src/gallium/drivers/softpipe/sp_fs_sse.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 #include "sp_context.h"
  30 #include "sp_state.h"
  31 #include "sp_fs.h"
  32 #include "sp_quad.h"
  33
  34
  35 #include "pipe/p_state.h"
  36 #include "pipe/p_defines.h"
  37 #include "util/u_memory.h"
  38 #include "tgsi/tgsi_exec.h"
  39 #include "tgsi/tgsi_sse2.h"
  40
  41
  42 #if defined(PIPE_ARCH_X86)
  43
  44 #include "rtasm/rtasm_x86sse.h"
  45
  46 /* Surely this should be defined somewhere in a tgsi header:
  47  */
  48 typedef void (PIPE_CDECL *codegen_function)(
  49    const struct tgsi_exec_vector *input,
  50    struct tgsi_exec_vector *output,
  51    const float (*constant)[4],
  52    struct tgsi_exec_vector *temporary,
  53    const struct tgsi_interp_coef *coef,
  54    float (*immediates)[4]
  55    //, const struct tgsi_exec_vector *quadPos
  56  );
  57
  58
  59 struct sp_sse_fragment_shader {
  60    struct sp_fragment_shader base;
  61    struct x86_function             sse2_program;
  62    codegen_function func;
  63    float immediates[TGSI_EXEC_NUM_IMMEDIATES][4];
  64 };
  65
  66
  67
  68 static void
  69 fs_sse_prepare( const struct sp_fragment_shader *base,
  70                 struct tgsi_exec_machine *machine,
  71                 struct tgsi_sampler **samplers )
  72 {
  73 }
  74
  75
  76 /* TODO: codegenerate the whole run function, skip this wrapper.
  77  * TODO: break dependency on tgsi_exec_machine struct
  78  * TODO: push Position calculation into the generated shader
  79  * TODO: process >1 quad at a time
  80  */
  81 static unsigned
  82 fs_sse_run( const struct sp_fragment_shader *base,
  83             struct tgsi_exec_machine *machine,
  84             struct quad_header *quad )
  85 {
  86    struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base;
  87
  88    /* Compute X, Y, Z, W vals for this quad -- place in temp[0] for now */
  89    sp_setup_pos_vector(quad->posCoef,
  90                        (float)quad->input.x0, (float)quad->input.y0,
  91                        machine->Temps);
  92
  93    /* init kill mask */
  94    tgsi_set_kill_mask(machine, 0x0);
  95    tgsi_set_exec_mask(machine, 1, 1, 1, 1);
  96
  97    shader->func( machine->Inputs,
  98                  machine->Outputs,
  99                  machine->Consts,
 100                  machine->Temps,
 101                  machine->InterpCoefs,
 102                  shader->immediates
 103                  //      , &machine->QuadPos
 104       );
 105
 106    return ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]);
 107 }
 108
 109
 110 static void
 111 fs_sse_delete( struct sp_fragment_shader *base )
 112 {
 113    struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base;
 114
 115    x86_release_func( &shader->sse2_program );
 116    FREE(shader);
 117 }
 118
 119
 120 struct sp_fragment_shader *
 121 softpipe_create_fs_sse(struct softpipe_context *softpipe,
 122                        const struct pipe_shader_state *templ)
 123 {
 124    struct sp_sse_fragment_shader *shader;
 125
 126    if (!softpipe->use_sse)
 127       return NULL;
 128
 129    shader = CALLOC_STRUCT(sp_sse_fragment_shader);
 130    if (!shader)
 131       return NULL;
 132
 133    x86_init_func( &shader->sse2_program );
 134
 135    if (!tgsi_emit_sse2( templ->tokens, &shader->sse2_program,
 136                         shader->immediates, FALSE )) {
 137       FREE(shader);
 138       return NULL;
 139    }
 140
 141    shader->func = (codegen_function) x86_get_func( &shader->sse2_program );
 142    if (!shader->func) {
 143       x86_release_func( &shader->sse2_program );
 144       FREE(shader);
 145       return NULL;
 146    }
 147
 148    shader->base.shader.tokens = NULL; /* don't hold reference to templ->tokens */
 149    shader->base.prepare = fs_sse_prepare;
 150    shader->base.run = fs_sse_run;
 151    shader->base.delete = fs_sse_delete;
 152
 153    return &shader->base;
 154 }
 155
 156
 157 #else
 158
 159 /* Maybe put this varient in the header file.
 160  */
 161 struct sp_fragment_shader *
 162 softpipe_create_fs_sse(struct softpipe_context *softpipe,
 163                        const struct pipe_shader_state *templ)
 164 {
 165    return NULL;
 166 }
 167
 168 #endif