*/
#define MAX_ARRAYS 256
-/* will be 4 for GLSL 4.00 */
-#define MAX_GLSL_TEXTURE_OFFSET 1
+#define MAX_GLSL_TEXTURE_OFFSET 4
class st_src_reg;
class st_dst_reg;
unsigned op;
st_dst_reg dst;
- st_src_reg src[3];
+ st_src_reg src[4];
/** Pointer to the ir source this tree came from for debugging */
ir_instruction *ir;
GLboolean cond_update;
int sampler; /**< sampler index */
int tex_target; /**< One of TEXTURE_*_INDEX */
GLboolean tex_shadow;
- struct tgsi_texture_offset tex_offsets[MAX_GLSL_TEXTURE_OFFSET];
+
+ st_src_reg tex_offsets[MAX_GLSL_TEXTURE_OFFSET];
unsigned tex_offset_num_offset;
int dead_mask; /**< Used in dead code elimination */
glsl_to_tgsi_instruction *emit(ir_instruction *ir, unsigned op,
st_dst_reg dst,
st_src_reg src0, st_src_reg src1, st_src_reg src2);
-
+
+ glsl_to_tgsi_instruction *emit(ir_instruction *ir, unsigned op,
+ st_dst_reg dst,
+ st_src_reg src0, st_src_reg src1,
+ st_src_reg src2, st_src_reg src3);
+
unsigned get_opcode(ir_instruction *ir, unsigned op,
st_dst_reg dst,
st_src_reg src0, st_src_reg src1);
int get_last_temp_write(int index);
void copy_propagate(void);
- void eliminate_dead_code(void);
- int eliminate_dead_code_advanced(void);
+ int eliminate_dead_code(void);
void merge_registers(void);
void renumber_registers(void);
glsl_to_tgsi_instruction *
glsl_to_tgsi_visitor::emit(ir_instruction *ir, unsigned op,
- st_dst_reg dst,
- st_src_reg src0, st_src_reg src1, st_src_reg src2)
+ st_dst_reg dst,
+ st_src_reg src0, st_src_reg src1,
+ st_src_reg src2, st_src_reg src3)
{
glsl_to_tgsi_instruction *inst = new(mem_ctx) glsl_to_tgsi_instruction();
int num_reladdr = 0, i;
num_reladdr += src0.reladdr != NULL || src0.reladdr2 != NULL;
num_reladdr += src1.reladdr != NULL || src1.reladdr2 != NULL;
num_reladdr += src2.reladdr != NULL || src2.reladdr2 != NULL;
+ num_reladdr += src3.reladdr != NULL || src3.reladdr2 != NULL;
+ reladdr_to_temp(ir, &src3, &num_reladdr);
reladdr_to_temp(ir, &src2, &num_reladdr);
reladdr_to_temp(ir, &src1, &num_reladdr);
reladdr_to_temp(ir, &src0, &num_reladdr);
inst->src[0] = src0;
inst->src[1] = src1;
inst->src[2] = src2;
+ inst->src[3] = src3;
inst->ir = ir;
inst->dead_mask = 0;
}
}
else {
- for (i=0; i<3; i++) {
+ for (i=0; i<4; i++) {
if(inst->src[i].reladdr) {
switch(inst->src[i].file) {
case PROGRAM_STATE_VAR:
return inst;
}
+glsl_to_tgsi_instruction *
+glsl_to_tgsi_visitor::emit(ir_instruction *ir, unsigned op,
+ st_dst_reg dst, st_src_reg src0,
+ st_src_reg src1, st_src_reg src2)
+{
+ return emit(ir, op, dst, src0, src1, src2, undef_src);
+}
glsl_to_tgsi_instruction *
glsl_to_tgsi_visitor::emit(ir_instruction *ir, unsigned op,
st_dst_reg dst, st_src_reg src0, st_src_reg src1)
{
- return emit(ir, op, dst, src0, src1, undef_src);
+ return emit(ir, op, dst, src0, src1, undef_src, undef_src);
}
glsl_to_tgsi_instruction *
st_dst_reg dst, st_src_reg src0)
{
assert(dst.writemask != 0);
- return emit(ir, op, dst, src0, undef_src, undef_src);
+ return emit(ir, op, dst, src0, undef_src, undef_src, undef_src);
}
glsl_to_tgsi_instruction *
glsl_to_tgsi_visitor::emit(ir_instruction *ir, unsigned op)
{
- return emit(ir, op, undef_dst, undef_src, undef_src, undef_src);
+ return emit(ir, op, undef_dst, undef_src, undef_src, undef_src, undef_src);
}
/**
st_src_reg src0, st_src_reg src1)
{
int type = GLSL_TYPE_FLOAT;
-
+
+ if (op == TGSI_OPCODE_MOV)
+ return op;
+
assert(src0.type != GLSL_TYPE_ARRAY);
assert(src0.type != GLSL_TYPE_STRUCT);
assert(src1.type != GLSL_TYPE_ARRAY);
case2fi(SSG, ISSG);
case3(ABS, IABS, IABS);
-
+
+ case2iu(IBFE, UBFE);
+ case2iu(IMSB, UMSB);
+ case2iu(IMUL_HI, UMUL_HI);
default: break;
}
/* Search immediate storage to see if we already have an identical
* immediate that we can use instead of adding a duplicate entry.
*/
- foreach_iter(exec_list_iterator, iter, this->immediates) {
- entry = (immediate_storage *)iter.get();
+ foreach_list(node, &this->immediates) {
+ entry = (immediate_storage *) node;
if (entry->size == size &&
entry->type == datatype &&
}
return size;
case GLSL_TYPE_SAMPLER:
+ case GLSL_TYPE_IMAGE:
/* Samplers take up one slot in UNIFORMS[], but they're baked in
* at link time.
*/
variable_storage *entry;
- foreach_iter(exec_list_iterator, iter, this->variables) {
- entry = (variable_storage *)iter.get();
+ foreach_list(node, &this->variables) {
+ entry = (variable_storage *) node;
if (entry->var == var)
return entry;
if (strcmp(ir->name, "gl_FragCoord") == 0) {
struct gl_fragment_program *fp = (struct gl_fragment_program *)this->prog;
- fp->OriginUpperLeft = ir->origin_upper_left;
- fp->PixelCenterInteger = ir->pixel_center_integer;
+ fp->OriginUpperLeft = ir->data.origin_upper_left;
+ fp->PixelCenterInteger = ir->data.pixel_center_integer;
}
- if (ir->mode == ir_var_uniform && strncmp(ir->name, "gl_", 3) == 0) {
+ if (ir->data.mode == ir_var_uniform && strncmp(ir->name, "gl_", 3) == 0) {
unsigned int i;
const ir_state_slot *const slots = ir->state_slots;
assert(ir->state_slots != NULL);
void
glsl_to_tgsi_visitor::visit(ir_loop *ir)
{
- /* Any normative loop bounds should have been lowered by
- * lower_bounded_loops().
- */
- assert(ir->normative_bound < 0);
-
emit(NULL, TGSI_OPCODE_BGNLOOP);
visit_exec_list(&ir->body_instructions, this);
assert(sig);
- foreach_iter(exec_list_iterator, iter, sig->body) {
- ir_instruction *ir = (ir_instruction *)iter.get();
+ foreach_list(node, &sig->body) {
+ ir_instruction *ir = (ir_instruction *) node;
ir->accept(this);
}
case ir_unop_any: {
assert(ir->operands[0]->type->is_vector());
- /* After the dot-product, the value will be an integer on the
- * range [0,4]. Zero stays zero, and positive values become 1.0.
- */
- glsl_to_tgsi_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 &&
- result_dst.type == GLSL_TYPE_FLOAT) {
- /* The clamping to [0,1] can be done for free in the fragment
- * shader with a saturate.
- */
- dp->saturate = true;
- } else if (result_dst.type == GLSL_TYPE_FLOAT) {
- /* 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.
- */
- st_src_reg slt_src = result_src;
- slt_src.negate = ~slt_src.negate;
- emit(ir, TGSI_OPCODE_SLT, result_dst, slt_src, st_src_reg_for_float(0.0));
- }
- else {
- /* Use SNE 0 if integers are being used as boolean values. */
- emit(ir, TGSI_OPCODE_SNE, result_dst, result_src, st_src_reg_for_int(0));
+ if (native_integers) {
+ int dst_swizzle = 0, op0_swizzle, i;
+ st_src_reg accum = op[0];
+
+ op0_swizzle = op[0].swizzle;
+ accum.swizzle = MAKE_SWIZZLE4(GET_SWZ(op0_swizzle, 0),
+ GET_SWZ(op0_swizzle, 0),
+ GET_SWZ(op0_swizzle, 0),
+ GET_SWZ(op0_swizzle, 0));
+ for (i = 0; i < 4; i++) {
+ if (result_dst.writemask & (1 << i)) {
+ dst_swizzle = MAKE_SWIZZLE4(i, i, i, i);
+ break;
+ }
+ }
+ assert(i != 4);
+ assert(ir->operands[0]->type->is_boolean());
+
+ /* OR all the components together, since they should be either 0 or ~0
+ */
+ switch (ir->operands[0]->type->vector_elements) {
+ case 4:
+ op[0].swizzle = MAKE_SWIZZLE4(GET_SWZ(op0_swizzle, 3),
+ GET_SWZ(op0_swizzle, 3),
+ GET_SWZ(op0_swizzle, 3),
+ GET_SWZ(op0_swizzle, 3));
+ emit(ir, TGSI_OPCODE_OR, result_dst, accum, op[0]);
+ accum = st_src_reg(result_dst);
+ accum.swizzle = dst_swizzle;
+ /* fallthrough */
+ case 3:
+ op[0].swizzle = MAKE_SWIZZLE4(GET_SWZ(op0_swizzle, 2),
+ GET_SWZ(op0_swizzle, 2),
+ GET_SWZ(op0_swizzle, 2),
+ GET_SWZ(op0_swizzle, 2));
+ emit(ir, TGSI_OPCODE_OR, result_dst, accum, op[0]);
+ accum = st_src_reg(result_dst);
+ accum.swizzle = dst_swizzle;
+ /* fallthrough */
+ case 2:
+ op[0].swizzle = MAKE_SWIZZLE4(GET_SWZ(op0_swizzle, 1),
+ GET_SWZ(op0_swizzle, 1),
+ GET_SWZ(op0_swizzle, 1),
+ GET_SWZ(op0_swizzle, 1));
+ emit(ir, TGSI_OPCODE_OR, result_dst, accum, op[0]);
+ break;
+ default:
+ assert(!"Unexpected vector size");
+ break;
+ }
+ } else {
+ /* After the dot-product, the value will be an integer on the
+ * range [0,4]. Zero stays zero, and positive values become 1.0.
+ */
+ glsl_to_tgsi_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 &&
+ result_dst.type == GLSL_TYPE_FLOAT) {
+ /* The clamping to [0,1] can be done for free in the fragment
+ * shader with a saturate.
+ */
+ dp->saturate = true;
+ } else if (result_dst.type == GLSL_TYPE_FLOAT) {
+ /* 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.
+ */
+ st_src_reg slt_src = result_src;
+ slt_src.negate = ~slt_src.negate;
+ emit(ir, TGSI_OPCODE_SLT, result_dst, slt_src, st_src_reg_for_float(0.0));
+ }
+ else {
+ /* Use SNE 0 if integers are being used as boolean values. */
+ emit(ir, TGSI_OPCODE_SNE, result_dst, result_src, st_src_reg_for_int(0));
+ }
}
break;
}
emit(ir, TGSI_OPCODE_CMP, result_dst, op[0], op[1], op[2]);
}
break;
+ case ir_triop_bitfield_extract:
+ emit(ir, TGSI_OPCODE_IBFE, result_dst, op[0], op[1], op[2]);
+ break;
+ case ir_quadop_bitfield_insert:
+ emit(ir, TGSI_OPCODE_BFI, result_dst, op[0], op[1], op[2], op[3]);
+ break;
+ case ir_unop_bitfield_reverse:
+ emit(ir, TGSI_OPCODE_BREV, result_dst, op[0]);
+ break;
+ case ir_unop_bit_count:
+ emit(ir, TGSI_OPCODE_POPC, result_dst, op[0]);
+ break;
+ case ir_unop_find_msb:
+ emit(ir, TGSI_OPCODE_IMSB, result_dst, op[0]);
+ break;
+ case ir_unop_find_lsb:
+ emit(ir, TGSI_OPCODE_LSB, result_dst, op[0]);
+ break;
+ case ir_binop_imul_high:
+ emit(ir, TGSI_OPCODE_IMUL_HI, result_dst, op[0], op[1]);
+ break;
+ case ir_triop_fma:
+ /* NOTE: Perhaps there should be a special opcode that enforces fused
+ * mul-add. Just use MAD for now.
+ */
+ emit(ir, TGSI_OPCODE_MAD, result_dst, op[0], op[1], op[2]);
+ break;
case ir_unop_pack_snorm_2x16:
case ir_unop_pack_unorm_2x16:
case ir_unop_pack_half_2x16:
case ir_unop_unpack_snorm_4x8:
case ir_unop_unpack_unorm_4x8:
case ir_binop_pack_half_2x16_split:
- case ir_unop_bitfield_reverse:
- case ir_unop_bit_count:
- case ir_unop_find_msb:
- case ir_unop_find_lsb:
case ir_binop_bfm:
- case ir_triop_fma:
case ir_triop_bfi:
- case ir_triop_bitfield_extract:
- case ir_quadop_bitfield_insert:
case ir_quadop_vector:
case ir_binop_vector_extract:
case ir_triop_vector_insert:
case ir_binop_ldexp:
case ir_binop_carry:
case ir_binop_borrow:
- case ir_binop_imul_high:
/* This operation is not supported, or should have already been handled.
*/
assert(!"Invalid ir opcode in glsl_to_tgsi_visitor::visit()");
ir_variable *var = ir->var;
if (!entry) {
- switch (var->mode) {
+ switch (var->data.mode) {
case ir_var_uniform:
entry = new(mem_ctx) variable_storage(var, PROGRAM_UNIFORM,
- var->location);
+ var->data.location);
this->variables.push_tail(entry);
break;
case ir_var_shader_in:
* generic attributes (glBindVertexLocation), and
* user-defined varyings.
*/
- assert(var->location != -1);
+ assert(var->data.location != -1);
entry = new(mem_ctx) variable_storage(var,
PROGRAM_INPUT,
- var->location);
+ var->data.location);
break;
case ir_var_shader_out:
- assert(var->location != -1);
+ assert(var->data.location != -1);
entry = new(mem_ctx) variable_storage(var,
PROGRAM_OUTPUT,
- var->location + var->index);
+ var->data.location
+ + var->data.index);
break;
case ir_var_system_value:
entry = new(mem_ctx) variable_storage(var,
PROGRAM_SYSTEM_VALUE,
- var->location);
+ var->data.location);
break;
case ir_var_auto:
case ir_var_temporary:
assert(!ir->lhs->type->is_scalar() && !ir->lhs->type->is_vector());
l.writemask = WRITEMASK_XYZW;
} else if (ir->lhs->type->is_scalar() &&
- ir->lhs->variable_referenced()->mode == ir_var_shader_out) {
+ ir->lhs->variable_referenced()->data.mode == ir_var_shader_out) {
/* FINISHME: This hack makes writing to gl_FragDepth, which lives in the
* FINISHME: W component of fragment shader output zero, work correctly.
*/
st_src_reg temp_base = get_temp(ir->type);
st_dst_reg temp = st_dst_reg(temp_base);
- foreach_iter(exec_list_iterator, iter, ir->components) {
- ir_constant *field_value = (ir_constant *)iter.get();
+ foreach_list(node, &ir->components) {
+ ir_constant *field_value = (ir_constant *) node;
int size = type_size(field_value->type);
assert(size > 0);
{
function_entry *entry;
- foreach_iter(exec_list_iterator, iter, this->function_signatures) {
- entry = (function_entry *)iter.get();
+ foreach_list(node, &this->function_signatures) {
+ entry = (function_entry *) node;
if (entry->sig == sig)
return entry;
entry->bgn_inst = NULL;
/* Allocate storage for all the parameters. */
- foreach_iter(exec_list_iterator, iter, sig->parameters) {
- ir_variable *param = (ir_variable *)iter.get();
+ foreach_list(node, &sig->parameters) {
+ ir_variable *param = (ir_variable *) node;
variable_storage *storage;
storage = find_variable_storage(param);
int i;
/* Process in parameters. */
- exec_list_iterator sig_iter = sig->parameters.iterator();
- foreach_iter(exec_list_iterator, iter, *ir) {
- ir_rvalue *param_rval = (ir_rvalue *)iter.get();
- ir_variable *param = (ir_variable *)sig_iter.get();
+ 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->mode == ir_var_function_in ||
- param->mode == ir_var_function_inout) {
+ if (param->data.mode == ir_var_function_in ||
+ param->data.mode == ir_var_function_inout) {
variable_storage *storage = find_variable_storage(param);
assert(storage);
r.index++;
}
}
-
- sig_iter.next();
}
- assert(!sig_iter.has_next());
/* Emit call instruction */
call_inst = emit(ir, TGSI_OPCODE_CAL);
call_inst->function = entry;
/* Process out parameters. */
- sig_iter = sig->parameters.iterator();
- foreach_iter(exec_list_iterator, iter, *ir) {
- ir_rvalue *param_rval = (ir_rvalue *)iter.get();
- ir_variable *param = (ir_variable *)sig_iter.get();
+ 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->mode == ir_var_function_out ||
- param->mode == ir_var_function_inout) {
+ if (param->data.mode == ir_var_function_out ||
+ param->data.mode == ir_var_function_inout) {
variable_storage *storage = find_variable_storage(param);
assert(storage);
r.index++;
}
}
-
- sig_iter.next();
}
- assert(!sig_iter.has_next());
/* Process return value. */
this->result = entry->return_reg;
void
glsl_to_tgsi_visitor::visit(ir_texture *ir)
{
- st_src_reg result_src, coord, cube_sc, lod_info, projector, dx, dy, offset, sample_index;
+ st_src_reg result_src, coord, cube_sc, lod_info, projector, dx, dy, offset[MAX_GLSL_TEXTURE_OFFSET], sample_index, component;
st_dst_reg result_dst, coord_dst, cube_sc_dst;
glsl_to_tgsi_instruction *inst = NULL;
unsigned opcode = TGSI_OPCODE_NOP;
const glsl_type *sampler_type = ir->sampler->type;
bool is_cube_array = false;
+ unsigned i;
/* if we are a cube array sampler */
if ((sampler_type->sampler_dimensionality == GLSL_SAMPLER_DIM_CUBE &&
opcode = (is_cube_array && ir->shadow_comparitor) ? TGSI_OPCODE_TEX2 : TGSI_OPCODE_TEX;
if (ir->offset) {
ir->offset->accept(this);
- offset = this->result;
+ offset[0] = this->result;
}
break;
case ir_txb:
lod_info = this->result;
if (ir->offset) {
ir->offset->accept(this);
- offset = this->result;
+ offset[0] = this->result;
}
break;
case ir_txl:
lod_info = this->result;
if (ir->offset) {
ir->offset->accept(this);
- offset = this->result;
+ offset[0] = this->result;
}
break;
case ir_txd:
dy = this->result;
if (ir->offset) {
ir->offset->accept(this);
- offset = this->result;
+ offset[0] = this->result;
}
break;
case ir_txs:
lod_info = this->result;
if (ir->offset) {
ir->offset->accept(this);
- offset = this->result;
+ offset[0] = this->result;
}
break;
case ir_txf_ms:
ir->lod_info.sample_index->accept(this);
sample_index = this->result;
break;
- case ir_lod:
- assert(!"Unexpected ir_lod opcode");
- break;
case ir_tg4:
- assert(!"Unexpected ir_tg4 opcode");
+ opcode = TGSI_OPCODE_TG4;
+ ir->lod_info.component->accept(this);
+ component = this->result;
+ if (ir->offset) {
+ ir->offset->accept(this);
+ if (ir->offset->type->base_type == GLSL_TYPE_ARRAY) {
+ const glsl_type *elt_type = ir->offset->type->fields.array;
+ for (i = 0; i < ir->offset->type->length; i++) {
+ offset[i] = this->result;
+ offset[i].index += i * type_size(elt_type);
+ offset[i].type = elt_type->base_type;
+ offset[i].swizzle = swizzle_for_size(elt_type->vector_elements);
+ }
+ } else {
+ offset[0] = this->result;
+ }
+ }
+ break;
+ case ir_lod:
+ opcode = TGSI_OPCODE_LODQ;
break;
case ir_query_levels:
assert(!"Unexpected ir_query_levels opcode");
inst = emit(ir, opcode, result_dst, coord, lod_info);
} else if (opcode == TGSI_OPCODE_TEX2) {
inst = emit(ir, opcode, result_dst, coord, cube_sc);
- } else
+ } else if (opcode == TGSI_OPCODE_TG4) {
+ if (is_cube_array && ir->shadow_comparitor) {
+ inst = emit(ir, opcode, result_dst, coord, cube_sc);
+ } else {
+ inst = emit(ir, opcode, result_dst, coord, component);
+ }
+ } else
inst = emit(ir, opcode, result_dst, coord);
if (ir->shadow_comparitor)
this->prog);
if (ir->offset) {
- inst->tex_offset_num_offset = 1;
- inst->tex_offsets[0].Index = offset.index;
- inst->tex_offsets[0].File = offset.file;
- inst->tex_offsets[0].SwizzleX = GET_SWZ(offset.swizzle, 0);
- inst->tex_offsets[0].SwizzleY = GET_SWZ(offset.swizzle, 1);
- inst->tex_offsets[0].SwizzleZ = GET_SWZ(offset.swizzle, 2);
+ for (i = 0; i < MAX_GLSL_TEXTURE_OFFSET && offset[i].file != PROGRAM_UNDEFINED; i++)
+ inst->tex_offsets[i] = offset[i];
+ inst->tex_offset_num_offset = i;
}
switch (sampler_type->sampler_dimensionality) {
{
v->samplers_used = 0;
- foreach_iter(exec_list_iterator, iter, v->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &v->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
if (is_tex_instruction(inst->op)) {
v->samplers_used |= 1 << inst->sampler;
memset(tempWrites, 0, sizeof(unsigned) * MAX_TEMPS);
memset(outputWrites, 0, sizeof(outputWrites));
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &this->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
unsigned prevWriteMask = 0;
/* Give up if we encounter relative addressing or flow control. */
void
glsl_to_tgsi_visitor::rename_temp_register(int index, int new_index)
{
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &this->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
unsigned j;
for (j=0; j < num_inst_src_regs(inst->op); j++) {
inst->src[j].index = new_index;
}
}
+
+ for (j=0; j < inst->tex_offset_num_offset; j++) {
+ if (inst->tex_offsets[j].file == PROGRAM_TEMPORARY &&
+ inst->tex_offsets[j].index == index) {
+ inst->tex_offsets[j].index = new_index;
+ }
+ }
if (inst->dst.file == PROGRAM_TEMPORARY && inst->dst.index == index) {
inst->dst.index = new_index;
int loop_start = -1; /* index of the first active BGNLOOP (if any) */
unsigned i = 0, j;
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &this->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
for (j=0; j < num_inst_src_regs(inst->op); j++) {
if (inst->src[j].file == PROGRAM_TEMPORARY &&
return (depth == 0) ? i : loop_start;
}
}
+ for (j=0; j < inst->tex_offset_num_offset; j++) {
+ if (inst->tex_offsets[j].file == PROGRAM_TEMPORARY &&
+ inst->tex_offsets[j].index == index) {
+ return (depth == 0) ? i : loop_start;
+ }
+ }
if (inst->op == TGSI_OPCODE_BGNLOOP) {
if(depth++ == 0)
int loop_start = -1; /* index of the first active BGNLOOP (if any) */
int i = 0;
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &this->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
if (inst->dst.file == PROGRAM_TEMPORARY && inst->dst.index == index) {
return (depth == 0) ? i : loop_start;
int last = -1; /* index of last instruction that reads the temporary */
unsigned i = 0, j;
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &this->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
for (j=0; j < num_inst_src_regs(inst->op); j++) {
if (inst->src[j].file == PROGRAM_TEMPORARY &&
last = (depth == 0) ? i : -2;
}
}
+ for (j=0; j < inst->tex_offset_num_offset; j++) {
+ if (inst->tex_offsets[j].file == PROGRAM_TEMPORARY &&
+ inst->tex_offsets[j].index == index)
+ last = (depth == 0) ? i : -2;
+ }
if (inst->op == TGSI_OPCODE_BGNLOOP)
depth++;
int last = -1; /* index of last instruction that writes to the temporary */
int i = 0;
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &this->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
if (inst->dst.file == PROGRAM_TEMPORARY && inst->dst.index == index)
last = (depth == 0) ? i : -2;
int *acp_level = rzalloc_array(mem_ctx, int, this->next_temp * 4);
int level = 0;
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &this->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
assert(inst->dst.file != PROGRAM_TEMPORARY
|| inst->dst.index < this->next_temp);
}
/*
- * Tracks available PROGRAM_TEMPORARY registers for dead code elimination.
+ * On a basic block basis, tracks available PROGRAM_TEMPORARY registers for dead
+ * code elimination.
*
* The glsl_to_tgsi_visitor lazily produces code assuming that this pass
* will occur. As an example, a TXP production after copy propagation but
* and after this pass:
*
* 0: TXP TEMP[2], INPUT[4].xyyw, texture[0], 2D;
- *
- * FIXME: assumes that all functions are inlined (no support for BGNSUB/ENDSUB)
- * FIXME: doesn't eliminate all dead code inside of loops; it steps around them
- */
-void
-glsl_to_tgsi_visitor::eliminate_dead_code(void)
-{
- int i;
-
- for (i=0; i < this->next_temp; i++) {
- int last_read = get_last_temp_read(i);
- int j = 0;
-
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
-
- if (inst->dst.file == PROGRAM_TEMPORARY && inst->dst.index == i &&
- j > last_read)
- {
- iter.remove();
- delete inst;
- }
-
- j++;
- }
- }
-}
-
-/*
- * On a basic block basis, tracks available PROGRAM_TEMPORARY registers for dead
- * code elimination. This is less primitive than eliminate_dead_code(), as it
- * is per-channel and can detect consecutive writes without a read between them
- * as dead code. However, there is some dead code that can be eliminated by
- * eliminate_dead_code() but not this function - for example, this function
- * cannot eliminate an instruction writing to a register that is never read and
- * is the only instruction writing to that register.
- *
- * The glsl_to_tgsi_visitor lazily produces code assuming that this pass
- * will occur.
*/
int
-glsl_to_tgsi_visitor::eliminate_dead_code_advanced(void)
+glsl_to_tgsi_visitor::eliminate_dead_code(void)
{
glsl_to_tgsi_instruction **writes = rzalloc_array(mem_ctx,
glsl_to_tgsi_instruction *,
int level = 0;
int removed = 0;
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &this->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
assert(inst->dst.file != PROGRAM_TEMPORARY
|| inst->dst.index < this->next_temp);
}
}
}
+ for (unsigned i = 0; i < inst->tex_offset_num_offset; i++) {
+ if (inst->tex_offsets[i].file == PROGRAM_TEMPORARY && inst->tex_offsets[i].reladdr){
+ /* Any temporary might be read, so no dead code elimination
+ * across this instruction.
+ */
+ memset(writes, 0, sizeof(*writes) * this->next_temp * 4);
+ } else if (inst->tex_offsets[i].file == PROGRAM_TEMPORARY) {
+ /* Clear where it's used as src. */
+ int src_chans = 1 << GET_SWZ(inst->tex_offsets[i].swizzle, 0);
+ src_chans |= 1 << GET_SWZ(inst->tex_offsets[i].swizzle, 1);
+ src_chans |= 1 << GET_SWZ(inst->tex_offsets[i].swizzle, 2);
+ src_chans |= 1 << GET_SWZ(inst->tex_offsets[i].swizzle, 3);
+
+ for (int c = 0; c < 4; c++) {
+ if (src_chans & (1 << c)) {
+ writes[4 * inst->tex_offsets[i].index + c] = NULL;
+ }
+ }
+ }
+ }
break;
}
/* Now actually remove the instructions that are completely dead and update
* the writemask of other instructions with dead channels.
*/
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list_safe(node, &this->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
if (!inst->dead_mask || !inst->dst.writemask)
continue;
else if ((inst->dst.writemask & ~inst->dead_mask) == 0) {
- iter.remove();
+ inst->remove();
delete inst;
removed++;
} else
/* Now copy the instructions from the original glsl_to_tgsi_visitor into the
* new visitor. */
- foreach_iter(exec_list_iterator, iter, original->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &original->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
glsl_to_tgsi_instruction *newinst;
st_src_reg src_regs[3];
/* Now copy the instructions from the original glsl_to_tgsi_visitor into the
* new visitor. */
- foreach_iter(exec_list_iterator, iter, original->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
+ foreach_list(node, &original->instructions) {
+ glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *) node;
glsl_to_tgsi_instruction *newinst;
st_src_reg src_regs[3];
struct ureg_dst address[2];
struct ureg_src samplers[PIPE_MAX_SAMPLERS];
struct ureg_src systemValues[SYSTEM_VALUE_MAX];
-
+ struct tgsi_texture_offset tex_offsets[MAX_GLSL_TEXTURE_OFFSET];
unsigned array_sizes[MAX_ARRAYS];
const GLuint *inputMapping;
static unsigned mesa_sysval_to_semantic[SYSTEM_VALUE_MAX] = {
TGSI_SEMANTIC_FACE,
TGSI_SEMANTIC_VERTEXID,
- TGSI_SEMANTIC_INSTANCEID
+ TGSI_SEMANTIC_INSTANCEID,
+ TGSI_SEMANTIC_SAMPLEID,
+ TGSI_SEMANTIC_SAMPLEPOS,
+ TGSI_SEMANTIC_SAMPLEMASK,
+ TGSI_SEMANTIC_INVOCATIONID,
};
/**
break;
case TGSI_PROCESSOR_FRAGMENT:
- if (dst_reg->index >= FRAG_RESULT_COLOR) {
+ if (dst_reg->index == FRAG_RESULT_COLOR ||
+ dst_reg->index >= FRAG_RESULT_DATA0) {
dst = ureg_saturate(dst);
}
break;
static struct tgsi_texture_offset
translate_tex_offset(struct st_translate *t,
- const struct tgsi_texture_offset *in_offset)
+ const st_src_reg *in_offset, int idx)
{
struct tgsi_texture_offset offset;
struct ureg_src imm_src;
+ struct ureg_dst dst;
+ int array;
- assert(in_offset->File == PROGRAM_IMMEDIATE);
- imm_src = t->immediates[in_offset->Index];
+ switch (in_offset->file) {
+ case PROGRAM_IMMEDIATE:
+ imm_src = t->immediates[in_offset->index];
+
+ offset.File = imm_src.File;
+ offset.Index = imm_src.Index;
+ offset.SwizzleX = imm_src.SwizzleX;
+ offset.SwizzleY = imm_src.SwizzleY;
+ offset.SwizzleZ = imm_src.SwizzleZ;
+ offset.Padding = 0;
+ break;
+ case PROGRAM_TEMPORARY:
+ imm_src = ureg_src(t->temps[in_offset->index]);
+ offset.File = imm_src.File;
+ offset.Index = imm_src.Index;
+ offset.SwizzleX = GET_SWZ(in_offset->swizzle, 0);
+ offset.SwizzleY = GET_SWZ(in_offset->swizzle, 1);
+ offset.SwizzleZ = GET_SWZ(in_offset->swizzle, 2);
+ offset.Padding = 0;
+ break;
+ case PROGRAM_ARRAY:
+ array = in_offset->index >> 16;
- offset.File = imm_src.File;
- offset.Index = imm_src.Index;
- offset.SwizzleX = imm_src.SwizzleX;
- offset.SwizzleY = imm_src.SwizzleY;
- offset.SwizzleZ = imm_src.SwizzleZ;
- offset.File = TGSI_FILE_IMMEDIATE;
- offset.Padding = 0;
+ assert(array >= 0);
+ assert(array < (int) Elements(t->arrays));
+ dst = t->arrays[array];
+ offset.File = dst.File;
+ offset.Index = dst.Index + (in_offset->index & 0xFFFF) - 0x8000;
+ offset.SwizzleX = GET_SWZ(in_offset->swizzle, 0);
+ offset.SwizzleY = GET_SWZ(in_offset->swizzle, 1);
+ offset.SwizzleZ = GET_SWZ(in_offset->swizzle, 2);
+ offset.Padding = 0;
+ break;
+ default:
+ break;
+ }
return offset;
}
case TGSI_OPCODE_TEX2:
case TGSI_OPCODE_TXB2:
case TGSI_OPCODE_TXL2:
+ case TGSI_OPCODE_TG4:
+ case TGSI_OPCODE_LODQ:
src[num_src++] = t->samplers[inst->sampler];
for (i = 0; i < inst->tex_offset_num_offset; i++) {
- texoffsets[i] = translate_tex_offset(t, &inst->tex_offsets[i]);
+ texoffsets[i] = translate_tex_offset(t, &inst->tex_offsets[i], i);
}
tex_target = st_translate_texture_target(inst->tex_target, inst->tex_shadow);
TGSI_SEMANTIC_COLOR,
outputSemanticIndex[i]);
break;
+ case TGSI_SEMANTIC_SAMPLEMASK:
+ t->outputs[i] = ureg_DECL_output(ureg,
+ TGSI_SEMANTIC_SAMPLEMASK,
+ outputSemanticIndex[i]);
+ /* TODO: If we ever support more than 32 samples, this will have
+ * to become an array.
+ */
+ t->outputs[i] = ureg_writemask(t->outputs[i], TGSI_WRITEMASK_X);
+ break;
default:
assert(!"fragment shader outputs must be POSITION/STENCIL/COLOR");
ret = PIPE_ERROR_BAD_INPUT;
goto out;
}
i = 0;
- foreach_iter(exec_list_iterator, iter, program->immediates) {
- immediate_storage *imm = (immediate_storage *)iter.get();
+ foreach_list(node, &program->immediates) {
+ immediate_storage *imm = (immediate_storage *) node;
assert(i < program->num_immediates);
t->immediates[i++] = emit_immediate(t, imm->values, imm->type, imm->size);
}
assert(i == program->num_immediates);
/* texture samplers */
- for (i = 0; i < ctx->Const.FragmentProgram.MaxTextureImageUnits; i++) {
+ for (i = 0; i < ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxTextureImageUnits; i++) {
if (program->samplers_used & (1 << i)) {
t->samplers[i] = ureg_DECL_sampler(ureg, i);
}
/* Emit each instruction in turn:
*/
- foreach_iter(exec_list_iterator, iter, program->instructions) {
+ foreach_list(n, &program->instructions) {
set_insn_start(t, ureg_get_instruction_number(ureg));
- compile_tgsi_instruction(t, (glsl_to_tgsi_instruction *)iter.get(),
- clamp_color);
+ compile_tgsi_instruction(t, (glsl_to_tgsi_instruction *) n, clamp_color);
}
/* Fix up all emitted labels:
* prog->ParameterValues to get reallocated (e.g., anything that adds a
* program constant) has to happen before creating this linkage.
*/
- for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
+ for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
if (program->shader_program->_LinkedShaders[i] == NULL)
continue;
}
/* ----------------------------- End TGSI code ------------------------------ */
+
+static unsigned
+shader_stage_to_ptarget(gl_shader_stage stage)
+{
+ switch (stage) {
+ case MESA_SHADER_VERTEX:
+ return PIPE_SHADER_VERTEX;
+ case MESA_SHADER_FRAGMENT:
+ return PIPE_SHADER_FRAGMENT;
+ case MESA_SHADER_GEOMETRY:
+ return PIPE_SHADER_GEOMETRY;
+ case MESA_SHADER_COMPUTE:
+ return PIPE_SHADER_COMPUTE;
+ }
+
+ assert(!"should not be reached");
+ return PIPE_SHADER_VERTEX;
+}
+
+
/**
* Convert a shader's GLSL IR into a Mesa gl_program, although without
* generating Mesa IR.
{
glsl_to_tgsi_visitor* v;
struct gl_program *prog;
- GLenum target;
+ GLenum target = _mesa_shader_stage_to_program(shader->Stage);
bool progress;
struct gl_shader_compiler_options *options =
- &ctx->ShaderCompilerOptions[_mesa_shader_type_to_index(shader->Type)];
+ &ctx->ShaderCompilerOptions[_mesa_shader_enum_to_shader_stage(shader->Type)];
struct pipe_screen *pscreen = ctx->st->pipe->screen;
- unsigned ptarget;
-
- switch (shader->Type) {
- case GL_VERTEX_SHADER:
- target = GL_VERTEX_PROGRAM_ARB;
- ptarget = PIPE_SHADER_VERTEX;
- break;
- case GL_FRAGMENT_SHADER:
- target = GL_FRAGMENT_PROGRAM_ARB;
- ptarget = PIPE_SHADER_FRAGMENT;
- break;
- case GL_GEOMETRY_SHADER:
- target = GL_GEOMETRY_PROGRAM_NV;
- ptarget = PIPE_SHADER_GEOMETRY;
- break;
- default:
- assert(!"should not be reached");
- return NULL;
- }
+ unsigned ptarget = shader_stage_to_ptarget(shader->Stage);
validate_ir_tree(shader->ir);
do {
progress = GL_FALSE;
- foreach_iter(exec_list_iterator, iter, v->function_signatures) {
- function_entry *entry = (function_entry *)iter.get();
+ foreach_list(node, &v->function_signatures) {
+ function_entry *entry = (function_entry *) node;
if (!entry->bgn_inst) {
v->current_function = entry;
/* Perform optimizations on the instructions in the glsl_to_tgsi_visitor. */
v->simplify_cmp();
v->copy_propagate();
- while (v->eliminate_dead_code_advanced());
+ while (v->eliminate_dead_code());
- v->eliminate_dead_code();
v->merge_registers();
v->renumber_registers();
/* Write the END instruction. */
v->emit(NULL, TGSI_OPCODE_END);
- if (ctx->Shader.Flags & GLSL_DUMP) {
+ if (ctx->_Shader->Flags & GLSL_DUMP) {
printf("\n");
printf("GLSL IR for linked %s program %d:\n",
- _mesa_glsl_shader_target_name(shader->Type),
+ _mesa_shader_stage_to_string(shader->Stage),
shader_program->Name);
- _mesa_print_ir(shader->ir, NULL);
+ _mesa_print_ir(stdout, shader->ir, NULL);
printf("\n");
printf("\n");
fflush(stdout);
prog->Instructions = NULL;
prog->NumInstructions = 0;
- do_set_program_inouts(shader->ir, prog, shader->Type);
+ do_set_program_inouts(shader->ir, prog, shader->Stage);
count_resources(v, prog);
_mesa_reference_program(ctx, &shader->Program, prog);
stgp->Base.InputType = shader_program->Geom.InputType;
stgp->Base.OutputType = shader_program->Geom.OutputType;
stgp->Base.VerticesOut = shader_program->Geom.VerticesOut;
+ stgp->Base.Invocations = shader_program->Geom.Invocations;
break;
default:
assert(!"should not be reached");
shader = rzalloc(NULL, struct gl_shader);
if (shader) {
shader->Type = type;
+ shader->Stage = _mesa_shader_enum_to_shader_stage(type);
shader->Name = name;
_mesa_init_shader(ctx, shader);
}
GLboolean
st_link_shader(struct gl_context *ctx, struct gl_shader_program *prog)
{
+ struct pipe_screen *pscreen = ctx->st->pipe->screen;
assert(prog->LinkStatus);
- for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
+ for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
if (prog->_LinkedShaders[i] == NULL)
continue;
bool progress;
exec_list *ir = prog->_LinkedShaders[i]->ir;
const struct gl_shader_compiler_options *options =
- &ctx->ShaderCompilerOptions[_mesa_shader_type_to_index(prog->_LinkedShaders[i]->Type)];
+ &ctx->ShaderCompilerOptions[_mesa_shader_enum_to_shader_stage(prog->_LinkedShaders[i]->Type)];
/* If there are forms of indirect addressing that the driver
* cannot handle, perform the lowering pass.
lower_packing_builtins(ir, lower_inst);
}
+ if (!pscreen->get_param(pscreen, PIPE_CAP_TEXTURE_GATHER_OFFSETS))
+ lower_offset_arrays(ir);
do_mat_op_to_vec(ir);
lower_instructions(ir,
MOD_TO_FRACT |
DIV_TO_MUL_RCP |
EXP_TO_EXP2 |
LOG_TO_LOG2 |
+ LDEXP_TO_ARITH |
+ CARRY_TO_ARITH |
+ BORROW_TO_ARITH |
(options->EmitNoPow ? POW_TO_EXP2 : 0) |
(!ctx->Const.NativeIntegers ? INT_DIV_TO_MUL_RCP : 0));
progress = do_lower_jumps(ir, true, true, options->EmitNoMainReturn, options->EmitNoCont, options->EmitNoLoops) || progress;
- progress = do_common_optimization(ir, true, true,
- options->MaxUnrollIterations, options)
+ progress = do_common_optimization(ir, true, true, options,
+ ctx->Const.NativeIntegers)
|| progress;
progress = lower_if_to_cond_assign(ir, options->MaxIfDepth) || progress;
} while (progress);
- lower_bounded_loops(ir);
-
validate_ir_tree(ir);
}
- for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
+ for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
struct gl_program *linked_prog;
if (prog->_LinkedShaders[i] == NULL)
_mesa_reference_program(ctx, &prog->_LinkedShaders[i]->Program,
linked_prog);
if (!ctx->Driver.ProgramStringNotify(ctx,
- _mesa_program_index_to_target(i),
+ _mesa_shader_stage_to_program(i),
linked_prog)) {
_mesa_reference_program(ctx, &prog->_LinkedShaders[i]->Program,
NULL);
projects / mesa.git / blobdiff