src/gallium/drivers/svga/svga_state_fs.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 "util/u_bitmask.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 static INLINE int compare_fs_keys( const struct svga_fs_compile_key *a,
  41                                    const struct svga_fs_compile_key *b )
  42 {
  43    unsigned keysize_a = svga_fs_key_size( a );
  44    unsigned keysize_b = svga_fs_key_size( b );
  45
  46    if (keysize_a != keysize_b) {
  47       return (int)(keysize_a - keysize_b);
  48    }
  49    return memcmp( a, b, keysize_a );
  50 }
  51
  52
  53 static struct svga_shader_result *search_fs_key( struct svga_fragment_shader *fs,
  54                                                  const struct svga_fs_compile_key *key )
  55 {
  56    struct svga_shader_result *result = fs->base.results;
  57
  58    assert(key);
  59
  60    for ( ; result; result = result->next) {
  61       if (compare_fs_keys( key, &result->key.fkey ) == 0)
  62          return result;
  63    }
  64
  65    return NULL;
  66 }
  67
  68
  69 static enum pipe_error compile_fs( struct svga_context *svga,
  70                                    struct svga_fragment_shader *fs,
  71                                    const struct svga_fs_compile_key *key,
  72                                    struct svga_shader_result **out_result )
  73 {
  74    struct svga_shader_result *result;
  75    enum pipe_error ret = PIPE_ERROR;
  76
  77    result = svga_translate_fragment_program( fs, key );
  78    if (result == NULL) {
  79       ret = PIPE_ERROR_OUT_OF_MEMORY;
  80       goto fail;
  81    }
  82
  83    result->id = util_bitmask_add(svga->fs_bm);
  84    if(result->id == UTIL_BITMASK_INVALID_INDEX) {
  85       ret = PIPE_ERROR_OUT_OF_MEMORY;
  86       goto fail;
  87    }
  88
  89    ret = SVGA3D_DefineShader(svga->swc,
  90                              result->id,
  91                              SVGA3D_SHADERTYPE_PS,
  92                              result->tokens,
  93                              result->nr_tokens * sizeof result->tokens[0]);
  94    if (ret)
  95       goto fail;
  96
  97    *out_result = result;
  98    result->next = fs->base.results;
  99    fs->base.results = result;
 100    return PIPE_OK;
 101
 102 fail:
 103    if (result) {
 104       if (result->id != UTIL_BITMASK_INVALID_INDEX)
 105          util_bitmask_clear( svga->fs_bm, result->id );
 106       svga_destroy_shader_result( result );
 107    }
 108    return ret;
 109 }
 110
 111
 112 /* SVGA_NEW_TEXTURE_BINDING
 113  * SVGA_NEW_RAST
 114  * SVGA_NEW_NEED_SWTNL
 115  * SVGA_NEW_SAMPLER
 116  */
 117 static int make_fs_key( const struct svga_context *svga,
 118                         struct svga_fs_compile_key *key )
 119 {
 120    int i;
 121    int idx = 0;
 122
 123    memset(key, 0, sizeof *key);
 124
 125    /* Only need fragment shader fixup for twoside lighting if doing
 126     * hwtnl.  Otherwise the draw module does the whole job for us.
 127     *
 128     * SVGA_NEW_SWTNL
 129     */
 130    if (!svga->state.sw.need_swtnl) {
 131       /* SVGA_NEW_RAST
 132        */
 133       key->light_twoside = svga->curr.rast->templ.light_twoside;
 134       key->front_cw = (svga->curr.rast->templ.front_winding ==
 135                        PIPE_WINDING_CW);
 136    }
 137
 138    /* The blend workaround for simulating logicop xor behaviour
 139     * requires that the incoming fragment color be white.  This change
 140     * achieves that by creating a varient of the current fragment
 141     * shader that overrides all output colors with 1,1,1,1
 142     *
 143     * This will work for most shaders, including those containing
 144     * TEXKIL and/or depth-write.  However, it will break on the
 145     * combination of xor-logicop plus alphatest.
 146     *
 147     * Ultimately, we could implement alphatest in the shader using
 148     * texkil prior to overriding the outgoing fragment color.
 149     *
 150     * SVGA_NEW_BLEND
 151     */
 152    if (svga->curr.blend->need_white_fragments) {
 153       key->white_fragments = 1;
 154    }
 155
 156    /* XXX: want to limit this to the textures that the shader actually
 157     * refers to.
 158     *
 159     * SVGA_NEW_TEXTURE_BINDING | SVGA_NEW_SAMPLER
 160     */
 161    for (i = 0; i < svga->curr.num_textures; i++) {
 162       if (svga->curr.texture[i]) {
 163          assert(svga->curr.sampler[i]);
 164          key->tex[i].texture_target = svga->curr.texture[i]->target;
 165          if (!svga->curr.sampler[i]->normalized_coords) {
 166             key->tex[i].width_height_idx = idx++;
 167             key->tex[i].unnormalized = TRUE;
 168             ++key->num_unnormalized_coords;
 169          }
 170       }
 171    }
 172    key->num_textures = svga->curr.num_textures;
 173
 174    idx = 0;
 175    for (i = 0; i < svga->curr.num_samplers; ++i) {
 176       if (svga->curr.sampler[i]) {
 177          key->tex[i].compare_mode = svga->curr.sampler[i]->compare_mode;
 178          key->tex[i].compare_func = svga->curr.sampler[i]->compare_func;
 179       }
 180    }
 181
 182    return 0;
 183 }
 184
 185
 186
 187 static int emit_hw_fs( struct svga_context *svga,
 188                        unsigned dirty )
 189 {
 190    struct svga_shader_result *result = NULL;
 191    unsigned id = SVGA3D_INVALID_ID;
 192    int ret = 0;
 193
 194    struct svga_fragment_shader *fs = svga->curr.fs;
 195    struct svga_fs_compile_key key;
 196
 197    /* SVGA_NEW_BLEND
 198     * SVGA_NEW_TEXTURE_BINDING
 199     * SVGA_NEW_RAST
 200     * SVGA_NEW_NEED_SWTNL
 201     * SVGA_NEW_SAMPLER
 202     */
 203    ret = make_fs_key( svga, &key );
 204    if (ret)
 205       return ret;
 206
 207    result = search_fs_key( fs, &key );
 208    if (!result) {
 209       ret = compile_fs( svga, fs, &key, &result );
 210       if (ret)
 211          return ret;
 212    }
 213
 214    assert (result);
 215    id = result->id;
 216
 217    assert(id != SVGA3D_INVALID_ID);
 218
 219    if (result != svga->state.hw_draw.fs) {
 220       ret = SVGA3D_SetShader(svga->swc,
 221                              SVGA3D_SHADERTYPE_PS,
 222                              id );
 223       if (ret)
 224          return ret;
 225
 226       svga->dirty |= SVGA_NEW_FS_RESULT;
 227       svga->state.hw_draw.fs = result;
 228    }
 229
 230    return 0;
 231 }
 232
 233 struct svga_tracked_state svga_hw_fs =
 234 {
 235    "fragment shader (hwtnl)",
 236    (SVGA_NEW_FS |
 237     SVGA_NEW_TEXTURE_BINDING |
 238     SVGA_NEW_NEED_SWTNL |
 239     SVGA_NEW_RAST |
 240     SVGA_NEW_SAMPLER |
 241     SVGA_NEW_BLEND),
 242    emit_hw_fs
 243 };
 244
 245
 246