Original Post
Hi! I don't have yet a grpahic card that supports ps_2_0 ... so y have laways run my programs in software instead of in hardware. Until I was programming using the fixed pipelinge, time for a simple transformation of a 256*256image tooks me 0.07seconds. As it was in fixed pipeline, I don't needed that my graphic card supports pixel shader, and i get to run it into my grpahic card taken 200 times less. But later, I did a program more complex, but not much more using pixel and vertex shader. The output is a multiple render target and the code is attach at the end. Executing in software it takes 1second!!!!!!! Any idea, how much time could take in a good graphic card ???? 200 times less? More, less....! I don't have any idea between the relation of exectution times between software and hardware using pixel shaders. Can some one gives me it ? Thanks, Jesus " coord=In.oTex+0.5/texsize;\n" " Output.original_value.xyzw= tex2D (FlotatingImageSampler, coord);\n"//FlotatingImageSampler " transformation = tex2D (ChampDeformationSampler, coord);\n" " coord = In.oTex1 + (float2)transformation/texsize;\n" " coord2[0] = coord*texsize;\n" " frac2 = frac (coord2[0]);\n" " floor2 = floor (coord2[0]);\n" " if ((coord2[0].x<0) || (coord2[0].y<0)|| (coord2[0].x>texsize.x-1) || (coord2[0].y>texsize.y-1))\n" " clip(-1);\n" " for (i=0; i<4; i++)\n" " {\n" " coord3 = (floor2 + sampleOffsets + 0.5)/texsize;\n" " values = tex2D (OriginalImageSampler, coord3);\n" //need this values (4floats) (I have made just a texture with grey levels " }\n" " Output.neightborhood_values.x=values[0];\n" " Output.neightborhood_values.y=values[1];\n" " Output.neightborhood_values.z=values[2];\n" " Output.neightborhood_values.w=values[3];\n" " Output.weights.x = (1-frac2.x)*(1-frac2.y);\n" " Output.weights.y = frac2.x*(1-frac2.y);\n" " Output.weights.z = (1-frac2.x)*frac2.y;\n" " Output.weights.w = frac2.x*frac2.y;\n" " return Output;\n"