16 files changed, 79 insertions, 59 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_clip_state.c b/src/mesa/drivers/dri/i965/brw_clip_state.c
index 28ed6346314..c8f24a94e4d 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_state.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_state.c
@@ -131,7 +131,7 @@ clip_unit_create_from_key(struct brw_context *brw,
    clip.clip5.api_mode = BRW_CLIP_API_OGL;
    clip.clip5.clip_mode = key->clip_mode;
 
-   if (BRW_IS_G4X(brw))
+   if (intel->is_g4x)
       clip.clip5.negative_w_clip_test = 1;
 
    clip.clip6.clipper_viewport_state_ptr = 0;
diff --git a/src/mesa/drivers/dri/i965/brw_clip_util.c b/src/mesa/drivers/dri/i965/brw_clip_util.c
index 091ebfc34cd..86fed59fa4b 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_util.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_util.c
@@ -135,6 +135,7 @@ 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;
 
@@ -142,7 +143,7 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
     */
    /*
     * After CLIP stage, only first 256 bits of the VUE are read
-    * back on IGDNG, so needn't change it
+    * back on Ironlake, so needn't change it
     */
    brw_copy_indirect_to_indirect(p, dest_ptr, v0_ptr, 1);
       
@@ -151,7 +152,7 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
    for (i = 0; i < c->nr_attrs; i++) {
       GLuint delta = i*16 + 32;
 
-      if (BRW_IS_IGDNG(p->brw))
+      if (intel->is_ironlake)
           delta = i * 16 + 32 * 3;
 
       if (delta == c->offset[VERT_RESULT_EDGE]) {
@@ -185,7 +186,7 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
    if (i & 1) {
       GLuint delta = i*16 + 32;
 
-      if (BRW_IS_IGDNG(p->brw))
+      if (intel->is_ironlake)
           delta = i * 16 + 32 * 3;
 
       brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(0));
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index 78bea829493..282566f8d52 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -158,6 +158,38 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis,
       MIN2(ctx->Const.FragmentProgram.MaxNativeParameters,
 	   ctx->Const.FragmentProgram.MaxEnvParams);
 
+   if (intel->is_ironlake || intel->is_g4x) {
+      brw->CMD_VF_STATISTICS = CMD_VF_STATISTICS_GM45;
+      brw->CMD_PIPELINE_SELECT = CMD_PIPELINE_SELECT_GM45;
+      brw->has_surface_tile_offset = GL_TRUE;
+      brw->has_compr4 = GL_TRUE;
+  } else {
+      brw->CMD_VF_STATISTICS = CMD_VF_STATISTICS_965;
+      brw->CMD_PIPELINE_SELECT = CMD_PIPELINE_SELECT_965;
+      brw->has_surface_tile_offset = GL_FALSE;
+      brw->has_compr4 = GL_FALSE;
+   }
+
+   /* WM maximum threads is number of EUs times number of threads per EU. */
+   if (intel->is_ironlake) {
+      brw->urb.size = 1024;
+      brw->vs_max_threads = 72;
+      brw->wm_max_threads = 12 * 6;
+   } else if (intel->is_g4x) {
+      brw->urb.size = 384;
+      brw->vs_max_threads = 32;
+      brw->wm_max_threads = 10 * 5;
+   } else {
+      brw->urb.size = 256;
+      brw->vs_max_threads = 16;
+      brw->wm_max_threads = 8 * 4;
+   }
+
+   if (INTEL_DEBUG & DEBUG_SINGLE_THREAD) {
+      brw->vs_max_threads = 1;
+      brw->wm_max_threads = 1;
+   }
+
    brw_init_state( brw );
 
    brw->state.dirty.mesa = ~0;
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index e73e21433c5..4ae48bba056 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -438,7 +438,9 @@ struct brw_context
    GLuint primitive;
 
    GLboolean emit_state_always;
-
+   GLboolean has_surface_tile_offset;
+   GLboolean has_compr4;
+;
    struct {
       struct brw_state_flags dirty;
 
@@ -514,6 +516,12 @@ struct brw_context
     */
    GLuint next_free_page;
 
+   /* hw-dependent 3DSTATE_VF_STATISTICS opcode */
+   uint32_t CMD_VF_STATISTICS;
+   /* hw-dependent 3DSTATE_PIPELINE_SELECT opcode */
+   uint32_t CMD_PIPELINE_SELECT;
+   int vs_max_threads;
+   int wm_max_threads;
 
    /* BRW_NEW_URB_ALLOCATIONS:
     */
@@ -541,6 +549,7 @@ struct brw_context
       GLuint clip_start;
       GLuint sf_start;
       GLuint cs_start;
+      GLuint size; /* Hardware URB size, in KB. */
    } urb;
 
    
diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
index c19510bbd42..2ee331a200e 100644
--- a/src/mesa/drivers/dri/i965/brw_defines.h
+++ b/src/mesa/drivers/dri/i965/brw_defines.h
@@ -832,12 +832,7 @@
 
 #include "intel_chipset.h"
 
-#define BRW_IS_G4X(brw)         (IS_G4X((brw)->intel.intelScreen->deviceID))
-#define BRW_IS_IGDNG(brw)         (IS_IGDNG((brw)->intel.intelScreen->deviceID))
-#define BRW_IS_965(brw)         (!(BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)))
-#define CMD_PIPELINE_SELECT(brw)        ((BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? CMD_PIPELINE_SELECT_GM45 : CMD_PIPELINE_SELECT_965)
-#define CMD_VF_STATISTICS(brw)          ((BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? CMD_VF_STATISTICS_GM45 : CMD_VF_STATISTICS_965)
-#define URB_SIZES(brw)                  (BRW_IS_IGDNG(brw) ? 1024 : \
-                                         (BRW_IS_G4X(brw) ? 384 : 256))  /* 512 bit units */
+#define BRW_IS_965(brw)         (!(brw)->intel.is_g4x && \
+				 !(brw)->intel.is_ironlake)
 
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index 67673e99115..3413c969281 100644
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -452,7 +452,7 @@ static void brw_set_sampler_message(struct brw_context *brw,
       insn->bits3.sampler_igdng.end_of_thread = eot;
       insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_SAMPLER;
       insn->bits2.send_igdng.end_of_thread = eot;
-   } else if (BRW_IS_G4X(brw)) {
+   } else if (intel->is_g4x) {
       insn->bits3.sampler_g4x.binding_table_index = binding_table_index;
       insn->bits3.sampler_g4x.sampler = sampler;
       insn->bits3.sampler_g4x.msg_type = msg_type;
diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c
index 562a17844b4..fe5c1ae2791 100644
--- a/src/mesa/drivers/dri/i965/brw_fallback.c
+++ b/src/mesa/drivers/dri/i965/brw_fallback.c
@@ -47,7 +47,6 @@
 
 static GLboolean do_check_fallback(struct brw_context *brw)
 {
-   struct intel_context *intel = &brw->intel;
    GLcontext *ctx = &brw->intel.ctx;
    GLuint i;
 
@@ -86,8 +85,7 @@ static GLboolean do_check_fallback(struct brw_context *brw)
    }
 
    /* _NEW_BUFFERS */
-   if (IS_965(intel->intelScreen->deviceID) &&
-       !IS_G4X(intel->intelScreen->deviceID)) {
+   if (!brw->has_surface_tile_offset) {
       for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
 	 struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i];
 	 struct intel_renderbuffer *irb = intel_renderbuffer(rb);
diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c
index 6cb2da28e6a..fd14eaa6128 100644
--- a/src/mesa/drivers/dri/i965/brw_misc_state.c
+++ b/src/mesa/drivers/dri/i965/brw_misc_state.c
@@ -212,7 +212,7 @@ static void emit_depthbuffer(struct brw_context *brw)
 {
    struct intel_context *intel = &brw->intel;
    struct intel_region *region = brw->state.depth_region;
-   unsigned int len = (BRW_IS_G4X(brw) || intel->is_ironlake) ? 6 : 5;
+   unsigned int len = (intel->is_g4x || intel->is_ironlake) ? 6 : 5;
 
    if (region == NULL) {
       BEGIN_BATCH(len, IGNORE_CLIPRECTS);
@@ -223,7 +223,7 @@ static void emit_depthbuffer(struct brw_context *brw)
       OUT_BATCH(0);
       OUT_BATCH(0);
 
-      if (BRW_IS_G4X(brw) || intel->is_ironlake)
+      if (intel->is_g4x || intel->is_ironlake)
          OUT_BATCH(0);
 
       ADVANCE_BATCH();
@@ -262,7 +262,7 @@ static void emit_depthbuffer(struct brw_context *brw)
 		((region->height - 1) << 19));
       OUT_BATCH(0);
 
-      if (BRW_IS_G4X(brw) || intel->is_ironlake)
+      if (intel->is_g4x || intel->is_ironlake)
          OUT_BATCH(0);
 
       ADVANCE_BATCH();
@@ -444,7 +444,7 @@ static void upload_invarient_state( struct brw_context *brw )
       struct brw_pipeline_select ps;
 
       memset(&ps, 0, sizeof(ps));
-      ps.header.opcode = CMD_PIPELINE_SELECT(brw);
+      ps.header.opcode = brw->CMD_PIPELINE_SELECT;
       ps.header.pipeline_select = 0;
       BRW_BATCH_STRUCT(brw, &ps);
    }
@@ -480,7 +480,7 @@ static void upload_invarient_state( struct brw_context *brw )
       struct brw_vf_statistics vfs;
       memset(&vfs, 0, sizeof(vfs));
 
-      vfs.opcode = CMD_VF_STATISTICS(brw);
+      vfs.opcode = brw->CMD_VF_STATISTICS;
       if (INTEL_DEBUG & DEBUG_STATS)
 	 vfs.statistics_enable = 1; 
 
diff --git a/src/mesa/drivers/dri/i965/brw_urb.c b/src/mesa/drivers/dri/i965/brw_urb.c
index e1e8a90ca77..f2cdb203b85 100644
--- a/src/mesa/drivers/dri/i965/brw_urb.c
+++ b/src/mesa/drivers/dri/i965/brw_urb.c
@@ -105,7 +105,8 @@ static GLboolean check_urb_layout( struct brw_context *brw )
    brw->urb.sf_start = brw->urb.clip_start + brw->urb.nr_clip_entries * brw->urb.vsize;
    brw->urb.cs_start = brw->urb.sf_start + brw->urb.nr_sf_entries * brw->urb.sfsize;
 
-   return brw->urb.cs_start + brw->urb.nr_cs_entries * brw->urb.csize <= URB_SIZES(brw);
+   return brw->urb.cs_start + brw->urb.nr_cs_entries *
+      brw->urb.csize <= brw->urb.size;
 }
 
 /* Most minimal update, forces re-emit of URB fence packet after GS
@@ -157,7 +158,7 @@ static void recalculate_urb_fence( struct brw_context *brw )
             brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries;
             brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries;
          }
-      } else if (BRW_IS_G4X(brw)) {
+      } else if (intel->is_g4x) {
 	 brw->urb.nr_vs_entries = 64;
 	 if (check_urb_layout(brw)) {
 	    goto done;
@@ -201,7 +202,7 @@ done:
 		      brw->urb.clip_start,
 		      brw->urb.sf_start,
 		      brw->urb.cs_start, 
-		      URB_SIZES(brw));
+		      brw->urb.size);
       
       brw->state.dirty.brw |= BRW_NEW_URB_FENCE;
    }
@@ -245,7 +246,7 @@ void brw_upload_urb_fence(struct brw_context *brw)
    uf.bits0.gs_fence  = brw->urb.clip_start; 
    uf.bits0.clp_fence = brw->urb.sf_start; 
    uf.bits1.sf_fence  = brw->urb.cs_start; 
-   uf.bits1.cs_fence  = URB_SIZES(brw);
+   uf.bits1.cs_fence  = brw->urb.size;
 
    BRW_BATCH_STRUCT(brw, &uf);
 }
diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c
index fdc55aa1b20..345ffa7ee11 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_state.c
@@ -85,7 +85,6 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
    struct intel_context *intel = &brw->intel;
    struct brw_vs_unit_state vs;
    dri_bo *bo;
-   int chipset_max_threads;
 
    memset(&vs, 0, sizeof(vs));
 
@@ -136,7 +135,7 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
       case 32:
 	 break;
       case 64:
-	 assert(BRW_IS_G4X(brw));
+	 assert(intel->is_g4x);
 	 break;
       default:
 	 assert(0);
@@ -146,17 +145,8 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
 
    vs.thread4.urb_entry_allocation_size = key->urb_size - 1;
 
-   if (intel->is_ironlake)
-      chipset_max_threads = 72;
-   else if (BRW_IS_G4X(brw))
-      chipset_max_threads = 32;
-   else
-      chipset_max_threads = 16;
    vs.thread4.max_threads = CLAMP(key->nr_urb_entries / 2,
-				  1, chipset_max_threads) - 1;
-
-   if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
-      vs.thread4.max_threads = 0;
+				  1, brw->vs_max_threads) - 1;
 
    /* No samplers for ARB_vp programs:
     */
diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c
index bcaa793de16..7811689d6ad 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c
@@ -1162,7 +1162,6 @@ void emit_fb_write(struct brw_wm_compile *c,
 {
    struct brw_compile *p = &c->func;
    struct brw_context *brw = p->brw;
-   struct intel_context *intel = &brw->intel;
    GLuint nr = 2;
    GLuint channel;
 
@@ -1177,7 +1176,7 @@ void emit_fb_write(struct brw_wm_compile *c,
    brw_push_insn_state(p);
 
    for (channel = 0; channel < 4; channel++) {
-      if (c->dispatch_width == 16 && (BRW_IS_G4X(brw) || intel->is_ironlake)) {
+      if (c->dispatch_width == 16 && brw->has_compr4) {
 	 /* By setting the high bit of the MRF register number, we indicate
 	  * that we want COMPR4 mode - instead of doing the usual destination
 	  * + 1 for the second half we get destination + 4.
diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c
index fe795fadf86..d3373ea79e8 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_state.c
@@ -49,8 +49,6 @@ struct brw_wm_unit_key {
    unsigned int curbe_offset;
    unsigned int urb_size;
 
-   unsigned int max_threads;
-
    unsigned int nr_surfaces, sampler_count;
    GLboolean uses_depth, computes_depth, uses_kill, is_glsl;
    GLboolean polygon_stipple, stats_wm, line_stipple, offset_enable;
@@ -67,18 +65,6 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key)
 
    memset(key, 0, sizeof(*key));
 
-   if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
-      key->max_threads = 1;
-   else {
-      /* WM maximum threads is number of EUs times number of threads per EU. */
-      if (intel->is_ironlake)
-         key->max_threads = 12 * 6;
-      else if (BRW_IS_G4X(brw))
-	 key->max_threads = 10 * 5;
-      else
-	 key->max_threads = 8 * 4;
-   }
-
    /* CACHE_NEW_WM_PROG */
    key->total_grf = brw->wm.prog_data->total_grf;
    key->urb_entry_read_length = brw->wm.prog_data->urb_read_length;
@@ -192,7 +178,7 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key,
    else
       wm.wm5.enable_16_pix = 1;
 
-   wm.wm5.max_threads = key->max_threads - 1;
+   wm.wm5.max_threads = brw->wm_max_threads - 1;
    wm.wm5.thread_dispatch_enable = 1;	/* AKA: color_write */
    wm.wm5.legacy_line_rast = 0;
    wm.wm5.legacy_global_depth_bias = 0;
@@ -269,7 +255,7 @@ static void upload_wm_unit( struct brw_context *brw )
     */
    assert(key.total_scratch <= 12 * 1024);
    if (key.total_scratch) {
-      GLuint total = key.total_scratch * key.max_threads;
+      GLuint total = key.total_scratch * brw->wm_max_threads;
 
       if (brw->wm.scratch_bo && total > brw->wm.scratch_bo->size) {
 	 dri_bo_unreference(brw->wm.scratch_bo);
diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
index 8335e5a650e..8810f2a3808 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
@@ -608,8 +608,7 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 
 	 surf.ss1.base_addr = key.draw_offset - tile_offset;
 
-	 assert(BRW_IS_G4X(brw) || tile_offset == 0);
-	 if (BRW_IS_G4X(brw)) {
+	 if (brw->has_surface_tile_offset) {
 	    if (key.tiling == I915_TILING_X) {
 	       /* Note that the low bits of these fields are missing, so
 		* there's the possibility of getting in trouble.
@@ -620,6 +619,8 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 	       surf.ss5.x_offset = (tile_offset % 128) / key.cpp / 4;
 	       surf.ss5.y_offset = tile_offset / 128 / 2;
 	    }
+	 } else {
+	    assert(tile_offset == 0);
 	 }
       }
       if (region_bo != NULL)
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 234f99a00e5..6f567dc4d05 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -623,6 +623,10 @@ intelInitContext(struct intel_context *intel,
    if (IS_IGDNG(intel->intelScreen->deviceID)) {
       intel->is_ironlake = GL_TRUE;
       intel->needs_ff_sync = GL_TRUE;
+      intel->has_luminance_srgb = GL_TRUE;
+   } else if (IS_G4X(intel->intelScreen->deviceID)) {
+      intel->has_luminance_srgb = GL_TRUE;
+      intel->is_g4x = GL_TRUE;
    }
 
    /* Dri stuff */
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index f7d9016d71d..e46764e99cf 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -178,6 +178,10 @@ struct intel_context
    int gen;
    GLboolean needs_ff_sync;
    GLboolean is_ironlake;
+   GLboolean is_g4x;
+   GLboolean has_luminance_srgb;
+
+   int urb_size;
 
    struct intel_region *front_region;
    struct intel_region *back_region;
diff --git a/src/mesa/drivers/dri/intel/intel_tex_format.c b/src/mesa/drivers/dri/intel/intel_tex_format.c
index 87efb72cc51..a7c6c45ffe4 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_format.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_format.c
@@ -173,13 +173,13 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
       return MESA_FORMAT_SARGB8;
    case GL_SLUMINANCE_EXT:
    case GL_SLUMINANCE8_EXT:
-      if (IS_G4X(intel->intelScreen->deviceID))
+      if (intel->has_luminance_srgb)
          return MESA_FORMAT_SL8;
       else
          return MESA_FORMAT_SARGB8;
    case GL_SLUMINANCE_ALPHA_EXT:
    case GL_SLUMINANCE8_ALPHA8_EXT:
-      if (IS_G4X(intel->intelScreen->deviceID))
+      if (intel->has_luminance_srgb)
          return MESA_FORMAT_SLA8;
       else
          return MESA_FORMAT_SARGB8;