src/gallium/drivers/svga/svga_state_vs.c

   1 /**********************************************************
   2  * Copyright 2008-2009 VMware, Inc.  All rights reserved.
   3  *
   4  * Permission is hereby granted, free of charge, to any person
   5  * obtaining a copy of this software and associated documentation
   6  * files (the "Software"), to deal in the Software without
   7  * restriction, including without limitation the rights to use, copy,
   8  * modify, merge, publish, distribute, sublicense, and/or sell copies
   9  * of the Software, and to permit persons to whom the Software is
  10  * furnished to do so, subject to the following conditions:
  11  *
  12  * The above copyright notice and this permission notice shall be
  13  * included in all copies or substantial portions of the Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  16  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  17  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  18  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
  19  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
  20  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  21  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  22  * SOFTWARE.
  23  *
  24  **********************************************************/
  25
  26 #include "util/u_inlines.h"
  27 #include "pipe/p_defines.h"
  28 #include "util/u_format.h"
  29 #include "util/u_math.h"
  30 #include "util/u_bitmask.h"
  31 #include "translate/translate.h"
  32
  33 #include "svga_context.h"
  34 #include "svga_state.h"
  35 #include "svga_cmd.h"
  36 #include "svga_tgsi.h"
  37
  38 #include "svga_hw_reg.h"
  39
  40 /***********************************************************************
  41  */
  42
  43
  44 static INLINE int compare_vs_keys( const struct svga_vs_compile_key *a,
  45                                    const struct svga_vs_compile_key *b )
  46 {
  47    unsigned keysize = svga_vs_key_size( a );
  48    return memcmp( a, b, keysize );
  49 }
  50
  51
  52 static struct svga_shader_result *search_vs_key( struct svga_vertex_shader *vs,
  53                                                  const struct svga_vs_compile_key *key )
  54 {
  55    struct svga_shader_result *result = vs->base.results;
  56
  57    assert(key);
  58
  59    for ( ; result; result = result->next) {
  60       if (compare_vs_keys( key, &result->key.vkey ) == 0)
  61          return result;
  62    }
  63
  64    return NULL;
  65 }
  66
  67
  68 static enum pipe_error compile_vs( struct svga_context *svga,
  69                                    struct svga_vertex_shader *vs,
  70                                    const struct svga_vs_compile_key *key,
  71                                    struct svga_shader_result **out_result )
  72 {
  73    struct svga_shader_result *result;
  74    enum pipe_error ret = PIPE_ERROR;
  75
  76    result = svga_translate_vertex_program( vs, key );
  77    if (result == NULL) {
  78       ret = PIPE_ERROR_OUT_OF_MEMORY;
  79       goto fail;
  80    }
  81
  82    result->id = util_bitmask_add(svga->vs_bm);
  83    if(result->id == UTIL_BITMASK_INVALID_INDEX) {
  84       ret = PIPE_ERROR_OUT_OF_MEMORY;
  85       goto fail;
  86    }
  87
  88    ret = SVGA3D_DefineShader(svga->swc,
  89                              result->id,
  90                              SVGA3D_SHADERTYPE_VS,
  91                              result->tokens,
  92                              result->nr_tokens * sizeof result->tokens[0]);
  93    if (ret)
  94       goto fail;
  95
  96    *out_result = result;
  97    result->next = vs->base.results;
  98    vs->base.results = result;
  99    return PIPE_OK;
 100
 101 fail:
 102    if (result) {
 103       if (result->id != UTIL_BITMASK_INVALID_INDEX)
 104          util_bitmask_clear( svga->vs_bm, result->id );
 105       svga_destroy_shader_result( result );
 106    }
 107    return ret;
 108 }
 109
 110 /* SVGA_NEW_PRESCALE, SVGA_NEW_RAST, SVGA_NEW_ZERO_STRIDE
 111  */
 112 static int make_vs_key( struct svga_context *svga,
 113                         struct svga_vs_compile_key *key )
 114 {
 115    memset(key, 0, sizeof *key);
 116    key->need_prescale = svga->state.hw_clear.prescale.enabled;
 117    key->allow_psiz = svga->curr.rast->templ.point_size_per_vertex;
 118    key->zero_stride_vertex_elements =
 119       svga->curr.zero_stride_vertex_elements;
 120    key->num_zero_stride_vertex_elements =
 121       svga->curr.num_zero_stride_vertex_elements;
 122    return 0;
 123 }
 124
 125
 126
 127 static int emit_hw_vs( struct svga_context *svga,
 128                        unsigned dirty )
 129 {
 130    struct svga_shader_result *result = NULL;
 131    unsigned id = SVGA3D_INVALID_ID;
 132    int ret = 0;
 133
 134    /* SVGA_NEW_NEED_SWTNL */
 135    if (!svga->state.sw.need_swtnl) {
 136       struct svga_vertex_shader *vs = svga->curr.vs;
 137       struct svga_vs_compile_key key;
 138
 139       ret = make_vs_key( svga, &key );
 140       if (ret)
 141          return ret;
 142
 143       result = search_vs_key( vs, &key );
 144       if (!result) {
 145          ret = compile_vs( svga, vs, &key, &result );
 146          if (ret)
 147             return ret;
 148       }
 149
 150       assert (result);
 151       id = result->id;
 152    }
 153
 154    if (result != svga->state.hw_draw.vs) {
 155       ret = SVGA3D_SetShader(svga->swc,
 156                              SVGA3D_SHADERTYPE_VS,
 157                              id );
 158       if (ret)
 159          return ret;
 160
 161       svga->dirty |= SVGA_NEW_VS_RESULT;
 162       svga->state.hw_draw.vs = result;
 163    }
 164
 165    return 0;
 166 }
 167
 168 struct svga_tracked_state svga_hw_vs =
 169 {
 170    "vertex shader (hwtnl)",
 171    (SVGA_NEW_VS |
 172     SVGA_NEW_PRESCALE |
 173     SVGA_NEW_NEED_SWTNL |
 174     SVGA_NEW_ZERO_STRIDE),
 175    emit_hw_vs
 176 };
 177
 178
 179 /***********************************************************************
 180  */
 181 static int update_zero_stride( struct svga_context *svga,
 182                                unsigned dirty )
 183 {
 184    unsigned i;
 185
 186    svga->curr.zero_stride_vertex_elements = 0;
 187    svga->curr.num_zero_stride_vertex_elements = 0;
 188
 189    for (i = 0; i < svga->curr.velems->count; i++) {
 190       const struct pipe_vertex_element *vel = &svga->curr.velems->velem[i];
 191       const struct pipe_vertex_buffer *vbuffer = &svga->curr.vb[
 192          vel->vertex_buffer_index];
 193       if (vbuffer->stride == 0) {
 194          unsigned const_idx =
 195             svga->curr.num_zero_stride_vertex_elements;
 196          struct translate *translate;
 197          struct translate_key key;
 198          void *mapped_buffer;
 199
 200          svga->curr.zero_stride_vertex_elements |= (1 << i);
 201          ++svga->curr.num_zero_stride_vertex_elements;
 202
 203          key.output_stride = 4 * sizeof(float);
 204          key.nr_elements = 1;
 205          key.element[0].type = TRANSLATE_ELEMENT_NORMAL;
 206          key.element[0].input_format = vel->src_format;
 207          key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
 208          key.element[0].input_buffer = vel->vertex_buffer_index;
 209          key.element[0].input_offset = vel->src_offset;
 210          key.element[0].instance_divisor = vel->instance_divisor;
 211          key.element[0].output_offset = const_idx * 4 * sizeof(float);
 212
 213          translate_key_sanitize(&key);
 214          /* translate_generic_create is technically private but
 215           * we don't want to code-generate, just want generic
 216           * translation */
 217          translate = translate_generic_create(&key);
 218
 219          assert(vel->src_offset == 0);
 220
 221          mapped_buffer = pipe_buffer_map_range(svga->pipe.screen,
 222                                                vbuffer->buffer,
 223                                                vel->src_offset,
 224                                                util_format_get_blocksize(vel->src_format),
 225                                                PIPE_BUFFER_USAGE_CPU_READ);
 226          translate->set_buffer(translate, vel->vertex_buffer_index,
 227                                mapped_buffer,
 228                                vbuffer->stride);
 229          translate->run(translate, 0, 1, 0,
 230                         svga->curr.zero_stride_constants);
 231
 232          pipe_buffer_unmap(svga->pipe.screen,
 233                            vbuffer->buffer);
 234          translate->release(translate);
 235       }
 236    }
 237
 238    if (svga->curr.num_zero_stride_vertex_elements)
 239       svga->dirty |= SVGA_NEW_ZERO_STRIDE;
 240
 241    return 0;
 242 }
 243
 244 struct svga_tracked_state svga_hw_update_zero_stride =
 245 {
 246    "update zero_stride",
 247    ( SVGA_NEW_VELEMENT |
 248      SVGA_NEW_VBUFFER ),
 249    update_zero_stride
 250 };