#include "glsl/ir_optimization.h"
#include "glsl/nir/glsl_to_nir.h"
#include "brw_fs.h"
+#include "brw_nir.h"
static void
nir_optimize(nir_shader *nir)
nir_validate_shader(nir);
progress |= nir_opt_constant_folding(nir);
nir_validate_shader(nir);
+ progress |= nir_opt_remove_phis(nir);
+ nir_validate_shader(nir);
} while (progress);
}
void
fs_visitor::emit_nir_code()
{
+ const nir_shader_compiler_options *options =
+ ctx->Const.ShaderCompilerOptions[stage].NirOptions;
+
/* first, lower the GLSL IR shader to NIR */
lower_output_reads(shader->base.ir);
- nir_shader *nir = glsl_to_nir(shader->base.ir, NULL, true);
+ nir_shader *nir = glsl_to_nir(&shader->base, options);
nir_validate_shader(nir);
nir_lower_global_vars_to_local(nir);
nir_lower_var_copies(nir);
nir_validate_shader(nir);
- nir_lower_io(nir);
- nir_validate_shader(nir);
+ /* Get rid of split copies */
+ nir_optimize(nir);
- nir_lower_locals_to_regs(nir);
+ nir_assign_var_locations_scalar_direct_first(nir, &nir->uniforms,
+ &num_direct_uniforms,
+ &nir->num_uniforms);
+ nir_assign_var_locations_scalar(&nir->inputs, &nir->num_inputs);
+ nir_assign_var_locations_scalar(&nir->outputs, &nir->num_outputs);
+
+ nir_lower_io(nir);
nir_validate_shader(nir);
nir_remove_dead_variables(nir);
nir_optimize(nir);
+ nir_lower_locals_to_regs(nir);
+ nir_validate_shader(nir);
+
nir_lower_to_source_mods(nir);
nir_validate_shader(nir);
nir_copy_prop(nir);
nir_validate_shader(nir);
- if (INTEL_DEBUG & DEBUG_WM) {
- fprintf(stderr, "NIR (SSA form) for fragment shader:\n");
+ if (unlikely(debug_enabled)) {
+ fprintf(stderr, "NIR (SSA form) for %s shader:\n", stage_name);
nir_print_shader(nir, stderr);
}
MESA_DEBUG_SOURCE_SHADER_COMPILER,
MESA_DEBUG_TYPE_OTHER,
MESA_DEBUG_SEVERITY_NOTIFICATION,
- "FS NIR shader: %d inst\n",
+ "%s NIR shader: %d inst\n",
+ stage_abbrev,
count_nir_instrs(nir));
}
nir_convert_from_ssa(nir);
nir_validate_shader(nir);
- nir_lower_vec_to_movs(nir);
- nir_validate_shader(nir);
+
+ /* This is the last pass we run before we start emitting stuff. It
+ * determines when we need to insert boolean resolves on Gen <= 5. We
+ * run it last because it stashes data in instr->pass_flags and we don't
+ * want that to be squashed by other NIR passes.
+ */
+ if (brw->gen <= 5)
+ brw_nir_analyze_boolean_resolves(nir);
/* emit the arrays used for inputs and outputs - load/store intrinsics will
* be converted to reads/writes of these arrays
}
if (nir->num_uniforms > 0) {
- nir_uniforms = fs_reg(UNIFORM, 0);
nir_setup_uniforms(nir);
}
nir_emit_impl(overload->impl);
}
- if (INTEL_DEBUG & DEBUG_WM) {
- fprintf(stderr, "NIR (final form) for fragment shader:\n");
+ if (unlikely(debug_enabled)) {
+ fprintf(stderr, "NIR (final form) for %s shader:\n", stage_name);
nir_print_shader(nir, stderr);
}
void
fs_visitor::nir_setup_inputs(nir_shader *shader)
{
- struct hash_entry *entry;
- hash_table_foreach(shader->inputs, entry) {
- nir_variable *var = (nir_variable *) entry->data;
- fs_reg varying = offset(nir_inputs, var->data.driver_location);
+ foreach_list_typed(nir_variable, var, node, &shader->inputs) {
+ enum brw_reg_type type = brw_type_for_base_type(var->type);
+ fs_reg input = offset(nir_inputs, var->data.driver_location);
fs_reg reg;
- if (!strcmp(var->name, "gl_FragCoord")) {
- reg = *emit_fragcoord_interpolation(var->data.pixel_center_integer,
- var->data.origin_upper_left);
- emit_percomp(MOV(varying, reg), 0xF);
- } else if (!strcmp(var->name, "gl_FrontFacing")) {
- reg = *emit_frontfacing_interpolation();
- emit(MOV(retype(varying, BRW_REGISTER_TYPE_UD), reg));
- } else {
- emit_general_interpolation(varying, var->name, var->type,
- (glsl_interp_qualifier) var->data.interpolation,
- var->data.location, var->data.centroid,
- var->data.sample);
+ switch (stage) {
+ case MESA_SHADER_VERTEX: {
+ /* Our ATTR file is indexed by VERT_ATTRIB_*, which is the value
+ * stored in nir_variable::location.
+ *
+ * However, NIR's load_input intrinsics use a different index - an
+ * offset into a single contiguous array containing all inputs.
+ * This index corresponds to the nir_variable::driver_location field.
+ *
+ * So, we need to copy from fs_reg(ATTR, var->location) to
+ * offset(nir_inputs, var->data.driver_location).
+ */
+ unsigned components = var->type->without_array()->components();
+ unsigned array_length = var->type->is_array() ? var->type->length : 1;
+ for (unsigned i = 0; i < array_length; i++) {
+ for (unsigned j = 0; j < components; j++) {
+ emit(MOV(retype(offset(input, components * i + j), type),
+ offset(fs_reg(ATTR, var->data.location + i, type), j)));
+ }
+ }
+ break;
+ }
+ case MESA_SHADER_GEOMETRY:
+ case MESA_SHADER_COMPUTE:
+ unreachable("fs_visitor not used for these stages yet.");
+ break;
+ case MESA_SHADER_FRAGMENT:
+ if (var->data.location == VARYING_SLOT_POS) {
+ reg = *emit_fragcoord_interpolation(var->data.pixel_center_integer,
+ var->data.origin_upper_left);
+ emit_percomp(MOV(input, reg), 0xF);
+ } else {
+ emit_general_interpolation(input, var->name, var->type,
+ (glsl_interp_qualifier) var->data.interpolation,
+ var->data.location, var->data.centroid,
+ var->data.sample);
+ }
+ break;
}
}
}
{
brw_wm_prog_key *key = (brw_wm_prog_key*) this->key;
- struct hash_entry *entry;
- hash_table_foreach(shader->outputs, entry) {
- nir_variable *var = (nir_variable *) entry->data;
+ foreach_list_typed(nir_variable, var, node, &shader->outputs) {
fs_reg reg = offset(nir_outputs, var->data.driver_location);
- if (var->data.index > 0) {
+ int vector_elements =
+ var->type->is_array() ? var->type->fields.array->vector_elements
+ : var->type->vector_elements;
+
+ if (stage == MESA_SHADER_VERTEX) {
+ for (int i = 0; i < ALIGN(type_size(var->type), 4) / 4; i++) {
+ int output = var->data.location + i;
+ this->outputs[output] = offset(reg, 4 * i);
+ this->output_components[output] = vector_elements;
+ }
+ } else if (var->data.index > 0) {
assert(var->data.location == FRAG_RESULT_DATA0);
assert(var->data.index == 1);
this->dual_src_output = reg;
assert(var->data.location >= FRAG_RESULT_DATA0 &&
var->data.location < FRAG_RESULT_DATA0 + BRW_MAX_DRAW_BUFFERS);
- int vector_elements =
- var->type->is_array() ? var->type->fields.array->vector_elements
- : var->type->vector_elements;
-
/* General color output. */
for (unsigned int i = 0; i < MAX2(1, var->type->length); i++) {
int output = var->data.location - FRAG_RESULT_DATA0 + i;
fs_visitor::nir_setup_uniforms(nir_shader *shader)
{
uniforms = shader->num_uniforms;
- param_size[0] = shader->num_uniforms;
+
+ /* We split the uniform register file in half. The first half is
+ * entirely direct uniforms. The second half is indirect.
+ */
+ param_size[0] = num_direct_uniforms;
+ if (shader->num_uniforms > num_direct_uniforms)
+ param_size[num_direct_uniforms] = shader->num_uniforms - num_direct_uniforms;
if (dispatch_width != 8)
return;
- struct hash_entry *entry;
- hash_table_foreach(shader->uniforms, entry) {
- nir_variable *var = (nir_variable *) entry->data;
-
+ foreach_list_typed(nir_variable, var, node, &shader->uniforms) {
/* UBO's and atomics don't take up space in the uniform file */
if (var->interface_type != NULL || var->type->contains_atomic())
nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);
switch (intrin->intrinsic) {
+ case nir_intrinsic_load_vertex_id:
+ unreachable("should be lowered by lower_vertex_id().");
+
+ case nir_intrinsic_load_vertex_id_zero_base:
+ assert(v->stage == MESA_SHADER_VERTEX);
+ reg = &v->nir_system_values[SYSTEM_VALUE_VERTEX_ID_ZERO_BASE];
+ if (reg->file == BAD_FILE)
+ *reg = *v->emit_vs_system_value(SYSTEM_VALUE_VERTEX_ID_ZERO_BASE);
+ break;
+
+ case nir_intrinsic_load_base_vertex:
+ assert(v->stage == MESA_SHADER_VERTEX);
+ reg = &v->nir_system_values[SYSTEM_VALUE_BASE_VERTEX];
+ if (reg->file == BAD_FILE)
+ *reg = *v->emit_vs_system_value(SYSTEM_VALUE_BASE_VERTEX);
+ break;
+
+ case nir_intrinsic_load_instance_id:
+ assert(v->stage == MESA_SHADER_VERTEX);
+ reg = &v->nir_system_values[SYSTEM_VALUE_INSTANCE_ID];
+ if (reg->file == BAD_FILE)
+ *reg = *v->emit_vs_system_value(SYSTEM_VALUE_INSTANCE_ID);
+ break;
+
case nir_intrinsic_load_sample_pos:
assert(v->stage == MESA_SHADER_FRAGMENT);
reg = &v->nir_system_values[SYSTEM_VALUE_SAMPLE_POS];
void
fs_visitor::nir_emit_cf_list(exec_list *list)
{
+ exec_list_validate(list);
foreach_list_typed(nir_cf_node, node, node, list) {
switch (node->type) {
case nir_cf_node_if:
/* first, put the condition into f0 */
fs_inst *inst = emit(MOV(reg_null_d,
retype(get_nir_src(if_stmt->condition),
- BRW_REGISTER_TYPE_UD)));
+ BRW_REGISTER_TYPE_D)));
inst->conditional_mod = BRW_CONDITIONAL_NZ;
emit(IF(BRW_PREDICATE_NORMAL));
brw_type_for_nir_type(nir_alu_type type)
{
switch (type) {
- case nir_type_bool:
case nir_type_unsigned:
return BRW_REGISTER_TYPE_UD;
+ case nir_type_bool:
case nir_type_int:
return BRW_REGISTER_TYPE_D;
case nir_type_float:
return BRW_REGISTER_TYPE_F;
}
+bool
+fs_visitor::optimize_frontfacing_ternary(nir_alu_instr *instr,
+ const fs_reg &result)
+{
+ if (instr->src[0].src.is_ssa ||
+ !instr->src[0].src.reg.reg ||
+ !instr->src[0].src.reg.reg->parent_instr)
+ return false;
+
+ if (instr->src[0].src.reg.reg->parent_instr->type !=
+ nir_instr_type_intrinsic)
+ return false;
+
+ nir_intrinsic_instr *src0 =
+ nir_instr_as_intrinsic(instr->src[0].src.reg.reg->parent_instr);
+
+ if (src0->intrinsic != nir_intrinsic_load_front_face)
+ return false;
+
+ nir_const_value *value1 = nir_src_as_const_value(instr->src[1].src);
+ if (!value1 || fabsf(value1->f[0]) != 1.0f)
+ return false;
+
+ nir_const_value *value2 = nir_src_as_const_value(instr->src[2].src);
+ if (!value2 || fabsf(value2->f[0]) != 1.0f)
+ return false;
+
+ fs_reg tmp = vgrf(glsl_type::int_type);
+
+ if (brw->gen >= 6) {
+ /* Bit 15 of g0.0 is 0 if the polygon is front facing. */
+ fs_reg g0 = fs_reg(retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_W));
+
+ /* For (gl_FrontFacing ? 1.0 : -1.0), emit:
+ *
+ * or(8) tmp.1<2>W g0.0<0,1,0>W 0x00003f80W
+ * and(8) dst<1>D tmp<8,8,1>D 0xbf800000D
+ *
+ * and negate g0.0<0,1,0>W for (gl_FrontFacing ? -1.0 : 1.0).
+ *
+ * This negation looks like it's safe in practice, because bits 0:4 will
+ * surely be TRIANGLES
+ */
+
+ if (value1->f[0] == -1.0f) {
+ g0.negate = true;
+ }
+
+ tmp.type = BRW_REGISTER_TYPE_W;
+ tmp.subreg_offset = 2;
+ tmp.stride = 2;
+
+ fs_inst *or_inst = emit(OR(tmp, g0, fs_reg(0x3f80)));
+ or_inst->src[1].type = BRW_REGISTER_TYPE_UW;
+
+ tmp.type = BRW_REGISTER_TYPE_D;
+ tmp.subreg_offset = 0;
+ tmp.stride = 1;
+ } else {
+ /* Bit 31 of g1.6 is 0 if the polygon is front facing. */
+ fs_reg g1_6 = fs_reg(retype(brw_vec1_grf(1, 6), BRW_REGISTER_TYPE_D));
+
+ /* For (gl_FrontFacing ? 1.0 : -1.0), emit:
+ *
+ * or(8) tmp<1>D g1.6<0,1,0>D 0x3f800000D
+ * and(8) dst<1>D tmp<8,8,1>D 0xbf800000D
+ *
+ * and negate g1.6<0,1,0>D for (gl_FrontFacing ? -1.0 : 1.0).
+ *
+ * This negation looks like it's safe in practice, because bits 0:4 will
+ * surely be TRIANGLES
+ */
+
+ if (value1->f[0] == -1.0f) {
+ g1_6.negate = true;
+ }
+
+ emit(OR(tmp, g1_6, fs_reg(0x3f800000)));
+ }
+ emit(AND(retype(result, BRW_REGISTER_TYPE_D), tmp, fs_reg(0xbf800000)));
+
+ return true;
+}
+
void
fs_visitor::nir_emit_alu(nir_alu_instr *instr)
{
struct brw_wm_prog_key *fs_key = (struct brw_wm_prog_key *) this->key;
fs_inst *inst;
- fs_reg op[3];
fs_reg result = get_nir_dest(instr->dest.dest);
result.type = brw_type_for_nir_type(nir_op_infos[instr->op].output_type);
- for (unsigned i = 0; i < nir_op_infos[instr->op].num_inputs; i++)
- op[i] = get_nir_alu_src(instr, i);
+ fs_reg op[4];
+ for (unsigned i = 0; i < nir_op_infos[instr->op].num_inputs; i++) {
+ op[i] = get_nir_src(instr->src[i].src);
+ op[i].type = brw_type_for_nir_type(nir_op_infos[instr->op].input_types[i]);
+ op[i].abs = instr->src[i].abs;
+ op[i].negate = instr->src[i].negate;
+ }
+
+ /* We get a bunch of mov's out of the from_ssa pass and they may still
+ * be vectorized. We'll handle them as a special-case. We'll also
+ * handle vecN here because it's basically the same thing.
+ */
+ switch (instr->op) {
+ case nir_op_imov:
+ case nir_op_fmov:
+ case nir_op_vec2:
+ case nir_op_vec3:
+ case nir_op_vec4: {
+ fs_reg temp = result;
+ bool need_extra_copy = false;
+ for (unsigned i = 0; i < nir_op_infos[instr->op].num_inputs; i++) {
+ if (!instr->src[i].src.is_ssa &&
+ instr->dest.dest.reg.reg == instr->src[i].src.reg.reg) {
+ need_extra_copy = true;
+ temp = retype(vgrf(4), result.type);
+ break;
+ }
+ }
+
+ for (unsigned i = 0; i < 4; i++) {
+ if (!(instr->dest.write_mask & (1 << i)))
+ continue;
+
+ if (instr->op == nir_op_imov || instr->op == nir_op_fmov) {
+ inst = emit(MOV(offset(temp, i),
+ offset(op[0], instr->src[0].swizzle[i])));
+ } else {
+ inst = emit(MOV(offset(temp, i),
+ offset(op[i], instr->src[i].swizzle[0])));
+ }
+ inst->saturate = instr->dest.saturate;
+ }
+
+ /* In this case the source and destination registers were the same,
+ * so we need to insert an extra set of moves in order to deal with
+ * any swizzling.
+ */
+ if (need_extra_copy) {
+ for (unsigned i = 0; i < 4; i++) {
+ if (!(instr->dest.write_mask & (1 << i)))
+ continue;
+ emit(MOV(offset(result, i), offset(temp, i)));
+ }
+ }
+ return;
+ }
+ default:
+ break;
+ }
+
+ /* At this point, we have dealt with any instruction that operates on
+ * more than a single channel. Therefore, we can just adjust the source
+ * and destination registers for that channel and emit the instruction.
+ */
+ unsigned channel = 0;
if (nir_op_infos[instr->op].output_size == 0) {
- /* We've already scalarized, so we know that we only have one
- * channel. The only question is which channel.
+ /* Since NIR is doing the scalarizing for us, we should only ever see
+ * vectorized operations with a single channel.
*/
assert(_mesa_bitcount(instr->dest.write_mask) == 1);
- unsigned off = ffs(instr->dest.write_mask) - 1;
- result = offset(result, off);
+ channel = ffs(instr->dest.write_mask) - 1;
+
+ result = offset(result, channel);
+ }
- for (unsigned i = 0; i < nir_op_infos[instr->op].num_inputs; i++)
- op[i] = offset(op[i], off);
+ for (unsigned i = 0; i < nir_op_infos[instr->op].num_inputs; i++) {
+ assert(nir_op_infos[instr->op].input_sizes[i] < 2);
+ op[i] = offset(op[i], instr->src[i].swizzle[channel]);
}
switch (instr->op) {
- case nir_op_fmov:
case nir_op_i2f:
case nir_op_u2f:
inst = emit(MOV(result, op[0]));
inst->saturate = instr->dest.saturate;
break;
- case nir_op_imov:
case nir_op_f2i:
case nir_op_f2u:
emit(MOV(result, op[0]));
break;
case nir_op_imul: {
- /* TODO put in the 16-bit constant optimization once we have SSA */
+ if (brw->gen >= 8) {
+ emit(MUL(result, op[0], op[1]));
+ break;
+ } else {
+ nir_const_value *value0 = nir_src_as_const_value(instr->src[0].src);
+ nir_const_value *value1 = nir_src_as_const_value(instr->src[1].src);
+
+ if (value0 && value0->u[0] < (1 << 16)) {
+ if (brw->gen < 7) {
+ emit(MUL(result, op[0], op[1]));
+ } else {
+ emit(MUL(result, op[1], op[0]));
+ }
+ break;
+ } else if (value1 && value1->u[0] < (1 << 16)) {
+ if (brw->gen < 7) {
+ emit(MUL(result, op[1], op[0]));
+ } else {
+ emit(MUL(result, op[0], op[1]));
+ }
+ break;
+ }
+ }
if (brw->gen >= 7)
no16("SIMD16 explicit accumulator operands unsupported\n");
break;
case nir_op_inot:
+ if (brw->gen >= 8) {
+ resolve_source_modifiers(&op[0]);
+ }
emit(NOT(result, op[0]));
break;
case nir_op_ixor:
+ if (brw->gen >= 8) {
+ resolve_source_modifiers(&op[0]);
+ resolve_source_modifiers(&op[1]);
+ }
emit(XOR(result, op[0], op[1]));
break;
case nir_op_ior:
+ if (brw->gen >= 8) {
+ resolve_source_modifiers(&op[0]);
+ resolve_source_modifiers(&op[1]);
+ }
emit(OR(result, op[0], op[1]));
break;
case nir_op_iand:
+ if (brw->gen >= 8) {
+ resolve_source_modifiers(&op[0]);
+ resolve_source_modifiers(&op[1]);
+ }
emit(AND(result, op[0], op[1]));
break;
case nir_op_fnoise4_4:
unreachable("not reached: should be handled by lower_noise");
- case nir_op_vec2:
- case nir_op_vec3:
- case nir_op_vec4:
- unreachable("not reached: should be handled by lower_quadop_vector");
-
case nir_op_ldexp:
unreachable("not reached: should be handled by ldexp_to_arith()");
} else {
emit(CMP(reg_null_d, op[0], op[1], BRW_CONDITIONAL_L));
inst = emit(SEL(result, op[0], op[1]));
+ inst->predicate = BRW_PREDICATE_NORMAL;
}
inst->saturate = instr->dest.saturate;
break;
} else {
emit(CMP(reg_null_d, op[0], op[1], BRW_CONDITIONAL_GE));
inst = emit(SEL(result, op[0], op[1]));
+ inst->predicate = BRW_PREDICATE_NORMAL;
}
inst->saturate = instr->dest.saturate;
break;
break;
case nir_op_fpow:
- inst = emit(SHADER_OPCODE_POW, result, op[0], op[1]);
+ inst = emit_math(SHADER_OPCODE_POW, result, op[0], op[1]);
inst->saturate = instr->dest.saturate;
break;
break;
case nir_op_ffma:
- emit(MAD(result, op[2], op[1], op[0]));
+ inst = emit(MAD(result, op[2], op[1], op[0]));
+ inst->saturate = instr->dest.saturate;
break;
case nir_op_flrp:
- /* TODO emulate for gen < 6 */
- emit(LRP(result, op[2], op[1], op[0]));
+ inst = emit_lrp(result, op[0], op[1], op[2]);
+ inst->saturate = instr->dest.saturate;
break;
case nir_op_bcsel:
+ if (optimize_frontfacing_ternary(instr, result))
+ return;
+
emit(CMP(reg_null_d, op[0], fs_reg(0), BRW_CONDITIONAL_NZ));
inst = emit(SEL(result, op[1], op[2]));
inst->predicate = BRW_PREDICATE_NORMAL;
default:
unreachable("unhandled instruction");
}
+
+ /* If we need to do a boolean resolve, replace the result with -(x & 1)
+ * to sign extend the low bit to 0/~0
+ */
+ if (brw->gen <= 5 &&
+ (instr->instr.pass_flags & BRW_NIR_BOOLEAN_MASK) == BRW_NIR_BOOLEAN_NEEDS_RESOLVE) {
+ fs_reg masked = vgrf(glsl_type::int_type);
+ emit(AND(masked, result, fs_reg(1)));
+ masked.negate = true;
+ emit(MOV(retype(result, BRW_REGISTER_TYPE_D), masked));
+ }
}
fs_reg
}
}
-fs_reg
-fs_visitor::get_nir_alu_src(nir_alu_instr *instr, unsigned src)
-{
- fs_reg reg = get_nir_src(instr->src[src].src);
-
- reg.type = brw_type_for_nir_type(nir_op_infos[instr->op].input_types[src]);
- reg.abs = instr->src[src].abs;
- reg.negate = instr->src[src].negate;
-
- bool needs_swizzle = false;
- unsigned num_components = 0;
- for (unsigned i = 0; i < 4; i++) {
- if (!nir_alu_instr_channel_used(instr, src, i))
- continue;
-
- if (instr->src[src].swizzle[i] != i)
- needs_swizzle = true;
-
- num_components = i + 1;
- }
-
- if (needs_swizzle) {
- /* resolve the swizzle through MOV's */
- fs_reg new_reg = vgrf(num_components);
- new_reg.type = reg.type;
-
- for (unsigned i = 0; i < 4; i++) {
- if (!nir_alu_instr_channel_used(instr, src, i))
- continue;
-
- emit(MOV(offset(new_reg, i),
- offset(reg, instr->src[src].swizzle[i])));
- }
-
- return new_reg;
- }
-
- return reg;
-}
-
fs_reg
fs_visitor::get_nir_dest(nir_dest dest)
{
bool has_indirect = false;
switch (instr->intrinsic) {
- case nir_intrinsic_discard: {
+ case nir_intrinsic_discard:
+ case nir_intrinsic_discard_if: {
/* We track our discarded pixels in f0.1. By predicating on it, we can
- * update just the flag bits that aren't yet discarded. By emitting a
- * CMP of g0 != g0, all our currently executing channels will get turned
- * off.
+ * update just the flag bits that aren't yet discarded. If there's no
+ * condition, we emit a CMP of g0 != g0, so all currently executing
+ * channels will get turned off.
*/
- fs_reg some_reg = fs_reg(retype(brw_vec8_grf(0, 0),
- BRW_REGISTER_TYPE_UW));
- fs_inst *cmp = emit(CMP(reg_null_f, some_reg, some_reg,
- BRW_CONDITIONAL_NZ));
+ fs_inst *cmp;
+ if (instr->intrinsic == nir_intrinsic_discard_if) {
+ cmp = emit(CMP(reg_null_f, get_nir_src(instr->src[0]),
+ fs_reg(0), BRW_CONDITIONAL_Z));
+ } else {
+ fs_reg some_reg = fs_reg(retype(brw_vec8_grf(0, 0),
+ BRW_REGISTER_TYPE_UW));
+ cmp = emit(CMP(reg_null_f, some_reg, some_reg, BRW_CONDITIONAL_NZ));
+ }
cmp->predicate = BRW_PREDICATE_NORMAL;
cmp->flag_subreg = 1;
if (brw->gen >= 6) {
- /* For performance, after a discard, jump to the end of the shader.
- * Only jump if all relevant channels have been discarded.
- */
- fs_inst *discard_jump = emit(FS_OPCODE_DISCARD_JUMP);
- discard_jump->flag_subreg = 1;
-
- discard_jump->predicate = (dispatch_width == 8)
- ? BRW_PREDICATE_ALIGN1_ANY8H
- : BRW_PREDICATE_ALIGN1_ANY16H;
- discard_jump->predicate_inverse = true;
+ emit_discard_jump();
}
-
break;
}
}
case nir_intrinsic_load_front_face:
- assert(!"TODO");
+ emit(MOV(retype(dest, BRW_REGISTER_TYPE_D),
+ *emit_frontfacing_interpolation()));
+ break;
+
+ case nir_intrinsic_load_vertex_id:
+ unreachable("should be lowered by lower_vertex_id()");
+
+ case nir_intrinsic_load_vertex_id_zero_base: {
+ fs_reg vertex_id = nir_system_values[SYSTEM_VALUE_VERTEX_ID_ZERO_BASE];
+ assert(vertex_id.file != BAD_FILE);
+ dest.type = vertex_id.type;
+ emit(MOV(dest, vertex_id));
+ break;
+ }
+
+ case nir_intrinsic_load_base_vertex: {
+ fs_reg base_vertex = nir_system_values[SYSTEM_VALUE_BASE_VERTEX];
+ assert(base_vertex.file != BAD_FILE);
+ dest.type = base_vertex.type;
+ emit(MOV(dest, base_vertex));
+ break;
+ }
+
+ case nir_intrinsic_load_instance_id: {
+ fs_reg instance_id = nir_system_values[SYSTEM_VALUE_INSTANCE_ID];
+ assert(instance_id.file != BAD_FILE);
+ dest.type = instance_id.type;
+ emit(MOV(dest, instance_id));
+ break;
+ }
case nir_intrinsic_load_sample_mask_in: {
fs_reg sample_mask_in = nir_system_values[SYSTEM_VALUE_SAMPLE_MASK_IN];
case nir_intrinsic_load_uniform_indirect:
has_indirect = true;
case nir_intrinsic_load_uniform: {
- unsigned index = 0;
+ unsigned index = instr->const_index[0];
+
+ fs_reg uniform_reg;
+ if (index < num_direct_uniforms) {
+ uniform_reg = fs_reg(UNIFORM, 0);
+ } else {
+ uniform_reg = fs_reg(UNIFORM, num_direct_uniforms);
+ index -= num_direct_uniforms;
+ }
+
for (int i = 0; i < instr->const_index[1]; i++) {
for (unsigned j = 0; j < instr->num_components; j++) {
- fs_reg src = offset(retype(nir_uniforms, dest.type),
- instr->const_index[0] + index);
+ fs_reg src = offset(retype(uniform_reg, dest.type), index);
if (has_indirect)
src.reladdr = new(mem_ctx) fs_reg(get_nir_src(instr->src[0]));
index++;
case nir_intrinsic_load_ubo_indirect:
has_indirect = true;
+ /* fallthrough */
case nir_intrinsic_load_ubo: {
nir_const_value *const_index = nir_src_as_const_value(instr->src[0]);
fs_reg surf_index;
case nir_intrinsic_load_input_indirect:
has_indirect = true;
+ /* fallthrough */
case nir_intrinsic_load_input: {
unsigned index = 0;
for (int i = 0; i < instr->const_index[1]; i++) {
void
fs_visitor::nir_emit_texture(nir_tex_instr *instr)
{
- brw_wm_prog_key *key = (brw_wm_prog_key*) this->key;
unsigned sampler = instr->sampler_index;
fs_reg sampler_reg(sampler);
bool is_cube_array = instr->sampler_dim == GLSL_SAMPLER_DIM_CUBE &&
instr->is_array;
- int lod_components, offset_components = 0;
+ int lod_components = 0, offset_components = 0;
fs_reg coordinate, shadow_comparitor, lod, lod2, sample_index, mcs, offset;
}
if (instr->op == nir_texop_txf_ms) {
- if (brw->gen >= 7 && key->tex.compressed_multisample_layout_mask & (1<<sampler))
+ if (brw->gen >= 7 &&
+ key_tex->compressed_multisample_layout_mask & (1 << sampler)) {
mcs = emit_mcs_fetch(coordinate, instr->coord_components, sampler_reg);
- else
+ } else {
mcs = fs_reg(0u);
+ }
}
for (unsigned i = 0; i < 3; i++) {
emit_texture(op, dest_type, coordinate, instr->coord_components,
shadow_comparitor, lod, lod2, lod_components, sample_index,
- offset, offset_components, mcs, gather_component,
+ offset, mcs, gather_component,
is_cube_array, is_rect, sampler, sampler_reg, texunit);
fs_reg dest = get_nir_dest(instr->dest);