src/gallium/drivers/cell/spu/spu_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   *   Ian Romanick <idr@us.ibm.com>
  33   */
  34
  35 #include <spu_mfcio.h>
  36
  37 #include "pipe/p_util.h"
  38 #include "pipe/p_state.h"
  39 #include "pipe/p_shader_tokens.h"
  40 #include "spu_vertex_shader.h"
  41 #include "spu_exec.h"
  42 #include "draw/draw_private.h"
  43 #include "draw/draw_context.h"
  44 #include "cell/common.h"
  45 #include "spu_main.h"
  46
  47 static INLINE unsigned
  48 compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr)
  49 {
  50    unsigned mask = 0;
  51    unsigned i;
  52
  53    /* Do the hardwired planes first:
  54     */
  55    if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT;
  56    if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT;
  57    if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT;
  58    if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT;
  59    if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT;
  60    if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT;
  61
  62    /* Followed by any remaining ones:
  63     */
  64    for (i = 6; i < nr; i++) {
  65       if (dot4(clip, plane[i]) < 0)
  66          mask |= (1<<i);
  67    }
  68
  69    return mask;
  70 }
  71
  72
  73 /**
  74  * Transform vertices with the current vertex program/shader
  75  * Up to four vertices can be shaded at a time.
  76  * \param vbuffer  the input vertex data
  77  * \param elts  indexes of four input vertices
  78  * \param count  number of vertices to shade [1..4]
  79  * \param vOut  array of pointers to four output vertices
  80  */
  81 static void
  82 run_vertex_program(struct spu_vs_context *draw,
  83                    unsigned elts[4], unsigned count,
  84                    const uint64_t *vOut)
  85 {
  86    struct spu_exec_machine *machine = &draw->machine;
  87    unsigned int j;
  88
  89    ALIGN16_DECL(struct spu_exec_vector, inputs, PIPE_MAX_ATTRIBS);
  90    ALIGN16_DECL(struct spu_exec_vector, outputs, PIPE_MAX_ATTRIBS);
  91    const float *scale = draw->viewport.scale;
  92    const float *trans = draw->viewport.translate;
  93
  94    assert(count <= 4);
  95
  96    machine->Processor = TGSI_PROCESSOR_VERTEX;
  97
  98    ASSERT_ALIGN16(draw->constants);
  99    machine->Consts = (float (*)[4]) draw->constants;
 100
 101    machine->Inputs = ALIGN16_ASSIGN(inputs);
 102    machine->Outputs = ALIGN16_ASSIGN(outputs);
 103
 104    spu_vertex_fetch( draw, machine, elts, count );
 105
 106    /* run shader */
 107    spu_exec_machine_run( machine );
 108
 109
 110    /* store machine results */
 111    for (j = 0; j < count; j++) {
 112       unsigned slot;
 113       float x, y, z, w;
 114       unsigned char buffer[sizeof(struct vertex_header)
 115           + MAX_VERTEX_SIZE] ALIGN16_ATTRIB;
 116       struct vertex_header *const tmpOut =
 117           (struct vertex_header *) buffer;
 118       const unsigned vert_size = ROUNDUP16(sizeof(struct vertex_header)
 119                                            + (sizeof(float) * 4
 120                                               * draw->num_vs_outputs));
 121
 122       mfc_get(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0);
 123       wait_on_mask(1 << TAG_VERTEX_BUFFER);
 124
 125
 126       /* Handle attr[0] (position) specially:
 127        *
 128        * XXX: Computing the clipmask should be done in the vertex
 129        * program as a set of DP4 instructions appended to the
 130        * user-provided code.
 131        */
 132       x = tmpOut->clip[0] = machine->Outputs[0].xyzw[0].f[j];
 133       y = tmpOut->clip[1] = machine->Outputs[0].xyzw[1].f[j];
 134       z = tmpOut->clip[2] = machine->Outputs[0].xyzw[2].f[j];
 135       w = tmpOut->clip[3] = machine->Outputs[0].xyzw[3].f[j];
 136
 137       tmpOut->clipmask = compute_clipmask(tmpOut->clip, draw->plane,
 138                                            draw->nr_planes);
 139       tmpOut->edgeflag = 1;
 140
 141       /* divide by w */
 142       w = 1.0f / w;
 143       x *= w;
 144       y *= w;
 145       z *= w;
 146
 147       /* Viewport mapping */
 148       tmpOut->data[0][0] = x * scale[0] + trans[0];
 149       tmpOut->data[0][1] = y * scale[1] + trans[1];
 150       tmpOut->data[0][2] = z * scale[2] + trans[2];
 151       tmpOut->data[0][3] = w;
 152
 153       /* Remaining attributes are packed into sequential post-transform
 154        * vertex attrib slots.
 155        */
 156       for (slot = 1; slot < draw->num_vs_outputs; slot++) {
 157          tmpOut->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
 158          tmpOut->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
 159          tmpOut->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
 160          tmpOut->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
 161       }
 162
 163       mfc_put(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0);
 164    } /* loop over vertices */
 165 }
 166
 167
 168 unsigned char immediates[(sizeof(float) * 4 * TGSI_EXEC_NUM_IMMEDIATES) + 32]
 169     ALIGN16_ATTRIB;
 170
 171
 172 void
 173 spu_bind_vertex_shader(struct spu_vs_context *draw,
 174                        struct cell_shader_info *vs)
 175 {
 176    const unsigned immediate_addr = vs->immediates;
 177    const unsigned immediate_size =
 178        ROUNDUP16((sizeof(float) * 4 * vs->num_immediates)
 179                  + (immediate_addr & 0x0f));
 180
 181
 182    mfc_get(immediates, immediate_addr & ~0x0f, immediate_size,
 183            TAG_VERTEX_BUFFER, 0, 0);
 184
 185    draw->machine.Instructions = (struct tgsi_full_instruction *)
 186        vs->instructions;
 187    draw->machine.NumInstructions = vs->num_instructions;
 188
 189    draw->machine.Declarations = (struct tgsi_full_declaration *)
 190        vs->declarations;
 191    draw->machine.NumDeclarations = vs->num_declarations;
 192
 193    draw->num_vs_outputs = vs->num_outputs;
 194
 195    /* specify the shader to interpret/execute */
 196    spu_exec_machine_init(&draw->machine,
 197                          PIPE_MAX_SAMPLERS,
 198                          NULL /*samplers*/,
 199                          PIPE_SHADER_VERTEX);
 200
 201    wait_on_mask(1 << TAG_VERTEX_BUFFER);
 202
 203    (void) memcpy(& draw->machine.Imms, &immediates[immediate_addr & 0x0f],
 204                  sizeof(float) * 4 * vs->num_immediates);
 205 }
 206
 207
 208 void
 209 spu_execute_vertex_shader(struct spu_vs_context *draw,
 210                           const struct cell_command_vs *vs)
 211 {
 212    unsigned i;
 213
 214    (void) memcpy(draw->plane, vs->plane, sizeof(float) * 4 * vs->nr_planes);
 215    draw->nr_planes = vs->nr_planes;
 216    draw->vertex_fetch.nr_attrs = vs->nr_attrs;
 217
 218    for (i = 0; i < vs->num_elts; i += 4) {
 219       const unsigned batch_size = MIN2(vs->num_elts - i, 4);
 220
 221       run_vertex_program(draw, & vs->elts[i], batch_size, &vs->vOut[i]);
 222    }
 223 }