ALU2_ACC(ADDC)
ALU2_ACC(SUBB)
ALU2(MAC)
+ALU1(DIM)
/** Gen4 predicated IF. */
vec4_instruction *
emit(VEC4_OPCODE_PACK_BYTES, dst, bytes);
}
-/**
- * Returns the minimum number of vec4 elements needed to pack a type.
- *
- * For simple types, it will return 1 (a single vec4); for matrices, the
- * number of columns; for array and struct, the sum of the vec4_size of
- * each of its elements; and for sampler and atomic, zero.
- *
- * This method is useful to calculate how much register space is needed to
- * store a particular type.
+/*
+ * Returns the minimum number of vec4 (as_vec4 == true) or dvec4 (as_vec4 ==
+ * false) elements needed to pack a type.
*/
-extern "C" int
-type_size_vec4(const struct glsl_type *type)
+static int
+type_size_xvec4(const struct glsl_type *type, bool as_vec4)
{
unsigned int i;
int size;
case GLSL_TYPE_INT:
case GLSL_TYPE_FLOAT:
case GLSL_TYPE_BOOL:
+ case GLSL_TYPE_DOUBLE:
if (type->is_matrix()) {
- return type->matrix_columns;
+ const glsl_type *col_type = type->column_type();
+ unsigned col_slots =
+ (as_vec4 && col_type->is_dual_slot()) ? 2 : 1;
+ return type->matrix_columns * col_slots;
} else {
- /* Regardless of size of vector, it gets a vec4. This is bad
- * packing for things like floats, but otherwise arrays become a
- * mess. Hopefully a later pass over the code can pack scalars
- * down if appropriate.
- */
- return 1;
+ /* Regardless of size of vector, it gets a vec4. This is bad
+ * packing for things like floats, but otherwise arrays become a
+ * mess. Hopefully a later pass over the code can pack scalars
+ * down if appropriate.
+ */
+ return (as_vec4 && type->is_dual_slot()) ? 2 : 1;
}
case GLSL_TYPE_ARRAY:
assert(type->length > 0);
- return type_size_vec4(type->fields.array) * type->length;
+ return type_size_xvec4(type->fields.array, as_vec4) * type->length;
case GLSL_TYPE_STRUCT:
size = 0;
for (i = 0; i < type->length; i++) {
- size += type_size_vec4(type->fields.structure[i].type);
+ size += type_size_xvec4(type->fields.structure[i].type, as_vec4);
}
return size;
case GLSL_TYPE_SUBROUTINE:
case GLSL_TYPE_IMAGE:
return DIV_ROUND_UP(BRW_IMAGE_PARAM_SIZE, 4);
case GLSL_TYPE_VOID:
- case GLSL_TYPE_DOUBLE:
case GLSL_TYPE_ERROR:
case GLSL_TYPE_INTERFACE:
case GLSL_TYPE_FUNCTION:
return 0;
}
+/**
+ * Returns the minimum number of vec4 elements needed to pack a type.
+ *
+ * For simple types, it will return 1 (a single vec4); for matrices, the
+ * number of columns; for array and struct, the sum of the vec4_size of
+ * each of its elements; and for sampler and atomic, zero.
+ *
+ * This method is useful to calculate how much register space is needed to
+ * store a particular type.
+ */
+extern "C" int
+type_size_vec4(const struct glsl_type *type)
+{
+ return type_size_xvec4(type, true);
+}
+
+/**
+ * Returns the minimum number of dvec4 elements needed to pack a type.
+ *
+ * For simple types, it will return 1 (a single dvec4); for matrices, the
+ * number of columns; for array and struct, the sum of the dvec4_size of
+ * each of its elements; and for sampler and atomic, zero.
+ *
+ * This method is useful to calculate how much register space is needed to
+ * store a particular type.
+ *
+ * Measuring double-precision vertex inputs as dvec4 is required because
+ * ARB_vertex_attrib_64bit states that these uses the same number of locations
+ * than the single-precision version. That is, two consecutives dvec4 would be
+ * located in location "x" and location "x+1", not "x+2".
+ *
+ * In order to map vec4/dvec4 vertex inputs in the proper ATTRs,
+ * remap_vs_attrs() will take in account both the location and also if the
+ * type fits in one or two vec4 slots.
+ */
+extern "C" int
+type_size_dvec4(const struct glsl_type *type)
+{
+ return type_size_xvec4(type, false);
+}
+
src_reg::src_reg(class vec4_visitor *v, const struct glsl_type *type)
{
init();
pull->mlen = 2;
pull->header_size = 1;
} else if (devinfo->gen >= 7) {
- dst_reg grf_offset = dst_reg(this, glsl_type::int_type);
+ dst_reg grf_offset = dst_reg(this, glsl_type::uint_type);
grf_offset.type = offset_reg.type;
uint32_t constant_offset,
src_reg offset_value,
src_reg mcs,
- bool is_cube_array,
uint32_t surface,
src_reg surface_reg,
uint32_t sampler,
/* fixup num layers (z) for cube arrays: hardware returns faces * layers;
* spec requires layers.
*/
- if (op == ir_txs) {
- if (is_cube_array) {
- emit_math(SHADER_OPCODE_INT_QUOTIENT,
- writemask(inst->dst, WRITEMASK_Z),
- src_reg(inst->dst), brw_imm_d(6));
- } else if (devinfo->gen < 7) {
- /* Gen4-6 return 0 instead of 1 for single layer surfaces. */
- emit_minmax(BRW_CONDITIONAL_GE, writemask(inst->dst, WRITEMASK_Z),
- src_reg(inst->dst), brw_imm_d(1));
- }
+ if (op == ir_txs && devinfo->gen < 7) {
+ /* Gen4-6 return 0 instead of 1 for single layer surfaces. */
+ emit_minmax(BRW_CONDITIONAL_GE, writemask(inst->dst, WRITEMASK_Z),
+ src_reg(inst->dst), brw_imm_d(1));
}
if (devinfo->gen == 6 && op == ir_tg4) {
assert(varying < VARYING_SLOT_MAX);
assert(output_reg[varying].type == reg.type);
current_annotation = output_reg_annotation[varying];
- if (output_reg[varying].file != BAD_FILE)
+ if (output_reg[varying].file != BAD_FILE) {
return emit(MOV(reg, src_reg(output_reg[varying])));
- else
+ } else
return NULL;
}
+void
+vec4_visitor::emit_generic_urb_slot(dst_reg reg, int varying, int component)
+{
+ assert(varying < VARYING_SLOT_MAX);
+ assert(varying >= VARYING_SLOT_VAR0);
+ varying = varying - VARYING_SLOT_VAR0;
+
+ unsigned num_comps = output_generic_num_components[varying][component];
+ if (num_comps == 0)
+ return;
+
+ assert(output_generic_reg[varying][component].type == reg.type);
+ current_annotation = output_reg_annotation[varying];
+ if (output_generic_reg[varying][component].file != BAD_FILE) {
+ src_reg src = src_reg(output_generic_reg[varying][component]);
+ src.swizzle = BRW_SWZ_COMP_OUTPUT(component);
+ reg.writemask =
+ brw_writemask_for_component_packing(num_comps, component);
+ emit(MOV(reg, src));
+ }
+}
+
void
vec4_visitor::emit_urb_slot(dst_reg reg, int varying)
{
/* No need to write to this slot */
break;
default:
- emit_generic_urb_slot(reg, varying);
+ if (varying >= VARYING_SLOT_VAR0) {
+ for (int i = 0; i < 4; i++) {
+ emit_generic_urb_slot(reg, varying, i);
+ }
+ } else {
+ emit_generic_urb_slot(reg, varying);
+ }
break;
}
}
}
}
-src_reg
-vec4_visitor::get_pull_constant_offset(bblock_t * block, vec4_instruction *inst,
- src_reg *reladdr, int reg_offset)
-{
- if (reladdr) {
- src_reg index = src_reg(this, glsl_type::int_type);
-
- emit_before(block, inst, ADD(dst_reg(index), *reladdr,
- brw_imm_d(reg_offset * 16)));
-
- return index;
- } else if (devinfo->gen >= 8) {
- /* Store the offset in a GRF so we can send-from-GRF. */
- src_reg offset = src_reg(this, glsl_type::int_type);
- emit_before(block, inst, MOV(dst_reg(offset), brw_imm_d(reg_offset * 16)));
- return offset;
- } else {
- return brw_imm_d(reg_offset * 16);
- }
-}
-
/**
* Emits an instruction before @inst to load the value named by @orig_src
* from scratch space at @base_offset to @temp.
void
vec4_visitor::emit_pull_constant_load(bblock_t *block, vec4_instruction *inst,
dst_reg temp, src_reg orig_src,
- int base_offset)
+ int base_offset, src_reg indirect)
{
int reg_offset = base_offset + orig_src.reg_offset;
const unsigned index = prog_data->base.binding_table.pull_constants_start;
- src_reg offset = get_pull_constant_offset(block, inst, orig_src.reladdr,
- reg_offset);
+
+ src_reg offset;
+ if (indirect.file != BAD_FILE) {
+ offset = src_reg(this, glsl_type::uint_type);
+
+ emit_before(block, inst, ADD(dst_reg(offset), indirect,
+ brw_imm_ud(reg_offset * 16)));
+ } else if (devinfo->gen >= 8) {
+ /* Store the offset in a GRF so we can send-from-GRF. */
+ offset = src_reg(this, glsl_type::uint_type);
+ emit_before(block, inst, MOV(dst_reg(offset), brw_imm_ud(reg_offset * 16)));
+ } else {
+ offset = brw_imm_d(reg_offset * 16);
+ }
emit_pull_constant_load_reg(temp,
brw_imm_ud(index),
void
vec4_visitor::move_uniform_array_access_to_pull_constants()
{
+ /* The vulkan dirver doesn't support pull constants other than UBOs so
+ * everything has to be pushed regardless.
+ */
+ if (stage_prog_data->pull_param == NULL) {
+ split_uniform_registers();
+ return;
+ }
+
int pull_constant_loc[this->uniforms];
memset(pull_constant_loc, -1, sizeof(pull_constant_loc));
- bool nested_reladdr;
- /* Walk through and find array access of uniforms. Put a copy of that
- * uniform in the pull constant buffer.
- *
- * Note that we don't move constant-indexed accesses to arrays. No
- * testing has been done of the performance impact of this choice.
+ /* First, walk through the instructions and determine which things need to
+ * be pulled. We mark something as needing to be pulled by setting
+ * pull_constant_loc to 0.
*/
- do {
- nested_reladdr = false;
-
- foreach_block_and_inst_safe(block, vec4_instruction, inst, cfg) {
- for (int i = 0 ; i < 3; i++) {
- if (inst->src[i].file != UNIFORM || !inst->src[i].reladdr)
- continue;
+ foreach_block_and_inst(block, vec4_instruction, inst, cfg) {
+ /* We only care about MOV_INDIRECT of a uniform */
+ if (inst->opcode != SHADER_OPCODE_MOV_INDIRECT ||
+ inst->src[0].file != UNIFORM)
+ continue;
- int uniform = inst->src[i].nr;
+ int uniform_nr = inst->src[0].nr + inst->src[0].reg_offset;
- if (inst->src[i].reladdr->reladdr)
- nested_reladdr = true; /* will need another pass */
+ for (unsigned j = 0; j < DIV_ROUND_UP(inst->src[2].ud, 16); j++)
+ pull_constant_loc[uniform_nr + j] = 0;
+ }
- /* If this array isn't already present in the pull constant buffer,
- * add it.
- */
- if (pull_constant_loc[uniform] == -1) {
- const gl_constant_value **values =
- &stage_prog_data->param[uniform * 4];
+ /* Next, we walk the list of uniforms and assign real pull constant
+ * locations and set their corresponding entries in pull_param.
+ */
+ for (int j = 0; j < this->uniforms; j++) {
+ if (pull_constant_loc[j] < 0)
+ continue;
- pull_constant_loc[uniform] = stage_prog_data->nr_pull_params / 4;
+ pull_constant_loc[j] = stage_prog_data->nr_pull_params / 4;
- assert(uniform < uniform_array_size);
- for (int j = 0; j < uniform_size[uniform] * 4; j++) {
- stage_prog_data->pull_param[stage_prog_data->nr_pull_params++]
- = values[j];
- }
- }
+ for (int i = 0; i < 4; i++) {
+ stage_prog_data->pull_param[stage_prog_data->nr_pull_params++]
+ = stage_prog_data->param[j * 4 + i];
+ }
+ }
- /* Set up the annotation tracking for new generated instructions. */
- base_ir = inst->ir;
- current_annotation = inst->annotation;
+ /* Finally, we can walk through the instructions and lower MOV_INDIRECT
+ * instructions to actual uniform pulls.
+ */
+ foreach_block_and_inst_safe(block, vec4_instruction, inst, cfg) {
+ /* We only care about MOV_INDIRECT of a uniform */
+ if (inst->opcode != SHADER_OPCODE_MOV_INDIRECT ||
+ inst->src[0].file != UNIFORM)
+ continue;
- dst_reg temp = dst_reg(this, glsl_type::vec4_type);
+ int uniform_nr = inst->src[0].nr + inst->src[0].reg_offset;
- emit_pull_constant_load(block, inst, temp, inst->src[i],
- pull_constant_loc[uniform]);
+ assert(inst->src[0].swizzle == BRW_SWIZZLE_NOOP);
- inst->src[i].file = temp.file;
- inst->src[i].nr = temp.nr;
- inst->src[i].reg_offset = temp.reg_offset;
- inst->src[i].reladdr = NULL;
- }
- }
- } while (nested_reladdr);
+ emit_pull_constant_load(block, inst, inst->dst, inst->src[0],
+ pull_constant_loc[uniform_nr], inst->src[1]);
+ inst->remove(block);
+ }
/* Now there are no accesses of the UNIFORM file with a reladdr, so
* no need to track them as larger-than-vec4 objects. This will be
this->current_annotation = NULL;
memset(this->output_reg_annotation, 0, sizeof(this->output_reg_annotation));
+ memset(this->output_generic_num_components, 0,
+ sizeof(this->output_generic_num_components));
+
this->virtual_grf_start = NULL;
this->virtual_grf_end = NULL;
this->live_intervals = NULL;
this->max_grf = devinfo->gen >= 7 ? GEN7_MRF_HACK_START : BRW_MAX_GRF;
this->uniforms = 0;
-
- /* Initialize uniform_array_size to at least 1 because pre-gen6 VS requires
- * at least one. See setup_uniforms() in brw_vec4.cpp.
- */
- this->uniform_array_size = 1;
- if (prog_data) {
- this->uniform_array_size =
- MAX2(DIV_ROUND_UP(stage_prog_data->nr_params, 4), 1);
- }
-
- this->uniform_size = rzalloc_array(mem_ctx, int, this->uniform_array_size);
}
vec4_visitor::~vec4_visitor()