src/mesa/drivers/dri/i965/brw_vec4_vs_visitor.cpp

   1 /*
   2  * Copyright © 2013 Intel Corporation
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
  21  * DEALINGS IN THE SOFTWARE.
  22  */
  23
  24
  25 #include "brw_vs.h"
  26
  27
  28 namespace brw {
  29
  30 void
  31 vec4_vs_visitor::emit_prolog()
  32 {
  33    dst_reg sign_recovery_shift;
  34    dst_reg normalize_factor;
  35    dst_reg es3_normalize_factor;
  36
  37    for (int i = 0; i < VERT_ATTRIB_MAX; i++) {
  38       if (vs_prog_data->inputs_read & BITFIELD64_BIT(i)) {
  39          uint8_t wa_flags = key->gl_attrib_wa_flags[i];
  40          dst_reg reg(ATTR, i);
  41          dst_reg reg_d = reg;
  42          reg_d.type = BRW_REGISTER_TYPE_D;
  43          dst_reg reg_ud = reg;
  44          reg_ud.type = BRW_REGISTER_TYPE_UD;
  45
  46          /* Do GL_FIXED rescaling for GLES2.0.  Our GL_FIXED attributes
  47           * come in as floating point conversions of the integer values.
  48           */
  49          if (wa_flags & BRW_ATTRIB_WA_COMPONENT_MASK) {
  50             dst_reg dst = reg;
  51             dst.type = brw_type_for_base_type(glsl_type::vec4_type);
  52             dst.writemask = (1 << (wa_flags & BRW_ATTRIB_WA_COMPONENT_MASK)) - 1;
  53             emit(MUL(dst, src_reg(dst), src_reg(1.0f / 65536.0f)));
  54          }
  55
  56          /* Do sign recovery for 2101010 formats if required. */
  57          if (wa_flags & BRW_ATTRIB_WA_SIGN) {
  58             if (sign_recovery_shift.file == BAD_FILE) {
  59                /* shift constant: <22,22,22,30> */
  60                sign_recovery_shift = dst_reg(this, glsl_type::uvec4_type);
  61                emit(MOV(writemask(sign_recovery_shift, WRITEMASK_XYZ), src_reg(22u)));
  62                emit(MOV(writemask(sign_recovery_shift, WRITEMASK_W), src_reg(30u)));
  63             }
  64
  65             emit(SHL(reg_ud, src_reg(reg_ud), src_reg(sign_recovery_shift)));
  66             emit(ASR(reg_d, src_reg(reg_d), src_reg(sign_recovery_shift)));
  67          }
  68
  69          /* Apply BGRA swizzle if required. */
  70          if (wa_flags & BRW_ATTRIB_WA_BGRA) {
  71             src_reg temp = src_reg(reg);
  72             temp.swizzle = BRW_SWIZZLE4(2,1,0,3);
  73             emit(MOV(reg, temp));
  74          }
  75
  76          if (wa_flags & BRW_ATTRIB_WA_NORMALIZE) {
  77             /* ES 3.0 has different rules for converting signed normalized
  78              * fixed-point numbers than desktop GL.
  79              */
  80             if ((wa_flags & BRW_ATTRIB_WA_SIGN) && !use_legacy_snorm_formula) {
  81                /* According to equation 2.2 of the ES 3.0 specification,
  82                 * signed normalization conversion is done by:
  83                 *
  84                 * f = c / (2^(b-1)-1)
  85                 */
  86                if (es3_normalize_factor.file == BAD_FILE) {
  87                   /* mul constant: 1 / (2^(b-1) - 1) */
  88                   es3_normalize_factor = dst_reg(this, glsl_type::vec4_type);
  89                   emit(MOV(writemask(es3_normalize_factor, WRITEMASK_XYZ),
  90                            src_reg(1.0f / ((1<<9) - 1))));
  91                   emit(MOV(writemask(es3_normalize_factor, WRITEMASK_W),
  92                            src_reg(1.0f / ((1<<1) - 1))));
  93                }
  94
  95                dst_reg dst = reg;
  96                dst.type = brw_type_for_base_type(glsl_type::vec4_type);
  97                emit(MOV(dst, src_reg(reg_d)));
  98                emit(MUL(dst, src_reg(dst), src_reg(es3_normalize_factor)));
  99                emit_minmax(BRW_CONDITIONAL_GE, dst, src_reg(dst), src_reg(-1.0f));
 100             } else {
 101                /* The following equations are from the OpenGL 3.2 specification:
 102                 *
 103                 * 2.1 unsigned normalization
 104                 * f = c/(2^n-1)
 105                 *
 106                 * 2.2 signed normalization
 107                 * f = (2c+1)/(2^n-1)
 108                 *
 109                 * Both of these share a common divisor, which is represented by
 110                 * "normalize_factor" in the code below.
 111                 */
 112                if (normalize_factor.file == BAD_FILE) {
 113                   /* 1 / (2^b - 1) for b=<10,10,10,2> */
 114                   normalize_factor = dst_reg(this, glsl_type::vec4_type);
 115                   emit(MOV(writemask(normalize_factor, WRITEMASK_XYZ),
 116                            src_reg(1.0f / ((1<<10) - 1))));
 117                   emit(MOV(writemask(normalize_factor, WRITEMASK_W),
 118                            src_reg(1.0f / ((1<<2) - 1))));
 119                }
 120
 121                dst_reg dst = reg;
 122                dst.type = brw_type_for_base_type(glsl_type::vec4_type);
 123                emit(MOV(dst, src_reg((wa_flags & BRW_ATTRIB_WA_SIGN) ? reg_d : reg_ud)));
 124
 125                /* For signed normalization, we want the numerator to be 2c+1. */
 126                if (wa_flags & BRW_ATTRIB_WA_SIGN) {
 127                   emit(MUL(dst, src_reg(dst), src_reg(2.0f)));
 128                   emit(ADD(dst, src_reg(dst), src_reg(1.0f)));
 129                }
 130
 131                emit(MUL(dst, src_reg(dst), src_reg(normalize_factor)));
 132             }
 133          }
 134
 135          if (wa_flags & BRW_ATTRIB_WA_SCALE) {
 136             dst_reg dst = reg;
 137             dst.type = brw_type_for_base_type(glsl_type::vec4_type);
 138             emit(MOV(dst, src_reg((wa_flags & BRW_ATTRIB_WA_SIGN) ? reg_d : reg_ud)));
 139          }
 140       }
 141    }
 142 }
 143
 144
 145 dst_reg *
 146 vec4_vs_visitor::make_reg_for_system_value(int location,
 147                                            const glsl_type *type)
 148 {
 149    /* VertexID is stored by the VF as the last vertex element, but
 150     * we don't represent it with a flag in inputs_read, so we call
 151     * it VERT_ATTRIB_MAX, which setup_attributes() picks up on.
 152     */
 153    dst_reg *reg = new(mem_ctx) dst_reg(ATTR, VERT_ATTRIB_MAX);
 154
 155    switch (location) {
 156    case SYSTEM_VALUE_BASE_VERTEX:
 157       reg->writemask = WRITEMASK_X;
 158       vs_prog_data->uses_vertexid = true;
 159       break;
 160    case SYSTEM_VALUE_VERTEX_ID:
 161    case SYSTEM_VALUE_VERTEX_ID_ZERO_BASE:
 162       reg->writemask = WRITEMASK_Z;
 163       vs_prog_data->uses_vertexid = true;
 164       break;
 165    case SYSTEM_VALUE_INSTANCE_ID:
 166       reg->writemask = WRITEMASK_W;
 167       vs_prog_data->uses_instanceid = true;
 168       break;
 169    default:
 170       unreachable("not reached");
 171    }
 172
 173    return reg;
 174 }
 175
 176
 177 void
 178 vec4_vs_visitor::emit_urb_write_header(int mrf)
 179 {
 180    /* No need to do anything for VS; an implied write to this MRF will be
 181     * performed by VS_OPCODE_URB_WRITE.
 182     */
 183    (void) mrf;
 184 }
 185
 186
 187 vec4_instruction *
 188 vec4_vs_visitor::emit_urb_write_opcode(bool complete)
 189 {
 190    /* For VS, the URB writes end the thread. */
 191    if (complete) {
 192       if (INTEL_DEBUG & DEBUG_SHADER_TIME)
 193          emit_shader_time_end();
 194    }
 195
 196    vec4_instruction *inst = emit(VS_OPCODE_URB_WRITE);
 197    inst->urb_write_flags = complete ?
 198       BRW_URB_WRITE_EOT_COMPLETE : BRW_URB_WRITE_NO_FLAGS;
 199
 200    return inst;
 201 }
 202
 203
 204 void
 205 vec4_vs_visitor::emit_urb_slot(dst_reg reg, int varying)
 206 {
 207    reg.type = BRW_REGISTER_TYPE_F;
 208    output_reg[varying].type = reg.type;
 209
 210    switch (varying) {
 211    case VARYING_SLOT_COL0:
 212    case VARYING_SLOT_COL1:
 213    case VARYING_SLOT_BFC0:
 214    case VARYING_SLOT_BFC1: {
 215       /* These built-in varyings are only supported in compatibility mode,
 216        * and we only support GS in core profile.  So, this must be a vertex
 217        * shader.
 218        */
 219       vec4_instruction *inst = emit_generic_urb_slot(reg, varying);
 220       if (key->clamp_vertex_color)
 221          inst->saturate = true;
 222       break;
 223    }
 224    default:
 225       return vec4_visitor::emit_urb_slot(reg, varying);
 226    }
 227 }
 228
 229
 230 void
 231 vec4_vs_visitor::emit_clip_distances(dst_reg reg, int offset)
 232 {
 233    /* From the GLSL 1.30 spec, section 7.1 (Vertex Shader Special Variables):
 234     *
 235     *     "If a linked set of shaders forming the vertex stage contains no
 236     *     static write to gl_ClipVertex or gl_ClipDistance, but the
 237     *     application has requested clipping against user clip planes through
 238     *     the API, then the coordinate written to gl_Position is used for
 239     *     comparison against the user clip planes."
 240     *
 241     * This function is only called if the shader didn't write to
 242     * gl_ClipDistance.  Accordingly, we use gl_ClipVertex to perform clipping
 243     * if the user wrote to it; otherwise we use gl_Position.
 244     */
 245    gl_varying_slot clip_vertex = VARYING_SLOT_CLIP_VERTEX;
 246    if (!(prog_data->vue_map.slots_valid & VARYING_BIT_CLIP_VERTEX)) {
 247       clip_vertex = VARYING_SLOT_POS;
 248    }
 249
 250    for (int i = 0; i + offset < key->base.nr_userclip_plane_consts && i < 4;
 251         ++i) {
 252       reg.writemask = 1 << i;
 253       emit(DP4(reg,
 254                src_reg(output_reg[clip_vertex]),
 255                src_reg(this->userplane[i + offset])));
 256    }
 257 }
 258
 259
 260 void
 261 vec4_vs_visitor::setup_uniform_clipplane_values()
 262 {
 263    for (int i = 0; i < key->base.nr_userclip_plane_consts; ++i) {
 264       assert(this->uniforms < uniform_array_size);
 265       this->uniform_vector_size[this->uniforms] = 4;
 266       this->userplane[i] = dst_reg(UNIFORM, this->uniforms);
 267       this->userplane[i].type = BRW_REGISTER_TYPE_F;
 268       for (int j = 0; j < 4; ++j) {
 269          stage_prog_data->param[this->uniforms * 4 + j] =
 270             (gl_constant_value *) &clip_planes[i][j];
 271       }
 272       ++this->uniforms;
 273    }
 274 }
 275
 276
 277 void
 278 vec4_vs_visitor::emit_thread_end()
 279 {
 280    setup_uniform_clipplane_values();
 281
 282    /* Lower legacy ff and ClipVertex clipping to clip distances */
 283    if (key->base.nr_userclip_plane_consts > 0) {
 284       current_annotation = "user clip distances";
 285
 286       output_reg[VARYING_SLOT_CLIP_DIST0] = dst_reg(this, glsl_type::vec4_type);
 287       output_reg[VARYING_SLOT_CLIP_DIST1] = dst_reg(this, glsl_type::vec4_type);
 288
 289       emit_clip_distances(output_reg[VARYING_SLOT_CLIP_DIST0], 0);
 290       emit_clip_distances(output_reg[VARYING_SLOT_CLIP_DIST1], 4);
 291    }
 292
 293    /* For VS, we always end the thread by emitting a single vertex.
 294     * emit_urb_write_opcode() will take care of setting the eot flag on the
 295     * SEND instruction.
 296     */
 297    emit_vertex();
 298 }
 299
 300
 301 vec4_vs_visitor::vec4_vs_visitor(const struct brw_compiler *compiler,
 302                                  void *log_data,
 303                                  const struct brw_vs_prog_key *key,
 304                                  struct brw_vs_prog_data *vs_prog_data,
 305                                  struct gl_vertex_program *vp,
 306                                  struct gl_shader_program *prog,
 307                                  gl_clip_plane *clip_planes,
 308                                  void *mem_ctx,
 309                                  int shader_time_index,
 310                                  bool use_legacy_snorm_formula)
 311    : vec4_visitor(compiler, log_data,
 312                   &vp->Base, &key->base, &vs_prog_data->base, prog,
 313                   MESA_SHADER_VERTEX,
 314                   mem_ctx, false /* no_spills */,
 315                   shader_time_index),
 316      key(key),
 317      vs_prog_data(vs_prog_data),
 318      vp(vp),
 319      clip_planes(clip_planes),
 320      use_legacy_snorm_formula(use_legacy_snorm_formula)
 321 {
 322 }
 323
 324
 325 } /* namespace brw */