#include <stdio.h>
#include "main/compiler.h"
+#include "main/macros.h"
#include "main/mtypes.h"
#include "main/shaderapi.h"
#include "main/shaderobj.h"
#include "main/uniforms.h"
-#include "glsl/ast.h"
-#include "glsl/ir.h"
-#include "glsl/ir_expression_flattening.h"
-#include "glsl/ir_visitor.h"
-#include "glsl/ir_optimization.h"
-#include "glsl/ir_uniform.h"
-#include "glsl/glsl_parser_extras.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 "glsl/linker.h"
-#include "glsl/program.h"
-#include "program/hash_table.h"
+#include "compiler/glsl/linker.h"
+#include "compiler/glsl/program.h"
+#include "compiler/glsl/shader_cache.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"
+#include "util/string_to_uint_map.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 1;
}
break;
+ case GLSL_TYPE_UINT64:
+ case GLSL_TYPE_INT64:
+ if (type->vector_elements > 2)
+ return 2;
+ else
+ return 1;
case GLSL_TYPE_ARRAY:
assert(type->length > 0);
return type_size(type->fields.array) * type->length;
case GLSL_TYPE_VOID:
case GLSL_TYPE_ERROR:
case GLSL_TYPE_INTERFACE:
+ case GLSL_TYPE_FUNCTION:
assert(!"Invalid type in type_size");
break;
}
ir_to_mesa_visitor::visit(ir_variable *ir)
{
if (strcmp(ir->name, "gl_FragCoord") == 0) {
- struct gl_fragment_program *fp = (struct gl_fragment_program *)this->prog;
-
- fp->OriginUpperLeft = ir->data.origin_upper_left;
- fp->PixelCenterInteger = ir->data.pixel_center_integer;
+ this->prog->OriginUpperLeft = ir->data.origin_upper_left;
+ this->prog->PixelCenterInteger = ir->data.pixel_center_integer;
}
if (ir->data.mode == ir_var_uniform && strncmp(ir->name, "gl_", 3) == 0) {
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_scalar(ir, OPCODE_EX2, result_dst, op[0]);
break;
case ir_unop_exp:
+ assert(!"not reached: should be handled by exp_to_exp2");
+ break;
case ir_unop_log:
- assert(!"not reached: should be handled by ir_explog_to_explog2");
+ assert(!"not reached: should be handled by log_to_log2");
break;
case ir_unop_log2:
emit_scalar(ir, OPCODE_LG2, result_dst, op[0]);
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:
ir->operands[1]->as_constant()->is_zero()) {
temp = op[0];
} else {
- 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
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_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:
emit(ir, OPCODE_LRP, result_dst, op[2], op[1], op[0]);
break;
+ case ir_triop_csel:
+ /* We assume that boolean true and false are 1.0 and 0.0. OPCODE_CMP
+ * selects src1 if src0 is < 0, src2 otherwise.
+ */
+ op[0].negate = ~op[0].negate;
+ emit(ir, OPCODE_CMP, result_dst, op[0], op[1], op[2]);
+ break;
+
case ir_binop_vector_extract:
case ir_triop_fma:
case ir_triop_bitfield_extract:
case ir_triop_vector_insert:
case ir_quadop_bitfield_insert:
case ir_binop_ldexp:
- case ir_triop_csel:
case ir_binop_carry:
case ir_binop_borrow:
case ir_binop_imul_high:
case ir_unop_dFdy_fine:
case ir_unop_subroutine_to_int:
case ir_unop_get_buffer_size:
+ case ir_unop_bitcast_u642d:
+ case ir_unop_bitcast_i642d:
+ case ir_unop_bitcast_d2u64:
+ case ir_unop_bitcast_d2i64:
+ case ir_unop_i642i:
+ case ir_unop_u642i:
+ case ir_unop_i642u:
+ case ir_unop_u642u:
+ case ir_unop_i642b:
+ case ir_unop_i642f:
+ case ir_unop_u642f:
+ case ir_unop_i642d:
+ case ir_unop_u642d:
+ case ir_unop_i2i64:
+ case ir_unop_u2i64:
+ case ir_unop_b2i64:
+ case ir_unop_f2i64:
+ case ir_unop_d2i64:
+ case ir_unop_i2u64:
+ case ir_unop_u2u64:
+ case ir_unop_f2u64:
+ case ir_unop_d2u64:
+ case ir_unop_u642i64:
+ case ir_unop_i642u64:
+ case ir_unop_pack_int_2x32:
+ case ir_unop_unpack_int_2x32:
+ case ir_unop_pack_uint_2x32:
+ case ir_unop_unpack_uint_2x32:
+ case ir_unop_pack_sampler_2x32:
+ case ir_unop_unpack_sampler_2x32:
+ case ir_unop_pack_image_2x32:
+ case ir_unop_unpack_image_2x32:
assert(!"not supported");
break;
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->data->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->data->UniformStorage[location].opaque[shader].active);
+ return shader_program->data->UniformStorage[location].opaque[shader].index +
+ offset;
+}
+
/**
* Process the condition of a conditional assignment
*
* get lucky, copy propagation will eliminate the extra moves.
*/
- if (ir->type->base_type == GLSL_TYPE_STRUCT) {
+ if (ir->type->is_record()) {
src_reg temp_base = get_temp(ir->type);
dst_reg temp = dst_reg(temp_base);
dst_reg mat_column = dst_reg(mat);
for (i = 0; i < ir->type->matrix_columns; i++) {
- assert(ir->type->base_type == GLSL_TYPE_FLOAT);
+ assert(ir->type->is_float());
values = &ir->value.f[i * ir->type->vector_elements];
src = src_reg(PROGRAM_CONSTANT, -1, NULL);
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.
*/
emit(ir, OPCODE_RCP, coord_dst, projector);
/* In the case where we have to project the coordinates "by hand,"
- * the shadow comparitor value must also be projected.
+ * the shadow comparator value must also be projected.
*/
src_reg tmp_src = coord;
- if (ir->shadow_comparitor) {
+ if (ir->shadow_comparator) {
/* Slot the shadow value in as the second to last component of the
* coord.
*/
- ir->shadow_comparitor->accept(this);
+ ir->shadow_comparator->accept(this);
tmp_src = get_temp(glsl_type::vec4_type);
dst_reg tmp_dst = dst_reg(tmp_src);
}
/* If projection is done and the opcode is not OPCODE_TXP, then the shadow
- * comparitor was put in the correct place (and projected) by the code,
+ * comparator was put in the correct place (and projected) by the code,
* above, that handles by-hand projection.
*/
- if (ir->shadow_comparitor && (!ir->projector || opcode == OPCODE_TXP)) {
+ if (ir->shadow_comparator && (!ir->projector || opcode == OPCODE_TXP)) {
/* Slot the shadow value in as the second to last component of the
* coord.
*/
- ir->shadow_comparitor->accept(this);
+ ir->shadow_comparator->accept(this);
/* XXX This will need to be updated for cubemap array samplers. */
if (sampler_type->sampler_dimensionality == GLSL_SAMPLER_DIM_2D &&
else
inst = emit(ir, opcode, result_dst, coord);
- if (ir->shadow_comparitor)
+ if (ir->shadow_comparator)
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:
virtual void visit_field(const glsl_type *type, const char *name,
- bool row_major);
+ bool row_major, const glsl_type *record_type,
+ const enum glsl_interface_packing packing,
+ bool last_field);
struct gl_shader_program *shader_program;
struct gl_program_parameter_list *params;
void
add_uniform_to_shader::visit_field(const glsl_type *type, const char *name,
- bool row_major)
+ bool /* row_major */,
+ const glsl_type * /* record_type */,
+ const enum glsl_interface_packing,
+ bool /* last_field */)
{
- unsigned int size;
-
- (void) row_major;
-
- if (type->is_vector() || type->is_scalar()) {
- size = type->vector_elements;
- if (type->is_double())
- size *= 2;
- } else {
- size = type_size(type) * 4;
- }
+ /* atomics don't get real storage */
+ if (type->contains_atomic())
+ return;
gl_register_file file;
if (type->without_array()->is_sampler()) {
file = PROGRAM_UNIFORM;
}
- int index = _mesa_lookup_parameter_index(params, -1, name);
+ int index = _mesa_lookup_parameter_index(params, name);
if (index < 0) {
+ unsigned size = type_size(type) * 4;
+
index = _mesa_add_parameter(params, file, name, size, type->gl_type,
NULL, NULL);
return;
struct gl_uniform_storage *storage =
- &this->shader_program->UniformStorage[location];
+ &this->shader_program->data->UniformStorage[location];
assert(storage->type->is_sampler() &&
storage->opaque[shader_type].active);
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)
{
void
_mesa_associate_uniform_storage(struct gl_context *ctx,
- struct gl_shader_program *shader_program,
- struct gl_program_parameter_list *params)
+ struct gl_shader_program *shader_program,
+ struct gl_program_parameter_list *params,
+ bool propagate_to_storage)
{
/* After adding each uniform to the parameter list, connect the storage for
* the parameter with the tracking structure used by the API for the
unsigned last_location = unsigned(~0);
for (unsigned i = 0; i < params->NumParameters; i++) {
if (params->Parameters[i].Type != PROGRAM_UNIFORM)
- continue;
+ continue;
unsigned location;
const bool found =
- shader_program->UniformHash->get(location, params->Parameters[i].Name);
+ shader_program->UniformHash->get(location, params->Parameters[i].Name);
assert(found);
if (!found)
- continue;
+ continue;
struct gl_uniform_storage *storage =
- &shader_program->UniformStorage[location];
+ &shader_program->data->UniformStorage[location];
/* Do not associate any uniform storage to built-in uniforms */
if (storage->builtin)
continue;
if (location != last_location) {
- enum gl_uniform_driver_format format = uniform_native;
-
- unsigned columns = 0;
- int dmul = 4 * sizeof(float);
- switch (storage->type->base_type) {
- case GLSL_TYPE_UINT:
- assert(ctx->Const.NativeIntegers);
- format = uniform_native;
- columns = 1;
- break;
- case GLSL_TYPE_INT:
- format =
- (ctx->Const.NativeIntegers) ? uniform_native : uniform_int_float;
- columns = 1;
- break;
+ enum gl_uniform_driver_format format = uniform_native;
+ unsigned columns = 0;
+ int dmul = 4 * sizeof(float);
- case GLSL_TYPE_DOUBLE:
- if (storage->type->vector_elements > 2)
+ switch (storage->type->base_type) {
+ case GLSL_TYPE_UINT64:
+ if (storage->type->vector_elements > 2)
dmul *= 2;
- /* fallthrough */
- case GLSL_TYPE_FLOAT:
- format = uniform_native;
- columns = storage->type->matrix_columns;
- break;
- case GLSL_TYPE_BOOL:
- format = uniform_native;
- columns = 1;
- break;
- case GLSL_TYPE_SAMPLER:
- case GLSL_TYPE_IMAGE:
+ /* fallthrough */
+ case GLSL_TYPE_UINT:
+ assert(ctx->Const.NativeIntegers);
+ format = uniform_native;
+ columns = 1;
+ break;
+ case GLSL_TYPE_INT64:
+ if (storage->type->vector_elements > 2)
+ dmul *= 2;
+ /* fallthrough */
+ case GLSL_TYPE_INT:
+ format =
+ (ctx->Const.NativeIntegers) ? uniform_native : uniform_int_float;
+ columns = 1;
+ break;
+ case GLSL_TYPE_DOUBLE:
+ if (storage->type->vector_elements > 2)
+ dmul *= 2;
+ /* fallthrough */
+ case GLSL_TYPE_FLOAT:
+ format = uniform_native;
+ columns = storage->type->matrix_columns;
+ break;
+ case GLSL_TYPE_BOOL:
+ format = uniform_native;
+ columns = 1;
+ break;
+ case GLSL_TYPE_SAMPLER:
+ case GLSL_TYPE_IMAGE:
case GLSL_TYPE_SUBROUTINE:
- format = uniform_native;
- columns = 1;
- break;
+ format = uniform_native;
+ columns = 1;
+ break;
case GLSL_TYPE_ATOMIC_UINT:
case GLSL_TYPE_ARRAY:
case GLSL_TYPE_VOID:
case GLSL_TYPE_STRUCT:
case GLSL_TYPE_ERROR:
case GLSL_TYPE_INTERFACE:
- assert(!"Should not get here.");
- break;
- }
-
- _mesa_uniform_attach_driver_storage(storage,
- dmul * columns,
- dmul,
- format,
- ¶ms->ParameterValues[i]);
+ case GLSL_TYPE_FUNCTION:
+ assert(!"Should not get here.");
+ break;
+ }
- /* After attaching the driver's storage to the uniform, propagate any
- * data from the linker's backing store. This will cause values from
- * initializers in the source code to be copied over.
- */
- _mesa_propagate_uniforms_to_driver_storage(storage,
- 0,
- MAX2(1, storage->array_elements));
+ _mesa_uniform_attach_driver_storage(storage, dmul * columns, dmul,
+ format,
+ ¶ms->ParameterValues[i]);
+
+ /* After attaching the driver's storage to the uniform, propagate any
+ * data from the linker's backing store. This will cause values from
+ * initializers in the source code to be copied over.
+ */
+ if (propagate_to_storage) {
+ unsigned array_elements = MAX2(1, storage->array_elements);
+ _mesa_propagate_uniforms_to_driver_storage(storage, 0,
+ array_elements);
+ }
- last_location = location;
+ last_location = location;
}
}
}
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;
validate_ir_tree(shader->ir);
- prog = ctx->Driver.NewProgram(ctx, target, shader_program->Name);
- if (!prog)
- return NULL;
+ prog = shader->Program;
prog->Parameters = _mesa_new_parameter_list();
v.ctx = ctx;
v.prog = prog;
visit_exec_list(shader->ir, &v);
v.emit(NULL, OPCODE_END);
- prog->NumTemporaries = v.next_temp;
+ prog->arb.NumTemporaries = v.next_temp;
unsigned num_instructions = v.instructions.length();
- mesa_instructions =
- (struct prog_instruction *)calloc(num_instructions,
- sizeof(*mesa_instructions));
+ mesa_instructions = rzalloc_array(prog, struct prog_instruction,
+ num_instructions);
mesa_instruction_annotation = ralloc_array(v.mem_ctx, ir_instruction *,
num_instructions);
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]);
/* Set IndirectRegisterFiles. */
if (mesa_inst->DstReg.RelAddr)
- prog->IndirectRegisterFiles |= 1 << mesa_inst->DstReg.File;
+ prog->arb.IndirectRegisterFiles |= 1 << mesa_inst->DstReg.File;
/* Update program's bitmask of indirectly accessed register files */
for (unsigned src = 0; src < 3; src++)
if (mesa_inst->SrcReg[src].RelAddr)
- prog->IndirectRegisterFiles |= 1 << mesa_inst->SrcReg[src].File;
+ prog->arb.IndirectRegisterFiles |= 1 << mesa_inst->SrcReg[src].File;
switch (mesa_inst->Opcode) {
case OPCODE_IF:
}
break;
case OPCODE_ARL:
- prog->NumAddressRegs = 1;
+ prog->arb.NumAddressRegs = 1;
break;
default:
break;
mesa_inst++;
i++;
- if (!shader_program->LinkStatus)
+ if (!shader_program->data->LinkStatus)
break;
}
- if (!shader_program->LinkStatus) {
+ if (!shader_program->data->LinkStatus) {
goto fail_exit;
}
fflush(stderr);
}
- prog->Instructions = mesa_instructions;
- prog->NumInstructions = num_instructions;
+ prog->arb.Instructions = mesa_instructions;
+ prog->arb.NumInstructions = num_instructions;
/* Setting this to NULL prevents a possible double free in the fail_exit
* path (far below).
do_set_program_inouts(shader->ir, prog, shader->Stage);
- prog->SamplersUsed = shader->active_samplers;
prog->ShadowSamplers = shader->shadow_samplers;
+ prog->ExternalSamplersUsed = gl_external_samplers(prog);
_mesa_update_shader_textures_used(shader_program, prog);
/* Set the gl_FragDepth layout. */
if (target == GL_FRAGMENT_PROGRAM_ARB) {
- struct gl_fragment_program *fp = (struct gl_fragment_program *)prog;
- fp->FragDepthLayout = shader_program->FragDepthLayout;
+ prog->info.fs.depth_layout = shader_program->FragDepthLayout;
}
- _mesa_reference_program(ctx, &shader->Program, prog);
-
- if ((ctx->_Shader->Flags & GLSL_NO_OPT) == 0) {
- _mesa_optimize_program(ctx, prog);
- }
+ _mesa_optimize_program(ctx, prog, prog);
/* This has to be done last. Any operation that can cause
* prog->ParameterValues to get reallocated (e.g., anything that adds a
* program constant) has to happen before creating this linkage.
*/
- _mesa_associate_uniform_storage(ctx, shader_program, prog->Parameters);
- if (!shader_program->LinkStatus) {
+ _mesa_associate_uniform_storage(ctx, shader_program, prog->Parameters,
+ true);
+ if (!shader_program->data->LinkStatus) {
goto fail_exit;
}
return prog;
fail_exit:
- free(mesa_instructions);
+ ralloc_free(mesa_instructions);
_mesa_reference_program(ctx, &shader->Program, NULL);
return NULL;
}
GLboolean
_mesa_ir_link_shader(struct gl_context *ctx, struct gl_shader_program *prog)
{
- assert(prog->LinkStatus);
+ assert(prog->data->LinkStatus);
for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
if (prog->_LinkedShaders[i] == NULL)
| LOG_TO_LOG2 | INT_DIV_TO_MUL_RCP
| ((options->EmitNoPow) ? POW_TO_EXP2 : 0)));
- progress = do_lower_jumps(ir, true, true, options->EmitNoMainReturn, options->EmitNoCont, options->EmitNoLoops) || progress;
-
progress = do_common_optimization(ir, true, true,
options, ctx->Const.NativeIntegers)
|| progress;
if (options->MaxIfDepth == 0)
progress = lower_discard(ir) || progress;
- progress = lower_if_to_cond_assign(ir, options->MaxIfDepth) || progress;
+ progress = lower_if_to_cond_assign((gl_shader_stage)i, ir,
+ options->MaxIfDepth) || progress;
- if (options->EmitNoNoise)
- progress = lower_noise(ir) || progress;
+ progress = lower_noise(ir) || progress;
/* If there are forms of indirect addressing that the driver
* cannot handle, perform the lowering pass.
linked_prog = get_mesa_program(ctx, prog, prog->_LinkedShaders[i]);
if (linked_prog) {
- _mesa_copy_linked_program_data((gl_shader_stage) i, prog, linked_prog);
+ _mesa_copy_linked_program_data(prog, prog->_LinkedShaders[i]);
- _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, &prog->_LinkedShaders[i]->Program,
+ NULL);
return GL_FALSE;
}
}
-
- _mesa_reference_program(ctx, &linked_prog, NULL);
}
- return prog->LinkStatus;
+ build_program_resource_list(ctx, prog);
+ return prog->data->LinkStatus;
}
/**
{
unsigned int i;
- _mesa_clear_shader_program_data(prog);
+ _mesa_clear_shader_program_data(ctx, prog);
- prog->LinkStatus = GL_TRUE;
+ prog->data->LinkStatus = linking_success;
for (i = 0; i < prog->NumShaders; i++) {
if (!prog->Shaders[i]->CompileStatus) {
}
}
- if (prog->LinkStatus) {
+ if (prog->data->LinkStatus) {
link_shaders(ctx, prog);
}
- if (prog->LinkStatus) {
+ if (prog->data->LinkStatus) {
+ /* Reset sampler validated to true, validation happens via the
+ * LinkShader call below.
+ */
+ prog->SamplersValidated = GL_TRUE;
+
if (!ctx->Driver.LinkShader(ctx, prog)) {
- prog->LinkStatus = GL_FALSE;
- } else {
- build_program_resource_list(prog);
+ prog->data->LinkStatus = linking_failure;
}
}
+ /* Return early if we are loading the shader from on-disk cache */
+ if (prog->data->LinkStatus == linking_skipped)
+ return;
+
if (ctx->_Shader->Flags & GLSL_DUMP) {
- if (!prog->LinkStatus) {
+ if (!prog->data->LinkStatus) {
fprintf(stderr, "GLSL shader program %d failed to link\n", prog->Name);
}
- if (prog->InfoLog && prog->InfoLog[0] != 0) {
+ if (prog->data->InfoLog && prog->data->InfoLog[0] != 0) {
fprintf(stderr, "GLSL shader program %d info log:\n", prog->Name);
- fprintf(stderr, "%s\n", prog->InfoLog);
+ fprintf(stderr, "%s\n", prog->data->InfoLog);
}
}
+
+#ifdef ENABLE_SHADER_CACHE
+ if (prog->data->LinkStatus)
+ shader_cache_write_program_metadata(ctx, prog);
+#endif
}
} /* extern "C" */
projects / mesa.git / blobdiff