From da319095f2ca8869657ebda0db54eb9b2f7393ce Mon Sep 17 00:00:00 2001
From: Keith Whitwell <keithw@vmware.com>
Date: Tue, 11 Aug 2009 18:06:16 +0100
Subject: [PATCH] softpipe: reduce textual differences between exec and sse
 shader paths

Unshare one function (setup_pos_vector) as we want to push this code
into the generated shader in the SSE case.
---
 src/gallium/drivers/softpipe/sp_fs_exec.c | 51 ++++++++++++-----------
 src/gallium/drivers/softpipe/sp_fs_sse.c  | 50 ++++++++++++++++++----
 2 files changed, 69 insertions(+), 32 deletions(-)

diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c
index 91e04687c5b..c469ac63407 100644
--- a/src/gallium/drivers/softpipe/sp_fs_exec.c
+++ b/src/gallium/drivers/softpipe/sp_fs_exec.c
@@ -59,15 +59,34 @@ sp_exec_fragment_shader(const struct sp_fragment_shader *base)
 }
 
 
+static void
+exec_prepare( const struct sp_fragment_shader *base,
+	      struct tgsi_exec_machine *machine,
+	      struct tgsi_sampler **samplers )
+{
+   /*
+    * Bind tokens/shader to the interpreter's machine state.
+    * Avoid redundant binding.
+    */
+   if (machine->Tokens != base->shader.tokens) {
+      tgsi_exec_machine_bind_shader( machine,
+                                     base->shader.tokens,
+                                     PIPE_MAX_SAMPLERS,
+                                     samplers );
+   }
+}
+
+
+
 /**
  * Compute quad X,Y,Z,W for the four fragments in a quad.
  *
  * This should really be part of the compiled shader.
  */
-void
-sp_setup_pos_vector(const struct tgsi_interp_coef *coef,
-		    float x, float y,
-		    struct tgsi_exec_vector *quadpos)
+static void
+setup_pos_vector(const struct tgsi_interp_coef *coef,
+                 float x, float y,
+                 struct tgsi_exec_vector *quadpos)
 {
    uint chan;
    /* do X */
@@ -95,24 +114,6 @@ sp_setup_pos_vector(const struct tgsi_interp_coef *coef,
 }
 
 
-static void
-exec_prepare( const struct sp_fragment_shader *base,
-	      struct tgsi_exec_machine *machine,
-	      struct tgsi_sampler **samplers )
-{
-   /*
-    * Bind tokens/shader to the interpreter's machine state.
-    * Avoid redundant binding.
-    */
-   if (machine->Tokens != base->shader.tokens) {
-      tgsi_exec_machine_bind_shader( machine,
-                                     base->shader.tokens,
-                                     PIPE_MAX_SAMPLERS,
-                                     samplers );
-   }
-}
-
-
 /* TODO: hide the machine struct in here somewhere, remove from this
  * interface:
  */
@@ -122,9 +123,9 @@ exec_run( const struct sp_fragment_shader *base,
 	  struct quad_header *quad )
 {
    /* Compute X, Y, Z, W vals for this quad */
-   sp_setup_pos_vector(quad->posCoef, 
-		       (float)quad->input.x0, (float)quad->input.y0, 
-		       &machine->QuadPos);
+   setup_pos_vector(quad->posCoef, 
+                    (float)quad->input.x0, (float)quad->input.y0, 
+                    &machine->QuadPos);
    
    quad->inout.mask &= tgsi_exec_machine_run( machine );
    if (quad->inout.mask == 0)
diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c
index 364bb94a5f6..9d3e4670eef 100644
--- a/src/gallium/drivers/softpipe/sp_fs_sse.c
+++ b/src/gallium/drivers/softpipe/sp_fs_sse.c
@@ -76,6 +76,43 @@ fs_sse_prepare( const struct sp_fragment_shader *base,
 }
 
 
+
+/**
+ * Compute quad X,Y,Z,W for the four fragments in a quad.
+ *
+ * This should really be part of the compiled shader.
+ */
+static void
+setup_pos_vector(const struct tgsi_interp_coef *coef,
+		    float x, float y,
+		    struct tgsi_exec_vector *quadpos)
+{
+   uint chan;
+   /* do X */
+   quadpos->xyzw[0].f[0] = x;
+   quadpos->xyzw[0].f[1] = x + 1;
+   quadpos->xyzw[0].f[2] = x;
+   quadpos->xyzw[0].f[3] = x + 1;
+
+   /* do Y */
+   quadpos->xyzw[1].f[0] = y;
+   quadpos->xyzw[1].f[1] = y;
+   quadpos->xyzw[1].f[2] = y + 1;
+   quadpos->xyzw[1].f[3] = y + 1;
+
+   /* do Z and W for all fragments in the quad */
+   for (chan = 2; chan < 4; chan++) {
+      const float dadx = coef->dadx[chan];
+      const float dady = coef->dady[chan];
+      const float a0 = coef->a0[chan] + dadx * x + dady * y;
+      quadpos->xyzw[chan].f[0] = a0;
+      quadpos->xyzw[chan].f[1] = a0 + dadx;
+      quadpos->xyzw[chan].f[2] = a0 + dady;
+      quadpos->xyzw[chan].f[3] = a0 + dadx + dady;
+   }
+}
+
+
 /* TODO: codegenerate the whole run function, skip this wrapper.
  * TODO: break dependency on tgsi_exec_machine struct
  * TODO: push Position calculation into the generated shader
@@ -89,9 +126,9 @@ fs_sse_run( const struct sp_fragment_shader *base,
    struct sp_sse_fragment_shader *shader = sp_sse_fragment_shader(base);
 
    /* Compute X, Y, Z, W vals for this quad -- place in temp[0] for now */
-   sp_setup_pos_vector(quad->posCoef, 
-		       (float)quad->input.x0, (float)quad->input.y0, 
-		       machine->Temps);
+   setup_pos_vector(quad->posCoef, 
+                    (float)quad->input.x0, (float)quad->input.y0, 
+                    machine->Temps);
 
    /* init kill mask */
    tgsi_set_kill_mask(machine, 0x0);
@@ -108,12 +145,11 @@ fs_sse_run( const struct sp_fragment_shader *base,
    if (quad->inout.mask == 0)
       return FALSE;
 
-
    /* store outputs */
    {
-      const ubyte *sem_name = shader->base.info.output_semantic_name;
-      const ubyte *sem_index = shader->base.info.output_semantic_index;
-      const uint n = shader->base.info.num_outputs;
+      const ubyte *sem_name = base->info.output_semantic_name;
+      const ubyte *sem_index = base->info.output_semantic_index;
+      const uint n = base->info.num_outputs;
       uint i;
       for (i = 0; i < n; i++) {
          switch (sem_name[i]) {
-- 
2.30.2