#include "st_glsl_to_tgsi.h"
-#include "glsl_parser_extras.h"
-#include "ir_optimization.h"
+#include "compiler/glsl/glsl_parser_extras.h"
+#include "compiler/glsl/ir_optimization.h"
+#include "compiler/glsl/program.h"
#include "main/errors.h"
#include "main/shaderobj.h"
#include "main/uniforms.h"
#include "main/shaderapi.h"
+#include "main/shaderimage.h"
#include "program/prog_instruction.h"
-#include "program/sampler.h"
#include "pipe/p_context.h"
#include "pipe/p_screen.h"
#include "util/u_memory.h"
#include "st_program.h"
#include "st_mesa_to_tgsi.h"
+#include "st_format.h"
+#include "st_glsl_types.h"
+#include "st_nir.h"
-#define PROGRAM_IMMEDIATE PROGRAM_FILE_MAX
#define PROGRAM_ANY_CONST ((1 << PROGRAM_STATE_VAR) | \
(1 << PROGRAM_CONSTANT) | \
(1 << PROGRAM_UNIFORM))
this->has_index2 = false;
this->double_reg2 = false;
this->array_id = 0;
+ this->is_double_vertex_input = false;
}
- st_src_reg(gl_register_file file, int index, int type)
+ st_src_reg(gl_register_file file, int index, enum glsl_base_type type)
{
this->type = type;
this->file = file;
this->has_index2 = false;
this->double_reg2 = false;
this->array_id = 0;
+ this->is_double_vertex_input = false;
}
- st_src_reg(gl_register_file file, int index, int type, int index2D)
+ st_src_reg(gl_register_file file, int index, enum glsl_base_type type, int index2D)
{
this->type = type;
this->file = file;
this->has_index2 = false;
this->double_reg2 = false;
this->array_id = 0;
+ this->is_double_vertex_input = false;
}
st_src_reg()
this->has_index2 = false;
this->double_reg2 = false;
this->array_id = 0;
+ this->is_double_vertex_input = false;
}
explicit st_src_reg(st_dst_reg reg);
int index2D;
GLuint swizzle; /**< SWIZZLE_XYZWONEZERO swizzles from Mesa. */
int negate; /**< NEGATE_XYZW mask from mesa */
- int type; /** GLSL_TYPE_* from GLSL IR (enum glsl_base_type) */
+ enum glsl_base_type type; /** GLSL_TYPE_* from GLSL IR (enum glsl_base_type) */
/** Register index should be offset by the integer in this reg. */
st_src_reg *reladdr;
st_src_reg *reladdr2;
*/
bool double_reg2;
unsigned array_id;
+ bool is_double_vertex_input;
};
class st_dst_reg {
public:
- st_dst_reg(gl_register_file file, int writemask, int type, int index)
+ st_dst_reg(gl_register_file file, int writemask, enum glsl_base_type type, int index)
{
this->file = file;
this->index = index;
this->index2D = 0;
this->writemask = writemask;
- this->cond_mask = COND_TR;
this->reladdr = NULL;
this->reladdr2 = NULL;
this->has_index2 = false;
this->array_id = 0;
}
- st_dst_reg(gl_register_file file, int writemask, int type)
+ st_dst_reg(gl_register_file file, int writemask, enum glsl_base_type type)
{
this->file = file;
this->index = 0;
this->index2D = 0;
this->writemask = writemask;
- this->cond_mask = COND_TR;
this->reladdr = NULL;
this->reladdr2 = NULL;
this->has_index2 = false;
this->index = 0;
this->index2D = 0;
this->writemask = 0;
- this->cond_mask = COND_TR;
this->reladdr = NULL;
this->reladdr2 = NULL;
this->has_index2 = false;
int index; /**< temporary index, VERT_ATTRIB_*, VARYING_SLOT_*, etc. */
int index2D;
int writemask; /**< Bitfield of WRITEMASK_[XYZW] */
- GLuint cond_mask:4;
- int type; /** GLSL_TYPE_* from GLSL IR (enum glsl_base_type) */
+ enum glsl_base_type type; /** GLSL_TYPE_* from GLSL IR (enum glsl_base_type) */
/** Register index should be offset by the integer in this reg. */
st_src_reg *reladdr;
st_src_reg *reladdr2;
this->has_index2 = reg.has_index2;
this->double_reg2 = false;
this->array_id = reg.array_id;
+ this->is_double_vertex_input = false;
}
st_dst_reg::st_dst_reg(st_src_reg reg)
this->file = reg.file;
this->index = reg.index;
this->writemask = WRITEMASK_XYZW;
- this->cond_mask = COND_TR;
this->reladdr = reg.reladdr;
this->index2D = reg.index2D;
this->reladdr2 = reg.reladdr2;
GLboolean cond_update;
bool saturate;
st_src_reg sampler; /**< sampler register */
+ int sampler_base;
int sampler_array_size; /**< 1-based size of sampler array, 1 if not array */
int tex_target; /**< One of TEXTURE_*_INDEX */
glsl_base_type tex_type;
GLboolean tex_shadow;
+ unsigned image_format;
st_src_reg tex_offsets[MAX_GLSL_TEXTURE_OFFSET];
unsigned tex_offset_num_offset;
int dead_mask; /**< Used in dead code elimination */
+ st_src_reg buffer; /**< buffer register */
+ unsigned buffer_access; /**< buffer access type */
+
class function_entry *function; /* Set on TGSI_OPCODE_CAL or TGSI_OPCODE_BGNSUB */
const struct tgsi_opcode_info *info;
};
unsigned mesa_index;
unsigned array_id;
unsigned array_size;
+ enum glsl_base_type array_type;
};
+static enum glsl_base_type
+find_array_type(struct array_decl *arrays, unsigned count, unsigned array_id)
+{
+ unsigned i;
+
+ for (i = 0; i < count; i++) {
+ struct array_decl *decl = &arrays[i];
+
+ if (array_id == decl->array_id) {
+ return decl->array_type;
+ }
+ }
+ return GLSL_TYPE_ERROR;
+}
+
struct rename_reg_pair {
int old_reg;
int new_reg;
struct gl_context *ctx;
struct gl_program *prog;
struct gl_shader_program *shader_program;
- struct gl_shader *shader;
+ struct gl_linked_shader *shader;
struct gl_shader_compiler_options *options;
int next_temp;
unsigned num_output_arrays;
int num_address_regs;
- int samplers_used;
+ uint32_t samplers_used;
glsl_base_type sampler_types[PIPE_MAX_SAMPLERS];
int sampler_targets[PIPE_MAX_SAMPLERS]; /**< One of TGSI_TEXTURE_* */
+ int buffers_used;
+ int images_used;
+ int image_targets[PIPE_MAX_SHADER_IMAGES];
+ unsigned image_formats[PIPE_MAX_SHADER_IMAGES];
bool indirect_addr_consts;
int wpos_transform_const;
bool native_integers;
bool have_sqrt;
bool have_fma;
+ bool use_shared_memory;
variable_storage *find_variable_storage(ir_variable *var);
st_src_reg st_src_reg_for_double(double val);
st_src_reg st_src_reg_for_float(float val);
st_src_reg st_src_reg_for_int(int val);
- st_src_reg st_src_reg_for_type(int type, int val);
+ st_src_reg st_src_reg_for_type(enum glsl_base_type type, int val);
/**
* \name Visit methods
virtual void visit(ir_barrier *);
/*@}*/
+ void visit_expression(ir_expression *, st_src_reg *) ATTRIBUTE_NOINLINE;
+
+ void visit_atomic_counter_intrinsic(ir_call *);
+ void visit_ssbo_intrinsic(ir_call *);
+ void visit_membar_intrinsic(ir_call *);
+ void visit_shared_intrinsic(ir_call *);
+ void visit_image_intrinsic(ir_call *);
+
st_src_reg result;
/** List of variable_storage */
void emit_arl(ir_instruction *ir, st_dst_reg dst, st_src_reg src0);
+ void get_deref_offsets(ir_dereference *ir,
+ unsigned *array_size,
+ unsigned *base,
+ unsigned *index,
+ st_src_reg *reladdr);
+ void calc_deref_offsets(ir_dereference *head,
+ ir_dereference *tail,
+ unsigned *array_elements,
+ unsigned *base,
+ unsigned *index,
+ st_src_reg *indirect,
+ unsigned *location);
+
bool try_emit_mad(ir_expression *ir,
int mul_operand);
bool try_emit_mad_for_and_not(ir_expression *ir,
return size_swizzles[size - 1];
}
+static bool
+is_resource_instruction(unsigned opcode)
+{
+ switch (opcode) {
+ case TGSI_OPCODE_RESQ:
+ case TGSI_OPCODE_LOAD:
+ case TGSI_OPCODE_ATOMUADD:
+ case TGSI_OPCODE_ATOMXCHG:
+ case TGSI_OPCODE_ATOMCAS:
+ case TGSI_OPCODE_ATOMAND:
+ case TGSI_OPCODE_ATOMOR:
+ case TGSI_OPCODE_ATOMXOR:
+ case TGSI_OPCODE_ATOMUMIN:
+ case TGSI_OPCODE_ATOMUMAX:
+ case TGSI_OPCODE_ATOMIMIN:
+ case TGSI_OPCODE_ATOMIMAX:
+ return true;
+ default:
+ return false;
+ }
+}
+
static unsigned
num_inst_dst_regs(const glsl_to_tgsi_instruction *op)
{
static unsigned
num_inst_src_regs(const glsl_to_tgsi_instruction *op)
{
- return op->info->is_tex ? op->info->num_src - 1 : op->info->num_src;
+ return op->info->is_tex || is_resource_instruction(op->op) ?
+ op->info->num_src - 1 : op->info->num_src;
}
glsl_to_tgsi_instruction *
{
glsl_to_tgsi_instruction *inst = new(mem_ctx) glsl_to_tgsi_instruction();
int num_reladdr = 0, i, j;
+ bool dst_is_64bit[2];
op = get_opcode(ir, op, dst, src0, src1);
}
}
- this->instructions.push_tail(inst);
-
/*
* This section contains the double processing.
* GLSL just represents doubles as single channel values,
* GLSL [0].z -> TGSI [1].xy
* GLSL [0].w -> TGSI [1].zw
*/
- if (inst->dst[0].type == GLSL_TYPE_DOUBLE || inst->dst[1].type == GLSL_TYPE_DOUBLE ||
- inst->src[0].type == GLSL_TYPE_DOUBLE) {
+ for (j = 0; j < 2; j++) {
+ dst_is_64bit[j] = glsl_base_type_is_64bit(inst->dst[j].type);
+ if (!dst_is_64bit[j] && inst->dst[j].file == PROGRAM_OUTPUT && inst->dst[j].type == GLSL_TYPE_ARRAY) {
+ enum glsl_base_type type = find_array_type(this->output_arrays, this->num_output_arrays, inst->dst[j].array_id);
+ if (glsl_base_type_is_64bit(type))
+ dst_is_64bit[j] = true;
+ }
+ }
+
+ if (dst_is_64bit[0] || dst_is_64bit[1] ||
+ glsl_base_type_is_64bit(inst->src[0].type)) {
glsl_to_tgsi_instruction *dinst = NULL;
int initial_src_swz[4], initial_src_idx[4];
int initial_dst_idx[2], initial_dst_writemask[2];
/* select the writemask for dst0 or dst1 */
- unsigned writemask = inst->dst[0].file == PROGRAM_UNDEFINED ? inst->dst[1].writemask : inst->dst[0].writemask;
+ unsigned writemask = inst->dst[1].file == PROGRAM_UNDEFINED ? inst->dst[0].writemask : inst->dst[1].writemask;
/* copy out the writemask, index and swizzles for all src/dsts. */
for (j = 0; j < 2; j++) {
* scan all the components in the dst writemask
* generate an instruction for each of them if required.
*/
+ st_src_reg addr;
while (writemask) {
int i = u_bit_scan(&writemask);
+ /* before emitting the instruction, see if we have to adjust store
+ * address */
+ if (i > 1 && 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);
+ emit_asm(ir, TGSI_OPCODE_UADD, st_dst_reg(addr),
+ inst->src[0], st_src_reg_for_int(16));
+ }
+
+
/* first time use previous instruction */
if (dinst == NULL) {
dinst = inst;
*dinst = *inst;
dinst->next = NULL;
dinst->prev = NULL;
- this->instructions.push_tail(dinst);
}
+ this->instructions.push_tail(dinst);
/* modify the destination if we are splitting */
for (j = 0; j < 2; j++) {
- if (dinst->dst[j].type == GLSL_TYPE_DOUBLE) {
+ if (dst_is_64bit[j]) {
dinst->dst[j].writemask = (i & 1) ? WRITEMASK_ZW : WRITEMASK_XY;
dinst->dst[j].index = initial_dst_idx[j];
- if (i > 1)
+ if (i > 1) {
+ if (dinst->op == TGSI_OPCODE_STORE) {
+ dinst->src[0] = addr;
+ } else {
dinst->dst[j].index++;
+ }
+ }
} else {
/* if we aren't writing to a double, just get the bit of the initial writemask
for this channel */
for (j = 0; j < 4; j++) {
int swz = GET_SWZ(initial_src_swz[j], i);
- if (dinst->src[j].type == GLSL_TYPE_DOUBLE) {
+ if (glsl_base_type_is_64bit(dinst->src[j].type)) {
dinst->src[j].index = initial_src_idx[j];
if (swz > 1) {
dinst->src[j].double_reg2 = true;
- F2D is a float src0, DLDEXP is integer src1 */
if (op == TGSI_OPCODE_F2D ||
op == TGSI_OPCODE_DLDEXP ||
- (op == TGSI_OPCODE_UCMP && dinst->dst[0].type == GLSL_TYPE_DOUBLE)) {
+ (op == TGSI_OPCODE_UCMP && dst_is_64bit[0])) {
dinst->src[j].swizzle = MAKE_SWIZZLE4(swz, swz, swz, swz);
}
}
}
}
inst = dinst;
+ } else {
+ this->instructions.push_tail(inst);
}
st_dst_reg dst,
st_src_reg src0, st_src_reg src1)
{
- int type = GLSL_TYPE_FLOAT;
+ enum glsl_base_type type = GLSL_TYPE_FLOAT;
if (op == TGSI_OPCODE_MOV)
return op;
assert(src1.type != GLSL_TYPE_ARRAY);
assert(src1.type != GLSL_TYPE_STRUCT);
- if (src0.type == GLSL_TYPE_DOUBLE || src1.type == GLSL_TYPE_DOUBLE)
+ if (is_resource_instruction(op))
+ type = src1.type;
+ else if (src0.type == GLSL_TYPE_DOUBLE || src1.type == GLSL_TYPE_DOUBLE)
type = GLSL_TYPE_DOUBLE;
else if (src0.type == GLSL_TYPE_FLOAT || src1.type == GLSL_TYPE_FLOAT)
type = GLSL_TYPE_FLOAT;
case3fid(FLR, FLR, DFLR);
case3fid(ROUND, ROUND, DROUND);
+ case2iu(ATOMIMAX, ATOMUMAX);
+ case2iu(ATOMIMIN, ATOMUMIN);
+
default: break;
}
}
st_src_reg
-glsl_to_tgsi_visitor::st_src_reg_for_type(int type, int val)
+glsl_to_tgsi_visitor::st_src_reg_for_type(enum glsl_base_type type, int val)
{
if (native_integers)
return type == GLSL_TYPE_FLOAT ? st_src_reg_for_float(val) :
}
static int
-type_size(const struct glsl_type *type)
+attrib_type_size(const struct glsl_type *type, bool is_vs_input)
{
- unsigned int i;
- int size;
-
- switch (type->base_type) {
- case GLSL_TYPE_UINT:
- case GLSL_TYPE_INT:
- case GLSL_TYPE_FLOAT:
- case GLSL_TYPE_BOOL:
- if (type->is_matrix()) {
- return type->matrix_columns;
- } else {
- /* Regardless of size of vector, it gets a vec4. This is bad
- * packing for things like floats, but otherwise arrays become a
- * mess. Hopefully a later pass over the code can pack scalars
- * down if appropriate.
- */
- return 1;
- }
- break;
- case GLSL_TYPE_DOUBLE:
- if (type->is_matrix()) {
- if (type->vector_elements <= 2)
- return type->matrix_columns;
- else
- return type->matrix_columns * 2;
- } else {
- /* For doubles if we have a double or dvec2 they fit in one
- * vec4, else they need 2 vec4s.
- */
- if (type->vector_elements <= 2)
- return 1;
- else
- return 2;
- }
- break;
- case GLSL_TYPE_ARRAY:
- assert(type->length > 0);
- return type_size(type->fields.array) * type->length;
- case GLSL_TYPE_STRUCT:
- size = 0;
- for (i = 0; i < type->length; i++) {
- size += type_size(type->fields.structure[i].type);
- }
- return size;
- case GLSL_TYPE_SAMPLER:
- case GLSL_TYPE_IMAGE:
- case GLSL_TYPE_SUBROUTINE:
- /* Samplers take up one slot in UNIFORMS[], but they're baked in
- * at link time.
- */
- return 1;
- case GLSL_TYPE_ATOMIC_UINT:
- case GLSL_TYPE_INTERFACE:
- case GLSL_TYPE_VOID:
- case GLSL_TYPE_ERROR:
- assert(!"Invalid type in type_size");
- break;
- }
- return 0;
+ return st_glsl_attrib_type_size(type, is_vs_input);
}
+static int
+type_size(const struct glsl_type *type)
+{
+ return st_glsl_type_size(type);
+}
/**
* If the given GLSL type is an array or matrix or a structure containing
if (reg->reladdr2) emit_arl(ir, address_reg2, *reg->reladdr2);
if (*num_reladdr != 1) {
- st_src_reg temp = get_temp(glsl_type::vec4_type);
+ st_src_reg temp = get_temp(reg->type == GLSL_TYPE_DOUBLE ? glsl_type::dvec4_type : glsl_type::vec4_type);
emit_asm(ir, TGSI_OPCODE_MOV, st_dst_reg(temp), *reg);
*reg = temp;
void
glsl_to_tgsi_visitor::visit(ir_expression *ir)
{
- unsigned int operand;
st_src_reg op[ARRAY_SIZE(ir->operands)];
- st_src_reg result_src;
- st_dst_reg result_dst;
/* Quick peephole: Emit MAD(a, b, c) instead of ADD(MUL(a, b), c)
*/
if (ir->operation == ir_quadop_vector)
assert(!"ir_quadop_vector should have been lowered");
- for (operand = 0; operand < ir->get_num_operands(); operand++) {
+ for (unsigned int operand = 0; operand < ir->get_num_operands(); operand++) {
this->result.file = PROGRAM_UNDEFINED;
ir->operands[operand]->accept(this);
if (this->result.file == PROGRAM_UNDEFINED) {
assert(!ir->operands[operand]->type->is_matrix());
}
+ visit_expression(ir, op);
+}
+
+/* The non-recursive part of the expression visitor lives in a separate
+ * function and should be prevented from being inlined, to avoid a stack
+ * explosion when deeply nested expressions are visited.
+ */
+void
+glsl_to_tgsi_visitor::visit_expression(ir_expression* ir, st_src_reg *op)
+{
+ st_src_reg result_src;
+ st_dst_reg result_dst;
+
int vector_elements = ir->operands[0]->type->vector_elements;
if (ir->operands[1]) {
vector_elements = MAX2(vector_elements,
}
break;
- case ir_unop_any: {
- assert(ir->operands[0]->type->is_vector());
-
- 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_asm(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_asm(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_asm(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_asm(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_asm(ir, TGSI_OPCODE_SNE, result_dst, result_src, st_src_reg_for_int(0));
- }
- }
- break;
- }
-
case ir_binop_logic_xor:
if (native_integers)
emit_asm(ir, TGSI_OPCODE_XOR, result_dst, op[0], op[1]);
if (have_sqrt) {
emit_scalar(ir, TGSI_OPCODE_SQRT, result_dst, op[0]);
} else {
- /* sqrt(x) = x * rsq(x). */
- emit_scalar(ir, TGSI_OPCODE_RSQ, result_dst, op[0]);
- emit_asm(ir, TGSI_OPCODE_MUL, result_dst, result_src, op[0]);
- /* For incoming channels <= 0, set the result to 0. */
- op[0].negate = ~op[0].negate;
- emit_asm(ir, TGSI_OPCODE_CMP, result_dst,
- op[0], result_src, st_src_reg_for_float(0.0));
+ /* This is the only instruction sequence that makes the game "Risen"
+ * render correctly. ABS is not required for the game, but since GLSL
+ * declares negative values as "undefined", allowing us to do whatever
+ * 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_RCP, result_dst, result_src);
}
break;
case ir_unop_rsq:
case ir_unop_u2i:
/* Converting between signed and unsigned integers is a no-op. */
result_src = op[0];
+ result_src.type = result_dst.type;
break;
case ir_unop_b2i:
if (native_integers) {
emit_asm(ir, TGSI_OPCODE_TRUNC, result_dst, op[0]);
break;
case ir_unop_bitcast_f2i:
- result_src = op[0];
- result_src.type = GLSL_TYPE_INT;
- break;
case ir_unop_bitcast_f2u:
- result_src = op[0];
- result_src.type = GLSL_TYPE_UINT;
+ /* Make sure we don't propagate the negate modifier to integer opcodes. */
+ if (op[0].negate)
+ emit_asm(ir, TGSI_OPCODE_MOV, result_dst, op[0]);
+ else
+ result_src = op[0];
+ result_src.type = ir->operation == ir_unop_bitcast_f2i ? GLSL_TYPE_INT :
+ GLSL_TYPE_UINT;
break;
case ir_unop_bitcast_i2f:
case ir_unop_bitcast_u2f:
}
cbuf.swizzle = swizzle_for_size(ir->type->vector_elements);
- if (cbuf.type == GLSL_TYPE_DOUBLE)
+ if (glsl_base_type_is_64bit(cbuf.type))
cbuf.swizzle += MAKE_SWIZZLE4(const_offset % 16 / 8,
const_offset % 16 / 8,
const_offset % 16 / 8,
case ir_unop_interpolate_at_centroid:
emit_asm(ir, TGSI_OPCODE_INTERP_CENTROID, result_dst, op[0]);
break;
- case ir_binop_interpolate_at_offset:
- emit_asm(ir, TGSI_OPCODE_INTERP_OFFSET, result_dst, op[0], op[1]);
+ case ir_binop_interpolate_at_offset: {
+ /* The y coordinate needs to be flipped for the default fb */
+ static const gl_state_index transform_y_state[STATE_LENGTH]
+ = { STATE_INTERNAL, STATE_FB_WPOS_Y_TRANSFORM };
+
+ unsigned transform_y_index =
+ _mesa_add_state_reference(this->prog->Parameters,
+ transform_y_state);
+
+ st_src_reg transform_y = st_src_reg(PROGRAM_STATE_VAR,
+ transform_y_index,
+ glsl_type::vec4_type);
+ transform_y.swizzle = SWIZZLE_XXXX;
+
+ st_src_reg temp = get_temp(glsl_type::vec2_type);
+ st_dst_reg temp_dst = st_dst_reg(temp);
+
+ emit_asm(ir, TGSI_OPCODE_MOV, temp_dst, op[1]);
+ temp_dst.writemask = WRITEMASK_Y;
+ emit_asm(ir, TGSI_OPCODE_MUL, temp_dst, transform_y, op[1]);
+ emit_asm(ir, TGSI_OPCODE_INTERP_OFFSET, result_dst, op[0], temp);
break;
+ }
case ir_binop_interpolate_at_sample:
emit_asm(ir, TGSI_OPCODE_INTERP_SAMPLE, result_dst, op[0], op[1]);
break;
}
break;
+ case ir_unop_pack_half_2x16:
+ emit_asm(ir, TGSI_OPCODE_PK2H, result_dst, op[0]);
+ break;
+ case ir_unop_unpack_half_2x16:
+ emit_asm(ir, TGSI_OPCODE_UP2H, result_dst, op[0]);
+ break;
+
+ case ir_unop_get_buffer_size: {
+ ir_constant *const_offset = ir->operands[0]->as_constant();
+ st_src_reg buffer(
+ PROGRAM_BUFFER,
+ ctx->Const.Program[shader->Stage].MaxAtomicBuffers +
+ (const_offset ? const_offset->value.u[0] : 0),
+ GLSL_TYPE_UINT);
+ if (!const_offset) {
+ buffer.reladdr = ralloc(mem_ctx, st_src_reg);
+ *buffer.reladdr = op[0];
+ emit_arl(ir, sampler_reladdr, op[0]);
+ }
+ emit_asm(ir, TGSI_OPCODE_RESQ, result_dst)->buffer = buffer;
+ break;
+ }
+
+ case ir_unop_vote_any:
+ emit_asm(ir, TGSI_OPCODE_VOTE_ANY, result_dst, op[0]);
+ break;
+ case ir_unop_vote_all:
+ emit_asm(ir, TGSI_OPCODE_VOTE_ALL, result_dst, op[0]);
+ break;
+ case ir_unop_vote_eq:
+ emit_asm(ir, TGSI_OPCODE_VOTE_EQ, result_dst, op[0]);
+ break;
+
case ir_unop_pack_snorm_2x16:
case ir_unop_pack_unorm_2x16:
- case ir_unop_pack_half_2x16:
case ir_unop_pack_snorm_4x8:
case ir_unop_pack_unorm_4x8:
case ir_unop_unpack_snorm_2x16:
case ir_unop_unpack_unorm_2x16:
- case ir_unop_unpack_half_2x16:
- case ir_unop_unpack_half_2x16_split_x:
- case ir_unop_unpack_half_2x16_split_y:
case ir_unop_unpack_snorm_4x8:
case ir_unop_unpack_unorm_4x8:
- case ir_binop_pack_half_2x16_split:
- case ir_binop_bfm:
- case ir_triop_bfi:
case ir_quadop_vector:
case ir_binop_vector_extract:
case ir_triop_vector_insert:
*/
assert(!"Invalid ir opcode in glsl_to_tgsi_visitor::visit()");
break;
-
- case ir_unop_get_buffer_size:
- assert(!"Not implemented yet");
- break;
}
this->result = result_src;
switch (var->data.mode) {
case ir_var_uniform:
entry = new(mem_ctx) variable_storage(var, PROGRAM_UNIFORM,
- var->data.location);
+ var->data.param_index);
this->variables.push_tail(entry);
break;
case ir_var_shader_in:
decl->mesa_index = var->data.location;
decl->array_id = num_input_arrays + 1;
- if (is_2d)
+ if (is_2d) {
decl->array_size = type_size(var->type->fields.array);
- else
+ decl->array_type = var->type->fields.array->without_array()->base_type;
+ } else {
decl->array_size = type_size(var->type);
+ decl->array_type = var->type->without_array()->base_type;
+ }
num_input_arrays++;
entry = new(mem_ctx) variable_storage(var,
decl->mesa_index = var->data.location;
decl->array_id = num_output_arrays + 1;
- if (is_2d)
+ if (is_2d) {
decl->array_size = type_size(var->type->fields.array);
- else
+ decl->array_type = var->type->fields.array->without_array()->base_type;
+ } else {
decl->array_size = type_size(var->type);
+ decl->array_type = var->type->without_array()->base_type;
+ }
num_output_arrays++;
entry = new(mem_ctx) variable_storage(var,
this->result = st_src_reg(entry->file, entry->index, var->type);
this->result.array_id = entry->array_id;
+ if (this->shader->Stage == MESA_SHADER_VERTEX && var->data.mode == ir_var_shader_in && var->type->is_double())
+ this->result.is_double_vertex_input = true;
if (!native_integers)
this->result.type = GLSL_TYPE_FLOAT;
}
static void
shrink_array_declarations(struct array_decl *arrays, unsigned count,
GLbitfield64 usage_mask,
+ GLbitfield64 double_usage_mask,
GLbitfield patch_usage_mask)
{
- unsigned i, j;
+ unsigned i;
+ int j;
/* Fix array declarations by removing unused array elements at both ends
* of the arrays. For example, mat4[3] where only mat[1] is used.
struct array_decl *decl = &arrays[i];
/* Shrink the beginning. */
- for (j = 0; j < decl->array_size; j++) {
+ for (j = 0; j < (int)decl->array_size; j++) {
if (decl->mesa_index >= VARYING_SLOT_PATCH0) {
if (patch_usage_mask &
BITFIELD64_BIT(decl->mesa_index - VARYING_SLOT_PATCH0 + j))
else {
if (usage_mask & BITFIELD64_BIT(decl->mesa_index+j))
break;
+ if (double_usage_mask & BITFIELD64_BIT(decl->mesa_index+j-1))
+ break;
}
decl->mesa_index++;
else {
if (usage_mask & BITFIELD64_BIT(decl->mesa_index+j))
break;
+ if (double_usage_mask & BITFIELD64_BIT(decl->mesa_index+j-1))
+ break;
}
decl->array_size--;
element_size = 1;
if (index) {
+
+ if (this->prog->Target == GL_VERTEX_PROGRAM_ARB &&
+ src.file == PROGRAM_INPUT)
+ element_size = attrib_type_size(ir->type, true);
if (is_2D) {
src.index2D = index->value.i[0];
src.has_index2 = true;
if (type->is_matrix()) {
const struct glsl_type *vec_type;
- vec_type = glsl_type::get_instance(GLSL_TYPE_FLOAT,
+ vec_type = glsl_type::get_instance(type->is_double() ? GLSL_TYPE_DOUBLE : GLSL_TYPE_FLOAT,
type->vector_elements, 1);
for (int i = 0; i < type->matrix_columns; i++) {
}
l->index++;
r->index++;
+ if (type->is_dual_slot()) {
+ l->index++;
+ if (r->is_double_vertex_input == false)
+ r->index++;
+ }
}
void
*/
if (ir->write_mask == 0) {
assert(!ir->lhs->type->is_scalar() && !ir->lhs->type->is_vector());
- l.writemask = WRITEMASK_XYZW;
+
+ if (ir->lhs->type->is_array() || ir->lhs->type->without_array()->is_matrix()) {
+ if (ir->lhs->type->without_array()->is_64bit()) {
+ switch (ir->lhs->type->without_array()->vector_elements) {
+ case 1:
+ l.writemask = WRITEMASK_X;
+ break;
+ case 2:
+ l.writemask = WRITEMASK_XY;
+ break;
+ case 3:
+ l.writemask = WRITEMASK_XYZ;
+ break;
+ case 4:
+ l.writemask = WRITEMASK_XYZW;
+ break;
+ }
+ } else
+ l.writemask = WRITEMASK_XYZW;
+ }
} else if (ir->lhs->type->is_scalar() &&
- !ir->lhs->type->is_double() &&
+ !ir->lhs->type->is_64bit() &&
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_dst_reg mat_column = st_dst_reg(mat);
for (i = 0; i < ir->type->matrix_columns; i++) {
- assert(ir->type->base_type == GLSL_TYPE_FLOAT);
- values = (gl_constant_value *) &ir->value.f[i * ir->type->vector_elements];
-
- src = st_src_reg(file, -1, ir->type->base_type);
- src.index = add_constant(file,
- values,
- ir->type->vector_elements,
- GL_FLOAT,
- &src.swizzle);
- emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ switch (ir->type->base_type) {
+ case GLSL_TYPE_FLOAT:
+ values = (gl_constant_value *) &ir->value.f[i * ir->type->vector_elements];
+ src = st_src_reg(file, -1, ir->type->base_type);
+ src.index = add_constant(file,
+ values,
+ ir->type->vector_elements,
+ GL_FLOAT,
+ &src.swizzle);
+ emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ break;
+ case GLSL_TYPE_DOUBLE:
+ values = (gl_constant_value *) &ir->value.d[i * ir->type->vector_elements];
+ src = st_src_reg(file, -1, ir->type->base_type);
+ src.index = add_constant(file,
+ values,
+ ir->type->vector_elements,
+ GL_DOUBLE,
+ &src.swizzle);
+ if (ir->type->vector_elements >= 2) {
+ mat_column.writemask = WRITEMASK_XY;
+ src.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_X, SWIZZLE_Y);
+ emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ } else {
+ mat_column.writemask = WRITEMASK_X;
+ src.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X);
+ emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ }
+ src.index++;
+ if (ir->type->vector_elements > 2) {
+ if (ir->type->vector_elements == 4) {
+ mat_column.writemask = WRITEMASK_ZW;
+ src.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_X, SWIZZLE_Y);
+ emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ } else {
+ mat_column.writemask = WRITEMASK_Z;
+ src.swizzle = MAKE_SWIZZLE4(SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y);
+ emit_asm(ir, TGSI_OPCODE_MOV, mat_column, src);
+ mat_column.writemask = WRITEMASK_XYZW;
+ src.swizzle = SWIZZLE_XYZW;
+ }
+ mat_column.index++;
+ }
+ break;
+ default:
+ unreachable("Illegal matrix constant type.\n");
+ break;
+ }
mat_column.index++;
}
-
this->result = mat;
return;
}
return entry;
}
+void
+glsl_to_tgsi_visitor::visit_atomic_counter_intrinsic(ir_call *ir)
+{
+ const char *callee = ir->callee->function_name();
+ exec_node *param = ir->actual_parameters.get_head();
+ ir_dereference *deref = static_cast<ir_dereference *>(param);
+ ir_variable *location = deref->variable_referenced();
+
+ st_src_reg buffer(
+ PROGRAM_BUFFER, location->data.binding, GLSL_TYPE_ATOMIC_UINT);
+
+ /* Calculate the surface offset */
+ st_src_reg offset;
+ unsigned array_size = 0, base = 0, index = 0;
+
+ get_deref_offsets(deref, &array_size, &base, &index, &offset);
+
+ if (offset.file != PROGRAM_UNDEFINED) {
+ emit_asm(ir, TGSI_OPCODE_MUL, st_dst_reg(offset),
+ offset, st_src_reg_for_int(ATOMIC_COUNTER_SIZE));
+ emit_asm(ir, TGSI_OPCODE_ADD, st_dst_reg(offset),
+ offset, st_src_reg_for_int(location->data.offset + index * ATOMIC_COUNTER_SIZE));
+ } else {
+ offset = st_src_reg_for_int(location->data.offset + index * ATOMIC_COUNTER_SIZE);
+ }
+
+ ir->return_deref->accept(this);
+ st_dst_reg dst(this->result);
+ dst.writemask = WRITEMASK_X;
+
+ glsl_to_tgsi_instruction *inst;
+
+ if (!strcmp("__intrinsic_atomic_read", callee)) {
+ inst = emit_asm(ir, TGSI_OPCODE_LOAD, dst, offset);
+ } else if (!strcmp("__intrinsic_atomic_increment", callee)) {
+ inst = emit_asm(ir, TGSI_OPCODE_ATOMUADD, dst, offset,
+ st_src_reg_for_int(1));
+ } else if (!strcmp("__intrinsic_atomic_predecrement", callee)) {
+ inst = emit_asm(ir, TGSI_OPCODE_ATOMUADD, dst, offset,
+ st_src_reg_for_int(-1));
+ emit_asm(ir, TGSI_OPCODE_ADD, dst, this->result, st_src_reg_for_int(-1));
+ } else {
+ param = param->get_next();
+ ir_rvalue *val = ((ir_instruction *)param)->as_rvalue();
+ val->accept(this);
+
+ st_src_reg data = this->result, data2 = undef_src;
+ unsigned opcode;
+ if (!strcmp("__intrinsic_atomic_add", callee))
+ opcode = TGSI_OPCODE_ATOMUADD;
+ else if (!strcmp("__intrinsic_atomic_min", callee))
+ opcode = TGSI_OPCODE_ATOMIMIN;
+ else if (!strcmp("__intrinsic_atomic_max", callee))
+ opcode = TGSI_OPCODE_ATOMIMAX;
+ else if (!strcmp("__intrinsic_atomic_and", callee))
+ opcode = TGSI_OPCODE_ATOMAND;
+ else if (!strcmp("__intrinsic_atomic_or", callee))
+ opcode = TGSI_OPCODE_ATOMOR;
+ else if (!strcmp("__intrinsic_atomic_xor", callee))
+ opcode = TGSI_OPCODE_ATOMXOR;
+ else if (!strcmp("__intrinsic_atomic_exchange", callee))
+ opcode = TGSI_OPCODE_ATOMXCHG;
+ else if (!strcmp("__intrinsic_atomic_comp_swap", callee)) {
+ opcode = TGSI_OPCODE_ATOMCAS;
+ param = param->get_next();
+ val = ((ir_instruction *)param)->as_rvalue();
+ val->accept(this);
+ data2 = this->result;
+ } else if (!strcmp("__intrinsic_atomic_sub", callee)) {
+ opcode = TGSI_OPCODE_ATOMUADD;
+ st_src_reg res = get_temp(glsl_type::uvec4_type);
+ st_dst_reg dstres = st_dst_reg(res);
+ dstres.writemask = dst.writemask;
+ emit_asm(ir, TGSI_OPCODE_INEG, dstres, data);
+ data = res;
+ } else {
+ assert(!"Unexpected intrinsic");
+ return;
+ }
+
+ inst = emit_asm(ir, opcode, dst, offset, data, data2);
+ }
+
+ inst->buffer = buffer;
+}
+
+void
+glsl_to_tgsi_visitor::visit_ssbo_intrinsic(ir_call *ir)
+{
+ const char *callee = ir->callee->function_name();
+ exec_node *param = ir->actual_parameters.get_head();
+
+ ir_rvalue *block = ((ir_instruction *)param)->as_rvalue();
+
+ param = param->get_next();
+ ir_rvalue *offset = ((ir_instruction *)param)->as_rvalue();
+
+ ir_constant *const_block = block->as_constant();
+
+ st_src_reg buffer(
+ PROGRAM_BUFFER,
+ ctx->Const.Program[shader->Stage].MaxAtomicBuffers +
+ (const_block ? const_block->value.u[0] : 0),
+ GLSL_TYPE_UINT);
+
+ if (!const_block) {
+ block->accept(this);
+ buffer.reladdr = ralloc(mem_ctx, st_src_reg);
+ *buffer.reladdr = this->result;
+ emit_arl(ir, sampler_reladdr, this->result);
+ }
+
+ /* Calculate the surface offset */
+ offset->accept(this);
+ st_src_reg off = this->result;
+
+ st_dst_reg dst = undef_dst;
+ if (ir->return_deref) {
+ ir->return_deref->accept(this);
+ dst = st_dst_reg(this->result);
+ dst.writemask = (1 << ir->return_deref->type->vector_elements) - 1;
+ }
+
+ glsl_to_tgsi_instruction *inst;
+
+ if (!strcmp("__intrinsic_load_ssbo", callee)) {
+ inst = emit_asm(ir, TGSI_OPCODE_LOAD, dst, off);
+ if (dst.type == GLSL_TYPE_BOOL)
+ emit_asm(ir, TGSI_OPCODE_USNE, dst, st_src_reg(dst), st_src_reg_for_int(0));
+ } else if (!strcmp("__intrinsic_store_ssbo", callee)) {
+ param = param->get_next();
+ ir_rvalue *val = ((ir_instruction *)param)->as_rvalue();
+ val->accept(this);
+
+ param = param->get_next();
+ ir_constant *write_mask = ((ir_instruction *)param)->as_constant();
+ assert(write_mask);
+ dst.writemask = write_mask->value.u[0];
+
+ dst.type = this->result.type;
+ inst = emit_asm(ir, TGSI_OPCODE_STORE, dst, off, this->result);
+ } else {
+ param = param->get_next();
+ ir_rvalue *val = ((ir_instruction *)param)->as_rvalue();
+ val->accept(this);
+
+ st_src_reg data = this->result, data2 = undef_src;
+ unsigned opcode;
+ if (!strcmp("__intrinsic_atomic_add_ssbo", callee))
+ opcode = TGSI_OPCODE_ATOMUADD;
+ else if (!strcmp("__intrinsic_atomic_min_ssbo", callee))
+ opcode = TGSI_OPCODE_ATOMIMIN;
+ else if (!strcmp("__intrinsic_atomic_max_ssbo", callee))
+ opcode = TGSI_OPCODE_ATOMIMAX;
+ else if (!strcmp("__intrinsic_atomic_and_ssbo", callee))
+ opcode = TGSI_OPCODE_ATOMAND;
+ else if (!strcmp("__intrinsic_atomic_or_ssbo", callee))
+ opcode = TGSI_OPCODE_ATOMOR;
+ else if (!strcmp("__intrinsic_atomic_xor_ssbo", callee))
+ opcode = TGSI_OPCODE_ATOMXOR;
+ else if (!strcmp("__intrinsic_atomic_exchange_ssbo", callee))
+ opcode = TGSI_OPCODE_ATOMXCHG;
+ else if (!strcmp("__intrinsic_atomic_comp_swap_ssbo", callee)) {
+ opcode = TGSI_OPCODE_ATOMCAS;
+ param = param->get_next();
+ val = ((ir_instruction *)param)->as_rvalue();
+ val->accept(this);
+ data2 = this->result;
+ } else {
+ assert(!"Unexpected intrinsic");
+ return;
+ }
+
+ inst = emit_asm(ir, opcode, dst, off, data, data2);
+ }
+
+ param = param->get_next();
+ ir_constant *access = NULL;
+ if (!param->is_tail_sentinel()) {
+ access = ((ir_instruction *)param)->as_constant();
+ assert(access);
+ }
+
+ /* The emit_asm() might have actually split the op into pieces, e.g. for
+ * double stores. We have to go back and fix up all the generated ops.
+ */
+ unsigned op = inst->op;
+ do {
+ inst->buffer = buffer;
+ if (access)
+ inst->buffer_access = access->value.u[0];
+ inst = (glsl_to_tgsi_instruction *)inst->get_prev();
+ if (inst->op == TGSI_OPCODE_UADD)
+ inst = (glsl_to_tgsi_instruction *)inst->get_prev();
+ } while (inst && inst->op == op && inst->buffer.file == PROGRAM_UNDEFINED);
+}
+
+void
+glsl_to_tgsi_visitor::visit_membar_intrinsic(ir_call *ir)
+{
+ const char *callee = ir->callee->function_name();
+
+ if (!strcmp("__intrinsic_memory_barrier", callee))
+ emit_asm(ir, TGSI_OPCODE_MEMBAR, undef_dst,
+ st_src_reg_for_int(TGSI_MEMBAR_SHADER_BUFFER |
+ TGSI_MEMBAR_ATOMIC_BUFFER |
+ TGSI_MEMBAR_SHADER_IMAGE |
+ TGSI_MEMBAR_SHARED));
+ else if (!strcmp("__intrinsic_memory_barrier_atomic_counter", callee))
+ emit_asm(ir, TGSI_OPCODE_MEMBAR, undef_dst,
+ st_src_reg_for_int(TGSI_MEMBAR_ATOMIC_BUFFER));
+ else if (!strcmp("__intrinsic_memory_barrier_buffer", callee))
+ emit_asm(ir, TGSI_OPCODE_MEMBAR, undef_dst,
+ st_src_reg_for_int(TGSI_MEMBAR_SHADER_BUFFER));
+ else if (!strcmp("__intrinsic_memory_barrier_image", callee))
+ emit_asm(ir, TGSI_OPCODE_MEMBAR, undef_dst,
+ st_src_reg_for_int(TGSI_MEMBAR_SHADER_IMAGE));
+ else if (!strcmp("__intrinsic_memory_barrier_shared", callee))
+ emit_asm(ir, TGSI_OPCODE_MEMBAR, undef_dst,
+ st_src_reg_for_int(TGSI_MEMBAR_SHARED));
+ else if (!strcmp("__intrinsic_group_memory_barrier", callee))
+ emit_asm(ir, TGSI_OPCODE_MEMBAR, undef_dst,
+ st_src_reg_for_int(TGSI_MEMBAR_SHADER_BUFFER |
+ TGSI_MEMBAR_ATOMIC_BUFFER |
+ TGSI_MEMBAR_SHADER_IMAGE |
+ TGSI_MEMBAR_SHARED |
+ TGSI_MEMBAR_THREAD_GROUP));
+ else
+ assert(!"Unexpected memory barrier intrinsic");
+}
+
+void
+glsl_to_tgsi_visitor::visit_shared_intrinsic(ir_call *ir)
+{
+ const char *callee = ir->callee->function_name();
+ exec_node *param = ir->actual_parameters.get_head();
+
+ ir_rvalue *offset = ((ir_instruction *)param)->as_rvalue();
+
+ st_src_reg buffer(PROGRAM_MEMORY, 0, GLSL_TYPE_UINT);
+
+ /* Calculate the surface offset */
+ offset->accept(this);
+ st_src_reg off = this->result;
+
+ st_dst_reg dst = undef_dst;
+ if (ir->return_deref) {
+ ir->return_deref->accept(this);
+ dst = st_dst_reg(this->result);
+ dst.writemask = (1 << ir->return_deref->type->vector_elements) - 1;
+ }
+
+ glsl_to_tgsi_instruction *inst;
+
+ if (!strcmp("__intrinsic_load_shared", callee)) {
+ inst = emit_asm(ir, TGSI_OPCODE_LOAD, dst, off);
+ inst->buffer = buffer;
+ } else if (!strcmp("__intrinsic_store_shared", callee)) {
+ param = param->get_next();
+ ir_rvalue *val = ((ir_instruction *)param)->as_rvalue();
+ val->accept(this);
+
+ param = param->get_next();
+ ir_constant *write_mask = ((ir_instruction *)param)->as_constant();
+ assert(write_mask);
+ dst.writemask = write_mask->value.u[0];
+
+ dst.type = this->result.type;
+ inst = emit_asm(ir, TGSI_OPCODE_STORE, dst, off, this->result);
+ inst->buffer = buffer;
+ } else {
+ param = param->get_next();
+ ir_rvalue *val = ((ir_instruction *)param)->as_rvalue();
+ val->accept(this);
+
+ st_src_reg data = this->result, data2 = undef_src;
+ unsigned opcode;
+ if (!strcmp("__intrinsic_atomic_add_shared", callee))
+ opcode = TGSI_OPCODE_ATOMUADD;
+ else if (!strcmp("__intrinsic_atomic_min_shared", callee))
+ opcode = TGSI_OPCODE_ATOMIMIN;
+ else if (!strcmp("__intrinsic_atomic_max_shared", callee))
+ opcode = TGSI_OPCODE_ATOMIMAX;
+ else if (!strcmp("__intrinsic_atomic_and_shared", callee))
+ opcode = TGSI_OPCODE_ATOMAND;
+ else if (!strcmp("__intrinsic_atomic_or_shared", callee))
+ opcode = TGSI_OPCODE_ATOMOR;
+ else if (!strcmp("__intrinsic_atomic_xor_shared", callee))
+ opcode = TGSI_OPCODE_ATOMXOR;
+ else if (!strcmp("__intrinsic_atomic_exchange_shared", callee))
+ opcode = TGSI_OPCODE_ATOMXCHG;
+ else if (!strcmp("__intrinsic_atomic_comp_swap_shared", callee)) {
+ opcode = TGSI_OPCODE_ATOMCAS;
+ param = param->get_next();
+ val = ((ir_instruction *)param)->as_rvalue();
+ val->accept(this);
+ data2 = this->result;
+ } else {
+ assert(!"Unexpected intrinsic");
+ return;
+ }
+
+ inst = emit_asm(ir, opcode, dst, off, data, data2);
+ inst->buffer = buffer;
+ }
+}
+
+void
+glsl_to_tgsi_visitor::visit_image_intrinsic(ir_call *ir)
+{
+ const char *callee = ir->callee->function_name();
+ exec_node *param = ir->actual_parameters.get_head();
+
+ ir_dereference *img = (ir_dereference *)param;
+ const ir_variable *imgvar = img->variable_referenced();
+ const glsl_type *type = imgvar->type->without_array();
+ unsigned sampler_array_size = 1, sampler_base = 0;
+
+ st_src_reg reladdr;
+ st_src_reg image(PROGRAM_IMAGE, 0, GLSL_TYPE_UINT);
+
+ get_deref_offsets(img, &sampler_array_size, &sampler_base,
+ (unsigned int *)&image.index, &reladdr);
+ if (reladdr.file != PROGRAM_UNDEFINED) {
+ image.reladdr = ralloc(mem_ctx, st_src_reg);
+ *image.reladdr = reladdr;
+ emit_arl(ir, sampler_reladdr, reladdr);
+ }
+
+ st_dst_reg dst = undef_dst;
+ if (ir->return_deref) {
+ ir->return_deref->accept(this);
+ dst = st_dst_reg(this->result);
+ dst.writemask = (1 << ir->return_deref->type->vector_elements) - 1;
+ }
+
+ glsl_to_tgsi_instruction *inst;
+
+ if (!strcmp("__intrinsic_image_size", callee)) {
+ dst.writemask = WRITEMASK_XYZ;
+ inst = emit_asm(ir, TGSI_OPCODE_RESQ, dst);
+ } else if (!strcmp("__intrinsic_image_samples", callee)) {
+ st_src_reg res = get_temp(glsl_type::ivec4_type);
+ st_dst_reg dstres = st_dst_reg(res);
+ dstres.writemask = WRITEMASK_W;
+ inst = emit_asm(ir, TGSI_OPCODE_RESQ, dstres);
+ res.swizzle = SWIZZLE_WWWW;
+ emit_asm(ir, TGSI_OPCODE_MOV, dst, res);
+ } else {
+ st_src_reg arg1 = undef_src, arg2 = undef_src;
+ st_src_reg coord;
+ st_dst_reg coord_dst;
+ coord = get_temp(glsl_type::ivec4_type);
+ coord_dst = st_dst_reg(coord);
+ coord_dst.writemask = (1 << type->coordinate_components()) - 1;
+ param = param->get_next();
+ ((ir_dereference *)param)->accept(this);
+ emit_asm(ir, TGSI_OPCODE_MOV, coord_dst, this->result);
+ coord.swizzle = SWIZZLE_XXXX;
+ switch (type->coordinate_components()) {
+ case 4: assert(!"unexpected coord count");
+ /* fallthrough */
+ case 3: coord.swizzle |= SWIZZLE_Z << 6;
+ /* fallthrough */
+ case 2: coord.swizzle |= SWIZZLE_Y << 3;
+ }
+
+ if (type->sampler_dimensionality == GLSL_SAMPLER_DIM_MS) {
+ param = param->get_next();
+ ((ir_dereference *)param)->accept(this);
+ st_src_reg sample = this->result;
+ sample.swizzle = SWIZZLE_XXXX;
+ coord_dst.writemask = WRITEMASK_W;
+ emit_asm(ir, TGSI_OPCODE_MOV, coord_dst, sample);
+ coord.swizzle |= SWIZZLE_W << 9;
+ }
+
+ param = param->get_next();
+ if (!param->is_tail_sentinel()) {
+ ((ir_dereference *)param)->accept(this);
+ arg1 = this->result;
+ param = param->get_next();
+ }
+
+ if (!param->is_tail_sentinel()) {
+ ((ir_dereference *)param)->accept(this);
+ arg2 = this->result;
+ param = param->get_next();
+ }
+
+ assert(param->is_tail_sentinel());
+
+ unsigned opcode;
+ if (!strcmp("__intrinsic_image_load", callee))
+ opcode = TGSI_OPCODE_LOAD;
+ else if (!strcmp("__intrinsic_image_store", callee))
+ opcode = TGSI_OPCODE_STORE;
+ else if (!strcmp("__intrinsic_image_atomic_add", callee))
+ opcode = TGSI_OPCODE_ATOMUADD;
+ else if (!strcmp("__intrinsic_image_atomic_min", callee))
+ opcode = TGSI_OPCODE_ATOMIMIN;
+ else if (!strcmp("__intrinsic_image_atomic_max", callee))
+ opcode = TGSI_OPCODE_ATOMIMAX;
+ else if (!strcmp("__intrinsic_image_atomic_and", callee))
+ opcode = TGSI_OPCODE_ATOMAND;
+ else if (!strcmp("__intrinsic_image_atomic_or", callee))
+ opcode = TGSI_OPCODE_ATOMOR;
+ else if (!strcmp("__intrinsic_image_atomic_xor", callee))
+ opcode = TGSI_OPCODE_ATOMXOR;
+ else if (!strcmp("__intrinsic_image_atomic_exchange", callee))
+ opcode = TGSI_OPCODE_ATOMXCHG;
+ else if (!strcmp("__intrinsic_image_atomic_comp_swap", callee))
+ opcode = TGSI_OPCODE_ATOMCAS;
+ else {
+ assert(!"Unexpected intrinsic");
+ return;
+ }
+
+ inst = emit_asm(ir, opcode, dst, coord, arg1, arg2);
+ if (opcode == TGSI_OPCODE_STORE)
+ inst->dst[0].writemask = WRITEMASK_XYZW;
+ }
+
+ inst->buffer = image;
+ inst->sampler_array_size = sampler_array_size;
+ inst->sampler_base = sampler_base;
+
+ switch (type->sampler_dimensionality) {
+ case GLSL_SAMPLER_DIM_1D:
+ inst->tex_target = (type->sampler_array)
+ ? TEXTURE_1D_ARRAY_INDEX : TEXTURE_1D_INDEX;
+ break;
+ case GLSL_SAMPLER_DIM_2D:
+ inst->tex_target = (type->sampler_array)
+ ? TEXTURE_2D_ARRAY_INDEX : TEXTURE_2D_INDEX;
+ break;
+ case GLSL_SAMPLER_DIM_3D:
+ inst->tex_target = TEXTURE_3D_INDEX;
+ break;
+ case GLSL_SAMPLER_DIM_CUBE:
+ inst->tex_target = (type->sampler_array)
+ ? TEXTURE_CUBE_ARRAY_INDEX : TEXTURE_CUBE_INDEX;
+ break;
+ case GLSL_SAMPLER_DIM_RECT:
+ inst->tex_target = TEXTURE_RECT_INDEX;
+ break;
+ case GLSL_SAMPLER_DIM_BUF:
+ inst->tex_target = TEXTURE_BUFFER_INDEX;
+ break;
+ case GLSL_SAMPLER_DIM_EXTERNAL:
+ inst->tex_target = TEXTURE_EXTERNAL_INDEX;
+ break;
+ case GLSL_SAMPLER_DIM_MS:
+ inst->tex_target = (type->sampler_array)
+ ? TEXTURE_2D_MULTISAMPLE_ARRAY_INDEX : TEXTURE_2D_MULTISAMPLE_INDEX;
+ break;
+ default:
+ assert(!"Should not get here.");
+ }
+
+ inst->image_format = st_mesa_format_to_pipe_format(st_context(ctx),
+ _mesa_get_shader_image_format(imgvar->data.image_format));
+
+ if (imgvar->data.image_coherent)
+ inst->buffer_access |= TGSI_MEMORY_COHERENT;
+ if (imgvar->data.image_restrict)
+ inst->buffer_access |= TGSI_MEMORY_RESTRICT;
+ if (imgvar->data.image_volatile)
+ inst->buffer_access |= TGSI_MEMORY_VOLATILE;
+}
+
void
glsl_to_tgsi_visitor::visit(ir_call *ir)
{
glsl_to_tgsi_instruction *call_inst;
ir_function_signature *sig = ir->callee;
- function_entry *entry = get_function_signature(sig);
+ const char *callee = sig->function_name();
+ function_entry *entry;
int i;
+ /* Filter out intrinsics */
+ if (!strcmp("__intrinsic_atomic_read", callee) ||
+ !strcmp("__intrinsic_atomic_increment", callee) ||
+ !strcmp("__intrinsic_atomic_predecrement", callee) ||
+ !strcmp("__intrinsic_atomic_add", callee) ||
+ !strcmp("__intrinsic_atomic_sub", callee) ||
+ !strcmp("__intrinsic_atomic_min", callee) ||
+ !strcmp("__intrinsic_atomic_max", callee) ||
+ !strcmp("__intrinsic_atomic_and", callee) ||
+ !strcmp("__intrinsic_atomic_or", callee) ||
+ !strcmp("__intrinsic_atomic_xor", callee) ||
+ !strcmp("__intrinsic_atomic_exchange", callee) ||
+ !strcmp("__intrinsic_atomic_comp_swap", callee)) {
+ visit_atomic_counter_intrinsic(ir);
+ return;
+ }
+
+ if (!strcmp("__intrinsic_load_ssbo", callee) ||
+ !strcmp("__intrinsic_store_ssbo", callee) ||
+ !strcmp("__intrinsic_atomic_add_ssbo", callee) ||
+ !strcmp("__intrinsic_atomic_min_ssbo", callee) ||
+ !strcmp("__intrinsic_atomic_max_ssbo", callee) ||
+ !strcmp("__intrinsic_atomic_and_ssbo", callee) ||
+ !strcmp("__intrinsic_atomic_or_ssbo", callee) ||
+ !strcmp("__intrinsic_atomic_xor_ssbo", callee) ||
+ !strcmp("__intrinsic_atomic_exchange_ssbo", callee) ||
+ !strcmp("__intrinsic_atomic_comp_swap_ssbo", callee)) {
+ visit_ssbo_intrinsic(ir);
+ return;
+ }
+
+ if (!strcmp("__intrinsic_memory_barrier", callee) ||
+ !strcmp("__intrinsic_memory_barrier_atomic_counter", callee) ||
+ !strcmp("__intrinsic_memory_barrier_buffer", callee) ||
+ !strcmp("__intrinsic_memory_barrier_image", callee) ||
+ !strcmp("__intrinsic_memory_barrier_shared", callee) ||
+ !strcmp("__intrinsic_group_memory_barrier", callee)) {
+ visit_membar_intrinsic(ir);
+ return;
+ }
+
+ if (!strcmp("__intrinsic_load_shared", callee) ||
+ !strcmp("__intrinsic_store_shared", callee) ||
+ !strcmp("__intrinsic_atomic_add_shared", callee) ||
+ !strcmp("__intrinsic_atomic_min_shared", callee) ||
+ !strcmp("__intrinsic_atomic_max_shared", callee) ||
+ !strcmp("__intrinsic_atomic_and_shared", callee) ||
+ !strcmp("__intrinsic_atomic_or_shared", callee) ||
+ !strcmp("__intrinsic_atomic_xor_shared", callee) ||
+ !strcmp("__intrinsic_atomic_exchange_shared", callee) ||
+ !strcmp("__intrinsic_atomic_comp_swap_shared", callee)) {
+ visit_shared_intrinsic(ir);
+ return;
+ }
+
+ if (!strcmp("__intrinsic_image_load", callee) ||
+ !strcmp("__intrinsic_image_store", callee) ||
+ !strcmp("__intrinsic_image_atomic_add", callee) ||
+ !strcmp("__intrinsic_image_atomic_min", callee) ||
+ !strcmp("__intrinsic_image_atomic_max", callee) ||
+ !strcmp("__intrinsic_image_atomic_and", callee) ||
+ !strcmp("__intrinsic_image_atomic_or", callee) ||
+ !strcmp("__intrinsic_image_atomic_xor", callee) ||
+ !strcmp("__intrinsic_image_atomic_exchange", callee) ||
+ !strcmp("__intrinsic_image_atomic_comp_swap", callee) ||
+ !strcmp("__intrinsic_image_size", callee) ||
+ !strcmp("__intrinsic_image_samples", callee)) {
+ visit_image_intrinsic(ir);
+ return;
+ }
+
+ entry = get_function_signature(sig);
/* Process in parameters. */
foreach_two_lists(formal_node, &sig->parameters,
actual_node, &ir->actual_parameters) {
l.index = storage->index;
l.reladdr = NULL;
l.writemask = WRITEMASK_XYZW;
- l.cond_mask = COND_TR;
for (i = 0; i < type_size(param->type); i++) {
emit_asm(ir, TGSI_OPCODE_MOV, l, r);
this->result = entry->return_reg;
}
+void
+glsl_to_tgsi_visitor::calc_deref_offsets(ir_dereference *head,
+ ir_dereference *tail,
+ unsigned *array_elements,
+ unsigned *base,
+ unsigned *index,
+ st_src_reg *indirect,
+ unsigned *location)
+{
+ switch (tail->ir_type) {
+ case ir_type_dereference_record: {
+ ir_dereference_record *deref_record = tail->as_dereference_record();
+ const glsl_type *struct_type = deref_record->record->type;
+ int field_index = deref_record->record->type->field_index(deref_record->field);
+
+ calc_deref_offsets(head, deref_record->record->as_dereference(), array_elements, base, index, indirect, location);
+
+ assert(field_index >= 0);
+ *location += struct_type->record_location_offset(field_index);
+ break;
+ }
+
+ case ir_type_dereference_array: {
+ ir_dereference_array *deref_arr = tail->as_dereference_array();
+ ir_constant *array_index = deref_arr->array_index->constant_expression_value();
+
+ if (!array_index) {
+ st_src_reg temp_reg;
+ st_dst_reg temp_dst;
+
+ temp_reg = get_temp(glsl_type::uint_type);
+ temp_dst = st_dst_reg(temp_reg);
+ temp_dst.writemask = 1;
+
+ deref_arr->array_index->accept(this);
+ if (*array_elements != 1)
+ emit_asm(NULL, TGSI_OPCODE_MUL, temp_dst, this->result, st_src_reg_for_int(*array_elements));
+ else
+ emit_asm(NULL, TGSI_OPCODE_MOV, temp_dst, this->result);
+
+ if (indirect->file == PROGRAM_UNDEFINED)
+ *indirect = temp_reg;
+ else {
+ temp_dst = st_dst_reg(*indirect);
+ temp_dst.writemask = 1;
+ emit_asm(NULL, TGSI_OPCODE_ADD, temp_dst, *indirect, temp_reg);
+ }
+ } else
+ *index += array_index->value.u[0] * *array_elements;
+
+ *array_elements *= deref_arr->array->type->length;
+
+ calc_deref_offsets(head, deref_arr->array->as_dereference(), array_elements, base, index, indirect, location);
+ break;
+ }
+ default:
+ break;
+ }
+}
+
+void
+glsl_to_tgsi_visitor::get_deref_offsets(ir_dereference *ir,
+ unsigned *array_size,
+ unsigned *base,
+ unsigned *index,
+ st_src_reg *reladdr)
+{
+ GLuint shader = _mesa_program_enum_to_shader_stage(this->prog->Target);
+ unsigned location = 0;
+ ir_variable *var = ir->variable_referenced();
+
+ memset(reladdr, 0, sizeof(*reladdr));
+ reladdr->file = PROGRAM_UNDEFINED;
+
+ *base = 0;
+ *array_size = 1;
+
+ assert(var);
+ location = var->data.location;
+ calc_deref_offsets(ir, ir, array_size, base, index, reladdr, &location);
+
+ /*
+ * If we end up with no indirect then adjust the base to the index,
+ * and set the array size to 1.
+ */
+ if (reladdr->file == PROGRAM_UNDEFINED) {
+ *base = *index;
+ *array_size = 1;
+ }
+
+ if (location != 0xffffffff) {
+ *base += this->shader_program->UniformStorage[location].opaque[shader].index;
+ *index += this->shader_program->UniformStorage[location].opaque[shader].index;
+ }
+}
+
void
glsl_to_tgsi_visitor::visit(ir_texture *ir)
{
st_src_reg result_src, coord, cube_sc, lod_info, projector, dx, dy;
st_src_reg offset[MAX_GLSL_TEXTURE_OFFSET], sample_index, component;
- st_src_reg levels_src;
+ st_src_reg levels_src, reladdr;
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;
- ir_rvalue *sampler_index =
- _mesa_get_sampler_array_nonconst_index(ir->sampler);
+ unsigned sampler_array_size = 1, sampler_index = 0, sampler_base = 0;
bool is_cube_array = false;
unsigned i;
ir->coordinate->accept(this);
/* Put our coords in a temp. We'll need to modify them for shadow,
- * projection, or LOD, so the only case we'd use it as is is if
+ * projection, or LOD, so the only case we'd use it as-is is if
* we're doing plain old texturing. The optimization passes on
* glsl_to_tgsi_visitor should handle cleaning up our mess in that case.
*/
coord_dst.writemask = WRITEMASK_XYZW;
}
- if (sampler_index) {
- sampler_index->accept(this);
- emit_arl(ir, sampler_reladdr, this->result);
- }
+ get_deref_offsets(ir->sampler, &sampler_array_size, &sampler_base,
+ &sampler_index, &reladdr);
+ if (reladdr.file != PROGRAM_UNDEFINED)
+ emit_arl(ir, sampler_reladdr, reladdr);
if (opcode == TGSI_OPCODE_TXD)
inst = emit_asm(ir, opcode, result_dst, coord, dx, dy);
if (ir->shadow_comparitor)
inst->tex_shadow = GL_TRUE;
- inst->sampler.index = _mesa_get_sampler_uniform_value(ir->sampler,
- this->shader_program,
- this->prog);
- if (sampler_index) {
+ inst->sampler.index = sampler_index;
+ inst->sampler_array_size = sampler_array_size;
+ inst->sampler_base = sampler_base;
+
+ if (reladdr.file != PROGRAM_UNDEFINED) {
inst->sampler.reladdr = ralloc(mem_ctx, st_src_reg);
- memcpy(inst->sampler.reladdr, &sampler_reladdr, sizeof(sampler_reladdr));
- inst->sampler_array_size =
- ir->sampler->as_dereference_array()->array->type->array_size();
- } else {
- inst->sampler_array_size = 1;
+ memcpy(inst->sampler.reladdr, &reladdr, sizeof(reladdr));
}
if (ir->offset) {
glsl_to_tgsi_visitor::glsl_to_tgsi_visitor()
{
+ STATIC_ASSERT(sizeof(samplers_used) * 8 >= PIPE_MAX_SAMPLERS);
+
result.file = PROGRAM_UNDEFINED;
next_temp = 1;
array_sizes = NULL;
current_function = NULL;
num_address_regs = 0;
samplers_used = 0;
+ buffers_used = 0;
+ images_used = 0;
indirect_addr_consts = false;
wpos_transform_const = -1;
glsl_version = 0;
options = NULL;
have_sqrt = false;
have_fma = false;
+ use_shared_memory = false;
}
glsl_to_tgsi_visitor::~glsl_to_tgsi_visitor()
count_resources(glsl_to_tgsi_visitor *v, gl_program *prog)
{
v->samplers_used = 0;
+ v->buffers_used = 0;
+ v->images_used = 0;
foreach_in_list(glsl_to_tgsi_instruction, inst, &v->instructions) {
if (inst->info->is_tex) {
for (int i = 0; i < inst->sampler_array_size; i++) {
- unsigned idx = inst->sampler.index + i;
- v->samplers_used |= 1 << idx;
+ unsigned idx = inst->sampler_base + i;
+ v->samplers_used |= 1u << idx;
debug_assert(idx < (int)ARRAY_SIZE(v->sampler_types));
v->sampler_types[idx] = inst->tex_type;
}
}
}
+ if (inst->buffer.file != PROGRAM_UNDEFINED && (
+ is_resource_instruction(inst->op) ||
+ inst->op == TGSI_OPCODE_STORE)) {
+ if (inst->buffer.file == PROGRAM_BUFFER) {
+ v->buffers_used |= 1 << inst->buffer.index;
+ } else if (inst->buffer.file == PROGRAM_MEMORY) {
+ v->use_shared_memory = true;
+ } else {
+ assert(inst->buffer.file == PROGRAM_IMAGE);
+ for (int i = 0; i < inst->sampler_array_size; i++) {
+ unsigned idx = inst->sampler_base + i;
+ v->images_used |= 1 << idx;
+ v->image_targets[idx] =
+ st_translate_texture_target(inst->tex_target, false);
+ v->image_formats[idx] = inst->image_format;
+ }
+ }
+ }
}
prog->SamplersUsed = v->samplers_used;
&& inst->dst[0].writemask == get_src_arg_mask(inst->dst[0], inst->src[2])) {
inst->op = TGSI_OPCODE_MOV;
+ inst->info = tgsi_get_opcode_info(inst->op);
inst->src[0] = inst->src[1];
}
}
last_reads[inst->src[j].index] = (depth == 0) ? i : -2;
}
for (j = 0; j < num_inst_dst_regs(inst); j++) {
- if (inst->dst[j].file == PROGRAM_TEMPORARY)
+ if (inst->dst[j].file == PROGRAM_TEMPORARY) {
if (first_writes[inst->dst[j].index] == -1)
first_writes[inst->dst[j].index] = (depth == 0) ? i : loop_start;
+ last_reads[inst->dst[j].index] = (depth == 0) ? i : -2;
+ }
}
for (j = 0; j < inst->tex_offset_num_offset; j++) {
if (inst->tex_offsets[j].file == PROGRAM_TEMPORARY)
foreach_in_list_safe(glsl_to_tgsi_instruction, inst, &this->instructions) {
if (!inst->dead_mask || !inst->dst[0].writemask)
continue;
- else if ((inst->dst[0].writemask & ~inst->dead_mask) == 0) {
+ /* No amount of dead masks should remove memory stores */
+ if (inst->info->is_store)
+ continue;
+
+ if ((inst->dst[0].writemask & ~inst->dead_mask) == 0) {
inst->remove();
delete inst;
removed++;
} else {
- if (inst->dst[0].type == GLSL_TYPE_DOUBLE) {
+ if (glsl_base_type_is_64bit(inst->dst[0].type)) {
if (inst->dead_mask == WRITEMASK_XY ||
inst->dead_mask == WRITEMASK_ZW)
inst->dst[0].writemask &= ~(inst->dead_mask);
/* Update the first_writes and last_reads arrays with the new
* values for the merged register index, and mark the newly unused
* register index as such. */
+ assert(last_reads[j] >= last_reads[i]);
last_reads[i] = last_reads[j];
first_writes[j] = -1;
last_reads[j] = -1;
struct ureg_src inputs[PIPE_MAX_SHADER_INPUTS];
struct ureg_dst address[3];
struct ureg_src samplers[PIPE_MAX_SAMPLERS];
+ struct ureg_src buffers[PIPE_MAX_SHADER_BUFFERS];
+ struct ureg_src images[PIPE_MAX_SHADER_IMAGES];
struct ureg_src systemValues[SYSTEM_VALUE_MAX];
+ struct ureg_src shared_memory;
struct tgsi_texture_offset tex_offsets[MAX_GLSL_TEXTURE_OFFSET];
unsigned *array_sizes;
struct array_decl *input_arrays;
unsigned insn_size;
unsigned insn_count;
- unsigned procType; /**< TGSI_PROCESSOR_VERTEX/FRAGMENT */
+ unsigned procType; /**< PIPE_SHADER_VERTEX/FRAGMENT */
boolean error;
};
/** Map Mesa's SYSTEM_VALUE_x to TGSI_SEMANTIC_x */
-const unsigned _mesa_sysval_to_semantic[SYSTEM_VALUE_MAX] = {
- /* Vertex shader
- */
- TGSI_SEMANTIC_VERTEXID,
- TGSI_SEMANTIC_INSTANCEID,
- TGSI_SEMANTIC_VERTEXID_NOBASE,
- TGSI_SEMANTIC_BASEVERTEX,
-
- /* Geometry shader
- */
- TGSI_SEMANTIC_INVOCATIONID,
-
- /* Fragment shader
- */
- TGSI_SEMANTIC_FACE,
- TGSI_SEMANTIC_SAMPLEID,
- TGSI_SEMANTIC_SAMPLEPOS,
- TGSI_SEMANTIC_SAMPLEMASK,
- TGSI_SEMANTIC_HELPER_INVOCATION,
+unsigned
+_mesa_sysval_to_semantic(unsigned sysval)
+{
+ switch (sysval) {
+ /* Vertex shader */
+ case SYSTEM_VALUE_VERTEX_ID:
+ return TGSI_SEMANTIC_VERTEXID;
+ case SYSTEM_VALUE_INSTANCE_ID:
+ return TGSI_SEMANTIC_INSTANCEID;
+ case SYSTEM_VALUE_VERTEX_ID_ZERO_BASE:
+ return TGSI_SEMANTIC_VERTEXID_NOBASE;
+ case SYSTEM_VALUE_BASE_VERTEX:
+ return TGSI_SEMANTIC_BASEVERTEX;
+ case SYSTEM_VALUE_BASE_INSTANCE:
+ return TGSI_SEMANTIC_BASEINSTANCE;
+ case SYSTEM_VALUE_DRAW_ID:
+ return TGSI_SEMANTIC_DRAWID;
+
+ /* Geometry shader */
+ case SYSTEM_VALUE_INVOCATION_ID:
+ return TGSI_SEMANTIC_INVOCATIONID;
+
+ /* Fragment shader */
+ case SYSTEM_VALUE_FRAG_COORD:
+ return TGSI_SEMANTIC_POSITION;
+ case SYSTEM_VALUE_FRONT_FACE:
+ return TGSI_SEMANTIC_FACE;
+ case SYSTEM_VALUE_SAMPLE_ID:
+ return TGSI_SEMANTIC_SAMPLEID;
+ case SYSTEM_VALUE_SAMPLE_POS:
+ return TGSI_SEMANTIC_SAMPLEPOS;
+ case SYSTEM_VALUE_SAMPLE_MASK_IN:
+ return TGSI_SEMANTIC_SAMPLEMASK;
+ case SYSTEM_VALUE_HELPER_INVOCATION:
+ return TGSI_SEMANTIC_HELPER_INVOCATION;
+
+ /* Tessellation shader */
+ case SYSTEM_VALUE_TESS_COORD:
+ return TGSI_SEMANTIC_TESSCOORD;
+ case SYSTEM_VALUE_VERTICES_IN:
+ return TGSI_SEMANTIC_VERTICESIN;
+ case SYSTEM_VALUE_PRIMITIVE_ID:
+ return TGSI_SEMANTIC_PRIMID;
+ case SYSTEM_VALUE_TESS_LEVEL_OUTER:
+ return TGSI_SEMANTIC_TESSOUTER;
+ case SYSTEM_VALUE_TESS_LEVEL_INNER:
+ return TGSI_SEMANTIC_TESSINNER;
+
+ /* Compute shader */
+ case SYSTEM_VALUE_LOCAL_INVOCATION_ID:
+ return TGSI_SEMANTIC_THREAD_ID;
+ case SYSTEM_VALUE_WORK_GROUP_ID:
+ return TGSI_SEMANTIC_BLOCK_ID;
+ case SYSTEM_VALUE_NUM_WORK_GROUPS:
+ return TGSI_SEMANTIC_GRID_SIZE;
+
+ /* Unhandled */
+ case SYSTEM_VALUE_LOCAL_INVOCATION_INDEX:
+ case SYSTEM_VALUE_GLOBAL_INVOCATION_ID:
+ case SYSTEM_VALUE_VERTEX_CNT:
+ default:
+ assert(!"Unexpected SYSTEM_VALUE_ enum");
+ return TGSI_SEMANTIC_COUNT;
+ }
+}
- /* Tessellation shaders
- */
- TGSI_SEMANTIC_TESSCOORD,
- TGSI_SEMANTIC_VERTICESIN,
- TGSI_SEMANTIC_PRIMID,
- TGSI_SEMANTIC_TESSOUTER,
- TGSI_SEMANTIC_TESSINNER,
-};
/**
* Make note of a branch to a label in the TGSI code.
case PROGRAM_TEMPORARY:
/* Allocate space for temporaries on demand. */
if (index >= t->temps_size) {
- const int inc = 4096;
+ const int inc = align(index - t->temps_size + 1, 4096);
t->temps = (struct ureg_dst*)
realloc(t->temps,
case PROGRAM_OUTPUT:
if (!array_id) {
- if (t->procType == TGSI_PROCESSOR_FRAGMENT)
+ if (t->procType == PIPE_SHADER_FRAGMENT)
assert(index < FRAG_RESULT_MAX);
- else if (t->procType == TGSI_PROCESSOR_TESS_CTRL ||
- t->procType == TGSI_PROCESSOR_TESS_EVAL)
+ else if (t->procType == PIPE_SHADER_TESS_CTRL ||
+ t->procType == PIPE_SHADER_TESS_EVAL)
assert(index < VARYING_SLOT_TESS_MAX);
else
assert(index < VARYING_SLOT_MAX);
if (!reg->array_id) {
assert(t->inputMapping[index] < ARRAY_SIZE(t->inputs));
assert(t->inputs[t->inputMapping[index]].File != TGSI_FILE_NULL);
- return t->inputs[t->inputMapping[index]];
+ return t->inputs[t->inputMapping[index] + double_reg2];
}
else {
struct array_decl *decl = &t->input_arrays[reg->array_id-1];
assert(slot != -1 && t->inputs[slot].File == TGSI_FILE_INPUT);
assert(t->inputs[slot].ArrayID == reg->array_id);
- return ureg_src_array_offset(t->inputs[slot], index - mesa_index);
+ return ureg_src_array_offset(t->inputs[slot], index + double_reg2 - mesa_index);
}
case PROGRAM_ADDRESS:
offset.SwizzleZ = imm_src.SwizzleZ;
offset.Padding = 0;
break;
+ case PROGRAM_INPUT:
+ imm_src = t->inputs[t->inputMapping[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_TEMPORARY:
imm_src = ureg_src(t->temps[in_offset->index]);
offset.File = imm_src.File;
const glsl_to_tgsi_instruction *inst)
{
struct ureg_program *ureg = t->ureg;
- GLuint i;
+ int i;
struct ureg_dst dst[2];
struct ureg_src src[4];
struct tgsi_texture_offset texoffsets[MAX_GLSL_TEXTURE_OFFSET];
- unsigned num_dst;
- unsigned num_src;
- unsigned tex_target;
+ int num_dst;
+ int num_src;
+ unsigned tex_target = 0;
num_dst = num_inst_dst_regs(inst);
num_src = num_inst_src_regs(inst);
src[num_src] =
ureg_src_indirect(src[num_src], ureg_src(t->address[2]));
num_src++;
- for (i = 0; i < inst->tex_offset_num_offset; i++) {
+ for (i = 0; i < (int)inst->tex_offset_num_offset; i++) {
texoffsets[i] = translate_tex_offset(t, &inst->tex_offsets[i], i);
}
tex_target = st_translate_texture_target(inst->tex_target, inst->tex_shadow);
src, num_src);
return;
+ case TGSI_OPCODE_RESQ:
+ case TGSI_OPCODE_LOAD:
+ case TGSI_OPCODE_ATOMUADD:
+ case TGSI_OPCODE_ATOMXCHG:
+ case TGSI_OPCODE_ATOMCAS:
+ case TGSI_OPCODE_ATOMAND:
+ case TGSI_OPCODE_ATOMOR:
+ case TGSI_OPCODE_ATOMXOR:
+ case TGSI_OPCODE_ATOMUMIN:
+ case TGSI_OPCODE_ATOMUMAX:
+ case TGSI_OPCODE_ATOMIMIN:
+ case TGSI_OPCODE_ATOMIMAX:
+ for (i = num_src - 1; i >= 0; i--)
+ src[i + 1] = src[i];
+ num_src++;
+ if (inst->buffer.file == PROGRAM_MEMORY) {
+ src[0] = t->shared_memory;
+ } else if (inst->buffer.file == PROGRAM_BUFFER) {
+ src[0] = t->buffers[inst->buffer.index];
+ } else {
+ src[0] = t->images[inst->buffer.index];
+ tex_target = st_translate_texture_target(inst->tex_target, inst->tex_shadow);
+ }
+ if (inst->buffer.reladdr)
+ src[0] = ureg_src_indirect(src[0], ureg_src(t->address[2]));
+ assert(src[0].File != TGSI_FILE_NULL);
+ ureg_memory_insn(ureg, inst->op, dst, num_dst, src, num_src,
+ inst->buffer_access,
+ tex_target, inst->image_format);
+ break;
+
+ case TGSI_OPCODE_STORE:
+ if (inst->buffer.file == PROGRAM_MEMORY) {
+ dst[0] = ureg_dst(t->shared_memory);
+ } else if (inst->buffer.file == PROGRAM_BUFFER) {
+ dst[0] = ureg_dst(t->buffers[inst->buffer.index]);
+ } else {
+ dst[0] = ureg_dst(t->images[inst->buffer.index]);
+ tex_target = st_translate_texture_target(inst->tex_target, inst->tex_shadow);
+ }
+ dst[0] = ureg_writemask(dst[0], inst->dst[0].writemask);
+ if (inst->buffer.reladdr)
+ dst[0] = ureg_dst_indirect(dst[0], ureg_src(t->address[2]));
+ assert(dst[0].File != TGSI_FILE_NULL);
+ ureg_memory_insn(ureg, inst->op, dst, num_dst, src, num_src,
+ inst->buffer_access,
+ tex_target, inst->image_format);
+ break;
+
case TGSI_OPCODE_SCS:
dst[0] = ureg_writemask(dst[0], TGSI_WRITEMASK_XY);
ureg_insn(ureg, inst->op, dst, num_dst, src, num_src);
* a FBO is bound (STATE_FB_WPOS_Y_TRANSFORM).
*/
static void
-emit_wpos_adjustment( struct st_translate *t,
- int wpos_transform_const,
- boolean invert,
- GLfloat adjX, GLfloat adjY[2])
+emit_wpos_adjustment(struct gl_context *ctx,
+ struct st_translate *t,
+ int wpos_transform_const,
+ boolean invert,
+ GLfloat adjX, GLfloat adjY[2])
{
struct ureg_program *ureg = t->ureg;
*/
struct ureg_src wpostrans = ureg_DECL_constant(ureg, wpos_transform_const);
struct ureg_dst wpos_temp = ureg_DECL_temporary( ureg );
- struct ureg_src wpos_input = t->inputs[t->inputMapping[VARYING_SLOT_POS]];
+ struct ureg_src *wpos =
+ ctx->Const.GLSLFragCoordIsSysVal ?
+ &t->systemValues[SYSTEM_VALUE_FRAG_COORD] :
+ &t->inputs[t->inputMapping[VARYING_SLOT_POS]];
+ struct ureg_src wpos_input = *wpos;
/* First, apply the coordinate shift: */
if (adjX || adjY[0] || adjY[1]) {
/* Use wpos_temp as position input from here on:
*/
- t->inputs[t->inputMapping[VARYING_SLOT_POS]] = ureg_src(wpos_temp);
+ *wpos = ureg_src(wpos_temp);
}
*
* The bias of the y-coordinate depends on whether y-inversion takes place
* (adjY[1]) or not (adjY[0]), which is in turn dependent on whether we are
- * drawing to an FBO (causes additional inversion), and whether the the pipe
+ * drawing to an FBO (causes additional inversion), and whether the pipe
* driver origin and the requested origin differ (the latter condition is
* stored in the 'invert' variable).
*
/* we invert after adjustment so that we avoid the MOV to temporary,
* and reuse the adjustment ADD instead */
- emit_wpos_adjustment(t, wpos_transform_const, invert, adjX, adjY);
+ emit_wpos_adjustment(st->ctx, t, wpos_transform_const, invert, adjX, adjY);
}
/**
return false;
}
+static void
+emit_compute_block_size(const struct gl_program *program,
+ 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]);
+ ureg_property(ureg, TGSI_PROPERTY_CS_FIXED_BLOCK_HEIGHT,
+ cp->LocalSize[1]);
+ ureg_property(ureg, TGSI_PROPERTY_CS_FIXED_BLOCK_DEPTH,
+ cp->LocalSize[2]);
+}
+
/**
* Translate intermediate IR (glsl_to_tgsi_instruction) to TGSI format.
* \param program the program to translate
{
struct st_translate *t;
unsigned i;
+ struct gl_program_constants *frag_const =
+ &ctx->Const.Program[MESA_SHADER_FRAGMENT];
enum pipe_error ret = PIPE_OK;
assert(numInputs <= ARRAY_SIZE(t->inputs));
assert(numOutputs <= ARRAY_SIZE(t->outputs));
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_FRONT_FACE] ==
- TGSI_SEMANTIC_FACE);
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_VERTEX_ID] ==
- TGSI_SEMANTIC_VERTEXID);
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_INSTANCE_ID] ==
- TGSI_SEMANTIC_INSTANCEID);
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_SAMPLE_ID] ==
- TGSI_SEMANTIC_SAMPLEID);
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_SAMPLE_POS] ==
- TGSI_SEMANTIC_SAMPLEPOS);
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_SAMPLE_MASK_IN] ==
- TGSI_SEMANTIC_SAMPLEMASK);
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_INVOCATION_ID] ==
- TGSI_SEMANTIC_INVOCATIONID);
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_VERTEX_ID_ZERO_BASE] ==
- TGSI_SEMANTIC_VERTEXID_NOBASE);
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_BASE_VERTEX] ==
- TGSI_SEMANTIC_BASEVERTEX);
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_TESS_COORD] ==
- TGSI_SEMANTIC_TESSCOORD);
- assert(_mesa_sysval_to_semantic[SYSTEM_VALUE_HELPER_INVOCATION] ==
- TGSI_SEMANTIC_HELPER_INVOCATION);
-
t = CALLOC_STRUCT(st_translate);
if (!t) {
ret = PIPE_ERROR_OUT_OF_MEMORY;
* Declare input attributes.
*/
switch (procType) {
- case TGSI_PROCESSOR_FRAGMENT:
+ case PIPE_SHADER_FRAGMENT:
for (i = 0; i < numInputs; i++) {
unsigned array_id = 0;
unsigned array_size;
}
}
break;
- case TGSI_PROCESSOR_GEOMETRY:
- case TGSI_PROCESSOR_TESS_EVAL:
- case TGSI_PROCESSOR_TESS_CTRL:
+ case PIPE_SHADER_GEOMETRY:
+ case PIPE_SHADER_TESS_EVAL:
+ case PIPE_SHADER_TESS_CTRL:
for (i = 0; i < numInputs; i++) {
unsigned array_id = 0;
unsigned array_size;
}
}
break;
- case TGSI_PROCESSOR_VERTEX:
+ case PIPE_SHADER_VERTEX:
for (i = 0; i < numInputs; i++) {
t->inputs[i] = ureg_DECL_vs_input(ureg, i);
}
break;
+ case PIPE_SHADER_COMPUTE:
+ break;
default:
assert(0);
}
* Declare output attributes.
*/
switch (procType) {
- case TGSI_PROCESSOR_FRAGMENT:
+ case PIPE_SHADER_FRAGMENT:
+ case PIPE_SHADER_COMPUTE:
break;
- case TGSI_PROCESSOR_GEOMETRY:
- case TGSI_PROCESSOR_TESS_EVAL:
- case TGSI_PROCESSOR_TESS_CTRL:
- case TGSI_PROCESSOR_VERTEX:
+ case PIPE_SHADER_GEOMETRY:
+ case PIPE_SHADER_TESS_EVAL:
+ case PIPE_SHADER_TESS_CTRL:
+ case PIPE_SHADER_VERTEX:
for (i = 0; i < numOutputs; i++) {
unsigned array_id = 0;
unsigned array_size;
assert(0);
}
- if (procType == TGSI_PROCESSOR_FRAGMENT) {
+ if (procType == PIPE_SHADER_FRAGMENT) {
+ if (program->shader->info.EarlyFragmentTests)
+ ureg_property(ureg, TGSI_PROPERTY_FS_EARLY_DEPTH_STENCIL, 1);
+
if (proginfo->InputsRead & VARYING_BIT_POS) {
/* Must do this after setting up t->inputs. */
emit_wpos(st_context(ctx), t, proginfo, ureg,
}
}
}
- else if (procType == TGSI_PROCESSOR_VERTEX) {
+ else if (procType == PIPE_SHADER_VERTEX) {
for (i = 0; i < numOutputs; i++) {
if (outputSemanticName[i] == TGSI_SEMANTIC_FOG) {
/* force register to contain a fog coordinate in the form (F, 0, 0, 1). */
}
}
+ if (procType == PIPE_SHADER_COMPUTE) {
+ emit_compute_block_size(proginfo, ureg);
+ }
+
/* Declare address register.
*/
if (program->num_address_regs > 0) {
*/
{
GLbitfield sysInputs = proginfo->SystemValuesRead;
- unsigned numSys = 0;
+
for (i = 0; sysInputs; i++) {
if (sysInputs & (1 << i)) {
- unsigned semName = _mesa_sysval_to_semantic[i];
- t->systemValues[i] = ureg_DECL_system_value(ureg, numSys, semName, 0);
+ unsigned semName = _mesa_sysval_to_semantic(i);
+
+ t->systemValues[i] = ureg_DECL_system_value(ureg, semName, 0);
+
if (semName == TGSI_SEMANTIC_INSTANCEID ||
semName == TGSI_SEMANTIC_VERTEXID) {
/* From Gallium perspective, these system values are always
*/
struct st_context *st = st_context(ctx);
struct pipe_screen *pscreen = st->pipe->screen;
- assert(procType == TGSI_PROCESSOR_VERTEX);
+ assert(procType == PIPE_SHADER_VERTEX);
assert(pscreen->get_shader_param(pscreen, PIPE_SHADER_VERTEX, PIPE_SHADER_CAP_INTEGERS));
(void) pscreen;
if (!ctx->Const.NativeIntegers) {
t->systemValues[i] = ureg_scalar(ureg_src(temp), 0);
}
}
- numSys++;
+
+ if (procType == PIPE_SHADER_FRAGMENT &&
+ semName == TGSI_SEMANTIC_POSITION)
+ emit_wpos(st_context(ctx), t, proginfo, ureg,
+ program->wpos_transform_const);
+
sysInputs &= ~(1 << i);
}
}
assert(i == program->num_immediates);
/* texture samplers */
- for (i = 0; i < ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxTextureImageUnits; i++) {
- if (program->samplers_used & (1 << i)) {
+ for (i = 0; i < frag_const->MaxTextureImageUnits; i++) {
+ if (program->samplers_used & (1u << i)) {
unsigned type;
t->samplers[i] = ureg_DECL_sampler(ureg, i);
}
}
+ for (i = 0; i < frag_const->MaxAtomicBuffers; i++) {
+ if (program->buffers_used & (1 << i)) {
+ t->buffers[i] = ureg_DECL_buffer(ureg, i, true);
+ }
+ }
+
+ for (; i < frag_const->MaxAtomicBuffers + frag_const->MaxShaderStorageBlocks;
+ i++) {
+ if (program->buffers_used & (1 << i)) {
+ t->buffers[i] = ureg_DECL_buffer(ureg, i, false);
+ }
+ }
+
+ if (program->use_shared_memory)
+ t->shared_memory = ureg_DECL_memory(ureg, TGSI_MEMORY_TYPE_SHARED);
+
+ for (i = 0; i < program->shader->NumImages; i++) {
+ if (program->images_used & (1 << i)) {
+ t->images[i] = ureg_DECL_image(ureg, i,
+ program->image_targets[i],
+ program->image_formats[i],
+ true, false);
+ }
+ }
+
/* Emit each instruction in turn:
*/
foreach_in_list(glsl_to_tgsi_instruction, inst, &program->instructions) {
t->insn[t->labels[i].branch_target]);
}
+ /* Set the next shader stage hint for VS and TES. */
+ switch (procType) {
+ case PIPE_SHADER_VERTEX:
+ case PIPE_SHADER_TESS_EVAL:
+ if (program->shader_program->SeparateShader)
+ break;
+
+ for (i = program->shader->Stage+1; i <= MESA_SHADER_FRAGMENT; i++) {
+ if (program->shader_program->_LinkedShaders[i]) {
+ unsigned next;
+
+ switch (i) {
+ case MESA_SHADER_TESS_CTRL:
+ next = PIPE_SHADER_TESS_CTRL;
+ break;
+ case MESA_SHADER_TESS_EVAL:
+ next = PIPE_SHADER_TESS_EVAL;
+ break;
+ case MESA_SHADER_GEOMETRY:
+ next = PIPE_SHADER_GEOMETRY;
+ break;
+ case MESA_SHADER_FRAGMENT:
+ next = PIPE_SHADER_FRAGMENT;
+ break;
+ default:
+ assert(0);
+ continue;
+ }
+
+ ureg_set_next_shader_processor(ureg, next);
+ break;
+ }
+ }
+ break;
+ }
+
out:
if (t) {
free(t->arrays);
* generating Mesa IR.
*/
static struct gl_program *
-get_mesa_program(struct gl_context *ctx,
- struct gl_shader_program *shader_program,
- struct gl_shader *shader)
+get_mesa_program_tgsi(struct gl_context *ctx,
+ struct gl_shader_program *shader_program,
+ struct gl_linked_shader *shader)
{
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[_mesa_shader_enum_to_shader_stage(shader->Type)];
+ &ctx->Const.ShaderCompilerOptions[shader->Stage];
struct pipe_screen *pscreen = ctx->st->pipe->screen;
unsigned ptarget = st_shader_stage_to_ptarget(shader->Stage);
/* Perform optimizations on the instructions in the glsl_to_tgsi_visitor. */
v->simplify_cmp();
- if (shader->Type != GL_TESS_CONTROL_SHADER &&
- shader->Type != GL_TESS_EVALUATION_SHADER)
+ if (shader->Stage != MESA_SHADER_TESS_CTRL &&
+ shader->Stage != MESA_SHADER_TESS_EVAL)
v->copy_propagate();
while (v->eliminate_dead_code());
do_set_program_inouts(shader->ir, prog, shader->Stage);
shrink_array_declarations(v->input_arrays, v->num_input_arrays,
- prog->InputsRead, prog->PatchInputsRead);
+ prog->InputsRead, prog->DoubleInputsRead, prog->PatchInputsRead);
shrink_array_declarations(v->output_arrays, v->num_output_arrays,
- prog->OutputsWritten, prog->PatchOutputsWritten);
+ prog->OutputsWritten, 0ULL, prog->PatchOutputsWritten);
count_resources(v, prog);
+ /* The GLSL IR won't be needed anymore. */
+ ralloc_free(shader->ir);
+ shader->ir = NULL;
+
/* This must be done before the uniform storage is associated. */
- if (shader->Type == GL_FRAGMENT_SHADER &&
- prog->InputsRead & VARYING_BIT_POS){
+ if (shader->Stage == MESA_SHADER_FRAGMENT &&
+ (prog->InputsRead & VARYING_BIT_POS ||
+ prog->SystemValuesRead & (1 << SYSTEM_VALUE_FRAG_COORD))) {
static const gl_state_index wposTransformState[STATE_LENGTH] = {
STATE_INTERNAL, STATE_FB_WPOS_Y_TRANSFORM
};
_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.
+ */
+ _mesa_reserve_parameter_storage(prog->Parameters, 8);
+
/* This has to be done last. Any operation the can cause
* prog->ParameterValues to get reallocated (e.g., anything that adds a
* program constant) has to happen before creating this linkage.
struct st_geometry_program *stgp;
struct st_tessctrl_program *sttcp;
struct st_tesseval_program *sttep;
+ struct st_compute_program *stcp;
- switch (shader->Type) {
- case GL_VERTEX_SHADER:
+ switch (shader->Stage) {
+ case MESA_SHADER_VERTEX:
stvp = (struct st_vertex_program *)prog;
stvp->glsl_to_tgsi = v;
break;
- case GL_FRAGMENT_SHADER:
+ case MESA_SHADER_FRAGMENT:
stfp = (struct st_fragment_program *)prog;
stfp->glsl_to_tgsi = v;
break;
- case GL_GEOMETRY_SHADER:
+ case MESA_SHADER_GEOMETRY:
stgp = (struct st_geometry_program *)prog;
stgp->glsl_to_tgsi = v;
break;
- case GL_TESS_CONTROL_SHADER:
+ case MESA_SHADER_TESS_CTRL:
sttcp = (struct st_tessctrl_program *)prog;
sttcp->glsl_to_tgsi = v;
break;
- case GL_TESS_EVALUATION_SHADER:
+ case MESA_SHADER_TESS_EVAL:
sttep = (struct st_tesseval_program *)prog;
sttep->glsl_to_tgsi = v;
break;
+ case MESA_SHADER_COMPUTE:
+ stcp = (struct st_compute_program *)prog;
+ stcp->glsl_to_tgsi = v;
+ break;
default:
assert(!"should not be reached");
return NULL;
return prog;
}
-extern "C" {
-
-static void
-st_dump_program_for_shader_db(struct gl_context *ctx,
- struct gl_shader_program *prog)
+static struct gl_program *
+get_mesa_program(struct gl_context *ctx,
+ struct gl_shader_program *shader_program,
+ struct gl_linked_shader *shader)
{
- /* Dump only successfully compiled and linked shaders to the specified
- * file. This is for shader-db.
- *
- * These options allow some pre-processing of shaders while dumping,
- * because some apps have ill-formed shaders.
- */
- const char *dump_filename = os_get_option("ST_DUMP_SHADERS");
- const char *insert_directives = os_get_option("ST_DUMP_INSERT");
-
- if (dump_filename && prog->Name != 0) {
- FILE *f = fopen(dump_filename, "a");
-
- if (f) {
- for (unsigned i = 0; i < prog->NumShaders; i++) {
- const struct gl_shader *sh = prog->Shaders[i];
- const char *source;
- bool skip_version = false;
-
- if (!sh)
- continue;
-
- source = sh->Source;
-
- /* This string mustn't be changed. shader-db uses it to find
- * where the shader begins.
- */
- fprintf(f, "GLSL %s shader %d source for linked program %d:\n",
- _mesa_shader_stage_to_string(sh->Stage),
- i, prog->Name);
-
- /* Dump the forced version if set. */
- if (ctx->Const.ForceGLSLVersion) {
- fprintf(f, "#version %i\n", ctx->Const.ForceGLSLVersion);
- skip_version = true;
- }
-
- /* Insert directives (optional). */
- if (insert_directives) {
- if (!ctx->Const.ForceGLSLVersion && prog->Version)
- fprintf(f, "#version %i\n", prog->Version);
- fprintf(f, "%s\n", insert_directives);
- skip_version = true;
- }
-
- if (skip_version && strncmp(source, "#version ", 9) == 0) {
- const char *next_line = strstr(source, "\n");
-
- if (next_line)
- source = next_line + 1;
- else
- continue;
- }
-
- fprintf(f, "%s", source);
- fprintf(f, "\n");
- }
- fclose(f);
+ struct pipe_screen *pscreen = ctx->st->pipe->screen;
+ unsigned ptarget = st_shader_stage_to_ptarget(shader->Stage);
+ enum pipe_shader_ir preferred_ir = (enum pipe_shader_ir)
+ pscreen->get_shader_param(pscreen, ptarget, PIPE_SHADER_CAP_PREFERRED_IR);
+ if (preferred_ir == PIPE_SHADER_IR_NIR) {
+ /* TODO only for GLSL VS/FS for now: */
+ switch (shader->Stage) {
+ case MESA_SHADER_VERTEX:
+ case MESA_SHADER_FRAGMENT:
+ return st_nir_get_mesa_program(ctx, shader_program, shader);
+ default:
+ break;
}
}
+ return get_mesa_program_tgsi(ctx, shader_program, shader);
}
+
+extern "C" {
+
/**
* Link a shader.
* Called via ctx->Driver.LinkShader()
bool progress;
exec_list *ir = prog->_LinkedShaders[i]->ir;
- gl_shader_stage stage = _mesa_shader_enum_to_shader_stage(prog->_LinkedShaders[i]->Type);
+ gl_shader_stage stage = prog->_LinkedShaders[i]->Stage;
const struct gl_shader_compiler_options *options =
&ctx->Const.ShaderCompilerOptions[stage];
unsigned ptarget = st_shader_stage_to_ptarget(stage);
LOWER_PACK_SNORM_4x8 |
LOWER_UNPACK_SNORM_4x8 |
LOWER_UNPACK_UNORM_4x8 |
- LOWER_PACK_UNORM_4x8 |
- LOWER_PACK_HALF_2x16 |
- LOWER_UNPACK_HALF_2x16;
+ LOWER_PACK_UNORM_4x8;
if (ctx->Extensions.ARB_gpu_shader5)
lower_inst |= LOWER_PACK_USE_BFI |
LOWER_PACK_USE_BFE;
+ if (!ctx->st->has_half_float_packing)
+ lower_inst |= LOWER_PACK_HALF_2x16 |
+ LOWER_UNPACK_HALF_2x16;
lower_packing_builtins(ir, lower_inst);
}
(have_dround ? 0 : DOPS_TO_DFRAC) |
(options->EmitNoPow ? POW_TO_EXP2 : 0) |
(!ctx->Const.NativeIntegers ? INT_DIV_TO_MUL_RCP : 0) |
- (options->EmitNoSat ? SAT_TO_CLAMP : 0));
+ (options->EmitNoSat ? SAT_TO_CLAMP : 0) |
+ /* Assume that if ARB_gpu_shader5 is not supported
+ * then all of the extended integer functions need
+ * lowering. It may be necessary to add some caps
+ * for individual instructions.
+ */
+ (!ctx->Extensions.ARB_gpu_shader5
+ ? BIT_COUNT_TO_MATH |
+ EXTRACT_TO_SHIFTS |
+ INSERT_TO_SHIFTS |
+ REVERSE_TO_SHIFTS |
+ FIND_LSB_TO_FLOAT_CAST |
+ FIND_MSB_TO_FLOAT_CAST |
+ IMUL_HIGH_TO_MUL
+ : 0));
do_vec_index_to_cond_assign(ir);
lower_vector_insert(ir, true);
validate_ir_tree(ir);
}
+ build_program_resource_list(ctx, prog);
+
for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
struct gl_program *linked_prog;
_mesa_reference_program(ctx, &linked_prog, NULL);
}
- st_dump_program_for_shader_db(ctx, prog);
return GL_TRUE;
}
const GLuint outputMapping[],
struct pipe_stream_output_info *so)
{
- unsigned i;
struct gl_transform_feedback_info *info =
&glsl_to_tgsi->shader_program->LinkedTransformFeedback;
+ st_translate_stream_output_info2(info, outputMapping, so);
+}
+
+void
+st_translate_stream_output_info2(struct gl_transform_feedback_info *info,
+ const GLuint outputMapping[],
+ struct pipe_stream_output_info *so)
+{
+ unsigned i;
for (i = 0; i < info->NumOutputs; i++) {
so->output[i].register_index =
}
for (i = 0; i < PIPE_MAX_SO_BUFFERS; i++) {
- so->stride[i] = info->BufferStride[i];
+ so->stride[i] = info->Buffers[i].Stride;
}
so->num_outputs = info->NumOutputs;
}
projects / mesa.git / blobdiff