#include "st_glsl_types.h"
#include "st_nir.h"
+#include <algorithm>
#define PROGRAM_ANY_CONST ((1 << PROGRAM_STATE_VAR) | \
(1 << PROGRAM_CONSTANT) | \
this->index = index;
this->swizzle = swizzle_for_type(type, component);
this->negate = 0;
+ this->abs = 0;
this->index2D = 0;
this->type = type ? type->base_type : GLSL_TYPE_ERROR;
this->reladdr = NULL;
this->index2D = 0;
this->swizzle = SWIZZLE_XYZW;
this->negate = 0;
+ this->abs = 0;
this->reladdr = NULL;
this->reladdr2 = NULL;
this->has_index2 = false;
this->index2D = index2D;
this->swizzle = SWIZZLE_XYZW;
this->negate = 0;
+ this->abs = 0;
this->reladdr = NULL;
this->reladdr2 = NULL;
this->has_index2 = false;
this->index2D = 0;
this->swizzle = 0;
this->negate = 0;
+ this->abs = 0;
this->reladdr = NULL;
this->reladdr2 = NULL;
this->has_index2 = false;
int16_t index2D;
uint16_t swizzle; /**< SWIZZLE_XYZWONEZERO swizzles from Mesa. */
int negate:4; /**< NEGATE_XYZW mask from mesa */
+ unsigned abs:1;
enum glsl_base_type type:4; /** GLSL_TYPE_* from GLSL IR (enum glsl_base_type) */
unsigned has_index2:1;
gl_register_file file:5; /**< PROGRAM_* from Mesa */
/** Register index should be offset by the integer in this reg. */
st_src_reg *reladdr;
st_src_reg *reladdr2;
+
+ st_src_reg get_abs()
+ {
+ st_src_reg reg = *this;
+ reg.negate = 0;
+ reg.abs = 1;
+ return reg;
+ }
};
class st_dst_reg {
this->index = reg.index;
this->swizzle = SWIZZLE_XYZW;
this->negate = 0;
+ this->abs = 0;
this->reladdr = reg.reladdr;
this->index2D = reg.index2D;
this->reladdr2 = reg.reladdr2;
unsigned dead_mask:4; /**< Used in dead code elimination */
unsigned buffer_access:3; /**< buffer access type */
- class function_entry *function; /* Set on TGSI_OPCODE_CAL or TGSI_OPCODE_BGNSUB */
const struct tgsi_opcode_info *info;
};
int type; /**< GL_DOUBLE, GL_FLOAT, GL_INT, GL_BOOL, or GL_UNSIGNED_INT */
};
-class function_entry : public exec_node {
-public:
- ir_function_signature *sig;
-
- /**
- * identifier of this function signature used by the program.
- *
- * At the point that TGSI instructions for function calls are
- * generated, we don't know the address of the first instruction of
- * the function body. So we make the BranchTarget that is called a
- * small integer and rewrite them during set_branchtargets().
- */
- int sig_id;
-
- /**
- * Pointer to first instruction of the function body.
- *
- * Set during function body emits after main() is processed.
- */
- glsl_to_tgsi_instruction *bgn_inst;
-
- /**
- * Index of the first instruction of the function body in actual TGSI.
- *
- * Set after conversion from glsl_to_tgsi_instruction to TGSI.
- */
- int inst;
-
- /** Storage for the return value. */
- st_src_reg return_reg;
-};
-
static st_src_reg undef_src = st_src_reg(PROGRAM_UNDEFINED, 0, GLSL_TYPE_ERROR);
static st_dst_reg undef_dst = st_dst_reg(PROGRAM_UNDEFINED, SWIZZLE_NOOP, GLSL_TYPE_ERROR);
unsigned mesa_index;
unsigned array_id; /* TGSI ArrayID; 1-based: 0 means not an array */
unsigned size;
+ unsigned interp_loc;
+ unsigned gs_out_streams;
+ enum glsl_interp_mode interp;
enum glsl_base_type base_type;
ubyte usage_mask; /* GLSL-style usage-mask, i.e. single bit per double */
};
glsl_to_tgsi_visitor();
~glsl_to_tgsi_visitor();
- function_entry *current_function;
-
struct gl_context *ctx;
struct gl_program *prog;
struct gl_shader_program *shader_program;
int add_constant(gl_register_file file, gl_constant_value values[8],
int size, int datatype, uint16_t *swizzle_out);
- function_entry *get_function_signature(ir_function_signature *sig);
-
st_src_reg get_temp(const glsl_type *type);
void reladdr_to_temp(ir_instruction *ir, st_src_reg *reg, int *num_reladdr);
exec_list immediates;
unsigned num_immediates;
- /** List of function_entry */
- exec_list function_signatures;
- int next_signature_id;
-
/** List of glsl_to_tgsi_instruction */
exec_list instructions;
{
va_list args;
va_start(args, fmt);
- ralloc_vasprintf_append(&prog->InfoLog, fmt, args);
+ ralloc_vasprintf_append(&prog->data->InfoLog, fmt, args);
va_end(args);
- prog->LinkStatus = GL_FALSE;
+ prog->data->LinkStatus = GL_FALSE;
}
static int
inst->is_64bit_expanded = false;
inst->ir = ir;
inst->dead_mask = 0;
+ inst->tex_offsets = NULL;
+ inst->tex_offset_num_offset = 0;
+ inst->saturate = 0;
+ inst->tex_shadow = 0;
/* default to float, for paths where this is not initialized
* (since 0==UINT which is likely wrong):
*/
inst->tex_type = GLSL_TYPE_FLOAT;
- inst->function = NULL;
-
/* Update indirect addressing status used by TGSI */
if (dst.reladdr || dst.reladdr2) {
switch(dst.file) {
int i = u_bit_scan(&writemask);
- /* before emitting the instruction, see if we have to adjust store
+ /* before emitting the instruction, see if we have to adjust load / store
* address */
- if (i > 1 && inst->op == TGSI_OPCODE_STORE &&
+ if (i > 1 && (inst->op == TGSI_OPCODE_LOAD || inst->op == TGSI_OPCODE_STORE) &&
addr.file == PROGRAM_UNDEFINED) {
/* We have to advance the buffer address by 16 */
addr = get_temp(glsl_type::uint_type);
inst->src[0], st_src_reg_for_int(16));
}
-
/* first time use previous instruction */
if (dinst == NULL) {
dinst = inst;
dinst->dst[j].writemask = (i & 1) ? WRITEMASK_ZW : WRITEMASK_XY;
dinst->dst[j].index = initial_dst_idx[j];
if (i > 1) {
- if (dinst->op == TGSI_OPCODE_STORE) {
+ if (dinst->op == TGSI_OPCODE_LOAD || dinst->op == TGSI_OPCODE_STORE)
dinst->src[0] = addr;
- } else {
+ if (dinst->op != TGSI_OPCODE_STORE)
dinst->dst[j].index++;
- }
}
} else {
/* if we aren't writing to a double, just get the bit of the initial writemask
case2iu(ISHR, USHR);
case3fid(SSG, ISSG, DSSG);
- case3fid(ABS, IABS, DABS);
case2iu(IBFE, UBFE);
case2iu(IMSB, UMSB);
src.type = native_integers ? type->base_type : GLSL_TYPE_FLOAT;
src.reladdr = NULL;
src.negate = 0;
+ src.abs = 0;
if (!options->EmitNoIndirectTemp && type_has_array_or_matrix(type)) {
if (next_array >= max_num_arrays) {
glsl_to_tgsi_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) {
emit_asm(ir, TGSI_OPCODE_MOV, result_dst, op[0]);
break;
case ir_unop_abs:
- emit_asm(ir, TGSI_OPCODE_ABS, result_dst, op[0]);
+ if (result_dst.type == GLSL_TYPE_FLOAT)
+ emit_asm(ir, TGSI_OPCODE_MOV, result_dst, op[0].get_abs());
+ else if (result_dst.type == GLSL_TYPE_DOUBLE)
+ emit_asm(ir, TGSI_OPCODE_DABS, result_dst, op[0]);
+ else
+ emit_asm(ir, TGSI_OPCODE_IABS, result_dst, op[0]);
break;
case ir_unop_sign:
emit_asm(ir, TGSI_OPCODE_SSG, result_dst, op[0]);
* we want, I choose to use ABS to match DX9 and pre-GLSL RSQ
* behavior.
*/
- emit_scalar(ir, TGSI_OPCODE_ABS, result_dst, op[0]);
- emit_scalar(ir, TGSI_OPCODE_RSQ, result_dst, result_src);
+ emit_scalar(ir, TGSI_OPCODE_RSQ, result_dst, op[0].get_abs());
emit_scalar(ir, TGSI_OPCODE_RCP, result_dst, result_src);
}
break;
case ir_unop_bitcast_f2i:
case ir_unop_bitcast_f2u:
/* Make sure we don't propagate the negate modifier to integer opcodes. */
- if (op[0].negate)
+ if (op[0].negate || op[0].abs)
emit_asm(ir, TGSI_OPCODE_MOV, result_dst, op[0]);
else
result_src = op[0];
cbuf.index = 0;
cbuf.reladdr = NULL;
cbuf.negate = 0;
+ cbuf.abs = 0;
assert(ir->type->is_vector() || ir->type->is_scalar());
return type->is_array() || type->is_matrix();
}
+static unsigned
+st_translate_interp_loc(ir_variable *var)
+{
+ if (var->data.centroid)
+ return TGSI_INTERPOLATE_LOC_CENTROID;
+ else if (var->data.sample)
+ return TGSI_INTERPOLATE_LOC_SAMPLE;
+ else
+ return TGSI_INTERPOLATE_LOC_CENTER;
+}
+
void
glsl_to_tgsi_visitor::visit(ir_dereference_variable *ir)
{
num_components = 4;
decl->mesa_index = var->data.location;
+ decl->interp = (glsl_interp_mode) var->data.interpolation;
+ decl->interp_loc = st_translate_interp_loc(var);
decl->base_type = type_without_array->base_type;
decl->usage_mask = u_bit_consecutive(component, num_components);
decl->mesa_index = var->data.location + FRAG_RESULT_MAX * var->data.index;
decl->base_type = type_without_array->base_type;
decl->usage_mask = u_bit_consecutive(component, num_components);
+ if (var->data.stream & (1u << 31)) {
+ decl->gs_out_streams = var->data.stream & ~(1u << 31);
+ } else {
+ assert(var->data.stream < 4);
+ decl->gs_out_streams = 0;
+ for (unsigned i = 0; i < num_components; ++i)
+ decl->gs_out_streams |= var->data.stream << (2 * (component + i));
+ }
if (is_inout_array(shader->Stage, var, &remove_array)) {
decl->array_id = num_output_arrays + 1;
assert(type->is_scalar() || type->is_vector());
+ l->type = type->base_type;
r->type = type->base_type;
if (cond) {
st_src_reg l_src = st_src_reg(*l);
} else if (ir->write_mask == 0) {
assert(!ir->lhs->type->is_scalar() && !ir->lhs->type->is_vector());
- if (ir->lhs->type->is_array() || ir->lhs->type->is_matrix()) {
- unsigned num_elements = ir->lhs->type->without_array()->vector_elements;
+ unsigned num_elements = ir->lhs->type->without_array()->vector_elements;
+
+ if (num_elements) {
l.writemask = u_bit_consecutive(0, num_elements);
} else {
+ /* The type is a struct or an array of (array of) structs. */
l.writemask = WRITEMASK_XYZW;
}
} else {
&this->result.swizzle);
}
-function_entry *
-glsl_to_tgsi_visitor::get_function_signature(ir_function_signature *sig)
-{
- foreach_in_list_use_after(function_entry, entry, &this->function_signatures) {
- if (entry->sig == sig)
- return entry;
- }
-
- entry = ralloc(mem_ctx, function_entry);
- entry->sig = sig;
- entry->sig_id = this->next_signature_id++;
- entry->bgn_inst = NULL;
-
- /* Allocate storage for all the parameters. */
- foreach_in_list(ir_variable, param, &sig->parameters) {
- variable_storage *storage;
-
- storage = find_variable_storage(param);
- assert(!storage);
-
- st_src_reg src = get_temp(param->type);
-
- storage = new(mem_ctx) variable_storage(param, src.file, src.index);
- this->variables.push_tail(storage);
- }
-
- if (!sig->return_type->is_void()) {
- entry->return_reg = get_temp(sig->return_type);
- } else {
- entry->return_reg = undef_src;
- }
-
- this->function_signatures.push_tail(entry);
- return entry;
-}
-
void
glsl_to_tgsi_visitor::visit_atomic_counter_intrinsic(ir_call *ir)
{
void
glsl_to_tgsi_visitor::visit(ir_call *ir)
{
- glsl_to_tgsi_instruction *call_inst;
ir_function_signature *sig = ir->callee;
- function_entry *entry;
- int i;
/* Filter out intrinsics */
switch (sig->intrinsic_id) {
- case ir_intrinsic_invalid:
- break;
-
case ir_intrinsic_atomic_counter_read:
case ir_intrinsic_atomic_counter_increment:
case ir_intrinsic_atomic_counter_predecrement:
visit_image_intrinsic(ir);
return;
+ case ir_intrinsic_invalid:
case ir_intrinsic_generic_load:
case ir_intrinsic_generic_store:
case ir_intrinsic_generic_atomic_add:
case ir_intrinsic_shader_clock:
unreachable("Invalid intrinsic");
}
-
- entry = get_function_signature(sig);
- /* Process in parameters. */
- foreach_two_lists(formal_node, &sig->parameters,
- actual_node, &ir->actual_parameters) {
- ir_rvalue *param_rval = (ir_rvalue *) actual_node;
- ir_variable *param = (ir_variable *) formal_node;
-
- if (param->data.mode == ir_var_function_in ||
- param->data.mode == ir_var_function_inout) {
- variable_storage *storage = find_variable_storage(param);
- assert(storage);
-
- param_rval->accept(this);
- st_src_reg r = this->result;
-
- st_dst_reg l;
- l.file = storage->file;
- l.index = storage->index;
- l.reladdr = NULL;
- l.writemask = WRITEMASK_XYZW;
-
- for (i = 0; i < type_size(param->type); i++) {
- emit_asm(ir, TGSI_OPCODE_MOV, l, r);
- l.index++;
- r.index++;
- }
- }
- }
-
- /* Emit call instruction */
- call_inst = emit_asm(ir, TGSI_OPCODE_CAL);
- call_inst->function = entry;
-
- /* Process out parameters. */
- foreach_two_lists(formal_node, &sig->parameters,
- actual_node, &ir->actual_parameters) {
- ir_rvalue *param_rval = (ir_rvalue *) actual_node;
- ir_variable *param = (ir_variable *) formal_node;
-
- if (param->data.mode == ir_var_function_out ||
- param->data.mode == ir_var_function_inout) {
- variable_storage *storage = find_variable_storage(param);
- assert(storage);
-
- st_src_reg r;
- r.file = storage->file;
- r.index = storage->index;
- r.reladdr = NULL;
- r.swizzle = SWIZZLE_NOOP;
- r.negate = 0;
-
- param_rval->accept(this);
- st_dst_reg l = st_dst_reg(this->result);
-
- for (i = 0; i < type_size(param->type); i++) {
- emit_asm(ir, TGSI_OPCODE_MOV, l, r);
- l.index++;
- r.index++;
- }
- }
- }
-
- /* Process return value. */
- this->result = entry->return_reg;
}
void
if (opaque) {
assert(location != 0xffffffff);
- *base += this->shader_program->UniformStorage[location].opaque[shader].index;
- *index += this->shader_program->UniformStorage[location].opaque[shader].index;
+ *base += this->shader_program->data->UniformStorage[location].opaque[shader].index;
+ *index += this->shader_program->data->UniformStorage[location].opaque[shader].index;
}
}
switch (ir->op) {
case ir_tex:
- opcode = (is_cube_array && ir->shadow_comparitor) ? TGSI_OPCODE_TEX2 : TGSI_OPCODE_TEX;
+ opcode = (is_cube_array && ir->shadow_comparator) ? TGSI_OPCODE_TEX2 : TGSI_OPCODE_TEX;
if (ir->offset) {
ir->offset->accept(this);
offset[0] = this->result;
* the shadow comparator value must also be projected.
*/
st_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);
st_dst_reg tmp_dst = st_dst_reg(tmp_src);
* 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 == TGSI_OPCODE_TXP)) {
+ if (ir->shadow_comparator && (!ir->projector || opcode == TGSI_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);
if (is_cube_array) {
cube_sc = get_temp(glsl_type::float_type);
} else if (opcode == TGSI_OPCODE_TEX2) {
inst = emit_asm(ir, opcode, result_dst, coord, cube_sc);
} else if (opcode == TGSI_OPCODE_TG4) {
- if (is_cube_array && ir->shadow_comparitor) {
+ if (is_cube_array && ir->shadow_comparator) {
inst = emit_asm(ir, opcode, result_dst, coord, cube_sc);
} else {
inst = emit_asm(ir, opcode, result_dst, coord, component);
} else
inst = emit_asm(ir, opcode, result_dst, coord);
- if (ir->shadow_comparitor)
+ if (ir->shadow_comparator)
inst->tex_shadow = GL_TRUE;
inst->resource.index = sampler_index;
void
glsl_to_tgsi_visitor::visit(ir_return *ir)
{
- if (ir->get_value()) {
- st_dst_reg l;
- int i;
-
- assert(current_function);
-
- ir->get_value()->accept(this);
- st_src_reg r = this->result;
-
- l = st_dst_reg(current_function->return_reg);
-
- for (i = 0; i < type_size(current_function->sig->return_type); i++) {
- emit_asm(ir, TGSI_OPCODE_MOV, l, r);
- l.index++;
- r.index++;
- }
- }
+ assert(!ir->get_value());
emit_asm(ir, TGSI_OPCODE_RET);
}
num_outputs = 0;
num_input_arrays = 0;
num_output_arrays = 0;
- next_signature_id = 1;
num_immediates = 0;
- current_function = NULL;
num_address_regs = 0;
samplers_used = 0;
buffers_used = 0;
if (inst->dst[0].reladdr || inst->dst[0].reladdr2 ||
inst->dst[1].reladdr || inst->dst[1].reladdr2 ||
tgsi_get_opcode_info(inst->op)->is_branch ||
- inst->op == TGSI_OPCODE_BGNSUB ||
inst->op == TGSI_OPCODE_CONT ||
inst->op == TGSI_OPCODE_END ||
- inst->op == TGSI_OPCODE_ENDSUB ||
inst->op == TGSI_OPCODE_RET) {
break;
}
inst->src[0].file != PROGRAM_ARRAY &&
!inst->src[0].reladdr &&
!inst->src[0].reladdr2 &&
- !inst->src[0].negate) {
+ !inst->src[0].negate &&
+ !inst->src[0].abs) {
for (int i = 0; i < 4; i++) {
if (inst->dst[0].writemask & (1 << i)) {
acp[4 * inst->dst[0].index + i] = inst;
}
/* ------------------------- TGSI conversion stuff -------------------------- */
-struct label {
- unsigned branch_target;
- unsigned token;
-};
/**
* Intermediate state used during shader translation.
const GLuint *inputMapping;
const GLuint *outputMapping;
- /* For every instruction that contains a label (eg CALL), keep
- * details so that we can go back afterwards and emit the correct
- * tgsi instruction number for each label.
- */
- struct label *labels;
- unsigned labels_size;
- unsigned labels_count;
-
- /* Keep a record of the tgsi instruction number that each mesa
- * instruction starts at, will be used to fix up labels after
- * translation.
- */
- unsigned *insn;
- unsigned insn_size;
- unsigned insn_count;
-
unsigned procType; /**< PIPE_SHADER_VERTEX/FRAGMENT */
-
- boolean error;
};
/** Map Mesa's SYSTEM_VALUE_x to TGSI_SEMANTIC_x */
}
}
-
-/**
- * Make note of a branch to a label in the TGSI code.
- * After we've emitted all instructions, we'll go over the list
- * of labels built here and patch the TGSI code with the actual
- * location of each label.
- */
-static unsigned *get_label(struct st_translate *t, unsigned branch_target)
-{
- unsigned i;
-
- if (t->labels_count + 1 >= t->labels_size) {
- t->labels_size = 1 << (util_logbase2(t->labels_size) + 1);
- t->labels = (struct label *)realloc(t->labels,
- t->labels_size * sizeof(struct label));
- if (t->labels == NULL) {
- static unsigned dummy;
- t->error = TRUE;
- return &dummy;
- }
- }
-
- i = t->labels_count++;
- t->labels[i].branch_target = branch_target;
- return &t->labels[i].token;
-}
-
-/**
- * Called prior to emitting the TGSI code for each instruction.
- * Allocate additional space for instructions if needed.
- * Update the insn[] array so the next glsl_to_tgsi_instruction points to
- * the next TGSI instruction.
- */
-static void set_insn_start(struct st_translate *t, unsigned start)
-{
- if (t->insn_count + 1 >= t->insn_size) {
- t->insn_size = 1 << (util_logbase2(t->insn_size) + 1);
- t->insn = (unsigned *)realloc(t->insn, t->insn_size * sizeof(t->insn[0]));
- if (t->insn == NULL) {
- t->error = TRUE;
- return;
- }
- }
-
- t->insn[t->insn_count++] = start;
-}
-
/**
* Map a glsl_to_tgsi constant/immediate to a TGSI immediate.
*/
case PROGRAM_TEMPORARY:
case PROGRAM_ARRAY:
- case PROGRAM_OUTPUT:
return ureg_src(dst_register(t, reg->file, reg->index, reg->array_id));
+ case PROGRAM_OUTPUT: {
+ struct ureg_dst dst = dst_register(t, reg->file, reg->index, reg->array_id);
+ assert(dst.WriteMask != 0);
+ unsigned shift = ffs(dst.WriteMask) - 1;
+ return ureg_swizzle(ureg_src(dst),
+ shift,
+ MIN2(shift + 1, 3),
+ MIN2(shift + 2, 3),
+ MIN2(shift + 3, 3));
+ }
+
case PROGRAM_UNIFORM:
assert(reg->index >= 0);
return reg->index < t->num_constants ?
GET_SWZ(src_reg->swizzle, 2) & 0x3,
GET_SWZ(src_reg->swizzle, 3) & 0x3);
+ if (src_reg->abs)
+ src = ureg_abs(src);
+
if ((src_reg->negate & 0xf) == NEGATE_XYZW)
src = ureg_negate(src);
switch(inst->op) {
case TGSI_OPCODE_BGNLOOP:
- case TGSI_OPCODE_CAL:
case TGSI_OPCODE_ELSE:
case TGSI_OPCODE_ENDLOOP:
case TGSI_OPCODE_IF:
case TGSI_OPCODE_UIF:
assert(num_dst == 0);
- ureg_label_insn(ureg,
- inst->op,
- src, num_src,
- get_label(t,
- inst->op == TGSI_OPCODE_CAL ? inst->function->sig_id : 0));
+ ureg_insn(ureg, inst->op, NULL, 0, src, num_src);
return;
case TGSI_OPCODE_TEX:
struct ureg_program *ureg,
int wpos_transform_const)
{
- const struct gl_fragment_program *fp =
- (const struct gl_fragment_program *) program;
struct pipe_screen *pscreen = st->pipe->screen;
GLfloat adjX = 0.0f;
GLfloat adjY[2] = { 0.0f, 0.0f };
* u,i -> l,h: (99.0 + 0.5) * -1 + 100 = 0.5
* u,h -> l,i: (99.5 + 0.5) * -1 + 100 = 0
*/
- if (fp->OriginUpperLeft) {
+ if (program->OriginUpperLeft) {
/* Fragment shader wants origin in upper-left */
if (pscreen->get_param(pscreen, PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT)) {
/* the driver supports upper-left origin */
assert(0);
}
- if (fp->PixelCenterInteger) {
+ if (program->PixelCenterInteger) {
/* Fragment shader wants pixel center integer */
if (pscreen->get_param(pscreen, PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER)) {
/* the driver supports pixel center integer */
}
static void
-emit_compute_block_size(const struct gl_program *program,
+emit_compute_block_size(const struct gl_program *prog,
struct ureg_program *ureg) {
- const struct gl_compute_program *cp =
- (const struct gl_compute_program *)program;
-
ureg_property(ureg, TGSI_PROPERTY_CS_FIXED_BLOCK_WIDTH,
- cp->LocalSize[0]);
+ prog->info.cs.local_size[0]);
ureg_property(ureg, TGSI_PROPERTY_CS_FIXED_BLOCK_HEIGHT,
- cp->LocalSize[1]);
+ prog->info.cs.local_size[1]);
ureg_property(ureg, TGSI_PROPERTY_CS_FIXED_BLOCK_DEPTH,
- cp->LocalSize[2]);
+ prog->info.cs.local_size[2]);
+}
+
+struct sort_inout_decls {
+ bool operator()(const struct inout_decl &a, const struct inout_decl &b) const {
+ return mapping[a.mesa_index] < mapping[b.mesa_index];
+ }
+
+ const GLuint *mapping;
+};
+
+/* Sort the given array of decls by the corresponding slot (TGSI file index).
+ *
+ * This is for the benefit of older drivers which are broken when the
+ * declarations aren't sorted in this way.
+ */
+static void
+sort_inout_decls_by_slot(struct inout_decl *decls,
+ unsigned count,
+ const GLuint mapping[])
+{
+ sort_inout_decls sorter;
+ sorter.mapping = mapping;
+ std::sort(decls, decls + count, sorter);
+}
+
+static unsigned
+st_translate_interp(enum glsl_interp_mode glsl_qual, GLuint varying)
+{
+ switch (glsl_qual) {
+ case INTERP_MODE_NONE:
+ if (varying == VARYING_SLOT_COL0 || varying == VARYING_SLOT_COL1)
+ return TGSI_INTERPOLATE_COLOR;
+ return TGSI_INTERPOLATE_PERSPECTIVE;
+ case INTERP_MODE_SMOOTH:
+ return TGSI_INTERPOLATE_PERSPECTIVE;
+ case INTERP_MODE_FLAT:
+ return TGSI_INTERPOLATE_CONSTANT;
+ case INTERP_MODE_NOPERSPECTIVE:
+ return TGSI_INTERPOLATE_LINEAR;
+ default:
+ assert(0 && "unexpected interp mode in st_translate_interp()");
+ return TGSI_INTERPOLATE_PERSPECTIVE;
+ }
}
/**
* \param inputSemanticIndex the semantic index (ex: which texcoord) for
* each input
* \param interpMode the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input
- * \param interpLocation the TGSI_INTERPOLATE_LOC_* location for each input
* \param numOutputs number of output registers used
* \param outputMapping maps Mesa fragment program outputs to TGSI
* generic outputs
const ubyte inputSemanticName[],
const ubyte inputSemanticIndex[],
const GLuint interpMode[],
- const GLuint interpLocation[],
GLuint numOutputs,
const GLuint outputMapping[],
const GLuint outputSlotToAttr[],
case PIPE_SHADER_GEOMETRY:
case PIPE_SHADER_TESS_EVAL:
case PIPE_SHADER_TESS_CTRL:
+ sort_inout_decls_by_slot(program->inputs, program->num_inputs, inputMapping);
+
for (i = 0; i < program->num_inputs; ++i) {
struct inout_decl *decl = &program->inputs[i];
unsigned slot = inputMapping[decl->mesa_index];
tgsi_usage_mask = TGSI_WRITEMASK_XYZW;
}
+ unsigned interp_mode = 0;
+ unsigned interp_location = 0;
+ if (procType == PIPE_SHADER_FRAGMENT) {
+ assert(interpMode);
+ interp_mode = interpMode[slot] != TGSI_INTERPOLATE_COUNT ?
+ interpMode[slot] :
+ st_translate_interp(decl->interp, inputSlotToAttr[slot]);
+
+ interp_location = decl->interp_loc;
+ }
+
src = ureg_DECL_fs_input_cyl_centroid_layout(ureg,
inputSemanticName[slot], inputSemanticIndex[slot],
- interpMode ? interpMode[slot] : 0, 0, interpLocation ? interpLocation[slot] : 0,
- slot, tgsi_usage_mask, decl->array_id, decl->size);
+ interp_mode, 0, interp_location, slot, tgsi_usage_mask,
+ decl->array_id, decl->size);
for (unsigned j = 0; j < decl->size; ++j) {
if (t->inputs[slot + j].File != TGSI_FILE_INPUT) {
case PIPE_SHADER_TESS_EVAL:
case PIPE_SHADER_TESS_CTRL:
case PIPE_SHADER_VERTEX:
+ sort_inout_decls_by_slot(program->outputs, program->num_outputs, outputMapping);
+
for (i = 0; i < program->num_outputs; ++i) {
struct inout_decl *decl = &program->outputs[i];
unsigned slot = outputMapping[decl->mesa_index];
dst = ureg_DECL_output_layout(ureg,
outputSemanticName[slot], outputSemanticIndex[slot],
+ decl->gs_out_streams,
slot, tgsi_usage_mask, decl->array_id, decl->size);
for (unsigned j = 0; j < decl->size; ++j) {
if (program->shader->info.EarlyFragmentTests)
ureg_property(ureg, TGSI_PROPERTY_FS_EARLY_DEPTH_STENCIL, 1);
- if (proginfo->InputsRead & VARYING_BIT_POS) {
+ if (proginfo->info.inputs_read & VARYING_BIT_POS) {
/* Must do this after setting up t->inputs. */
emit_wpos(st_context(ctx), t, proginfo, ureg,
program->wpos_transform_const);
}
- if (proginfo->InputsRead & VARYING_BIT_FACE)
+ if (proginfo->info.inputs_read & VARYING_BIT_FACE)
emit_face_var(ctx, t);
for (i = 0; i < numOutputs; i++) {
/* Declare misc input registers
*/
{
- GLbitfield sysInputs = proginfo->SystemValuesRead;
+ GLbitfield sysInputs = proginfo->info.system_values_read;
for (i = 0; sysInputs; i++) {
if (sysInputs & (1 << i)) {
/* Emit each instruction in turn:
*/
- foreach_in_list(glsl_to_tgsi_instruction, inst, &program->instructions) {
- set_insn_start(t, ureg_get_instruction_number(ureg));
+ foreach_in_list(glsl_to_tgsi_instruction, inst, &program->instructions)
compile_tgsi_instruction(t, inst);
- }
-
- /* Fix up all emitted labels:
- */
- for (i = 0; i < t->labels_count; i++) {
- ureg_fixup_label(ureg, t->labels[i].token,
- t->insn[t->labels[i].branch_target]);
- }
/* Set the next shader stage hint for VS and TES. */
switch (procType) {
if (t) {
free(t->arrays);
free(t->temps);
- free(t->insn);
- free(t->labels);
free(t->constants);
t->num_constants = 0;
free(t->immediates);
t->num_immediates = 0;
-
- if (t->error) {
- debug_printf("%s: translate error flag set\n", __func__);
- }
-
FREE(t);
}
{
glsl_to_tgsi_visitor* v;
struct gl_program *prog;
- GLenum target = _mesa_shader_stage_to_program(shader->Stage);
- bool progress;
struct gl_shader_compiler_options *options =
&ctx->Const.ShaderCompilerOptions[shader->Stage];
struct pipe_screen *pscreen = ctx->st->pipe->screen;
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 = new glsl_to_tgsi_visitor();
v->ctx = ctx;
v->have_fma = pscreen->get_shader_param(pscreen, ptarget,
PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED);
- _mesa_copy_linked_program_data(shader->Stage, shader_program, prog);
_mesa_generate_parameters_list_for_uniforms(shader_program, shader,
prog->Parameters);
/* Remove reads from output registers. */
- lower_output_reads(shader->Stage, shader->ir);
+ if (!pscreen->get_param(pscreen, PIPE_CAP_TGSI_CAN_READ_OUTPUTS))
+ lower_output_reads(shader->Stage, shader->ir);
/* Emit intermediate IR for main(). */
visit_exec_list(shader->ir, v);
- /* Now emit bodies for any functions that were used. */
- do {
- progress = GL_FALSE;
-
- foreach_in_list(function_entry, entry, &v->function_signatures) {
- if (!entry->bgn_inst) {
- v->current_function = entry;
-
- entry->bgn_inst = v->emit_asm(NULL, TGSI_OPCODE_BGNSUB);
- entry->bgn_inst->function = entry;
-
- visit_exec_list(&entry->sig->body, v);
-
- glsl_to_tgsi_instruction *last;
- last = (glsl_to_tgsi_instruction *)v->instructions.get_tail();
- if (last->op != TGSI_OPCODE_RET)
- v->emit_asm(NULL, TGSI_OPCODE_RET);
-
- glsl_to_tgsi_instruction *end;
- end = v->emit_asm(NULL, TGSI_OPCODE_ENDSUB);
- end->function = entry;
-
- progress = GL_TRUE;
- }
- }
- } while (progress);
-
#if 0
/* Print out some information (for debugging purposes) used by the
* optimization passes. */
_mesa_log("\n\n");
}
- prog->Instructions = NULL;
- prog->NumInstructions = 0;
-
do_set_program_inouts(shader->ir, prog, shader->Stage);
+ _mesa_copy_linked_program_data(shader_program, shader);
shrink_array_declarations(v->inputs, v->num_inputs,
- &prog->InputsRead, prog->DoubleInputsRead, &prog->PatchInputsRead);
+ &prog->info.inputs_read,
+ prog->info.double_inputs_read,
+ &prog->info.patch_inputs_read);
shrink_array_declarations(v->outputs, v->num_outputs,
- &prog->OutputsWritten, 0ULL, &prog->PatchOutputsWritten);
+ &prog->info.outputs_written, 0ULL,
+ &prog->info.patch_outputs_written);
count_resources(v, prog);
/* The GLSL IR won't be needed anymore. */
/* This must be done before the uniform storage is associated. */
if (shader->Stage == MESA_SHADER_FRAGMENT &&
- (prog->InputsRead & VARYING_BIT_POS ||
- prog->SystemValuesRead & (1 << SYSTEM_VALUE_FRAG_COORD))) {
+ (prog->info.inputs_read & VARYING_BIT_POS ||
+ prog->info.system_values_read & (1 << SYSTEM_VALUE_FRAG_COORD))) {
static const gl_state_index wposTransformState[STATE_LENGTH] = {
STATE_INTERNAL, STATE_FB_WPOS_Y_TRANSFORM
};
wposTransformState);
}
- _mesa_reference_program(ctx, &shader->Program, prog);
-
/* Avoid reallocation of the program parameter list, because the uniform
* storage is only associated with the original parameter list.
* This should be enough for Bitmap and DrawPixels constants.
* program constant) has to happen before creating this linkage.
*/
_mesa_associate_uniform_storage(ctx, shader_program, prog->Parameters);
- if (!shader_program->LinkStatus) {
+ if (!shader_program->data->LinkStatus) {
free_glsl_to_tgsi_visitor(v);
+ _mesa_reference_program(ctx, &shader->Program, NULL);
return NULL;
}
if (shader->NumShaderStorageBlocks)
*states |= new_ssbos;
- if (shader->NumAtomicBuffers)
+ if (prog->info.num_abos)
*states |= new_atomics;
}
return prog;
}
+/* See if there are unsupported control flow statements. */
+class ir_control_flow_info_visitor : public ir_hierarchical_visitor {
+private:
+ const struct gl_shader_compiler_options *options;
+public:
+ ir_control_flow_info_visitor(const struct gl_shader_compiler_options *options)
+ : options(options),
+ unsupported(false)
+ {
+ }
+
+ virtual ir_visitor_status visit_enter(ir_function *ir)
+ {
+ /* Other functions are skipped (same as glsl_to_tgsi). */
+ if (strcmp(ir->name, "main") == 0)
+ return visit_continue;
+
+ return visit_continue_with_parent;
+ }
+
+ virtual ir_visitor_status visit_enter(ir_call *ir)
+ {
+ if (!ir->callee->is_intrinsic()) {
+ unsupported = true; /* it's a function call */
+ return visit_stop;
+ }
+ return visit_continue;
+ }
+
+ virtual ir_visitor_status visit_enter(ir_return *ir)
+ {
+ if (options->EmitNoMainReturn) {
+ unsupported = true;
+ return visit_stop;
+ }
+ return visit_continue;
+ }
+
+ bool unsupported;
+};
+
+static bool
+has_unsupported_control_flow(exec_list *ir,
+ const struct gl_shader_compiler_options *options)
+{
+ ir_control_flow_info_visitor visitor(options);
+ visit_list_elements(&visitor, ir);
+ return visitor.unsupported;
+}
extern "C" {
st_link_shader(struct gl_context *ctx, struct gl_shader_program *prog)
{
struct pipe_screen *pscreen = ctx->st->pipe->screen;
- assert(prog->LinkStatus);
+ assert(prog->data->LinkStatus);
for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
if (prog->_LinkedShaders[i] == NULL)
continue;
- bool progress;
exec_list *ir = prog->_LinkedShaders[i]->ir;
gl_shader_stage stage = prog->_LinkedShaders[i]->Stage;
const struct gl_shader_compiler_options *options =
PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED);
bool have_dfrexp = pscreen->get_shader_param(pscreen, ptarget,
PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED);
+ unsigned if_threshold = pscreen->get_shader_param(pscreen, ptarget,
+ PIPE_SHADER_CAP_LOWER_IF_THRESHOLD);
/* If there are forms of indirect addressing that the driver
* cannot handle, perform the lowering pass.
lower_discard(ir);
}
- do {
- progress = false;
-
- 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;
-
- progress = lower_if_to_cond_assign(ir, options->MaxIfDepth) || progress;
-
- } while (progress);
+ if (ctx->Const.GLSLOptimizeConservatively) {
+ /* Do it once and repeat only if there's unsupported control flow. */
+ do {
+ do_common_optimization(ir, true, true, options,
+ ctx->Const.NativeIntegers);
+ lower_if_to_cond_assign((gl_shader_stage)i, ir,
+ options->MaxIfDepth, if_threshold);
+ } while (has_unsupported_control_flow(ir, options));
+ } else {
+ /* Repeat it until it stops making changes. */
+ bool progress;
+ do {
+ progress = do_common_optimization(ir, true, true, options,
+ ctx->Const.NativeIntegers);
+ progress |= lower_if_to_cond_assign((gl_shader_stage)i, ir,
+ options->MaxIfDepth, if_threshold);
+ } while (progress);
+ }
validate_ir_tree(ir);
}
linked_prog = get_mesa_program(ctx, prog, prog->_LinkedShaders[i]);
if (linked_prog) {
- _mesa_reference_program(ctx, &prog->_LinkedShaders[i]->Program,
- linked_prog);
if (!ctx->Driver.ProgramStringNotify(ctx,
_mesa_shader_stage_to_program(i),
linked_prog)) {
_mesa_reference_program(ctx, &prog->_LinkedShaders[i]->Program,
NULL);
- _mesa_reference_program(ctx, &linked_prog, NULL);
return GL_FALSE;
}
}
-
- _mesa_reference_program(ctx, &linked_prog, NULL);
}
return GL_TRUE;
struct pipe_stream_output_info *so)
{
struct gl_transform_feedback_info *info =
- &glsl_to_tgsi->shader_program->LinkedTransformFeedback;
+ glsl_to_tgsi->shader_program->xfb_program->sh.LinkedTransformFeedback;
st_translate_stream_output_info2(info, outputMapping, so);
}