aboutsummaryrefslogtreecommitdiffstats
path: root/src/mesa
diff options
context:
space:
mode:
authorKenneth Graunke <[email protected]>2014-05-16 02:21:51 -0700
committerKenneth Graunke <[email protected]>2015-01-19 13:13:34 -0800
commitd1533d87cc7e2c39e7ce9dc838b45a2c39c96e33 (patch)
tree881f381238c2eddd15bf9327ed193e514cb19be7 /src/mesa
parent5b01512df3328711968dea7240ae817721c79774 (diff)
i965: Replace fs_reg(fs_visitor, type) with fs_visitor::vgrf(type).
I dislike how fs_reg has a constructor that knows about fs_visitor. Apart from that, it stands alone, with no need to interact with the rest of the compiler. Which is sensible - a class that represents a register should do just that. Allocating virtual register numbers should be left up to the compiler (fs_visitor). This patch replaces the constructor with a new fs_visitor::vgrf method, eliminating fs_reg's dependency on fs_visitor. It ends up being no more code. v2: Rebase from May 2014 -> January 2015. Signed-off-by: Kenneth Graunke <[email protected]> Reviewed-by: Jason Ekstrand <[email protected]> Reviewed-by: Matt Turner <[email protected]>
Diffstat (limited to 'src/mesa')
-rw-r--r--src/mesa/drivers/dri/i965/brw_fs.cpp50
-rw-r--r--src/mesa/drivers/dri/i965/brw_fs.h2
-rw-r--r--src/mesa/drivers/dri/i965/brw_fs_fp.cpp46
-rw-r--r--src/mesa/drivers/dri/i965/brw_fs_nir.cpp18
-rw-r--r--src/mesa/drivers/dri/i965/brw_fs_sel_peephole.cpp2
-rw-r--r--src/mesa/drivers/dri/i965/brw_fs_visitor.cpp132
6 files changed, 122 insertions, 128 deletions
diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index 35639de0db7..817ca2e860f 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -406,7 +406,7 @@ fs_visitor::VARYING_PULL_CONSTANT_LOAD(const fs_reg &dst,
* CSE can later notice that those loads are all the same and eliminate
* the redundant ones.
*/
- fs_reg vec4_offset = fs_reg(this, glsl_type::int_type);
+ fs_reg vec4_offset = vgrf(glsl_type::int_type);
instructions.push_tail(ADD(vec4_offset,
varying_offset, fs_reg(const_offset & ~3)));
@@ -778,9 +778,9 @@ fs_visitor::emit_shader_time_write(enum shader_time_shader_type type,
fs_reg payload;
if (dispatch_width == 8)
- payload = fs_reg(this, glsl_type::uvec2_type);
+ payload = vgrf(glsl_type::uvec2_type);
else
- payload = fs_reg(this, glsl_type::uint_type);
+ payload = vgrf(glsl_type::uint_type);
emit(new(mem_ctx) fs_inst(SHADER_OPCODE_SHADER_TIME_ADD,
fs_reg(), payload, offset, value));
@@ -1032,6 +1032,14 @@ fs_visitor::virtual_grf_alloc(int size)
return virtual_grf_count++;
}
+fs_reg
+fs_visitor::vgrf(const glsl_type *const type)
+{
+ int reg_width = dispatch_width / 8;
+ return fs_reg(GRF, virtual_grf_alloc(type_size(type) * reg_width),
+ brw_type_for_base_type(type), dispatch_width);
+}
+
/** Fixed HW reg constructor. */
fs_reg::fs_reg(enum register_file file, int reg)
{
@@ -1077,20 +1085,6 @@ fs_reg::fs_reg(enum register_file file, int reg, enum brw_reg_type type,
this->width = width;
}
-/** Automatic reg constructor. */
-fs_reg::fs_reg(fs_visitor *v, const struct glsl_type *type)
-{
- init();
- int reg_width = v->dispatch_width / 8;
-
- this->file = GRF;
- this->reg = v->virtual_grf_alloc(v->type_size(type) * reg_width);
- this->reg_offset = 0;
- this->type = brw_type_for_base_type(type);
- this->width = v->dispatch_width;
- assert(this->width == 8 || this->width == 16);
-}
-
fs_reg *
fs_visitor::variable_storage(ir_variable *var)
{
@@ -1208,7 +1202,7 @@ fs_visitor::emit_fragcoord_interpolation(bool pixel_center_integer,
{
assert(stage == MESA_SHADER_FRAGMENT);
brw_wm_prog_key *key = (brw_wm_prog_key*) this->key;
- fs_reg *reg = new(this->mem_ctx) fs_reg(this, glsl_type::vec4_type);
+ fs_reg *reg = new(this->mem_ctx) fs_reg(vgrf(glsl_type::vec4_type));
fs_reg wpos = *reg;
bool flip = !origin_upper_left ^ key->render_to_fbo;
@@ -1394,7 +1388,7 @@ fs_visitor::emit_general_interpolation(fs_reg attr, const char *name,
fs_reg *
fs_visitor::emit_frontfacing_interpolation()
{
- fs_reg *reg = new(this->mem_ctx) fs_reg(this, glsl_type::bool_type);
+ fs_reg *reg = new(this->mem_ctx) fs_reg(vgrf(glsl_type::bool_type));
if (brw->gen >= 6) {
/* Bit 15 of g0.0 is 0 if the polygon is front facing. We want to create
@@ -1461,10 +1455,10 @@ fs_visitor::emit_samplepos_setup()
assert(brw->gen >= 6);
this->current_annotation = "compute sample position";
- fs_reg *reg = new(this->mem_ctx) fs_reg(this, glsl_type::vec2_type);
+ fs_reg *reg = new(this->mem_ctx) fs_reg(vgrf(glsl_type::vec2_type));
fs_reg pos = *reg;
- fs_reg int_sample_x = fs_reg(this, glsl_type::int_type);
- fs_reg int_sample_y = fs_reg(this, glsl_type::int_type);
+ fs_reg int_sample_x = vgrf(glsl_type::int_type);
+ fs_reg int_sample_y = vgrf(glsl_type::int_type);
/* WM will be run in MSDISPMODE_PERSAMPLE. So, only one of SIMD8 or SIMD16
* mode will be enabled.
@@ -1512,11 +1506,11 @@ fs_visitor::emit_sampleid_setup()
assert(brw->gen >= 6);
this->current_annotation = "compute sample id";
- fs_reg *reg = new(this->mem_ctx) fs_reg(this, glsl_type::int_type);
+ fs_reg *reg = new(this->mem_ctx) fs_reg(vgrf(glsl_type::int_type));
if (key->compute_sample_id) {
- fs_reg t1 = fs_reg(this, glsl_type::int_type);
- fs_reg t2 = fs_reg(this, glsl_type::int_type);
+ fs_reg t1 = vgrf(glsl_type::int_type);
+ fs_reg t2 = vgrf(glsl_type::int_type);
t2.type = BRW_REGISTER_TYPE_UW;
/* The PS will be run in MSDISPMODE_PERSAMPLE. For example with
@@ -1584,7 +1578,7 @@ fs_visitor::fix_math_operand(fs_reg src)
if (brw->gen >= 7 && src.file != IMM)
return src;
- fs_reg expanded = fs_reg(this, glsl_type::float_type);
+ fs_reg expanded = vgrf(glsl_type::float_type);
expanded.type = src.type;
emit(BRW_OPCODE_MOV, expanded, src);
return expanded;
@@ -2261,7 +2255,7 @@ fs_visitor::demote_pull_constants()
current_annotation = inst->annotation;
fs_reg surf_index(stage_prog_data->binding_table.pull_constants_start);
- fs_reg dst = fs_reg(this, glsl_type::float_type);
+ fs_reg dst = vgrf(glsl_type::float_type);
/* Generate a pull load into dst. */
if (inst->src[i].reladdr) {
@@ -2997,7 +2991,7 @@ fs_visitor::lower_uniform_pull_constant_loads()
assert(const_offset_reg.file == IMM &&
const_offset_reg.type == BRW_REGISTER_TYPE_UD);
const_offset_reg.fixed_hw_reg.dw1.ud /= 4;
- fs_reg payload = fs_reg(this, glsl_type::uint_type);
+ fs_reg payload = vgrf(glsl_type::uint_type);
/* We have to use a message header on Skylake to get SIMD4x2 mode.
* Reserve space for the register.
diff --git a/src/mesa/drivers/dri/i965/brw_fs.h b/src/mesa/drivers/dri/i965/brw_fs.h
index 2aa58ebf0cb..466f90f476c 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.h
+++ b/src/mesa/drivers/dri/i965/brw_fs.h
@@ -83,7 +83,6 @@ public:
fs_reg(enum register_file file, int reg);
fs_reg(enum register_file file, int reg, enum brw_reg_type type);
fs_reg(enum register_file file, int reg, enum brw_reg_type type, uint8_t width);
- fs_reg(fs_visitor *v, const struct glsl_type *type);
bool equals(const fs_reg &r) const;
bool is_contiguous() const;
@@ -323,6 +322,7 @@ public:
fs_reg *variable_storage(ir_variable *var);
int virtual_grf_alloc(int size);
+ fs_reg vgrf(const glsl_type *const type);
void import_uniforms(fs_visitor *v);
void setup_uniform_clipplane_values();
void compute_clip_distance();
diff --git a/src/mesa/drivers/dri/i965/brw_fs_fp.cpp b/src/mesa/drivers/dri/i965/brw_fs_fp.cpp
index 4639590c8b9..7f2874d9258 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_fp.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_fp.cpp
@@ -105,7 +105,7 @@ fs_visitor::emit_fp_scalar_math(enum opcode opcode,
const struct prog_instruction *fpi,
fs_reg dst, fs_reg src)
{
- fs_reg temp = fs_reg(this, glsl_type::float_type);
+ fs_reg temp = vgrf(glsl_type::float_type);
emit_math(opcode, temp, src);
emit_fp_scalar_write(fpi, dst, temp);
}
@@ -125,7 +125,7 @@ fs_visitor::emit_fragment_program_code()
* mov dst 0.0
* mov.f0 dst 1.0
*/
- fs_reg one = fs_reg(this, glsl_type::float_type);
+ fs_reg one = vgrf(glsl_type::float_type);
emit(MOV(one, fs_reg(1.0f)));
for (unsigned int insn = 0; insn < prog->NumInstructions; insn++) {
@@ -138,7 +138,7 @@ fs_visitor::emit_fragment_program_code()
/* We always emit into a temporary destination register to avoid
* aliasing issues.
*/
- dst = fs_reg(this, glsl_type::vec4_type);
+ dst = vgrf(glsl_type::vec4_type);
for (int i = 0; i < 3; i++)
src[i] = get_fp_src_reg(&fpi->SrcReg[i]);
@@ -177,8 +177,8 @@ fs_visitor::emit_fragment_program_code()
case OPCODE_DP3:
case OPCODE_DP4:
case OPCODE_DPH: {
- fs_reg mul = fs_reg(this, glsl_type::float_type);
- fs_reg acc = fs_reg(this, glsl_type::float_type);
+ fs_reg mul = vgrf(glsl_type::float_type);
+ fs_reg acc = vgrf(glsl_type::float_type);
int count;
switch (fpi->Opcode) {
@@ -316,7 +316,7 @@ fs_visitor::emit_fragment_program_code()
case OPCODE_MAD:
for (int i = 0; i < 4; i++) {
if (fpi->DstReg.WriteMask & (1 << i)) {
- fs_reg temp = fs_reg(this, glsl_type::float_type);
+ fs_reg temp = vgrf(glsl_type::float_type);
emit(MUL(temp, offset(src[0], i), offset(src[1], i)));
emit(ADD(offset(dst, i), temp, offset(src[2], i)));
}
@@ -340,7 +340,7 @@ fs_visitor::emit_fragment_program_code()
break;
case OPCODE_POW: {
- fs_reg temp = fs_reg(this, glsl_type::float_type);
+ fs_reg temp = vgrf(glsl_type::float_type);
emit_math(SHADER_OPCODE_POW, temp, src[0], src[1]);
emit_fp_scalar_write(fpi, dst, temp);
break;
@@ -404,8 +404,8 @@ fs_visitor::emit_fragment_program_code()
case OPCODE_TXP: {
op = ir_tex;
- coordinate = fs_reg(this, glsl_type::vec3_type);
- fs_reg invproj = fs_reg(this, glsl_type::float_type);
+ coordinate = vgrf(glsl_type::vec3_type);
+ fs_reg invproj = vgrf(glsl_type::float_type);
emit_math(SHADER_OPCODE_RCP, invproj, offset(src[0], 3));
for (int i = 0; i < 3; i++) {
emit(MUL(offset(coordinate, i),
@@ -442,8 +442,8 @@ fs_visitor::emit_fragment_program_code()
case TEXTURE_CUBE_INDEX: {
coord_components = 4;
- fs_reg temp = fs_reg(this, glsl_type::float_type);
- fs_reg cubecoord = fs_reg(this, glsl_type::vec3_type);
+ fs_reg temp = vgrf(glsl_type::float_type);
+ fs_reg cubecoord = vgrf(glsl_type::vec3_type);
fs_reg abscoord = coordinate;
abscoord.negate = false;
abscoord.abs = true;
@@ -495,7 +495,7 @@ fs_visitor::emit_fragment_program_code()
int i1 = (i + 1) % 3;
int i2 = (i + 2) % 3;
- fs_reg temp = fs_reg(this, glsl_type::float_type);
+ fs_reg temp = vgrf(glsl_type::float_type);
fs_reg neg_src1_1 = offset(src[1], i1);
neg_src1_1.negate = !neg_src1_1.negate;
emit(MUL(temp, offset(src[0], i2), neg_src1_1));
@@ -537,7 +537,7 @@ fs_visitor::emit_fragment_program_code()
*/
this->current_annotation = "result.depth write";
if (frag_depth.file != BAD_FILE) {
- fs_reg temp = fs_reg(this, glsl_type::float_type);
+ fs_reg temp = vgrf(glsl_type::float_type);
emit(MOV(temp, offset(frag_depth, 2)));
frag_depth = temp;
}
@@ -550,7 +550,7 @@ fs_visitor::setup_fp_regs()
int num_temp = prog->NumTemporaries;
fp_temp_regs = rzalloc_array(mem_ctx, fs_reg, num_temp);
for (int i = 0; i < num_temp; i++)
- fp_temp_regs[i] = fs_reg(this, glsl_type::vec4_type);
+ fp_temp_regs[i] = vgrf(glsl_type::vec4_type);
/* PROGRAM_STATE_VAR etc. */
if (dispatch_width == 8) {
@@ -583,7 +583,7 @@ fs_visitor::setup_fp_regs()
fp_input_regs[i] = *emit_frontfacing_interpolation();
break;
default:
- fp_input_regs[i] = fs_reg(this, glsl_type::vec4_type);
+ fp_input_regs[i] = vgrf(glsl_type::vec4_type);
emit_general_interpolation(fp_input_regs[i], "fp_input",
glsl_type::vec4_type,
INTERP_QUALIFIER_NONE,
@@ -616,11 +616,11 @@ fs_visitor::get_fp_dst_reg(const prog_dst_register *dst)
case PROGRAM_OUTPUT:
if (dst->Index == FRAG_RESULT_DEPTH) {
if (frag_depth.file == BAD_FILE)
- frag_depth = fs_reg(this, glsl_type::vec4_type);
+ frag_depth = vgrf(glsl_type::vec4_type);
return frag_depth;
} else if (dst->Index == FRAG_RESULT_COLOR) {
if (outputs[0].file == BAD_FILE) {
- outputs[0] = fs_reg(this, glsl_type::vec4_type);
+ outputs[0] = vgrf(glsl_type::vec4_type);
output_components[0] = 4;
/* Tell emit_fb_writes() to smear fragment.color across all the
@@ -635,7 +635,7 @@ fs_visitor::get_fp_dst_reg(const prog_dst_register *dst)
} else {
int output_index = dst->Index - FRAG_RESULT_DATA0;
if (outputs[output_index].file == BAD_FILE) {
- outputs[output_index] = fs_reg(this, glsl_type::vec4_type);
+ outputs[output_index] = vgrf(glsl_type::vec4_type);
}
output_components[output_index] = 4;
return outputs[output_index];
@@ -647,7 +647,7 @@ fs_visitor::get_fp_dst_reg(const prog_dst_register *dst)
default:
_mesa_problem(ctx, "bad dst register file: %s\n",
_mesa_register_file_name((gl_register_file)dst->File));
- return fs_reg(this, glsl_type::vec4_type);
+ return vgrf(glsl_type::vec4_type);
}
}
@@ -680,7 +680,7 @@ fs_visitor::get_fp_src_reg(const prog_src_register *src)
*/
switch (plist->Parameters[src->Index].Type) {
case PROGRAM_CONSTANT: {
- result = fs_reg(this, glsl_type::vec4_type);
+ result = vgrf(glsl_type::vec4_type);
for (int i = 0; i < 4; i++) {
emit(MOV(offset(result, i),
@@ -697,19 +697,19 @@ fs_visitor::get_fp_src_reg(const prog_src_register *src)
default:
_mesa_problem(ctx, "bad uniform src register file: %s\n",
_mesa_register_file_name((gl_register_file)src->File));
- return fs_reg(this, glsl_type::vec4_type);
+ return vgrf(glsl_type::vec4_type);
}
break;
default:
_mesa_problem(ctx, "bad src register file: %s\n",
_mesa_register_file_name((gl_register_file)src->File));
- return fs_reg(this, glsl_type::vec4_type);
+ return vgrf(glsl_type::vec4_type);
}
if (src->Swizzle != SWIZZLE_NOOP || src->Negate) {
fs_reg unswizzled = result;
- result = fs_reg(this, glsl_type::vec4_type);
+ result = vgrf(glsl_type::vec4_type);
for (int i = 0; i < 4; i++) {
bool negate = src->Negate & (1 << i);
/* The ZERO, ONE, and Negate options are only used for OPCODE_SWZ,
diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
index e8f398a7138..7d7c04cedc7 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
@@ -858,7 +858,7 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr)
}
case nir_op_fceil: {
op[0].negate = !op[0].negate;
- fs_reg temp = fs_reg(this, glsl_type::vec4_type);
+ fs_reg temp = vgrf(glsl_type::vec4_type);
emit_percomp(RNDD(temp, op[0]), instr->dest.write_mask);
temp.negate = true;
fs_inst *inst = MOV(result, temp);
@@ -1393,7 +1393,7 @@ fs_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr)
* per-channel and add the base UBO index; the generator will select
* a value from any live channel.
*/
- surf_index = fs_reg(this, glsl_type::uint_type);
+ surf_index = vgrf(glsl_type::uint_type);
emit(ADD(surf_index, get_nir_src(instr->src[0]),
fs_reg(stage_prog_data->binding_table.ubo_start)))
->force_writemask_all = true;
@@ -1408,7 +1408,7 @@ fs_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr)
if (has_indirect) {
/* Turn the byte offset into a dword offset. */
- fs_reg base_offset = fs_reg(this, glsl_type::int_type);
+ fs_reg base_offset = vgrf(glsl_type::int_type);
emit(SHR(base_offset, retype(get_nir_src(instr->src[1]),
BRW_REGISTER_TYPE_D),
fs_reg(2)));
@@ -1418,7 +1418,7 @@ fs_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr)
emit(VARYING_PULL_CONSTANT_LOAD(offset(dest, i), surf_index,
base_offset, vec4_offset + i));
} else {
- fs_reg packed_consts = fs_reg(this, glsl_type::float_type);
+ fs_reg packed_consts = vgrf(glsl_type::float_type);
packed_consts.type = dest.type;
fs_reg const_offset_reg((unsigned) instr->const_index[0] & ~15);
@@ -1494,7 +1494,7 @@ fs_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr)
* requires mlen==1 even when there is no payload. in the per-slot
* offset case, we'll replace this with the proper source data.
*/
- fs_reg src(this, glsl_type::float_type);
+ fs_reg src = vgrf(glsl_type::float_type);
int mlen = 1; /* one reg unless overriden */
fs_inst *inst;
@@ -1523,13 +1523,13 @@ fs_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr)
inst = emit(FS_OPCODE_INTERPOLATE_AT_SHARED_OFFSET, dst_x, src,
fs_reg(off_x | (off_y << 4)));
} else {
- src = fs_reg(this, glsl_type::ivec2_type);
+ src = vgrf(glsl_type::ivec2_type);
fs_reg offset_src = retype(get_nir_src(instr->src[0]),
BRW_REGISTER_TYPE_F);
for (int i = 0; i < 2; i++) {
- fs_reg temp(this, glsl_type::float_type);
+ fs_reg temp = vgrf(glsl_type::float_type);
emit(MUL(temp, offset(offset_src, i), fs_reg(16.0f)));
- fs_reg itemp(this, glsl_type::int_type);
+ fs_reg itemp = vgrf(glsl_type::int_type);
emit(MOV(itemp, temp)); /* float to int */
/* Clamp the upper end of the range to +7/16.
@@ -1691,7 +1691,7 @@ fs_visitor::nir_emit_texture(nir_tex_instr *instr)
brw_mark_surface_used(prog_data, max_used);
/* Emit code to evaluate the actual indexing expression */
- sampler_reg = fs_reg(this, glsl_type::uint_type);
+ sampler_reg = vgrf(glsl_type::uint_type);
emit(ADD(sampler_reg, src, fs_reg(sampler)))
->force_writemask_all = true;
break;
diff --git a/src/mesa/drivers/dri/i965/brw_fs_sel_peephole.cpp b/src/mesa/drivers/dri/i965/brw_fs_sel_peephole.cpp
index f18589cbe99..ee485fa8dda 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_sel_peephole.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_sel_peephole.cpp
@@ -196,7 +196,7 @@ fs_visitor::opt_peephole_sel()
*/
fs_reg src0(then_mov[i]->src[0]);
if (src0.file == IMM) {
- src0 = fs_reg(this, glsl_type::float_type);
+ src0 = vgrf(glsl_type::float_type);
src0.type = then_mov[i]->src[0].type;
mov_imm_inst[i] = MOV(src0, then_mov[i]->src[0]);
}
diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
index 38cf29fb753..9805b552f87 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
@@ -98,7 +98,7 @@ fs_visitor::visit(ir_variable *ir)
} else if (!strcmp(ir->name, "gl_FrontFacing")) {
reg = emit_frontfacing_interpolation();
} else {
- reg = new(this->mem_ctx) fs_reg(this, ir->type);
+ reg = new(this->mem_ctx) fs_reg(vgrf(ir->type));
emit_general_interpolation(*reg, ir->name, ir->type,
(glsl_interp_qualifier) ir->data.interpolation,
ir->data.location, ir->data.centroid,
@@ -108,7 +108,7 @@ fs_visitor::visit(ir_variable *ir)
hash_table_insert(this->variable_ht, reg, ir);
return;
} else if (ir->data.mode == ir_var_shader_out) {
- reg = new(this->mem_ctx) fs_reg(this, ir->type);
+ reg = new(this->mem_ctx) fs_reg(vgrf(ir->type));
if (stage == MESA_SHADER_VERTEX) {
int vector_elements =
@@ -210,7 +210,7 @@ fs_visitor::visit(ir_variable *ir)
}
if (!reg)
- reg = new(this->mem_ctx) fs_reg(this, ir->type);
+ reg = new(this->mem_ctx) fs_reg(vgrf(ir->type));
hash_table_insert(this->variable_ht, reg, ir);
}
@@ -279,7 +279,7 @@ fs_visitor::visit(ir_dereference_array *ir)
ir->array_index->accept(this);
fs_reg index_reg;
- index_reg = fs_reg(this, glsl_type::int_type);
+ index_reg = vgrf(glsl_type::int_type);
emit(BRW_OPCODE_MUL, index_reg, this->result, fs_reg(element_size));
if (src.reladdr) {
@@ -298,9 +298,9 @@ fs_visitor::emit_lrp(const fs_reg &dst, const fs_reg &x, const fs_reg &y,
{
if (brw->gen < 6) {
/* We can't use the LRP instruction. Emit x*(1-a) + y*a. */
- fs_reg y_times_a = fs_reg(this, glsl_type::float_type);
- fs_reg one_minus_a = fs_reg(this, glsl_type::float_type);
- fs_reg x_times_one_minus_a = fs_reg(this, glsl_type::float_type);
+ fs_reg y_times_a = vgrf(glsl_type::float_type);
+ fs_reg one_minus_a = vgrf(glsl_type::float_type);
+ fs_reg x_times_one_minus_a = vgrf(glsl_type::float_type);
emit(MUL(y_times_a, y, a));
@@ -408,11 +408,11 @@ fs_visitor::try_emit_line(ir_expression *ir)
non_const_mul->accept(this);
fs_reg src1 = this->result;
- fs_reg src0 = fs_reg(this, ir->type);
+ fs_reg src0 = vgrf(ir->type);
emit(BRW_OPCODE_MOV, src0,
fs_reg((uint8_t)mul_operand_vf, 0, 0, (uint8_t)add_operand_vf));
- this->result = fs_reg(this, ir->type);
+ this->result = vgrf(ir->type);
emit(BRW_OPCODE_LINE, this->result, src0, src1);
return true;
}
@@ -477,7 +477,7 @@ fs_visitor::try_emit_mad(ir_expression *ir)
if (mul_abs)
src2.negate = false;
- this->result = fs_reg(this, ir->type);
+ this->result = vgrf(ir->type);
emit(BRW_OPCODE_MAD, this->result, src0, src1, src2);
return true;
@@ -520,13 +520,13 @@ fs_visitor::emit_interpolate_expression(ir_expression *ir)
/* 1. collect interpolation factors */
- fs_reg dst_x = fs_reg(this, glsl_type::get_instance(ir->type->base_type, 2, 1));
+ fs_reg dst_x = vgrf(glsl_type::get_instance(ir->type->base_type, 2, 1));
fs_reg dst_y = offset(dst_x, 1);
/* for most messages, we need one reg of ignored data; the hardware requires mlen==1
* even when there is no payload. in the per-slot offset case, we'll replace this with
* the proper source data. */
- fs_reg src = fs_reg(this, glsl_type::float_type);
+ fs_reg src = vgrf(glsl_type::float_type);
int mlen = 1; /* one reg unless overriden */
int reg_width = dispatch_width / 8;
fs_inst *inst;
@@ -555,10 +555,10 @@ fs_visitor::emit_interpolate_expression(ir_expression *ir)
} else {
/* pack the operands: hw wants offsets as 4 bit signed ints */
ir->operands[1]->accept(this);
- src = fs_reg(this, glsl_type::ivec2_type);
+ src = vgrf(glsl_type::ivec2_type);
fs_reg src2 = src;
for (int i = 0; i < 2; i++) {
- fs_reg temp = fs_reg(this, glsl_type::float_type);
+ fs_reg temp = vgrf(glsl_type::float_type);
emit(MUL(temp, this->result, fs_reg(16.0f)));
emit(MOV(src2, temp)); /* float to int */
@@ -600,7 +600,7 @@ fs_visitor::emit_interpolate_expression(ir_expression *ir)
/* 2. emit linterp */
- fs_reg res(this, ir->type);
+ fs_reg res = vgrf(ir->type);
this->result = res;
for (int i = 0; i < ir->type->vector_elements; i++) {
@@ -642,7 +642,7 @@ fs_visitor::visit(ir_expression *ir)
emit_bool_to_cond_code(ir->operands[0]);
- this->result = fs_reg(this, ir->type);
+ this->result = vgrf(ir->type);
inst = emit(SEL(this->result, op[1], op[2]));
inst->predicate = BRW_PREDICATE_NORMAL;
return;
@@ -680,7 +680,7 @@ fs_visitor::visit(ir_expression *ir)
/* Storage for our result. If our result goes into an assignment, it will
* just get copy-propagated out, so no worries.
*/
- this->result = fs_reg(this, ir->type);
+ this->result = vgrf(ir->type);
switch (ir->operation) {
case ir_unop_logic_not:
@@ -984,7 +984,7 @@ fs_visitor::visit(ir_expression *ir)
emit(RNDZ(this->result, op[0]));
break;
case ir_unop_ceil: {
- fs_reg tmp = fs_reg(this, ir->type);
+ fs_reg tmp = vgrf(ir->type);
op[0].negate = !op[0].negate;
emit(RNDD(tmp, op[0]));
tmp.negate = true;
@@ -1037,7 +1037,7 @@ fs_visitor::visit(ir_expression *ir)
emit(CBIT(this->result, op[0]));
break;
case ir_unop_find_msb:
- temp = fs_reg(this, glsl_type::uint_type);
+ temp = vgrf(glsl_type::uint_type);
emit(FBH(temp, op[0]));
/* FBH counts from the MSB side, while GLSL's findMSB() wants the count
@@ -1121,7 +1121,7 @@ fs_visitor::visit(ir_expression *ir)
* per-channel and add the base UBO index; the generator will select
* a value from any live channel.
*/
- surf_index = fs_reg(this, glsl_type::uint_type);
+ surf_index = vgrf(glsl_type::uint_type);
emit(ADD(surf_index, op[0],
fs_reg(stage_prog_data->binding_table.ubo_start)))
->force_writemask_all = true;
@@ -1135,7 +1135,7 @@ fs_visitor::visit(ir_expression *ir)
}
if (const_offset) {
- fs_reg packed_consts = fs_reg(this, glsl_type::float_type);
+ fs_reg packed_consts = vgrf(glsl_type::float_type);
packed_consts.type = result.type;
fs_reg const_offset_reg = fs_reg(const_offset->value.u[0] & ~15);
@@ -1163,7 +1163,7 @@ fs_visitor::visit(ir_expression *ir)
}
} else {
/* Turn the byte offset into a dword offset. */
- fs_reg base_offset = fs_reg(this, glsl_type::int_type);
+ fs_reg base_offset = vgrf(glsl_type::int_type);
emit(SHR(base_offset, op[1], fs_reg(2)));
for (int i = 0; i < ir->type->vector_elements; i++) {
@@ -1655,7 +1655,7 @@ fs_visitor::emit_texture_gen7(ir_texture_opcode op, fs_reg dst,
fs_reg *sources = ralloc_array(mem_ctx, fs_reg, MAX_SAMPLER_MESSAGE_SIZE);
for (int i = 0; i < MAX_SAMPLER_MESSAGE_SIZE; i++) {
- sources[i] = fs_reg(this, glsl_type::float_type);
+ sources[i] = vgrf(glsl_type::float_type);
}
int length = 0;
@@ -1956,7 +1956,7 @@ fs_visitor::rescale_texcoord(fs_reg coordinate, int coord_components,
* for clamping, but we don't care enough to make a new
* parameter type, so just invert back.
*/
- fs_reg limit = fs_reg(this, glsl_type::float_type);
+ fs_reg limit = vgrf(glsl_type::float_type);
emit(MOV(limit, i == 0 ? scale_x : scale_y));
emit(SHADER_OPCODE_RCP, limit, limit);
@@ -1987,12 +1987,12 @@ fs_visitor::emit_mcs_fetch(fs_reg coordinate, int components, fs_reg sampler)
int reg_width = dispatch_width / 8;
fs_reg payload = fs_reg(GRF, virtual_grf_alloc(components * reg_width),
BRW_REGISTER_TYPE_F);
- fs_reg dest = fs_reg(this, glsl_type::uvec4_type);
+ fs_reg dest = vgrf(glsl_type::uvec4_type);
fs_reg *sources = ralloc_array(mem_ctx, fs_reg, components);
/* parameters are: u, v, r; missing parameters are treated as zero */
for (int i = 0; i < components; i++) {
- sources[i] = fs_reg(this, glsl_type::float_type);
+ sources[i] = vgrf(glsl_type::float_type);
emit(MOV(retype(sources[i], BRW_REGISTER_TYPE_D), coordinate));
coordinate = offset(coordinate, 1);
}
@@ -2036,7 +2036,7 @@ fs_visitor::emit_texture(ir_texture_opcode op,
int swiz = GET_SWZ(tex->swizzles[sampler], gather_component);
if (swiz == SWIZZLE_ZERO || swiz == SWIZZLE_ONE) {
- fs_reg res = fs_reg(this, glsl_type::vec4_type);
+ fs_reg res = vgrf(glsl_type::vec4_type);
this->result = res;
for (int i=0; i<4; i++) {
@@ -2058,7 +2058,7 @@ fs_visitor::emit_texture(ir_texture_opcode op,
/* Writemasking doesn't eliminate channels on SIMD8 texture
* samples, so don't worry about them.
*/
- fs_reg dst(this, glsl_type::get_instance(dest_type->base_type, 4, 1));
+ fs_reg dst = vgrf(glsl_type::get_instance(dest_type->base_type, 4, 1));
if (brw->gen >= 7) {
inst = emit_texture_gen7(op, dst, coordinate, coord_components,
@@ -2093,7 +2093,7 @@ fs_visitor::emit_texture(ir_texture_opcode op,
/* fixup #layers for cube map arrays */
if (op == ir_txs && is_cube_array) {
fs_reg depth = offset(dst, 2);
- fs_reg fixed_depth = fs_reg(this, glsl_type::int_type);
+ fs_reg fixed_depth = vgrf(glsl_type::int_type);
emit_math(SHADER_OPCODE_INT_QUOTIENT, fixed_depth, depth, fs_reg(6));
fs_reg *fixed_payload = ralloc_array(mem_ctx, fs_reg, inst->regs_written);
@@ -2142,7 +2142,7 @@ fs_visitor::visit(ir_texture *ir)
/* Emit code to evaluate the actual indexing expression */
nonconst_sampler_index->accept(this);
- fs_reg temp(this, glsl_type::uint_type);
+ fs_reg temp = vgrf(glsl_type::uint_type);
emit(ADD(temp, this->result, fs_reg(sampler)))
->force_writemask_all = true;
sampler_reg = temp;
@@ -2342,7 +2342,7 @@ fs_visitor::swizzle_result(ir_texture_opcode op, int dest_components,
if (dest_components == 1) {
/* Ignore DEPTH_TEXTURE_MODE swizzling. */
} else if (tex->swizzles[sampler] != SWIZZLE_NOOP) {
- fs_reg swizzled_result = fs_reg(this, glsl_type::vec4_type);
+ fs_reg swizzled_result = vgrf(glsl_type::vec4_type);
swizzled_result.type = orig_val.type;
for (int i = 0; i < 4; i++) {
@@ -2374,7 +2374,7 @@ fs_visitor::visit(ir_swizzle *ir)
return;
}
- fs_reg result = fs_reg(this, ir->type);
+ fs_reg result = vgrf(ir->type);
this->result = result;
for (unsigned int i = 0; i < ir->type->vector_elements; i++) {
@@ -2442,7 +2442,7 @@ fs_visitor::visit(ir_constant *ir)
* Make reg constant so that it doesn't get accidentally modified along the
* way. Yes, I actually had this problem. :(
*/
- const fs_reg reg(this, ir->type);
+ const fs_reg reg = vgrf(ir->type);
fs_reg dst_reg = reg;
if (ir->type->is_array()) {
@@ -2536,7 +2536,7 @@ fs_visitor::emit_bool_to_cond_code(ir_rvalue *ir)
case ir_binop_logic_xor:
if (brw->gen <= 5) {
- fs_reg temp = fs_reg(this, ir->type);
+ fs_reg temp = vgrf(ir->type);
emit(XOR(temp, op[0], op[1]));
inst = emit(AND(reg_null_d, temp, fs_reg(1)));
} else {
@@ -2547,7 +2547,7 @@ fs_visitor::emit_bool_to_cond_code(ir_rvalue *ir)
case ir_binop_logic_or:
if (brw->gen <= 5) {
- fs_reg temp = fs_reg(this, ir->type);
+ fs_reg temp = vgrf(ir->type);
emit(OR(temp, op[0], op[1]));
inst = emit(AND(reg_null_d, temp, fs_reg(1)));
} else {
@@ -2558,7 +2558,7 @@ fs_visitor::emit_bool_to_cond_code(ir_rvalue *ir)
case ir_binop_logic_and:
if (brw->gen <= 5) {
- fs_reg temp = fs_reg(this, ir->type);
+ fs_reg temp = vgrf(ir->type);
emit(AND(temp, op[0], op[1]));
inst = emit(AND(reg_null_d, temp, fs_reg(1)));
} else {
@@ -2608,7 +2608,7 @@ fs_visitor::emit_bool_to_cond_code(ir_rvalue *ir)
inst->conditional_mod = BRW_CONDITIONAL_NZ;
/* Select which boolean to return. */
- fs_reg temp(this, expr->operands[1]->type);
+ fs_reg temp = vgrf(expr->operands[1]->type);
inst = emit(SEL(temp, op[1], op[2]));
inst->predicate = BRW_PREDICATE_NORMAL;
@@ -2655,13 +2655,13 @@ fs_visitor::emit_if_gen6(ir_if *ir)
return;
case ir_binop_logic_or:
- temp = fs_reg(this, glsl_type::bool_type);
+ temp = vgrf(glsl_type::bool_type);
emit(OR(temp, op[0], op[1]));
emit(IF(temp, fs_reg(0), BRW_CONDITIONAL_NZ));
return;
case ir_binop_logic_and:
- temp = fs_reg(this, glsl_type::bool_type);
+ temp = vgrf(glsl_type::bool_type);
emit(AND(temp, op[0], op[1]));
emit(IF(temp, fs_reg(0), BRW_CONDITIONAL_NZ));
return;
@@ -2698,7 +2698,7 @@ fs_visitor::emit_if_gen6(ir_if *ir)
inst->conditional_mod = BRW_CONDITIONAL_NZ;
/* Select which boolean to use as the result. */
- fs_reg temp(this, expr->operands[1]->type);
+ fs_reg temp = vgrf(expr->operands[1]->type);
inst = emit(SEL(temp, op[1], op[2]));
inst->predicate = BRW_PREDICATE_NORMAL;
@@ -2779,7 +2779,7 @@ fs_visitor::try_replace_with_sel()
*/
fs_reg src0(then_mov->src[0]);
if (src0.file == IMM) {
- src0 = fs_reg(this, glsl_type::float_type);
+ src0 = vgrf(glsl_type::float_type);
src0.type = then_mov->src[0].type;
emit(MOV(src0, then_mov->src[0]));
}
@@ -2881,13 +2881,13 @@ fs_visitor::visit_atomic_counter_intrinsic(ir_call *ir)
location->data.binding);
/* Calculate the surface offset */
- fs_reg offset(this, glsl_type::uint_type);
+ fs_reg offset = vgrf(glsl_type::uint_type);
ir_dereference_array *deref_array = deref->as_dereference_array();
if (deref_array) {
deref_array->array_index->accept(this);
- fs_reg tmp(this, glsl_type::uint_type);
+ fs_reg tmp = vgrf(glsl_type::uint_type);
emit(MUL(tmp, this->result, fs_reg(ATOMIC_COUNTER_SIZE)));
emit(ADD(offset, tmp, fs_reg(location->data.atomic.offset)));
} else {
@@ -3000,19 +3000,19 @@ fs_visitor::emit_untyped_atomic(unsigned atomic_op, unsigned surf_index,
length++;
/* Set the atomic operation offset. */
- sources[1] = fs_reg(this, glsl_type::uint_type);
+ sources[1] = vgrf(glsl_type::uint_type);
emit(MOV(sources[1], offset));
length++;
/* Set the atomic operation arguments. */
if (src0.file != BAD_FILE) {
- sources[length] = fs_reg(this, glsl_type::uint_type);
+ sources[length] = vgrf(glsl_type::uint_type);
emit(MOV(sources[length], src0));
length++;
}
if (src1.file != BAD_FILE) {
- sources[length] = fs_reg(this, glsl_type::uint_type);
+ sources[length] = vgrf(glsl_type::uint_type);
emit(MOV(sources[length], src1));
length++;
}
@@ -3054,7 +3054,7 @@ fs_visitor::emit_untyped_surface_read(unsigned surf_index, fs_reg dst,
}
/* Set the surface read offset. */
- sources[1] = fs_reg(this, glsl_type::uint_type);
+ sources[1] = vgrf(glsl_type::uint_type);
emit(MOV(sources[1], offset));
int mlen = 1 + reg_width;
@@ -3157,8 +3157,8 @@ void
fs_visitor::emit_interpolation_setup_gen4()
{
this->current_annotation = "compute pixel centers";
- this->pixel_x = fs_reg(this, glsl_type::uint_type);
- this->pixel_y = fs_reg(this, glsl_type::uint_type);
+ this->pixel_x = vgrf(glsl_type::uint_type);
+ this->pixel_y = vgrf(glsl_type::uint_type);
this->pixel_x.type = BRW_REGISTER_TYPE_UW;
this->pixel_y.type = BRW_REGISTER_TYPE_UW;
@@ -3168,14 +3168,14 @@ fs_visitor::emit_interpolation_setup_gen4()
this->current_annotation = "compute pixel deltas from v0";
if (brw->has_pln) {
this->delta_x[BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC] =
- fs_reg(this, glsl_type::vec2_type);
+ vgrf(glsl_type::vec2_type);
this->delta_y[BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC] =
offset(this->delta_x[BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC], 1);
} else {
this->delta_x[BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC] =
- fs_reg(this, glsl_type::float_type);
+ vgrf(glsl_type::float_type);
this->delta_y[BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC] =
- fs_reg(this, glsl_type::float_type);
+ vgrf(glsl_type::float_type);
}
emit(ADD(this->delta_x[BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC],
this->pixel_x, fs_reg(negate(brw_vec1_grf(1, 0)))));
@@ -3186,13 +3186,13 @@ fs_visitor::emit_interpolation_setup_gen4()
/* Compute wpos.w. It's always in our setup, since it's needed to
* interpolate the other attributes.
*/
- this->wpos_w = fs_reg(this, glsl_type::float_type);
+ this->wpos_w = vgrf(glsl_type::float_type);
emit(FS_OPCODE_LINTERP, wpos_w,
this->delta_x[BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC],
this->delta_y[BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC],
interp_reg(VARYING_SLOT_POS, 3));
/* Compute the pixel 1/W value from wpos.w. */
- this->pixel_w = fs_reg(this, glsl_type::float_type);
+ this->pixel_w = vgrf(glsl_type::float_type);
emit_math(SHADER_OPCODE_RCP, this->pixel_w, wpos_w);
this->current_annotation = NULL;
}
@@ -3205,8 +3205,8 @@ fs_visitor::emit_interpolation_setup_gen6()
/* If the pixel centers end up used, the setup is the same as for gen4. */
this->current_annotation = "compute pixel centers";
- fs_reg int_pixel_x = fs_reg(this, glsl_type::uint_type);
- fs_reg int_pixel_y = fs_reg(this, glsl_type::uint_type);
+ fs_reg int_pixel_x = vgrf(glsl_type::uint_type);
+ fs_reg int_pixel_y = vgrf(glsl_type::uint_type);
int_pixel_x.type = BRW_REGISTER_TYPE_UW;
int_pixel_y.type = BRW_REGISTER_TYPE_UW;
emit(ADD(int_pixel_x,
@@ -3220,14 +3220,14 @@ fs_visitor::emit_interpolation_setup_gen6()
* to turn the integer pixel centers into floats for their actual
* use.
*/
- this->pixel_x = fs_reg(this, glsl_type::float_type);
- this->pixel_y = fs_reg(this, glsl_type::float_type);
+ this->pixel_x = vgrf(glsl_type::float_type);
+ this->pixel_y = vgrf(glsl_type::float_type);
emit(MOV(this->pixel_x, int_pixel_x));
emit(MOV(this->pixel_y, int_pixel_y));
this->current_annotation = "compute pos.w";
this->pixel_w = fs_reg(brw_vec8_grf(payload.source_w_reg, 0));
- this->wpos_w = fs_reg(this, glsl_type::float_type);
+ this->wpos_w = vgrf(glsl_type::float_type);
emit_math(SHADER_OPCODE_RCP, this->wpos_w, this->pixel_w);
for (int i = 0; i < BRW_WM_BARYCENTRIC_INTERP_MODE_COUNT; ++i) {
@@ -3458,7 +3458,7 @@ fs_visitor::emit_single_fb_write(fs_reg color0, fs_reg color1,
no16("Missing support for simd16 depth writes on gen6\n");
}
- sources[length] = fs_reg(this, glsl_type::float_type);
+ sources[length] = vgrf(glsl_type::float_type);
if (prog->OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) {
/* Hand over gl_FragDepth. */
assert(this->frag_depth.file != BAD_FILE);
@@ -3472,7 +3472,7 @@ fs_visitor::emit_single_fb_write(fs_reg color0, fs_reg color1,
}
if (payload.dest_depth_reg) {
- sources[length] = fs_reg(this, glsl_type::float_type);
+ sources[length] = vgrf(glsl_type::float_type);
emit(MOV(sources[length],
fs_reg(brw_vec8_grf(payload.dest_depth_reg, 0))));
length++;
@@ -3615,8 +3615,8 @@ void fs_visitor::compute_clip_distance()
current_annotation = "user clip distances";
- this->outputs[VARYING_SLOT_CLIP_DIST0] = fs_reg(this, glsl_type::vec4_type);
- this->outputs[VARYING_SLOT_CLIP_DIST1] = fs_reg(this, glsl_type::vec4_type);
+ this->outputs[VARYING_SLOT_CLIP_DIST0] = vgrf(glsl_type::vec4_type);
+ this->outputs[VARYING_SLOT_CLIP_DIST1] = vgrf(glsl_type::vec4_type);
for (int i = 0; i < key->nr_userclip_plane_consts; i++) {
fs_reg u = userplane[i];
@@ -3802,7 +3802,7 @@ fs_visitor::resolve_ud_negate(fs_reg *reg)
!reg->negate)
return;
- fs_reg temp = fs_reg(this, glsl_type::uint_type);
+ fs_reg temp = vgrf(glsl_type::uint_type);
emit(MOV(temp, *reg));
*reg = temp;
}
@@ -3821,8 +3821,8 @@ fs_visitor::resolve_bool_comparison(ir_rvalue *rvalue, fs_reg *reg)
if (rvalue->type != glsl_type::bool_type)
return;
- fs_reg and_result = fs_reg(this, glsl_type::bool_type);
- fs_reg neg_result = fs_reg(this, glsl_type::bool_type);
+ fs_reg and_result = vgrf(glsl_type::bool_type);
+ fs_reg neg_result = vgrf(glsl_type::bool_type);
emit(AND(and_result, *reg, fs_reg(1)));
emit(MOV(neg_result, negate(and_result)));
*reg = neg_result;