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 "pipe/p_inlines.h"
  27 #include "pipe/p_defines.h"
  28 #include "util/u_math.h"
  29 #include "translate/translate.h"
  30
  31 #include "svga_context.h"
  32 #include "svga_state.h"
  33 #include "svga_cmd.h"
  34 #include "svga_tgsi.h"
  35
  36 #include "svga_hw_reg.h"
  37
  38 /***********************************************************************
  39  */
  40
  41
  42 static INLINE int compare_vs_keys( const struct svga_vs_compile_key *a,
  43                                    const struct svga_vs_compile_key *b )
  44 {
  45    unsigned keysize = svga_vs_key_size( a );
  46    return memcmp( a, b, keysize );
  47 }
  48
  49
  50 static struct svga_shader_result *search_vs_key( struct svga_vertex_shader *vs,
  51                                                  const struct svga_vs_compile_key *key )
  52 {
  53    struct svga_shader_result *result = vs->base.results;
  54
  55    assert(key);
  56
  57    for ( ; result; result = result->next) {
  58       if (compare_vs_keys( key, &result->key.vkey ) == 0)
  59          return result;
  60    }
  61
  62    return NULL;
  63 }
  64
  65
  66 static enum pipe_error compile_vs( struct svga_context *svga,
  67                                    struct svga_vertex_shader *vs,
  68                                    const struct svga_vs_compile_key *key,
  69                                    struct svga_shader_result **out_result )
  70 {
  71    struct svga_shader_result *result;
  72    enum pipe_error ret = PIPE_OK;
  73
  74    result = svga_translate_vertex_program( vs, key );
  75    if (result == NULL) {
  76       ret = PIPE_ERROR_OUT_OF_MEMORY;
  77       goto fail;
  78    }
  79
  80    ret = SVGA3D_DefineShader(svga->swc,
  81                              svga->state.next_vs_id,
  82                              SVGA3D_SHADERTYPE_VS,
  83                              result->tokens,
  84                              result->nr_tokens * sizeof result->tokens[0]);
  85    if (ret)
  86       goto fail;
  87
  88    *out_result = result;
  89    result->id = svga->state.next_vs_id++;
  90    result->next = vs->base.results;
  91    vs->base.results = result;
  92    return PIPE_OK;
  93
  94 fail:
  95    if (result)
  96       svga_destroy_shader_result( result );
  97    return ret;
  98 }
  99
 100 /* SVGA_NEW_PRESCALE, SVGA_NEW_RAST, SVGA_NEW_ZERO_STRIDE
 101  */
 102 static int make_vs_key( struct svga_context *svga,
 103                         struct svga_vs_compile_key *key )
 104 {
 105    memset(key, 0, sizeof *key);
 106    key->need_prescale = svga->state.hw_clear.prescale.enabled;
 107    key->allow_psiz = svga->curr.rast->templ.point_size_per_vertex;
 108    key->zero_stride_vertex_elements =
 109       svga->curr.zero_stride_vertex_elements;
 110    key->num_zero_stride_vertex_elements =
 111       svga->curr.num_zero_stride_vertex_elements;
 112    return 0;
 113 }
 114
 115
 116
 117 static int emit_hw_vs( struct svga_context *svga,
 118                        unsigned dirty )
 119 {
 120    struct svga_shader_result *result = NULL;
 121    unsigned id = SVGA3D_INVALID_ID;
 122    int ret = 0;
 123
 124    /* SVGA_NEW_NEED_SWTNL */
 125    if (!svga->state.sw.need_swtnl) {
 126       struct svga_vertex_shader *vs = svga->curr.vs;
 127       struct svga_vs_compile_key key;
 128
 129       ret = make_vs_key( svga, &key );
 130       if (ret)
 131          return ret;
 132
 133       result = search_vs_key( vs, &key );
 134       if (!result) {
 135          ret = compile_vs( svga, vs, &key, &result );
 136          if (ret)
 137             return ret;
 138       }
 139
 140       assert (result);
 141       id = result->id;
 142    }
 143
 144    if (id != svga->state.hw_draw.shader_id[PIPE_SHADER_VERTEX]) {
 145       ret = SVGA3D_SetShader(svga->swc,
 146                              SVGA3D_SHADERTYPE_VS,
 147                              id );
 148       if (ret)
 149          return ret;
 150
 151       svga->dirty |= SVGA_NEW_VS_RESULT;
 152       svga->state.hw_draw.shader_id[PIPE_SHADER_VERTEX] = id;
 153       svga->state.hw_draw.vs = result;
 154    }
 155
 156    return 0;
 157 }
 158
 159 struct svga_tracked_state svga_hw_vs =
 160 {
 161    "vertex shader (hwtnl)",
 162    (SVGA_NEW_VS |
 163     SVGA_NEW_PRESCALE |
 164     SVGA_NEW_NEED_SWTNL |
 165     SVGA_NEW_ZERO_STRIDE),
 166    emit_hw_vs
 167 };
 168
 169
 170 /***********************************************************************
 171  */
 172 static int update_zero_stride( struct svga_context *svga,
 173                                unsigned dirty )
 174 {
 175    unsigned i;
 176
 177    svga->curr.zero_stride_vertex_elements = 0;
 178    svga->curr.num_zero_stride_vertex_elements = 0;
 179
 180    for (i = 0; i < svga->curr.num_vertex_elements; i++) {
 181       const struct pipe_vertex_element *vel = &svga->curr.ve[i];
 182       const struct pipe_vertex_buffer *vbuffer = &svga->curr.vb[
 183          vel->vertex_buffer_index];
 184       if (vbuffer->stride == 0) {
 185          unsigned const_idx =
 186             svga->curr.num_zero_stride_vertex_elements;
 187          struct translate *translate;
 188          struct translate_key key;
 189          void *mapped_buffer;
 190
 191          svga->curr.zero_stride_vertex_elements |= (1 << i);
 192          ++svga->curr.num_zero_stride_vertex_elements;
 193
 194          key.output_stride = 4 * sizeof(float);
 195          key.nr_elements = 1;
 196          key.element[0].input_format = vel->src_format;
 197          key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
 198          key.element[0].input_buffer = vel->vertex_buffer_index;
 199          key.element[0].input_offset = vel->src_offset;
 200          key.element[0].output_offset = const_idx * 4 * sizeof(float);
 201
 202          translate_key_sanitize(&key);
 203          /* translate_generic_create is technically private but
 204           * we don't want to code-generate, just want generic
 205           * translation */
 206          translate = translate_generic_create(&key);
 207
 208          assert(vel->src_offset == 0);
 209
 210          mapped_buffer = pipe_buffer_map_range(svga->pipe.screen,
 211                                                vbuffer->buffer,
 212                                                vel->src_offset,
 213                                                pf_get_size(vel->src_format),
 214                                                PIPE_BUFFER_USAGE_CPU_READ);
 215          translate->set_buffer(translate, vel->vertex_buffer_index,
 216                                mapped_buffer,
 217                                vbuffer->stride);
 218          translate->run(translate, 0, 1,
 219                         svga->curr.zero_stride_constants);
 220
 221          pipe_buffer_unmap(svga->pipe.screen,
 222                            vbuffer->buffer);
 223          translate->release(translate);
 224       }
 225    }
 226
 227    if (svga->curr.num_zero_stride_vertex_elements)
 228       svga->dirty |= SVGA_NEW_ZERO_STRIDE;
 229
 230    return 0;
 231 }
 232
 233 struct svga_tracked_state svga_hw_update_zero_stride =
 234 {
 235    "update zero_stride",
 236    ( SVGA_NEW_VELEMENT |
 237      SVGA_NEW_VBUFFER ),
 238    update_zero_stride
 239 };