intel: Replace IS_G4X() across the driver with context structure usage.

[mesa.git] / src / mesa / drivers / dri / i965 / brw_clip_util.c

diff --git a/src/mesa/drivers/dri/i965/brw_clip_util.c b/src/mesa/drivers/dri/i965/brw_clip_util.c
index 19bef19801a14d3aee9f99f00db82d438168ea51..86fed59fa4b89dfbc90b7747a3407d72d7a38b95 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_clip_util.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_util.c
@@ -30,11 +30,11 @@
   */
 
 
-#include "glheader.h"
-#include "macros.h"
-#include "enums.h"
-
+#include "main/glheader.h"
+#include "main/macros.h"
+#include "main/enums.h"
 #include "shader/program.h"
+
 #include "intel_batchbuffer.h"
 
 #include "brw_defines.h"
@@ -46,8 +46,7 @@
 
 
 
-
-static struct brw_reg get_tmp( struct brw_clip_compile *c )
+struct brw_reg get_tmp( struct brw_clip_compile *c )
 {
    struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0);
 
@@ -90,7 +89,7 @@ void brw_clip_init_planes( struct brw_clip_compile *c )
 
 /* Project 'pos' to screen space (or back again), overwrite with results:
  */
-static void brw_clip_project_position(struct brw_clip_compile *c, struct brw_reg pos )
+void brw_clip_project_position(struct brw_clip_compile *c, struct brw_reg pos )
 {
    struct brw_compile *p = &c->func;
 
@@ -136,11 +135,16 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
                             GLboolean force_edgeflag)
 {
    struct brw_compile *p = &c->func;
+   struct intel_context *intel = &p->brw->intel;
    struct brw_reg tmp = get_tmp(c);
    GLuint i;
 
    /* Just copy the vertex header:
     */
+   /*
+    * After CLIP stage, only first 256 bits of the VUE are read
+    * back on Ironlake, so needn't change it
+    */
    brw_copy_indirect_to_indirect(p, dest_ptr, v0_ptr, 1);
       
    /* Iterate over each attribute (could be done in pairs?)
@@ -148,6 +152,9 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
    for (i = 0; i < c->nr_attrs; i++) {
       GLuint delta = i*16 + 32;
 
+      if (intel->is_ironlake)
+          delta = i * 16 + 32 * 3;
+
       if (delta == c->offset[VERT_RESULT_EDGE]) {
         if (force_edgeflag) 
            brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(1));
@@ -178,6 +185,10 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
 
    if (i & 1) {
       GLuint delta = i*16 + 32;
+
+      if (intel->is_ironlake)
+          delta = i * 16 + 32 * 3;
+
       brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(0));
    }
 
@@ -203,6 +214,8 @@ void brw_clip_emit_vue(struct brw_clip_compile *c,
    struct brw_compile *p = &c->func;
    GLuint start = c->last_mrf;
 
+   brw_clip_ff_sync(c);
+
    assert(!(allocate && eot));
    
    /* Cycle through mrf regs - probably futile as we have to wait for
@@ -253,6 +266,7 @@ void brw_clip_kill_thread(struct brw_clip_compile *c)
 {
    struct brw_compile *p = &c->func;
 
+   brw_clip_ff_sync(c);
    /* Send an empty message to kill the thread and release any
     * allocated urb entry:
     */
@@ -262,7 +276,7 @@ void brw_clip_kill_thread(struct brw_clip_compile *c)
                 c->reg.R0,
                 0,             /* allocate */
                 0,             /* used */
-                0,             /* msg len */
+                1,             /* msg len */
                 0,             /* response len */
                 1,             /* eot */
                 1,             /* writes complete */
@@ -272,6 +286,7 @@ void brw_clip_kill_thread(struct brw_clip_compile *c)
 
 
 
+
 struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c )
 {
    return brw_address(c->reg.fixed_planes);
@@ -327,8 +342,7 @@ void brw_clip_init_clipmask( struct brw_clip_compile *c )
    
    /* Shift so that lowest outcode bit is rightmost: 
     */
-   brw_MOV(p, c->reg.planemask, incoming);
-   brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(26));
+   brw_SHR(p, c->reg.planemask, incoming, brw_imm_ud(26));
 
    if (c->key.nr_userclip) {
       struct brw_reg tmp = retype(vec1(get_tmp(c)), BRW_REGISTER_TYPE_UD);
@@ -342,13 +356,46 @@ void brw_clip_init_clipmask( struct brw_clip_compile *c )
       
       release_tmp(c, tmp);
    }
+}
 
-   /* Test for -ve rhw workaround 
-    */
-   brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
-   brw_AND(p, vec1(brw_null_reg()), incoming, brw_imm_ud(1<<20));
-   brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(0x3f));
-   brw_set_predicate_control(p, BRW_PREDICATE_NONE);
-
+void brw_clip_ff_sync(struct brw_clip_compile *c)
+{
+    struct intel_context *intel = &c->func.brw->intel;
+
+    if (intel->needs_ff_sync) {
+        struct brw_compile *p = &c->func;
+        struct brw_instruction *need_ff_sync;
+
+        brw_set_conditionalmod(p, BRW_CONDITIONAL_Z);
+        brw_AND(p, brw_null_reg(), c->reg.ff_sync, brw_imm_ud(0x1));
+        need_ff_sync = brw_IF(p, BRW_EXECUTE_1);
+        {
+            brw_OR(p, c->reg.ff_sync, c->reg.ff_sync, brw_imm_ud(0x1));
+            brw_ff_sync(p, 
+                    c->reg.R0,
+                    0,
+                    c->reg.R0,
+                    1, 
+                    1,         /* used */
+                    1,         /* msg length */
+                    1,         /* response length */
+                    0,         /* eot */
+                    1,         /* write compelete */
+                    0,         /* urb offset */
+                    BRW_URB_SWIZZLE_NONE);
+        }
+        brw_ENDIF(p, need_ff_sync);
+        brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    }
 }
 
+void brw_clip_init_ff_sync(struct brw_clip_compile *c)
+{
+    struct intel_context *intel = &c->func.brw->intel;
+
+    if (intel->needs_ff_sync) {
+       struct brw_compile *p = &c->func;
+        
+        brw_MOV(p, c->reg.ff_sync, brw_imm_ud(0));
+    }
+}