From c4285be9a5bd1adaa89050989374b95a9a601cdc Mon Sep 17 00:00:00 2001
From: Ian Romanick <ian.d.romanick@intel.com>
Date: Wed, 24 Nov 2010 22:21:10 -0800
Subject: glsl: Lower ir_binop_pow to a sequence of EXP2 and LOG2

---
 src/glsl/ir_optimization.h      |  5 +++--
 src/glsl/lower_instructions.cpp | 26 ++++++++++++++++++++++++++
 2 files changed, 29 insertions(+), 2 deletions(-)

(limited to 'src/glsl')

diff --git a/src/glsl/ir_optimization.h b/src/glsl/ir_optimization.h
index 1048ff982e..f264265f4b 100644
--- a/src/glsl/ir_optimization.h
+++ b/src/glsl/ir_optimization.h
@@ -32,8 +32,9 @@
 #define SUB_TO_ADD_NEG 0x01
 #define DIV_TO_MUL_RCP 0x02
 #define EXP_TO_EXP2    0x04
-#define LOG_TO_LOG2    0x08
-#define MOD_TO_FRACT   0x10
+#define POW_TO_EXP2    0x08
+#define LOG_TO_LOG2    0x10
+#define MOD_TO_FRACT   0x20
 
 bool do_common_optimization(exec_list *ir, bool linked, unsigned max_unroll_iterations);
 
diff --git a/src/glsl/lower_instructions.cpp b/src/glsl/lower_instructions.cpp
index 0d9374d73b..a5f61f213d 100644
--- a/src/glsl/lower_instructions.cpp
+++ b/src/glsl/lower_instructions.cpp
@@ -33,6 +33,7 @@
  * - SUB_TO_ADD_NEG
  * - DIV_TO_MUL_RCP
  * - EXP_TO_EXP2
+ * - POW_TO_EXP2
  * - LOG_TO_LOG2
  * - MOD_TO_FRACT
  *
@@ -61,6 +62,11 @@
  * do have base 2 versions, so this pass converts exp and log to exp2
  * and log2 operations.
  *
+ * POW_TO_EXP2:
+ * -----------
+ * Many older GPUs don't have an x**y instruction.  For these GPUs, convert
+ * x**y to 2**(y * log2(x)).
+ *
  * MOD_TO_FRACT:
  * -------------
  * Breaks an ir_unop_mod expression down to (op1 * fract(op0 / op1))
@@ -91,6 +97,7 @@ private:
    void div_to_mul_rcp(ir_expression *);
    void mod_to_fract(ir_expression *);
    void exp_to_exp2(ir_expression *);
+   void pow_to_exp2(ir_expression *);
    void log_to_log2(ir_expression *);
 };
 
@@ -180,6 +187,20 @@ lower_instructions_visitor::exp_to_exp2(ir_expression *ir)
    this->progress = true;
 }
 
+void
+lower_instructions_visitor::pow_to_exp2(ir_expression *ir)
+{
+   ir_expression *const log2_x =
+      new(ir) ir_expression(ir_unop_log2, ir->operands[0]->type,
+			    ir->operands[0]);
+
+   ir->operation = ir_unop_exp2;
+   ir->operands[0] = new(ir) ir_expression(ir_binop_mul, ir->operands[1]->type,
+					   ir->operands[1], log2_x);
+   ir->operands[1] = NULL;
+   this->progress = true;
+}
+
 void
 lower_instructions_visitor::log_to_log2(ir_expression *ir)
 {
@@ -254,6 +275,11 @@ lower_instructions_visitor::visit_leave(ir_expression *ir)
 	 mod_to_fract(ir);
       break;
 
+   case ir_binop_pow:
+      if (lowering(POW_TO_EXP2))
+	 pow_to_exp2(ir);
+      break;
+
    default:
       return visit_continue;
    }
-- 
cgit v1.2.3