src/mesa/pipe/draw/draw_vertex_shader.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
   4  * All Rights Reserved.
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a
   7  * copy of this software and associated documentation files (the
   8  * "Software"), to deal in the Software without restriction, including
   9  * without limitation the rights to use, copy, modify, merge, publish,
  10  * distribute, sub license, and/or sell copies of the Software, and to
  11  * permit persons to whom the Software is furnished to do so, subject to
  12  * the following conditions:
  13  *
  14  * The above copyright notice and this permission notice (including the
  15  * next paragraph) shall be included in all copies or substantial portions
  16  * of the Software.
  17  *
  18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  19  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  20  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
  21  * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
  22  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
  23  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
  24  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  25  *
  26  **************************************************************************/
  27
  28  /*
  29   * Authors:
  30   *   Keith Whitwell <keith@tungstengraphics.com>
  31   *   Brian Paul
  32   */
  33
  34 #include "pipe/p_util.h"
  35 #include "draw_private.h"
  36 #include "draw_context.h"
  37 #include "draw_vertex.h"
  38
  39 #include "pipe/tgsi/exec/tgsi_core.h"
  40
  41 static INLINE unsigned
  42 compute_clipmask(float cx, float cy, float cz, float cw)
  43 {
  44    unsigned mask = 0;
  45
  46    if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT;
  47    if ( cx + cw < 0) mask |= CLIP_LEFT_BIT;
  48    if (-cy + cw < 0) mask |= CLIP_TOP_BIT;
  49    if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT;
  50    if (-cz + cw < 0) mask |= CLIP_FAR_BIT;
  51    if ( cz + cw < 0) mask |= CLIP_NEAR_BIT;
  52
  53    return mask;
  54 }
  55
  56
  57
  58
  59 #if !defined(XSTDCALL)
  60 #if defined(WIN32)
  61 #define XSTDCALL __stdcall
  62 #else
  63 #define XSTDCALL
  64 #endif
  65 #endif
  66
  67 #if defined(USE_X86_ASM) || defined(SLANG_X86)
  68 typedef void (XSTDCALL *sse2_function)(
  69    const struct tgsi_exec_vector *input,
  70    struct tgsi_exec_vector *output,
  71    float (*constant)[4],
  72    struct tgsi_exec_vector *temporary );
  73 #endif
  74
  75 /**
  76  * Transform vertices with the current vertex program/shader
  77  * Up to four vertices can be shaded at a time.
  78  * \param vbuffer  the input vertex data
  79  * \param elts  indexes of four input vertices
  80  * \param count  number of vertices to shade [1..4]
  81  * \param vOut  array of pointers to four output vertices
  82  */
  83 static void
  84 run_vertex_program(struct draw_context *draw,
  85                    unsigned elts[4], unsigned count,
  86                    struct vertex_header *vOut[])
  87 {
  88    struct tgsi_exec_machine machine;
  89    unsigned int j;
  90
  91    ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX);
  92    ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX);
  93    const float *scale = draw->viewport.scale;
  94    const float *trans = draw->viewport.translate;
  95
  96    assert(count <= 4);
  97    assert(draw->vertex_shader.outputs_written & (1 << TGSI_ATTRIB_POS));
  98
  99 #ifdef DEBUG
 100    memset( &machine, 0, sizeof( machine ) );
 101 #endif
 102
 103    /* init machine state */
 104    tgsi_exec_machine_init(&machine,
 105                           draw->vertex_shader.tokens,
 106                           PIPE_MAX_SAMPLERS,
 107                           NULL /*samplers*/ );
 108
 109    /* Consts does not require 16 byte alignment. */
 110    machine.Consts = (float (*)[4]) draw->mapped_constants;
 111
 112    machine.Inputs = ALIGN16_ASSIGN(inputs);
 113    machine.Outputs = ALIGN16_ASSIGN(outputs);
 114
 115    draw_vertex_fetch( draw, &machine, elts, count );
 116
 117    /* run shader */
 118    if( draw->vertex_shader.executable != NULL ) {
 119 #if defined(USE_X86_ASM) || defined(SLANG_X86)
 120       sse2_function func = (sse2_function) draw->vertex_shader.executable;
 121       func(
 122          machine.Inputs,
 123          machine.Outputs,
 124          machine.Consts,
 125          machine.Temps );
 126 #else
 127       assert( 0 );
 128 #endif
 129    }
 130    else {
 131       tgsi_exec_machine_run( &machine );
 132    }
 133
 134
 135    /* store machine results */
 136    for (j = 0; j < count; j++) {
 137       unsigned slot;
 138       float x, y, z, w;
 139
 140       /* Handle attr[0] (position) specially: */
 141       x = vOut[j]->clip[0] = machine.Outputs[0].xyzw[0].f[j];
 142       y = vOut[j]->clip[1] = machine.Outputs[0].xyzw[1].f[j];
 143       z = vOut[j]->clip[2] = machine.Outputs[0].xyzw[2].f[j];
 144       w = vOut[j]->clip[3] = machine.Outputs[0].xyzw[3].f[j];
 145
 146       vOut[j]->clipmask = compute_clipmask(x, y, z, w) | draw->user_clipmask;
 147       vOut[j]->edgeflag = 1;
 148
 149       /* divide by w */
 150       w = 1.0f / w;
 151       x *= w;
 152       y *= w;
 153       z *= w;
 154
 155       /* Viewport mapping */
 156       vOut[j]->data[0][0] = x * scale[0] + trans[0];
 157       vOut[j]->data[0][1] = y * scale[1] + trans[1];
 158       vOut[j]->data[0][2] = z * scale[2] + trans[2];
 159       vOut[j]->data[0][3] = w;
 160
 161       /* Remaining attributes are packed into sequential post-transform
 162        * vertex attrib slots.
 163        * Skip 0 since we just did it above.
 164        * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs.
 165        */
 166       for (slot = 1; slot < draw->vertex_info.num_attribs - 2; slot++) {
 167          vOut[j]->data[slot][0] = machine.Outputs[slot].xyzw[0].f[j];
 168          vOut[j]->data[slot][1] = machine.Outputs[slot].xyzw[1].f[j];
 169          vOut[j]->data[slot][2] = machine.Outputs[slot].xyzw[2].f[j];
 170          vOut[j]->data[slot][3] = machine.Outputs[slot].xyzw[3].f[j];
 171          /*
 172          printf("output %d: %f %f %f %f\n", slot,
 173                 vOut[j]->data[slot][0],
 174                 vOut[j]->data[slot][1],
 175                 vOut[j]->data[slot][2],
 176                 vOut[j]->data[slot][3]);
 177          */
 178       }
 179    } /* loop over vertices */
 180 }
 181
 182
 183 /**
 184  * Called by the draw module when the vertx cache needs to be flushed.
 185  * This involves running the vertex shader.
 186  */
 187 void draw_vertex_shader_queue_flush( struct draw_context *draw )
 188 {
 189    unsigned i, j;
 190
 191    /* run vertex shader on vertex cache entries, four per invokation */
 192    for (i = 0; i < draw->vs.queue_nr; i += 4) {
 193       struct vertex_header *dests[4];
 194       unsigned elts[4];
 195       int n;
 196
 197       for (j = 0; j < 4; j++) {
 198          elts[j] = draw->vs.queue[i + j].elt;
 199          dests[j] = draw->vs.queue[i + j].dest;
 200       }
 201
 202       n = MIN2(4, draw->vs.queue_nr - i);
 203       assert(n > 0);
 204       assert(n <= 4);
 205
 206       run_vertex_program(draw, elts, n, dests);
 207    }
 208
 209    draw->vs.queue_nr = 0;
 210 }
 211