i965/vs: Add support for LRP instruction.

Only 13 affected programs in shader-db, but they were all helped.

total instructions in shared programs: 368877 -> 368851 (-0.01%)
instructions in affected programs:     1576 -> 1550 (-1.65%)

Reviewed-by: Chris Forbes <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>

5 files changed, 22 insertions, 3 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp
index b3bd1b97667..5addff67318 100644
--- a/src/mesa/drivers/dri/i965/brw_shader.cpp
+++ b/src/mesa/drivers/dri/i965/brw_shader.cpp
@@ -152,8 +152,7 @@ brw_link_shader(struct gl_context *ctx, struct gl_shader_program *shProg)
        */
       brw_lower_packing_builtins(brw, (gl_shader_type) stage, shader->ir);
       do_mat_op_to_vec(shader->ir);
-      const int lrp_to_arith = (intel->gen < 6 || stage != MESA_SHADER_FRAGMENT)
-                                ? LRP_TO_ARITH : 0;
+      const int lrp_to_arith = intel->gen < 6 ? LRP_TO_ARITH : 0;
       lower_instructions(shader->ir,
 			 MOD_TO_FRACT |
 			 DIV_TO_MUL_RCP |
diff --git a/src/mesa/drivers/dri/i965/brw_vec4.h b/src/mesa/drivers/dri/i965/brw_vec4.h
index c28092244a4..d34ed35ebc6 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4.h
+++ b/src/mesa/drivers/dri/i965/brw_vec4.h
@@ -386,6 +386,7 @@ public:
    vec4_instruction *PULL_CONSTANT_LOAD(dst_reg dst, src_reg index);
    vec4_instruction *SCRATCH_READ(dst_reg dst, src_reg index);
    vec4_instruction *SCRATCH_WRITE(dst_reg dst, src_reg src, src_reg index);
+   vec4_instruction *LRP(dst_reg dst, src_reg a, src_reg y, src_reg x);
 
    int implied_mrf_writes(vec4_instruction *inst);
 
diff --git a/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp b/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp
index 51ee4750490..f2c6cd60911 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp
@@ -215,6 +215,9 @@ vec4_visitor::try_copy_propagation(struct intel_context *intel,
    if (has_source_modifiers && !can_do_source_mods(inst))
       return false;
 
+   if (inst->opcode == BRW_OPCODE_LRP && value.file == UNIFORM)
+      return false;
+
    /* We can't copy-propagate a UD negation into a condmod
     * instruction, because the condmod ends up looking at the 33-bit
     * signed accumulator value instead of the 32-bit value we wanted
diff --git a/src/mesa/drivers/dri/i965/brw_vec4_emit.cpp b/src/mesa/drivers/dri/i965/brw_vec4_emit.cpp
index c9963bff80a..96b4965045b 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_emit.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_emit.cpp
@@ -838,6 +838,10 @@ vec4_generator::generate_code(exec_list *instructions)
          brw_F16TO32(p, dst, src[0]);
          break;
 
+      case BRW_OPCODE_LRP:
+         brw_LRP(p, dst, src[0], src[1], src[2]);
+         break;
+
       case BRW_OPCODE_IF:
 	 if (inst->src[0].file != BAD_FILE) {
 	    /* The instruction has an embedded compare (only allowed on gen6) */
diff --git a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
index 69e805d1e1a..88c435ca292 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
@@ -107,6 +107,14 @@ vec4_visitor::emit(enum opcode opcode)
 					   src0, src1);			\
    }
 
+#define ALU3(op)							\
+   vec4_instruction *							\
+   vec4_visitor::op(dst_reg dst, src_reg src0, src_reg src1, src_reg src2)\
+   {									\
+      return new(mem_ctx) vec4_instruction(this, BRW_OPCODE_##op, dst,	\
+					   src0, src1, src2);		\
+   }
+
 ALU1(NOT)
 ALU1(MOV)
 ALU1(FRC)
@@ -127,6 +135,7 @@ ALU2(DPH)
 ALU2(SHL)
 ALU2(SHR)
 ALU2(ASR)
+ALU3(LRP)
 
 /** Gen4 predicated IF. */
 vec4_instruction *
@@ -1619,7 +1628,10 @@ vec4_visitor::visit(ir_expression *ir)
    }
 
    case ir_triop_lrp:
-      assert(!"not reached: should be handled by lrp_to_arith");
+      op[0] = fix_3src_operand(op[0]);
+      op[1] = fix_3src_operand(op[1]);
+      op[2] = fix_3src_operand(op[2]);
+      emit(LRP(result_dst, op[0], op[1], op[2]));
       break;
 
    case ir_quadop_vector: