Redoing the way we handle vertex shaders for the draw module.

17 files changed, 211 insertions, 135 deletions

diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c
index 66c66ff698e..5efb173228f 100644
--- a/src/mesa/pipe/draw/draw_context.c
+++ b/src/mesa/pipe/draw/draw_context.c
@@ -155,14 +155,6 @@ void draw_set_viewport_state( struct draw_context *draw,
 }
 
 
-void
-draw_set_vertex_shader(struct draw_context *draw,
-                       const struct pipe_shader_state *shader)
-{
-   draw_flush( draw );
-   draw->vertex_shader = *shader;
-}
-
 
 void
 draw_set_vertex_buffer(struct draw_context *draw,
diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h
index 0ccf5f60466..398e96d94d2 100644
--- a/src/mesa/pipe/draw/draw_context.h
+++ b/src/mesa/pipe/draw/draw_context.h
@@ -92,9 +92,12 @@ void draw_set_rasterize_stage( struct draw_context *draw,
                                struct draw_stage *stage );
 
 
-void
-draw_set_vertex_shader(struct draw_context *draw,
-                       const struct pipe_shader_state *shader);
+void * draw_create_vertex_shader(struct draw_context *draw,
+                                 const struct pipe_shader_state *shader);
+void draw_bind_vertex_shader(struct draw_context *draw,
+                             void *vcso);
+void draw_delete_vertex_shader(struct draw_context *draw,
+                               void *vcso);
 
 
 void
diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h
index ebef5347ab9..2f52299f74a 100644
--- a/src/mesa/pipe/draw/draw_private.h
+++ b/src/mesa/pipe/draw/draw_private.h
@@ -46,6 +46,7 @@
 
 #include "draw_vertex.h"
 
+#include "x86/rtasm/x86sse.h"
 
 /**
  * Basic vertex info.
@@ -104,7 +105,7 @@ struct draw_stage
 
    void (*tri)( struct draw_stage *,
 		struct prim_header * );
-   
+
    void (*end)( struct draw_stage * );
 
    void (*reset_stipple_counter)( struct draw_stage * );
@@ -116,6 +117,15 @@ struct draw_stage
 #define VCACHE_OVERFLOW        4
 #define VS_QUEUE_LENGTH        (VCACHE_SIZE + VCACHE_OVERFLOW + 1)	/* can never fill up */
 
+/**
+ * Private version of the compiled vertex_shader
+ */
+struct draw_vertex_shader {
+   const struct pipe_shader_state *state;
+#if defined(__i386__) || defined(__386__)
+   struct x86_function  sse2_program;
+#endif
+};
 
 /**
  * Private context for the drawing module.
@@ -145,7 +155,7 @@ struct draw_context
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX];
    struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX];
-   struct pipe_shader_state vertex_shader;
+   const struct draw_vertex_shader *vertex_shader;
    struct pipe_vertex_buffer feedback_buffer[PIPE_ATTRIB_MAX];
    struct pipe_vertex_element feedback_element[PIPE_ATTRIB_MAX];
 
diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c
index 2b839d641eb..bc37e18c341 100644
--- a/src/mesa/pipe/draw/draw_vertex_fetch.c
+++ b/src/mesa/pipe/draw/draw_vertex_fetch.c
@@ -85,7 +85,7 @@ void draw_vertex_fetch( struct draw_context *draw,
       /*printf("fetch vertex %u: \n", j);*/
 
       /* loop over vertex attributes (vertex shader inputs) */
-      for (attr = 0; attr < draw->vertex_shader.num_inputs; attr++) {
+      for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) {
 
          unsigned buf = draw->vertex_element[attr].vertex_buffer_index;
          const void *src
diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c
index a2e1cdc472d..f7ddbd19093 100644
--- a/src/mesa/pipe/draw/draw_vertex_shader.c
+++ b/src/mesa/pipe/draw/draw_vertex_shader.c
@@ -36,6 +36,8 @@
 #include "draw_context.h"
 #include "draw_vertex.h"
 
+#include "x86/rtasm/x86sse.h"
+
 #include "pipe/tgsi/exec/tgsi_core.h"
 
 static INLINE unsigned
@@ -70,6 +72,7 @@ typedef void (XSTDCALL *codegen_function) (
    float (*constant)[4],
    struct tgsi_exec_vector *temporary );
 
+
 /**
  * Transform vertices with the current vertex program/shader
  * Up to four vertices can be shaded at a time.
@@ -92,7 +95,7 @@ run_vertex_program(struct draw_context *draw,
    const float *trans = draw->viewport.translate;
 
    assert(count <= 4);
-   assert(draw->vertex_shader.output_semantic_name[0]
+   assert(draw->vertex_shader->state->output_semantic_name[0]
           == TGSI_SEMANTIC_POSITION);
 
 #ifdef DEBUG
@@ -101,7 +104,7 @@ run_vertex_program(struct draw_context *draw,
 
    /* init machine state */
    tgsi_exec_machine_init(&machine,
-                          draw->vertex_shader.tokens,
+                          draw->vertex_shader->state->tokens,
                           PIPE_MAX_SAMPLERS,
                           NULL /*samplers*/ );
 
@@ -114,8 +117,8 @@ run_vertex_program(struct draw_context *draw,
    draw_vertex_fetch( draw, &machine, elts, count );
 
    /* run shader */
-   if( draw->vertex_shader.executable != NULL ) {
-      codegen_function func = (codegen_function) draw->vertex_shader.executable;
+   if( draw->vertex_shader->state->executable != NULL ) {
+      codegen_function func = (codegen_function) draw->vertex_shader->state->executable;
       func(
          machine.Inputs,
          machine.Outputs,
@@ -206,3 +209,42 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw )
    draw->vs.queue_nr = 0;
 }
 
+
+void *
+draw_create_vertex_shader(struct draw_context *draw,
+                          const struct pipe_shader_state *shader)
+{
+   struct draw_vertex_shader *vs = calloc(1, sizeof(struct draw_vertex_shader));
+
+   vs->state = shader;
+#if defined(__i386__) || defined(__386__)
+   x86_init_func(&vs->sse2_program);
+
+   tgsi_emit_sse2(shader->tokens, &vs->sse2_program);
+
+   ((struct pipe_shader_state*)(vs->state))->executable =
+      x86_get_func(&vs->sse2_program);
+#endif
+
+   return vs;
+}
+
+void draw_bind_vertex_shader(struct draw_context *draw,
+                             void *vcso)
+{
+   draw_flush(draw);
+   draw->vertex_shader = (struct draw_vertex_shader*)(vcso);
+}
+
+void draw_delete_vertex_shader(struct draw_context *draw,
+                               void *vcso)
+{
+   struct draw_vertex_shader *vs = (struct draw_vertex_shader*)(vcso);
+#if defined(__i386__) || defined(__386__)
+   x86_release_func(&vs->sse2_program);
+#endif
+   free(vcso);
+}
+
+
+
diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c
index 5c6d37a4661..19ca5e575fb 100644
--- a/src/mesa/pipe/i915simple/i915_state.c
+++ b/src/mesa/pipe/i915simple/i915_state.c
@@ -443,16 +443,13 @@ static void i915_set_polygon_stipple( struct pipe_context *pipe,
 {
 }
 
-
-static void *
-i915_create_shader_state(struct pipe_context *pipe,
-                         const struct pipe_shader_state *templ)
+static void * i915_create_fs_state(struct pipe_context *pipe,
+                                   const struct pipe_shader_state *templ)
 {
    return 0;
 }
 
-static void i915_bind_fs_state( struct pipe_context *pipe,
-                               void *fs )
+static void i915_bind_fs_state(struct pipe_context *pipe, void *fs)
 {
    struct i915_context *i915 = i915_context(pipe);
 
@@ -461,20 +458,35 @@ static void i915_bind_fs_state( struct pipe_context *pipe,
    i915->dirty |= I915_NEW_FS;
 }
 
+static void i915_delete_fs_state(struct pipe_context *pipe, void *shader)
+{
+   /*do nothing*/
+}
 
-static void i915_bind_vs_state(struct pipe_context *pipe,
-                               void *vs)
+static void *
+i915_create_vs_state(struct pipe_context *pipe,
+                     const struct pipe_shader_state *templ)
 {
    struct i915_context *i915 = i915_context(pipe);
 
    /* just pass-through to draw module */
-   draw_set_vertex_shader(i915->draw, (const struct pipe_shader_state *)vs);
+   return draw_create_vertex_shader(i915->draw, templ);
 }
 
-static void i915_delete_shader_state(struct pipe_context *pipe,
-                                     void *shader)
+static void i915_bind_vs_state(struct pipe_context *pipe, void *vs)
 {
-   /*do nothing*/
+   struct i915_context *i915 = i915_context(pipe);
+
+   /* just pass-through to draw module */
+   draw_bind_vertex_shader(i915->draw, vs);
+}
+
+static void i915_delete_vs_state(struct pipe_context *pipe, void *shader)
+{
+   struct i915_context *i915 = i915_context(pipe);
+
+   /* just pass-through to draw module */
+   draw_delete_vertex_shader(i915->draw, shader);
 }
 
 static void i915_set_constant_buffer(struct pipe_context *pipe,
@@ -707,12 +719,12 @@ i915_init_state_functions( struct i915_context *i915 )
    i915->pipe.create_rasterizer_state = i915_create_rasterizer_state;
    i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state;
    i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state;
-   i915->pipe.create_fs_state = i915_create_shader_state;
+   i915->pipe.create_fs_state = i915_create_fs_state;
    i915->pipe.bind_fs_state = i915_bind_fs_state;
-   i915->pipe.delete_fs_state = i915_delete_shader_state;
-   i915->pipe.create_vs_state = i915_create_shader_state;
+   i915->pipe.delete_fs_state = i915_delete_fs_state;
+   i915->pipe.create_vs_state = i915_create_vs_state;
    i915->pipe.bind_vs_state = i915_bind_vs_state;
-   i915->pipe.delete_vs_state = i915_delete_shader_state;
+   i915->pipe.delete_vs_state = i915_delete_vs_state;
 
    i915->pipe.set_blend_color = i915_set_blend_color;
    i915->pipe.set_clip_state = i915_set_clip_state;
diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c
index e415966dae1..695bf1a9e02 100644
--- a/src/mesa/pipe/softpipe/sp_context.c
+++ b/src/mesa/pipe/softpipe/sp_context.c
@@ -275,12 +275,12 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys,
    softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state;
    softpipe->pipe.bind_rasterizer_state   = softpipe_bind_rasterizer_state;
    softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state;
-   softpipe->pipe.create_fs_state = softpipe_create_shader_state;
+   softpipe->pipe.create_fs_state = softpipe_create_fs_state;
    softpipe->pipe.bind_fs_state   = softpipe_bind_fs_state;
-   softpipe->pipe.delete_fs_state = softpipe_delete_shader_state;
-   softpipe->pipe.create_vs_state = softpipe_create_shader_state;
+   softpipe->pipe.delete_fs_state = softpipe_delete_fs_state;
+   softpipe->pipe.create_vs_state = softpipe_create_vs_state;
    softpipe->pipe.bind_vs_state   = softpipe_bind_vs_state;
-   softpipe->pipe.delete_vs_state = softpipe_delete_shader_state;
+   softpipe->pipe.delete_vs_state = softpipe_delete_vs_state;
 
    softpipe->pipe.set_blend_color = softpipe_set_blend_color;
    softpipe->pipe.set_clip_state = softpipe_set_clip_state;
diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h
index c0a681f3d6b..ccf29b5683d 100644
--- a/src/mesa/pipe/softpipe/sp_context.h
+++ b/src/mesa/pipe/softpipe/sp_context.h
@@ -62,8 +62,12 @@ struct draw_stage;
 #define SP_NEW_VS            0x2000
 #define SP_NEW_CONSTANTS     0x4000
 
+struct sp_vertex_shader_state {
+   const struct pipe_shader_state *state;
+   void  *draw_data;
+};
 
-struct softpipe_context {     
+struct softpipe_context {
    struct pipe_context pipe;  /**< base class */
    struct softpipe_winsys *winsys;	/**< window system interface */
 
@@ -76,7 +80,7 @@ struct softpipe_context {
    const struct pipe_depth_stencil_state   *depth_stencil;
    const struct pipe_rasterizer_state *rasterizer;
    const struct pipe_shader_state *fs;
-   const struct pipe_shader_state *vs;
+   const struct sp_vertex_shader_state *vs;
 
    struct pipe_blend_color blend_color;
    struct pipe_clear_color_state clear_color;
diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h
index f0e1461d258..f9061e86e53 100644
--- a/src/mesa/pipe/softpipe/sp_state.h
+++ b/src/mesa/pipe/softpipe/sp_state.h
@@ -87,12 +87,14 @@ void softpipe_set_constant_buffer(struct pipe_context *,
 void softpipe_set_feedback_state( struct pipe_context *,
                                   const struct pipe_feedback_state * );
 
-void *
-softpipe_create_shader_state( struct pipe_context *,
-                              const struct pipe_shader_state * );
-void softpipe_bind_fs_state( struct pipe_context *, void * );
-void softpipe_bind_vs_state( struct pipe_context *, void * );
-void softpipe_delete_shader_state( struct pipe_context *, void * );
+void *softpipe_create_fs_state(struct pipe_context *,
+                               const struct pipe_shader_state *);
+void softpipe_bind_fs_state(struct pipe_context *, void *);
+void softpipe_delete_fs_state(struct pipe_context *, void *);
+void *softpipe_create_vs_state(struct pipe_context *,
+                               const struct pipe_shader_state *);
+void softpipe_bind_vs_state(struct pipe_context *, void *);
+void softpipe_delete_vs_state(struct pipe_context *, void *);
 
 void softpipe_set_polygon_stipple( struct pipe_context *,
 				  const struct pipe_poly_stipple * );
diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c
index 6c6e7980698..b6145df8e25 100644
--- a/src/mesa/pipe/softpipe/sp_state_derived.c
+++ b/src/mesa/pipe/softpipe/sp_state_derived.c
@@ -43,7 +43,7 @@
  */
 static void calculate_vertex_layout( struct softpipe_context *softpipe )
 {
-   const struct pipe_shader_state *vs = softpipe->vs;
+   const struct pipe_shader_state *vs = softpipe->vs->state;
    const struct pipe_shader_state *fs = softpipe->fs;
    const interp_mode colorInterp
       = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR;
diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c
index 8306a95f449..f1bec2c73ac 100644
--- a/src/mesa/pipe/softpipe/sp_state_fs.c
+++ b/src/mesa/pipe/softpipe/sp_state_fs.c
@@ -33,10 +33,13 @@
 #include "pipe/draw/draw_context.h"
 
 
-void * softpipe_create_shader_state(struct pipe_context *pipe,
-                                    const struct pipe_shader_state *templ)
+void * softpipe_create_fs_state(struct pipe_context *pipe,
+                                const struct pipe_shader_state *templ)
 {
-   /* we just want the pipe_shader_state template in the bind calls */
+   /* Decide whether we'll be codegenerating this shader and if so do
+    * that now.
+    */
+
    return 0;
 }
 
@@ -49,25 +52,51 @@ void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
    softpipe->dirty |= SP_NEW_FS;
 }
 
+void softpipe_delete_fs_state(struct pipe_context *pipe,
+                              void *shader)
+{
+}
+
+
+void * softpipe_create_vs_state(struct pipe_context *pipe,
+                                const struct pipe_shader_state *templ)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+   struct sp_vertex_shader_state *state =
+      malloc(sizeof(struct sp_vertex_shader_state));
+
+   state->state = templ;
+   state->draw_data = draw_create_vertex_shader(softpipe->draw,
+                                                state->state);
+
+   return state;
+}
 
 void softpipe_bind_vs_state(struct pipe_context *pipe, void *vs)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
 
-   softpipe->vs = (struct pipe_shader_state *)vs;
+   softpipe->vs = (const struct sp_vertex_shader_state *)vs;
 
-   softpipe->dirty |= SP_NEW_VS;
+   draw_bind_vertex_shader(softpipe->draw, softpipe->vs->draw_data);
 
-   draw_set_vertex_shader(softpipe->draw, (struct pipe_shader_state *)vs);
+   softpipe->dirty |= SP_NEW_VS;
 }
 
-
-void softpipe_delete_shader_state( struct pipe_context *pipe,
-                                   void *shader )
+void softpipe_delete_vs_state(struct pipe_context *pipe,
+                              void *vs)
 {
-   /* do nothing */
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   struct sp_vertex_shader_state *state =
+      (struct sp_vertex_shader_state *)vs;
+
+   draw_delete_vertex_shader(softpipe->draw, state->draw_data);
+   free(state);
 }
 
+
+
 void softpipe_set_constant_buffer(struct pipe_context *pipe,
                                   uint shader, uint index,
                                   const struct pipe_constant_buffer *buf)
diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c
index 25dbc2c2442..3a7ce9405e8 100644
--- a/src/mesa/state_tracker/st_cb_program.c
+++ b/src/mesa/state_tracker/st_cb_program.c
@@ -86,10 +86,6 @@ static struct gl_program *st_new_program( GLcontext *ctx,
 
       prog->serialNo = 1;
 
-#if defined(__i386__) || defined(__386__)
-      x86_init_func( &prog->sse2_program );
-#endif
-
       return _mesa_init_vertex_program( ctx, 
 					&prog->Base,
 					target, 
@@ -129,9 +125,6 @@ static void st_delete_program( GLcontext *ctx,
    case GL_VERTEX_PROGRAM_ARB:
       {
          struct st_vertex_program *stvp = (struct st_vertex_program *) prog;
-#if defined(__i386__) || defined(__386__)
-         x86_release_func( &stvp->sse2_program );
-#endif
          st_remove_vertex_program(st, stvp);
       }
       break;
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
index ce5bf0c8a94..7c836ba2c13 100644
--- a/src/mesa/state_tracker/st_draw.c
+++ b/src/mesa/state_tracker/st_draw.c
@@ -405,7 +405,7 @@ st_feedback_draw_vbo(GLcontext *ctx,
    draw_set_viewport_state(draw, &st->state.viewport);
    draw_set_clip_state(draw, &st->state.clip);
    draw_set_rasterizer_state(draw, &st->state.rasterizer->state);
-   draw_set_vertex_shader(draw, &st->state.vs->state);
+   draw_bind_vertex_shader(draw, st->state.vs->data);
    /* XXX need to set vertex info too */
 
 
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index 44cc33b09fb..116b59a067a 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -253,14 +253,6 @@ st_translate_vertex_program(struct st_context *st,
    if (TGSI_DEBUG)
       tgsi_dump( tokensOut, 0 );
 
-#if defined(__i386__) || defined(__386__)
-   if (stvp->sse2_program.csr == stvp->sse2_program.store)
-      tgsi_emit_sse2( tokensOut, &stvp->sse2_program );
-
-   if (!cso->state.executable)
-      ((struct cso_vertex_shader*)cso)->state.executable = (void *) x86_get_func( &stvp->sse2_program );
-#endif
-
    return cso;
 }
 
diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h
index 6e106c16b02..2b792013131 100644
--- a/src/mesa/state_tracker/st_program.h
+++ b/src/mesa/state_tracker/st_program.h
@@ -79,10 +79,6 @@ struct st_vertex_program
    /** The program in TGSI format */
    struct tgsi_token tokens[ST_FP_MAX_TOKENS];
 
-#if defined(__i386__) || defined(__386__)
-   struct x86_function  sse2_program;
-#endif
-
    /** Pointer to the corresponding cached shader */
    const struct cso_vertex_shader *vs;
 
diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c
index a6cfa40f91f..f9a9352e468 100644
--- a/src/mesa/x86/rtasm/x86sse.c
+++ b/src/mesa/x86/rtasm/x86sse.c
@@ -6,22 +6,22 @@
 #define DISASSEM 0
 #define X86_TWOB 0x0f
 
-static GLubyte *cptr( void (*label)() )
+static unsigned char *cptr( void (*label)() )
 {
-   return (char *)(unsigned long)label;
+   return (unsigned char *)(unsigned long)label;
 }
 
 
 /* Emit bytes to the instruction stream:
  */
-static void emit_1b( struct x86_function *p, GLbyte b0 )
+static void emit_1b( struct x86_function *p, char b0 )
 {
-   *(GLbyte *)(p->csr++) = b0;
+   *(char *)(p->csr++) = b0;
 }
 
-static void emit_1i( struct x86_function *p, GLint i0 )
+static void emit_1i( struct x86_function *p, int i0 )
 {
-   *(GLint *)(p->csr) = i0;
+   *(int *)(p->csr) = i0;
    p->csr += 4;
 }
 
@@ -35,20 +35,20 @@ static void disassem( struct x86_function *p, const char *fn )
 #endif
 }
 
-static void emit_1ub_fn( struct x86_function *p, GLubyte b0, const char *fn )
+static void emit_1ub_fn( struct x86_function *p, unsigned char b0, const char *fn )
 {
    disassem(p, fn);
    *(p->csr++) = b0;
 }
 
-static void emit_2ub_fn( struct x86_function *p, GLubyte b0, GLubyte b1, const char *fn )
+static void emit_2ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, const char *fn )
 {
    disassem(p, fn);
    *(p->csr++) = b0;
    *(p->csr++) = b1;
 }
 
-static void emit_3ub_fn( struct x86_function *p, GLubyte b0, GLubyte b1, GLubyte b2, const char *fn )
+static void emit_3ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2, const char *fn )
 {
    disassem(p, fn);
    *(p->csr++) = b0;
@@ -69,7 +69,7 @@ static void emit_modrm( struct x86_function *p,
 			struct x86_reg reg, 
 			struct x86_reg regmem )
 {
-   GLubyte val = 0;
+   unsigned char val = 0;
    
    assert(reg.mod == mod_REG);
    
@@ -104,7 +104,7 @@ static void emit_modrm( struct x86_function *p,
 
 
 static void emit_modrm_noreg( struct x86_function *p,
-			      GLuint op,
+			      unsigned op,
 			      struct x86_reg regmem )
 {
    struct x86_reg dummy = x86_make_reg(file_REG32, op);
@@ -117,8 +117,8 @@ static void emit_modrm_noreg( struct x86_function *p,
  * the arguments presented.
  */
 static void emit_op_modrm( struct x86_function *p,
-			   GLubyte op_dst_is_reg, 
-			   GLubyte op_dst_is_mem,
+			   unsigned char op_dst_is_reg, 
+			   unsigned char op_dst_is_mem,
 			   struct x86_reg dst,
 			   struct x86_reg src )
 {  
@@ -162,7 +162,7 @@ struct x86_reg x86_make_reg( enum x86_reg_file file,
 }
 
 struct x86_reg x86_make_disp( struct x86_reg reg,
-			      GLint disp )
+			      int disp )
 {
    assert(reg.file == file_REG32);
 
@@ -191,7 +191,7 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg )
    return x86_make_reg( reg.file, reg.idx );
 }
 
-GLubyte *x86_get_label( struct x86_function *p )
+unsigned char *x86_get_label( struct x86_function *p )
 {
    return p->csr;
 }
@@ -205,13 +205,13 @@ GLubyte *x86_get_label( struct x86_function *p )
 
 void x86_jcc( struct x86_function *p,
 	      enum x86_cc cc,
-	      GLubyte *label )
+	      unsigned char *label )
 {
-   GLint offset = label - (x86_get_label(p) + 2);
+   int offset = label - (x86_get_label(p) + 2);
    
    if (offset <= 127 && offset >= -128) {
       emit_1ub(p, 0x70 + cc);
-      emit_1b(p, (GLbyte) offset);
+      emit_1b(p, (char) offset);
    }
    else {
       offset = label - (x86_get_label(p) + 6);
@@ -222,7 +222,7 @@ void x86_jcc( struct x86_function *p,
 
 /* Always use a 32bit offset for forward jumps:
  */
-GLubyte *x86_jcc_forward( struct x86_function *p,
+unsigned char *x86_jcc_forward( struct x86_function *p,
 			  enum x86_cc cc )
 {
    emit_2ub(p, 0x0f, 0x80 + cc);
@@ -230,14 +230,14 @@ GLubyte *x86_jcc_forward( struct x86_function *p,
    return x86_get_label(p);
 }
 
-GLubyte *x86_jmp_forward( struct x86_function *p)
+unsigned char *x86_jmp_forward( struct x86_function *p)
 {
    emit_1ub(p, 0xe9);
    emit_1i(p, 0);
    return x86_get_label(p);
 }
 
-GLubyte *x86_call_forward( struct x86_function *p)
+unsigned char *x86_call_forward( struct x86_function *p)
 {
    emit_1ub(p, 0xe8);
    emit_1i(p, 0);
@@ -247,12 +247,12 @@ GLubyte *x86_call_forward( struct x86_function *p)
 /* Fixup offset from forward jump:
  */
 void x86_fixup_fwd_jump( struct x86_function *p,
-			 GLubyte *fixup )
+			 unsigned char *fixup )
 {
    *(int *)(fixup - 4) = x86_get_label(p) - fixup;
 }
 
-void x86_jmp( struct x86_function *p, GLubyte *label)
+void x86_jmp( struct x86_function *p, unsigned char *label)
 {
    emit_1ub(p, 0xe9);
    emit_1i(p, label - x86_get_label(p) - 4);
@@ -268,7 +268,7 @@ void x86_call( struct x86_function *p, void (*label)())
  * Temporary. As I need immediate operands, and dont want to mess with the codegen,
  * I load the immediate into general purpose register and use it.
  */
-void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, GLint imm )
+void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm )
 {
    assert(dst.mod == mod_REG);
    emit_1ub(p, 0xb8 + dst.idx);
@@ -595,7 +595,7 @@ void sse_cvtps2pi( struct x86_function *p,
 void sse_shufps( struct x86_function *p,
 		 struct x86_reg dest,
 		 struct x86_reg arg0,
-		 GLubyte shuf) 
+		 unsigned char shuf) 
 {
    emit_2ub(p, X86_TWOB, 0xC6);
    emit_modrm(p, dest, arg0);
@@ -605,7 +605,7 @@ void sse_shufps( struct x86_function *p,
 void sse_cmpps( struct x86_function *p,
 		struct x86_reg dest,
 		struct x86_reg arg0,
-		GLubyte cc) 
+		unsigned char cc) 
 {
    emit_2ub(p, X86_TWOB, 0xC2);
    emit_modrm(p, dest, arg0);
@@ -630,7 +630,7 @@ void sse_pmovmskb( struct x86_function *p,
 void sse2_pshufd( struct x86_function *p,
 		  struct x86_reg dest,
 		  struct x86_reg arg0,
-		  GLubyte shuf) 
+		  unsigned char shuf) 
 {
    emit_3ub(p, 0x66, X86_TWOB, 0x70);
    emit_modrm(p, dest, arg0);
@@ -772,11 +772,11 @@ void x87_fclex( struct x86_function *p )
 
 
 static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86_reg arg,
-			  GLubyte dst0ub0,
-			  GLubyte dst0ub1,
-			  GLubyte arg0ub0,
-			  GLubyte arg0ub1,
-			  GLubyte argmem_noreg)
+			  unsigned char dst0ub0,
+			  unsigned char dst0ub1,
+			  unsigned char arg0ub0,
+			  unsigned char arg0ub1,
+			  unsigned char argmem_noreg)
 {
    assert(dst.file == file_x87);
 
@@ -1116,7 +1116,7 @@ void mmx_movq( struct x86_function *p,
  * account any push/pop activity:
  */
 struct x86_reg x86_fn_arg( struct x86_function *p,
-			   GLuint arg )
+			   unsigned arg )
 {
    return x86_make_disp(x86_make_reg(file_REG32, reg_SP), 
 			p->stack_offset + arg * 4);	/* ??? */
@@ -1128,7 +1128,7 @@ void x86_init_func( struct x86_function *p )
    x86_init_func_size(p, 1024);
 }
 
-void x86_init_func_size( struct x86_function *p, GLuint code_size )
+void x86_init_func_size( struct x86_function *p, unsigned code_size )
 {
    p->store = _mesa_exec_malloc(code_size);
    p->csr = p->store;
diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h
index c1ca06088bc..63b9a36392a 100644
--- a/src/mesa/x86/rtasm/x86sse.h
+++ b/src/mesa/x86/rtasm/x86sse.h
@@ -4,24 +4,22 @@
 
 #if defined(__i386__) || defined(__386__)
 
-#include "glheader.h"
-
 /* It is up to the caller to ensure that instructions issued are
  * suitable for the host cpu.  There are no checks made in this module
  * for mmx/sse/sse2 support on the cpu.
  */
 struct x86_reg {
-   GLuint file:3;
-   GLuint idx:3;
-   GLuint mod:2;		/* mod_REG if this is just a register */
-   GLint  disp:24;		/* only +/- 23bits of offset - should be enough... */
+   unsigned file:3;
+   unsigned idx:3;
+   unsigned mod:2;		/* mod_REG if this is just a register */
+   int      disp:24;		/* only +/- 23bits of offset - should be enough... */
 };
 
 struct x86_function {
-   GLubyte *store;
-   GLubyte *csr;
-   GLuint stack_offset;
-   GLint need_emms;
+   unsigned char *store;
+   unsigned char *csr;
+   unsigned stack_offset;
+   int need_emms;
    const char *fn;
 };
 
@@ -81,7 +79,7 @@ enum sse_cc {
 
 
 void x86_init_func( struct x86_function *p );
-void x86_init_func_size( struct x86_function *p, GLuint code_size );
+void x86_init_func_size( struct x86_function *p, unsigned code_size );
 void x86_release_func( struct x86_function *p );
 void (*x86_get_func( struct x86_function *p ))( void );
 
@@ -93,7 +91,7 @@ struct x86_reg x86_make_reg( enum x86_reg_file file,
 			     enum x86_reg_name idx );
 
 struct x86_reg x86_make_disp( struct x86_reg reg,
-			      GLint disp );
+			      int disp );
 
 struct x86_reg x86_deref( struct x86_reg reg );
 
@@ -102,23 +100,23 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg );
 
 /* Labels, jumps and fixup:
  */
-GLubyte *x86_get_label( struct x86_function *p );
+unsigned char *x86_get_label( struct x86_function *p );
 
 void x86_jcc( struct x86_function *p,
 	      enum x86_cc cc,
-	      GLubyte *label );
+	      unsigned char *label );
 
-GLubyte *x86_jcc_forward( struct x86_function *p,
+unsigned char *x86_jcc_forward( struct x86_function *p,
 			  enum x86_cc cc );
 
-GLubyte *x86_jmp_forward( struct x86_function *p);
+unsigned char *x86_jmp_forward( struct x86_function *p);
 
-GLubyte *x86_call_forward( struct x86_function *p);
+unsigned char *x86_call_forward( struct x86_function *p);
 
 void x86_fixup_fwd_jump( struct x86_function *p,
-			 GLubyte *fixup );
+			 unsigned char *fixup );
 
-void x86_jmp( struct x86_function *p, GLubyte *label );
+void x86_jmp( struct x86_function *p, unsigned char *label );
 
 void x86_call( struct x86_function *p, void (*label)() );
 
@@ -126,7 +124,7 @@ void x86_call( struct x86_function *p, void (*label)() );
  * Temporary. As I need immediate operands, and dont want to mess with the codegen,
  * I load the immediate into general purpose register and use it.
  */
-void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, GLint imm );
+void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm );
 
 
 /* Macro for sse_shufps() and sse2_pshufd():
@@ -147,7 +145,8 @@ void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
 void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
-void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf );
+void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0,
+                  unsigned char shuf );
 void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 
@@ -157,7 +156,8 @@ void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg sr
 void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
-void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, GLubyte cc );
+void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src,
+                unsigned char cc );
 void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
@@ -175,7 +175,8 @@ void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
 void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
-void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf );
+void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0,
+                 unsigned char shuf );
 void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src );
 
 void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
@@ -247,7 +248,7 @@ void x87_fucom( struct x86_function *p, struct x86_reg arg );
  * account any push/pop activity.  Note - doesn't track explict
  * manipulation of ESP by other instructions.
  */
-struct x86_reg x86_fn_arg( struct x86_function *p, GLuint arg );
+struct x86_reg x86_fn_arg( struct x86_function *p, unsigned arg );
 
 #endif
 #endif