#include <stdio.h>
#include "main/compiler.h"
-#include "ir.h"
-#include "ir_visitor.h"
-#include "ir_expression_flattening.h"
-#include "ir_uniform.h"
-#include "glsl_types.h"
-#include "glsl_parser_extras.h"
-#include "../glsl/program.h"
-#include "ir_optimization.h"
-#include "ast.h"
-#include "linker.h"
-
+#include "main/macros.h"
#include "main/mtypes.h"
#include "main/shaderapi.h"
#include "main/shaderobj.h"
#include "main/uniforms.h"
-
+#include "compiler/glsl/ast.h"
+#include "compiler/glsl/ir.h"
+#include "compiler/glsl/ir_expression_flattening.h"
+#include "compiler/glsl/ir_visitor.h"
+#include "compiler/glsl/ir_optimization.h"
+#include "compiler/glsl/ir_uniform.h"
+#include "compiler/glsl/glsl_parser_extras.h"
+#include "compiler/glsl_types.h"
+#include "compiler/glsl/linker.h"
+#include "compiler/glsl/program.h"
#include "program/hash_table.h"
#include "program/prog_instruction.h"
#include "program/prog_optimize.h"
#include "program/prog_print.h"
#include "program/program.h"
#include "program/prog_parameter.h"
-#include "program/sampler.h"
static int swizzle_for_size(int size);
this->file = file;
this->index = 0;
this->writemask = writemask;
- this->cond_mask = COND_TR;
this->reladdr = NULL;
}
this->file = PROGRAM_UNDEFINED;
this->index = 0;
this->writemask = 0;
- this->cond_mask = COND_TR;
this->reladdr = NULL;
}
gl_register_file file; /**< PROGRAM_* from Mesa */
int index; /**< temporary index, VERT_ATTRIB_*, VARYING_SLOT_*, etc. */
int writemask; /**< Bitfield of WRITEMASK_[XYZW] */
- GLuint cond_mask:4;
/** Register index should be offset by the integer in this reg. */
src_reg *reladdr;
};
this->file = reg.file;
this->index = reg.index;
this->writemask = WRITEMASK_XYZW;
- this->cond_mask = COND_TR;
this->reladdr = reg.reladdr;
}
src_reg src[3];
/** Pointer to the ir source this tree came from for debugging */
ir_instruction *ir;
- GLboolean cond_update;
bool saturate;
int sampler; /**< sampler index */
int tex_target; /**< One of TEXTURE_*_INDEX */
void emit_swz(ir_expression *ir);
+ void emit_equality_comparison(ir_expression *ir, enum prog_opcode op,
+ dst_reg dst,
+ const src_reg &src0, const src_reg &src1);
+
+ inline void emit_sne(ir_expression *ir, dst_reg dst,
+ const src_reg &src0, const src_reg &src1)
+ {
+ emit_equality_comparison(ir, OPCODE_SLT, dst, src0, src1);
+ }
+
+ inline void emit_seq(ir_expression *ir, dst_reg dst,
+ const src_reg &src0, const src_reg &src1)
+ {
+ emit_equality_comparison(ir, OPCODE_SGE, dst, src0, src1);
+ }
+
bool process_move_condition(ir_rvalue *ir);
void copy_propagate(void);
return size;
case GLSL_TYPE_SAMPLER:
case GLSL_TYPE_IMAGE:
+ case GLSL_TYPE_SUBROUTINE:
/* Samplers take up one slot in UNIFORMS[], but they're baked in
* at link time.
*/
case GLSL_TYPE_VOID:
case GLSL_TYPE_ERROR:
case GLSL_TYPE_INTERFACE:
+ case GLSL_TYPE_FUNCTION:
assert(!"Invalid type in type_size");
break;
}
this->result = result_src;
}
+void
+ir_to_mesa_visitor::emit_equality_comparison(ir_expression *ir,
+ enum prog_opcode op,
+ dst_reg dst,
+ const src_reg &src0,
+ const src_reg &src1)
+{
+ src_reg difference;
+ src_reg abs_difference = get_temp(glsl_type::vec4_type);
+ const src_reg zero = src_reg_for_float(0.0);
+
+ /* x == y is equivalent to -abs(x-y) >= 0. Since all of the code that
+ * consumes the generated IR is pretty dumb, take special care when one
+ * of the operands is zero.
+ *
+ * Similarly, x != y is equivalent to -abs(x-y) < 0.
+ */
+ if (src0.file == zero.file &&
+ src0.index == zero.index &&
+ src0.swizzle == zero.swizzle) {
+ difference = src1;
+ } else if (src1.file == zero.file &&
+ src1.index == zero.index &&
+ src1.swizzle == zero.swizzle) {
+ difference = src0;
+ } else {
+ difference = get_temp(glsl_type::vec4_type);
+
+ src_reg tmp_src = src0;
+ tmp_src.negate = ~tmp_src.negate;
+
+ emit(ir, OPCODE_ADD, dst_reg(difference), tmp_src, src1);
+ }
+
+ emit(ir, OPCODE_ABS, dst_reg(abs_difference), difference);
+
+ abs_difference.negate = ~abs_difference.negate;
+ emit(ir, op, dst, abs_difference, zero);
+}
+
void
ir_to_mesa_visitor::visit(ir_expression *ir)
{
emit(ir, OPCODE_SLT, result_dst, op[0], op[1]);
break;
case ir_binop_greater:
- emit(ir, OPCODE_SGT, result_dst, op[0], op[1]);
+ /* Negating the operands (as opposed to switching the order of the
+ * operands) produces the correct result when both are +/-Inf.
+ */
+ op[0].negate = ~op[0].negate;
+ op[1].negate = ~op[1].negate;
+ emit(ir, OPCODE_SLT, result_dst, op[0], op[1]);
break;
case ir_binop_lequal:
- emit(ir, OPCODE_SLE, result_dst, op[0], op[1]);
+ /* Negating the operands (as opposed to switching the order of the
+ * operands) produces the correct result when both are +/-Inf.
+ */
+ op[0].negate = ~op[0].negate;
+ op[1].negate = ~op[1].negate;
+ emit(ir, OPCODE_SGE, result_dst, op[0], op[1]);
break;
case ir_binop_gequal:
emit(ir, OPCODE_SGE, result_dst, op[0], op[1]);
break;
case ir_binop_equal:
- emit(ir, OPCODE_SEQ, result_dst, op[0], op[1]);
+ emit_seq(ir, result_dst, op[0], op[1]);
break;
case ir_binop_nequal:
- emit(ir, OPCODE_SNE, result_dst, op[0], op[1]);
+ emit_sne(ir, result_dst, op[0], op[1]);
break;
case ir_binop_all_equal:
/* "==" operator producing a scalar boolean. */
if (ir->operands[0]->type->is_vector() ||
ir->operands[1]->type->is_vector()) {
src_reg temp = get_temp(glsl_type::vec4_type);
- emit(ir, OPCODE_SNE, dst_reg(temp), op[0], op[1]);
+ emit_sne(ir, dst_reg(temp), op[0], op[1]);
/* After the dot-product, the value will be an integer on the
* range [0,4]. Zero becomes 1.0, and positive values become zero.
sge_src.negate = ~sge_src.negate;
emit(ir, OPCODE_SGE, result_dst, sge_src, src_reg_for_float(0.0));
} else {
- emit(ir, OPCODE_SEQ, result_dst, op[0], op[1]);
+ emit_seq(ir, result_dst, op[0], op[1]);
}
break;
case ir_binop_any_nequal:
if (ir->operands[0]->type->is_vector() ||
ir->operands[1]->type->is_vector()) {
src_reg temp = get_temp(glsl_type::vec4_type);
- emit(ir, OPCODE_SNE, dst_reg(temp), op[0], op[1]);
+ if (ir->operands[0]->type->is_boolean() &&
+ ir->operands[1]->as_constant() &&
+ ir->operands[1]->as_constant()->is_zero()) {
+ temp = op[0];
+ } else {
+ emit_sne(ir, dst_reg(temp), op[0], op[1]);
+ }
/* After the dot-product, the value will be an integer on the
* range [0,4]. Zero stays zero, and positive values become 1.0.
emit(ir, OPCODE_SLT, result_dst, slt_src, src_reg_for_float(0.0));
}
} else {
- emit(ir, OPCODE_SNE, result_dst, op[0], op[1]);
+ emit_sne(ir, result_dst, op[0], op[1]);
}
break;
- case ir_unop_any: {
- assert(ir->operands[0]->type->is_vector());
-
- /* After the dot-product, the value will be an integer on the
- * range [0,4]. Zero stays zero, and positive values become 1.0.
- */
- ir_to_mesa_instruction *const dp =
- emit_dp(ir, result_dst, op[0], op[0],
- ir->operands[0]->type->vector_elements);
- if (this->prog->Target == GL_FRAGMENT_PROGRAM_ARB) {
- /* The clamping to [0,1] can be done for free in the fragment
- * shader with a saturate.
- */
- dp->saturate = true;
- } else {
- /* Negating the result of the dot-product gives values on the range
- * [-4, 0]. Zero stays zero, and negative values become 1.0. This
- * is achieved using SLT.
- */
- src_reg slt_src = result_src;
- slt_src.negate = ~slt_src.negate;
- emit(ir, OPCODE_SLT, result_dst, slt_src, src_reg_for_float(0.0));
- }
- break;
- }
-
case ir_binop_logic_xor:
- emit(ir, OPCODE_SNE, result_dst, op[0], op[1]);
+ emit_sne(ir, result_dst, op[0], op[1]);
break;
case ir_binop_logic_or: {
- /* After the addition, the value will be an integer on the
- * range [0,2]. Zero stays zero, and positive values become 1.0.
- */
- ir_to_mesa_instruction *add =
- emit(ir, OPCODE_ADD, result_dst, op[0], op[1]);
if (this->prog->Target == GL_FRAGMENT_PROGRAM_ARB) {
- /* The clamping to [0,1] can be done for free in the fragment
- * shader with a saturate.
- */
+ /* After the addition, the value will be an integer on the
+ * range [0,2]. Zero stays zero, and positive values become 1.0.
+ */
+ ir_to_mesa_instruction *add =
+ emit(ir, OPCODE_ADD, result_dst, op[0], op[1]);
add->saturate = true;
} else {
- /* Negating the result of the addition gives values on the range
- * [-2, 0]. Zero stays zero, and negative values become 1.0. This
- * is achieved using SLT.
- */
- src_reg slt_src = result_src;
- slt_src.negate = ~slt_src.negate;
- emit(ir, OPCODE_SLT, result_dst, slt_src, src_reg_for_float(0.0));
+ /* The Boolean arguments are stored as float 0.0 and 1.0. If either
+ * value is 1.0, the result of the logcal-or should be 1.0. If both
+ * values are 0.0, the result should be 0.0. This is exactly what
+ * MAX does.
+ */
+ emit(ir, OPCODE_MAX, result_dst, op[0], op[1]);
}
break;
}
break;
case ir_unop_f2b:
case ir_unop_i2b:
- emit(ir, OPCODE_SNE, result_dst,
- op[0], src_reg_for_float(0.0));
+ emit_sne(ir, result_dst, op[0], src_reg_for_float(0.0));
break;
case ir_unop_bitcast_f2i: // Ignore these 4, they can't happen here anyway
case ir_unop_bitcast_f2u:
case ir_unop_unpack_unorm_2x16:
case ir_unop_unpack_unorm_4x8:
case ir_unop_unpack_half_2x16:
- case ir_unop_unpack_half_2x16_split_x:
- case ir_unop_unpack_half_2x16_split_y:
case ir_unop_unpack_double_2x32:
- case ir_binop_pack_half_2x16_split:
case ir_unop_bitfield_reverse:
case ir_unop_bit_count:
case ir_unop_find_msb:
break;
case ir_binop_vector_extract:
- case ir_binop_bfm:
case ir_triop_fma:
- case ir_triop_bfi:
case ir_triop_bitfield_extract:
case ir_triop_vector_insert:
case ir_quadop_bitfield_insert:
case ir_unop_dFdx_fine:
case ir_unop_dFdy_coarse:
case ir_unop_dFdy_fine:
+ case ir_unop_subroutine_to_int:
+ case ir_unop_get_buffer_size:
+ case ir_unop_vote_any:
+ case ir_unop_vote_all:
+ case ir_unop_vote_eq:
assert(!"not supported");
break;
+ case ir_unop_ssbo_unsized_array_length:
case ir_quadop_vector:
/* This operation should have already been handled.
*/
switch (var->data.mode) {
case ir_var_uniform:
entry = new(mem_ctx) variable_storage(var, PROGRAM_UNIFORM,
- var->data.location);
+ var->data.param_index);
this->variables.push_tail(entry);
break;
case ir_var_shader_in:
return dst_reg(v->result);
}
+/* Calculate the sampler index and also calculate the base uniform location
+ * for struct members.
+ */
+static void
+calc_sampler_offsets(struct gl_shader_program *prog, ir_dereference *deref,
+ unsigned *offset, unsigned *array_elements,
+ unsigned *location)
+{
+ if (deref->ir_type == ir_type_dereference_variable)
+ return;
+
+ switch (deref->ir_type) {
+ case ir_type_dereference_array: {
+ ir_dereference_array *deref_arr = deref->as_dereference_array();
+ ir_constant *array_index =
+ deref_arr->array_index->constant_expression_value();
+
+ if (!array_index) {
+ /* GLSL 1.10 and 1.20 allowed variable sampler array indices,
+ * while GLSL 1.30 requires that the array indices be
+ * constant integer expressions. We don't expect any driver
+ * to actually work with a really variable array index, so
+ * all that would work would be an unrolled loop counter that ends
+ * up being constant above.
+ */
+ ralloc_strcat(&prog->InfoLog,
+ "warning: Variable sampler array index unsupported.\n"
+ "This feature of the language was removed in GLSL 1.20 "
+ "and is unlikely to be supported for 1.10 in Mesa.\n");
+ } else {
+ *offset += array_index->value.u[0] * *array_elements;
+ }
+
+ *array_elements *= deref_arr->array->type->length;
+
+ calc_sampler_offsets(prog, deref_arr->array->as_dereference(),
+ offset, array_elements, location);
+ break;
+ }
+
+ case ir_type_dereference_record: {
+ ir_dereference_record *deref_record = deref->as_dereference_record();
+ unsigned field_index =
+ deref_record->record->type->field_index(deref_record->field);
+ *location +=
+ deref_record->record->type->record_location_offset(field_index);
+ calc_sampler_offsets(prog, deref_record->record->as_dereference(),
+ offset, array_elements, location);
+ break;
+ }
+
+ default:
+ unreachable("Invalid deref type");
+ break;
+ }
+}
+
+static int
+get_sampler_uniform_value(class ir_dereference *sampler,
+ struct gl_shader_program *shader_program,
+ const struct gl_program *prog)
+{
+ GLuint shader = _mesa_program_enum_to_shader_stage(prog->Target);
+ ir_variable *var = sampler->variable_referenced();
+ unsigned location = var->data.location;
+ unsigned array_elements = 1;
+ unsigned offset = 0;
+
+ calc_sampler_offsets(shader_program, sampler, &offset, &array_elements,
+ &location);
+
+ assert(shader_program->UniformStorage[location].opaque[shader].active);
+ return shader_program->UniformStorage[location].opaque[shader].index +
+ offset;
+}
+
/**
* Process the condition of a conditional assignment
*
ir->coordinate->accept(this);
/* Put our coords in a temp. We'll need to modify them for shadow,
- * projection, or LOD, so the only case we'd use it as is is if
+ * projection, or LOD, so the only case we'd use it as-is is if
* we're doing plain old texturing. Mesa IR optimization should
* handle cleaning up our mess in that case.
*/
case ir_query_levels:
assert(!"Unexpected ir_query_levels opcode");
break;
+ case ir_samples_identical:
+ unreachable("Unexpected ir_samples_identical opcode");
+ case ir_texture_samples:
+ unreachable("Unexpected ir_texture_samples opcode");
}
const glsl_type *sampler_type = ir->sampler->type;
if (ir->shadow_comparitor)
inst->tex_shadow = GL_TRUE;
- inst->sampler = _mesa_get_sampler_uniform_value(ir->sampler,
- this->shader_program,
- this->prog);
+ inst->sampler = get_sampler_uniform_value(ir->sampler, shader_program,
+ prog);
switch (sampler_type->sampler_dimensionality) {
case GLSL_SAMPLER_DIM_1D:
void
ir_to_mesa_visitor::visit(ir_discard *ir)
{
- if (ir->condition) {
- ir->condition->accept(this);
- this->result.negate = ~this->result.negate;
- emit(ir, OPCODE_KIL, undef_dst, this->result);
- } else {
- emit(ir, OPCODE_KIL_NV);
- }
+ if (!ir->condition)
+ ir->condition = new(mem_ctx) ir_constant(true);
+
+ ir->condition->accept(this);
+ this->result.negate = ~this->result.negate;
+ emit(ir, OPCODE_KIL, undef_dst, this->result);
}
void
ir_to_mesa_visitor::visit(ir_if *ir)
{
- ir_to_mesa_instruction *cond_inst, *if_inst;
- ir_to_mesa_instruction *prev_inst;
-
- prev_inst = (ir_to_mesa_instruction *)this->instructions.get_tail();
+ ir_to_mesa_instruction *if_inst;
ir->condition->accept(this);
assert(this->result.file != PROGRAM_UNDEFINED);
- if (this->options->EmitCondCodes) {
- cond_inst = (ir_to_mesa_instruction *)this->instructions.get_tail();
-
- /* See if we actually generated any instruction for generating
- * the condition. If not, then cook up a move to a temp so we
- * have something to set cond_update on.
- */
- if (cond_inst == prev_inst) {
- src_reg temp = get_temp(glsl_type::bool_type);
- cond_inst = emit(ir->condition, OPCODE_MOV, dst_reg(temp), result);
- }
- cond_inst->cond_update = GL_TRUE;
-
- if_inst = emit(ir->condition, OPCODE_IF);
- if_inst->dst.cond_mask = COND_NE;
- } else {
- if_inst = emit(ir->condition, OPCODE_IF, undef_dst, this->result);
- }
+ if_inst = emit(ir->condition, OPCODE_IF, undef_dst, this->result);
this->instructions.push_tail(if_inst);
mesa_reg.Swizzle = reg.swizzle;
mesa_reg.RelAddr = reg.reladdr != NULL;
mesa_reg.Negate = reg.negate;
- mesa_reg.Abs = 0;
- mesa_reg.HasIndex2 = GL_FALSE;
- mesa_reg.RelAddr2 = 0;
- mesa_reg.Index2 = 0;
return mesa_reg;
}
{
this->idx = -1;
this->program_resource_visitor::process(var);
-
- var->data.location = this->idx;
+ var->data.param_index = this->idx;
}
private:
(void) row_major;
+ /* atomics don't get real storage */
+ if (type->contains_atomic())
+ return;
+
if (type->is_vector() || type->is_scalar()) {
size = type->vector_elements;
- if (type->is_double())
+ if (type->is_64bit())
size *= 2;
} else {
size = type_size(type) * 4;
file = PROGRAM_UNIFORM;
}
- int index = _mesa_lookup_parameter_index(params, -1, name);
+ int index = _mesa_lookup_parameter_index(params, name);
if (index < 0) {
index = _mesa_add_parameter(params, file, name, size, type->gl_type,
NULL, NULL);
struct gl_uniform_storage *storage =
&this->shader_program->UniformStorage[location];
- assert(storage->sampler[shader_type].active);
+ assert(storage->type->is_sampler() &&
+ storage->opaque[shader_type].active);
for (unsigned int j = 0; j < size / 4; j++)
params->ParameterValues[index + j][0].f =
- storage->sampler[shader_type].index + j;
+ storage->opaque[shader_type].index + j;
}
}
void
_mesa_generate_parameters_list_for_uniforms(struct gl_shader_program
*shader_program,
- struct gl_shader *sh,
+ struct gl_linked_shader *sh,
struct gl_program_parameter_list
*params)
{
ir_variable *var = node->as_variable();
if ((var == NULL) || (var->data.mode != ir_var_uniform)
- || var->is_in_uniform_block() || (strncmp(var->name, "gl_", 3) == 0))
+ || var->is_in_buffer_block() || (strncmp(var->name, "gl_", 3) == 0))
continue;
add.process(var);
break;
case GLSL_TYPE_SAMPLER:
case GLSL_TYPE_IMAGE:
+ case GLSL_TYPE_SUBROUTINE:
format = uniform_native;
columns = 1;
break;
case GLSL_TYPE_STRUCT:
case GLSL_TYPE_ERROR:
case GLSL_TYPE_INTERFACE:
+ case GLSL_TYPE_FUNCTION:
assert(!"Should not get here.");
break;
}
static struct gl_program *
get_mesa_program(struct gl_context *ctx,
struct gl_shader_program *shader_program,
- struct gl_shader *shader)
+ struct gl_linked_shader *shader)
{
ir_to_mesa_visitor v;
struct prog_instruction *mesa_instructions, *mesa_inst;
i = 0;
foreach_in_list(const ir_to_mesa_instruction, inst, &v.instructions) {
mesa_inst->Opcode = inst->op;
- mesa_inst->CondUpdate = inst->cond_update;
if (inst->saturate)
mesa_inst->Saturate = GL_TRUE;
mesa_inst->DstReg.File = inst->dst.file;
mesa_inst->DstReg.Index = inst->dst.index;
- mesa_inst->DstReg.CondMask = inst->dst.cond_mask;
mesa_inst->DstReg.WriteMask = inst->dst.writemask;
mesa_inst->DstReg.RelAddr = inst->dst.reladdr != NULL;
mesa_inst->SrcReg[0] = mesa_src_reg_from_ir_src_reg(inst->src[0]);
if (options->EmitNoIndirectInput || options->EmitNoIndirectOutput
|| options->EmitNoIndirectTemp || options->EmitNoIndirectUniform)
progress =
- lower_variable_index_to_cond_assign(ir,
+ lower_variable_index_to_cond_assign(prog->_LinkedShaders[i]->Stage, ir,
options->EmitNoIndirectInput,
options->EmitNoIndirectOutput,
options->EmitNoIndirectTemp,
if (linked_prog) {
_mesa_copy_linked_program_data((gl_shader_stage) i, prog, linked_prog);
- _mesa_reference_program(ctx, &prog->_LinkedShaders[i]->Program,
- linked_prog);
if (!ctx->Driver.ProgramStringNotify(ctx,
_mesa_shader_stage_to_program(i),
linked_prog)) {
_mesa_reference_program(ctx, &linked_prog, NULL);
}
+ build_program_resource_list(ctx, prog);
return prog->LinkStatus;
}