intel: Annotate debug printout checks with unlikely().

This provides the optimizer with hints about code hotness, which we're
quite certain about for debug printouts (or, rather, while we
developers often hit the checks for debug printouts, we don't care
about performance while doing so).

34 files changed, 116 insertions, 129 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_cc.c b/src/mesa/drivers/dri/i965/brw_cc.c
index 00418760da3..a8369b07c35 100644
--- a/src/mesa/drivers/dri/i965/brw_cc.c
+++ b/src/mesa/drivers/dri/i965/brw_cc.c
@@ -204,7 +204,7 @@ static void upload_cc_unit(struct brw_context *brw)
       cc.cc2.depth_write_enable = ctx->Depth.Mask;
    }
 
-   if (intel->stats_wm || (INTEL_DEBUG & DEBUG_STATS))
+   if (intel->stats_wm || unlikely(INTEL_DEBUG & DEBUG_STATS))
       cc.cc5.statistics_enable = 1;
 
    /* CACHE_NEW_CC_VP */
diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c
index 15e60bf3ce3..1be165cc9a1 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.c
+++ b/src/mesa/drivers/dri/i965/brw_clip.c
@@ -133,13 +133,13 @@ static void compile_clip_prog( struct brw_context *brw,
     */
    program = brw_get_program(&c.func, &program_size);
 
-    if (INTEL_DEBUG & DEBUG_CLIP) {
+   if (unlikely(INTEL_DEBUG & DEBUG_CLIP)) {
       printf("clip:\n");
       for (i = 0; i < program_size / sizeof(struct brw_instruction); i++)
 	 brw_disasm(stdout, &((struct brw_instruction *)program)[i],
 		    intel->gen);
       printf("\n");
-    }
+   }
 
    /* Upload
     */
diff --git a/src/mesa/drivers/dri/i965/brw_clip_state.c b/src/mesa/drivers/dri/i965/brw_clip_state.c
index 885167da908..60fd5fa7d9e 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_state.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_state.c
@@ -114,10 +114,10 @@ clip_unit_create_from_key(struct brw_context *brw,
       clip.thread4.max_threads = 1 - 1;
    }
 
-   if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
+   if (unlikely(INTEL_DEBUG & DEBUG_SINGLE_THREAD))
       clip.thread4.max_threads = 0;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (unlikely(INTEL_DEBUG & DEBUG_STATS))
       clip.thread4.stats_enable = 1;
 
    clip.clip5.userclip_enable_flags = 0x7f;
diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
index 04bc8cb2db0..a1f403ca4e6 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.c
+++ b/src/mesa/drivers/dri/i965/brw_draw.c
@@ -42,7 +42,7 @@
 
 #include "intel_batchbuffer.h"
 
-#define FILE_DEBUG_FLAG DEBUG_BATCH
+#define FILE_DEBUG_FLAG DEBUG_PRIMS
 
 static GLuint prim_to_hw_prim[GL_POLYGON+1] = {
    _3DPRIM_POINTLIST,
@@ -83,8 +83,7 @@ static GLuint brw_set_prim(struct brw_context *brw,
    struct gl_context *ctx = &brw->intel.ctx;
    GLenum mode = prim->mode;
 
-   if (INTEL_DEBUG & DEBUG_PRIMS)
-      printf("PRIM: %s\n", _mesa_lookup_enum_by_nr(prim->mode));
+   DBG("PRIM: %s\n", _mesa_lookup_enum_by_nr(prim->mode));
 
    /* Slight optimization to avoid the GS program when not needed:
     */
@@ -133,9 +132,8 @@ static void brw_emit_prim(struct brw_context *brw,
    struct brw_3d_primitive prim_packet;
    struct intel_context *intel = &brw->intel;
 
-   if (INTEL_DEBUG & DEBUG_PRIMS)
-      printf("PRIM: %s %d %d\n", _mesa_lookup_enum_by_nr(prim->mode), 
-		   prim->start, prim->count);
+   DBG("PRIM: %s %d %d\n", _mesa_lookup_enum_by_nr(prim->mode),
+       prim->start, prim->count);
 
    prim_packet.header.opcode = CMD_3D_PRIM;
    prim_packet.header.length = sizeof(prim_packet)/4 - 2;
diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c
index c4654360d46..2cefe614dd2 100644
--- a/src/mesa/drivers/dri/i965/brw_draw_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c
@@ -168,7 +168,7 @@ static GLuint byte_types_scale[5] = {
 static GLuint get_surface_type( GLenum type, GLuint size,
                                 GLenum format, GLboolean normalized )
 {
-   if (INTEL_DEBUG & DEBUG_VERTS)
+   if (unlikely(INTEL_DEBUG & DEBUG_VERTS))
       printf("type %s size %d normalized %d\n", 
 		   _mesa_lookup_enum_by_nr(type), size, normalized);
 
diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index 2ed59d3f5d4..3caddae4a10 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -3129,7 +3129,7 @@ fs_visitor::generate_code()
    const char *last_annotation_string = NULL;
    ir_instruction *last_annotation_ir = NULL;
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       printf("Native code for fragment shader %d:\n",
 	     ctx->Shader.CurrentFragmentProgram->Name);
    }
@@ -3141,7 +3141,7 @@ fs_visitor::generate_code()
       fs_inst *inst = (fs_inst *)iter.get();
       struct brw_reg src[3], dst;
 
-      if (INTEL_DEBUG & DEBUG_WM) {
+      if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
 	 if (last_annotation_ir != inst->ir) {
 	    last_annotation_ir = inst->ir;
 	    if (last_annotation_ir) {
@@ -3335,7 +3335,7 @@ fs_visitor::generate_code()
 	 this->fail = true;
       }
 
-      if (INTEL_DEBUG & DEBUG_WM) {
+      if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
 	 for (unsigned int i = last_native_inst; i < p->nr_insn; i++) {
 	    if (0) {
 	       printf("0x%08x 0x%08x 0x%08x 0x%08x ",
@@ -3376,7 +3376,7 @@ brw_wm_fs_emit(struct brw_context *brw, struct brw_wm_compile *c)
     */
    c->dispatch_width = 8;
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       printf("GLSL IR for native fragment shader %d:\n", prog->Name);
       _mesa_print_ir(shader->ir, NULL);
       printf("\n");
diff --git a/src/mesa/drivers/dri/i965/brw_gs.c b/src/mesa/drivers/dri/i965/brw_gs.c
index cfcc8ea4d6a..b0c76f4094d 100644
--- a/src/mesa/drivers/dri/i965/brw_gs.c
+++ b/src/mesa/drivers/dri/i965/brw_gs.c
@@ -127,8 +127,8 @@ static void compile_gs_prog( struct brw_context *brw,
     */
    program = brw_get_program(&c.func, &program_size);
 
-    if (INTEL_DEBUG & DEBUG_GS) {
-       int i;
+   if (unlikely(INTEL_DEBUG & DEBUG_GS)) {
+      int i;
 
       printf("gs:\n");
       for (i = 0; i < program_size / sizeof(struct brw_instruction); i++)
diff --git a/src/mesa/drivers/dri/i965/brw_gs_state.c b/src/mesa/drivers/dri/i965/brw_gs_state.c
index 63562ebcfc2..69a5f7a6667 100644
--- a/src/mesa/drivers/dri/i965/brw_gs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_gs_state.c
@@ -101,7 +101,7 @@ gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key)
    if (intel->gen == 5)
       gs.thread4.rendering_enable = 1;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (unlikely(INTEL_DEBUG & DEBUG_STATS))
       gs.thread4.stats_enable = 1;
 
    bo = brw_upload_cache(&brw->cache, BRW_GS_UNIT,
diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c
index 24041e57b00..1d350bc0413 100644
--- a/src/mesa/drivers/dri/i965/brw_misc_state.c
+++ b/src/mesa/drivers/dri/i965/brw_misc_state.c
@@ -555,7 +555,7 @@ static void upload_invarient_state( struct brw_context *brw )
       memset(&vfs, 0, sizeof(vfs));
 
       vfs.opcode = brw->CMD_VF_STATISTICS;
-      if (INTEL_DEBUG & DEBUG_STATS)
+      if (unlikely(INTEL_DEBUG & DEBUG_STATS))
 	 vfs.statistics_enable = 1; 
 
       BRW_BATCH_STRUCT(brw, &vfs);
diff --git a/src/mesa/drivers/dri/i965/brw_sf.c b/src/mesa/drivers/dri/i965/brw_sf.c
index 7dbd70daaea..6da155b1a9b 100644
--- a/src/mesa/drivers/dri/i965/brw_sf.c
+++ b/src/mesa/drivers/dri/i965/brw_sf.c
@@ -108,7 +108,7 @@ static void compile_sf_prog( struct brw_context *brw,
     */
    program = brw_get_program(&c.func, &program_size);
 
-   if (INTEL_DEBUG & DEBUG_SF) {
+   if (unlikely(INTEL_DEBUG & DEBUG_SF)) {
       printf("sf:\n");
       for (i = 0; i < program_size / sizeof(struct brw_instruction); i++)
 	 brw_disasm(stdout, &((struct brw_instruction *)program)[i],
diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c
index 6ad9e1b48a4..bd3a21ed9e2 100644
--- a/src/mesa/drivers/dri/i965/brw_sf_state.c
+++ b/src/mesa/drivers/dri/i965/brw_sf_state.c
@@ -210,10 +210,10 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
 
    sf.thread4.max_threads = MIN2(chipset_max_threads, key->nr_urb_entries) - 1;
 
-   if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
+   if (unlikely(INTEL_DEBUG & DEBUG_SINGLE_THREAD))
       sf.thread4.max_threads = 0;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (unlikely(INTEL_DEBUG & DEBUG_STATS))
       sf.thread4.stats_enable = 1;
 
    /* CACHE_NEW_SF_VP */
diff --git a/src/mesa/drivers/dri/i965/brw_state_cache.c b/src/mesa/drivers/dri/i965/brw_state_cache.c
index b31d84953a1..58ff528d44b 100644
--- a/src/mesa/drivers/dri/i965/brw_state_cache.c
+++ b/src/mesa/drivers/dri/i965/brw_state_cache.c
@@ -61,6 +61,7 @@
 #include "intel_batchbuffer.h"
 #include "brw_wm.h"
 
+#define FILE_DEBUG_FLAG DEBUG_STATE
 
 static GLuint
 hash_key(struct brw_cache_item *item)
@@ -265,10 +266,9 @@ brw_upload_cache_with_auxdata(struct brw_cache *cache,
       *(void **)aux_return = (void *)((char *)item->key + item->key_size);
    }
 
-   if (INTEL_DEBUG & DEBUG_STATE)
-      printf("upload %s: %d bytes to cache id %d\n",
-		   cache->name[cache_id],
-		   data_size, cache_id);
+   DBG("upload %s: %d bytes to cache id %d\n",
+       cache->name[cache_id],
+       data_size, cache_id);
 
    /* Copy data to the buffer */
    drm_intel_bo_subdata(bo, 0, data_size, data);
@@ -407,8 +407,7 @@ brw_clear_cache(struct brw_context *brw, struct brw_cache *cache)
    struct brw_cache_item *c, *next;
    GLuint i;
 
-   if (INTEL_DEBUG & DEBUG_STATE)
-      printf("%s\n", __FUNCTION__);
+   DBG("%s\n", __FUNCTION__);
 
    for (i = 0; i < cache->size; i++) {
       for (c = cache->items[i]; c; c = next) {
@@ -434,8 +433,7 @@ brw_clear_cache(struct brw_context *brw, struct brw_cache *cache)
 void
 brw_state_cache_check_size(struct brw_context *brw)
 {
-   if (INTEL_DEBUG & DEBUG_STATE)
-      printf("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items);
+   DBG("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items);
 
    /* un-tuned guess.  Each object is generally a page, so 1000 of them is 4 MB of
     * state cache.
@@ -450,8 +448,7 @@ brw_destroy_cache(struct brw_context *brw, struct brw_cache *cache)
 {
    GLuint i;
 
-   if (INTEL_DEBUG & DEBUG_STATE)
-      printf("%s\n", __FUNCTION__);
+   DBG("%s\n", __FUNCTION__);
 
    brw_clear_cache(brw, cache);
    for (i = 0; i < BRW_MAX_CACHE; i++) {
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index f3b6a90f61a..338f3876b31 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -435,7 +435,7 @@ void brw_upload_state(struct brw_context *brw)
 
    brw_clear_validated_bos(brw);
 
-   if (INTEL_DEBUG) {
+   if (unlikely(INTEL_DEBUG)) {
       /* Debug version which enforces various sanity checks on the
        * state flags which are generated and checked to help ensure
        * state atoms are ordered correctly in the list.
@@ -487,7 +487,7 @@ void brw_upload_state(struct brw_context *brw)
       }
    }
 
-   if (INTEL_DEBUG & DEBUG_STATE) {
+   if (unlikely(INTEL_DEBUG & DEBUG_STATE)) {
       brw_update_dirty_count(mesa_bits, state->mesa);
       brw_update_dirty_count(brw_bits, state->brw);
       brw_update_dirty_count(cache_bits, state->cache);
diff --git a/src/mesa/drivers/dri/i965/brw_urb.c b/src/mesa/drivers/dri/i965/brw_urb.c
index 0f597184b42..dfc1551aca6 100644
--- a/src/mesa/drivers/dri/i965/brw_urb.c
+++ b/src/mesa/drivers/dri/i965/brw_urb.c
@@ -190,12 +190,12 @@ static void recalculate_urb_fence( struct brw_context *brw )
 	    exit(1);
 	 }
 	 
-	 if (INTEL_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS))
+	 if (unlikely(INTEL_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS)))
 	    printf("URB CONSTRAINED\n");
       }
 
 done:
-      if (INTEL_DEBUG & DEBUG_URB)
+      if (unlikely(INTEL_DEBUG & DEBUG_URB))
 	 printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n",
 		      brw->urb.vs_start,
 		      brw->urb.gs_start,
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index e560cdac1f9..aa8048c9fc3 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -394,7 +394,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 
    c->prog_data.total_grf = reg;
 
-   if (INTEL_DEBUG & DEBUG_VS) {
+   if (unlikely(INTEL_DEBUG & DEBUG_VS)) {
       printf("%s NumAddrRegs %d\n", __FUNCTION__, c->vp->program.Base.NumAddressRegs);
       printf("%s NumTemps %d\n", __FUNCTION__, c->vp->program.Base.NumTemporaries);
       printf("%s reg = %d\n", __FUNCTION__, reg);
@@ -1678,7 +1678,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
    GLuint index;
    GLuint file;
 
-   if (INTEL_DEBUG & DEBUG_VS) {
+   if (unlikely(INTEL_DEBUG & DEBUG_VS)) {
       printf("vs-mesa:\n");
       _mesa_fprint_program_opt(stdout, &c->vp->program.Base, PROG_PRINT_DEBUG,
 			       GL_TRUE);
@@ -2048,7 +2048,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
 
    brw_optimize(p);
 
-   if (INTEL_DEBUG & DEBUG_VS) {
+   if (unlikely(INTEL_DEBUG & DEBUG_VS)) {
       int i;
 
       printf("vs-native:\n");
diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c
index ebae94269f9..be923138617 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_state.c
@@ -154,7 +154,7 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
     */
    vs.vs5.sampler_count = 0;
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (unlikely(INTEL_DEBUG & DEBUG_STATS))
       vs.thread4.stats_enable = 1;
 
    /* Vertex program always enabled:
diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c
index 7f3ba5f0581..a6d2a2377f6 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.c
+++ b/src/mesa/drivers/dri/i965/brw_wm.c
@@ -198,7 +198,7 @@ static void do_wm_prog( struct brw_context *brw,
       c->prog_data.total_scratch = 0;
    }
 
-   if (INTEL_DEBUG & DEBUG_WM)
+   if (unlikely(INTEL_DEBUG & DEBUG_WM))
       fprintf(stderr, "\n");
 
    /* get the program
diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c
index d06c49fd5be..272476219a2 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c
@@ -1920,7 +1920,7 @@ void brw_wm_emit( struct brw_wm_compile *c )
 	brw_remove_grf_to_mrf_moves(p);
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       int i;
 
      printf("wm-native:\n");
diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c
index 15a238cda62..938fb80857b 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_fp.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c
@@ -1058,7 +1058,7 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
    struct brw_fragment_program *fp = c->fp;
    GLuint insn;
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       printf("pre-fp:\n");
       _mesa_fprint_program_opt(stdout, &fp->program.Base, PROG_PRINT_DEBUG,
 			       GL_TRUE);
@@ -1174,7 +1174,7 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
       }
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       printf("pass_fp:\n");
       print_insns( c->prog_instructions, c->nr_fp_insns );
       printf("\n");
diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
index d325f85ce00..7fe8ab1f334 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
@@ -19,7 +19,7 @@ GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp)
 {
     int i;
 
-    if (INTEL_DEBUG & DEBUG_GLSL_FORCE)
+    if (unlikely(INTEL_DEBUG & DEBUG_GLSL_FORCE))
        return GL_TRUE;
 
     for (i = 0; i < fp->Base.NumInstructions; i++) {
@@ -1002,7 +1002,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
     }
     post_wm_emit(c);
 
-    if (INTEL_DEBUG & DEBUG_WM) {
+    if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       printf("wm-native:\n");
       for (i = 0; i < p->nr_insn; i++)
 	 brw_disasm(stdout, &p->store[i], intel->gen);
@@ -1016,7 +1016,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
  */
 void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c)
 {
-    if (INTEL_DEBUG & DEBUG_WM) {
+    if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
         printf("brw_wm_glsl_emit:\n");
     }
 
@@ -1026,7 +1026,7 @@ void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c)
     /* actual code generation */
     brw_wm_emit_glsl(brw, c);
 
-    if (INTEL_DEBUG & DEBUG_WM) {
+    if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
         brw_wm_print_program(c, "brw_wm_glsl_emit done");
     }
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass0.c b/src/mesa/drivers/dri/i965/brw_wm_pass0.c
index d6aa9f957a2..83152526b3a 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass0.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass0.c
@@ -440,7 +440,7 @@ void brw_wm_pass0( struct brw_wm_compile *c )
       }
    }
  
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       brw_wm_print_program(c, "pass0");
    }
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass1.c b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
index 962515a99e9..3a2874b6ddf 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass1.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
@@ -291,7 +291,7 @@ void brw_wm_pass1( struct brw_wm_compile *c )
       track_arg(c, inst, 2, read2);
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       brw_wm_print_program(c, "pass1");
    }
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass2.c b/src/mesa/drivers/dri/i965/brw_wm_pass2.c
index 54acb3038b5..44e39538145 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass2.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass2.c
@@ -331,13 +331,13 @@ void brw_wm_pass2( struct brw_wm_compile *c )
       }
    }
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
       brw_wm_print_program(c, "pass2");
    }
 
    c->state = PASS2_DONE;
 
-   if (INTEL_DEBUG & DEBUG_WM) {
+   if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
        brw_wm_print_program(c, "pass2/done");
    }
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c
index 9a27b937103..76de7b7b6f6 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_state.c
@@ -249,7 +249,7 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key,
 
    wm.wm5.line_stipple = key->line_stipple;
 
-   if (INTEL_DEBUG & DEBUG_STATS || key->stats_wm)
+   if (unlikely(INTEL_DEBUG & DEBUG_STATS) || key->stats_wm)
       wm.wm4.stats_enable = 1;
 
    bo = brw_upload_cache(&brw->cache, BRW_WM_UNIT,
diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
index aaf3a573cf7..4b498f8c5b2 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
@@ -97,7 +97,7 @@ do_flush_locked(struct intel_batchbuffer *batch, GLuint used)
 			(x_off & 0xffff) | (y_off << 16));
    }
 
-   if (INTEL_DEBUG & DEBUG_BATCH) {
+   if (unlikely(INTEL_DEBUG & DEBUG_BATCH)) {
       drm_intel_bo_map(batch->buf, GL_FALSE);
       intel_decode(batch->buf->virtual, used / 4, batch->buf->offset,
 		   intel->intelScreen->deviceID, GL_TRUE);
@@ -128,7 +128,7 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
    if (used == 0)
       return;
 
-   if (INTEL_DEBUG & DEBUG_BATCH)
+   if (unlikely(INTEL_DEBUG & DEBUG_BATCH))
       fprintf(stderr, "%s:%d: Batchbuffer flush with %db used\n", file, line,
 	      used);
 
@@ -172,7 +172,7 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
 
    do_flush_locked(batch, used);
 
-   if (INTEL_DEBUG & DEBUG_SYNC) {
+   if (unlikely(INTEL_DEBUG & DEBUG_SYNC)) {
       fprintf(stderr, "waiting for idle\n");
       drm_intel_bo_map(batch->buf, GL_TRUE);
       drm_intel_bo_unmap(batch->buf);
diff --git a/src/mesa/drivers/dri/intel/intel_clear.c b/src/mesa/drivers/dri/intel/intel_clear.c
index d7814635b72..fa451f0045e 100644
--- a/src/mesa/drivers/dri/intel/intel_clear.c
+++ b/src/mesa/drivers/dri/intel/intel_clear.c
@@ -58,6 +58,21 @@ static const char *buffer_names[] = {
    [BUFFER_COLOR7] = "color7",
 };
 
+static void
+debug_mask(const char *name, GLbitfield mask)
+{
+   GLuint i;
+
+   if (unlikely(INTEL_DEBUG & DEBUG_BLIT)) {
+      DBG("%s clear:", name);
+      for (i = 0; i < BUFFER_COUNT; i++) {
+	 if (mask & (1 << i))
+	    DBG(" %s", buffer_names[i]);
+      }
+      DBG("\n");
+   }
+}
+
 /**
  * Called by ctx->Driver.Clear.
  */
@@ -70,7 +85,6 @@ intelClear(struct gl_context *ctx, GLbitfield mask)
    GLbitfield blit_mask = 0;
    GLbitfield swrast_mask = 0;
    struct gl_framebuffer *fb = ctx->DrawBuffer;
-   GLuint i;
 
    if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_FRONT_RIGHT)) {
       intel->front_buffer_dirty = GL_TRUE;
@@ -162,39 +176,17 @@ intelClear(struct gl_context *ctx, GLbitfield mask)
    }
 
    if (blit_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("blit clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (blit_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
+      debug_mask("blit", blit_mask);
       intelClearWithBlit(ctx, blit_mask);
    }
 
    if (tri_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("tri clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (tri_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
-
+      debug_mask("tri", tri_mask);
       _mesa_meta_Clear(&intel->ctx, tri_mask);
    }
 
    if (swrast_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("swrast clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (swrast_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
+      debug_mask("swrast", swrast_mask);
       _swrast_Clear(ctx, swrast_mask);
    }
 }
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 7ace50bde97..de4bcde20b6 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -249,7 +249,7 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
     * thus ignore the invalidate. */
    drawable->lastStamp = drawable->dri2.stamp;
 
-   if (INTEL_DEBUG & DEBUG_DRI)
+   if (unlikely(INTEL_DEBUG & DEBUG_DRI))
       fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable);
 
    screen = intel->intelScreen->driScrnPriv;
@@ -378,14 +378,14 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
        if (rb->region && rb->region->name == buffers[i].name)
 	     continue;
 
-       if (INTEL_DEBUG & DEBUG_DRI)
+       if (unlikely(INTEL_DEBUG & DEBUG_DRI))
 	  fprintf(stderr,
 		  "attaching buffer %d, at %d, cpp %d, pitch %d\n",
 		  buffers[i].name, buffers[i].attachment,
 		  buffers[i].cpp, buffers[i].pitch);
        
        if (buffers[i].attachment == __DRI_BUFFER_STENCIL && depth_region) {
-	  if (INTEL_DEBUG & DEBUG_DRI)
+	  if (unlikely(INTEL_DEBUG & DEBUG_DRI))
 	     fprintf(stderr, "(reusing depth buffer as stencil)\n");
 	  intel_region_reference(&region, depth_region);
        }
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index 46d10d74ba3..2ec65b83026 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -98,6 +98,16 @@ extern void intelFallback(struct intel_context *intel, GLbitfield bit,
 
 #define INTEL_MAX_FIXUP 64
 
+#ifndef likely
+#ifdef __GNUC__
+#define likely(expr) (__builtin_expect(expr, 1))
+#define unlikely(expr) (__builtin_expect(expr, 0))
+#else
+#define likely(expr) (expr)
+#define unlikely(expr) (expr)
+#endif
+#endif
+
 struct intel_sync_object {
    struct gl_sync_object Base;
 
@@ -359,10 +369,15 @@ extern int INTEL_DEBUG;
 #define DEBUG_CLIP      0x8000000
 
 #define DBG(...) do {						\
-	if (INTEL_DEBUG & FILE_DEBUG_FLAG)			\
+	if (unlikely(INTEL_DEBUG & FILE_DEBUG_FLAG))		\
 		printf(__VA_ARGS__);			\
 } while(0)
 
+#define fallback_debug(...) do {				\
+	if (unlikely(INTEL_DEBUG & DEBUG_FALLBACKS))		\
+		printf(__VA_ARGS__);				\
+} while(0)
+
 #define PCI_CHIP_845_G			0x2562
 #define PCI_CHIP_I830_M			0x3577
 #define PCI_CHIP_I855_GM		0x3582
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c
index 60583ef4c0d..d5c35775ce4 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel.c
@@ -147,10 +147,9 @@ intel_check_blit_format(struct intel_region * region,
       return GL_TRUE;
    }
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s: bad format for blit (cpp %d, type %s format %s)\n",
-              __FUNCTION__, region->cpp,
-              _mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format));
+   DBG("%s: bad format for blit (cpp %d, type %s format %s)\n",
+       __FUNCTION__, region->cpp,
+       _mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format));
 
    return GL_FALSE;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index 63fb4b37b18..e7356a6da0d 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -113,9 +113,8 @@ static GLuint get_bitmap_rect(GLsizei width, GLsizei height,
    GLint incr;
    GLuint count = 0;
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      printf("%s %d,%d %dx%d bitmap %dx%d skip %d src_offset %d mask %d\n",
-		   __FUNCTION__, x,y,w,h,width,height,unpack->SkipPixels, src_offset, mask);
+   DBG("%s %d,%d %dx%d bitmap %dx%d skip %d src_offset %d mask %d\n",
+       __FUNCTION__, x,y,w,h,width,height,unpack->SkipPixels, src_offset, mask);
 
    if (invert) {
       first = h-1;
@@ -285,7 +284,7 @@ do_blit_bitmap( struct gl_context *ctx,
    }
 out:
 
-   if (INTEL_DEBUG & DEBUG_SYNC)
+   if (unlikely(INTEL_DEBUG & DEBUG_SYNC))
       intel_batchbuffer_flush(intel->batch);
 
    if (_mesa_is_bufferobj(unpack->BufferObj)) {
@@ -299,6 +298,7 @@ out:
    return GL_TRUE;
 }
 
+
 /* There are a large number of possible ways to implement bitmap on
  * this hardware, most of them have some sort of drawback.  Here are a
  * few that spring to mind:
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_copy.c b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
index c6b36ed4291..a7ca780e944 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
@@ -119,8 +119,7 @@ do_blit_copypixels(struct gl_context * ctx,
    GLboolean flip = GL_FALSE;
 
    if (type == GL_DEPTH || type == GL_STENCIL) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "glCopyPixels() fallback: GL_DEPTH || GL_STENCIL\n");
+      fallback_debug("glCopyPixels() fallback: GL_DEPTH || GL_STENCIL\n");
       return GL_FALSE;
    }
 
@@ -203,8 +202,7 @@ intelCopyPixels(struct gl_context * ctx,
                 GLsizei width, GLsizei height,
                 GLint destx, GLint desty, GLenum type)
 {
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   DBG("%s\n", __FUNCTION__);
 
    if (do_blit_copypixels(ctx, srcx, srcy, width, height, destx, desty, type))
       return;
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_read.c b/src/mesa/drivers/dri/intel/intel_pixel_read.c
index b249f9a5a0b..54da29236d2 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_read.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_read.c
@@ -42,6 +42,8 @@
 #include "intel_pixel.h"
 #include "intel_buffer_objects.h"
 
+#define FILE_DEBUG_FLAG DEBUG_PIXEL
+
 /* For many applications, the new ability to pull the source buffers
  * back out of the GTT and then do the packing/conversion operations
  * in software will be as much of an improvement as trying to get the
@@ -79,8 +81,7 @@ do_blit_readpixels(struct gl_context * ctx,
    GLboolean all;
    GLint dst_x, dst_y;
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      printf("%s\n", __FUNCTION__);
+   DBG("%s\n", __FUNCTION__);
 
    if (!src)
       return GL_FALSE;
@@ -88,22 +89,19 @@ do_blit_readpixels(struct gl_context * ctx,
    if (!_mesa_is_bufferobj(pack->BufferObj)) {
       /* PBO only for now:
        */
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         printf("%s - not PBO\n", __FUNCTION__);
+      DBG("%s - not PBO\n", __FUNCTION__);
       return GL_FALSE;
    }
 
 
    if (ctx->_ImageTransferState ||
        !intel_check_blit_format(src, format, type)) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         printf("%s - bad format for blit\n", __FUNCTION__);
+      DBG("%s - bad format for blit\n", __FUNCTION__);
       return GL_FALSE;
    }
 
    if (pack->Alignment != 1 || pack->SwapBytes || pack->LsbFirst) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         printf("%s: bad packing params\n", __FUNCTION__);
+      DBG("%s: bad packing params\n", __FUNCTION__);
       return GL_FALSE;
    }
 
@@ -113,8 +111,7 @@ do_blit_readpixels(struct gl_context * ctx,
       rowLength = width;
 
    if (pack->Invert) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         printf("%s: MESA_PACK_INVERT not done yet\n", __FUNCTION__);
+      DBG("%s: MESA_PACK_INVERT not done yet\n", __FUNCTION__);
       return GL_FALSE;
    }
    else {
@@ -158,8 +155,7 @@ do_blit_readpixels(struct gl_context * ctx,
       return GL_FALSE;
    }
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      printf("%s - DONE\n", __FUNCTION__);
+   DBG("%s - DONE\n", __FUNCTION__);
 
    return GL_TRUE;
 }
@@ -173,8 +169,7 @@ intelReadPixels(struct gl_context * ctx,
    struct intel_context *intel = intel_context(ctx);
    GLboolean dirty;
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   DBG("%s\n", __FUNCTION__);
 
    intel_flush(ctx);
 
@@ -188,8 +183,7 @@ intelReadPixels(struct gl_context * ctx,
        (ctx, x, y, width, height, format, type, pack, pixels))
       return;
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      printf("%s: fallback to swrast\n", __FUNCTION__);
+   fallback_debug("%s: fallback to swrast\n", __FUNCTION__);
 
    /* Update Mesa state before calling down into _swrast_ReadPixels, as
     * the spans code requires the computed buffer states to be up to date,
diff --git a/src/mesa/drivers/dri/intel/intel_tex.c b/src/mesa/drivers/dri/intel/intel_tex.c
index 3d9a2549db0..743a0432e43 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.c
+++ b/src/mesa/drivers/dri/intel/intel_tex.c
@@ -158,8 +158,7 @@ intelGenerateMipmap(struct gl_context *ctx, GLenum target,
       struct intel_context *intel = intel_context(ctx);
       struct intel_texture_object *intelObj = intel_texture_object(texObj);
 
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__);
+      fallback_debug("%s - fallback to swrast\n", __FUNCTION__);
 
       intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel);
       _mesa_generate_mipmap(ctx, target, texObj);
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index 2d046fd52d9..284ba19e8a3 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -105,16 +105,15 @@ do_copy_texsubimage(struct intel_context *intel,
    const struct intel_region *src = get_teximage_source(intel, internalFormat);
 
    if (!intelImage->mt || !src || !src->buffer) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+      if (unlikely(INTEL_DEBUG & DEBUG_FALLBACKS))
 	 fprintf(stderr, "%s fail %p %p (0x%08x)\n",
 		 __FUNCTION__, intelImage->mt, src, internalFormat);
       return GL_FALSE;
    }
 
    if (intelImage->mt->cpp != src->cpp) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s fail %d vs %d cpp\n",
-		 __FUNCTION__, intelImage->mt->cpp, src->cpp);
+      fallback_debug("%s fail %d vs %d cpp\n",
+		     __FUNCTION__, intelImage->mt->cpp, src->cpp);
       return GL_FALSE;
    }
 
@@ -212,8 +211,7 @@ intelCopyTexImage1D(struct gl_context * ctx, GLenum target, GLint level,
    return;
 
  fail:
-   if (INTEL_DEBUG & DEBUG_FALLBACKS)
-      fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__);
+   fallback_debug("%s - fallback to swrast\n", __FUNCTION__);
    _mesa_meta_CopyTexImage1D(ctx, target, level, internalFormat, x, y,
                              width, border);
 }
@@ -261,8 +259,7 @@ intelCopyTexImage2D(struct gl_context * ctx, GLenum target, GLint level,
    return;
 
  fail:
-   if (INTEL_DEBUG & DEBUG_FALLBACKS)
-      fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__);
+   fallback_debug("%s - fallback to swrast\n", __FUNCTION__);
    _mesa_meta_CopyTexImage2D(ctx, target, level, internalFormat, x, y,
                              width, height, border);
 }
@@ -287,8 +284,7 @@ intelCopyTexSubImage1D(struct gl_context * ctx, GLenum target, GLint level,
    if (!do_copy_texsubimage(intel_context(ctx), target,
                             intel_texture_image(texImage),
                             internalFormat, xoffset, 0, x, y, width, 1)) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-         fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__);
+      fallback_debug("%s - fallback to swrast\n", __FUNCTION__);
       _mesa_meta_CopyTexSubImage1D(ctx, target, level, xoffset, x, y, width);
    }
 }
@@ -314,8 +310,7 @@ intelCopyTexSubImage2D(struct gl_context * ctx, GLenum target, GLint level,
                             internalFormat,
                             xoffset, yoffset, x, y, width, height)) {
 
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-         fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__);
+      fallback_debug("%s - fallback to swrast\n", __FUNCTION__);
       _mesa_meta_CopyTexSubImage2D(ctx, target, level,
                                    xoffset, yoffset, x, y, width, height);
    }