diff options
Diffstat (limited to 'mesalib/src/mesa/program')
| -rw-r--r-- | mesalib/src/mesa/program/ir_to_mesa.cpp | 136 | ||||
| -rw-r--r-- | mesalib/src/mesa/program/prog_execute.c | 4 | ||||
| -rw-r--r-- | mesalib/src/mesa/program/prog_execute.h | 1 | ||||
| -rw-r--r-- | mesalib/src/mesa/program/prog_print.c | 5 |
4 files changed, 146 insertions, 0 deletions
diff --git a/mesalib/src/mesa/program/ir_to_mesa.cpp b/mesalib/src/mesa/program/ir_to_mesa.cpp index 02ca83d47..ae2742afd 100644 --- a/mesalib/src/mesa/program/ir_to_mesa.cpp +++ b/mesalib/src/mesa/program/ir_to_mesa.cpp @@ -295,6 +295,8 @@ public: bool process_move_condition(ir_rvalue *ir);
+ void copy_propagate(void);
+
void *mem_ctx;
};
@@ -1460,6 +1462,7 @@ ir_to_mesa_visitor::visit(ir_dereference_variable *ir) case ir_var_in:
case ir_var_out:
case ir_var_inout:
+ case ir_var_system_value:
/* The linker assigns locations for varyings and attributes,
* including deprecated builtins (like gl_Color), user-assign
* generic attributes (glBindVertexLocation), and
@@ -1482,6 +1485,10 @@ ir_to_mesa_visitor::visit(ir_dereference_variable *ir) ir->var->type->gl_type,
ir->var->location - VERT_ATTRIB_GENERIC0);
}
+ } else if (ir->var->mode == ir_var_system_value) {
+ entry = new(mem_ctx) variable_storage(ir->var,
+ PROGRAM_SYSTEM_VALUE,
+ ir->var->location);
} else {
entry = new(mem_ctx) variable_storage(ir->var,
PROGRAM_OUTPUT,
@@ -2616,6 +2623,133 @@ set_uniform_initializers(struct gl_context *ctx, talloc_free(mem_ctx);
}
+/*
+ * On a basic block basis, tracks available PROGRAM_TEMPORARY register
+ * channels for copy propagation and updates following instructions to
+ * use the original versions.
+ *
+ * The ir_to_mesa_visitor lazily produces code assuming that this pass
+ * will occur. As an example, a TXP production before this pass:
+ *
+ * 0: MOV TEMP[1], INPUT[4].xyyy;
+ * 1: MOV TEMP[1].w, INPUT[4].wwww;
+ * 2: TXP TEMP[2], TEMP[1], texture[0], 2D;
+ *
+ * and after:
+ *
+ * 0: MOV TEMP[1], INPUT[4].xyyy;
+ * 1: MOV TEMP[1].w, INPUT[4].wwww;
+ * 2: TXP TEMP[2], INPUT[4].xyyw, texture[0], 2D;
+ *
+ * which allows for dead code elimination on TEMP[1]'s writes.
+ */
+void
+ir_to_mesa_visitor::copy_propagate(void)
+{
+ ir_to_mesa_instruction **acp = talloc_zero_array(mem_ctx,
+ ir_to_mesa_instruction *,
+ this->next_temp * 4);
+
+ foreach_iter(exec_list_iterator, iter, this->instructions) {
+ ir_to_mesa_instruction *inst = (ir_to_mesa_instruction *)iter.get();
+
+ /* First, do any copy propagation possible into the src regs. */
+ for (int r = 0; r < 3; r++) {
+ ir_to_mesa_instruction *first = NULL;
+ bool good = true;
+ int acp_base = inst->src_reg[r].index * 4;
+
+ if (inst->src_reg[r].file != PROGRAM_TEMPORARY ||
+ inst->src_reg[r].reladdr)
+ continue;
+
+ /* See if we can find entries in the ACP consisting of MOVs
+ * from the same src register for all the swizzled channels
+ * of this src register reference.
+ */
+ for (int i = 0; i < 4; i++) {
+ int src_chan = GET_SWZ(inst->src_reg[r].swizzle, i);
+ ir_to_mesa_instruction *copy_chan = acp[acp_base + src_chan];
+
+ if (!copy_chan) {
+ good = false;
+ break;
+ }
+
+ if (!first) {
+ first = copy_chan;
+ } else {
+ if (first->src_reg[0].file != copy_chan->src_reg[0].file ||
+ first->src_reg[0].index != copy_chan->src_reg[0].index) {
+ good = false;
+ break;
+ }
+ }
+ }
+
+ if (good) {
+ /* We've now validated that we can copy-propagate to
+ * replace this src register reference. Do it.
+ */
+ inst->src_reg[r].file = first->src_reg[0].file;
+ inst->src_reg[r].index = first->src_reg[0].index;
+
+ int swizzle = 0;
+ for (int i = 0; i < 4; i++) {
+ int src_chan = GET_SWZ(inst->src_reg[r].swizzle, i);
+ ir_to_mesa_instruction *copy_inst = acp[acp_base + src_chan];
+ swizzle |= (GET_SWZ(copy_inst->src_reg[0].swizzle, src_chan) <<
+ (3 * i));
+ }
+ inst->src_reg[r].swizzle = swizzle;
+ }
+ }
+
+ switch (inst->op) {
+ case OPCODE_BGNLOOP:
+ case OPCODE_ENDLOOP:
+ case OPCODE_ELSE:
+ case OPCODE_ENDIF:
+ /* End of a basic block, clear the ACP entirely. */
+ memset(acp, 0, sizeof(*acp) * this->next_temp * 4);
+ break;
+
+ default:
+ /* Continuing the block, clear any written channels from
+ * the ACP.
+ */
+ if (inst->dst_reg.file == PROGRAM_TEMPORARY) {
+ if (inst->dst_reg.reladdr) {
+ memset(acp, 0, sizeof(*acp) * this->next_temp * 4);
+ } else {
+ for (int i = 0; i < 4; i++) {
+ if (inst->dst_reg.writemask & (1 << i)) {
+ acp[4 * inst->dst_reg.index + i] = NULL;
+ }
+ }
+ }
+ }
+ break;
+ }
+
+ /* If this is a copy, add it to the ACP. */
+ if (inst->op == OPCODE_MOV &&
+ inst->dst_reg.file == PROGRAM_TEMPORARY &&
+ !inst->dst_reg.reladdr &&
+ !inst->saturate &&
+ !inst->src_reg[0].reladdr &&
+ !inst->src_reg[0].negate) {
+ for (int i = 0; i < 4; i++) {
+ if (inst->dst_reg.writemask & (1 << i)) {
+ acp[4 * inst->dst_reg.index + i] = inst;
+ }
+ }
+ }
+ }
+
+ talloc_free(acp);
+}
+
/**
* Convert a shader's GLSL IR into a Mesa gl_program.
@@ -2715,6 +2849,8 @@ get_mesa_program(struct gl_context *ctx, mesa_instruction_annotation = talloc_array(v.mem_ctx, ir_instruction *,
num_instructions);
+ v.copy_propagate();
+
/* Convert ir_mesa_instructions into prog_instructions.
*/
mesa_inst = mesa_instructions;
diff --git a/mesalib/src/mesa/program/prog_execute.c b/mesalib/src/mesa/program/prog_execute.c index 4faed2663..28a3e1e20 100644 --- a/mesalib/src/mesa/program/prog_execute.c +++ b/mesalib/src/mesa/program/prog_execute.c @@ -159,6 +159,10 @@ get_src_register_pointer(const struct prog_src_register *source, return ZeroVec;
return prog->Parameters->ParameterValues[reg];
+ case PROGRAM_SYSTEM_VALUE:
+ assert(reg < Elements(machine->SystemValues));
+ return machine->SystemValues[reg];
+
default:
_mesa_problem(NULL,
"Invalid src register file %d in get_src_register_pointer()",
diff --git a/mesalib/src/mesa/program/prog_execute.h b/mesalib/src/mesa/program/prog_execute.h index ac71c0da3..6365b0741 100644 --- a/mesalib/src/mesa/program/prog_execute.h +++ b/mesalib/src/mesa/program/prog_execute.h @@ -61,6 +61,7 @@ struct gl_program_machine GLfloat (*EnvParams)[4]; /**< Vertex or Fragment env parameters */
GLuint CondCodes[4]; /**< COND_* value for x/y/z/w */
GLint AddressReg[MAX_PROGRAM_ADDRESS_REGS][4];
+ GLfloat SystemValues[SYSTEM_VALUE_MAX][4];
const GLubyte *Samplers; /** Array mapping sampler var to tex unit */
diff --git a/mesalib/src/mesa/program/prog_print.c b/mesalib/src/mesa/program/prog_print.c index de0ca1c4f..c1383f882 100644 --- a/mesalib/src/mesa/program/prog_print.c +++ b/mesalib/src/mesa/program/prog_print.c @@ -72,6 +72,8 @@ _mesa_register_file_name(gl_register_file f) return "ADDR";
case PROGRAM_SAMPLER:
return "SAMPLER";
+ case PROGRAM_SYSTEM_VALUE:
+ return "SYSVAL";
case PROGRAM_UNDEFINED:
return "UNDEFINED";
default:
@@ -310,6 +312,9 @@ reg_string(gl_register_file f, GLint index, gl_prog_print_mode mode, case PROGRAM_UNIFORM: /* extension */
sprintf(str, "uniform[%s%d]", addr, index);
break;
+ case PROGRAM_SYSTEM_VALUE:
+ sprintf(str, "sysvalue[%s%d]", addr, index);
+ break;
case PROGRAM_STATE_VAR:
{
struct gl_program_parameter *param
|