#include "program/sampler.h"
}
+class src_reg;
+class dst_reg;
+
static int swizzle_for_size(int size);
/**
* This struct is a corresponding struct to Mesa prog_src_register, with
* wider fields.
*/
-typedef struct ir_to_mesa_src_reg {
- ir_to_mesa_src_reg(int file, int index, const glsl_type *type)
+class src_reg {
+public:
+ src_reg(gl_register_file file, int index, const glsl_type *type)
{
- this->file = (gl_register_file) file;
+ this->file = file;
this->index = index;
if (type && (type->is_scalar() || type->is_vector() || type->is_matrix()))
this->swizzle = swizzle_for_size(type->vector_elements);
this->reladdr = NULL;
}
- ir_to_mesa_src_reg()
+ src_reg()
{
this->file = PROGRAM_UNDEFINED;
this->index = 0;
this->reladdr = NULL;
}
+ explicit src_reg(dst_reg reg);
+
gl_register_file file; /**< PROGRAM_* from Mesa */
int index; /**< temporary index, VERT_ATTRIB_*, FRAG_ATTRIB_*, etc. */
GLuint swizzle; /**< SWIZZLE_XYZWONEZERO swizzles from Mesa. */
int negate; /**< NEGATE_XYZW mask from mesa */
/** Register index should be offset by the integer in this reg. */
- ir_to_mesa_src_reg *reladdr;
-} ir_to_mesa_src_reg;
+ src_reg *reladdr;
+};
+
+class dst_reg {
+public:
+ dst_reg(gl_register_file file, int writemask)
+ {
+ this->file = file;
+ this->index = 0;
+ this->writemask = writemask;
+ this->cond_mask = COND_TR;
+ this->reladdr = NULL;
+ }
+
+ dst_reg()
+ {
+ this->file = PROGRAM_UNDEFINED;
+ this->index = 0;
+ this->writemask = 0;
+ this->cond_mask = COND_TR;
+ this->reladdr = NULL;
+ }
+
+ explicit dst_reg(src_reg reg);
-typedef struct ir_to_mesa_dst_reg {
- int file; /**< PROGRAM_* from Mesa */
+ gl_register_file file; /**< PROGRAM_* from Mesa */
int index; /**< temporary index, VERT_ATTRIB_*, FRAG_ATTRIB_*, etc. */
int writemask; /**< Bitfield of WRITEMASK_[XYZW] */
GLuint cond_mask:4;
/** Register index should be offset by the integer in this reg. */
- ir_to_mesa_src_reg *reladdr;
-} ir_to_mesa_dst_reg;
+ src_reg *reladdr;
+};
-extern ir_to_mesa_src_reg ir_to_mesa_undef;
+src_reg::src_reg(dst_reg reg)
+{
+ this->file = reg.file;
+ this->index = reg.index;
+ this->swizzle = SWIZZLE_XYZW;
+ this->negate = 0;
+ this->reladdr = reg.reladdr;
+}
+
+dst_reg::dst_reg(src_reg reg)
+{
+ this->file = reg.file;
+ this->index = reg.index;
+ this->writemask = WRITEMASK_XYZW;
+ this->cond_mask = COND_TR;
+ this->reladdr = reg.reladdr;
+}
class ir_to_mesa_instruction : public exec_node {
public:
- /* Callers of this talloc-based new need not call delete. It's
- * easier to just talloc_free 'ctx' (or any of its ancestors). */
+ /* Callers of this ralloc-based new need not call delete. It's
+ * easier to just ralloc_free 'ctx' (or any of its ancestors). */
static void* operator new(size_t size, void *ctx)
{
void *node;
- node = talloc_zero_size(ctx, size);
+ node = rzalloc_size(ctx, size);
assert(node != NULL);
return node;
}
enum prog_opcode op;
- ir_to_mesa_dst_reg dst_reg;
- ir_to_mesa_src_reg src_reg[3];
+ dst_reg dst;
+ src_reg src[3];
/** Pointer to the ir source this tree came from for debugging */
ir_instruction *ir;
GLboolean cond_update;
int inst;
/** Storage for the return value. */
- ir_to_mesa_src_reg return_reg;
+ src_reg return_reg;
};
class ir_to_mesa_visitor : public ir_visitor {
function_entry *get_function_signature(ir_function_signature *sig);
- ir_to_mesa_src_reg get_temp(const glsl_type *type);
- void reladdr_to_temp(ir_instruction *ir,
- ir_to_mesa_src_reg *reg, int *num_reladdr);
+ src_reg get_temp(const glsl_type *type);
+ void reladdr_to_temp(ir_instruction *ir, src_reg *reg, int *num_reladdr);
- struct ir_to_mesa_src_reg src_reg_for_float(float val);
+ src_reg src_reg_for_float(float val);
/**
* \name Visit methods
virtual void visit(ir_if *);
/*@}*/
- struct ir_to_mesa_src_reg result;
+ src_reg result;
/** List of variable_storage */
exec_list variables;
/** List of ir_to_mesa_instruction */
exec_list instructions;
- ir_to_mesa_instruction *ir_to_mesa_emit_op0(ir_instruction *ir,
- enum prog_opcode op);
+ ir_to_mesa_instruction *emit(ir_instruction *ir, enum prog_opcode op);
- ir_to_mesa_instruction *ir_to_mesa_emit_op1(ir_instruction *ir,
- enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0);
+ ir_to_mesa_instruction *emit(ir_instruction *ir, enum prog_opcode op,
+ dst_reg dst, src_reg src0);
- ir_to_mesa_instruction *ir_to_mesa_emit_op2(ir_instruction *ir,
- enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0,
- ir_to_mesa_src_reg src1);
+ ir_to_mesa_instruction *emit(ir_instruction *ir, enum prog_opcode op,
+ dst_reg dst, src_reg src0, src_reg src1);
- ir_to_mesa_instruction *ir_to_mesa_emit_op3(ir_instruction *ir,
- enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0,
- ir_to_mesa_src_reg src1,
- ir_to_mesa_src_reg src2);
+ ir_to_mesa_instruction *emit(ir_instruction *ir, enum prog_opcode op,
+ dst_reg dst,
+ src_reg src0, src_reg src1, src_reg src2);
/**
* Emit the correct dot-product instruction for the type of arguments
- *
- * \sa ir_to_mesa_emit_op2
*/
- void ir_to_mesa_emit_dp(ir_instruction *ir,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0,
- ir_to_mesa_src_reg src1,
- unsigned elements);
-
- void ir_to_mesa_emit_scalar_op1(ir_instruction *ir,
- enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0);
-
- void ir_to_mesa_emit_scalar_op2(ir_instruction *ir,
- enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0,
- ir_to_mesa_src_reg src1);
+ ir_to_mesa_instruction * emit_dp(ir_instruction *ir,
+ dst_reg dst,
+ src_reg src0,
+ src_reg src1,
+ unsigned elements);
+
+ void emit_scalar(ir_instruction *ir, enum prog_opcode op,
+ dst_reg dst, src_reg src0);
+
+ void emit_scalar(ir_instruction *ir, enum prog_opcode op,
+ dst_reg dst, src_reg src0, src_reg src1);
void emit_scs(ir_instruction *ir, enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- const ir_to_mesa_src_reg &src);
+ dst_reg dst, const src_reg &src);
GLboolean try_emit_mad(ir_expression *ir,
int mul_operand);
void *mem_ctx;
};
-ir_to_mesa_src_reg ir_to_mesa_undef = ir_to_mesa_src_reg(PROGRAM_UNDEFINED, 0, NULL);
+src_reg undef_src = src_reg(PROGRAM_UNDEFINED, 0, NULL);
-ir_to_mesa_dst_reg ir_to_mesa_undef_dst = {
- PROGRAM_UNDEFINED, 0, SWIZZLE_NOOP, COND_TR, NULL,
-};
+dst_reg undef_dst = dst_reg(PROGRAM_UNDEFINED, SWIZZLE_NOOP);
-ir_to_mesa_dst_reg ir_to_mesa_address_reg = {
- PROGRAM_ADDRESS, 0, WRITEMASK_X, COND_TR, NULL
-};
-
-static void
-fail_link(struct gl_shader_program *prog, const char *fmt, ...) PRINTFLIKE(2, 3);
-
-static void
-fail_link(struct gl_shader_program *prog, const char *fmt, ...)
-{
- va_list args;
- va_start(args, fmt);
- prog->InfoLog = talloc_vasprintf_append(prog->InfoLog, fmt, args);
- va_end(args);
-
- prog->LinkStatus = GL_FALSE;
-}
+dst_reg address_reg = dst_reg(PROGRAM_ADDRESS, WRITEMASK_X);
static int
swizzle_for_size(int size)
}
ir_to_mesa_instruction *
-ir_to_mesa_visitor::ir_to_mesa_emit_op3(ir_instruction *ir,
- enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0,
- ir_to_mesa_src_reg src1,
- ir_to_mesa_src_reg src2)
+ir_to_mesa_visitor::emit(ir_instruction *ir, enum prog_opcode op,
+ dst_reg dst,
+ src_reg src0, src_reg src1, src_reg src2)
{
ir_to_mesa_instruction *inst = new(mem_ctx) ir_to_mesa_instruction();
int num_reladdr = 0;
reladdr_to_temp(ir, &src0, &num_reladdr);
if (dst.reladdr) {
- ir_to_mesa_emit_op1(ir, OPCODE_ARL, ir_to_mesa_address_reg,
- *dst.reladdr);
-
+ emit(ir, OPCODE_ARL, address_reg, *dst.reladdr);
num_reladdr--;
}
assert(num_reladdr == 0);
inst->op = op;
- inst->dst_reg = dst;
- inst->src_reg[0] = src0;
- inst->src_reg[1] = src1;
- inst->src_reg[2] = src2;
+ inst->dst = dst;
+ inst->src[0] = src0;
+ inst->src[1] = src1;
+ inst->src[2] = src2;
inst->ir = ir;
inst->function = NULL;
ir_to_mesa_instruction *
-ir_to_mesa_visitor::ir_to_mesa_emit_op2(ir_instruction *ir,
- enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0,
- ir_to_mesa_src_reg src1)
+ir_to_mesa_visitor::emit(ir_instruction *ir, enum prog_opcode op,
+ dst_reg dst, src_reg src0, src_reg src1)
{
- return ir_to_mesa_emit_op3(ir, op, dst, src0, src1, ir_to_mesa_undef);
+ return emit(ir, op, dst, src0, src1, undef_src);
}
ir_to_mesa_instruction *
-ir_to_mesa_visitor::ir_to_mesa_emit_op1(ir_instruction *ir,
- enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0)
+ir_to_mesa_visitor::emit(ir_instruction *ir, enum prog_opcode op,
+ dst_reg dst, src_reg src0)
{
assert(dst.writemask != 0);
- return ir_to_mesa_emit_op3(ir, op, dst,
- src0, ir_to_mesa_undef, ir_to_mesa_undef);
+ return emit(ir, op, dst, src0, undef_src, undef_src);
}
ir_to_mesa_instruction *
-ir_to_mesa_visitor::ir_to_mesa_emit_op0(ir_instruction *ir,
- enum prog_opcode op)
+ir_to_mesa_visitor::emit(ir_instruction *ir, enum prog_opcode op)
{
- return ir_to_mesa_emit_op3(ir, op, ir_to_mesa_undef_dst,
- ir_to_mesa_undef,
- ir_to_mesa_undef,
- ir_to_mesa_undef);
+ return emit(ir, op, undef_dst, undef_src, undef_src, undef_src);
}
-void
-ir_to_mesa_visitor::ir_to_mesa_emit_dp(ir_instruction *ir,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0,
- ir_to_mesa_src_reg src1,
- unsigned elements)
+ir_to_mesa_instruction *
+ir_to_mesa_visitor::emit_dp(ir_instruction *ir,
+ dst_reg dst, src_reg src0, src_reg src1,
+ unsigned elements)
{
static const gl_inst_opcode dot_opcodes[] = {
OPCODE_DP2, OPCODE_DP3, OPCODE_DP4
};
- ir_to_mesa_emit_op3(ir, dot_opcodes[elements - 2],
- dst, src0, src1, ir_to_mesa_undef);
-}
-
-inline ir_to_mesa_dst_reg
-ir_to_mesa_dst_reg_from_src(ir_to_mesa_src_reg reg)
-{
- ir_to_mesa_dst_reg dst_reg;
-
- dst_reg.file = reg.file;
- dst_reg.index = reg.index;
- dst_reg.writemask = WRITEMASK_XYZW;
- dst_reg.cond_mask = COND_TR;
- dst_reg.reladdr = reg.reladdr;
-
- return dst_reg;
-}
-
-inline ir_to_mesa_src_reg
-ir_to_mesa_src_reg_from_dst(ir_to_mesa_dst_reg reg)
-{
- return ir_to_mesa_src_reg(reg.file, reg.index, NULL);
+ return emit(ir, dot_opcodes[elements - 2], dst, src0, src1);
}
/**
* to produce dest channels.
*/
void
-ir_to_mesa_visitor::ir_to_mesa_emit_scalar_op2(ir_instruction *ir,
- enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg orig_src0,
- ir_to_mesa_src_reg orig_src1)
+ir_to_mesa_visitor::emit_scalar(ir_instruction *ir, enum prog_opcode op,
+ dst_reg dst,
+ src_reg orig_src0, src_reg orig_src1)
{
int i, j;
int done_mask = ~dst.writemask;
for (i = 0; i < 4; i++) {
GLuint this_mask = (1 << i);
ir_to_mesa_instruction *inst;
- ir_to_mesa_src_reg src0 = orig_src0;
- ir_to_mesa_src_reg src1 = orig_src1;
+ src_reg src0 = orig_src0;
+ src_reg src1 = orig_src1;
if (done_mask & this_mask)
continue;
src1.swizzle = MAKE_SWIZZLE4(src1_swiz, src1_swiz,
src1_swiz, src1_swiz);
- inst = ir_to_mesa_emit_op2(ir, op,
- dst,
- src0,
- src1);
- inst->dst_reg.writemask = this_mask;
+ inst = emit(ir, op, dst, src0, src1);
+ inst->dst.writemask = this_mask;
done_mask |= this_mask;
}
}
void
-ir_to_mesa_visitor::ir_to_mesa_emit_scalar_op1(ir_instruction *ir,
- enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- ir_to_mesa_src_reg src0)
+ir_to_mesa_visitor::emit_scalar(ir_instruction *ir, enum prog_opcode op,
+ dst_reg dst, src_reg src0)
{
- ir_to_mesa_src_reg undef = ir_to_mesa_undef;
+ src_reg undef = undef_src;
undef.swizzle = SWIZZLE_XXXX;
- ir_to_mesa_emit_scalar_op2(ir, op, dst, src0, undef);
+ emit_scalar(ir, op, dst, src0, undef);
}
/**
*/
void
ir_to_mesa_visitor::emit_scs(ir_instruction *ir, enum prog_opcode op,
- ir_to_mesa_dst_reg dst,
- const ir_to_mesa_src_reg &src)
+ dst_reg dst,
+ const src_reg &src)
{
/* Vertex programs cannot use the SCS opcode.
*/
if (this->prog->Target == GL_VERTEX_PROGRAM_ARB) {
- ir_to_mesa_emit_scalar_op1(ir, op, dst, src);
+ emit_scalar(ir, op, dst, src);
return;
}
const unsigned component = (op == OPCODE_SIN) ? 0 : 1;
const unsigned scs_mask = (1U << component);
int done_mask = ~dst.writemask;
- ir_to_mesa_src_reg tmp;
+ src_reg tmp;
assert(op == OPCODE_SIN || op == OPCODE_COS);
for (unsigned i = 0; i < 4; i++) {
unsigned this_mask = (1U << i);
- ir_to_mesa_src_reg src0 = src;
+ src_reg src0 = src;
if ((done_mask & this_mask) != 0)
continue;
if (this_mask != scs_mask) {
ir_to_mesa_instruction *inst;
- ir_to_mesa_dst_reg tmp_dst = ir_to_mesa_dst_reg_from_src(tmp);
+ dst_reg tmp_dst = dst_reg(tmp);
/* Emit the SCS instruction.
*/
- inst = ir_to_mesa_emit_op1(ir, OPCODE_SCS, tmp_dst, src0);
- inst->dst_reg.writemask = scs_mask;
+ inst = emit(ir, OPCODE_SCS, tmp_dst, src0);
+ inst->dst.writemask = scs_mask;
/* Move the result of the SCS instruction to the desired location in
* the destination.
*/
tmp.swizzle = MAKE_SWIZZLE4(component, component,
component, component);
- inst = ir_to_mesa_emit_op1(ir, OPCODE_SCS, dst, tmp);
- inst->dst_reg.writemask = this_mask;
+ inst = emit(ir, OPCODE_SCS, dst, tmp);
+ inst->dst.writemask = this_mask;
} else {
/* Emit the SCS instruction to write directly to the destination.
*/
- ir_to_mesa_instruction *inst =
- ir_to_mesa_emit_op1(ir, OPCODE_SCS, dst, src0);
- inst->dst_reg.writemask = scs_mask;
+ ir_to_mesa_instruction *inst = emit(ir, OPCODE_SCS, dst, src0);
+ inst->dst.writemask = scs_mask;
}
done_mask |= this_mask;
}
}
-struct ir_to_mesa_src_reg
+struct src_reg
ir_to_mesa_visitor::src_reg_for_float(float val)
{
- ir_to_mesa_src_reg src_reg(PROGRAM_CONSTANT, -1, NULL);
+ src_reg src(PROGRAM_CONSTANT, -1, NULL);
- src_reg.index = _mesa_add_unnamed_constant(this->prog->Parameters,
- &val, 1, &src_reg.swizzle);
+ src.index = _mesa_add_unnamed_constant(this->prog->Parameters,
+ (const gl_constant_value *)&val, 1, &src.swizzle);
- return src_reg;
+ return src;
}
static int
return 1;
}
case GLSL_TYPE_ARRAY:
+ assert(type->length > 0);
return type_size(type->fields.array) * type->length;
case GLSL_TYPE_STRUCT:
size = 0;
* storage). Actual register allocation for the Mesa VM occurs in a
* pass over the Mesa IR later.
*/
-ir_to_mesa_src_reg
+src_reg
ir_to_mesa_visitor::get_temp(const glsl_type *type)
{
- ir_to_mesa_src_reg src_reg;
- int swizzle[4];
- int i;
+ src_reg src;
- src_reg.file = PROGRAM_TEMPORARY;
- src_reg.index = next_temp;
- src_reg.reladdr = NULL;
+ src.file = PROGRAM_TEMPORARY;
+ src.index = next_temp;
+ src.reladdr = NULL;
next_temp += type_size(type);
if (type->is_array() || type->is_record()) {
- src_reg.swizzle = SWIZZLE_NOOP;
+ src.swizzle = SWIZZLE_NOOP;
} else {
- for (i = 0; i < type->vector_elements; i++)
- swizzle[i] = i;
- for (; i < 4; i++)
- swizzle[i] = type->vector_elements - 1;
- src_reg.swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1],
- swizzle[2], swizzle[3]);
+ src.swizzle = swizzle_for_size(type->vector_elements);
}
- src_reg.negate = 0;
+ src.negate = 0;
- return src_reg;
+ return src;
}
variable_storage *
fp->OriginUpperLeft = ir->origin_upper_left;
fp->PixelCenterInteger = ir->pixel_center_integer;
+
+ } else if (strcmp(ir->name, "gl_FragDepth") == 0) {
+ struct gl_fragment_program *fp = (struct gl_fragment_program *)this->prog;
+ switch (ir->depth_layout) {
+ case ir_depth_layout_none:
+ fp->FragDepthLayout = FRAG_DEPTH_LAYOUT_NONE;
+ break;
+ case ir_depth_layout_any:
+ fp->FragDepthLayout = FRAG_DEPTH_LAYOUT_ANY;
+ break;
+ case ir_depth_layout_greater:
+ fp->FragDepthLayout = FRAG_DEPTH_LAYOUT_GREATER;
+ break;
+ case ir_depth_layout_less:
+ fp->FragDepthLayout = FRAG_DEPTH_LAYOUT_LESS;
+ break;
+ case ir_depth_layout_unchanged:
+ fp->FragDepthLayout = FRAG_DEPTH_LAYOUT_UNCHANGED;
+ break;
+ default:
+ assert(0);
+ break;
+ }
}
if (ir->mode == ir_var_uniform && strncmp(ir->name, "gl_", 3) == 0) {
unsigned int i;
- const struct gl_builtin_uniform_desc *statevar;
-
- for (i = 0; _mesa_builtin_uniform_desc[i].name; i++) {
- if (strcmp(ir->name, _mesa_builtin_uniform_desc[i].name) == 0)
- break;
- }
-
- if (!_mesa_builtin_uniform_desc[i].name) {
- fail_link(this->shader_program,
- "Failed to find builtin uniform `%s'\n", ir->name);
- return;
- }
-
- statevar = &_mesa_builtin_uniform_desc[i];
-
- int array_count;
- if (ir->type->is_array()) {
- array_count = ir->type->length;
- } else {
- array_count = 1;
- }
+ const ir_state_slot *const slots = ir->state_slots;
+ assert(ir->state_slots != NULL);
/* Check if this statevar's setup in the STATE file exactly
* matches how we'll want to reference it as a
* temporary storage and hope that it'll get copy-propagated
* out.
*/
- for (i = 0; i < statevar->num_elements; i++) {
- if (statevar->elements[i].swizzle != SWIZZLE_XYZW) {
+ for (i = 0; i < ir->num_state_slots; i++) {
+ if (slots[i].swizzle != SWIZZLE_XYZW) {
break;
}
}
struct variable_storage *storage;
- ir_to_mesa_dst_reg dst;
- if (i == statevar->num_elements) {
+ dst_reg dst;
+ if (i == ir->num_state_slots) {
/* We'll set the index later. */
storage = new(mem_ctx) variable_storage(ir, PROGRAM_STATE_VAR, -1);
this->variables.push_tail(storage);
- dst = ir_to_mesa_undef_dst;
+ dst = undef_dst;
} else {
+ /* The variable_storage constructor allocates slots based on the size
+ * of the type. However, this had better match the number of state
+ * elements that we're going to copy into the new temporary.
+ */
+ assert((int) ir->num_state_slots == type_size(ir->type));
+
storage = new(mem_ctx) variable_storage(ir, PROGRAM_TEMPORARY,
this->next_temp);
this->variables.push_tail(storage);
this->next_temp += type_size(ir->type);
- dst = ir_to_mesa_dst_reg_from_src(ir_to_mesa_src_reg(PROGRAM_TEMPORARY,
- storage->index,
- NULL));
+ dst = dst_reg(src_reg(PROGRAM_TEMPORARY, storage->index, NULL));
}
- for (int a = 0; a < array_count; a++) {
- for (unsigned int i = 0; i < statevar->num_elements; i++) {
- struct gl_builtin_uniform_element *element = &statevar->elements[i];
- int tokens[STATE_LENGTH];
+ for (unsigned int i = 0; i < ir->num_state_slots; i++) {
+ int index = _mesa_add_state_reference(this->prog->Parameters,
+ (gl_state_index *)slots[i].tokens);
- memcpy(tokens, element->tokens, sizeof(element->tokens));
- if (ir->type->is_array()) {
- tokens[1] = a;
- }
-
- int index = _mesa_add_state_reference(this->prog->Parameters,
- (gl_state_index *)tokens);
-
- if (storage->file == PROGRAM_STATE_VAR) {
- if (storage->index == -1) {
- storage->index = index;
- } else {
- assert(index ==
- (int)(storage->index + a * statevar->num_elements + i));
- }
+ if (storage->file == PROGRAM_STATE_VAR) {
+ if (storage->index == -1) {
+ storage->index = index;
} else {
- ir_to_mesa_src_reg src(PROGRAM_STATE_VAR, index, NULL);
- src.swizzle = element->swizzle;
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, dst, src);
- /* even a float takes up a whole vec4 reg in a struct/array. */
- dst.index++;
+ assert(index == storage->index + (int)i);
}
+ } else {
+ src_reg src(PROGRAM_STATE_VAR, index, NULL);
+ src.swizzle = slots[i].swizzle;
+ emit(ir, OPCODE_MOV, dst, src);
+ /* even a float takes up a whole vec4 reg in a struct/array. */
+ dst.index++;
}
}
+
if (storage->file == PROGRAM_TEMPORARY &&
- dst.index != storage->index + type_size(ir->type)) {
- fail_link(this->shader_program,
- "failed to load builtin uniform `%s' (%d/%d regs loaded)\n",
- ir->name, dst.index - storage->index,
- type_size(ir->type));
+ dst.index != storage->index + (int) ir->num_state_slots) {
+ linker_error(this->shader_program,
+ "failed to load builtin uniform `%s' "
+ "(%d/%d regs loaded)\n",
+ ir->name, dst.index - storage->index,
+ type_size(ir->type));
}
}
}
ir_dereference_variable *counter = NULL;
if (ir->counter != NULL)
- counter = new(ir) ir_dereference_variable(ir->counter);
+ counter = new(mem_ctx) ir_dereference_variable(ir->counter);
if (ir->from != NULL) {
assert(ir->counter != NULL);
- ir_assignment *a = new(ir) ir_assignment(counter, ir->from, NULL);
+ ir_assignment *a =
+ new(mem_ctx) ir_assignment(counter, ir->from, NULL);
a->accept(this);
- delete a;
}
- ir_to_mesa_emit_op0(NULL, OPCODE_BGNLOOP);
+ emit(NULL, OPCODE_BGNLOOP);
if (ir->to) {
ir_expression *e =
- new(ir) ir_expression(ir->cmp, glsl_type::bool_type,
- counter, ir->to);
- ir_if *if_stmt = new(ir) ir_if(e);
+ new(mem_ctx) ir_expression(ir->cmp, glsl_type::bool_type,
+ counter, ir->to);
+ ir_if *if_stmt = new(mem_ctx) ir_if(e);
- ir_loop_jump *brk = new(ir) ir_loop_jump(ir_loop_jump::jump_break);
+ ir_loop_jump *brk =
+ new(mem_ctx) ir_loop_jump(ir_loop_jump::jump_break);
if_stmt->then_instructions.push_tail(brk);
if_stmt->accept(this);
-
- delete if_stmt;
- delete e;
- delete brk;
}
visit_exec_list(&ir->body_instructions, this);
if (ir->increment) {
ir_expression *e =
- new(ir) ir_expression(ir_binop_add, counter->type,
- counter, ir->increment);
+ new(mem_ctx) ir_expression(ir_binop_add, counter->type,
+ counter, ir->increment);
- ir_assignment *a = new(ir) ir_assignment(counter, e, NULL);
+ ir_assignment *a =
+ new(mem_ctx) ir_assignment(counter, e, NULL);
a->accept(this);
- delete a;
- delete e;
}
- ir_to_mesa_emit_op0(NULL, OPCODE_ENDLOOP);
+ emit(NULL, OPCODE_ENDLOOP);
}
void
{
switch (ir->mode) {
case ir_loop_jump::jump_break:
- ir_to_mesa_emit_op0(NULL, OPCODE_BRK);
+ emit(NULL, OPCODE_BRK);
break;
case ir_loop_jump::jump_continue:
- ir_to_mesa_emit_op0(NULL, OPCODE_CONT);
+ emit(NULL, OPCODE_CONT);
break;
}
}
ir_to_mesa_visitor::try_emit_mad(ir_expression *ir, int mul_operand)
{
int nonmul_operand = 1 - mul_operand;
- ir_to_mesa_src_reg a, b, c;
+ src_reg a, b, c;
ir_expression *expr = ir->operands[mul_operand]->as_expression();
if (!expr || expr->operation != ir_binop_mul)
c = this->result;
this->result = get_temp(ir->type);
- ir_to_mesa_emit_op3(ir, OPCODE_MAD,
- ir_to_mesa_dst_reg_from_src(this->result), a, b, c);
+ emit(ir, OPCODE_MAD, dst_reg(this->result), a, b, c);
return true;
}
return false;
sat_src->accept(this);
- ir_to_mesa_src_reg src = this->result;
+ src_reg src = this->result;
- this->result = get_temp(ir->type);
- ir_to_mesa_instruction *inst;
- inst = ir_to_mesa_emit_op1(ir, OPCODE_MOV,
- ir_to_mesa_dst_reg_from_src(this->result),
- src);
- inst->saturate = true;
+ /* If we generated an expression instruction into a temporary in
+ * processing the saturate's operand, apply the saturate to that
+ * instruction. Otherwise, generate a MOV to do the saturate.
+ *
+ * Note that we have to be careful to only do this optimization if
+ * the instruction in question was what generated src->result. For
+ * example, ir_dereference_array might generate a MUL instruction
+ * to create the reladdr, and return us a src reg using that
+ * reladdr. That MUL result is not the value we're trying to
+ * saturate.
+ */
+ ir_expression *sat_src_expr = sat_src->as_expression();
+ ir_to_mesa_instruction *new_inst;
+ new_inst = (ir_to_mesa_instruction *)this->instructions.get_tail();
+ if (sat_src_expr && (sat_src_expr->operation == ir_binop_mul ||
+ sat_src_expr->operation == ir_binop_add ||
+ sat_src_expr->operation == ir_binop_dot)) {
+ new_inst->saturate = true;
+ } else {
+ this->result = get_temp(ir->type);
+ ir_to_mesa_instruction *inst;
+ inst = emit(ir, OPCODE_MOV, dst_reg(this->result), src);
+ inst->saturate = true;
+ }
return true;
}
void
ir_to_mesa_visitor::reladdr_to_temp(ir_instruction *ir,
- ir_to_mesa_src_reg *reg, int *num_reladdr)
+ src_reg *reg, int *num_reladdr)
{
if (!reg->reladdr)
return;
- ir_to_mesa_emit_op1(ir, OPCODE_ARL, ir_to_mesa_address_reg, *reg->reladdr);
+ emit(ir, OPCODE_ARL, address_reg, *reg->reladdr);
if (*num_reladdr != 1) {
- ir_to_mesa_src_reg temp = get_temp(glsl_type::vec4_type);
+ src_reg temp = get_temp(glsl_type::vec4_type);
- ir_to_mesa_emit_op1(ir, OPCODE_MOV,
- ir_to_mesa_dst_reg_from_src(temp), *reg);
+ emit(ir, OPCODE_MOV, dst_reg(temp), *reg);
*reg = temp;
}
exit(1);
}
- ir_to_mesa_src_reg src;
+ src_reg src;
src = this->result;
src.swizzle = MAKE_SWIZZLE4(components[0],
/* Storage for our result. Ideally for an assignment we'd be using the
* actual storage for the result here, instead.
*/
- const ir_to_mesa_src_reg result_src = get_temp(ir->type);
- ir_to_mesa_dst_reg result_dst = ir_to_mesa_dst_reg_from_src(result_src);
+ const src_reg result_src = get_temp(ir->type);
+ dst_reg result_dst = dst_reg(result_src);
/* Limit writes to the channels that will be used by result_src later.
* This does limit this temp's use as a temporary for multi-instruction
*/
result_dst.writemask = (1 << ir->type->vector_elements) - 1;
- ir_to_mesa_emit_op1(ir, OPCODE_SWZ, result_dst, src);
+ emit(ir, OPCODE_SWZ, result_dst, src);
this->result = result_src;
}
ir_to_mesa_visitor::visit(ir_expression *ir)
{
unsigned int operand;
- struct ir_to_mesa_src_reg op[Elements(ir->operands)];
- struct ir_to_mesa_src_reg result_src;
- struct ir_to_mesa_dst_reg result_dst;
+ src_reg op[Elements(ir->operands)];
+ src_reg result_src;
+ dst_reg result_dst;
/* Quick peephole: Emit OPCODE_MAD(a, b, c) instead of ADD(MUL(a, b), c)
*/
*/
result_src = get_temp(ir->type);
/* convenience for the emit functions below. */
- result_dst = ir_to_mesa_dst_reg_from_src(result_src);
+ result_dst = dst_reg(result_src);
/* Limit writes to the channels that will be used by result_src later.
* This does limit this temp's use as a temporary for multi-instruction
* sequences.
switch (ir->operation) {
case ir_unop_logic_not:
- ir_to_mesa_emit_op2(ir, OPCODE_SEQ, result_dst,
- op[0], src_reg_for_float(0.0));
+ /* Previously 'SEQ dst, src, 0.0' was used for this. However, many
+ * older GPUs implement SEQ using multiple instructions (i915 uses two
+ * SGE instructions and a MUL instruction). Since our logic values are
+ * 0.0 and 1.0, 1-x also implements !x.
+ */
+ op[0].negate = ~op[0].negate;
+ emit(ir, OPCODE_ADD, result_dst, op[0], src_reg_for_float(1.0));
break;
case ir_unop_neg:
op[0].negate = ~op[0].negate;
result_src = op[0];
break;
case ir_unop_abs:
- ir_to_mesa_emit_op1(ir, OPCODE_ABS, result_dst, op[0]);
+ emit(ir, OPCODE_ABS, result_dst, op[0]);
break;
case ir_unop_sign:
- ir_to_mesa_emit_op1(ir, OPCODE_SSG, result_dst, op[0]);
+ emit(ir, OPCODE_SSG, result_dst, op[0]);
break;
case ir_unop_rcp:
- ir_to_mesa_emit_scalar_op1(ir, OPCODE_RCP, result_dst, op[0]);
+ emit_scalar(ir, OPCODE_RCP, result_dst, op[0]);
break;
case ir_unop_exp2:
- ir_to_mesa_emit_scalar_op1(ir, OPCODE_EX2, result_dst, op[0]);
+ emit_scalar(ir, OPCODE_EX2, result_dst, op[0]);
break;
case ir_unop_exp:
case ir_unop_log:
assert(!"not reached: should be handled by ir_explog_to_explog2");
break;
case ir_unop_log2:
- ir_to_mesa_emit_scalar_op1(ir, OPCODE_LG2, result_dst, op[0]);
+ emit_scalar(ir, OPCODE_LG2, result_dst, op[0]);
break;
case ir_unop_sin:
- ir_to_mesa_emit_scalar_op1(ir, OPCODE_SIN, result_dst, op[0]);
+ emit_scalar(ir, OPCODE_SIN, result_dst, op[0]);
break;
case ir_unop_cos:
- ir_to_mesa_emit_scalar_op1(ir, OPCODE_COS, result_dst, op[0]);
+ emit_scalar(ir, OPCODE_COS, result_dst, op[0]);
break;
case ir_unop_sin_reduced:
emit_scs(ir, OPCODE_SIN, result_dst, op[0]);
break;
case ir_unop_dFdx:
- ir_to_mesa_emit_op1(ir, OPCODE_DDX, result_dst, op[0]);
+ emit(ir, OPCODE_DDX, result_dst, op[0]);
break;
case ir_unop_dFdy:
- ir_to_mesa_emit_op1(ir, OPCODE_DDY, result_dst, op[0]);
+ emit(ir, OPCODE_DDY, result_dst, op[0]);
break;
case ir_unop_noise: {
+ (ir->operands[0]->type->vector_elements) - 1);
assert((opcode >= OPCODE_NOISE1) && (opcode <= OPCODE_NOISE4));
- ir_to_mesa_emit_op1(ir, opcode, result_dst, op[0]);
+ emit(ir, opcode, result_dst, op[0]);
break;
}
case ir_binop_add:
- ir_to_mesa_emit_op2(ir, OPCODE_ADD, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_ADD, result_dst, op[0], op[1]);
break;
case ir_binop_sub:
- ir_to_mesa_emit_op2(ir, OPCODE_SUB, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_SUB, result_dst, op[0], op[1]);
break;
case ir_binop_mul:
- ir_to_mesa_emit_op2(ir, OPCODE_MUL, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_MUL, result_dst, op[0], op[1]);
break;
case ir_binop_div:
assert(!"not reached: should be handled by ir_div_to_mul_rcp");
break;
case ir_binop_less:
- ir_to_mesa_emit_op2(ir, OPCODE_SLT, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_SLT, result_dst, op[0], op[1]);
break;
case ir_binop_greater:
- ir_to_mesa_emit_op2(ir, OPCODE_SGT, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_SGT, result_dst, op[0], op[1]);
break;
case ir_binop_lequal:
- ir_to_mesa_emit_op2(ir, OPCODE_SLE, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_SLE, result_dst, op[0], op[1]);
break;
case ir_binop_gequal:
- ir_to_mesa_emit_op2(ir, OPCODE_SGE, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_SGE, result_dst, op[0], op[1]);
break;
case ir_binop_equal:
- ir_to_mesa_emit_op2(ir, OPCODE_SEQ, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_SEQ, result_dst, op[0], op[1]);
break;
case ir_binop_nequal:
- ir_to_mesa_emit_op2(ir, OPCODE_SNE, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_SNE, 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()) {
- ir_to_mesa_src_reg temp = get_temp(glsl_type::vec4_type);
- ir_to_mesa_emit_op2(ir, OPCODE_SNE,
- ir_to_mesa_dst_reg_from_src(temp), op[0], op[1]);
- ir_to_mesa_emit_dp(ir, result_dst, temp, temp, vector_elements);
- ir_to_mesa_emit_op2(ir, OPCODE_SEQ,
- result_dst, result_src, src_reg_for_float(0.0));
+ src_reg temp = get_temp(glsl_type::vec4_type);
+ emit(ir, OPCODE_SNE, dst_reg(temp), op[0], op[1]);
+ emit_dp(ir, result_dst, temp, temp, vector_elements);
+ emit(ir, OPCODE_SEQ, result_dst, result_src, src_reg_for_float(0.0));
} else {
- ir_to_mesa_emit_op2(ir, OPCODE_SEQ, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_SEQ, result_dst, op[0], op[1]);
}
break;
case ir_binop_any_nequal:
/* "!=" operator producing a scalar boolean. */
if (ir->operands[0]->type->is_vector() ||
ir->operands[1]->type->is_vector()) {
- ir_to_mesa_src_reg temp = get_temp(glsl_type::vec4_type);
- ir_to_mesa_emit_op2(ir, OPCODE_SNE,
- ir_to_mesa_dst_reg_from_src(temp), op[0], op[1]);
- ir_to_mesa_emit_dp(ir, result_dst, temp, temp, vector_elements);
- ir_to_mesa_emit_op2(ir, OPCODE_SNE,
- result_dst, result_src, src_reg_for_float(0.0));
+ src_reg temp = get_temp(glsl_type::vec4_type);
+ emit(ir, OPCODE_SNE, dst_reg(temp), op[0], op[1]);
+ emit_dp(ir, result_dst, temp, temp, vector_elements);
+ emit(ir, OPCODE_SNE, result_dst, result_src, src_reg_for_float(0.0));
} else {
- ir_to_mesa_emit_op2(ir, OPCODE_SNE, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_SNE, result_dst, op[0], op[1]);
}
break;
- case ir_unop_any:
+ case ir_unop_any: {
assert(ir->operands[0]->type->is_vector());
- ir_to_mesa_emit_dp(ir, result_dst, op[0], op[0],
- ir->operands[0]->type->vector_elements);
- ir_to_mesa_emit_op2(ir, OPCODE_SNE,
- result_dst, result_src, src_reg_for_float(0.0));
+
+ /* 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:
- ir_to_mesa_emit_op2(ir, OPCODE_SNE, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_SNE, result_dst, op[0], op[1]);
break;
- case ir_binop_logic_or:
- /* This could be a saturated add and skip the SNE. */
- ir_to_mesa_emit_op2(ir, OPCODE_ADD,
- result_dst,
- op[0], op[1]);
-
- ir_to_mesa_emit_op2(ir, OPCODE_SNE,
- result_dst,
- result_src, src_reg_for_float(0.0));
+ 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.
+ */
+ 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));
+ }
break;
+ }
case ir_binop_logic_and:
/* the bool args are stored as float 0.0 or 1.0, so "mul" gives us "and". */
- ir_to_mesa_emit_op2(ir, OPCODE_MUL,
- result_dst,
- op[0], op[1]);
+ emit(ir, OPCODE_MUL, result_dst, op[0], op[1]);
break;
case ir_binop_dot:
assert(ir->operands[0]->type->is_vector());
assert(ir->operands[0]->type == ir->operands[1]->type);
- ir_to_mesa_emit_dp(ir, result_dst, op[0], op[1],
- ir->operands[0]->type->vector_elements);
+ emit_dp(ir, result_dst, op[0], op[1],
+ ir->operands[0]->type->vector_elements);
break;
case ir_unop_sqrt:
/* sqrt(x) = x * rsq(x). */
- ir_to_mesa_emit_scalar_op1(ir, OPCODE_RSQ, result_dst, op[0]);
- ir_to_mesa_emit_op2(ir, OPCODE_MUL, result_dst, result_src, op[0]);
+ emit_scalar(ir, OPCODE_RSQ, result_dst, op[0]);
+ emit(ir, OPCODE_MUL, result_dst, result_src, op[0]);
/* For incoming channels <= 0, set the result to 0. */
op[0].negate = ~op[0].negate;
- ir_to_mesa_emit_op3(ir, OPCODE_CMP, result_dst,
+ emit(ir, OPCODE_CMP, result_dst,
op[0], result_src, src_reg_for_float(0.0));
break;
case ir_unop_rsq:
- ir_to_mesa_emit_scalar_op1(ir, OPCODE_RSQ, result_dst, op[0]);
+ emit_scalar(ir, OPCODE_RSQ, result_dst, op[0]);
break;
case ir_unop_i2f:
+ case ir_unop_u2f:
case ir_unop_b2f:
case ir_unop_b2i:
+ case ir_unop_i2u:
+ case ir_unop_u2i:
/* Mesa IR lacks types, ints are stored as truncated floats. */
result_src = op[0];
break;
case ir_unop_f2i:
- ir_to_mesa_emit_op1(ir, OPCODE_TRUNC, result_dst, op[0]);
+ emit(ir, OPCODE_TRUNC, result_dst, op[0]);
break;
case ir_unop_f2b:
case ir_unop_i2b:
- ir_to_mesa_emit_op2(ir, OPCODE_SNE, result_dst,
+ emit(ir, OPCODE_SNE, result_dst,
op[0], src_reg_for_float(0.0));
break;
case ir_unop_trunc:
- ir_to_mesa_emit_op1(ir, OPCODE_TRUNC, result_dst, op[0]);
+ emit(ir, OPCODE_TRUNC, result_dst, op[0]);
break;
case ir_unop_ceil:
op[0].negate = ~op[0].negate;
- ir_to_mesa_emit_op1(ir, OPCODE_FLR, result_dst, op[0]);
+ emit(ir, OPCODE_FLR, result_dst, op[0]);
result_src.negate = ~result_src.negate;
break;
case ir_unop_floor:
- ir_to_mesa_emit_op1(ir, OPCODE_FLR, result_dst, op[0]);
+ emit(ir, OPCODE_FLR, result_dst, op[0]);
break;
case ir_unop_fract:
- ir_to_mesa_emit_op1(ir, OPCODE_FRC, result_dst, op[0]);
+ emit(ir, OPCODE_FRC, result_dst, op[0]);
break;
case ir_binop_min:
- ir_to_mesa_emit_op2(ir, OPCODE_MIN, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_MIN, result_dst, op[0], op[1]);
break;
case ir_binop_max:
- ir_to_mesa_emit_op2(ir, OPCODE_MAX, result_dst, op[0], op[1]);
+ emit(ir, OPCODE_MAX, result_dst, op[0], op[1]);
break;
case ir_binop_pow:
- ir_to_mesa_emit_scalar_op2(ir, OPCODE_POW, result_dst, op[0], op[1]);
+ emit_scalar(ir, OPCODE_POW, result_dst, op[0], op[1]);
break;
case ir_unop_bit_not:
- case ir_unop_u2f:
case ir_binop_lshift:
case ir_binop_rshift:
case ir_binop_bit_and:
void
ir_to_mesa_visitor::visit(ir_swizzle *ir)
{
- ir_to_mesa_src_reg src_reg;
+ src_reg src;
int i;
int swizzle[4];
*/
ir->val->accept(this);
- src_reg = this->result;
- assert(src_reg.file != PROGRAM_UNDEFINED);
+ src = this->result;
+ assert(src.file != PROGRAM_UNDEFINED);
for (i = 0; i < 4; i++) {
if (i < ir->type->vector_elements) {
switch (i) {
case 0:
- swizzle[i] = GET_SWZ(src_reg.swizzle, ir->mask.x);
+ swizzle[i] = GET_SWZ(src.swizzle, ir->mask.x);
break;
case 1:
- swizzle[i] = GET_SWZ(src_reg.swizzle, ir->mask.y);
+ swizzle[i] = GET_SWZ(src.swizzle, ir->mask.y);
break;
case 2:
- swizzle[i] = GET_SWZ(src_reg.swizzle, ir->mask.z);
+ swizzle[i] = GET_SWZ(src.swizzle, ir->mask.z);
break;
case 3:
- swizzle[i] = GET_SWZ(src_reg.swizzle, ir->mask.w);
+ swizzle[i] = GET_SWZ(src.swizzle, ir->mask.w);
break;
}
} else {
}
}
- src_reg.swizzle = MAKE_SWIZZLE4(swizzle[0],
- swizzle[1],
- swizzle[2],
- swizzle[3]);
+ src.swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], swizzle[3]);
- this->result = src_reg;
+ this->result = src;
}
void
ir_to_mesa_visitor::visit(ir_dereference_variable *ir)
{
variable_storage *entry = find_variable_storage(ir->var);
+ ir_variable *var = ir->var;
if (!entry) {
- switch (ir->var->mode) {
+ switch (var->mode) {
case ir_var_uniform:
- entry = new(mem_ctx) variable_storage(ir->var, PROGRAM_UNIFORM,
- ir->var->location);
+ entry = new(mem_ctx) variable_storage(var, PROGRAM_UNIFORM,
+ var->location);
this->variables.push_tail(entry);
break;
case ir_var_in:
- case ir_var_out:
case ir_var_inout:
/* The linker assigns locations for varyings and attributes,
- * including deprecated builtins (like gl_Color), user-assign
- * generic attributes (glBindVertexLocation), and
- * user-defined varyings.
+ * including deprecated builtins (like gl_Color),
+ * user-assigned generic attributes (glBindVertexLocation),
+ * and user-defined varyings.
*
* FINISHME: We would hit this path for function arguments. Fix!
*/
- assert(ir->var->location != -1);
- if (ir->var->mode == ir_var_in ||
- ir->var->mode == ir_var_inout) {
- entry = new(mem_ctx) variable_storage(ir->var,
- PROGRAM_INPUT,
- ir->var->location);
-
- if (this->prog->Target == GL_VERTEX_PROGRAM_ARB &&
- ir->var->location >= VERT_ATTRIB_GENERIC0) {
- _mesa_add_attribute(prog->Attributes,
- ir->var->name,
- _mesa_sizeof_glsl_type(ir->var->type->gl_type),
- ir->var->type->gl_type,
- ir->var->location - VERT_ATTRIB_GENERIC0);
- }
- } else {
- entry = new(mem_ctx) variable_storage(ir->var,
- PROGRAM_OUTPUT,
- ir->var->location);
- }
-
+ assert(var->location != -1);
+ entry = new(mem_ctx) variable_storage(var,
+ PROGRAM_INPUT,
+ var->location);
+ if (this->prog->Target == GL_VERTEX_PROGRAM_ARB &&
+ var->location >= VERT_ATTRIB_GENERIC0) {
+ _mesa_add_attribute(this->prog->Attributes,
+ var->name,
+ _mesa_sizeof_glsl_type(var->type->gl_type),
+ var->type->gl_type,
+ var->location - VERT_ATTRIB_GENERIC0);
+ }
+ break;
+ case ir_var_out:
+ assert(var->location != -1);
+ entry = new(mem_ctx) variable_storage(var,
+ PROGRAM_OUTPUT,
+ var->location);
break;
+ case ir_var_system_value:
+ entry = new(mem_ctx) variable_storage(var,
+ PROGRAM_SYSTEM_VALUE,
+ var->location);
+ break;
case ir_var_auto:
case ir_var_temporary:
- entry = new(mem_ctx) variable_storage(ir->var, PROGRAM_TEMPORARY,
+ entry = new(mem_ctx) variable_storage(var, PROGRAM_TEMPORARY,
this->next_temp);
this->variables.push_tail(entry);
- next_temp += type_size(ir->var->type);
+ next_temp += type_size(var->type);
break;
}
if (!entry) {
- printf("Failed to make storage for %s\n", ir->var->name);
+ printf("Failed to make storage for %s\n", var->name);
exit(1);
}
}
- this->result = ir_to_mesa_src_reg(entry->file, entry->index, ir->var->type);
+ this->result = src_reg(entry->file, entry->index, var->type);
}
void
ir_to_mesa_visitor::visit(ir_dereference_array *ir)
{
ir_constant *index;
- ir_to_mesa_src_reg src_reg;
+ src_reg src;
int element_size = type_size(ir->type);
index = ir->array_index->constant_expression_value();
ir->array->accept(this);
- src_reg = this->result;
+ src = this->result;
if (index) {
- src_reg.index += index->value.i[0] * element_size;
+ src.index += index->value.i[0] * element_size;
} else {
- ir_to_mesa_src_reg array_base = this->result;
/* Variable index array dereference. It eats the "vec4" of the
* base of the array and an index that offsets the Mesa register
* index.
*/
ir->array_index->accept(this);
- ir_to_mesa_src_reg index_reg;
+ src_reg index_reg;
if (element_size == 1) {
index_reg = this->result;
} else {
index_reg = get_temp(glsl_type::float_type);
- ir_to_mesa_emit_op2(ir, OPCODE_MUL,
- ir_to_mesa_dst_reg_from_src(index_reg),
- this->result, src_reg_for_float(element_size));
+ emit(ir, OPCODE_MUL, dst_reg(index_reg),
+ this->result, src_reg_for_float(element_size));
+ }
+
+ /* If there was already a relative address register involved, add the
+ * new and the old together to get the new offset.
+ */
+ if (src.reladdr != NULL) {
+ src_reg accum_reg = get_temp(glsl_type::float_type);
+
+ emit(ir, OPCODE_ADD, dst_reg(accum_reg),
+ index_reg, *src.reladdr);
+
+ index_reg = accum_reg;
}
- src_reg.reladdr = talloc(mem_ctx, ir_to_mesa_src_reg);
- memcpy(src_reg.reladdr, &index_reg, sizeof(index_reg));
+ src.reladdr = ralloc(mem_ctx, src_reg);
+ memcpy(src.reladdr, &index_reg, sizeof(index_reg));
}
/* If the type is smaller than a vec4, replicate the last channel out. */
if (ir->type->is_scalar() || ir->type->is_vector())
- src_reg.swizzle = swizzle_for_size(ir->type->vector_elements);
+ src.swizzle = swizzle_for_size(ir->type->vector_elements);
else
- src_reg.swizzle = SWIZZLE_NOOP;
+ src.swizzle = SWIZZLE_NOOP;
- this->result = src_reg;
+ this->result = src;
}
void
* instead of potentially using a temporary like we might with the
* ir_dereference handler.
*/
-static struct ir_to_mesa_dst_reg
+static dst_reg
get_assignment_lhs(ir_dereference *ir, ir_to_mesa_visitor *v)
{
/* The LHS must be a dereference. If the LHS is a variable indexed array
* swizzles in it and write swizzles using writemask, though.
*/
ir->accept(v);
- return ir_to_mesa_dst_reg_from_src(v->result);
+ return dst_reg(v->result);
}
/**
void
ir_to_mesa_visitor::visit(ir_assignment *ir)
{
- struct ir_to_mesa_dst_reg l;
- struct ir_to_mesa_src_reg r;
+ dst_reg l;
+ src_reg r;
int i;
ir->rhs->accept(this);
if (ir->condition) {
const bool switch_order = this->process_move_condition(ir->condition);
- ir_to_mesa_src_reg condition = this->result;
+ src_reg condition = this->result;
for (i = 0; i < type_size(ir->lhs->type); i++) {
if (switch_order) {
- ir_to_mesa_emit_op3(ir, OPCODE_CMP, l,
- condition, ir_to_mesa_src_reg_from_dst(l), r);
+ emit(ir, OPCODE_CMP, l, condition, src_reg(l), r);
} else {
- ir_to_mesa_emit_op3(ir, OPCODE_CMP, l,
- condition, r, ir_to_mesa_src_reg_from_dst(l));
+ emit(ir, OPCODE_CMP, l, condition, r, src_reg(l));
}
l.index++;
}
} else {
for (i = 0; i < type_size(ir->lhs->type); i++) {
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, l, r);
+ emit(ir, OPCODE_MOV, l, r);
l.index++;
r.index++;
}
void
ir_to_mesa_visitor::visit(ir_constant *ir)
{
- ir_to_mesa_src_reg src_reg;
+ src_reg src;
GLfloat stack_vals[4] = { 0 };
GLfloat *values = stack_vals;
unsigned int i;
*/
if (ir->type->base_type == GLSL_TYPE_STRUCT) {
- ir_to_mesa_src_reg temp_base = get_temp(ir->type);
- ir_to_mesa_dst_reg temp = ir_to_mesa_dst_reg_from_src(temp_base);
+ src_reg temp_base = get_temp(ir->type);
+ dst_reg temp = dst_reg(temp_base);
foreach_iter(exec_list_iterator, iter, ir->components) {
ir_constant *field_value = (ir_constant *)iter.get();
assert(size > 0);
field_value->accept(this);
- src_reg = this->result;
+ src = this->result;
for (i = 0; i < (unsigned int)size; i++) {
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, temp, src_reg);
+ emit(ir, OPCODE_MOV, temp, src);
- src_reg.index++;
+ src.index++;
temp.index++;
}
}
}
if (ir->type->is_array()) {
- ir_to_mesa_src_reg temp_base = get_temp(ir->type);
- ir_to_mesa_dst_reg temp = ir_to_mesa_dst_reg_from_src(temp_base);
+ src_reg temp_base = get_temp(ir->type);
+ dst_reg temp = dst_reg(temp_base);
int size = type_size(ir->type->fields.array);
assert(size > 0);
for (i = 0; i < ir->type->length; i++) {
ir->array_elements[i]->accept(this);
- src_reg = this->result;
+ src = this->result;
for (int j = 0; j < size; j++) {
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, temp, src_reg);
+ emit(ir, OPCODE_MOV, temp, src);
- src_reg.index++;
+ src.index++;
temp.index++;
}
}
}
if (ir->type->is_matrix()) {
- ir_to_mesa_src_reg mat = get_temp(ir->type);
- ir_to_mesa_dst_reg mat_column = ir_to_mesa_dst_reg_from_src(mat);
+ src_reg mat = get_temp(ir->type);
+ dst_reg mat_column = dst_reg(mat);
for (i = 0; i < ir->type->matrix_columns; i++) {
assert(ir->type->base_type == GLSL_TYPE_FLOAT);
values = &ir->value.f[i * ir->type->vector_elements];
- src_reg = ir_to_mesa_src_reg(PROGRAM_CONSTANT, -1, NULL);
- src_reg.index = _mesa_add_unnamed_constant(this->prog->Parameters,
- values,
+ src = src_reg(PROGRAM_CONSTANT, -1, NULL);
+ src.index = _mesa_add_unnamed_constant(this->prog->Parameters,
+ (gl_constant_value *) values,
ir->type->vector_elements,
- &src_reg.swizzle);
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, mat_column, src_reg);
+ &src.swizzle);
+ emit(ir, OPCODE_MOV, mat_column, src);
mat_column.index++;
}
return;
}
- src_reg.file = PROGRAM_CONSTANT;
+ src.file = PROGRAM_CONSTANT;
switch (ir->type->base_type) {
case GLSL_TYPE_FLOAT:
values = &ir->value.f[0];
assert(!"Non-float/uint/int/bool constant");
}
- this->result = ir_to_mesa_src_reg(PROGRAM_CONSTANT, -1, ir->type);
+ this->result = src_reg(PROGRAM_CONSTANT, -1, ir->type);
this->result.index = _mesa_add_unnamed_constant(this->prog->Parameters,
- values,
+ (gl_constant_value *) values,
ir->type->vector_elements,
&this->result.swizzle);
}
return entry;
}
- entry = talloc(mem_ctx, function_entry);
+ entry = ralloc(mem_ctx, function_entry);
entry->sig = sig;
entry->sig_id = this->next_signature_id++;
entry->bgn_inst = NULL;
if (!sig->return_type->is_void()) {
entry->return_reg = get_temp(sig->return_type);
} else {
- entry->return_reg = ir_to_mesa_undef;
+ entry->return_reg = undef_src;
}
this->function_signatures.push_tail(entry);
assert(storage);
param_rval->accept(this);
- ir_to_mesa_src_reg r = this->result;
+ src_reg r = this->result;
- ir_to_mesa_dst_reg l;
+ dst_reg l;
l.file = storage->file;
l.index = storage->index;
l.reladdr = NULL;
l.cond_mask = COND_TR;
for (i = 0; i < type_size(param->type); i++) {
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, l, r);
+ emit(ir, OPCODE_MOV, l, r);
l.index++;
r.index++;
}
assert(!sig_iter.has_next());
/* Emit call instruction */
- call_inst = ir_to_mesa_emit_op1(ir, OPCODE_CAL,
- ir_to_mesa_undef_dst, ir_to_mesa_undef);
+ call_inst = emit(ir, OPCODE_CAL);
call_inst->function = entry;
/* Process out parameters. */
variable_storage *storage = find_variable_storage(param);
assert(storage);
- ir_to_mesa_src_reg r;
+ src_reg r;
r.file = storage->file;
r.index = storage->index;
r.reladdr = NULL;
r.negate = 0;
param_rval->accept(this);
- ir_to_mesa_dst_reg l = ir_to_mesa_dst_reg_from_src(this->result);
+ dst_reg l = dst_reg(this->result);
for (i = 0; i < type_size(param->type); i++) {
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, l, r);
+ emit(ir, OPCODE_MOV, l, r);
l.index++;
r.index++;
}
void
ir_to_mesa_visitor::visit(ir_texture *ir)
{
- ir_to_mesa_src_reg result_src, coord, lod_info, projector;
- ir_to_mesa_dst_reg result_dst, coord_dst;
+ src_reg result_src, coord, lod_info, projector, dx, dy;
+ dst_reg result_dst, coord_dst;
ir_to_mesa_instruction *inst = NULL;
prog_opcode opcode = OPCODE_NOP;
* handle cleaning up our mess in that case.
*/
coord = get_temp(glsl_type::vec4_type);
- coord_dst = ir_to_mesa_dst_reg_from_src(coord);
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, coord_dst,
- this->result);
+ coord_dst = dst_reg(coord);
+ emit(ir, OPCODE_MOV, coord_dst, this->result);
if (ir->projector) {
ir->projector->accept(this);
* the actual storage for the result here, instead.
*/
result_src = get_temp(glsl_type::vec4_type);
- result_dst = ir_to_mesa_dst_reg_from_src(result_src);
+ result_dst = dst_reg(result_src);
switch (ir->op) {
case ir_tex:
lod_info = this->result;
break;
case ir_txd:
+ opcode = OPCODE_TXD;
+ ir->lod_info.grad.dPdx->accept(this);
+ dx = this->result;
+ ir->lod_info.grad.dPdy->accept(this);
+ dy = this->result;
+ break;
case ir_txf:
assert(!"GLSL 1.30 features unsupported");
break;
if (opcode == OPCODE_TEX) {
/* Slot the projector in as the last component of the coord. */
coord_dst.writemask = WRITEMASK_W;
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, coord_dst, projector);
+ emit(ir, OPCODE_MOV, coord_dst, projector);
coord_dst.writemask = WRITEMASK_XYZW;
opcode = OPCODE_TXP;
} else {
- ir_to_mesa_src_reg coord_w = coord;
+ src_reg coord_w = coord;
coord_w.swizzle = SWIZZLE_WWWW;
/* For the other TEX opcodes there's no projective version
* projective divide now.
*/
coord_dst.writemask = WRITEMASK_W;
- ir_to_mesa_emit_op1(ir, OPCODE_RCP, coord_dst, projector);
+ 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.
+ */
+ src_reg tmp_src = coord;
+ if (ir->shadow_comparitor) {
+ /* Slot the shadow value in as the second to last component of the
+ * coord.
+ */
+ ir->shadow_comparitor->accept(this);
+
+ tmp_src = get_temp(glsl_type::vec4_type);
+ dst_reg tmp_dst = dst_reg(tmp_src);
+
+ tmp_dst.writemask = WRITEMASK_Z;
+ emit(ir, OPCODE_MOV, tmp_dst, this->result);
+
+ tmp_dst.writemask = WRITEMASK_XY;
+ emit(ir, OPCODE_MOV, tmp_dst, coord);
+ }
coord_dst.writemask = WRITEMASK_XYZ;
- ir_to_mesa_emit_op2(ir, OPCODE_MUL, coord_dst, coord, coord_w);
+ emit(ir, OPCODE_MUL, coord_dst, tmp_src, coord_w);
coord_dst.writemask = WRITEMASK_XYZW;
coord.swizzle = SWIZZLE_XYZW;
}
}
- if (ir->shadow_comparitor) {
+ /* 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,
+ * above, that handles by-hand projection.
+ */
+ if (ir->shadow_comparitor && (!ir->projector || opcode == OPCODE_TXP)) {
/* Slot the shadow value in as the second to last component of the
* coord.
*/
ir->shadow_comparitor->accept(this);
coord_dst.writemask = WRITEMASK_Z;
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, coord_dst, this->result);
+ emit(ir, OPCODE_MOV, coord_dst, this->result);
coord_dst.writemask = WRITEMASK_XYZW;
}
if (opcode == OPCODE_TXL || opcode == OPCODE_TXB) {
/* Mesa IR stores lod or lod bias in the last channel of the coords. */
coord_dst.writemask = WRITEMASK_W;
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, coord_dst, lod_info);
+ emit(ir, OPCODE_MOV, coord_dst, lod_info);
coord_dst.writemask = WRITEMASK_XYZW;
}
- inst = ir_to_mesa_emit_op1(ir, opcode, result_dst, coord);
+ if (opcode == OPCODE_TXD)
+ inst = emit(ir, opcode, result_dst, coord, dx, dy);
+ else
+ inst = emit(ir, opcode, result_dst, coord);
if (ir->shadow_comparitor)
inst->tex_shadow = GL_TRUE;
ir_to_mesa_visitor::visit(ir_return *ir)
{
if (ir->get_value()) {
- ir_to_mesa_dst_reg l;
+ dst_reg l;
int i;
assert(current_function);
ir->get_value()->accept(this);
- ir_to_mesa_src_reg r = this->result;
+ src_reg r = this->result;
- l = ir_to_mesa_dst_reg_from_src(current_function->return_reg);
+ l = dst_reg(current_function->return_reg);
for (i = 0; i < type_size(current_function->sig->return_type); i++) {
- ir_to_mesa_emit_op1(ir, OPCODE_MOV, l, r);
+ emit(ir, OPCODE_MOV, l, r);
l.index++;
r.index++;
}
}
- ir_to_mesa_emit_op0(ir, OPCODE_RET);
+ emit(ir, OPCODE_RET);
}
void
if (ir->condition) {
ir->condition->accept(this);
this->result.negate = ~this->result.negate;
- ir_to_mesa_emit_op1(ir, OPCODE_KIL, ir_to_mesa_undef_dst, this->result);
+ emit(ir, OPCODE_KIL, undef_dst, this->result);
} else {
- ir_to_mesa_emit_op0(ir, OPCODE_KIL_NV);
+ emit(ir, OPCODE_KIL_NV);
}
fp->UsesKill = GL_TRUE;
void
ir_to_mesa_visitor::visit(ir_if *ir)
{
- ir_to_mesa_instruction *cond_inst, *if_inst, *else_inst = NULL;
+ ir_to_mesa_instruction *cond_inst, *if_inst;
ir_to_mesa_instruction *prev_inst;
prev_inst = (ir_to_mesa_instruction *)this->instructions.get_tail();
* have something to set cond_update on.
*/
if (cond_inst == prev_inst) {
- ir_to_mesa_src_reg temp = get_temp(glsl_type::bool_type);
- cond_inst = ir_to_mesa_emit_op1(ir->condition, OPCODE_MOV,
- ir_to_mesa_dst_reg_from_src(temp),
- result);
+ 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 = ir_to_mesa_emit_op0(ir->condition, OPCODE_IF);
- if_inst->dst_reg.cond_mask = COND_NE;
+ if_inst = emit(ir->condition, OPCODE_IF);
+ if_inst->dst.cond_mask = COND_NE;
} else {
- if_inst = ir_to_mesa_emit_op1(ir->condition,
- OPCODE_IF, ir_to_mesa_undef_dst,
- this->result);
+ if_inst = emit(ir->condition, OPCODE_IF, undef_dst, this->result);
}
this->instructions.push_tail(if_inst);
visit_exec_list(&ir->then_instructions, this);
if (!ir->else_instructions.is_empty()) {
- else_inst = ir_to_mesa_emit_op0(ir->condition, OPCODE_ELSE);
+ emit(ir->condition, OPCODE_ELSE);
visit_exec_list(&ir->else_instructions, this);
}
- if_inst = ir_to_mesa_emit_op1(ir->condition, OPCODE_ENDIF,
- ir_to_mesa_undef_dst, ir_to_mesa_undef);
+ if_inst = emit(ir->condition, OPCODE_ENDIF);
}
ir_to_mesa_visitor::ir_to_mesa_visitor()
next_temp = 1;
next_signature_id = 1;
current_function = NULL;
- mem_ctx = talloc_new(NULL);
+ mem_ctx = ralloc_context(NULL);
}
ir_to_mesa_visitor::~ir_to_mesa_visitor()
{
- talloc_free(mem_ctx);
+ ralloc_free(mem_ctx);
}
static struct prog_src_register
-mesa_src_reg_from_ir_src_reg(ir_to_mesa_src_reg reg)
+mesa_src_reg_from_ir_src_reg(src_reg reg)
{
struct prog_src_register mesa_reg;
}
}
- if_stack = talloc_zero_array(v->mem_ctx, int, if_count);
- loop_stack = talloc_zero_array(v->mem_ctx, int, loop_count);
+ if_stack = rzalloc_array(v->mem_ctx, int, if_count);
+ loop_stack = rzalloc_array(v->mem_ctx, int, loop_count);
for (i = 0; i < num_instructions; i++) {
switch (mesa_instructions[i].Opcode) {
}
}
+
+/**
+ * Count resources used by the given gpu program (number of texture
+ * samplers, etc).
+ */
static void
count_resources(struct gl_program *prog)
{
_mesa_update_shader_textures_used(prog);
}
+
+/**
+ * Check if the given vertex/fragment/shader program is within the
+ * resource limits of the context (number of texture units, etc).
+ * If any of those checks fail, record a linker error.
+ *
+ * XXX more checks are needed...
+ */
+static void
+check_resources(const struct gl_context *ctx,
+ struct gl_shader_program *shader_program,
+ struct gl_program *prog)
+{
+ switch (prog->Target) {
+ case GL_VERTEX_PROGRAM_ARB:
+ if (_mesa_bitcount(prog->SamplersUsed) >
+ ctx->Const.MaxVertexTextureImageUnits) {
+ linker_error(shader_program,
+ "Too many vertex shader texture samplers");
+ }
+ if (prog->Parameters->NumParameters > MAX_UNIFORMS) {
+ linker_error(shader_program, "Too many vertex shader constants");
+ }
+ break;
+ case MESA_GEOMETRY_PROGRAM:
+ if (_mesa_bitcount(prog->SamplersUsed) >
+ ctx->Const.MaxGeometryTextureImageUnits) {
+ linker_error(shader_program,
+ "Too many geometry shader texture samplers");
+ }
+ if (prog->Parameters->NumParameters >
+ MAX_GEOMETRY_UNIFORM_COMPONENTS / 4) {
+ linker_error(shader_program, "Too many geometry shader constants");
+ }
+ break;
+ case GL_FRAGMENT_PROGRAM_ARB:
+ if (_mesa_bitcount(prog->SamplersUsed) >
+ ctx->Const.MaxTextureImageUnits) {
+ linker_error(shader_program,
+ "Too many fragment shader texture samplers");
+ }
+ if (prog->Parameters->NumParameters > MAX_UNIFORMS) {
+ linker_error(shader_program, "Too many fragment shader constants");
+ }
+ break;
+ default:
+ _mesa_problem(ctx, "unexpected program type in check_resources()");
+ }
+}
+
+
+
struct uniform_sort {
struct gl_uniform *u;
int pos;
unsigned int next_sampler = 0, num_uniforms = 0;
struct uniform_sort *sorted_uniforms;
- sorted_uniforms = talloc_array(NULL, struct uniform_sort,
+ sorted_uniforms = ralloc_array(NULL, struct uniform_sort,
shader_program->Uniforms->NumUniforms);
for (i = 0; i < shader_program->Uniforms->NumUniforms; i++) {
*/
if (file == PROGRAM_SAMPLER) {
for (unsigned int j = 0; j < size / 4; j++)
- prog->Parameters->ParameterValues[index + j][0] = next_sampler++;
+ prog->Parameters->ParameterValues[index + j][0].f = next_sampler++;
}
/* The location chosen in the Parameters list here (returned
* from _mesa_add_uniform) has to match what the linker chose.
*/
if (index != parameter_index) {
- fail_link(shader_program, "Allocation of uniform `%s' to target "
- "failed (%d vs %d)\n",
- uniform->Name, index, parameter_index);
+ linker_error(shader_program,
+ "Allocation of uniform `%s' to target failed "
+ "(%d vs %d)\n",
+ uniform->Name, index, parameter_index);
}
}
}
- talloc_free(sorted_uniforms);
+ ralloc_free(sorted_uniforms);
}
static void
for (unsigned int i = 0; i < type->length; i++) {
const glsl_type *field_type = type->fields.structure[i].type;
- const char *field_name = talloc_asprintf(mem_ctx, "%s.%s", name,
+ const char *field_name = ralloc_asprintf(mem_ctx, "%s.%s", name,
type->fields.structure[i].name);
set_uniform_initializer(ctx, mem_ctx, shader_program, field_name,
field_type, field_constant);
int loc = _mesa_get_uniform_location(ctx, shader_program, name);
if (loc == -1) {
- fail_link(shader_program,
- "Couldn't find uniform for initializer %s\n", name);
+ linker_error(shader_program,
+ "Couldn't find uniform for initializer %s\n", name);
return;
}
void *values;
if (element_type->base_type == GLSL_TYPE_BOOL) {
- int *conv = talloc_array(mem_ctx, int, element_type->components());
+ int *conv = ralloc_array(mem_ctx, int, element_type->components());
for (unsigned int j = 0; j < element_type->components(); j++) {
conv[j] = element->value.b[j];
}
continue;
if (!mem_ctx)
- mem_ctx = talloc_new(NULL);
+ mem_ctx = ralloc_context(NULL);
set_uniform_initializer(ctx, mem_ctx, shader_program, var->name,
var->type, var->constant_value);
}
}
- talloc_free(mem_ctx);
+ ralloc_free(mem_ctx);
}
/*
void
ir_to_mesa_visitor::copy_propagate(void)
{
- ir_to_mesa_instruction **acp = new ir_to_mesa_instruction *[this->next_temp * 4];
-
- memset(acp, 0, sizeof(acp) * this->next_temp * 4);
+ ir_to_mesa_instruction **acp = rzalloc_array(mem_ctx,
+ ir_to_mesa_instruction *,
+ this->next_temp * 4);
+ int *acp_level = rzalloc_array(mem_ctx, int, this->next_temp * 4);
+ int level = 0;
foreach_iter(exec_list_iterator, iter, this->instructions) {
ir_to_mesa_instruction *inst = (ir_to_mesa_instruction *)iter.get();
+ assert(inst->dst.file != PROGRAM_TEMPORARY
+ || inst->dst.index < this->next_temp);
+
/* First, do any copy propagation possible into the src regs. */
for (int r = 0; r < 3; r++) {
ir_to_mesa_instruction *first = NULL;
bool good = true;
- int acp_base = inst->src_reg[r].index * 4;
+ int acp_base = inst->src[r].index * 4;
- if (inst->src_reg[r].file != PROGRAM_TEMPORARY ||
- inst->src_reg[r].reladdr)
+ if (inst->src[r].file != PROGRAM_TEMPORARY ||
+ inst->src[r].reladdr)
continue;
/* See if we can find entries in the ACP consisting of MOVs
* of this src register reference.
*/
for (int i = 0; i < 4; i++) {
- int src_chan = GET_SWZ(inst->src_reg[r].swizzle, i);
+ int src_chan = GET_SWZ(inst->src[r].swizzle, i);
ir_to_mesa_instruction *copy_chan = acp[acp_base + src_chan];
if (!copy_chan) {
break;
}
+ assert(acp_level[acp_base + src_chan] <= level);
+
if (!first) {
first = copy_chan;
} else {
- if (first->src_reg[0].file != copy_chan->src_reg[0].file ||
- first->src_reg[0].index != copy_chan->src_reg[0].index) {
+ if (first->src[0].file != copy_chan->src[0].file ||
+ first->src[0].index != copy_chan->src[0].index) {
good = false;
break;
}
/* We've now validated that we can copy-propagate to
* replace this src register reference. Do it.
*/
- inst->src_reg[r].file = first->src_reg[0].file;
- inst->src_reg[r].index = first->src_reg[0].index;
+ inst->src[r].file = first->src[0].file;
+ inst->src[r].index = first->src[0].index;
int swizzle = 0;
for (int i = 0; i < 4; i++) {
- int src_chan = GET_SWZ(inst->src_reg[r].swizzle, i);
+ int src_chan = GET_SWZ(inst->src[r].swizzle, i);
ir_to_mesa_instruction *copy_inst = acp[acp_base + src_chan];
- swizzle |= (GET_SWZ(copy_inst->src_reg[0].swizzle, src_chan) <<
+ swizzle |= (GET_SWZ(copy_inst->src[0].swizzle, src_chan) <<
(3 * i));
}
- inst->src_reg[r].swizzle = swizzle;
+ inst->src[r].swizzle = swizzle;
}
}
switch (inst->op) {
case OPCODE_BGNLOOP:
case OPCODE_ENDLOOP:
- case OPCODE_ELSE:
- case OPCODE_ENDIF:
/* End of a basic block, clear the ACP entirely. */
- memset(&acp, 0, sizeof(acp));
+ memset(acp, 0, sizeof(*acp) * this->next_temp * 4);
+ break;
+
+ case OPCODE_IF:
+ ++level;
+ break;
+
+ case OPCODE_ENDIF:
+ case OPCODE_ELSE:
+ /* Clear all channels written inside the block from the ACP, but
+ * leaving those that were not touched.
+ */
+ for (int r = 0; r < this->next_temp; r++) {
+ for (int c = 0; c < 4; c++) {
+ if (!acp[4 * r + c])
+ continue;
+
+ if (acp_level[4 * r + c] >= level)
+ acp[4 * r + c] = NULL;
+ }
+ }
+ if (inst->op == OPCODE_ENDIF)
+ --level;
break;
default:
/* Continuing the block, clear any written channels from
* the ACP.
*/
- if (inst->dst_reg.file == PROGRAM_TEMPORARY) {
- if (inst->dst_reg.reladdr) {
- memset(&acp, 0, sizeof(acp));
- } else {
- for (int i = 0; i < 4; i++) {
- if (inst->dst_reg.writemask & (1 << i)) {
- acp[4 * inst->dst_reg.index + i] = NULL;
+ if (inst->dst.file == PROGRAM_TEMPORARY && inst->dst.reladdr) {
+ /* Any temporary might be written, so no copy propagation
+ * across this instruction.
+ */
+ memset(acp, 0, sizeof(*acp) * this->next_temp * 4);
+ } else if (inst->dst.file == PROGRAM_OUTPUT &&
+ inst->dst.reladdr) {
+ /* Any output might be written, so no copy propagation
+ * from outputs across this instruction.
+ */
+ for (int r = 0; r < this->next_temp; r++) {
+ for (int c = 0; c < 4; c++) {
+ if (!acp[4 * r + c])
+ continue;
+
+ if (acp[4 * r + c]->src[0].file == PROGRAM_OUTPUT)
+ acp[4 * r + c] = NULL;
+ }
+ }
+ } else if (inst->dst.file == PROGRAM_TEMPORARY ||
+ inst->dst.file == PROGRAM_OUTPUT) {
+ /* Clear where it's used as dst. */
+ if (inst->dst.file == PROGRAM_TEMPORARY) {
+ for (int c = 0; c < 4; c++) {
+ if (inst->dst.writemask & (1 << c)) {
+ acp[4 * inst->dst.index + c] = NULL;
+ }
+ }
+ }
+
+ /* Clear where it's used as src. */
+ for (int r = 0; r < this->next_temp; r++) {
+ for (int c = 0; c < 4; c++) {
+ if (!acp[4 * r + c])
+ continue;
+
+ int src_chan = GET_SWZ(acp[4 * r + c]->src[0].swizzle, c);
+
+ if (acp[4 * r + c]->src[0].file == inst->dst.file &&
+ acp[4 * r + c]->src[0].index == inst->dst.index &&
+ inst->dst.writemask & (1 << src_chan))
+ {
+ acp[4 * r + c] = NULL;
}
}
}
/* If this is a copy, add it to the ACP. */
if (inst->op == OPCODE_MOV &&
- inst->dst_reg.file == PROGRAM_TEMPORARY &&
- !inst->dst_reg.reladdr &&
+ inst->dst.file == PROGRAM_TEMPORARY &&
+ !inst->dst.reladdr &&
!inst->saturate &&
- !inst->src_reg[0].reladdr &&
- !inst->src_reg[0].negate) {
+ !inst->src[0].reladdr &&
+ !inst->src[0].negate) {
for (int i = 0; i < 4; i++) {
- if (inst->dst_reg.writemask & (1 << i)) {
- acp[4 * inst->dst_reg.index + i] = inst;
+ if (inst->dst.writemask & (1 << i)) {
+ acp[4 * inst->dst.index + i] = inst;
+ acp_level[4 * inst->dst.index + i] = level;
}
}
}
}
- delete [] acp;
+ ralloc_free(acp_level);
+ ralloc_free(acp);
}
/* Emit Mesa IR for main(). */
visit_exec_list(shader->ir, &v);
- v.ir_to_mesa_emit_op0(NULL, OPCODE_END);
+ v.emit(NULL, OPCODE_END);
/* Now emit bodies for any functions that were used. */
do {
if (!entry->bgn_inst) {
v.current_function = entry;
- entry->bgn_inst = v.ir_to_mesa_emit_op0(NULL, OPCODE_BGNSUB);
+ entry->bgn_inst = v.emit(NULL, OPCODE_BGNSUB);
entry->bgn_inst->function = entry;
visit_exec_list(&entry->sig->body, &v);
ir_to_mesa_instruction *last;
last = (ir_to_mesa_instruction *)v.instructions.get_tail();
if (last->op != OPCODE_RET)
- v.ir_to_mesa_emit_op0(NULL, OPCODE_RET);
+ v.emit(NULL, OPCODE_RET);
ir_to_mesa_instruction *end;
- end = v.ir_to_mesa_emit_op0(NULL, OPCODE_ENDSUB);
+ end = v.emit(NULL, OPCODE_ENDSUB);
end->function = entry;
progress = GL_TRUE;
mesa_instructions =
(struct prog_instruction *)calloc(num_instructions,
sizeof(*mesa_instructions));
- mesa_instruction_annotation = talloc_array(v.mem_ctx, ir_instruction *,
+ mesa_instruction_annotation = ralloc_array(v.mem_ctx, ir_instruction *,
num_instructions);
v.copy_propagate();
mesa_inst->CondUpdate = inst->cond_update;
if (inst->saturate)
mesa_inst->SaturateMode = SATURATE_ZERO_ONE;
- mesa_inst->DstReg.File = inst->dst_reg.file;
- mesa_inst->DstReg.Index = inst->dst_reg.index;
- mesa_inst->DstReg.CondMask = inst->dst_reg.cond_mask;
- mesa_inst->DstReg.WriteMask = inst->dst_reg.writemask;
- mesa_inst->DstReg.RelAddr = inst->dst_reg.reladdr != NULL;
- mesa_inst->SrcReg[0] = mesa_src_reg_from_ir_src_reg(inst->src_reg[0]);
- mesa_inst->SrcReg[1] = mesa_src_reg_from_ir_src_reg(inst->src_reg[1]);
- mesa_inst->SrcReg[2] = mesa_src_reg_from_ir_src_reg(inst->src_reg[2]);
+ 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]);
+ mesa_inst->SrcReg[1] = mesa_src_reg_from_ir_src_reg(inst->src[1]);
+ mesa_inst->SrcReg[2] = mesa_src_reg_from_ir_src_reg(inst->src[2]);
mesa_inst->TexSrcUnit = inst->sampler;
mesa_inst->TexSrcTarget = inst->tex_target;
mesa_inst->TexShadow = inst->tex_shadow;
if (mesa_inst->SrcReg[src].RelAddr)
prog->IndirectRegisterFiles |= 1 << mesa_inst->SrcReg[src].File;
- if (options->EmitNoIfs && mesa_inst->Opcode == OPCODE_IF) {
- fail_link(shader_program, "Couldn't flatten if statement\n");
- }
-
switch (mesa_inst->Opcode) {
+ case OPCODE_IF:
+ if (options->EmitNoIfs) {
+ linker_warning(shader_program,
+ "Couldn't flatten if-statement. "
+ "This will likely result in software "
+ "rasterization.\n");
+ }
+ break;
+ case OPCODE_BGNLOOP:
+ if (options->EmitNoLoops) {
+ linker_warning(shader_program,
+ "Couldn't unroll loop. "
+ "This will likely result in software "
+ "rasterization.\n");
+ }
+ break;
+ case OPCODE_CONT:
+ if (options->EmitNoCont) {
+ linker_warning(shader_program,
+ "Couldn't lower continue-statement. "
+ "This will likely result in software "
+ "rasterization.\n");
+ }
+ break;
case OPCODE_BGNSUB:
inst->function->inst = i;
mesa_inst->Comment = strdup(inst->function->sig->function_name());
do_set_program_inouts(shader->ir, prog);
count_resources(prog);
+ check_resources(ctx, shader_program, prog);
+
_mesa_reference_program(ctx, &shader->Program, prog);
if ((ctx->Shader.Flags & GLSL_NO_OPT) == 0) {
extern "C" {
-/**
- * Called via ctx->Driver.CompilerShader().
- * This is a no-op.
- * XXX can we remove the ctx->Driver.CompileShader() hook?
- */
-GLboolean
-_mesa_ir_compile_shader(struct gl_context *ctx, struct gl_shader *shader)
-{
- assert(shader->CompileStatus);
- (void) ctx;
-
- return GL_TRUE;
-}
-
-
/**
* Link a shader.
* Called via ctx->Driver.LinkShader()
&ctx->Extensions, ctx->API);
if (ctx->Shader.Flags & GLSL_DUMP) {
- printf("GLSL source for shader %d:\n", shader->Name);
+ printf("GLSL source for %s shader %d:\n",
+ _mesa_glsl_shader_target_name(state->target), shader->Name);
printf("%s\n", shader->Source);
}
_mesa_glsl_lexer_dtor(state);
}
- talloc_free(shader->ir);
+ ralloc_free(shader->ir);
shader->ir = new(shader) exec_list;
if (!state->error && !state->translation_unit.is_empty())
_mesa_ast_to_hir(shader->ir, state);
/* Retain any live IR, but trash the rest. */
reparent_ir(shader->ir, shader->ir);
- talloc_free(state);
-
- if (shader->CompileStatus) {
- if (!ctx->Driver.CompileShader(ctx, shader))
- shader->CompileStatus = GL_FALSE;
- }
+ ralloc_free(state);
}
for (i = 0; i < prog->NumShaders; i++) {
if (!prog->Shaders[i]->CompileStatus) {
- fail_link(prog, "linking with uncompiled shader");
+ linker_error(prog, "linking with uncompiled shader");
prog->LinkStatus = GL_FALSE;
}
}
projects / mesa.git / blobdiff