Mesa (master): i965: Allow FS constants to be used as immediates instead of push/pull.

[Eric Anholt]

Module: Mesa
Branch: master
Commit: 4fc57322258a750c0a9cabc77372b5ccde1fa877
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=4fc57322258a750c0a9cabc77372b5ccde1fa877

Author: Eric Anholt <eric at anholt.net>
Date:   Fri Mar 19 12:34:53 2010 -0700

i965: Allow FS constants to be used as immediates instead of push/pull.

The hope is to later take advantage of the reduced constant usage to
free up regs.  This only covers the GLSL path at the moment, because
the brw_wm_emit path doesn't get the information as to whether a float
value is a constant or a uniform.

---

 src/mesa/drivers/dri/i965/brw_wm.h      |    2 +-
 src/mesa/drivers/dri/i965/brw_wm_emit.c |   38 +++++++++++++++++++++++++++++++
 src/mesa/drivers/dri/i965/brw_wm_glsl.c |   25 +++++++++++++++-----
 3 files changed, 58 insertions(+), 7 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h
index 47b764d..5aade1c 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.h
+++ b/src/mesa/drivers/dri/i965/brw_wm.h
@@ -286,7 +286,7 @@ void brw_wm_pass0( struct brw_wm_compile *c );
 void brw_wm_pass1( struct brw_wm_compile *c );
 void brw_wm_pass2( struct brw_wm_compile *c );
 void brw_wm_emit( struct brw_wm_compile *c );
-
+GLboolean brw_wm_arg_can_be_immediate(enum prog_opcode, int arg);
 void brw_wm_print_value( struct brw_wm_compile *c,
 			 struct brw_wm_value *value );
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c
index 5abc067..f79de5d 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c
@@ -61,6 +61,44 @@ static INLINE struct brw_reg sechalf( struct brw_reg reg )
    return reg;
 }
 
+/* Return the SrcReg index of the channels that can be immediate float operands
+ * instead of usage of PROGRAM_CONSTANT values through push/pull.
+ */
+GLboolean
+brw_wm_arg_can_be_immediate(enum prog_opcode opcode, int arg)
+{
+   int opcode_array[] = {
+      [OPCODE_ADD] = 2,
+      [OPCODE_CMP] = 3,
+      [OPCODE_DP3] = 2,
+      [OPCODE_DP4] = 2,
+      [OPCODE_DPH] = 2,
+      [OPCODE_MAX] = 2,
+      [OPCODE_MIN] = 2,
+      [OPCODE_MOV] = 1,
+      [OPCODE_MUL] = 2,
+      [OPCODE_SEQ] = 2,
+      [OPCODE_SGE] = 2,
+      [OPCODE_SGT] = 2,
+      [OPCODE_SLE] = 2,
+      [OPCODE_SLT] = 2,
+      [OPCODE_SNE] = 2,
+      [OPCODE_XPD] = 2,
+   };
+
+   /* These opcodes get broken down in a way that allow two
+    * args to be immediates.
+    */
+   if (opcode == OPCODE_MAD || opcode == OPCODE_LRP) {
+      if (arg == 1 || arg == 2)
+	 return GL_TRUE;
+   }
+
+   if (opcode > ARRAY_SIZE(opcode_array))
+      return GL_FALSE;
+
+   return arg == opcode_array[opcode] - 1;
+}
 
 /**
  * Computes the screen-space x,y position of the pixels.
diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
index d9d6dda..d78fb4e 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
@@ -570,12 +570,25 @@ static struct brw_reg get_src_reg(struct brw_wm_compile *c,
     const GLuint nr = 1;
     const GLuint component = GET_SWZ(src->Swizzle, channel);
 
-    /* Extended swizzle terms */
-    if (component == SWIZZLE_ZERO) {
-       return brw_imm_f(0.0F);
-    }
-    else if (component == SWIZZLE_ONE) {
-       return brw_imm_f(1.0F);
+    /* Only one immediate value can be used per native opcode, and it
+     * has be in the src1 slot, so not all Mesa instructions will get
+     * to take advantage of immediate constants.
+     */
+    if (brw_wm_arg_can_be_immediate(inst->Opcode, srcRegIndex)) {
+       const struct gl_program_parameter_list *params;
+
+       params = c->fp->program.Base.Parameters;
+
+       /* Extended swizzle terms */
+       if (component == SWIZZLE_ZERO) {
+	  return brw_imm_f(0.0F);
+       } else if (component == SWIZZLE_ONE) {
+	  return brw_imm_f(1.0F);
+       }
+
+       if (src->File == PROGRAM_CONSTANT) {
+	  return brw_imm_f(params->ParameterValues[src->Index][component]);
+       }
     }
 
     if (c->fp->use_const_buffer &&