseFormat; + /* 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 (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; + 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]); - UNCLAMPED_FLOAT_TO_CHAN(Gc, texUnit->EnvColor[1]); - UNCLAMPED_FLOAT_TO_CHAN(Bc, texUnit->EnvColor[2]); - UNCLAMPED_FLOAT_TO_CHAN(Ac, texUnit->EnvColor[3]); - switch (format) { - case GL_ALPHA: - for (i=0;iend * 4 * sizeof(GLfloat)); - default: - _mesa_problem(ctx, "Bad env mode in texture_apply"); - return; + 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); - GLchan 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) - MEMCPY(primary_rgba, span->array->rgba, 4 * span->end * sizeof(GLchan)); + if (swrast->_TextureCombinePrimary) { + GLuint i; + for (i = 0; i < span->end; i++) { + primary_rgba[i][RCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][RCOMP]); + primary_rgba[i][GCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][GCOMP]); + primary_rgba[i][BCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][BCOMP]); + primary_rgba[i][ACOMP] = CHAN_TO_FLOAT(span->array->rgba[i][ACOMP]); + } + } /* * 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) { - 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]; - GLchan (*texels)[4] = (GLchan (*)[4]) - (swrast->TexelBuffer + unit * (span->end * 4 * sizeof(GLchan))); + 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; @@ -1246,10 +671,11 @@ _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]; @@ -1257,21 +683,23 @@ _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) { -#if CHAN_TYPE == GL_UNSIGNED_BYTE - _mesa_lookup_rgba_ubyte(&texUnit->ColorTable, span->end, texels); -#elif CHAN_TYPE == GL_UNSIGNED_SHORT - _mesa_lookup_rgba_ubyte(&texUnit->ColorTable, span->end, texels); -#else - _mesa_lookup_rgba_float(&texUnit->ColorTable, span->end, texels); -#endif - } + swrast->TextureSample[unit]( ctx, samp, + ctx->Texture.Unit[unit]._Current, + span->end, texcoords, lambda, texels ); /* GL_EXT_texture_swizzle */ if (curObj->_Swizzle != SWIZZLE_NOOP) { @@ -1280,31 +708,14 @@ _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 GLchan (*)[4]) primary_rgba, - swrast->TexelBuffer, - span->array->rgba ); - } - else { - /* conventional texture blend */ - const GLchan (*texels)[4] = (const GLchan (*)[4]) - (swrast->TexelBuffer + unit * - (span->end * 4 * sizeof(GLchan))); - - - texture_apply( ctx, texUnit, span->end, - (CONST GLchan (*)[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); }