Merge branch 'wip/i965-separate-sampler-tex' into vulkan

[mesa.git] / src / mesa / drivers / dri / i965 / brw_vec4_visitor.cpp

diff --git a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
index 9b04acb8ed64fd4be852ee89e15db0c4a79bc03e..6b8798da71ccddf95696d292e646d70c160d96b2 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
@@ -237,8 +237,6 @@ vec4_visitor::CMP(dst_reg dst, src_reg src0, src_reg src1,
     * type to match src0 so we can compact the instruction.
     */
    dst.type = src0.type;
-   if (dst.file == HW_REG)
-      dst.fixed_hw_reg.type = dst.type;
 
    resolve_ud_negate(&src0);
    resolve_ud_negate(&src1);
@@ -624,6 +622,7 @@ type_size_vec4(const struct glsl_type *type)
    case GLSL_TYPE_DOUBLE:
    case GLSL_TYPE_ERROR:
    case GLSL_TYPE_INTERFACE:
+   case GLSL_TYPE_FUNCTION:
       unreachable("not reached");
    }
 
@@ -634,8 +633,8 @@ src_reg::src_reg(class vec4_visitor *v, const struct glsl_type *type)
 {
    init();
 
-   this->file = GRF;
-   this->reg = v->alloc.allocate(type_size_vec4(type));
+   this->file = VGRF;
+   this->nr = v->alloc.allocate(type_size_vec4(type));
 
    if (type->is_array() || type->is_record()) {
       this->swizzle = BRW_SWIZZLE_NOOP;
@@ -652,8 +651,8 @@ src_reg::src_reg(class vec4_visitor *v, const struct glsl_type *type, int size)
 
    init();
 
-   this->file = GRF;
-   this->reg = v->alloc.allocate(type_size_vec4(type) * size);
+   this->file = VGRF;
+   this->nr = v->alloc.allocate(type_size_vec4(type) * size);
 
    this->swizzle = BRW_SWIZZLE_NOOP;
 
@@ -664,8 +663,8 @@ dst_reg::dst_reg(class vec4_visitor *v, const struct glsl_type *type)
 {
    init();
 
-   this->file = GRF;
-   this->reg = v->alloc.allocate(type_size_vec4(type));
+   this->file = VGRF;
+   this->nr = v->alloc.allocate(type_size_vec4(type));
 
    if (type->is_array() || type->is_record()) {
       this->writemask = WRITEMASK_XYZW;
@@ -863,7 +862,7 @@ vec4_visitor::is_high_sampler(src_reg sampler)
    if (devinfo->gen < 8 && !devinfo->is_haswell)
       return false;
 
-   return sampler.file != IMM || sampler.fixed_hw_reg.ud >= 16;
+   return sampler.file != IMM || sampler.ud >= 16;
 }
 
 void
@@ -879,6 +878,8 @@ vec4_visitor::emit_texture(ir_texture_opcode op,
                            src_reg offset_value,
                            src_reg mcs,
                            bool is_cube_array,
+                           uint32_t surface,
+                           src_reg surface_reg,
                            uint32_t sampler,
                            src_reg sampler_reg)
 {
@@ -938,7 +939,8 @@ vec4_visitor::emit_texture(ir_texture_opcode op,
    inst->dst.writemask = WRITEMASK_XYZW;
    inst->shadow_compare = shadow_comparitor.file != BAD_FILE;
 
-   inst->src[1] = sampler_reg;
+   inst->src[1] = surface_reg;
+   inst->src[2] = sampler_reg;
 
    /* MRF for the first parameter */
    int param_base = inst->base_mrf + inst->header_size;
@@ -1064,7 +1066,7 @@ vec4_visitor::emit_texture(ir_texture_opcode op,
    }
 
    if (devinfo->gen == 6 && op == ir_tg4) {
-      emit_gen6_gather_wa(key_tex->gen6_gather_wa[sampler], inst->dst);
+      emit_gen6_gather_wa(key_tex->gen6_gather_wa[surface], inst->dst);
    }
 
    swizzle_result(op, dest,
@@ -1102,7 +1104,8 @@ vec4_visitor::emit_gen6_gather_wa(uint8_t wa, dst_reg dst)
  * Set up the gather channel based on the swizzle, for gather4.
  */
 uint32_t
-vec4_visitor::gather_channel(unsigned gather_component, uint32_t sampler)
+vec4_visitor::gather_channel(unsigned gather_component,
+                             uint32_t surface, uint32_t sampler)
 {
    int swiz = GET_SWZ(key_tex->swizzles[sampler], gather_component);
    switch (swiz) {
@@ -1111,7 +1114,7 @@ vec4_visitor::gather_channel(unsigned gather_component, uint32_t sampler)
          /* gather4 sampler is broken for green channel on RG32F --
           * we must ask for blue instead.
           */
-         if (key_tex->gather_channel_quirk_mask & (1 << sampler))
+         if (key_tex->gather_channel_quirk_mask & (1 << surface))
             return 2;
          return 1;
       case SWIZZLE_Z: return 2;
@@ -1617,7 +1620,7 @@ vec4_visitor::emit_scratch_write(bblock_t *block, vec4_instruction *inst,
    inst->insert_after(block, write);
 
    inst->dst.file = temp.file;
-   inst->dst.reg = temp.reg;
+   inst->dst.nr = temp.nr;
    inst->dst.reg_offset = temp.reg_offset;
    inst->dst.reladdr = NULL;
 }
@@ -1644,10 +1647,10 @@ vec4_visitor::emit_resolve_reladdr(int scratch_loc[], bblock_t *block,
                                           *src.reladdr);
 
    /* Now handle scratch access on src */
-   if (src.file == GRF && scratch_loc[src.reg] != -1) {
+   if (src.file == VGRF && scratch_loc[src.nr] != -1) {
       dst_reg temp = dst_reg(this, glsl_type::vec4_type);
-      emit_scratch_read(block, inst, temp, src, scratch_loc[src.reg]);
-      src.reg = temp.reg;
+      emit_scratch_read(block, inst, temp, src, scratch_loc[src.nr]);
+      src.nr = temp.nr;
       src.reg_offset = temp.reg_offset;
       src.reladdr = NULL;
    }
@@ -1672,18 +1675,18 @@ vec4_visitor::move_grf_array_access_to_scratch()
     * scratch.
     */
    foreach_block_and_inst(block, vec4_instruction, inst, cfg) {
-      if (inst->dst.file == GRF && inst->dst.reladdr) {
-         if (scratch_loc[inst->dst.reg] == -1) {
-            scratch_loc[inst->dst.reg] = last_scratch;
-            last_scratch += this->alloc.sizes[inst->dst.reg];
+      if (inst->dst.file == VGRF && inst->dst.reladdr) {
+         if (scratch_loc[inst->dst.nr] == -1) {
+            scratch_loc[inst->dst.nr] = last_scratch;
+            last_scratch += this->alloc.sizes[inst->dst.nr];
          }
 
          for (src_reg *iter = inst->dst.reladdr;
               iter->reladdr;
               iter = iter->reladdr) {
-            if (iter->file == GRF && scratch_loc[iter->reg] == -1) {
-               scratch_loc[iter->reg] = last_scratch;
-               last_scratch += this->alloc.sizes[iter->reg];
+            if (iter->file == VGRF && scratch_loc[iter->nr] == -1) {
+               scratch_loc[iter->nr] = last_scratch;
+               last_scratch += this->alloc.sizes[iter->nr];
             }
          }
       }
@@ -1692,9 +1695,9 @@ vec4_visitor::move_grf_array_access_to_scratch()
          for (src_reg *iter = &inst->src[i];
               iter->reladdr;
               iter = iter->reladdr) {
-            if (iter->file == GRF && scratch_loc[iter->reg] == -1) {
-               scratch_loc[iter->reg] = last_scratch;
-               last_scratch += this->alloc.sizes[iter->reg];
+            if (iter->file == VGRF && scratch_loc[iter->nr] == -1) {
+               scratch_loc[iter->nr] = last_scratch;
+               last_scratch += this->alloc.sizes[iter->nr];
             }
          }
       }
@@ -1720,8 +1723,8 @@ vec4_visitor::move_grf_array_access_to_scratch()
       /* Now that we have handled any (possibly recursive) reladdr scratch
        * accesses for dst we can safely do the scratch write for dst itself
        */
-      if (inst->dst.file == GRF && scratch_loc[inst->dst.reg] != -1)
-         emit_scratch_write(block, inst, scratch_loc[inst->dst.reg]);
+      if (inst->dst.file == VGRF && scratch_loc[inst->dst.nr] != -1)
+         emit_scratch_write(block, inst, scratch_loc[inst->dst.nr]);
 
       /* Now handle scratch access on any src. In this case, since inst->src[i]
        * already is a src_reg, we can just call emit_resolve_reladdr with
@@ -1790,7 +1793,7 @@ vec4_visitor::move_uniform_array_access_to_pull_constants()
             if (inst->src[i].file != UNIFORM || !inst->src[i].reladdr)
                continue;
 
-            int uniform = inst->src[i].reg;
+            int uniform = inst->src[i].nr;
 
             if (inst->src[i].reladdr->reladdr)
                nested_reladdr = true;  /* will need another pass */
@@ -1821,7 +1824,7 @@ vec4_visitor::move_uniform_array_access_to_pull_constants()
                                     pull_constant_loc[uniform]);
 
             inst->src[i].file = temp.file;
-            inst->src[i].reg = temp.reg;
+            inst->src[i].nr = temp.nr;
             inst->src[i].reg_offset = temp.reg_offset;
             inst->src[i].reladdr = NULL;
          }