From 3445f476977ae403cef9ca15661fa0f96ff50eca Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sun, 25 Oct 2009 13:53:25 +0100
Subject: r300g: VBOs WIP

---
 src/gallium/drivers/r300/r300_emit.c | 108 +++++++++++++++++++++++++++++++++--
 1 file changed, 103 insertions(+), 5 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 69ce5966e8..92e6ec606c 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -582,7 +582,48 @@ void r300_emit_texture(struct r300_context* r300,
     END_CS;
 }
 
-void r300_emit_vertex_buffer(struct r300_context* r300)
+void r300_emit_aos(struct r300_context* r300, unsigned offset)
+{
+    struct pipe_vertex_buffer *vbuf = r300->vertex_buffer;
+    struct pipe_vertex_element *velem = r300->vertex_element;
+    CS_LOCALS(r300);
+    int i;
+    unsigned packet_size = (r300->aos_count * 3 + 1) / 2;
+    BEGIN_CS(2 + packet_size + r300->aos_count * 2);
+    OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, packet_size);
+    OUT_CS(r300->aos_count);
+    for (i = 0; i < r300->aos_count - 1; i += 2) {
+        int buf_num1 = velem[i].vertex_buffer_index;
+        int buf_num2 = velem[i+1].vertex_buffer_index;
+        assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0);
+        assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_size(velem[i+1].src_format) % 4 == 0);
+        OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) |
+               (pf_get_size(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22));
+        OUT_CS(vbuf[buf_num1].buffer_offset + velem[i].src_offset +
+               offset * vbuf[buf_num1].stride);
+        OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset +
+               offset * vbuf[buf_num2].stride);
+    }
+    if (r300->aos_count & 1) {
+        int buf_num = velem[i].vertex_buffer_index;
+        assert(vbuf[buf_num].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0);
+        OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6));
+        OUT_CS(vbuf[buf_num].buffer_offset + velem[i].src_offset +
+               offset * vbuf[buf_num].stride);
+    }
+
+    for (i = 0; i < r300->aos_count; i++) {
+        cs_winsys->write_cs_reloc(cs_winsys,
+                                  vbuf[velem[i].vertex_buffer_index].buffer,
+                                  RADEON_GEM_DOMAIN_GTT,
+                                  0,
+                                  0);
+        cs_count -= 2;
+    }
+    END_CS;
+}
+#if 0
+void r300_emit_draw_packet(struct r300_context* r300)
 {
     CS_LOCALS(r300);
 
@@ -605,6 +646,65 @@ void r300_emit_vertex_buffer(struct r300_context* r300)
     OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
     END_CS;
 }
+#endif
+void r300_emit_draw_arrays(struct r300_context *r300,
+                           unsigned count)
+{
+    CS_LOCALS(r300);
+    assert(count < 65536);
+
+    BEGIN_CS(4);
+    OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, count);
+    OUT_CS_PKT3(R300_PACKET3_3D_DRAW_VBUF_2, 0);
+    OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (count << 16) |
+           r300->hw_prim);
+    END_CS;
+}
+
+void r300_emit_draw_elements(struct r300_context *r300,
+                             struct pipe_buffer* indexBuffer,
+                             unsigned indexSize,
+                             unsigned minIndex,
+                             unsigned maxIndex,
+                             unsigned start,
+                             unsigned count)
+{
+    CS_LOCALS(r300);
+    assert(indexSize == 4 || indexSize == 2);
+    assert(count < 65536);
+    assert((start * indexSize)  % 4 == 0);
+
+    uint32_t size_dwords;
+    uint32_t skip_dwords = indexSize * start / sizeof(uint32_t);
+    assert(skip_dwords == 0);
+
+    BEGIN_CS(10);
+    OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, maxIndex);
+    OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, 0);
+    if (indexSize == 4) {
+        size_dwords = count + start;
+        OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) |
+               R300_VAP_VF_CNTL__INDEX_SIZE_32bit | r300->hw_prim);
+    } else {
+        size_dwords = (count + start + 1) / 2;
+        OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+               (count << 16) | r300->hw_prim);
+    }
+
+    OUT_CS_PKT3(R300_PACKET3_INDX_BUFFER, 2);
+    OUT_CS(R300_INDX_BUFFER_ONE_REG_WR | (R300_VAP_PORT_IDX0 >> 2) |
+           (0 << R300_INDX_BUFFER_SKIP_SHIFT));
+    OUT_CS(skip_dwords);
+    OUT_CS(size_dwords);
+    cs_winsys->write_cs_reloc(cs_winsys,
+                              indexBuffer,
+                              RADEON_GEM_DOMAIN_GTT,
+                              0,
+                              0);
+    cs_count -= 2;
+
+    END_CS;
+}
 
 void r300_emit_vertex_format_state(struct r300_context* r300)
 {
@@ -771,8 +871,6 @@ void r300_emit_dirty_state(struct r300_context* r300)
         return;
     }
 
-    r300_update_derived_state(r300);
-
     /* Clean out BOs. */
     r300->winsys->reset_bos(r300->winsys);
 
@@ -823,7 +921,7 @@ validate:
             goto validate;
         }
     } else {
-        debug_printf("No VBO while emitting dirty state!\n");
+        // debug_printf("No VBO while emitting dirty state!\n");
     }
     if (!r300->winsys->validate(r300->winsys)) {
         r300->context.flush(&r300->context, 0, NULL);
@@ -951,7 +1049,7 @@ validate:
     */
 
     /* Finally, emit the VBO. */
-    r300_emit_vertex_buffer(r300);
+    //r300_emit_vertex_buffer(r300);
 
     r300->dirty_hw++;
 }
-- 
cgit v1.2.3


From 96b729f926fafeca6479eed0933bc4275fb7843b Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sat, 7 Nov 2009 09:47:01 -0800
Subject: r300g: Don't pass hw_prim around in the context.

And some other fixes.
---
 src/gallium/drivers/r300/r300_context.h |  1 -
 src/gallium/drivers/r300/r300_emit.c    | 58 -------------------------
 src/gallium/drivers/r300/r300_emit.h    | 10 -----
 src/gallium/drivers/r300/r300_render.c  | 76 +++++++++++++++++++++++++++++----
 src/gallium/drivers/r300/r300_vbo.c     | 10 +++--
 5 files changed, 74 insertions(+), 81 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index a6748852d8..8d14c53f49 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -299,7 +299,6 @@ struct r300_context {
     /* Vertex elements for Gallium. */
     struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
     int aos_count;
-    unsigned hw_prim;
 
     /* Bitmask of dirty state objects. */
     uint32_t dirty_state;
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 92e6ec606c..ec1d521800 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -647,64 +647,6 @@ void r300_emit_draw_packet(struct r300_context* r300)
     END_CS;
 }
 #endif
-void r300_emit_draw_arrays(struct r300_context *r300,
-                           unsigned count)
-{
-    CS_LOCALS(r300);
-    assert(count < 65536);
-
-    BEGIN_CS(4);
-    OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, count);
-    OUT_CS_PKT3(R300_PACKET3_3D_DRAW_VBUF_2, 0);
-    OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (count << 16) |
-           r300->hw_prim);
-    END_CS;
-}
-
-void r300_emit_draw_elements(struct r300_context *r300,
-                             struct pipe_buffer* indexBuffer,
-                             unsigned indexSize,
-                             unsigned minIndex,
-                             unsigned maxIndex,
-                             unsigned start,
-                             unsigned count)
-{
-    CS_LOCALS(r300);
-    assert(indexSize == 4 || indexSize == 2);
-    assert(count < 65536);
-    assert((start * indexSize)  % 4 == 0);
-
-    uint32_t size_dwords;
-    uint32_t skip_dwords = indexSize * start / sizeof(uint32_t);
-    assert(skip_dwords == 0);
-
-    BEGIN_CS(10);
-    OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, maxIndex);
-    OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, 0);
-    if (indexSize == 4) {
-        size_dwords = count + start;
-        OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) |
-               R300_VAP_VF_CNTL__INDEX_SIZE_32bit | r300->hw_prim);
-    } else {
-        size_dwords = (count + start + 1) / 2;
-        OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
-               (count << 16) | r300->hw_prim);
-    }
-
-    OUT_CS_PKT3(R300_PACKET3_INDX_BUFFER, 2);
-    OUT_CS(R300_INDX_BUFFER_ONE_REG_WR | (R300_VAP_PORT_IDX0 >> 2) |
-           (0 << R300_INDX_BUFFER_SKIP_SHIFT));
-    OUT_CS(skip_dwords);
-    OUT_CS(size_dwords);
-    cs_winsys->write_cs_reloc(cs_winsys,
-                              indexBuffer,
-                              RADEON_GEM_DOMAIN_GTT,
-                              0,
-                              0);
-    cs_count -= 2;
-
-    END_CS;
-}
 
 void r300_emit_vertex_format_state(struct r300_context* r300)
 {
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index b4fdfecde0..7c83c5166d 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -40,16 +40,6 @@ void r300_emit_blend_color_state(struct r300_context* r300,
 void r300_emit_clip_state(struct r300_context* r300,
                           struct pipe_clip_state* clip);
 
-void r300_emit_draw_arrays(struct r300_context *r300, unsigned count);
-
-void r300_emit_draw_elements(struct r300_context *r300,
-                             struct pipe_buffer* indexBuffer,
-                             unsigned indexSize,
-                             unsigned minIndex,
-                             unsigned maxIndex,
-                             unsigned start,
-                             unsigned count);
-
 void r300_emit_dsa_state(struct r300_context* r300,
                          struct r300_dsa_state* dsa);
 
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index cbda30227d..6f7c645334 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -31,13 +31,13 @@
 #include "util/u_memory.h"
 #include "util/u_prim.h"
 
-#include "r300_vbo.h"
 #include "r300_cs.h"
 #include "r300_context.h"
 #include "r300_emit.h"
 #include "r300_reg.h"
 #include "r300_render.h"
 #include "r300_state_derived.h"
+#include "r300_vbo.h"
 
 /* r300_render: Vertex and index buffer primitive emission. */
 #define R300_MAX_VBO_SIZE  (1024 * 1024)
@@ -70,6 +70,70 @@ uint32_t r300_translate_primitive(unsigned prim)
     }
 }
 
+static void r300_emit_draw_arrays(struct r300_context *r300,
+                                  unsigned mode,
+                                  unsigned count)
+{
+    CS_LOCALS(r300);
+    assert(count < 65536);
+
+    BEGIN_CS(4);
+    OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, count);
+    OUT_CS_PKT3(R300_PACKET3_3D_DRAW_VBUF_2, 0);
+    OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (count << 16) |
+           r300_translate_primitive(mode));
+    END_CS;
+}
+
+static void r300_emit_draw_elements(struct r300_context *r300,
+                                    struct pipe_buffer* indexBuffer,
+                                    unsigned indexSize,
+                                    unsigned minIndex,
+                                    unsigned maxIndex,
+                                    unsigned mode,
+                                    unsigned start,
+                                    unsigned count)
+{
+    CS_LOCALS(r300);
+    assert(indexSize == 4 || indexSize == 2);
+    assert(count < 65536);
+    assert((start * indexSize)  % 4 == 0);
+
+    uint32_t size_dwords;
+    uint32_t skip_dwords = indexSize * start / sizeof(uint32_t);
+    assert(skip_dwords == 0);
+
+    BEGIN_CS(10);
+    OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, maxIndex);
+    OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, 0);
+    if (indexSize == 4) {
+        size_dwords = count + start;
+        OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) |
+               R300_VAP_VF_CNTL__INDEX_SIZE_32bit |
+               r300_translate_primitive(mode));
+    } else {
+        size_dwords = (count + start + 1) / 2;
+        OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) |
+               r300_translate_primitive(mode));
+    }
+
+    OUT_CS_PKT3(R300_PACKET3_INDX_BUFFER, 2);
+    OUT_CS(R300_INDX_BUFFER_ONE_REG_WR | (R300_VAP_PORT_IDX0 >> 2) |
+           (0 << R300_INDX_BUFFER_SKIP_SHIFT));
+    OUT_CS(skip_dwords);
+    OUT_CS(size_dwords);
+    /* XXX hax */
+    cs_winsys->write_cs_reloc(cs_winsys,
+                              indexBuffer,
+                              RADEON_GEM_DOMAIN_GTT,
+                              0,
+                              0);
+    cs_count -= 2;
+
+    END_CS;
+}
+
+
 static boolean setup_vertex_buffers(struct r300_context *r300)
 {
     unsigned vbuf_count = r300->aos_count;
@@ -123,14 +187,12 @@ boolean r300_draw_range_elements(struct pipe_context* pipe,
 
     setup_index_buffer(r300, indexBuffer, indexSize);
 
-    r300->hw_prim = r300_translate_primitive(mode);
-
     r300_emit_dirty_state(r300);
 
     r300_emit_aos(r300, 0);
 
     r300_emit_draw_elements(r300, indexBuffer, indexSize, minIndex, maxIndex,
-                            start, count);
+                            mode, start, count);
 
     return TRUE;
 }
@@ -159,13 +221,11 @@ boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode,
 
     setup_vertex_attributes(r300);
 
-    r300->hw_prim = r300_translate_primitive(mode);
-
     r300_emit_dirty_state(r300);
 
     r300_emit_aos(r300, start);
 
-    r300_emit_draw_arrays(r300, count);
+    r300_emit_draw_arrays(r300, mode, count);
 
     return TRUE;
 }
@@ -186,8 +246,8 @@ boolean r300_swtcl_draw_range_elements(struct pipe_context* pipe,
                                        unsigned count)
 {
     assert(0);
-    struct r300_context* r300 = r300_context(pipe);
 #if 0
+    struct r300_context* r300 = r300_context(pipe);
     int i;
 
     if (!u_trim_pipe_prim(mode, &count)) {
diff --git a/src/gallium/drivers/r300/r300_vbo.c b/src/gallium/drivers/r300/r300_vbo.c
index 37b5c9224f..ab6f5c5942 100644
--- a/src/gallium/drivers/r300/r300_vbo.c
+++ b/src/gallium/drivers/r300/r300_vbo.c
@@ -56,7 +56,8 @@ static INLINE void setup_vertex_attribute(struct r300_vertex_info *vinfo,
 static void finish_vertex_attribs_setup(struct r300_vertex_info *vinfo,
                                         unsigned attribs_num)
 {
-    uint32_t last_vec_bit = (attribs_num % 2 == 0) ? (R300_LAST_VEC << 16) : R300_LAST_VEC;
+    uint32_t last_vec_bit = (attribs_num % 2 == 0) ?
+        (R300_LAST_VEC << 16) : R300_LAST_VEC;
 
     assert(attribs_num > 0 && attribs_num <= 16);
     vinfo->vap_prog_stream_cntl[(attribs_num - 1) >> 1] |= last_vec_bit;
@@ -64,10 +65,11 @@ static void finish_vertex_attribs_setup(struct r300_vertex_info *vinfo,
 
 void setup_vertex_attributes(struct r300_context *r300)
 {
-    for (int i=0; i<r300->aos_count; i++)
-    {
-        struct pipe_vertex_element *vert_elem = &r300->vertex_element[i];
+    struct pipe_vertex_element *vert_elem;
+    int i;
 
+    for (i = 0; i < r300->aos_count; i++) {
+        vert_elem = &r300->vertex_element[i];
         setup_vertex_attribute(r300->vertex_info, vert_elem, i);
     }
 
-- 
cgit v1.2.3


From b6f93e2607f1bbc5b2f478f0a57d7786dd7d73a5 Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 8 Nov 2009 11:32:32 -0800
Subject: r300g: Enable PSC/RS dump with new debugging flags.

---
 src/gallium/drivers/r300/r300_emit.c | 28 ++++++++++++++++------------
 1 file changed, 16 insertions(+), 12 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index ec1d521800..b3d9db676a 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -510,6 +510,8 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
     CS_LOCALS(r300);
 
+    DBG(r300, DBG_DRAW, "r300: RS emit:\n");
+
     BEGIN_CS(21);
     if (r300screen->caps->is_r500) {
         OUT_CS_REG_SEQ(R500_RS_IP_0, 8);
@@ -518,7 +520,7 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     }
     for (i = 0; i < 8; i++) {
         OUT_CS(rs->ip[i]);
-        /* debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]); */
+        DBG(r300, DBG_DRAW, "    : ip %d: 0x%08x\n", i, rs->ip[i]);
     }
 
     OUT_CS_REG_SEQ(R300_RS_COUNT, 2);
@@ -532,11 +534,11 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     }
     for (i = 0; i < 8; i++) {
         OUT_CS(rs->inst[i]);
-        /* debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]); */
+        DBG(r300, DBG_DRAW, "    : inst %d: 0x%08x\n", i, rs->inst[i]);
     }
 
-    /* debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count,
-     *        rs->inst_count); */
+    DBG(r300, DBG_DRAW, "    : count: 0x%08x inst_count: 0x%08x\n",
+        rs->count, rs->inst_count);
 
     END_CS;
 }
@@ -653,6 +655,8 @@ void r300_emit_vertex_format_state(struct r300_context* r300)
     int i;
     CS_LOCALS(r300);
 
+    DBG(r300, DBG_DRAW, "r300: VAP/PSC emit:\n");
+
     BEGIN_CS(26);
     OUT_CS_REG(R300_VAP_VTX_SIZE, r300->vertex_info->vinfo.size);
 
@@ -662,22 +666,22 @@ void r300_emit_vertex_format_state(struct r300_context* r300)
     OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2);
     OUT_CS(r300->vertex_info->vinfo.hwfmt[2]);
     OUT_CS(r300->vertex_info->vinfo.hwfmt[3]);
-    /* for (i = 0; i < 4; i++) {
-     *    debug_printf("hwfmt%d: 0x%08x\n", i,
-     *            r300->vertex_info->vinfo.hwfmt[i]);
-     * } */
+    for (i = 0; i < 4; i++) {
+       DBG(r300, DBG_DRAW, "    : hwfmt%d: 0x%08x\n", i,
+               r300->vertex_info->vinfo.hwfmt[i]);
+    }
 
     OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0, 8);
     for (i = 0; i < 8; i++) {
         OUT_CS(r300->vertex_info->vap_prog_stream_cntl[i]);
-        /* debug_printf("prog_stream_cntl%d: 0x%08x\n", i,
-         *        r300->vertex_info->vap_prog_stream_cntl[i]); */
+        DBG(r300, DBG_DRAW, "    : prog_stream_cntl%d: 0x%08x\n", i,
+               r300->vertex_info->vap_prog_stream_cntl[i]);
     }
     OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0, 8);
     for (i = 0; i < 8; i++) {
         OUT_CS(r300->vertex_info->vap_prog_stream_cntl_ext[i]);
-        /* debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i,
-         *        r300->vertex_info->vap_prog_stream_cntl_ext[i]); */
+        DBG(r300, DBG_DRAW, "    : prog_stream_cntl_ext%d: 0x%08x\n", i,
+               r300->vertex_info->vap_prog_stream_cntl_ext[i]);
     }
     END_CS;
 }
-- 
cgit v1.2.3


From 11d9edf4c9c75d5a41fb0a1757441ad315330bea Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 8 Nov 2009 11:45:57 -0800
Subject: r300g: Unify context names for counts.

From the SW TCL fixups.
---
 src/gallium/drivers/r300/r300_context.c |  2 +-
 src/gallium/drivers/r300/r300_context.h |  4 ++--
 src/gallium/drivers/r300/r300_emit.c    | 16 ++++++++++------
 src/gallium/drivers/r300/r300_render.c  |  2 +-
 src/gallium/drivers/r300/r300_state.c   |  4 ++--
 src/gallium/drivers/r300/r300_vbo.c     |  5 +++--
 6 files changed, 19 insertions(+), 14 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 43d7ff3ed3..ae23329b83 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -158,6 +158,7 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     /* Open up the OQ BO. */
     r300->oqbo = screen->buffer_create(screen, 4096,
             PIPE_BUFFER_USAGE_VERTEX, 4096);
+    make_empty_list(&r300->query_list);
 
     r300_init_flush_functions(r300);
 
@@ -172,6 +173,5 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     r300->winsys->set_flush_cb(r300->winsys, r300_flush_cb, r300);
     r300->dirty_state = R300_NEW_KITCHEN_SINK;
     r300->dirty_hw++;
-    make_empty_list(&r300->query_list);
     return &r300->context;
 }
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 8d14c53f49..f954ba7f9a 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -295,10 +295,10 @@ struct r300_context {
 
     /* Vertex buffers for Gallium. */
     struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
-    int vbuf_count;
+    int vertex_buffer_count;
     /* Vertex elements for Gallium. */
     struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
-    int aos_count;
+    int vertex_element_count;
 
     /* Bitmask of dirty state objects. */
     uint32_t dirty_state;
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index b3d9db676a..eeb97a2d37 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -584,17 +584,20 @@ void r300_emit_texture(struct r300_context* r300,
     END_CS;
 }
 
+/* XXX I can't read this and that's not good */
 void r300_emit_aos(struct r300_context* r300, unsigned offset)
 {
     struct pipe_vertex_buffer *vbuf = r300->vertex_buffer;
     struct pipe_vertex_element *velem = r300->vertex_element;
     CS_LOCALS(r300);
     int i;
-    unsigned packet_size = (r300->aos_count * 3 + 1) / 2;
-    BEGIN_CS(2 + packet_size + r300->aos_count * 2);
+    unsigned aos_count = r300->vertex_element_count;
+
+    unsigned packet_size = (aos_count * 3 + 1) / 2;
+    BEGIN_CS(2 + packet_size + aos_count * 2);
     OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, packet_size);
-    OUT_CS(r300->aos_count);
-    for (i = 0; i < r300->aos_count - 1; i += 2) {
+    OUT_CS(aos_count);
+    for (i = 0; i < aos_count - 1; i += 2) {
         int buf_num1 = velem[i].vertex_buffer_index;
         int buf_num2 = velem[i+1].vertex_buffer_index;
         assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0);
@@ -606,7 +609,7 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset)
         OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset +
                offset * vbuf[buf_num2].stride);
     }
-    if (r300->aos_count & 1) {
+    if (aos_count & 1) {
         int buf_num = velem[i].vertex_buffer_index;
         assert(vbuf[buf_num].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0);
         OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6));
@@ -614,7 +617,8 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset)
                offset * vbuf[buf_num].stride);
     }
 
-    for (i = 0; i < r300->aos_count; i++) {
+    /* XXX bare CS reloc */
+    for (i = 0; i < aos_count; i++) {
         cs_winsys->write_cs_reloc(cs_winsys,
                                   vbuf[velem[i].vertex_buffer_index].buffer,
                                   RADEON_GEM_DOMAIN_GTT,
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index fa057324f8..1ff3e64b44 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -140,7 +140,7 @@ static boolean r300_setup_vertex_buffers(struct r300_context *r300)
     struct pipe_vertex_element *velem = r300->vertex_element;
 
 validate:
-    for (int i = 0; i < r300->aos_count; i++) {
+    for (int i = 0; i < r300->vertex_element_count; i++) {
         if (!r300->winsys->add_buffer(r300->winsys,
                 vbuf[velem[i].vertex_buffer_index].buffer,
             RADEON_GEM_DOMAIN_GTT, 0)) {
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index e0b85ab768..d1eced61db 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -668,7 +668,7 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe,
 
     memcpy(r300->vertex_buffer, buffers,
         sizeof(struct pipe_vertex_buffer) * count);
-    r300->vbuf_count = count;
+    r300->vertex_buffer_count = count;
 
     if (r300->draw) {
         draw_flush(r300->draw);
@@ -685,7 +685,7 @@ static void r300_set_vertex_elements(struct pipe_context* pipe,
     memcpy(r300->vertex_element,
            elements,
            sizeof(struct pipe_vertex_element) * count);
-    r300->aos_count = count;
+    r300->vertex_element_count = count;
 
     if (r300->draw) {
         draw_flush(r300->draw);
diff --git a/src/gallium/drivers/r300/r300_vbo.c b/src/gallium/drivers/r300/r300_vbo.c
index 5ad6b9c215..a6a159667a 100644
--- a/src/gallium/drivers/r300/r300_vbo.c
+++ b/src/gallium/drivers/r300/r300_vbo.c
@@ -71,12 +71,13 @@ void setup_vertex_attributes(struct r300_context *r300)
     struct pipe_vertex_element *vert_elem;
     int i;
 
-    for (i = 0; i < r300->aos_count; i++) {
+    for (i = 0; i < r300->vertex_element_count; i++) {
         vert_elem = &r300->vertex_element[i];
         setup_vertex_attribute(r300->vertex_info, vert_elem, i);
     }
 
-    finish_vertex_attribs_setup(r300->vertex_info, r300->aos_count);
+    finish_vertex_attribs_setup(r300->vertex_info,
+        r300->vertex_element_count);
 }
 
 static INLINE int get_buffer_offset(struct r300_context *r300,
-- 
cgit v1.2.3


From beea241374a91b8aab81db175b28e98c2b4835d9 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Thu, 19 Nov 2009 01:35:08 +0100
Subject: r300g: set better values in the R300_VAP_CNTL register

---
 src/gallium/drivers/r300/r300_emit.c | 33 ++++++++++++++++++++++++++++-----
 1 file changed, 28 insertions(+), 5 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index eeb97a2d37..2a8d32242b 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -690,12 +690,35 @@ void r300_emit_vertex_format_state(struct r300_context* r300)
     END_CS;
 }
 
+/* XXX This should probably go to util ... */
+/* Return the number of bits set in the given number. */
+static unsigned bitcount(unsigned n)
+{
+    unsigned bits;
+    for (bits = 0; n > 0; n = n >> 1) {
+        bits += n & 1;
+    }
+    return bits;
+}
+
+/* XXX ... and this one too. */
+#define MIN3(x, y, z) MIN2(MIN2(x, y), z)
+
 void r300_emit_vertex_program_code(struct r300_context* r300,
                                    struct r300_vertex_program_code* code)
 {
     int i;
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
     unsigned instruction_count = code->length / 4;
+
+    int vtx_mem_size = r300screen->caps->is_r500 ? 128 : 72;
+    int input_count = MAX2(bitcount(code->InputsRead), 1);
+    int output_count = MAX2(bitcount(code->OutputsWritten), 1);
+    int temp_count = MAX2(code->num_temporaries, 1);
+    int pvs_num_slots = MIN3(vtx_mem_size / input_count,
+                             vtx_mem_size / output_count, 10);
+    int pvs_num_controllers = MIN2(6, vtx_mem_size / temp_count);
+
     CS_LOCALS(r300);
 
     if (!r300screen->caps->has_tcl) {
@@ -708,8 +731,7 @@ void r300_emit_vertex_program_code(struct r300_context* r300,
     /* R300_VAP_PVS_CODE_CNTL_0
      * R300_VAP_PVS_CONST_CNTL
      * R300_VAP_PVS_CODE_CNTL_1
-     * See the r5xx docs for instructions on how to use these.
-     * XXX these could be optimized to select better values... */
+     * See the r5xx docs for instructions on how to use these. */
     OUT_CS_REG_SEQ(R300_VAP_PVS_CODE_CNTL_0, 3);
     OUT_CS(R300_PVS_FIRST_INST(0) |
             R300_PVS_XYZW_VALID_INST(instruction_count - 1) |
@@ -722,10 +744,11 @@ void r300_emit_vertex_program_code(struct r300_context* r300,
     for (i = 0; i < code->length; i++)
         OUT_CS(code->body.d[i]);
 
-    OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(10) |
-            R300_PVS_NUM_CNTLRS(5) |
+    OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(pvs_num_slots) |
+            R300_PVS_NUM_CNTLRS(pvs_num_controllers) |
             R300_PVS_NUM_FPUS(r300screen->caps->num_vert_fpus) |
-            R300_PVS_VF_MAX_VTX_NUM(12));
+            R300_PVS_VF_MAX_VTX_NUM(12) |
+            (r300screen->caps->is_r500 ? R500_TCL_STATE_OPTIMIZATION : 0));
     END_CS;
 }
 
-- 
cgit v1.2.3


From 6a3eb1f91b4ccd4ee7ac6b91505e0dfa476922d4 Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Fri, 20 Nov 2009 14:10:45 -0800
Subject: r300g: Use MAX3 and MIN3.

---
 src/gallium/drivers/r300/r300_emit.c          | 3 ---
 src/gallium/drivers/r300/r300_state_derived.c | 2 +-
 2 files changed, 1 insertion(+), 4 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 2a8d32242b..4cd5074379 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -701,9 +701,6 @@ static unsigned bitcount(unsigned n)
     return bits;
 }
 
-/* XXX ... and this one too. */
-#define MIN3(x, y, z) MIN2(MIN2(x, y), z)
-
 void r300_emit_vertex_program_code(struct r300_context* r300,
                                    struct r300_vertex_program_code* code)
 {
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index 7166694edf..b4d0eeaf8c 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -458,7 +458,7 @@ static void r300_update_rs_block(struct r300_context* r300,
     rs->count = (rs_tex_comp) | (col_count << R300_IC_COUNT_SHIFT) |
         R300_HIRES_EN;
 
-    rs->inst_count = MAX2(MAX2(col_count - 1, tex_count - 1), 0);
+    rs->inst_count = MAX3(col_count - 1, tex_count - 1, 0);
 }
 
 /* Update the vertex format. */
-- 
cgit v1.2.3


From 06ec216d191e160494dd0a922ab0395418a78402 Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Fri, 20 Nov 2009 14:10:59 -0800
Subject: r300g: Clean up bitcounting function.

I didn't see this in u_math; surely somebody else has this wheel
reinvented elsewhere.
---
 src/gallium/drivers/r300/r300_emit.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 4cd5074379..c50c989f01 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -690,14 +690,19 @@ void r300_emit_vertex_format_state(struct r300_context* r300)
     END_CS;
 }
 
-/* XXX This should probably go to util ... */
+/* XXX This should go to util ... */
 /* Return the number of bits set in the given number. */
 static unsigned bitcount(unsigned n)
 {
-    unsigned bits;
-    for (bits = 0; n > 0; n = n >> 1) {
-        bits += n & 1;
+    unsigned bits = 0;
+
+    while (n) {
+        if (n & 1) {
+            bits++;
+        }
+        n >>= 1;
     }
+
     return bits;
 }
 
@@ -714,7 +719,7 @@ void r300_emit_vertex_program_code(struct r300_context* r300,
     int temp_count = MAX2(code->num_temporaries, 1);
     int pvs_num_slots = MIN3(vtx_mem_size / input_count,
                              vtx_mem_size / output_count, 10);
-    int pvs_num_controllers = MIN2(6, vtx_mem_size / temp_count);
+    int pvs_num_controllers = MIN2(vtx_mem_size / temp_count, 6);
 
     CS_LOCALS(r300);
 
-- 
cgit v1.2.3


From f4041b37e2d305cff0a97eb836250e9f8b1840a8 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Sat, 14 Nov 2009 22:14:42 +0100
Subject: r300g: fix rectangle textures on r3xx

Adapted from Maciej Cencora's patch.
---
 src/gallium/drivers/r300/r300_emit.c  | 22 ++++++++++++++++++++--
 src/gallium/drivers/r300/r300_state.c |  8 ++++++++
 2 files changed, 28 insertions(+), 2 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index c50c989f01..0bdf58202f 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -129,7 +129,9 @@ static const float * get_shader_constant(
     struct rc_constant * constant,
     struct r300_constant_buffer * externals)
 {
-    static const float zero[4] = { 0.0, 0.0, 0.0, 0.0 };
+    static float vec[4] = { 0.0, 0.0, 0.0, 0.0 };
+    struct pipe_texture *tex;
+
     switch(constant->Type) {
         case RC_CONSTANT_EXTERNAL:
             return externals->constants[constant->u.External];
@@ -137,10 +139,26 @@ static const float * get_shader_constant(
         case RC_CONSTANT_IMMEDIATE:
             return constant->u.Immediate;
 
+        case RC_CONSTANT_STATE:
+            switch (constant->u.State[0])
+            {
+                /* R3xx-specific */
+                case RC_STATE_R300_TEXRECT_FACTOR:
+                    tex = &r300->textures[constant->u.State[1]]->tex;
+                    vec[0] = 1.0 / tex->width[0];
+                    vec[1] = 1.0 / tex->height[0];
+                    vec[2] = vec[3] = 1;
+                    break;
+
+                default:
+                    assert(0);
+            }
+            return vec;
+
         default:
             debug_printf("r300: Implementation error: Unhandled constant type %i\n",
                 constant->Type);
-            return zero;
+            return vec;
     }
 }
 
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index d1eced61db..00f10ffd73 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -571,6 +571,7 @@ static void r300_set_sampler_textures(struct pipe_context* pipe,
                                       struct pipe_texture** texture)
 {
     struct r300_context* r300 = r300_context(pipe);
+    boolean is_r500 = r300_screen(r300->context.screen)->caps->is_r500;
     int i;
 
     /* XXX magic num */
@@ -585,6 +586,13 @@ static void r300_set_sampler_textures(struct pipe_context* pipe,
             pipe_texture_reference((struct pipe_texture**)&r300->textures[i],
                 texture[i]);
             r300->dirty_state |= (R300_NEW_TEXTURE << i);
+
+            /* R300-specific - set the texrect factor in a fragment shader */
+            if (!is_r500 && r300->textures[i]->is_npot) {
+                /* XXX It would be nice to re-emit just 1 constant,
+                 * XXX not all of them */
+                r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
+            }
         }
     }
 
-- 
cgit v1.2.3


From 6a95996abb33a040f957ffedf3824afcc98a9e71 Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Fri, 20 Nov 2009 14:55:22 -0800
Subject: r300g: Texrect factor cleanup.

(0, 0, 0, 1) is a much saner default value, and texrect factors only need
to be (1/s, 1/t, 0, 1).
---
 src/gallium/drivers/r300/r300_emit.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 0bdf58202f..37e75ba061 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -129,7 +129,7 @@ static const float * get_shader_constant(
     struct rc_constant * constant,
     struct r300_constant_buffer * externals)
 {
-    static float vec[4] = { 0.0, 0.0, 0.0, 0.0 };
+    static float vec[4] = { 0.0, 0.0, 0.0, 1.0 };
     struct pipe_texture *tex;
 
     switch(constant->Type) {
@@ -140,26 +140,30 @@ static const float * get_shader_constant(
             return constant->u.Immediate;
 
         case RC_CONSTANT_STATE:
-            switch (constant->u.State[0])
-            {
-                /* R3xx-specific */
+            switch (constant->u.State[0]) {
+                /* Factor for converting rectangle coords to
+                 * normalized coords. Should only show up on non-r500. */
                 case RC_STATE_R300_TEXRECT_FACTOR:
                     tex = &r300->textures[constant->u.State[1]]->tex;
                     vec[0] = 1.0 / tex->width[0];
                     vec[1] = 1.0 / tex->height[0];
-                    vec[2] = vec[3] = 1;
                     break;
 
                 default:
-                    assert(0);
+                    debug_printf("r300: Implementation error: "
+                        "Unknown RC_CONSTANT type %d\n", constant->u.State[0]);
             }
-            return vec;
+            break;
 
         default:
-            debug_printf("r300: Implementation error: Unhandled constant type %i\n",
-                constant->Type);
-            return vec;
+            debug_printf("r300: Implementation error: "
+                "Unhandled constant type %d\n", constant->Type);
     }
+
+    /* This should either be (0, 0, 0, 1), which should be a relatively safe
+     * RGBA or STRQ value, or it could be one of the RC_CONSTANT_STATE
+     * state factors. */
+    return vec;
 }
 
 /* Convert a normal single-precision float into the 7.16 format
-- 
cgit v1.2.3


From 015e7e7724a64d3d9e02e57f6a8eb88a6441f596 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Fri, 20 Nov 2009 05:17:00 +0100
Subject: r300g: emit R300_TEX_ENABLE to indicate there are no textures bound

Previously, this reg wasn't emitted at all if texture_count == 0.
---
 src/gallium/drivers/r300/r300_emit.c | 15 +++++++++++++--
 src/gallium/drivers/r300/r300_emit.h |  2 ++
 2 files changed, 15 insertions(+), 2 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 37e75ba061..6d702c0027 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -837,13 +837,22 @@ void r300_emit_viewport_state(struct r300_context* r300,
     END_CS;
 }
 
+void r300_emit_texture_count(struct r300_context* r300)
+{
+    CS_LOCALS(r300);
+
+    BEGIN_CS(2);
+    OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1);
+    END_CS;
+
+}
+
 void r300_flush_textures(struct r300_context* r300)
 {
     CS_LOCALS(r300);
 
-    BEGIN_CS(4);
+    BEGIN_CS(2);
     OUT_CS_REG(R300_TX_INVALTAGS, 0);
-    OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1);
     END_CS;
 }
 
@@ -997,6 +1006,8 @@ validate:
     /* Samplers and textures are tracked separately but emitted together. */
     if (r300->dirty_state &
             (R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES)) {
+        r300_emit_texture_count(r300);
+
         for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) {
   	    if (r300->dirty_state &
 		((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) {
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 7c83c5166d..3797d3d332 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -92,6 +92,8 @@ void r300_emit_vertex_shader(struct r300_context* r300,
 void r300_emit_viewport_state(struct r300_context* r300,
                               struct r300_viewport_state* viewport);
 
+void r300_emit_texture_count(struct r300_context* r300);
+
 void r300_flush_textures(struct r300_context* r300);
 
 /* Emit all dirty state. */
-- 
cgit v1.2.3


From b7078a88119e248b0196f7446abe029c22f1ee28 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Sat, 14 Nov 2009 23:27:20 +0100
Subject: r300g: add texture lod clamping

These now work:
piglit/lodclamp
piglit/levelclamp
---
 src/gallium/drivers/r300/r300_context.h |  4 ++++
 src/gallium/drivers/r300/r300_emit.c    | 11 ++++++++++-
 src/gallium/drivers/r300/r300_reg.h     |  5 +++--
 src/gallium/drivers/r300/r300_state.c   |  5 +++++
 src/gallium/drivers/r300/r300_texture.c |  3 +--
 5 files changed, 23 insertions(+), 5 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index f954ba7f9a..60ef415caa 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -92,6 +92,10 @@ struct r300_sampler_state {
     uint32_t filter0;      /* R300_TX_FILTER0: 0x4400 */
     uint32_t filter1;      /* R300_TX_FILTER1: 0x4440 */
     uint32_t border_color; /* R300_TX_BORDER_COLOR: 0x45c0 */
+
+    /* Min/max LOD must be clamped to [0, last_level], thus
+     * it's dependent on a currently bound texture */
+    unsigned min_lod, max_lod;
 };
 
 struct r300_scissor_state {
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 6d702c0027..ad7dff36be 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -583,6 +583,8 @@ void r300_emit_texture(struct r300_context* r300,
                        unsigned offset)
 {
     uint32_t filter0 = sampler->filter0;
+    uint32_t format0 = tex->state.format0;
+    unsigned min_level, max_level;
     CS_LOCALS(r300);
 
     /* to emulate 1D textures through 2D ones correctly */
@@ -591,13 +593,20 @@ void r300_emit_texture(struct r300_context* r300,
         filter0 |= R300_TX_WRAP_T(R300_TX_CLAMP_TO_EDGE);
     }
 
+    /* determine min/max levels */
+    /* the MAX_MIP level is the largest (finest) one */
+    max_level = MIN2(sampler->max_lod, tex->tex.last_level);
+    min_level = MIN2(sampler->min_lod, max_level);
+    format0 |= R300_TX_NUM_LEVELS(max_level);
+    filter0 |= R300_TX_MAX_MIP_LEVEL(min_level);
+
     BEGIN_CS(16);
     OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), filter0 |
         (offset << 28));
     OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1);
     OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color);
 
-    OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), tex->state.format0);
+    OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), format0);
     OUT_CS_REG(R300_TX_FORMAT1_0 + (offset * 4), tex->state.format1);
     OUT_CS_REG(R300_TX_FORMAT2_0 + (offset * 4), tex->state.format2);
     OUT_CS_REG_SEQ(R300_TX_OFFSET_0 + (offset * 4), 1);
diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h
index 8ca785cb58..66fdada221 100644
--- a/src/gallium/drivers/r300/r300_reg.h
+++ b/src/gallium/drivers/r300/r300_reg.h
@@ -1463,6 +1463,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_TX_MIN_FILTER_MIP_NEAREST    (1 << 13)
 #	define R300_TX_MIN_FILTER_MIP_LINEAR     (2 << 13)
 #	define R300_TX_MIN_FILTER_MIP_MASK       (3 << 13)
+#       define R300_TX_MAX_MIP_LEVEL_SHIFT       17
+#       define R300_TX_MAX_MIP_LEVEL_MASK        (0xf << 17)
 #	define R300_TX_MAX_ANISO_1_TO_1          (0 << 21)
 #	define R300_TX_MAX_ANISO_2_TO_1          (1 << 21)
 #	define R300_TX_MAX_ANISO_4_TO_1          (2 << 21)
@@ -1471,6 +1473,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_TX_MAX_ANISO_MASK            (7 << 21)
 #       define R300_TX_WRAP_S(x)                 ((x) << 0)
 #       define R300_TX_WRAP_T(x)                 ((x) << 3)
+#       define R300_TX_MAX_MIP_LEVEL(x)          ((x) << 17)
 
 #define R300_TX_FILTER1_0                      0x4440
 #	define R300_CHROMA_KEY_MODE_DISABLE    0
@@ -1500,8 +1503,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #       define R300_TX_HEIGHTMASK_MASK           (2047 << 11)
 #	define R300_TX_DEPTHMASK_SHIFT           22
 #	define R300_TX_DEPTHMASK_MASK            (0xf << 22)
-#       define R300_TX_MAX_MIP_LEVEL_SHIFT       26
-#       define R300_TX_MAX_MIP_LEVEL_MASK        (0xf << 26)
 #       define R300_TX_SIZE_PROJECTED            (1 << 30)
 #       define R300_TX_PITCH_EN                  (1 << 31)
 #       define R300_TX_WIDTH(x)                  ((x) << 0)
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 5422a2cc9c..f2867675f0 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -523,6 +523,11 @@ static void*
                                                    state->mag_img_filter,
                                                    state->min_mip_filter);
 
+    /* Unfortunately, r300-r500 don't support floating-point mipmap lods. */
+    /* We must pass these to the emit function to clamp them properly. */
+    sampler->min_lod = MAX2((unsigned)state->min_lod, 0);
+    sampler->max_lod = MAX2((unsigned)ceilf(state->max_lod), 0);
+
     lod_bias = CLAMP((int)(state->lod_bias * 32), -(1 << 9), (1 << 9) - 1);
 
     sampler->filter1 |= lod_bias << R300_LOD_BIAS_SHIFT;
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index aea25cf71d..d13aa8f036 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -43,8 +43,7 @@ static void r300_setup_texture_state(struct r300_texture* tex, boolean is_r500)
         state->format2 = (tex->pitch[0] - 1) & 0x1fff;
     } else {
         /* power of two textures (3D, mipmaps, and no pitch) */
-        state->format0 |= R300_TX_DEPTH(util_logbase2(pt->depth[0]) & 0xf) |
-                          R300_TX_NUM_LEVELS(pt->last_level & 0xf);
+        state->format0 |= R300_TX_DEPTH(util_logbase2(pt->depth[0]) & 0xf);
     }
 
     state->format1 = r300_translate_texformat(pt->format);
-- 
cgit v1.2.3


From 8d80b5400a1bbf4e959cd8257d11dfe0483e93db Mon Sep 17 00:00:00 2001
From: Brian Paul <brianp@vmware.com>
Date: Mon, 23 Nov 2009 18:06:19 -0700
Subject: r300g: use util_bitcount()

---
 src/gallium/drivers/r300/r300_emit.c | 19 ++-----------------
 1 file changed, 2 insertions(+), 17 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index ad7dff36be..e6ab8e4af1 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -721,21 +721,6 @@ void r300_emit_vertex_format_state(struct r300_context* r300)
     END_CS;
 }
 
-/* XXX This should go to util ... */
-/* Return the number of bits set in the given number. */
-static unsigned bitcount(unsigned n)
-{
-    unsigned bits = 0;
-
-    while (n) {
-        if (n & 1) {
-            bits++;
-        }
-        n >>= 1;
-    }
-
-    return bits;
-}
 
 void r300_emit_vertex_program_code(struct r300_context* r300,
                                    struct r300_vertex_program_code* code)
@@ -745,8 +730,8 @@ void r300_emit_vertex_program_code(struct r300_context* r300,
     unsigned instruction_count = code->length / 4;
 
     int vtx_mem_size = r300screen->caps->is_r500 ? 128 : 72;
-    int input_count = MAX2(bitcount(code->InputsRead), 1);
-    int output_count = MAX2(bitcount(code->OutputsWritten), 1);
+    int input_count = MAX2(util_bitcount(code->InputsRead), 1);
+    int output_count = MAX2(util_bitcount(code->OutputsWritten), 1);
     int temp_count = MAX2(code->num_temporaries, 1);
     int pvs_num_slots = MIN3(vtx_mem_size / input_count,
                              vtx_mem_size / output_count, 10);
-- 
cgit v1.2.3


From b748a9f574361273df6b05c06c647ac4fd9b3e41 Mon Sep 17 00:00:00 2001
From: Roland Scheidegger <sroland@vmware.com>
Date: Fri, 27 Nov 2009 17:40:24 +0100
Subject: r300g,llvmpipe: fix some more merge problems

---
 src/gallium/drivers/llvmpipe/lp_texture.c | 2 +-
 src/gallium/drivers/r300/r300_emit.c      | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c
index 0a0f31f8a3..65d62fd072 100644
--- a/src/gallium/drivers/llvmpipe/lp_texture.c
+++ b/src/gallium/drivers/llvmpipe/lp_texture.c
@@ -169,7 +169,7 @@ llvmpipe_texture_blanket(struct pipe_screen * screen,
    /* Only supports one type */
    if (base->target != PIPE_TEXTURE_2D ||
        base->last_level != 0 ||
-       base->depth[0] != 1) {
+       base->depth0 != 1) {
       return NULL;
    }
 
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index e6ab8e4af1..98a39390bf 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -145,8 +145,8 @@ static const float * get_shader_constant(
                  * normalized coords. Should only show up on non-r500. */
                 case RC_STATE_R300_TEXRECT_FACTOR:
                     tex = &r300->textures[constant->u.State[1]]->tex;
-                    vec[0] = 1.0 / tex->width[0];
-                    vec[1] = 1.0 / tex->height[0];
+                    vec[0] = 1.0 / tex->width0;
+                    vec[1] = 1.0 / tex->height0;
                     break;
 
                 default:
-- 
cgit v1.2.3


From c78748a5274e58bcbb122923edf81065be9bbe16 Mon Sep 17 00:00:00 2001
From: Roland Scheidegger <sroland@vmware.com>
Date: Wed, 2 Dec 2009 02:08:26 +0100
Subject: gallium: adapt drivers to interface cleanups

---
 src/gallium/drivers/cell/ppu/cell_texture.c    |  30 +++---
 src/gallium/drivers/i915/i915_surface.c        |  21 +++--
 src/gallium/drivers/i915/i915_texture.c        | 125 ++++++++++---------------
 src/gallium/drivers/llvmpipe/lp_setup.c        |   2 +-
 src/gallium/drivers/llvmpipe/lp_tex_cache.c    |   2 +-
 src/gallium/drivers/llvmpipe/lp_texture.c      |  48 +++++-----
 src/gallium/drivers/llvmpipe/lp_tile_cache.c   |   6 +-
 src/gallium/drivers/r300/r300_emit.c           |  12 +--
 src/gallium/drivers/r300/r300_screen.c         |   9 +-
 src/gallium/drivers/r300/r300_texture.c        |  12 +--
 src/gallium/drivers/svga/svga_screen_texture.c |  37 ++++----
 src/gallium/drivers/svga/svga_state_vs.c       |   2 +-
 12 files changed, 138 insertions(+), 168 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c
index e6b8a87045..77a57aef14 100644
--- a/src/gallium/drivers/cell/ppu/cell_texture.c
+++ b/src/gallium/drivers/cell/ppu/cell_texture.c
@@ -65,14 +65,11 @@ cell_texture_layout(struct cell_texture *ct)
       w_tile = align(width, TILE_SIZE);
       h_tile = align(height, TILE_SIZE);
 
-      pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w_tile);  
-      pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h_tile);  
-
-      ct->stride[level] = pt->nblocksx[level] * pt->block.size;
+      ct->stride[level] = pf_get_stride(pt->format, w_tile);
 
       ct->level_offset[level] = ct->buffer_size;
 
-      size = pt->nblocksx[level] * pt->nblocksy[level] * pt->block.size;
+      size = ct->stride[level] * pf_get_nblocksy(pt->format, h_tile);
       if (pt->target == PIPE_TEXTURE_CUBE)
          size *= 6;
       else
@@ -283,10 +280,12 @@ cell_get_tex_surface(struct pipe_screen *screen,
       ps->zslice = zslice;
 
       if (pt->target == PIPE_TEXTURE_CUBE) {
-         ps->offset += face * pt->nblocksy[level] * ct->stride[level];
+         unsigned h_tile = align(ps->height, TILE_SIZE);
+         ps->offset += face * pf_get_nblocksy(ps->format, h_tile) * ct->stride[level];
       }
       else if (pt->target == PIPE_TEXTURE_3D) {
-         ps->offset += zslice * pt->nblocksy[level] * ct->stride[level];
+         unsigned h_tile = align(ps->height, TILE_SIZE);
+         ps->offset += zslice * pf_get_nblocksy(ps->format, h_tile) * ct->stride[level];
       }
       else {
          assert(face == 0);
@@ -327,14 +326,10 @@ cell_get_tex_transfer(struct pipe_screen *screen,
    if (ctrans) {
       struct pipe_transfer *pt = &ctrans->base;
       pipe_texture_reference(&pt->texture, texture);
-      pt->format = texture->format;
-      pt->block = texture->block;
       pt->x = x;
       pt->y = y;
       pt->width = w;
       pt->height = h;
-      pt->nblocksx = texture->nblocksx[level];
-      pt->nblocksy = texture->nblocksy[level];
       pt->stride = ct->stride[level];
       pt->usage = usage;
       pt->face = face;
@@ -344,10 +339,12 @@ cell_get_tex_transfer(struct pipe_screen *screen,
       ctrans->offset = ct->level_offset[level];
 
       if (texture->target == PIPE_TEXTURE_CUBE) {
-         ctrans->offset += face * pt->nblocksy * pt->stride;
+         unsigned h_tile = align(u_minify(texture->height0, level), TILE_SIZE);
+         ctrans->offset += face * pf_get_nblocksy(texture->format, h_tile) * pt->stride;
       }
       else if (texture->target == PIPE_TEXTURE_3D) {
-         ctrans->offset += zslice * pt->nblocksy * pt->stride;
+         unsigned h_tile = align(u_minify(texture->height0, level), TILE_SIZE);
+         ctrans->offset += zslice * pf_get_nblocksy(texture->format, h_tile) * pt->stride;
       }
       else {
          assert(face == 0);
@@ -400,7 +397,8 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer)
     * Create a buffer of ordinary memory for the linear texture.
     * This is the memory that the user will read/write.
     */
-   size = pt->nblocksx[level] * pt->nblocksy[level] * pt->block.size;
+   size = pf_get_stride(pt->format, align(texWidth, TILE_SIZE)) *
+          pf_get_nblocksy(pt->format, align(texHeight, TILE_SIZE));
 
    ctrans->map = align_malloc(size, 16);
    if (!ctrans->map)
@@ -408,7 +406,7 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer)
 
    if (transfer->usage & PIPE_TRANSFER_READ) {
       /* need to untwiddle the texture to make a linear version */
-      const uint bpp = pf_get_size(ct->base.format);
+      const uint bpp = pf_get_blocksize(ct->base.format);
       if (bpp == 4) {
          const uint *src = (uint *) (ct->mapped + ctrans->offset);
          uint *dst = ctrans->map;
@@ -451,7 +449,7 @@ cell_transfer_unmap(struct pipe_screen *screen,
       /* The user wrote new texture data into the mapped buffer.
        * We need to convert the new linear data into the twiddled/tiled format.
        */
-      const uint bpp = pf_get_size(ct->base.format);
+      const uint bpp = pf_get_blocksize(ct->base.format);
       if (bpp == 4) {
          const uint *src = ctrans->map;
          uint *dst = (uint *) (ct->mapped + ctrans->offset);
diff --git a/src/gallium/drivers/i915/i915_surface.c b/src/gallium/drivers/i915/i915_surface.c
index ab8331f3e6..24e1024aaa 100644
--- a/src/gallium/drivers/i915/i915_surface.c
+++ b/src/gallium/drivers/i915/i915_surface.c
@@ -48,17 +48,19 @@ i915_surface_copy(struct pipe_context *pipe,
 {
    struct i915_texture *dst_tex = (struct i915_texture *)dst->texture;
    struct i915_texture *src_tex = (struct i915_texture *)src->texture;
+   struct pipe_texture *dpt = &dst_tex->base;
+   struct pipe_texture *spt = &src_tex->base;
 
    assert( dst != src );
-   assert( dst_tex->base.block.size == src_tex->base.block.size );
-   assert( dst_tex->base.block.width == src_tex->base.block.height );
-   assert( dst_tex->base.block.height == src_tex->base.block.height );
-   assert( dst_tex->base.block.width == 1 );
-   assert( dst_tex->base.block.height == 1 );
+   assert( pf_get_blocksize(dpt->format) == pf_get_blocksize(spt->format) );
+   assert( pf_get_blockwidth(dpt->format) == pf_get_blockwidth(spt->format) );
+   assert( pf_get_blockheight(dpt->format) == pf_get_blockheight(spt->format) );
+   assert( pf_get_blockwidth(dpt->format) == 1 );
+   assert( pf_get_blockheight(dpt->format) == 1 );
 
    i915_copy_blit( i915_context(pipe),
                    FALSE,
-                   dst_tex->base.block.size,
+                   pf_get_blocksize(dpt->format),
                    (unsigned short) src_tex->stride, src_tex->buffer, src->offset,
                    (unsigned short) dst_tex->stride, dst_tex->buffer, dst->offset,
                    (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height );
@@ -72,12 +74,13 @@ i915_surface_fill(struct pipe_context *pipe,
 		  unsigned width, unsigned height, unsigned value)
 {
    struct i915_texture *tex = (struct i915_texture *)dst->texture;
+   struct pipe_texture *pt = &tex->base;
 
-   assert(tex->base.block.width == 1);
-   assert(tex->base.block.height == 1);
+   assert(pf_get_blockwidth(pt->format) == 1);
+   assert(pf_get_blockheight(pt->format) == 1);
 
    i915_fill_blit( i915_context(pipe),
-                   tex->base.block.size,
+                   pf_get_blocksize(pt->format),
                    (unsigned short) tex->stride,
                    tex->buffer, dst->offset,
                    (short) dstx, (short) dsty,
diff --git a/src/gallium/drivers/i915/i915_texture.c b/src/gallium/drivers/i915/i915_texture.c
index c7b86dd4c5..b28b413771 100644
--- a/src/gallium/drivers/i915/i915_texture.c
+++ b/src/gallium/drivers/i915/i915_texture.c
@@ -74,6 +74,9 @@ static const int step_offsets[6][2] = {
    {-1, 1}
 };
 
+/* XXX really need twice the size if x is already pot?
+   Otherwise just use util_next_power_of_two?
+*/
 static unsigned
 power_of_two(unsigned x)
 {
@@ -83,13 +86,6 @@ power_of_two(unsigned x)
    return value;
 }
 
-static unsigned
-round_up(unsigned n, unsigned multiple)
-{
-   return (n + multiple - 1) & ~(multiple - 1);
-}
-
-
 /*
  * More advanced helper funcs
  */
@@ -101,13 +97,8 @@ i915_miptree_set_level_info(struct i915_texture *tex,
                              unsigned nr_images,
                              unsigned w, unsigned h, unsigned d)
 {
-   struct pipe_texture *pt = &tex->base;
-
    assert(level < PIPE_MAX_TEXTURE_LEVELS);
 
-   pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w);
-   pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h);
-
    tex->nr_images[level] = nr_images;
 
    /*
@@ -138,7 +129,7 @@ i915_miptree_set_image_offset(struct i915_texture *tex,
 
    assert(img < tex->nr_images[level]);
 
-   tex->image_offset[level][img] = y * tex->stride + x * tex->base.block.size;
+   tex->image_offset[level][img] = y * tex->stride + x * pf_get_blocksize(tex->base.format);
 
    /*
    printf("%s level %d img %d pos %d,%d image_offset %x\n",
@@ -160,28 +151,28 @@ i915_scanout_layout(struct i915_texture *tex)
 {
    struct pipe_texture *pt = &tex->base;
 
-   if (pt->last_level > 0 || pt->block.size != 4)
+   if (pt->last_level > 0 || pf_get_blocksize(pt->format) != 4)
       return FALSE;
 
    i915_miptree_set_level_info(tex, 0, 1,
-                               tex->base.width0,
-                               tex->base.height0,
+                               pt->width0,
+                               pt->height0,
                                1);
    i915_miptree_set_image_offset(tex, 0, 0, 0, 0);
 
-   if (tex->base.width0 >= 240) {
-      tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size);
-      tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8);
+   if (pt->width0 >= 240) {
+      tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0));
+      tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8);
       tex->hw_tiled = INTEL_TILE_X;
-   } else if (tex->base.width0 == 64 && tex->base.height0 == 64) {
-      tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size);
-      tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8);
+   } else if (pt->width0 == 64 && pt->height0 == 64) {
+      tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0));
+      tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8);
    } else {
       return FALSE;
    }
 
    debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
-      tex->base.width0, tex->base.height0, pt->block.size,
+      pt->width0, pt->height0, pf_get_blocksize(pt->format),
       tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy);
 
    return TRUE;
@@ -195,25 +186,25 @@ i915_display_target_layout(struct i915_texture *tex)
 {
    struct pipe_texture *pt = &tex->base;
 
-   if (pt->last_level > 0 || pt->block.size != 4)
+   if (pt->last_level > 0 || pf_get_blocksize(pt->format) != 4)
       return FALSE;
 
    /* fallback to normal textures for small textures */
-   if (tex->base.width0 < 240)
+   if (pt->width0 < 240)
       return FALSE;
 
    i915_miptree_set_level_info(tex, 0, 1,
-                               tex->base.width0,
-                               tex->base.height0,
+                               pt->width0,
+                               pt->height0,
                                1);
    i915_miptree_set_image_offset(tex, 0, 0, 0, 0);
 
-   tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size);
-   tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8);
+   tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0));
+   tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8);
    tex->hw_tiled = INTEL_TILE_X;
 
    debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
-      tex->base.width0, tex->base.height0, pt->block.size,
+      pt->width0, pt->height0, pf_get_blocksize(pt->format),
       tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy);
 
    return TRUE;
@@ -226,34 +217,32 @@ i915_miptree_layout_2d(struct i915_texture *tex)
    unsigned level;
    unsigned width = pt->width0;
    unsigned height = pt->height0;
-   unsigned nblocksx = pt->nblocksx[0];
-   unsigned nblocksy = pt->nblocksy[0];
+   unsigned nblocksy = pf_get_nblocksy(pt->format, pt->width0);
 
    /* used for scanouts that need special layouts */
-   if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY)
+   if (pt->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY)
       if (i915_scanout_layout(tex))
          return;
 
    /* for shared buffers we use some very like scanout */
-   if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET)
+   if (pt->tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET)
       if (i915_display_target_layout(tex))
          return;
 
-   tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+   tex->stride = align(pf_get_stride(pt->format, pt->width0), 4);
    tex->total_nblocksy = 0;
 
    for (level = 0; level <= pt->last_level; level++) {
       i915_miptree_set_level_info(tex, level, 1, width, height, 1);
       i915_miptree_set_image_offset(tex, level, 0, 0, tex->total_nblocksy);
 
-      nblocksy = round_up(MAX2(2, nblocksy), 2);
+      nblocksy = align(MAX2(2, nblocksy), 2);
 
       tex->total_nblocksy += nblocksy;
 
       width = u_minify(width, 1);
       height = u_minify(height, 1);
-      nblocksx = pf_get_nblocksx(&pt->block, width);
-      nblocksy = pf_get_nblocksy(&pt->block, height);
+      nblocksy = pf_get_nblocksy(pt->format, height);
    }
 }
 
@@ -266,13 +255,12 @@ i915_miptree_layout_3d(struct i915_texture *tex)
    unsigned width = pt->width0;
    unsigned height = pt->height0;
    unsigned depth = pt->depth0;
-   unsigned nblocksx = pt->nblocksx[0];
-   unsigned nblocksy = pt->nblocksy[0];
+   unsigned nblocksy = pf_get_nblocksy(pt->format, pt->height0);
    unsigned stack_nblocksy = 0;
 
    /* Calculate the size of a single slice. 
     */
-   tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+   tex->stride = align(pf_get_stride(pt->format, pt->width0), 4);
 
    /* XXX: hardware expects/requires 9 levels at minimum.
     */
@@ -283,8 +271,7 @@ i915_miptree_layout_3d(struct i915_texture *tex)
 
       width = u_minify(width, 1);
       height = u_minify(height, 1);
-      nblocksx = pf_get_nblocksx(&pt->block, width);
-      nblocksy = pf_get_nblocksy(&pt->block, height);
+      nblocksy = pf_get_nblocksy(pt->format, height);
    }
 
    /* Fixup depth image_offsets: 
@@ -309,14 +296,14 @@ i915_miptree_layout_cube(struct i915_texture *tex)
 {
    struct pipe_texture *pt = &tex->base;
    unsigned width = pt->width0, height = pt->height0;
-   const unsigned nblocks = pt->nblocksx[0];
+   const unsigned nblocks = pf_get_nblocksx(pt->format, pt->width0);
    unsigned level;
    unsigned face;
 
    assert(width == height); /* cubemap images are square */
 
    /* double pitch for cube layouts */
-   tex->stride = round_up(nblocks * pt->block.size * 2, 4);
+   tex->stride = align(nblocks * pf_get_blocksize(pt->format) * 2, 4);
    tex->total_nblocksy = nblocks * 4;
 
    for (level = 0; level <= pt->last_level; level++) {
@@ -379,8 +366,8 @@ i945_miptree_layout_2d(struct i915_texture *tex)
    unsigned y = 0;
    unsigned width = pt->width0;
    unsigned height = pt->height0;
-   unsigned nblocksx = pt->nblocksx[0];
-   unsigned nblocksy = pt->nblocksy[0];
+   unsigned nblocksx = pf_get_nblocksx(pt->format, pt->width0);
+   unsigned nblocksy = pf_get_nblocksy(pt->format, pt->height0);
 
    /* used for scanouts that need special layouts */
    if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY)
@@ -392,7 +379,7 @@ i945_miptree_layout_2d(struct i915_texture *tex)
       if (i915_display_target_layout(tex))
          return;
 
-   tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+   tex->stride = align(pf_get_stride(pt->format, pt->width0), 4);
 
    /* May need to adjust pitch to accomodate the placement of
     * the 2nd mipmap level.  This occurs when the alignment
@@ -401,11 +388,11 @@ i945_miptree_layout_2d(struct i915_texture *tex)
     */
    if (pt->last_level > 0) {
       unsigned mip1_nblocksx 
-         = align(pf_get_nblocksx(&pt->block, u_minify(width, 1)), align_x)
-         + pf_get_nblocksx(&pt->block, u_minify(width, 2));
+         = align(pf_get_nblocksx(pt->format, u_minify(width, 1)), align_x)
+         + pf_get_nblocksx(pt->format, u_minify(width, 2));
 
       if (mip1_nblocksx > nblocksx)
-         tex->stride = mip1_nblocksx * pt->block.size;
+         tex->stride = mip1_nblocksx * pf_get_blocksize(pt->format);
    }
 
    /* Pitch must be a whole number of dwords
@@ -435,8 +422,8 @@ i945_miptree_layout_2d(struct i915_texture *tex)
 
       width  = u_minify(width, 1);
       height = u_minify(height, 1);
-      nblocksx = pf_get_nblocksx(&pt->block, width);
-      nblocksy = pf_get_nblocksy(&pt->block, height);
+      nblocksx = pf_get_nblocksx(pt->format, width);
+      nblocksy = pf_get_nblocksy(pt->format, height);
    }
 }
 
@@ -447,17 +434,16 @@ i945_miptree_layout_3d(struct i915_texture *tex)
    unsigned width = pt->width0;
    unsigned height = pt->height0;
    unsigned depth = pt->depth0;
-   unsigned nblocksx = pt->nblocksx[0];
-   unsigned nblocksy = pt->nblocksy[0];
+   unsigned nblocksy = pf_get_nblocksy(pt->format, pt->width0);
    unsigned pack_x_pitch, pack_x_nr;
    unsigned pack_y_pitch;
    unsigned level;
 
-   tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+   tex->stride = align(pf_get_stride(pt->format, pt->width0), 4);
    tex->total_nblocksy = 0;
 
-   pack_y_pitch = MAX2(pt->nblocksy[0], 2);
-   pack_x_pitch = tex->stride / pt->block.size;
+   pack_y_pitch = MAX2(nblocksy, 2);
+   pack_x_pitch = tex->stride / pf_get_blocksize(pt->format);
    pack_x_nr = 1;
 
    for (level = 0; level <= pt->last_level; level++) {
@@ -482,7 +468,7 @@ i945_miptree_layout_3d(struct i915_texture *tex)
       if (pack_x_pitch > 4) {
          pack_x_pitch >>= 1;
          pack_x_nr <<= 1;
-         assert(pack_x_pitch * pack_x_nr * pt->block.size <= tex->stride);
+         assert(pack_x_pitch * pack_x_nr * pf_get_blocksize(pt->format) <= tex->stride);
       }
 
       if (pack_y_pitch > 2) {
@@ -492,8 +478,7 @@ i945_miptree_layout_3d(struct i915_texture *tex)
       width = u_minify(width, 1);
       height = u_minify(height, 1);
       depth = u_minify(depth, 1);
-      nblocksx = pf_get_nblocksx(&pt->block, width);
-      nblocksy = pf_get_nblocksy(&pt->block, height);
+      nblocksy = pf_get_nblocksy(pt->format, height);
    }
 }
 
@@ -503,7 +488,7 @@ i945_miptree_layout_cube(struct i915_texture *tex)
    struct pipe_texture *pt = &tex->base;
    unsigned level;
 
-   const unsigned nblocks = pt->nblocksx[0];
+   const unsigned nblocks = pf_get_nblocksx(pt->format, pt->width0);
    unsigned face;
    unsigned width = pt->width0;
    unsigned height = pt->height0;
@@ -523,9 +508,9 @@ i945_miptree_layout_cube(struct i915_texture *tex)
     * or the final row of 4x4, 2x2 and 1x1 faces below this.
     */
    if (nblocks > 32)
-      tex->stride = round_up(nblocks * pt->block.size * 2, 4);
+      tex->stride = align(nblocks * pf_get_blocksize(pt->format) * 2, 4);
    else
-      tex->stride = 14 * 8 * pt->block.size;
+      tex->stride = 14 * 8 * pf_get_blocksize(pt->format);
 
    tex->total_nblocksy = nblocks * 4;
 
@@ -645,9 +630,6 @@ i915_texture_create(struct pipe_screen *screen,
    pipe_reference_init(&tex->base.reference, 1);
    tex->base.screen = screen;
 
-   tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width0);
-   tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height0);
-   
    if (is->is_i945) {
       if (!i945_miptree_layout(tex))
          goto fail;
@@ -829,14 +811,10 @@ i915_get_tex_transfer(struct pipe_screen *screen,
    trans = CALLOC_STRUCT(i915_transfer);
    if (trans) {
       pipe_texture_reference(&trans->base.texture, texture);
-      trans->base.format = trans->base.format;
       trans->base.x = x;
       trans->base.y = y;
       trans->base.width = w;
       trans->base.height = h;
-      trans->base.block = texture->block;
-      trans->base.nblocksx = texture->nblocksx[level];
-      trans->base.nblocksy = texture->nblocksy[level];
       trans->base.stride = tex->stride;
       trans->offset = offset;
       trans->base.usage = usage;
@@ -852,6 +830,7 @@ i915_transfer_map(struct pipe_screen *screen,
    struct intel_winsys *iws = i915_screen(tex->base.screen)->iws;
    char *map;
    boolean write = FALSE;
+   enum pipe_format format = tex->base.format;
 
    if (transfer->usage & PIPE_TRANSFER_WRITE)
       write = TRUE;
@@ -861,8 +840,8 @@ i915_transfer_map(struct pipe_screen *screen,
       return NULL;
 
    return map + i915_transfer(transfer)->offset +
-      transfer->y / transfer->block.height * transfer->stride +
-      transfer->x / transfer->block.width * transfer->block.size;
+      transfer->y / pf_get_blockheight(format) * transfer->stride +
+      transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format);
 }
 
 static void
diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c
index ffcbc9a379..b4aabd4d7c 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup.c
@@ -166,7 +166,7 @@ shade_quads(struct llvmpipe_context *llvmpipe,
       assert((y % 2) == 0);
       depth = llvmpipe->zsbuf_map +
               y*llvmpipe->zsbuf_transfer->stride +
-              2*x*llvmpipe->zsbuf_transfer->block.size;
+              2*x*pf_get_blocksize(llvmpipe->zsbuf_transfer->texture->format);
    }
    else
       depth = NULL;
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_cache.c b/src/gallium/drivers/llvmpipe/lp_tex_cache.c
index c7c4143bc6..5dbc597d2c 100644
--- a/src/gallium/drivers/llvmpipe/lp_tex_cache.c
+++ b/src/gallium/drivers/llvmpipe/lp_tex_cache.c
@@ -291,7 +291,7 @@ lp_find_cached_tex_tile(struct llvmpipe_tex_tile_cache *tc,
             assert(0);
          }
 
-         util_format_read_4ub(tc->tex_trans->format,
+         util_format_read_4ub(tc->tex_trans->texture->format,
                               (uint8_t *)tile->color, sizeof tile->color[0],
                               tc->tex_trans_map, tc->tex_trans->stride,
                               x, y, w, h);
diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c
index 65d62fd072..f099f903bd 100644
--- a/src/gallium/drivers/llvmpipe/lp_texture.c
+++ b/src/gallium/drivers/llvmpipe/lp_texture.c
@@ -48,7 +48,6 @@
 /* Simple, maximally packed layout.
  */
 
-
 /* Conventional allocation path for non-display textures:
  */
 static boolean
@@ -63,20 +62,15 @@ llvmpipe_texture_layout(struct llvmpipe_screen *screen,
 
    unsigned buffer_size = 0;
 
-   pf_get_block(lpt->base.format, &lpt->base.block);
-
    for (level = 0; level <= pt->last_level; level++) {
       unsigned nblocksx, nblocksy;
 
-      pt->nblocksx[level] = pf_get_nblocksx(&pt->block, width);  
-      pt->nblocksy[level] = pf_get_nblocksy(&pt->block, height);
-
       /* Allocate storage for whole quads. This is particularly important
        * for depth surfaces, which are currently stored in a swizzled format. */
-      nblocksx = pf_get_nblocksx(&pt->block, align(width, 2));
-      nblocksy = pf_get_nblocksy(&pt->block, align(height, 2));
+      nblocksx = pf_get_nblocksx(pt->format, align(width, 2));
+      nblocksy = pf_get_nblocksy(pt->format, align(height, 2));
 
-      lpt->stride[level] = align(nblocksx*pt->block.size, 16);
+      lpt->stride[level] = align(nblocksx * pf_get_blocksize(pt->format), 16);
 
       lpt->level_offset[level] = buffer_size;
 
@@ -100,10 +94,6 @@ llvmpipe_displaytarget_layout(struct llvmpipe_screen *screen,
 {
    struct llvmpipe_winsys *winsys = screen->winsys;
 
-   pf_get_block(lpt->base.format, &lpt->base.block);
-   lpt->base.nblocksx[0] = pf_get_nblocksx(&lpt->base.block, lpt->base.width0);  
-   lpt->base.nblocksy[0] = pf_get_nblocksy(&lpt->base.block, lpt->base.height0);  
-
    lpt->dt = winsys->displaytarget_create(winsys,
                                           lpt->base.format,
                                           lpt->base.width0,
@@ -180,8 +170,6 @@ llvmpipe_texture_blanket(struct pipe_screen * screen,
    lpt->base = *base;
    pipe_reference_init(&lpt->base.reference, 1);
    lpt->base.screen = screen;
-   lpt->base.nblocksx[0] = pf_get_nblocksx(&lpt->base.block, lpt->base.width0);  
-   lpt->base.nblocksy[0] = pf_get_nblocksy(&lpt->base.block, lpt->base.height0);  
    lpt->stride[0] = stride[0];
 
    pipe_buffer_reference(&lpt->buffer, buffer);
@@ -255,11 +243,17 @@ llvmpipe_get_tex_surface(struct pipe_screen *screen,
       ps->level = level;
       ps->zslice = zslice;
 
+      /* XXX shouldn't that rather be
+         tex_height = align(ps->height, 2);
+         to account for alignment done in llvmpipe_texture_layout ?
+      */
       if (pt->target == PIPE_TEXTURE_CUBE) {
-         ps->offset += face * pt->nblocksy[level] * lpt->stride[level];
+         unsigned tex_height = ps->height;
+         ps->offset += face * pf_get_nblocksy(pt->format, tex_height) * lpt->stride[level];
       }
       else if (pt->target == PIPE_TEXTURE_3D) {
-         ps->offset += zslice * pt->nblocksy[level] * lpt->stride[level];
+         unsigned tex_height = ps->height;
+         ps->offset += zslice * pf_get_nblocksy(pt->format, tex_height) * lpt->stride[level];
       }
       else {
          assert(face == 0);
@@ -300,14 +294,10 @@ llvmpipe_get_tex_transfer(struct pipe_screen *screen,
    if (lpt) {
       struct pipe_transfer *pt = &lpt->base;
       pipe_texture_reference(&pt->texture, texture);
-      pt->format = texture->format;
-      pt->block = texture->block;
       pt->x = x;
       pt->y = y;
       pt->width = w;
       pt->height = h;
-      pt->nblocksx = texture->nblocksx[level];
-      pt->nblocksy = texture->nblocksy[level];
       pt->stride = lptex->stride[level];
       pt->usage = usage;
       pt->face = face;
@@ -316,11 +306,17 @@ llvmpipe_get_tex_transfer(struct pipe_screen *screen,
 
       lpt->offset = lptex->level_offset[level];
 
+      /* XXX shouldn't that rather be
+         tex_height = align(u_minify(texture->height0, level), 2)
+         to account for alignment done in llvmpipe_texture_layout ?
+      */
       if (texture->target == PIPE_TEXTURE_CUBE) {
-         lpt->offset += face * pt->nblocksy * pt->stride;
+         unsigned tex_height = u_minify(texture->height0, level);
+         lpt->offset += face *  pf_get_nblocksy(texture->format, tex_height) * pt->stride;
       }
       else if (texture->target == PIPE_TEXTURE_3D) {
-         lpt->offset += zslice * pt->nblocksy * pt->stride;
+         unsigned tex_height = u_minify(texture->height0, level);
+         lpt->offset += zslice * pf_get_nblocksy(texture->format, tex_height) * pt->stride;
       }
       else {
          assert(face == 0);
@@ -352,9 +348,11 @@ llvmpipe_transfer_map( struct pipe_screen *_screen,
    struct llvmpipe_screen *screen = llvmpipe_screen(_screen);
    ubyte *map, *xfer_map;
    struct llvmpipe_texture *lpt;
+   enum pipe_format format;
 
    assert(transfer->texture);
    lpt = llvmpipe_texture(transfer->texture);
+   format = lpt->base.format;
 
    if(lpt->dt) {
       struct llvmpipe_winsys *winsys = screen->winsys;
@@ -379,8 +377,8 @@ llvmpipe_transfer_map( struct pipe_screen *_screen,
    }
    
    xfer_map = map + llvmpipe_transfer(transfer)->offset +
-      transfer->y / transfer->block.height * transfer->stride +
-      transfer->x / transfer->block.width * transfer->block.size;
+      transfer->y / pf_get_blockheight(format) * transfer->stride +
+      transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format);
    /*printf("map = %p  xfer map = %p\n", map, xfer_map);*/
    return xfer_map;
 }
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_cache.c b/src/gallium/drivers/llvmpipe/lp_tile_cache.c
index ec3e002d62..50891c4227 100644
--- a/src/gallium/drivers/llvmpipe/lp_tile_cache.c
+++ b/src/gallium/drivers/llvmpipe/lp_tile_cache.c
@@ -252,13 +252,13 @@ lp_flush_tile_cache(struct llvmpipe_tile_cache *tc)
                case LP_TILE_STATUS_CLEAR:
                   /* Actually clear the tiles which were flagged as being in a
                    * clear state. */
-                  util_fill_rect(tc->transfer_map, &pt->block, pt->stride,
+                  util_fill_rect(tc->transfer_map, pt->texture->format, pt->stride,
                                  x, y, w, h,
                                  tc->clear_val);
                   break;
 
                case LP_TILE_STATUS_DEFINED:
-                  lp_tile_write_4ub(pt->format,
+                  lp_tile_write_4ub(pt->texture->format,
                                     tile->color,
                                     tc->transfer_map, pt->stride,
                                     x, y, w, h);
@@ -306,7 +306,7 @@ lp_get_cached_tile(struct llvmpipe_tile_cache *tc,
       y &= ~(TILE_SIZE - 1);
 
       if (!pipe_clip_tile(x, y, &w, &h, tc->transfer))
-         lp_tile_read_4ub(pt->format,
+         lp_tile_read_4ub(pt->texture->format,
                           tile->color,
                           tc->transfer_map, tc->transfer->stride,
                           x, y, w, h);
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 98a39390bf..a479842f9e 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -631,10 +631,10 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset)
     for (i = 0; i < aos_count - 1; i += 2) {
         int buf_num1 = velem[i].vertex_buffer_index;
         int buf_num2 = velem[i+1].vertex_buffer_index;
-        assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0);
-        assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_size(velem[i+1].src_format) % 4 == 0);
-        OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) |
-               (pf_get_size(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22));
+        assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_blocksize(velem[i].src_format) % 4 == 0);
+        assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_blocksize(velem[i+1].src_format) % 4 == 0);
+        OUT_CS((pf_get_blocksize(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) |
+               (pf_get_blocksize(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22));
         OUT_CS(vbuf[buf_num1].buffer_offset + velem[i].src_offset +
                offset * vbuf[buf_num1].stride);
         OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset +
@@ -642,8 +642,8 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset)
     }
     if (aos_count & 1) {
         int buf_num = velem[i].vertex_buffer_index;
-        assert(vbuf[buf_num].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0);
-        OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6));
+        assert(vbuf[buf_num].stride % 4 == 0 && pf_get_blocksize(velem[i].src_format) % 4 == 0);
+        OUT_CS((pf_get_blocksize(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6));
         OUT_CS(vbuf[buf_num].buffer_offset + velem[i].src_offset +
                offset * vbuf[buf_num].stride);
     }
diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c
index 390b63007e..032fa69ec0 100644
--- a/src/gallium/drivers/r300/r300_screen.c
+++ b/src/gallium/drivers/r300/r300_screen.c
@@ -311,14 +311,10 @@ r300_get_tex_transfer(struct pipe_screen *screen,
     trans = CALLOC_STRUCT(r300_transfer);
     if (trans) {
         pipe_texture_reference(&trans->transfer.texture, texture);
-        trans->transfer.format = texture->format;
         trans->transfer.x = x;
         trans->transfer.y = y;
         trans->transfer.width = w;
         trans->transfer.height = h;
-        trans->transfer.block = texture->block;
-        trans->transfer.nblocksx = texture->nblocksx[level];
-        trans->transfer.nblocksy = texture->nblocksy[level];
         trans->transfer.stride = r300_texture_get_stride(tex, level);
         trans->transfer.usage = usage;
 
@@ -344,6 +340,7 @@ static void* r300_transfer_map(struct pipe_screen* screen,
 {
     struct r300_texture* tex = (struct r300_texture*)transfer->texture;
     char* map;
+    enum pipe_format format = tex->tex.format;
 
     map = pipe_buffer_map(screen, tex->buffer,
                           pipe_transfer_buffer_flags(transfer));
@@ -353,8 +350,8 @@ static void* r300_transfer_map(struct pipe_screen* screen,
     }
 
     return map + r300_transfer(transfer)->offset +
-        transfer->y / transfer->block.height * transfer->stride +
-        transfer->x / transfer->block.width * transfer->block.size;
+        transfer->y / pf_get_blockheight(format) * transfer->stride +
+        transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format);
 }
 
 static void r300_transfer_unmap(struct pipe_screen* screen,
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index 093a21ebe2..63fc6a235a 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -105,7 +105,7 @@ unsigned r300_texture_get_stride(struct r300_texture* tex, unsigned level)
         return 0;
     }
 
-    return align(pf_get_stride(&tex->tex.block, u_minify(tex->tex.width0, level)), 32);
+    return align(pf_get_stride(tex->tex.format, u_minify(tex->tex.width0, level)), 32);
 }
 
 static void r300_setup_miptree(struct r300_texture* tex)
@@ -115,11 +115,10 @@ static void r300_setup_miptree(struct r300_texture* tex)
     int i;
 
     for (i = 0; i <= base->last_level; i++) {
-        base->nblocksx[i] = pf_get_nblocksx(&base->block, u_minify(base->width0, i));
-        base->nblocksy[i] = pf_get_nblocksy(&base->block, u_minify(base->height0, i));
+        unsigned nblocksy = pf_get_nblocksy(base->format, u_minify(base->height0, i));
 
         stride = r300_texture_get_stride(tex, i);
-        layer_size = stride * base->nblocksy[i];
+        layer_size = stride * nblocksy;
 
         if (base->target == PIPE_TEXTURE_CUBE)
             size = layer_size * 6;
@@ -129,7 +128,7 @@ static void r300_setup_miptree(struct r300_texture* tex)
         tex->offset[i] = align(tex->size, 32);
         tex->size = tex->offset[i] + size;
         tex->layer_size[i] = layer_size;
-        tex->pitch[i] = stride / base->block.size;
+        tex->pitch[i] = stride / pf_get_blocksize(base->format);
 
         debug_printf("r300: Texture miptree: Level %d "
                 "(%dx%dx%d px, pitch %d bytes)\n",
@@ -245,7 +244,7 @@ static struct pipe_texture*
     tex->tex.screen = screen;
 
     tex->stride_override = *stride;
-    tex->pitch[0] = *stride / base->block.size;
+    tex->pitch[0] = *stride / pf_get_blocksize(base->format);
 
     r300_setup_flags(tex);
     r300_setup_texture_state(tex, r300_screen(screen)->caps->is_r500);
@@ -283,7 +282,6 @@ r300_video_surface_create(struct pipe_screen *screen,
     template.width0 = util_next_power_of_two(width);
     template.height0 = util_next_power_of_two(height);
     template.depth0 = 1;
-    pf_get_block(template.format, &template.block);
     template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER |
                          PIPE_TEXTURE_USAGE_RENDER_TARGET;
 
diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c
index fb11b80dcf..410adf881b 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.c
+++ b/src/gallium/drivers/svga/svga_screen_texture.c
@@ -158,7 +158,8 @@ svga_transfer_dma_band(struct svga_transfer *st,
                 st->base.x + st->base.width,
                 y + h,
                 st->base.zslice + 1,
-                texture->base.block.size*8/(texture->base.block.width*texture->base.block.height));
+                pf_get_blocksize(texture->base.format)*8/
+                (pf_get_blockwidth(texture->base.format)*pf_get_blockheight(texture->base.format)));
    
    box.x = st->base.x;
    box.y = y;
@@ -208,7 +209,8 @@ svga_transfer_dma(struct svga_transfer *st,
    }
    else {
       unsigned y, h, srcy;
-      h = st->hw_nblocksy * st->base.block.height;
+      unsigned blockheight = pf_get_blockheight(st->base.texture->format);
+      h = st->hw_nblocksy * blockheight;
       srcy = 0;
       for(y = 0; y < st->base.height; y += h) {
          unsigned offset, length;
@@ -218,11 +220,11 @@ svga_transfer_dma(struct svga_transfer *st,
             h = st->base.height - y;
 
          /* Transfer band must be aligned to pixel block boundaries */
-         assert(y % st->base.block.height == 0);
-         assert(h % st->base.block.height == 0);
+         assert(y % blockheight == 0);
+         assert(h % blockheight == 0);
          
-         offset = y * st->base.stride / st->base.block.height;
-         length = h * st->base.stride / st->base.block.height;
+         offset = y * st->base.stride / blockheight;
+         length = h * st->base.stride / blockheight;
 
          sw = (uint8_t *)st->swbuf + offset;
          
@@ -291,8 +293,6 @@ svga_texture_create(struct pipe_screen *screen,
    height = templat->height0;
    depth = templat->depth0;
    for(level = 0; level <= templat->last_level; ++level) {
-      tex->base.nblocksx[level] = pf_get_nblocksx(&tex->base.block, width);  
-      tex->base.nblocksy[level] = pf_get_nblocksy(&tex->base.block, height);  
       width = u_minify(width, 1);
       height = u_minify(height, 1);
       depth = u_minify(depth, 1);
@@ -750,6 +750,8 @@ svga_get_tex_transfer(struct pipe_screen *screen,
    struct svga_screen *ss = svga_screen(screen);
    struct svga_winsys_screen *sws = ss->sws;
    struct svga_transfer *st;
+   unsigned nblocksx = pf_get_nblocksx(texture->format, w);
+   unsigned nblocksy = pf_get_nblocksy(texture->format, h);
 
    /* We can't map texture storage directly */
    if (usage & PIPE_TRANSFER_MAP_DIRECTLY)
@@ -759,21 +761,17 @@ svga_get_tex_transfer(struct pipe_screen *screen,
    if (!st)
       return NULL;
    
-   st->base.format = texture->format;
-   st->base.block = texture->block;
    st->base.x = x;
    st->base.y = y;
    st->base.width = w;
    st->base.height = h;
-   st->base.nblocksx = pf_get_nblocksx(&texture->block, w);
-   st->base.nblocksy = pf_get_nblocksy(&texture->block, h);
-   st->base.stride = st->base.nblocksx*st->base.block.size;
+   st->base.stride = nblocksx*pf_get_blocksize(texture->format);
    st->base.usage = usage;
    st->base.face = face;
    st->base.level = level;
    st->base.zslice = zslice;
 
-   st->hw_nblocksy = st->base.nblocksy;
+   st->hw_nblocksy = nblocksy;
    
    st->hwbuf = svga_winsys_buffer_create(ss, 
                                          1, 
@@ -789,15 +787,15 @@ svga_get_tex_transfer(struct pipe_screen *screen,
    if(!st->hwbuf)
       goto no_hwbuf;
 
-   if(st->hw_nblocksy < st->base.nblocksy) {
+   if(st->hw_nblocksy < nblocksy) {
       /* We couldn't allocate a hardware buffer big enough for the transfer, 
        * so allocate regular malloc memory instead */
       debug_printf("%s: failed to allocate %u KB of DMA, splitting into %u x %u KB DMA transfers\n",
                    __FUNCTION__,
-                   (st->base.nblocksy*st->base.stride + 1023)/1024,
-                   (st->base.nblocksy + st->hw_nblocksy - 1)/st->hw_nblocksy,
+                   (nblocksy*st->base.stride + 1023)/1024,
+                   (nblocksy + st->hw_nblocksy - 1)/st->hw_nblocksy,
                    (st->hw_nblocksy*st->base.stride + 1023)/1024);
-      st->swbuf = MALLOC(st->base.nblocksy*st->base.stride);
+      st->swbuf = MALLOC(nblocksy*st->base.stride);
       if(!st->swbuf)
          goto no_swbuf;
    }
@@ -1046,8 +1044,7 @@ svga_screen_buffer_from_texture(struct pipe_texture *texture,
        svga_translate_format(texture->format),
        stex->handle);
 
-   *stride = pf_get_nblocksx(&texture->block, texture->width0) *
-      texture->block.size;
+   *stride = pf_get_stride(texture->format, texture->width0);
 
    return *buffer != NULL;
 }
diff --git a/src/gallium/drivers/svga/svga_state_vs.c b/src/gallium/drivers/svga/svga_state_vs.c
index a947745732..f1b0daf9f6 100644
--- a/src/gallium/drivers/svga/svga_state_vs.c
+++ b/src/gallium/drivers/svga/svga_state_vs.c
@@ -210,7 +210,7 @@ static int update_zero_stride( struct svga_context *svga,
          mapped_buffer = pipe_buffer_map_range(svga->pipe.screen, 
                                                vbuffer->buffer,
                                                vel->src_offset,
-                                               pf_get_size(vel->src_format),
+                                               pf_get_blocksize(vel->src_format),
                                                PIPE_BUFFER_USAGE_CPU_READ);
          translate->set_buffer(translate, vel->vertex_buffer_index,
                                mapped_buffer,
-- 
cgit v1.2.3


From cceeab39ea541b1be1521114316d660a77769c2a Mon Sep 17 00:00:00 2001
From: Michal Krol <michal@vmware.com>
Date: Thu, 3 Dec 2009 11:17:37 +0100
Subject: Move pf_get_bits/size() to u_format auxiliary module.

---
 src/gallium/auxiliary/util/u_format.h        | 29 ++++++++++++++++++++++
 src/gallium/auxiliary/util/u_gen_mipmap.c    |  5 ++--
 src/gallium/drivers/cell/ppu/cell_texture.c  |  6 +++--
 src/gallium/drivers/r300/r300_emit.c         | 13 +++++-----
 src/gallium/drivers/softpipe/sp_tile_cache.c |  3 ++-
 src/gallium/drivers/svga/svga_state_vs.c     |  3 ++-
 src/gallium/include/pipe/p_format.h          | 36 ----------------------------
 src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c |  5 ++--
 src/gallium/winsys/gdi/gdi_softpipe_winsys.c |  4 ++--
 src/mesa/state_tracker/st_cb_texture.c       |  2 +-
 10 files changed, 53 insertions(+), 53 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h
index 583b62e606..3ac5384024 100644
--- a/src/gallium/auxiliary/util/u_format.h
+++ b/src/gallium/auxiliary/util/u_format.h
@@ -183,6 +183,35 @@ util_format_get_block(enum pipe_format format,
    block->height = desc->block.height;
 }
 
+/**
+ * Return total bits needed for the pixel format.
+ */
+static INLINE uint
+util_format_get_bits(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(format);
+   if (!format) {
+      return 0;
+   }
+
+   return desc->block.bits / (desc->block.width * desc->block.height);
+}
+
+/**
+ * Return bytes per pixel for the given format.
+ */
+static INLINE uint
+util_format_get_size(enum pipe_format format)
+{
+   uint bits = util_format_get_bits(format);
+
+   assert(bits % 8 == 0);
+
+   return bits / 8;
+}
+
 
 /*
  * Format access functions.
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
index f67f1e458d..70ec925d15 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.c
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -41,6 +41,7 @@
 #include "pipe/p_shader_tokens.h"
 #include "pipe/p_state.h"
 
+#include "util/u_format.h"
 #include "util/u_memory.h"
 #include "util/u_draw_quad.h"
 #include "util/u_gen_mipmap.h"
@@ -996,7 +997,7 @@ reduce_2d(enum pipe_format pformat,
 {
    enum dtype datatype;
    uint comps;
-   const int bpt = pf_get_size(pformat);
+   const int bpt = util_format_get_size(pformat);
    const ubyte *srcA, *srcB;
    ubyte *dst;
    int row;
@@ -1035,7 +1036,7 @@ reduce_3d(enum pipe_format pformat,
           int dstWidth, int dstHeight, int dstDepth,
           int dstRowStride, ubyte *dstPtr)
 {
-   const int bpt = pf_get_size(pformat);
+   const int bpt = util_format_get_size(pformat);
    const int border = 0;
    int img, row;
    int bytesPerSrcImage, bytesPerDstImage;
diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c
index e6b8a87045..605d53a948 100644
--- a/src/gallium/drivers/cell/ppu/cell_texture.c
+++ b/src/gallium/drivers/cell/ppu/cell_texture.c
@@ -35,6 +35,8 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "pipe/internal/p_winsys_screen.h"
+
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 
@@ -408,7 +410,7 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer)
 
    if (transfer->usage & PIPE_TRANSFER_READ) {
       /* need to untwiddle the texture to make a linear version */
-      const uint bpp = pf_get_size(ct->base.format);
+      const uint bpp = util_format_get_size(ct->base.format);
       if (bpp == 4) {
          const uint *src = (uint *) (ct->mapped + ctrans->offset);
          uint *dst = ctrans->map;
@@ -451,7 +453,7 @@ cell_transfer_unmap(struct pipe_screen *screen,
       /* The user wrote new texture data into the mapped buffer.
        * We need to convert the new linear data into the twiddled/tiled format.
        */
-      const uint bpp = pf_get_size(ct->base.format);
+      const uint bpp = util_format_get_size(ct->base.format);
       if (bpp == 4) {
          const uint *src = ctrans->map;
          uint *dst = (uint *) (ct->mapped + ctrans->offset);
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 98a39390bf..171859b8e4 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -22,6 +22,7 @@
 
 /* r300_emit: Functions for emitting state. */
 
+#include "util/u_format.h"
 #include "util/u_math.h"
 
 #include "r300_context.h"
@@ -631,10 +632,10 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset)
     for (i = 0; i < aos_count - 1; i += 2) {
         int buf_num1 = velem[i].vertex_buffer_index;
         int buf_num2 = velem[i+1].vertex_buffer_index;
-        assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0);
-        assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_size(velem[i+1].src_format) % 4 == 0);
-        OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) |
-               (pf_get_size(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22));
+        assert(vbuf[buf_num1].stride % 4 == 0 && util_format_get_size(velem[i].src_format) % 4 == 0);
+        assert(vbuf[buf_num2].stride % 4 == 0 && util_format_get_size(velem[i+1].src_format) % 4 == 0);
+        OUT_CS((util_format_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) |
+               (util_format_get_size(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22));
         OUT_CS(vbuf[buf_num1].buffer_offset + velem[i].src_offset +
                offset * vbuf[buf_num1].stride);
         OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset +
@@ -642,8 +643,8 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset)
     }
     if (aos_count & 1) {
         int buf_num = velem[i].vertex_buffer_index;
-        assert(vbuf[buf_num].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0);
-        OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6));
+        assert(vbuf[buf_num].stride % 4 == 0 && util_format_get_size(velem[i].src_format) % 4 == 0);
+        OUT_CS((util_format_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6));
         OUT_CS(vbuf[buf_num].buffer_offset + velem[i].src_offset +
                offset * vbuf[buf_num].stride);
     }
diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c
index 65872cecc4..cde2260073 100644
--- a/src/gallium/drivers/softpipe/sp_tile_cache.c
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.c
@@ -33,6 +33,7 @@
  */
 
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 #include "util/u_memory.h"
 #include "util/u_tile.h"
 #include "sp_tile_cache.h"
@@ -238,7 +239,7 @@ clear_tile(struct softpipe_cached_tile *tile,
 {
    uint i, j;
 
-   switch (pf_get_size(format)) {
+   switch (util_format_get_size(format)) {
    case 1:
       memset(tile->data.any, clear_value, TILE_SIZE * TILE_SIZE);
       break;
diff --git a/src/gallium/drivers/svga/svga_state_vs.c b/src/gallium/drivers/svga/svga_state_vs.c
index a947745732..9e339577c7 100644
--- a/src/gallium/drivers/svga/svga_state_vs.c
+++ b/src/gallium/drivers/svga/svga_state_vs.c
@@ -25,6 +25,7 @@
 
 #include "pipe/p_inlines.h"
 #include "pipe/p_defines.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "translate/translate.h"
 
@@ -210,7 +211,7 @@ static int update_zero_stride( struct svga_context *svga,
          mapped_buffer = pipe_buffer_map_range(svga->pipe.screen, 
                                                vbuffer->buffer,
                                                vel->src_offset,
-                                               pf_get_size(vel->src_format),
+                                               util_format_get_size(vel->src_format),
                                                PIPE_BUFFER_USAGE_CPU_READ);
          translate->set_buffer(translate, vel->vertex_buffer_index,
                                mapped_buffer,
diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h
index 3be5b18a25..5fd073c95f 100644
--- a/src/gallium/include/pipe/p_format.h
+++ b/src/gallium/include/pipe/p_format.h
@@ -422,42 +422,6 @@ static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp )
    return size << (pf_mixed_scale8( format ) * 3);
 }
 
-/**
- * Return total bits needed for the pixel format.
- */
-static INLINE uint pf_get_bits( enum pipe_format format )
-{
-   switch (pf_layout(format)) {
-   case PIPE_FORMAT_LAYOUT_RGBAZS:
-   case PIPE_FORMAT_LAYOUT_MIXED:
-      return
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_0 ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_1 ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_B ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_S );
-   case PIPE_FORMAT_LAYOUT_YCBCR:
-      assert( format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV );
-      /* return effective bits per pixel */
-      return 16; 
-   default:
-      assert( 0 );
-      return 0;
-   }
-}
-
-/**
- * Return bytes per pixel for the given format.
- */
-static INLINE uint pf_get_size( enum pipe_format format )
-{
-   assert(pf_get_bits(format) % 8 == 0);
-   return pf_get_bits(format) / 8;
-}
-
 /**
  * Describe accurately the pixel format.
  * 
diff --git a/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c b/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c
index e8bc0f55ac..9cb77f3ad3 100644
--- a/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c
+++ b/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c
@@ -39,6 +39,7 @@
 #include "pipe/p_format.h"
 #include "pipe/p_context.h"
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 #include "llvmpipe/lp_winsys.h"
@@ -147,8 +148,8 @@ gdi_llvmpipe_displaytarget_create(struct llvmpipe_winsys *winsys,
    gdt->width = width;
    gdt->height = height;
 
-   bpp = pf_get_bits(format);
-   cpp = pf_get_size(format);
+   bpp = util_format_get_bits(format);
+   cpp = util_format_get_size(format);
    
    gdt->stride = round_up(width * cpp, alignment);
    gdt->size = gdt->stride * height;
diff --git a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c
index d9fb2080a1..a58648ddb9 100644
--- a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c
+++ b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c
@@ -284,10 +284,10 @@ gdi_softpipe_present(struct pipe_screen *screen,
 
     memset(&bmi, 0, sizeof(BITMAPINFO));
     bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER);
-    bmi.bmiHeader.biWidth = texture->stride[surface->level] / pf_get_size(surface->format);
+    bmi.bmiHeader.biWidth = texture->stride[surface->level] / util_format_get_size(surface->format);
     bmi.bmiHeader.biHeight= -(long)surface->height;
     bmi.bmiHeader.biPlanes = 1;
-    bmi.bmiHeader.biBitCount = pf_get_bits(surface->format);
+    bmi.bmiHeader.biBitCount = util_format_get_bits(surface->format);
     bmi.bmiHeader.biCompression = BI_RGB;
     bmi.bmiHeader.biSizeImage = 0;
     bmi.bmiHeader.biXPelsPerMeter = 0;
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index 6084ded72d..b41b36f98f 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -834,7 +834,7 @@ decompress_with_blit(GLcontext * ctx, GLenum target, GLint level,
    /* copy/pack data into user buffer */
    if (st_equal_formats(stImage->pt->format, format, type)) {
       /* memcpy */
-      const uint bytesPerRow = width * pf_get_size(stImage->pt->format);
+      const uint bytesPerRow = width * util_format_get_size(stImage->pt->format);
       ubyte *map = screen->transfer_map(screen, tex_xfer);
       GLuint row;
       for (row = 0; row < height; row++) {
-- 
cgit v1.2.3


From 7679447b5835fd73ab44b3d77b12a034c95af5c5 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Wed, 2 Dec 2009 17:15:27 +0100
Subject: r300g, radeong: fix the CS overflow

---
 src/gallium/drivers/r300/r300_cs.h               | 2 +-
 src/gallium/drivers/r300/r300_emit.c             | 9 ++++++++-
 src/gallium/winsys/drm/radeon/core/radeon_r300.c | 5 +++--
 3 files changed, 12 insertions(+), 4 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h
index 8b100375fd..9fcf3ab538 100644
--- a/src/gallium/drivers/r300/r300_cs.h
+++ b/src/gallium/drivers/r300/r300_cs.h
@@ -55,7 +55,7 @@
     int cs_count = 0;
 
 #define CHECK_CS(size) \
-    cs_winsys->check_cs(cs_winsys, (size))
+    assert(cs_winsys->check_cs(cs_winsys, (size)))
 
 #define BEGIN_CS(size) do { \
     CHECK_CS(size); \
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index a479842f9e..3bb42f9e43 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -871,10 +871,17 @@ void r300_emit_dirty_state(struct r300_context* r300)
         return;
     }
 
+    /* Check size of CS. */
+    /* Make sure we have at least 8*1024 spare dwords. */
+    /* XXX It would be nice to know the number of dwords we really need to
+     * XXX emit. */
+    if (!r300->winsys->check_cs(r300->winsys, 8*1024)) {
+        r300->context.flush(&r300->context, 0, NULL);
+    }
+
     /* Clean out BOs. */
     r300->winsys->reset_bos(r300->winsys);
 
-    /* XXX check size */
 validate:
     /* Color buffers... */
     for (i = 0; i < r300->framebuffer_state.nr_cbufs; i++) {
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c
index 7362279b77..ba0596c30d 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c
@@ -52,8 +52,9 @@ static boolean radeon_validate(struct radeon_winsys* winsys)
 
 static boolean radeon_check_cs(struct radeon_winsys* winsys, int size)
 {
-    /* XXX check size here, lazy ass! */
-    return radeon_validate(winsys);
+    struct radeon_cs* cs = winsys->priv->cs;
+
+    return radeon_validate(winsys) && cs->cdw + size <= cs->ndw;
 }
 
 static void radeon_begin_cs(struct radeon_winsys* winsys,
-- 
cgit v1.2.3


From 042b524d48ebb15215430149b9b1653f4b46dee3 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Fri, 4 Dec 2009 15:54:29 +0100
Subject: radeong: flush CS if a buffer being mapped is referenced by it

Also, overlapping occlusion queries seems to work now.
---
 src/gallium/drivers/r300/r300_emit.c               | 2 --
 src/gallium/winsys/drm/radeon/core/radeon_buffer.c | 5 +++++
 2 files changed, 5 insertions(+), 2 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 3bb42f9e43..60be03f54f 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -382,8 +382,6 @@ static void r300_emit_query_start(struct r300_context *r300)
     if (!query)
 	return;
 
-    /* XXX This will almost certainly not return good results
-     * for overlapping queries. */
     BEGIN_CS(4);
     if (caps->family == CHIP_FAMILY_RV530) {
         OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
index 0ca7b39255..2a8daed051 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
@@ -140,10 +140,15 @@ static void *radeon_buffer_map(struct pipe_winsys *ws,
                                struct pipe_buffer *buffer,
                                unsigned flags)
 {
+    struct radeon_winsys_priv *priv = ((struct radeon_winsys *)ws)->priv;
     struct radeon_pipe_buffer *radeon_buffer =
         (struct radeon_pipe_buffer*)buffer;
     int write = 0;
 
+    if (radeon_bo_is_referenced_by_cs(radeon_buffer->bo, priv->cs)) {
+        priv->cs->space_flush_fn(priv->cs->space_flush_data);
+    }
+
     if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) {
         uint32_t domain;
 
-- 
cgit v1.2.3


From 56a4342a0493ad1d502d4791ab941ef171d36e60 Mon Sep 17 00:00:00 2001
From: Michel Dänzer <daenzer@vmware.com>
Date: Sat, 5 Dec 2009 17:48:00 +0100
Subject: r300g: Need to emit a hardware scissor rectangle even if scissor is
 disabled.

Just make it cover the whole framebuffer in that case. Otherwise the kernel CS
checker may complain, e.g. running progs/demos/gearbox. That runs fast now
here, but doesn't look right yet.
---
 src/gallium/drivers/r300/r300_context.h |  2 ++
 src/gallium/drivers/r300/r300_emit.c    |  9 +++++++--
 src/gallium/drivers/r300/r300_state.c   | 19 +++++++++++++++++++
 3 files changed, 28 insertions(+), 2 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index dd3f6ac143..11cd9f855f 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -101,6 +101,8 @@ struct r300_sampler_state {
 struct r300_scissor_state {
     uint32_t scissor_top_left;     /* R300_SC_SCISSORS_TL: 0x43e0 */
     uint32_t scissor_bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */
+    uint32_t no_scissor_top_left;     /* R300_SC_SCISSORS_TL: 0x43e0 */
+    uint32_t no_scissor_bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */
 };
 
 struct r300_texture_state {
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 60be03f54f..04dca29216 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -570,8 +570,13 @@ void r300_emit_scissor_state(struct r300_context* r300,
 
     BEGIN_CS(3);
     OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL, 2);
-    OUT_CS(scissor->scissor_top_left);
-    OUT_CS(scissor->scissor_bottom_right);
+    if (r300->rs_state->rs.scissor) {
+       OUT_CS(scissor->scissor_top_left);
+       OUT_CS(scissor->scissor_bottom_right);
+    } else {
+       OUT_CS(scissor->no_scissor_top_left);
+       OUT_CS(scissor->no_scissor_bottom_right);
+    }
     END_CS;
 }
 
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 442af70e14..2bc2b79c02 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -302,6 +302,25 @@ static void
     r300->framebuffer_state = *state;
 
     r300->dirty_state |= R300_NEW_FRAMEBUFFERS;
+
+    if (r300_screen(r300->context.screen)->caps->is_r500) {
+        r300->scissor_state->no_scissor_top_left =
+            (0 << R300_SCISSORS_X_SHIFT) |
+            (0 << R300_SCISSORS_Y_SHIFT);
+        r300->scissor_state->no_scissor_bottom_right =
+            ((state->width - 1) << R300_SCISSORS_X_SHIFT) |
+            ((state->height - 1) << R300_SCISSORS_Y_SHIFT);
+    } else {
+        /* Offset of 1440 in non-R500 chipsets. */
+        r300->scissor_state->no_scissor_top_left =
+            ((0 + 1440) << R300_SCISSORS_X_SHIFT) |
+            ((0 + 1440) << R300_SCISSORS_Y_SHIFT);
+        r300->scissor_state->no_scissor_bottom_right =
+            (((state->width - 1) + 1440) << R300_SCISSORS_X_SHIFT) |
+            (((state->height - 1) + 1440) << R300_SCISSORS_Y_SHIFT);
+    }
+
+    r300->dirty_state |= R300_NEW_SCISSOR;
 }
 
 /* Create fragment shader state. */
-- 
cgit v1.2.3


From e1380cae885df37d4a211d0271f59487d9f2db78 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Sat, 5 Dec 2009 19:17:20 +0100
Subject: r300g: remove redundant code and clean up

---
 src/gallium/drivers/r300/r300_context.h | 11 +++--
 src/gallium/drivers/r300/r300_emit.c    | 19 +++++----
 src/gallium/drivers/r300/r300_state.c   | 73 +++++++++++++++++----------------
 3 files changed, 57 insertions(+), 46 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 11cd9f855f..23ea32c57e 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -98,11 +98,14 @@ struct r300_sampler_state {
     unsigned min_lod, max_lod;
 };
 
+struct r300_scissor_regs {
+    uint32_t top_left;     /* R300_SC_SCISSORS_TL: 0x43e0 */
+    uint32_t bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */
+};
+
 struct r300_scissor_state {
-    uint32_t scissor_top_left;     /* R300_SC_SCISSORS_TL: 0x43e0 */
-    uint32_t scissor_bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */
-    uint32_t no_scissor_top_left;     /* R300_SC_SCISSORS_TL: 0x43e0 */
-    uint32_t no_scissor_bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */
+    struct r300_scissor_regs framebuffer;
+    struct r300_scissor_regs scissor;
 };
 
 struct r300_texture_state {
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 04dca29216..dbf316a9b5 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -563,21 +563,26 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     END_CS;
 }
 
-void r300_emit_scissor_state(struct r300_context* r300,
-                             struct r300_scissor_state* scissor)
+static void r300_emit_scissor_regs(struct r300_context* r300,
+                                   struct r300_scissor_regs* scissor)
 {
     CS_LOCALS(r300);
 
     BEGIN_CS(3);
     OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL, 2);
+    OUT_CS(scissor->top_left);
+    OUT_CS(scissor->bottom_right);
+    END_CS;
+}
+
+void r300_emit_scissor_state(struct r300_context* r300,
+                             struct r300_scissor_state* scissor)
+{
     if (r300->rs_state->rs.scissor) {
-       OUT_CS(scissor->scissor_top_left);
-       OUT_CS(scissor->scissor_bottom_right);
+        r300_emit_scissor_regs(r300, &scissor->scissor);
     } else {
-       OUT_CS(scissor->no_scissor_top_left);
-       OUT_CS(scissor->no_scissor_bottom_right);
+        r300_emit_scissor_regs(r300, &scissor->framebuffer);
     }
-    END_CS;
 }
 
 void r300_emit_texture(struct r300_context* r300,
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 2bc2b79c02..d3233557ce 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -289,11 +289,34 @@ static void r300_set_edgeflags(struct pipe_context* pipe,
     /* XXX and even worse, I have no idea WTF the bitfield is */
 }
 
+static void r300_set_scissor_regs(const struct pipe_scissor_state* state,
+                                  struct r300_scissor_regs *scissor,
+                                  boolean is_r500)
+{
+    if (is_r500) {
+        scissor->top_left =
+            (state->minx << R300_SCISSORS_X_SHIFT) |
+            (state->miny << R300_SCISSORS_Y_SHIFT);
+        scissor->bottom_right =
+            ((state->maxx - 1) << R300_SCISSORS_X_SHIFT) |
+            ((state->maxy - 1) << R300_SCISSORS_Y_SHIFT);
+    } else {
+        /* Offset of 1440 in non-R500 chipsets. */
+        scissor->top_left =
+            ((state->minx + 1440) << R300_SCISSORS_X_SHIFT) |
+            ((state->miny + 1440) << R300_SCISSORS_Y_SHIFT);
+        scissor->bottom_right =
+            (((state->maxx - 1) + 1440) << R300_SCISSORS_X_SHIFT) |
+            (((state->maxy - 1) + 1440) << R300_SCISSORS_Y_SHIFT);
+    }
+}
+
 static void
     r300_set_framebuffer_state(struct pipe_context* pipe,
                                const struct pipe_framebuffer_state* state)
 {
     struct r300_context* r300 = r300_context(pipe);
+    struct pipe_scissor_state scissor;
 
     if (r300->draw) {
         draw_flush(r300->draw);
@@ -301,26 +324,17 @@ static void
 
     r300->framebuffer_state = *state;
 
-    r300->dirty_state |= R300_NEW_FRAMEBUFFERS;
+    scissor.minx = scissor.miny = 0;
+    scissor.maxx = state->width;
+    scissor.maxy = state->height;
+    r300_set_scissor_regs(&scissor, &r300->scissor_state->framebuffer,
+                          r300_screen(r300->context.screen)->caps->is_r500);
 
-    if (r300_screen(r300->context.screen)->caps->is_r500) {
-        r300->scissor_state->no_scissor_top_left =
-            (0 << R300_SCISSORS_X_SHIFT) |
-            (0 << R300_SCISSORS_Y_SHIFT);
-        r300->scissor_state->no_scissor_bottom_right =
-            ((state->width - 1) << R300_SCISSORS_X_SHIFT) |
-            ((state->height - 1) << R300_SCISSORS_Y_SHIFT);
-    } else {
-        /* Offset of 1440 in non-R500 chipsets. */
-        r300->scissor_state->no_scissor_top_left =
-            ((0 + 1440) << R300_SCISSORS_X_SHIFT) |
-            ((0 + 1440) << R300_SCISSORS_Y_SHIFT);
-        r300->scissor_state->no_scissor_bottom_right =
-            (((state->width - 1) + 1440) << R300_SCISSORS_X_SHIFT) |
-            (((state->height - 1) + 1440) << R300_SCISSORS_Y_SHIFT);
+    /* Don't rely on the order of states being set for the first time. */
+    if (!r300->rs_state || !r300->rs_state->rs.scissor) {
+        r300->dirty_state |= R300_NEW_SCISSOR;
     }
-
-    r300->dirty_state |= R300_NEW_SCISSOR;
+    r300->dirty_state |= R300_NEW_FRAMEBUFFERS;
 }
 
 /* Create fragment shader state. */
@@ -642,24 +656,13 @@ static void r300_set_scissor_state(struct pipe_context* pipe,
 {
     struct r300_context* r300 = r300_context(pipe);
 
-    if (r300_screen(r300->context.screen)->caps->is_r500) {
-        r300->scissor_state->scissor_top_left =
-            (state->minx << R300_SCISSORS_X_SHIFT) |
-            (state->miny << R300_SCISSORS_Y_SHIFT);
-        r300->scissor_state->scissor_bottom_right =
-            ((state->maxx - 1) << R300_SCISSORS_X_SHIFT) |
-            ((state->maxy - 1) << R300_SCISSORS_Y_SHIFT);
-    } else {
-        /* Offset of 1440 in non-R500 chipsets. */
-        r300->scissor_state->scissor_top_left =
-            ((state->minx + 1440) << R300_SCISSORS_X_SHIFT) |
-            ((state->miny + 1440) << R300_SCISSORS_Y_SHIFT);
-        r300->scissor_state->scissor_bottom_right =
-            (((state->maxx - 1) + 1440) << R300_SCISSORS_X_SHIFT) |
-            (((state->maxy - 1) + 1440) << R300_SCISSORS_Y_SHIFT);
-    }
+    r300_set_scissor_regs(state, &r300->scissor_state->scissor,
+                          r300_screen(r300->context.screen)->caps->is_r500);
 
-    r300->dirty_state |= R300_NEW_SCISSOR;
+    /* Don't rely on the order of states being set for the first time. */
+    if (!r300->rs_state || r300->rs_state->rs.scissor) {
+        r300->dirty_state |= R300_NEW_SCISSOR;
+    }
 }
 
 static void r300_set_viewport_state(struct pipe_context* pipe,
-- 
cgit v1.2.3


From 8cc570a48c2e8e18622027cbd76f16a746b430bc Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Wed, 9 Dec 2009 00:55:51 +0100
Subject: r300g: clean up r300_emit_aos

---
 src/gallium/drivers/r300/r300_cs.h   |  9 +++++
 src/gallium/drivers/r300/r300_emit.c | 71 +++++++++++++++++++++++-------------
 src/gallium/drivers/r300/r300_reg.h  |  5 +++
 3 files changed, 59 insertions(+), 26 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h
index 9fcf3ab538..d142fee050 100644
--- a/src/gallium/drivers/r300/r300_cs.h
+++ b/src/gallium/drivers/r300/r300_cs.h
@@ -115,6 +115,15 @@
     cs_count -= 3; \
 } while (0)
 
+#define OUT_CS_RELOC_NO_OFFSET(bo, rd, wd, flags) do { \
+    DBG(cs_context_copy, DBG_CS, "r300: writing relocation for buffer %p, " \
+            "domains (%d, %d, %d)\n", \
+        bo, rd, wd, flags); \
+    assert(bo); \
+    cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \
+    cs_count -= 2; \
+} while (0)
+
 #define END_CS do { \
     if (VERY_VERBOSE_CS) { \
         DBG(cs_context_copy, DBG_CS, "r300: END_CS in %s (%s:%d)\n", __FUNCTION__, \
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index dbf316a9b5..7620c73cac 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -1,5 +1,6 @@
 /*
  * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ * Copyright 2009 Marek Olšák <maraeo@gmail.com>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -623,50 +624,68 @@ void r300_emit_texture(struct r300_context* r300,
     END_CS;
 }
 
-/* XXX I can't read this and that's not good */
-void r300_emit_aos(struct r300_context* r300, unsigned offset)
+static boolean r300_validate_aos(struct r300_context *r300)
 {
     struct pipe_vertex_buffer *vbuf = r300->vertex_buffer;
     struct pipe_vertex_element *velem = r300->vertex_element;
-    CS_LOCALS(r300);
     int i;
-    unsigned aos_count = r300->vertex_element_count;
 
+    /* Check if formats and strides are aligned to the size of DWORD. */
+    for (i = 0; i < r300->vertex_element_count; i++) {
+        if (vbuf[velem[i].vertex_buffer_index].stride % 4 != 0 ||
+            pf_get_blocksize(velem[i].src_format) % 4 != 0) {
+            return FALSE;
+        }
+    }
+    return TRUE;
+}
+
+void r300_emit_aos(struct r300_context* r300, unsigned offset)
+{
+    struct pipe_vertex_buffer *vb1, *vb2, *vbuf = r300->vertex_buffer;
+    struct pipe_vertex_element *velem = r300->vertex_element;
+    int i;
+    unsigned size1, size2, aos_count = r300->vertex_element_count;
     unsigned packet_size = (aos_count * 3 + 1) / 2;
+    CS_LOCALS(r300);
+
+    /* XXX Move this checking to a more approriate place. */
+    if (!r300_validate_aos(r300)) {
+        /* XXX We should fallback using Draw. */
+        assert(0);
+    }
+
     BEGIN_CS(2 + packet_size + aos_count * 2);
     OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, packet_size);
     OUT_CS(aos_count);
+
     for (i = 0; i < aos_count - 1; i += 2) {
-        int buf_num1 = velem[i].vertex_buffer_index;
-        int buf_num2 = velem[i+1].vertex_buffer_index;
-        assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_blocksize(velem[i].src_format) % 4 == 0);
-        assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_blocksize(velem[i+1].src_format) % 4 == 0);
-        OUT_CS((pf_get_blocksize(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) |
-               (pf_get_blocksize(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22));
-        OUT_CS(vbuf[buf_num1].buffer_offset + velem[i].src_offset +
-               offset * vbuf[buf_num1].stride);
-        OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset +
-               offset * vbuf[buf_num2].stride);
+        vb1 = &vbuf[velem[i].vertex_buffer_index];
+        vb2 = &vbuf[velem[i+1].vertex_buffer_index];
+        size1 = pf_get_blocksize(velem[i].src_format);
+        size2 = pf_get_blocksize(velem[i+1].src_format);
+
+        OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride) |
+               R300_VBPNTR_SIZE1(size2) | R300_VBPNTR_STRIDE1(vb2->stride));
+        OUT_CS(vb1->buffer_offset + velem[i].src_offset   + offset * vb1->stride);
+        OUT_CS(vb2->buffer_offset + velem[i+1].src_offset + offset * vb2->stride);
     }
+
     if (aos_count & 1) {
-        int buf_num = velem[i].vertex_buffer_index;
-        assert(vbuf[buf_num].stride % 4 == 0 && pf_get_blocksize(velem[i].src_format) % 4 == 0);
-        OUT_CS((pf_get_blocksize(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6));
-        OUT_CS(vbuf[buf_num].buffer_offset + velem[i].src_offset +
-               offset * vbuf[buf_num].stride);
+        vb1 = &vbuf[velem[i].vertex_buffer_index];
+        size1 = pf_get_blocksize(velem[i].src_format);
+
+        OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride));
+        OUT_CS(vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride);
     }
 
-    /* XXX bare CS reloc */
     for (i = 0; i < aos_count; i++) {
-        cs_winsys->write_cs_reloc(cs_winsys,
-                                  vbuf[velem[i].vertex_buffer_index].buffer,
-                                  RADEON_GEM_DOMAIN_GTT,
-                                  0,
-                                  0);
-        cs_count -= 2;
+        OUT_CS_RELOC_NO_OFFSET(vbuf[velem[i].vertex_buffer_index].buffer,
+                               RADEON_GEM_DOMAIN_GTT, 0, 0);
     }
     END_CS;
 }
+
 #if 0
 void r300_emit_draw_packet(struct r300_context* r300)
 {
diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h
index 85b1ea568a..c1ea87d11e 100644
--- a/src/gallium/drivers/r300/r300_reg.h
+++ b/src/gallium/drivers/r300/r300_reg.h
@@ -3293,6 +3293,11 @@ enum {
  */
 #define R300_PACKET3_3D_LOAD_VBPNTR         0x00002F00
 
+#   define R300_VBPNTR_SIZE0(x)    ((x) >> 2)
+#   define R300_VBPNTR_STRIDE0(x)  (((x) >> 2) << 8)
+#   define R300_VBPNTR_SIZE1(x)    (((x) >> 2) << 16)
+#   define R300_VBPNTR_STRIDE1(x)  (((x) >> 2) << 24)
+
 #define R300_PACKET3_INDX_BUFFER            0x00003300
 #    define R300_INDX_BUFFER_DST_SHIFT          0
 #    define R300_INDX_BUFFER_SKIP_SHIFT         16
-- 
cgit v1.2.3


From 87b822e024797ef2fdb51ec9364f21eeb4d07161 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Tue, 8 Dec 2009 04:55:32 +0100
Subject: r300g: make pow(0,0) return 1 instead of NaN in the R500 fragment
 shader

Unfortunately we can't fix this easily in the R300 fragment shader,
and it's probably not worth the effort.
---
 src/gallium/drivers/r300/r300_emit.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 7620c73cac..55bc2b3528 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -277,7 +277,7 @@ void r500_emit_fragment_program_code(struct r300_context* r300,
 
     BEGIN_CS(13 +
              ((code->inst_end + 1) * 6));
-    OUT_CS_REG(R500_US_CONFIG, 0);
+    OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
     OUT_CS_REG(R500_US_PIXSIZE, code->max_temp_idx);
     OUT_CS_REG(R500_US_CODE_RANGE,
                R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(code->inst_end));
-- 
cgit v1.2.3


From 6de7ac73bf027b9ace6f5f0c8063cbf724d95cee Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Tue, 8 Dec 2009 21:53:19 +0100
Subject: r300g: always disable unused colorbuffers

---
 src/gallium/drivers/r300/r300_emit.c | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 55bc2b3528..f784e1fa8e 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -331,7 +331,13 @@ void r300_emit_fb_state(struct r300_context* r300,
     int i;
     CS_LOCALS(r300);
 
-    BEGIN_CS((10 * fb->nr_cbufs) + (fb->zsbuf ? 10 : 0) + 4);
+    /* Shouldn't fail unless there is a bug in the state tracker. */
+    assert(fb->nr_cbufs <= 4);
+
+    BEGIN_CS((10 * fb->nr_cbufs) + (2 * (4 - fb->nr_cbufs)) +
+             (fb->zsbuf ? 10 : 0) + 4);
+
+    /* Flush and free renderbuffer caches. */
     OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT,
         R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS |
         R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
@@ -339,6 +345,7 @@ void r300_emit_fb_state(struct r300_context* r300,
         R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
         R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
 
+    /* Set up colorbuffers. */
     for (i = 0; i < fb->nr_cbufs; i++) {
         surf = fb->cbufs[i];
         tex = (struct r300_texture*)surf->texture;
@@ -356,6 +363,12 @@ void r300_emit_fb_state(struct r300_context* r300,
             r300_translate_out_fmt(surf->format));
     }
 
+    /* Disable unused colorbuffers. */
+    for (; i < 4; i++) {
+        OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i), R300_US_OUT_FMT_UNUSED);
+    }
+
+    /* Set up a zbuffer. */
     if (fb->zsbuf) {
         surf = fb->zsbuf;
         tex = (struct r300_texture*)surf->texture;
-- 
cgit v1.2.3


From cf85bf9cd0c168caed6210a896df285c3d86db03 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Mon, 14 Dec 2009 05:47:54 +0100
Subject: r300g: set the number of colorbuffers in RB3D_CCTL

---
 src/gallium/drivers/r300/r300_emit.c            | 5 ++++-
 src/gallium/drivers/r300/r300_reg.h             | 1 +
 src/gallium/drivers/r300/r300_state_invariant.c | 3 +--
 3 files changed, 6 insertions(+), 3 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index f784e1fa8e..9644efb717 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -335,7 +335,7 @@ void r300_emit_fb_state(struct r300_context* r300,
     assert(fb->nr_cbufs <= 4);
 
     BEGIN_CS((10 * fb->nr_cbufs) + (2 * (4 - fb->nr_cbufs)) +
-             (fb->zsbuf ? 10 : 0) + 4);
+             (fb->zsbuf ? 10 : 0) + 6);
 
     /* Flush and free renderbuffer caches. */
     OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT,
@@ -345,6 +345,9 @@ void r300_emit_fb_state(struct r300_context* r300,
         R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
         R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
 
+    /* Set the number of colorbuffers. */
+    OUT_CS_REG(R300_RB3D_CCTL, R300_RB3D_CCTL_NUM_MULTIWRITES(fb->nr_cbufs));
+
     /* Set up colorbuffers. */
     for (i = 0; i < fb->nr_cbufs; i++) {
         surf = fb->cbufs[i];
diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h
index c1ea87d11e..d8d08fbe26 100644
--- a/src/gallium/drivers/r300/r300_reg.h
+++ b/src/gallium/drivers/r300/r300_reg.h
@@ -2145,6 +2145,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 /* Unpipelined. */
 #define R300_RB3D_CCTL                      0x4e00
+#	define R300_RB3D_CCTL_NUM_MULTIWRITES(x)       (MAX2(((x)-1), 0) << 5)
 #	define R300_RB3D_CCTL_NUM_MULTIWRITES_1_BUFFER                (0 << 5)
 #	define R300_RB3D_CCTL_NUM_MULTIWRITES_2_BUFFERS               (1 << 5)
 #	define R300_RB3D_CCTL_NUM_MULTIWRITES_3_BUFFERS               (2 << 5)
diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c
index 46d1cb39b5..3320d43b27 100644
--- a/src/gallium/drivers/r300/r300_state_invariant.c
+++ b/src/gallium/drivers/r300/r300_state_invariant.c
@@ -84,7 +84,7 @@ void r300_emit_invariant_state(struct r300_context* r300)
     END_CS;
 
     /* XXX unsorted stuff from surface_fill */
-    BEGIN_CS(56 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0));
+    BEGIN_CS(54 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0));
     /* Flush PVS. */
     OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0);
 
@@ -123,7 +123,6 @@ void r300_emit_invariant_state(struct r300_context* r300)
     OUT_CS_REG(R300_SU_DEPTH_OFFSET, 0x00000000);
     OUT_CS_REG(R300_SC_HYPERZ, 0x0000001C);
     OUT_CS_REG(R300_SC_EDGERULE, 0x2DA49525);
-    OUT_CS_REG(R300_RB3D_CCTL, 0x00000000);
     OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, 0x00000000);
     if (caps->is_r500) {
         OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000);
-- 
cgit v1.2.3


From a6d701d1c6ed8e0a649d62104aeded8fb25c66d8 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Mon, 14 Dec 2009 05:29:12 +0100
Subject: r300g: if no colorbuffers are set, disable blending and set the color
 mask to 0

This seems to be the only way to disable the first colorbuffer.
---
 src/gallium/drivers/r300/r300_emit.c  | 13 ++++++++++---
 src/gallium/drivers/r300/r300_state.c |  1 +
 2 files changed, 11 insertions(+), 3 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 9644efb717..55e4f94afe 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -41,9 +41,16 @@ void r300_emit_blend_state(struct r300_context* r300,
     CS_LOCALS(r300);
     BEGIN_CS(8);
     OUT_CS_REG_SEQ(R300_RB3D_CBLEND, 3);
-    OUT_CS(blend->blend_control);
-    OUT_CS(blend->alpha_blend_control);
-    OUT_CS(blend->color_channel_mask);
+    if (r300->framebuffer_state.nr_cbufs) {
+        OUT_CS(blend->blend_control);
+        OUT_CS(blend->alpha_blend_control);
+        OUT_CS(blend->color_channel_mask);
+    } else {
+        OUT_CS(0);
+        OUT_CS(0);
+        OUT_CS(0);
+        /* XXX also disable fastfill here once it's supported */
+    }
     OUT_CS_REG(R300_RB3D_ROPCNTL, blend->rop);
     OUT_CS_REG(R300_RB3D_DITHER_CTL, blend->dither);
     END_CS;
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 3cfa2e63f9..91cf972ede 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -339,6 +339,7 @@ static void
         r300->dirty_state |= R300_NEW_SCISSOR;
     }
     r300->dirty_state |= R300_NEW_FRAMEBUFFERS;
+    r300->dirty_state |= R300_NEW_BLEND;
 }
 
 /* Create fragment shader state. */
-- 
cgit v1.2.3


From 2ddee2cfbcef59128b2a251d5391ddc2b4aea4bb Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Tue, 15 Dec 2009 05:35:03 +0100
Subject: r300g: fix emission of which textures are enabled

It fixes most of the "Bad CS" issues in piglit/texCombine and piglit/fbo.
Some other issues of this kind will get fixed in the kernel soon (depth-only
rendering, S3TC, and RGTC).
---
 src/gallium/drivers/r300/r300_emit.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 55e4f94afe..55c8aa07bd 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -883,10 +883,21 @@ void r300_emit_viewport_state(struct r300_context* r300,
 
 void r300_emit_texture_count(struct r300_context* r300)
 {
+    uint32_t tx_enable = 0;
+    int i;
     CS_LOCALS(r300);
 
+    /* Notice that texture_count and sampler_count are just sizes
+     * of the respective arrays. We still have to check for the individual
+     * elements. */
+    for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) {
+        if (r300->textures[i]) {
+            tx_enable |= 1 << i;
+        }
+    }
+
     BEGIN_CS(2);
-    OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1);
+    OUT_CS_REG(R300_TX_ENABLE, tx_enable);
     END_CS;
 
 }
-- 
cgit v1.2.3


From b1ed72ebe2599ec178f51d86fd42f26486b9a19b Mon Sep 17 00:00:00 2001
From: Michal Krol <michal@vmware.com>
Date: Thu, 17 Dec 2009 23:41:57 +0100
Subject: Move the remaining format pf_get_* functions to u_format.h.

Previously they depended on format blocks, but after removing those
they started depending on format encoding.
---
 progs/rbug/bin_to_bmp.c                            |  13 +-
 src/gallium/auxiliary/util/u_blitter.c             |   4 +-
 src/gallium/auxiliary/util/u_debug.c               |   7 +-
 src/gallium/auxiliary/util/u_format.h              |  83 ++++++++++++-
 src/gallium/auxiliary/util/u_gen_mipmap.c          |  12 +-
 src/gallium/auxiliary/util/u_rect.c                |  21 ++--
 src/gallium/auxiliary/util/u_tile.c                |   9 +-
 src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c   |   4 +-
 src/gallium/drivers/cell/ppu/cell_texture.c        |  20 +--
 src/gallium/drivers/i915/i915_surface.c            |  19 +--
 src/gallium/drivers/i915/i915_texture.c            |  75 ++++++------
 src/gallium/drivers/llvmpipe/lp_setup.c            |   3 +-
 src/gallium/drivers/llvmpipe/lp_texture.c          |  18 +--
 src/gallium/drivers/nv04/nv04_surface_2d.c         |  13 +-
 src/gallium/drivers/nv04/nv04_transfer.c           |   3 +-
 src/gallium/drivers/nv10/nv10_miptree.c            |   5 +-
 src/gallium/drivers/nv10/nv10_transfer.c           |   3 +-
 src/gallium/drivers/nv20/nv20_miptree.c            |   5 +-
 src/gallium/drivers/nv20/nv20_transfer.c           |   3 +-
 src/gallium/drivers/nv30/nv30_miptree.c            |   5 +-
 src/gallium/drivers/nv30/nv30_transfer.c           |   3 +-
 src/gallium/drivers/nv40/nv40_miptree.c            |   5 +-
 src/gallium/drivers/nv40/nv40_transfer.c           |   3 +-
 src/gallium/drivers/nv50/nv50_miptree.c            |   9 +-
 src/gallium/drivers/nv50/nv50_transfer.c           |  23 ++--
 src/gallium/drivers/r300/r300_emit.c               |   8 +-
 src/gallium/drivers/r300/r300_screen.c             |   5 +-
 src/gallium/drivers/r300/r300_texture.c            |   8 +-
 src/gallium/drivers/softpipe/sp_texture.c          |  14 +--
 src/gallium/drivers/softpipe/sp_tile_cache.c       |   2 +-
 src/gallium/drivers/svga/svga_screen_texture.c     |  14 +--
 src/gallium/drivers/svga/svga_state_vs.c           |   2 +-
 src/gallium/drivers/trace/tr_rbug.c                |  15 +--
 src/gallium/drivers/trace/tr_screen.c              |   3 +-
 src/gallium/include/pipe/p_format.h                | 134 ---------------------
 src/gallium/state_trackers/python/st_sample.c      |  15 +--
 .../state_trackers/python/st_softpipe_winsys.c     |   4 +-
 .../winsys/drm/nouveau/drm/nouveau_drm_api.c       |   2 +-
 src/gallium/winsys/drm/radeon/core/radeon_buffer.c |   5 +-
 src/gallium/winsys/egl_xlib/sw_winsys.c            |   4 +-
 src/gallium/winsys/g3dvl/xlib/xsp_winsys.c         |   4 +-
 src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c       |   4 +-
 src/gallium/winsys/gdi/gdi_softpipe_winsys.c       |   8 +-
 src/gallium/winsys/xlib/xlib_cell.c                |   4 +-
 src/gallium/winsys/xlib/xlib_llvmpipe.c            |  10 +-
 src/gallium/winsys/xlib/xlib_softpipe.c            |  10 +-
 src/mesa/state_tracker/st_cb_drawpixels.c          |   4 +-
 src/mesa/state_tracker/st_cb_fbo.c                 |   4 +-
 src/mesa/state_tracker/st_cb_texture.c             |  26 ++--
 src/mesa/state_tracker/st_gen_mipmap.c             |   5 +-
 50 files changed, 322 insertions(+), 360 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/progs/rbug/bin_to_bmp.c b/progs/rbug/bin_to_bmp.c
index cdae3486ce..03ff622fee 100644
--- a/progs/rbug/bin_to_bmp.c
+++ b/progs/rbug/bin_to_bmp.c
@@ -25,6 +25,7 @@
 #include "pipe/p_compiler.h"
 #include "pipe/p_format.h"
 #include "pipe/p_state.h"
+#include "util/u_format.h"
 #include "util/u_memory.h"
 #include "util/u_debug.h"
 #include "util/u_network.h"
@@ -54,10 +55,7 @@ static void dump(unsigned width, unsigned height,
                  unsigned src_stride, enum pipe_format src_format,
                  uint8_t *data, unsigned src_size)
 {
-   struct pipe_format_block src_block;
-
    enum pipe_format dst_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
-   struct pipe_format_block dst_block;
    unsigned dst_stride;
    unsigned dst_size;
    float *rgba;
@@ -65,14 +63,11 @@ static void dump(unsigned width, unsigned height,
    char filename[512];
 
    {
-      pf_get_block(src_format, &src_block);
-      assert(src_stride >= pf_get_stride(&src_block, width));
-      assert(src_size >= pf_get_2d_size(&src_block, src_stride, width));
+      assert(src_stride >= util_format_get_stride(src_format, width));
    }
    {
-      pf_get_block(dst_format, &dst_block);
-      dst_stride = pf_get_stride(&dst_block, width);
-      dst_size = pf_get_2d_size(&dst_block, dst_stride, width);
+      dst_stride = util_format_get_stride(dst_format, width);
+      dst_size = util_format_get_2d_size(dst_format, dst_stride, width);
       rgba = MALLOC(dst_size);
    }
 
diff --git a/src/gallium/auxiliary/util/u_blitter.c b/src/gallium/auxiliary/util/u_blitter.c
index 895af2c8d0..0242b79615 100644
--- a/src/gallium/auxiliary/util/u_blitter.c
+++ b/src/gallium/auxiliary/util/u_blitter.c
@@ -579,8 +579,8 @@ void util_blitter_copy(struct blitter_context *blitter,
    if (!dst->texture || !src->texture)
       return;
 
-   is_depth = pf_get_component_bits(src->format, PIPE_FORMAT_COMP_Z) != 0;
-   is_stencil = pf_get_component_bits(src->format, PIPE_FORMAT_COMP_S) != 0;
+   is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0;
+   is_stencil = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 1) != 0;
    dst_tex_usage = is_depth || is_stencil ? PIPE_TEXTURE_USAGE_DEPTH_STENCIL :
                                             PIPE_TEXTURE_USAGE_RENDER_TARGET;
 
diff --git a/src/gallium/auxiliary/util/u_debug.c b/src/gallium/auxiliary/util/u_debug.c
index 40633574b0..27e0b0d159 100644
--- a/src/gallium/auxiliary/util/u_debug.c
+++ b/src/gallium/auxiliary/util/u_debug.c
@@ -64,6 +64,7 @@
 #include "pipe/p_format.h" 
 #include "pipe/p_state.h" 
 #include "pipe/p_inlines.h" 
+#include "util/u_format.h"
 #include "util/u_memory.h" 
 #include "util/u_string.h" 
 #include "util/u_stream.h" 
@@ -670,9 +671,9 @@ void debug_dump_surface(const char *prefix,
    
    debug_dump_image(prefix, 
                     texture->format,
-                    pf_get_blocksize(texture->format), 
-                    pf_get_nblocksx(texture->format, transfer->width),
-                    pf_get_nblocksy(texture->format, transfer->height),
+                    util_format_get_blocksize(texture->format), 
+                    util_format_get_nblocksx(texture->format, transfer->width),
+                    util_format_get_nblocksy(texture->format, transfer->height),
                     transfer->stride,
                     data);
    
diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h
index 72da2a44c7..97e4d959bc 100644
--- a/src/gallium/auxiliary/util/u_format.h
+++ b/src/gallium/auxiliary/util/u_format.h
@@ -200,7 +200,7 @@ util_format_is_depth_and_stencil(enum pipe_format format)
  * Return total bits needed for the pixel format.
  */
 static INLINE uint
-util_format_get_bits(enum pipe_format format)
+util_format_get_blocksizebits(enum pipe_format format)
 {
    const struct util_format_description *desc = util_format_description(format);
 
@@ -216,15 +216,92 @@ util_format_get_bits(enum pipe_format format)
  * Return bytes per pixel for the given format.
  */
 static INLINE uint
-util_format_get_size(enum pipe_format format)
+util_format_get_blocksize(enum pipe_format format)
 {
-   uint bits = util_format_get_bits(format);
+   uint bits = util_format_get_blocksizebits(format);
 
    assert(bits % 8 == 0);
 
    return bits / 8;
 }
 
+static INLINE uint
+util_format_get_blockwidth(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(format);
+   if (!format) {
+      return 1;
+   }
+
+   switch (desc->layout) {
+   case UTIL_FORMAT_LAYOUT_YUV:
+      return 2;
+   case UTIL_FORMAT_LAYOUT_DXT:
+      return 4;
+   default:
+      return 1;
+   }
+}
+
+static INLINE uint
+util_format_get_blockheight(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(format);
+   if (!format) {
+      return 1;
+   }
+
+   switch (desc->layout) {
+   case UTIL_FORMAT_LAYOUT_DXT:
+      return 4;
+   default:
+      return 1;
+   }
+}
+
+static INLINE unsigned
+util_format_get_nblocksx(enum pipe_format format,
+                         unsigned x)
+{
+   unsigned blockwidth = util_format_get_blockwidth(format);
+   return (x + blockwidth - 1) / blockwidth;
+}
+
+static INLINE unsigned
+util_format_get_nblocksy(enum pipe_format format,
+                         unsigned y)
+{
+   unsigned blockheight = util_format_get_blockheight(format);
+   return (y + blockheight - 1) / blockheight;
+}
+
+static INLINE unsigned
+util_format_get_nblocks(enum pipe_format format,
+                        unsigned width,
+                        unsigned height)
+{
+   return util_format_get_nblocksx(format, width) * util_format_get_nblocksy(format, height);
+}
+
+static INLINE size_t
+util_format_get_stride(enum pipe_format format,
+                       unsigned width)
+{
+   return util_format_get_nblocksx(format, width) * util_format_get_blocksize(format);
+}
+
+static INLINE size_t
+util_format_get_2d_size(enum pipe_format format,
+                        size_t stride,
+                        unsigned height)
+{
+   return util_format_get_nblocksy(format, height) * stride;
+}
+
 static INLINE uint
 util_format_get_component_bits(enum pipe_format format,
                                enum util_format_colorspace colorspace,
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
index 2931dfac47..0dad6ccbc0 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.c
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -998,7 +998,7 @@ reduce_2d(enum pipe_format pformat,
 {
    enum dtype datatype;
    uint comps;
-   const int bpt = util_format_get_size(pformat);
+   const int bpt = util_format_get_blocksize(pformat);
    const ubyte *srcA, *srcB;
    ubyte *dst;
    int row;
@@ -1037,7 +1037,7 @@ reduce_3d(enum pipe_format pformat,
           int dstWidth, int dstHeight, int dstDepth,
           int dstRowStride, ubyte *dstPtr)
 {
-   const int bpt = util_format_get_size(pformat);
+   const int bpt = util_format_get_blocksize(pformat);
    const int border = 0;
    int img, row;
    int bytesPerSrcImage, bytesPerDstImage;
@@ -1161,8 +1161,8 @@ make_2d_mipmap(struct gen_mipmap_state *ctx,
    const uint zslice = 0;
    uint dstLevel;
    
-   assert(pf_get_blockwidth(pt->format) == 1);
-   assert(pf_get_blockheight(pt->format) == 1);
+   assert(util_format_get_blockwidth(pt->format) == 1);
+   assert(util_format_get_blockheight(pt->format) == 1);
 
    for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
       const uint srcLevel = dstLevel - 1;
@@ -1206,8 +1206,8 @@ make_3d_mipmap(struct gen_mipmap_state *ctx,
    struct pipe_screen *screen = pipe->screen;
    uint dstLevel, zslice = 0;
 
-   assert(pf_get_blockwidth(pt->format) == 1);
-   assert(pf_get_blockheight(pt->format) == 1);
+   assert(util_format_get_blockwidth(pt->format) == 1);
+   assert(util_format_get_blockheight(pt->format) == 1);
 
    for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
       const uint srcLevel = dstLevel - 1;
diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c
index 72725b59d2..298fbacecb 100644
--- a/src/gallium/auxiliary/util/u_rect.c
+++ b/src/gallium/auxiliary/util/u_rect.c
@@ -34,6 +34,7 @@
 #include "pipe/p_format.h"
 #include "pipe/p_context.h"
 #include "pipe/p_screen.h"
+#include "util/u_format.h"
 #include "util/u_rect.h"
 
 
@@ -57,9 +58,9 @@ util_copy_rect(ubyte * dst,
 {
    unsigned i;
    int src_stride_pos = src_stride < 0 ? -src_stride : src_stride;
-   int blocksize = pf_get_blocksize(format);
-   int blockwidth = pf_get_blockwidth(format);
-   int blockheight = pf_get_blockheight(format);
+   int blocksize = util_format_get_blocksize(format);
+   int blockwidth = util_format_get_blockwidth(format);
+   int blockheight = util_format_get_blockheight(format);
 
    assert(blocksize > 0);
    assert(blockwidth > 0);
@@ -105,9 +106,9 @@ util_fill_rect(ubyte * dst,
 {
    unsigned i, j;
    unsigned width_size;
-   int blocksize = pf_get_blocksize(format);
-   int blockwidth = pf_get_blockwidth(format);
-   int blockheight = pf_get_blockheight(format);
+   int blocksize = util_format_get_blocksize(format);
+   int blockwidth = util_format_get_blockwidth(format);
+   int blockheight = util_format_get_blockheight(format);
 
    assert(blocksize > 0);
    assert(blockwidth > 0);
@@ -203,9 +204,9 @@ util_surface_copy(struct pipe_context *pipe,
                                         PIPE_TRANSFER_WRITE,
                                         dst_x, dst_y, w, h);
 
-   assert(pf_get_blocksize(dst_format) == pf_get_blocksize(src_format));
-   assert(pf_get_blockwidth(dst_format) == pf_get_blockwidth(src_format));
-   assert(pf_get_blockheight(dst_format) == pf_get_blockheight(src_format));
+   assert(util_format_get_blocksize(dst_format) == util_format_get_blocksize(src_format));
+   assert(util_format_get_blockwidth(dst_format) == util_format_get_blockwidth(src_format));
+   assert(util_format_get_blockheight(dst_format) == util_format_get_blockheight(src_format));
 
    src_map = pipe->screen->transfer_map(screen, src_trans);
    dst_map = pipe->screen->transfer_map(screen, dst_trans);
@@ -270,7 +271,7 @@ util_surface_fill(struct pipe_context *pipe,
    if (dst_map) {
       assert(dst_trans->stride > 0);
 
-      switch (pf_get_blocksize(dst_trans->texture->format)) {
+      switch (util_format_get_blocksize(dst_trans->texture->format)) {
       case 1:
       case 2:
       case 4:
diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c
index 88c9a1f097..5b8dd1abb9 100644
--- a/src/gallium/auxiliary/util/u_tile.c
+++ b/src/gallium/auxiliary/util/u_tile.c
@@ -34,6 +34,7 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 #include "util/u_rect.h"
@@ -52,7 +53,7 @@ pipe_get_tile_raw(struct pipe_transfer *pt,
    const void *src;
 
    if (dst_stride == 0)
-      dst_stride = pf_get_stride(pt->texture->format, w);
+      dst_stride = util_format_get_stride(pt->texture->format, w);
 
    if (pipe_clip_tile(x, y, &w, &h, pt))
       return;
@@ -81,7 +82,7 @@ pipe_put_tile_raw(struct pipe_transfer *pt,
    enum pipe_format format = pt->texture->format;
 
    if (src_stride == 0)
-      src_stride = pf_get_stride(format, w);
+      src_stride = util_format_get_stride(format, w);
 
    if (pipe_clip_tile(x, y, &w, &h, pt))
       return;
@@ -1275,7 +1276,7 @@ pipe_get_tile_rgba(struct pipe_transfer *pt,
    if (pipe_clip_tile(x, y, &w, &h, pt))
       return;
 
-   packed = MALLOC(pf_get_nblocks(format, w, h) * pf_get_blocksize(format));
+   packed = MALLOC(util_format_get_nblocks(format, w, h) * util_format_get_blocksize(format));
 
    if (!packed)
       return;
@@ -1303,7 +1304,7 @@ pipe_put_tile_rgba(struct pipe_transfer *pt,
    if (pipe_clip_tile(x, y, &w, &h, pt))
       return;
 
-   packed = MALLOC(pf_get_nblocks(format, w, h) * pf_get_blocksize(format));
+   packed = MALLOC(util_format_get_nblocks(format, w, h) * util_format_get_blocksize(format));
 
    if (!packed)
       return;
diff --git a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
index 8f0185b42a..ab196c21f8 100644
--- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
+++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
@@ -1444,7 +1444,7 @@ grab_blocks(struct vl_mpeg12_mc_renderer *r, unsigned mbx, unsigned mby,
    assert(r);
    assert(blocks);
 
-   tex_pitch = r->tex_transfer[0]->stride / pf_get_blocksize(r->tex_transfer[0]->texture->format);
+   tex_pitch = r->tex_transfer[0]->stride / util_format_get_blocksize(r->tex_transfer[0]->texture->format);
    texels = r->texels[0] + mbpy * tex_pitch + mbpx;
 
    for (y = 0; y < 2; ++y) {
@@ -1483,7 +1483,7 @@ grab_blocks(struct vl_mpeg12_mc_renderer *r, unsigned mbx, unsigned mby,
    mbpy /= 2;
 
    for (tb = 0; tb < 2; ++tb) {
-      tex_pitch = r->tex_transfer[tb + 1]->stride / pf_get_blocksize(r->tex_transfer[tb + 1]->texture->format);
+      tex_pitch = r->tex_transfer[tb + 1]->stride / util_format_get_blocksize(r->tex_transfer[tb + 1]->texture->format);
       texels = r->texels[tb + 1] + mbpy * tex_pitch + mbpx;
 
       if ((cbp >> (1 - tb)) & 1) {
diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c
index bc2e625f54..998944f77a 100644
--- a/src/gallium/drivers/cell/ppu/cell_texture.c
+++ b/src/gallium/drivers/cell/ppu/cell_texture.c
@@ -67,11 +67,11 @@ cell_texture_layout(struct cell_texture *ct)
       w_tile = align(width, TILE_SIZE);
       h_tile = align(height, TILE_SIZE);
 
-      ct->stride[level] = pf_get_stride(pt->format, w_tile);
+      ct->stride[level] = util_format_get_stride(pt->format, w_tile);
 
       ct->level_offset[level] = ct->buffer_size;
 
-      size = ct->stride[level] * pf_get_nblocksy(pt->format, h_tile);
+      size = ct->stride[level] * util_format_get_nblocksy(pt->format, h_tile);
       if (pt->target == PIPE_TEXTURE_CUBE)
          size *= 6;
       else
@@ -283,11 +283,11 @@ cell_get_tex_surface(struct pipe_screen *screen,
 
       if (pt->target == PIPE_TEXTURE_CUBE) {
          unsigned h_tile = align(ps->height, TILE_SIZE);
-         ps->offset += face * pf_get_nblocksy(ps->format, h_tile) * ct->stride[level];
+         ps->offset += face * util_format_get_nblocksy(ps->format, h_tile) * ct->stride[level];
       }
       else if (pt->target == PIPE_TEXTURE_3D) {
          unsigned h_tile = align(ps->height, TILE_SIZE);
-         ps->offset += zslice * pf_get_nblocksy(ps->format, h_tile) * ct->stride[level];
+         ps->offset += zslice * util_format_get_nblocksy(ps->format, h_tile) * ct->stride[level];
       }
       else {
          assert(face == 0);
@@ -342,11 +342,11 @@ cell_get_tex_transfer(struct pipe_screen *screen,
 
       if (texture->target == PIPE_TEXTURE_CUBE) {
          unsigned h_tile = align(u_minify(texture->height0, level), TILE_SIZE);
-         ctrans->offset += face * pf_get_nblocksy(texture->format, h_tile) * pt->stride;
+         ctrans->offset += face * util_format_get_nblocksy(texture->format, h_tile) * pt->stride;
       }
       else if (texture->target == PIPE_TEXTURE_3D) {
          unsigned h_tile = align(u_minify(texture->height0, level), TILE_SIZE);
-         ctrans->offset += zslice * pf_get_nblocksy(texture->format, h_tile) * pt->stride;
+         ctrans->offset += zslice * util_format_get_nblocksy(texture->format, h_tile) * pt->stride;
       }
       else {
          assert(face == 0);
@@ -399,8 +399,8 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer)
     * Create a buffer of ordinary memory for the linear texture.
     * This is the memory that the user will read/write.
     */
-   size = pf_get_stride(pt->format, align(texWidth, TILE_SIZE)) *
-          pf_get_nblocksy(pt->format, align(texHeight, TILE_SIZE));
+   size = util_format_get_stride(pt->format, align(texWidth, TILE_SIZE)) *
+          util_format_get_nblocksy(pt->format, align(texHeight, TILE_SIZE));
 
    ctrans->map = align_malloc(size, 16);
    if (!ctrans->map)
@@ -408,7 +408,7 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer)
 
    if (transfer->usage & PIPE_TRANSFER_READ) {
       /* need to untwiddle the texture to make a linear version */
-      const uint bpp = util_format_get_size(ct->base.format);
+      const uint bpp = util_format_get_blocksize(ct->base.format);
       if (bpp == 4) {
          const uint *src = (uint *) (ct->mapped + ctrans->offset);
          uint *dst = ctrans->map;
@@ -451,7 +451,7 @@ cell_transfer_unmap(struct pipe_screen *screen,
       /* The user wrote new texture data into the mapped buffer.
        * We need to convert the new linear data into the twiddled/tiled format.
        */
-      const uint bpp = util_format_get_size(ct->base.format);
+      const uint bpp = util_format_get_blocksize(ct->base.format);
       if (bpp == 4) {
          const uint *src = ctrans->map;
          uint *dst = (uint *) (ct->mapped + ctrans->offset);
diff --git a/src/gallium/drivers/i915/i915_surface.c b/src/gallium/drivers/i915/i915_surface.c
index 24e1024aaa..c693eb30e8 100644
--- a/src/gallium/drivers/i915/i915_surface.c
+++ b/src/gallium/drivers/i915/i915_surface.c
@@ -32,6 +32,7 @@
 #include "pipe/p_inlines.h"
 #include "pipe/p_inlines.h"
 #include "pipe/internal/p_winsys_screen.h"
+#include "util/u_format.h"
 #include "util/u_tile.h"
 #include "util/u_rect.h"
 
@@ -52,15 +53,15 @@ i915_surface_copy(struct pipe_context *pipe,
    struct pipe_texture *spt = &src_tex->base;
 
    assert( dst != src );
-   assert( pf_get_blocksize(dpt->format) == pf_get_blocksize(spt->format) );
-   assert( pf_get_blockwidth(dpt->format) == pf_get_blockwidth(spt->format) );
-   assert( pf_get_blockheight(dpt->format) == pf_get_blockheight(spt->format) );
-   assert( pf_get_blockwidth(dpt->format) == 1 );
-   assert( pf_get_blockheight(dpt->format) == 1 );
+   assert( util_format_get_blocksize(dpt->format) == util_format_get_blocksize(spt->format) );
+   assert( util_format_get_blockwidth(dpt->format) == util_format_get_blockwidth(spt->format) );
+   assert( util_format_get_blockheight(dpt->format) == util_format_get_blockheight(spt->format) );
+   assert( util_format_get_blockwidth(dpt->format) == 1 );
+   assert( util_format_get_blockheight(dpt->format) == 1 );
 
    i915_copy_blit( i915_context(pipe),
                    FALSE,
-                   pf_get_blocksize(dpt->format),
+                   util_format_get_blocksize(dpt->format),
                    (unsigned short) src_tex->stride, src_tex->buffer, src->offset,
                    (unsigned short) dst_tex->stride, dst_tex->buffer, dst->offset,
                    (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height );
@@ -76,11 +77,11 @@ i915_surface_fill(struct pipe_context *pipe,
    struct i915_texture *tex = (struct i915_texture *)dst->texture;
    struct pipe_texture *pt = &tex->base;
 
-   assert(pf_get_blockwidth(pt->format) == 1);
-   assert(pf_get_blockheight(pt->format) == 1);
+   assert(util_format_get_blockwidth(pt->format) == 1);
+   assert(util_format_get_blockheight(pt->format) == 1);
 
    i915_fill_blit( i915_context(pipe),
-                   pf_get_blocksize(pt->format),
+                   util_format_get_blocksize(pt->format),
                    (unsigned short) tex->stride,
                    tex->buffer, dst->offset,
                    (short) dstx, (short) dsty,
diff --git a/src/gallium/drivers/i915/i915_texture.c b/src/gallium/drivers/i915/i915_texture.c
index b28b413771..50a9e19094 100644
--- a/src/gallium/drivers/i915/i915_texture.c
+++ b/src/gallium/drivers/i915/i915_texture.c
@@ -35,6 +35,7 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "pipe/internal/p_winsys_screen.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 
@@ -129,7 +130,7 @@ i915_miptree_set_image_offset(struct i915_texture *tex,
 
    assert(img < tex->nr_images[level]);
 
-   tex->image_offset[level][img] = y * tex->stride + x * pf_get_blocksize(tex->base.format);
+   tex->image_offset[level][img] = y * tex->stride + x * util_format_get_blocksize(tex->base.format);
 
    /*
    printf("%s level %d img %d pos %d,%d image_offset %x\n",
@@ -151,7 +152,7 @@ i915_scanout_layout(struct i915_texture *tex)
 {
    struct pipe_texture *pt = &tex->base;
 
-   if (pt->last_level > 0 || pf_get_blocksize(pt->format) != 4)
+   if (pt->last_level > 0 || util_format_get_blocksize(pt->format) != 4)
       return FALSE;
 
    i915_miptree_set_level_info(tex, 0, 1,
@@ -161,18 +162,18 @@ i915_scanout_layout(struct i915_texture *tex)
    i915_miptree_set_image_offset(tex, 0, 0, 0, 0);
 
    if (pt->width0 >= 240) {
-      tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0));
-      tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8);
+      tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0));
+      tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8);
       tex->hw_tiled = INTEL_TILE_X;
    } else if (pt->width0 == 64 && pt->height0 == 64) {
-      tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0));
-      tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8);
+      tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0));
+      tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8);
    } else {
       return FALSE;
    }
 
    debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
-      pt->width0, pt->height0, pf_get_blocksize(pt->format),
+      pt->width0, pt->height0, util_format_get_blocksize(pt->format),
       tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy);
 
    return TRUE;
@@ -186,7 +187,7 @@ i915_display_target_layout(struct i915_texture *tex)
 {
    struct pipe_texture *pt = &tex->base;
 
-   if (pt->last_level > 0 || pf_get_blocksize(pt->format) != 4)
+   if (pt->last_level > 0 || util_format_get_blocksize(pt->format) != 4)
       return FALSE;
 
    /* fallback to normal textures for small textures */
@@ -199,12 +200,12 @@ i915_display_target_layout(struct i915_texture *tex)
                                1);
    i915_miptree_set_image_offset(tex, 0, 0, 0, 0);
 
-   tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0));
-   tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8);
+   tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0));
+   tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8);
    tex->hw_tiled = INTEL_TILE_X;
 
    debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
-      pt->width0, pt->height0, pf_get_blocksize(pt->format),
+      pt->width0, pt->height0, util_format_get_blocksize(pt->format),
       tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy);
 
    return TRUE;
@@ -217,7 +218,7 @@ i915_miptree_layout_2d(struct i915_texture *tex)
    unsigned level;
    unsigned width = pt->width0;
    unsigned height = pt->height0;
-   unsigned nblocksy = pf_get_nblocksy(pt->format, pt->width0);
+   unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->width0);
 
    /* used for scanouts that need special layouts */
    if (pt->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY)
@@ -229,7 +230,7 @@ i915_miptree_layout_2d(struct i915_texture *tex)
       if (i915_display_target_layout(tex))
          return;
 
-   tex->stride = align(pf_get_stride(pt->format, pt->width0), 4);
+   tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4);
    tex->total_nblocksy = 0;
 
    for (level = 0; level <= pt->last_level; level++) {
@@ -242,7 +243,7 @@ i915_miptree_layout_2d(struct i915_texture *tex)
 
       width = u_minify(width, 1);
       height = u_minify(height, 1);
-      nblocksy = pf_get_nblocksy(pt->format, height);
+      nblocksy = util_format_get_nblocksy(pt->format, height);
    }
 }
 
@@ -255,12 +256,12 @@ i915_miptree_layout_3d(struct i915_texture *tex)
    unsigned width = pt->width0;
    unsigned height = pt->height0;
    unsigned depth = pt->depth0;
-   unsigned nblocksy = pf_get_nblocksy(pt->format, pt->height0);
+   unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->height0);
    unsigned stack_nblocksy = 0;
 
    /* Calculate the size of a single slice. 
     */
-   tex->stride = align(pf_get_stride(pt->format, pt->width0), 4);
+   tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4);
 
    /* XXX: hardware expects/requires 9 levels at minimum.
     */
@@ -271,7 +272,7 @@ i915_miptree_layout_3d(struct i915_texture *tex)
 
       width = u_minify(width, 1);
       height = u_minify(height, 1);
-      nblocksy = pf_get_nblocksy(pt->format, height);
+      nblocksy = util_format_get_nblocksy(pt->format, height);
    }
 
    /* Fixup depth image_offsets: 
@@ -296,14 +297,14 @@ i915_miptree_layout_cube(struct i915_texture *tex)
 {
    struct pipe_texture *pt = &tex->base;
    unsigned width = pt->width0, height = pt->height0;
-   const unsigned nblocks = pf_get_nblocksx(pt->format, pt->width0);
+   const unsigned nblocks = util_format_get_nblocksx(pt->format, pt->width0);
    unsigned level;
    unsigned face;
 
    assert(width == height); /* cubemap images are square */
 
    /* double pitch for cube layouts */
-   tex->stride = align(nblocks * pf_get_blocksize(pt->format) * 2, 4);
+   tex->stride = align(nblocks * util_format_get_blocksize(pt->format) * 2, 4);
    tex->total_nblocksy = nblocks * 4;
 
    for (level = 0; level <= pt->last_level; level++) {
@@ -366,8 +367,8 @@ i945_miptree_layout_2d(struct i915_texture *tex)
    unsigned y = 0;
    unsigned width = pt->width0;
    unsigned height = pt->height0;
-   unsigned nblocksx = pf_get_nblocksx(pt->format, pt->width0);
-   unsigned nblocksy = pf_get_nblocksy(pt->format, pt->height0);
+   unsigned nblocksx = util_format_get_nblocksx(pt->format, pt->width0);
+   unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->height0);
 
    /* used for scanouts that need special layouts */
    if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY)
@@ -379,7 +380,7 @@ i945_miptree_layout_2d(struct i915_texture *tex)
       if (i915_display_target_layout(tex))
          return;
 
-   tex->stride = align(pf_get_stride(pt->format, pt->width0), 4);
+   tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4);
 
    /* May need to adjust pitch to accomodate the placement of
     * the 2nd mipmap level.  This occurs when the alignment
@@ -388,11 +389,11 @@ i945_miptree_layout_2d(struct i915_texture *tex)
     */
    if (pt->last_level > 0) {
       unsigned mip1_nblocksx 
-         = align(pf_get_nblocksx(pt->format, u_minify(width, 1)), align_x)
-         + pf_get_nblocksx(pt->format, u_minify(width, 2));
+         = align(util_format_get_nblocksx(pt->format, u_minify(width, 1)), align_x)
+         + util_format_get_nblocksx(pt->format, u_minify(width, 2));
 
       if (mip1_nblocksx > nblocksx)
-         tex->stride = mip1_nblocksx * pf_get_blocksize(pt->format);
+         tex->stride = mip1_nblocksx * util_format_get_blocksize(pt->format);
    }
 
    /* Pitch must be a whole number of dwords
@@ -422,8 +423,8 @@ i945_miptree_layout_2d(struct i915_texture *tex)
 
       width  = u_minify(width, 1);
       height = u_minify(height, 1);
-      nblocksx = pf_get_nblocksx(pt->format, width);
-      nblocksy = pf_get_nblocksy(pt->format, height);
+      nblocksx = util_format_get_nblocksx(pt->format, width);
+      nblocksy = util_format_get_nblocksy(pt->format, height);
    }
 }
 
@@ -434,16 +435,16 @@ i945_miptree_layout_3d(struct i915_texture *tex)
    unsigned width = pt->width0;
    unsigned height = pt->height0;
    unsigned depth = pt->depth0;
-   unsigned nblocksy = pf_get_nblocksy(pt->format, pt->width0);
+   unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->width0);
    unsigned pack_x_pitch, pack_x_nr;
    unsigned pack_y_pitch;
    unsigned level;
 
-   tex->stride = align(pf_get_stride(pt->format, pt->width0), 4);
+   tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4);
    tex->total_nblocksy = 0;
 
    pack_y_pitch = MAX2(nblocksy, 2);
-   pack_x_pitch = tex->stride / pf_get_blocksize(pt->format);
+   pack_x_pitch = tex->stride / util_format_get_blocksize(pt->format);
    pack_x_nr = 1;
 
    for (level = 0; level <= pt->last_level; level++) {
@@ -468,7 +469,7 @@ i945_miptree_layout_3d(struct i915_texture *tex)
       if (pack_x_pitch > 4) {
          pack_x_pitch >>= 1;
          pack_x_nr <<= 1;
-         assert(pack_x_pitch * pack_x_nr * pf_get_blocksize(pt->format) <= tex->stride);
+         assert(pack_x_pitch * pack_x_nr * util_format_get_blocksize(pt->format) <= tex->stride);
       }
 
       if (pack_y_pitch > 2) {
@@ -478,7 +479,7 @@ i945_miptree_layout_3d(struct i915_texture *tex)
       width = u_minify(width, 1);
       height = u_minify(height, 1);
       depth = u_minify(depth, 1);
-      nblocksy = pf_get_nblocksy(pt->format, height);
+      nblocksy = util_format_get_nblocksy(pt->format, height);
    }
 }
 
@@ -488,7 +489,7 @@ i945_miptree_layout_cube(struct i915_texture *tex)
    struct pipe_texture *pt = &tex->base;
    unsigned level;
 
-   const unsigned nblocks = pf_get_nblocksx(pt->format, pt->width0);
+   const unsigned nblocks = util_format_get_nblocksx(pt->format, pt->width0);
    unsigned face;
    unsigned width = pt->width0;
    unsigned height = pt->height0;
@@ -508,9 +509,9 @@ i945_miptree_layout_cube(struct i915_texture *tex)
     * or the final row of 4x4, 2x2 and 1x1 faces below this.
     */
    if (nblocks > 32)
-      tex->stride = align(nblocks * pf_get_blocksize(pt->format) * 2, 4);
+      tex->stride = align(nblocks * util_format_get_blocksize(pt->format) * 2, 4);
    else
-      tex->stride = 14 * 8 * pf_get_blocksize(pt->format);
+      tex->stride = 14 * 8 * util_format_get_blocksize(pt->format);
 
    tex->total_nblocksy = nblocks * 4;
 
@@ -840,8 +841,8 @@ i915_transfer_map(struct pipe_screen *screen,
       return NULL;
 
    return map + i915_transfer(transfer)->offset +
-      transfer->y / pf_get_blockheight(format) * transfer->stride +
-      transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format);
+      transfer->y / util_format_get_blockheight(format) * transfer->stride +
+      transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format);
 }
 
 static void
diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c
index b4aabd4d7c..b18f17c0cd 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup.c
@@ -41,6 +41,7 @@
 #include "draw/draw_vertex.h"
 #include "pipe/p_shader_tokens.h"
 #include "pipe/p_thread.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 #include "lp_bld_debug.h"
@@ -166,7 +167,7 @@ shade_quads(struct llvmpipe_context *llvmpipe,
       assert((y % 2) == 0);
       depth = llvmpipe->zsbuf_map +
               y*llvmpipe->zsbuf_transfer->stride +
-              2*x*pf_get_blocksize(llvmpipe->zsbuf_transfer->texture->format);
+              2*x*util_format_get_blocksize(llvmpipe->zsbuf_transfer->texture->format);
    }
    else
       depth = NULL;
diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c
index 9e41bc4074..2c135029ea 100644
--- a/src/gallium/drivers/llvmpipe/lp_texture.c
+++ b/src/gallium/drivers/llvmpipe/lp_texture.c
@@ -69,10 +69,10 @@ llvmpipe_texture_layout(struct llvmpipe_screen *screen,
 
       /* Allocate storage for whole quads. This is particularly important
        * for depth surfaces, which are currently stored in a swizzled format. */
-      nblocksx = pf_get_nblocksx(pt->format, align(width, 2));
-      nblocksy = pf_get_nblocksy(pt->format, align(height, 2));
+      nblocksx = util_format_get_nblocksx(pt->format, align(width, 2));
+      nblocksy = util_format_get_nblocksy(pt->format, align(height, 2));
 
-      lpt->stride[level] = align(nblocksx * pf_get_blocksize(pt->format), 16);
+      lpt->stride[level] = align(nblocksx * util_format_get_blocksize(pt->format), 16);
 
       lpt->level_offset[level] = buffer_size;
 
@@ -251,11 +251,11 @@ llvmpipe_get_tex_surface(struct pipe_screen *screen,
       */
       if (pt->target == PIPE_TEXTURE_CUBE) {
          unsigned tex_height = ps->height;
-         ps->offset += face * pf_get_nblocksy(pt->format, tex_height) * lpt->stride[level];
+         ps->offset += face * util_format_get_nblocksy(pt->format, tex_height) * lpt->stride[level];
       }
       else if (pt->target == PIPE_TEXTURE_3D) {
          unsigned tex_height = ps->height;
-         ps->offset += zslice * pf_get_nblocksy(pt->format, tex_height) * lpt->stride[level];
+         ps->offset += zslice * util_format_get_nblocksy(pt->format, tex_height) * lpt->stride[level];
       }
       else {
          assert(face == 0);
@@ -314,11 +314,11 @@ llvmpipe_get_tex_transfer(struct pipe_screen *screen,
       */
       if (texture->target == PIPE_TEXTURE_CUBE) {
          unsigned tex_height = u_minify(texture->height0, level);
-         lpt->offset += face *  pf_get_nblocksy(texture->format, tex_height) * pt->stride;
+         lpt->offset += face *  util_format_get_nblocksy(texture->format, tex_height) * pt->stride;
       }
       else if (texture->target == PIPE_TEXTURE_3D) {
          unsigned tex_height = u_minify(texture->height0, level);
-         lpt->offset += zslice * pf_get_nblocksy(texture->format, tex_height) * pt->stride;
+         lpt->offset += zslice * util_format_get_nblocksy(texture->format, tex_height) * pt->stride;
       }
       else {
          assert(face == 0);
@@ -379,8 +379,8 @@ llvmpipe_transfer_map( struct pipe_screen *_screen,
    }
    
    xfer_map = map + llvmpipe_transfer(transfer)->offset +
-      transfer->y / pf_get_blockheight(format) * transfer->stride +
-      transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format);
+      transfer->y / util_format_get_blockheight(format) * transfer->stride +
+      transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format);
    /*printf("map = %p  xfer map = %p\n", map, xfer_map);*/
    return xfer_map;
 }
diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.c b/src/gallium/drivers/nv04/nv04_surface_2d.c
index 3020806c5d..12df7fd199 100644
--- a/src/gallium/drivers/nv04/nv04_surface_2d.c
+++ b/src/gallium/drivers/nv04/nv04_surface_2d.c
@@ -1,5 +1,6 @@
 #include "pipe/p_context.h"
 #include "pipe/p_format.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 
@@ -158,10 +159,10 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx,
 	    sub_w = MIN2(sub_w, w - x);
 
 	    /* Must be 64-byte aligned */
-	    assert(!((dst->offset + nv04_swizzle_bits(dx+x, dy+y) * pf_get_blocksize(dst->texture->format)) & 63));
+	    assert(!((dst->offset + nv04_swizzle_bits(dx+x, dy+y) * util_format_get_blocksize(dst->texture->format)) & 63));
 
 	    BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_OFFSET, 1);
-	    OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(dx+x, dy+y) * pf_get_blocksize(dst->texture->format),
+	    OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(dx+x, dy+y) * util_format_get_blocksize(dst->texture->format),
                              NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
 
 	    BEGIN_RING(chan, sifm, NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION, 9);
@@ -180,7 +181,7 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx,
 	    OUT_RING  (chan, src_pitch |
 			     NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_CENTER |
 			     NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_POINT_SAMPLE);
-	    OUT_RELOCl(chan, src_bo, src->offset + (sy+y) * src_pitch + (sx+x) * pf_get_blocksize(src->texture->format),
+	    OUT_RELOCl(chan, src_bo, src->offset + (sy+y) * src_pitch + (sx+x) * util_format_get_blocksize(src->texture->format),
                              NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
 	    OUT_RING  (chan, 0);
 	  }
@@ -201,9 +202,9 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx,
 	unsigned src_pitch = ((struct nv04_surface *)src)->pitch;
 	unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch;
 	unsigned dst_offset = dst->offset + dy * dst_pitch +
-	                      dx * pf_get_blocksize(dst->texture->format);
+	                      dx * util_format_get_blocksize(dst->texture->format);
 	unsigned src_offset = src->offset + sy * src_pitch +
-	                      sx * pf_get_blocksize(src->texture->format);
+	                      sx * util_format_get_blocksize(src->texture->format);
 
 	MARK_RING (chan, 3 + ((h / 2047) + 1) * 9, 2 + ((h / 2047) + 1) * 2);
 	BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_BUFFER_IN, 2);
@@ -222,7 +223,7 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx,
 			   NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_WR);
 		OUT_RING  (chan, src_pitch);
 		OUT_RING  (chan, dst_pitch);
-		OUT_RING  (chan, w * pf_get_blocksize(src->texture->format));
+		OUT_RING  (chan, w * util_format_get_blocksize(src->texture->format));
 		OUT_RING  (chan, count);
 		OUT_RING  (chan, 0x0101);
 		OUT_RING  (chan, 0);
diff --git a/src/gallium/drivers/nv04/nv04_transfer.c b/src/gallium/drivers/nv04/nv04_transfer.c
index d66d6c6346..8446073ae8 100644
--- a/src/gallium/drivers/nv04/nv04_transfer.c
+++ b/src/gallium/drivers/nv04/nv04_transfer.c
@@ -1,6 +1,7 @@
 #include <pipe/p_state.h>
 #include <pipe/p_defines.h>
 #include <pipe/p_inlines.h>
+#include <util/u_format.h>
 #include <util/u_memory.h>
 #include <util/u_math.h>
 #include <nouveau/nouveau_winsys.h>
@@ -151,7 +152,7 @@ nv04_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
 	                            pipe_transfer_buffer_flags(ptx));
 
 	return map + ns->base.offset +
-	       ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format);
+	       ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format);
 }
 
 static void
diff --git a/src/gallium/drivers/nv10/nv10_miptree.c b/src/gallium/drivers/nv10/nv10_miptree.c
index 6a52b6af36..908482ad85 100644
--- a/src/gallium/drivers/nv10/nv10_miptree.c
+++ b/src/gallium/drivers/nv10/nv10_miptree.c
@@ -1,6 +1,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 
 #include "nv10_context.h"
@@ -23,9 +24,9 @@ nv10_miptree_layout(struct nv10_miptree *nv10mt)
 	
 	for (l = 0; l <= pt->last_level; l++) {
 		if (swizzled)
-			nv10mt->level[l].pitch = pf_get_stride(pt->format, width);
+			nv10mt->level[l].pitch = util_format_get_stride(pt->format, width);
 		else
-			nv10mt->level[l].pitch = pf_get_stride(pt->format, pt->width0);
+			nv10mt->level[l].pitch = util_format_get_stride(pt->format, pt->width0);
 		nv10mt->level[l].pitch = (nv10mt->level[l].pitch + 63) & ~63;
 
 		nv10mt->level[l].image_offset =
diff --git a/src/gallium/drivers/nv10/nv10_transfer.c b/src/gallium/drivers/nv10/nv10_transfer.c
index 06bb513417..c664973e90 100644
--- a/src/gallium/drivers/nv10/nv10_transfer.c
+++ b/src/gallium/drivers/nv10/nv10_transfer.c
@@ -1,6 +1,7 @@
 #include <pipe/p_state.h>
 #include <pipe/p_defines.h>
 #include <pipe/p_inlines.h>
+#include <util/u_format.h>
 #include <util/u_memory.h>
 #include <util/u_math.h>
 #include <nouveau/nouveau_winsys.h>
@@ -151,7 +152,7 @@ nv10_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
 	                            pipe_transfer_buffer_flags(ptx));
 
 	return map + ns->base.offset +
-	       ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format);
+	       ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format);
 }
 
 static void
diff --git a/src/gallium/drivers/nv20/nv20_miptree.c b/src/gallium/drivers/nv20/nv20_miptree.c
index e2e01bd849..d1291a92e0 100644
--- a/src/gallium/drivers/nv20/nv20_miptree.c
+++ b/src/gallium/drivers/nv20/nv20_miptree.c
@@ -1,6 +1,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 
 #include "nv20_context.h"
@@ -27,9 +28,9 @@ nv20_miptree_layout(struct nv20_miptree *nv20mt)
 	
 	for (l = 0; l <= pt->last_level; l++) {
 		if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR))
-			nv20mt->level[l].pitch = align(pf_get_stride(pt->format, pt->width0), 64);
+			nv20mt->level[l].pitch = align(util_format_get_stride(pt->format, pt->width0), 64);
 		else
-			nv20mt->level[l].pitch = pf_get_stride(pt->format, width);
+			nv20mt->level[l].pitch = util_format_get_stride(pt->format, width);
 
 		nv20mt->level[l].image_offset =
 			CALLOC(nr_faces, sizeof(unsigned));
diff --git a/src/gallium/drivers/nv20/nv20_transfer.c b/src/gallium/drivers/nv20/nv20_transfer.c
index 26a73c5143..69b79c809f 100644
--- a/src/gallium/drivers/nv20/nv20_transfer.c
+++ b/src/gallium/drivers/nv20/nv20_transfer.c
@@ -1,6 +1,7 @@
 #include <pipe/p_state.h>
 #include <pipe/p_defines.h>
 #include <pipe/p_inlines.h>
+#include <util/u_format.h>
 #include <util/u_memory.h>
 #include <util/u_math.h>
 #include <nouveau/nouveau_winsys.h>
@@ -151,7 +152,7 @@ nv20_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
 	                            pipe_transfer_buffer_flags(ptx));
 
 	return map + ns->base.offset +
-	       ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format);
+	       ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format);
 }
 
 static void
diff --git a/src/gallium/drivers/nv30/nv30_miptree.c b/src/gallium/drivers/nv30/nv30_miptree.c
index 920fe64c32..ce95d9700f 100644
--- a/src/gallium/drivers/nv30/nv30_miptree.c
+++ b/src/gallium/drivers/nv30/nv30_miptree.c
@@ -1,6 +1,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 
 #include "nv30_context.h"
@@ -29,9 +30,9 @@ nv30_miptree_layout(struct nv30_miptree *nv30mt)
 
 	for (l = 0; l <= pt->last_level; l++) {
 		if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR))
-			nv30mt->level[l].pitch = align(pf_get_stride(pt->format, pt->width0), 64);
+			nv30mt->level[l].pitch = align(util_format_get_stride(pt->format, pt->width0), 64);
 		else
-			nv30mt->level[l].pitch = pf_get_stride(pt->format, width);
+			nv30mt->level[l].pitch = util_format_get_stride(pt->format, width);
 
 		nv30mt->level[l].image_offset =
 			CALLOC(nr_faces, sizeof(unsigned));
diff --git a/src/gallium/drivers/nv30/nv30_transfer.c b/src/gallium/drivers/nv30/nv30_transfer.c
index e29bfbd3ef..2255a02cae 100644
--- a/src/gallium/drivers/nv30/nv30_transfer.c
+++ b/src/gallium/drivers/nv30/nv30_transfer.c
@@ -1,6 +1,7 @@
 #include <pipe/p_state.h>
 #include <pipe/p_defines.h>
 #include <pipe/p_inlines.h>
+#include <util/u_format.h>
 #include <util/u_memory.h>
 #include <util/u_math.h>
 #include <nouveau/nouveau_winsys.h>
@@ -151,7 +152,7 @@ nv30_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
 	                            pipe_transfer_buffer_flags(ptx));
 
 	return map + ns->base.offset +
-	       ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format);
+	       ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format);
 }
 
 static void
diff --git a/src/gallium/drivers/nv40/nv40_miptree.c b/src/gallium/drivers/nv40/nv40_miptree.c
index 89ddf373e9..b974e68a07 100644
--- a/src/gallium/drivers/nv40/nv40_miptree.c
+++ b/src/gallium/drivers/nv40/nv40_miptree.c
@@ -1,6 +1,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 
 #include "nv40_context.h"
@@ -31,9 +32,9 @@ nv40_miptree_layout(struct nv40_miptree *mt)
 
 	for (l = 0; l <= pt->last_level; l++) {
 		if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR))
-			mt->level[l].pitch = align(pf_get_stride(pt->format, pt->width0), 64);
+			mt->level[l].pitch = align(util_format_get_stride(pt->format, pt->width0), 64);
 		else
-			mt->level[l].pitch = pf_get_stride(pt->format, width);
+			mt->level[l].pitch = util_format_get_stride(pt->format, width);
 
 		mt->level[l].image_offset =
 			CALLOC(nr_faces, sizeof(unsigned));
diff --git a/src/gallium/drivers/nv40/nv40_transfer.c b/src/gallium/drivers/nv40/nv40_transfer.c
index ed5be1cf87..b084a38b48 100644
--- a/src/gallium/drivers/nv40/nv40_transfer.c
+++ b/src/gallium/drivers/nv40/nv40_transfer.c
@@ -1,6 +1,7 @@
 #include <pipe/p_state.h>
 #include <pipe/p_defines.h>
 #include <pipe/p_inlines.h>
+#include <util/u_format.h>
 #include <util/u_memory.h>
 #include <util/u_math.h>
 #include <nouveau/nouveau_winsys.h>
@@ -151,7 +152,7 @@ nv40_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
 	                            pipe_transfer_buffer_flags(ptx));
 
 	return map + ns->base.offset +
-	       ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format);
+	       ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format);
 }
 
 static void
diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c
index 9e083b662d..3f1edf0a13 100644
--- a/src/gallium/drivers/nv50/nv50_miptree.c
+++ b/src/gallium/drivers/nv50/nv50_miptree.c
@@ -23,6 +23,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 
 #include "nv50_context.h"
 
@@ -105,10 +106,10 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp)
 
 	for (l = 0; l <= pt->last_level; l++) {
 		struct nv50_miptree_level *lvl = &mt->level[l];
-		unsigned nblocksy = pf_get_nblocksy(pt->format, height);
+		unsigned nblocksy = util_format_get_nblocksy(pt->format, height);
 
 		lvl->image_offset = CALLOC(mt->image_nr, sizeof(int));
-		lvl->pitch = align(pf_get_stride(pt->format, width), 64);
+		lvl->pitch = align(util_format_get_stride(pt->format, width), 64);
 		lvl->tile_mode = get_tile_mode(nblocksy, depth);
 
 		width = u_minify(width, 1);
@@ -130,7 +131,7 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp)
 			unsigned tile_d = get_tile_depth(lvl->tile_mode);
 
 			size  = lvl->pitch;
-			size *= align(pf_get_nblocksy(pt->format, u_minify(pt->height0, l)), tile_h);
+			size *= align(util_format_get_nblocksy(pt->format, u_minify(pt->height0, l)), tile_h);
 			size *= align(u_minify(pt->depth0, l), tile_d);
 
 			lvl->image_offset[i] = mt->total_size;
@@ -222,7 +223,7 @@ nv50_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 	ps->offset = lvl->image_offset[img];
 
 	if (pt->target == PIPE_TEXTURE_3D) {
-		unsigned nb_h = pf_get_nblocksy(pt->format, ps->height);
+		unsigned nb_h = util_format_get_nblocksy(pt->format, ps->height);
 		ps->offset += get_zslice_offset(lvl->tile_mode, zslice,
 						lvl->pitch, nb_h);
 	}
diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c
index 6240a0c757..4d9afa6fed 100644
--- a/src/gallium/drivers/nv50/nv50_transfer.c
+++ b/src/gallium/drivers/nv50/nv50_transfer.c
@@ -1,6 +1,7 @@
 
 #include "pipe/p_context.h"
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 
 #include "nv50_context.h"
@@ -140,11 +141,11 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 		return NULL;
 
 	pipe_texture_reference(&tx->base.texture, pt);
-	tx->nblocksx = pf_get_nblocksx(pt->format, u_minify(pt->width0, level));
-	tx->nblocksy = pf_get_nblocksy(pt->format, u_minify(pt->height0, level));
+	tx->nblocksx = util_format_get_nblocksx(pt->format, u_minify(pt->width0, level));
+	tx->nblocksy = util_format_get_nblocksy(pt->format, u_minify(pt->height0, level));
 	tx->base.width = w;
 	tx->base.height = h;
-	tx->base.stride = tx->nblocksx * pf_get_blocksize(pt->format);
+	tx->base.stride = tx->nblocksx * util_format_get_blocksize(pt->format);
 	tx->base.usage = usage;
 
 	tx->level_pitch = lvl->pitch;
@@ -154,8 +155,8 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 	tx->level_offset = lvl->image_offset[image];
 	tx->level_tiling = lvl->tile_mode;
 	tx->level_z = zslice;
-	tx->level_x = pf_get_nblocksx(pt->format, x);
-	tx->level_y = pf_get_nblocksy(pt->format, y);
+	tx->level_x = util_format_get_nblocksx(pt->format, x);
+	tx->level_y = util_format_get_nblocksy(pt->format, y);
 	ret = nouveau_bo_new(dev, NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0,
 			     tx->nblocksy * tx->base.stride, &tx->bo);
 	if (ret) {
@@ -164,8 +165,8 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 	}
 
 	if (usage & PIPE_TRANSFER_READ) {
-		nx = pf_get_nblocksx(pt->format, tx->base.width);
-		ny = pf_get_nblocksy(pt->format, tx->base.height);
+		nx = util_format_get_nblocksx(pt->format, tx->base.width);
+		ny = util_format_get_nblocksy(pt->format, tx->base.height);
 
 		nv50_transfer_rect_m2mf(pscreen, mt->base.bo, tx->level_offset,
 					tx->level_pitch, tx->level_tiling,
@@ -176,7 +177,7 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 					tx->base.stride, tx->bo->tile_mode,
 					0, 0, 0,
 					tx->nblocksx, tx->nblocksy, 1,
-					pf_get_blocksize(pt->format), nx, ny,
+					util_format_get_blocksize(pt->format), nx, ny,
 					NOUVEAU_BO_VRAM | NOUVEAU_BO_GART,
 					NOUVEAU_BO_GART);
 	}
@@ -191,8 +192,8 @@ nv50_transfer_del(struct pipe_transfer *ptx)
 	struct nv50_miptree *mt = nv50_miptree(ptx->texture);
 	struct pipe_texture *pt = ptx->texture;
 
-	unsigned nx = pf_get_nblocksx(pt->format, tx->base.width);
-	unsigned ny = pf_get_nblocksy(pt->format, tx->base.height);
+	unsigned nx = util_format_get_nblocksx(pt->format, tx->base.width);
+	unsigned ny = util_format_get_nblocksy(pt->format, tx->base.height);
 
 	if (ptx->usage & PIPE_TRANSFER_WRITE) {
 		struct pipe_screen *pscreen = pt->screen;
@@ -206,7 +207,7 @@ nv50_transfer_del(struct pipe_transfer *ptx)
 					tx->level_x, tx->level_y, tx->level_z,
 					tx->nblocksx, tx->nblocksy,
 					tx->level_depth,
-					pf_get_blocksize(pt->format), nx, ny,
+					util_format_get_blocksize(pt->format), nx, ny,
 					NOUVEAU_BO_GART, NOUVEAU_BO_VRAM |
 					NOUVEAU_BO_GART);
 	}
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 52a8388ead..d7b6511d6d 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -657,7 +657,7 @@ static boolean r300_validate_aos(struct r300_context *r300)
     /* Check if formats and strides are aligned to the size of DWORD. */
     for (i = 0; i < r300->vertex_element_count; i++) {
         if (vbuf[velem[i].vertex_buffer_index].stride % 4 != 0 ||
-            pf_get_blocksize(velem[i].src_format) % 4 != 0) {
+            util_format_get_blocksize(velem[i].src_format) % 4 != 0) {
             return FALSE;
         }
     }
@@ -686,8 +686,8 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset)
     for (i = 0; i < aos_count - 1; i += 2) {
         vb1 = &vbuf[velem[i].vertex_buffer_index];
         vb2 = &vbuf[velem[i+1].vertex_buffer_index];
-        size1 = util_format_get_size(velem[i].src_format);
-        size2 = util_format_get_size(velem[i+1].src_format);
+        size1 = util_format_get_blocksize(velem[i].src_format);
+        size2 = util_format_get_blocksize(velem[i+1].src_format);
 
         OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride) |
                R300_VBPNTR_SIZE1(size2) | R300_VBPNTR_STRIDE1(vb2->stride));
@@ -697,7 +697,7 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset)
 
     if (aos_count & 1) {
         vb1 = &vbuf[velem[i].vertex_buffer_index];
-        size1 = util_format_get_size(velem[i].src_format);
+        size1 = util_format_get_blocksize(velem[i].src_format);
 
         OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride));
         OUT_CS(vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride);
diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c
index feb571a23d..a7ef3dbcc2 100644
--- a/src/gallium/drivers/r300/r300_screen.c
+++ b/src/gallium/drivers/r300/r300_screen.c
@@ -21,6 +21,7 @@
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 #include "util/u_memory.h"
 #include "util/u_simple_screen.h"
 
@@ -357,8 +358,8 @@ static void* r300_transfer_map(struct pipe_screen* screen,
     }
 
     return map + r300_transfer(transfer)->offset +
-        transfer->y / pf_get_blockheight(format) * transfer->stride +
-        transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format);
+        transfer->y / util_format_get_blockheight(format) * transfer->stride +
+        transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format);
 }
 
 static void r300_transfer_unmap(struct pipe_screen* screen,
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index 170483b9bb..9a96206a4d 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -106,7 +106,7 @@ unsigned r300_texture_get_stride(struct r300_texture* tex, unsigned level)
         return 0;
     }
 
-    return align(pf_get_stride(tex->tex.format, u_minify(tex->tex.width0, level)), 32);
+    return align(util_format_get_stride(tex->tex.format, u_minify(tex->tex.width0, level)), 32);
 }
 
 static void r300_setup_miptree(struct r300_texture* tex)
@@ -116,7 +116,7 @@ static void r300_setup_miptree(struct r300_texture* tex)
     int i;
 
     for (i = 0; i <= base->last_level; i++) {
-        unsigned nblocksy = pf_get_nblocksy(base->format, u_minify(base->height0, i));
+        unsigned nblocksy = util_format_get_nblocksy(base->format, u_minify(base->height0, i));
 
         stride = r300_texture_get_stride(tex, i);
         layer_size = stride * nblocksy;
@@ -129,7 +129,7 @@ static void r300_setup_miptree(struct r300_texture* tex)
         tex->offset[i] = align(tex->size, 32);
         tex->size = tex->offset[i] + size;
         tex->layer_size[i] = layer_size;
-        tex->pitch[i] = stride / pf_get_blocksize(base->format);
+        tex->pitch[i] = stride / util_format_get_blocksize(base->format);
 
         debug_printf("r300: Texture miptree: Level %d "
                 "(%dx%dx%d px, pitch %d bytes)\n",
@@ -245,7 +245,7 @@ static struct pipe_texture*
     tex->tex.screen = screen;
 
     tex->stride_override = *stride;
-    tex->pitch[0] = *stride / pf_get_blocksize(base->format);
+    tex->pitch[0] = *stride / util_format_get_blocksize(base->format);
 
     r300_setup_flags(tex);
     r300_setup_texture_state(tex, r300_screen(screen)->caps->is_r500);
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index 9e83410fcd..a9436a3394 100644
--- a/src/gallium/drivers/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -65,11 +65,11 @@ softpipe_texture_layout(struct pipe_screen *screen,
    pt->depth0 = depth;
 
    for (level = 0; level <= pt->last_level; level++) {
-      spt->stride[level] = pf_get_stride(pt->format, width);
+      spt->stride[level] = util_format_get_stride(pt->format, width);
 
       spt->level_offset[level] = buffer_size;
 
-      buffer_size += (pf_get_nblocksy(pt->format, height) *
+      buffer_size += (util_format_get_nblocksy(pt->format, height) *
                       ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) *
                       spt->stride[level]);
 
@@ -239,11 +239,11 @@ softpipe_get_tex_surface(struct pipe_screen *screen,
       ps->zslice = zslice;
 
       if (pt->target == PIPE_TEXTURE_CUBE) {
-         ps->offset += face * pf_get_nblocksy(pt->format, u_minify(pt->height0, level)) *
+         ps->offset += face * util_format_get_nblocksy(pt->format, u_minify(pt->height0, level)) *
                        spt->stride[level];
       }
       else if (pt->target == PIPE_TEXTURE_3D) {
-         ps->offset += zslice * pf_get_nblocksy(pt->format, u_minify(pt->height0, level)) *
+         ps->offset += zslice * util_format_get_nblocksy(pt->format, u_minify(pt->height0, level)) *
                        spt->stride[level];
       }
       else {
@@ -299,7 +299,7 @@ softpipe_get_tex_transfer(struct pipe_screen *screen,
    spt = CALLOC_STRUCT(softpipe_transfer);
    if (spt) {
       struct pipe_transfer *pt = &spt->base;
-      int nblocksy = pf_get_nblocksy(texture->format, u_minify(texture->height0, level));
+      int nblocksy = util_format_get_nblocksy(texture->format, u_minify(texture->height0, level));
       pipe_texture_reference(&pt->texture, texture);
       pt->x = x;
       pt->y = y;
@@ -376,8 +376,8 @@ softpipe_transfer_map( struct pipe_screen *screen,
    }
 
    xfer_map = map + softpipe_transfer(transfer)->offset +
-      transfer->y / pf_get_blockheight(format) * transfer->stride +
-      transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format);
+      transfer->y / util_format_get_blockheight(format) * transfer->stride +
+      transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format);
    /*printf("map = %p  xfer map = %p\n", map, xfer_map);*/
    return xfer_map;
 }
diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c
index 49e1ab0ca7..112a6fe0cf 100644
--- a/src/gallium/drivers/softpipe/sp_tile_cache.c
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.c
@@ -239,7 +239,7 @@ clear_tile(struct softpipe_cached_tile *tile,
 {
    uint i, j;
 
-   switch (util_format_get_size(format)) {
+   switch (util_format_get_blocksize(format)) {
    case 1:
       memset(tile->data.any, clear_value, TILE_SIZE * TILE_SIZE);
       break;
diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c
index ab93dab223..2224c2d394 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.c
+++ b/src/gallium/drivers/svga/svga_screen_texture.c
@@ -159,8 +159,8 @@ svga_transfer_dma_band(struct svga_transfer *st,
                 st->base.x + st->base.width,
                 y + h,
                 st->base.zslice + 1,
-                pf_get_blocksize(texture->base.format)*8/
-                (pf_get_blockwidth(texture->base.format)*pf_get_blockheight(texture->base.format)));
+                util_format_get_blocksize(texture->base.format)*8/
+                (util_format_get_blockwidth(texture->base.format)*util_format_get_blockheight(texture->base.format)));
    
    box.x = st->base.x;
    box.y = y;
@@ -210,7 +210,7 @@ svga_transfer_dma(struct svga_transfer *st,
    }
    else {
       unsigned y, h, srcy;
-      unsigned blockheight = pf_get_blockheight(st->base.texture->format);
+      unsigned blockheight = util_format_get_blockheight(st->base.texture->format);
       h = st->hw_nblocksy * blockheight;
       srcy = 0;
       for(y = 0; y < st->base.height; y += h) {
@@ -772,8 +772,8 @@ svga_get_tex_transfer(struct pipe_screen *screen,
    struct svga_screen *ss = svga_screen(screen);
    struct svga_winsys_screen *sws = ss->sws;
    struct svga_transfer *st;
-   unsigned nblocksx = pf_get_nblocksx(texture->format, w);
-   unsigned nblocksy = pf_get_nblocksy(texture->format, h);
+   unsigned nblocksx = util_format_get_nblocksx(texture->format, w);
+   unsigned nblocksy = util_format_get_nblocksy(texture->format, h);
 
    /* We can't map texture storage directly */
    if (usage & PIPE_TRANSFER_MAP_DIRECTLY)
@@ -787,7 +787,7 @@ svga_get_tex_transfer(struct pipe_screen *screen,
    st->base.y = y;
    st->base.width = w;
    st->base.height = h;
-   st->base.stride = nblocksx*pf_get_blocksize(texture->format);
+   st->base.stride = nblocksx*util_format_get_blocksize(texture->format);
    st->base.usage = usage;
    st->base.face = face;
    st->base.level = level;
@@ -1071,7 +1071,7 @@ svga_screen_buffer_from_texture(struct pipe_texture *texture,
        svga_translate_format(texture->format),
        stex->handle);
 
-   *stride = pf_get_stride(texture->format, texture->width0);
+   *stride = util_format_get_stride(texture->format, texture->width0);
 
    return *buffer != NULL;
 }
diff --git a/src/gallium/drivers/svga/svga_state_vs.c b/src/gallium/drivers/svga/svga_state_vs.c
index 9e339577c7..44b7ceb4fa 100644
--- a/src/gallium/drivers/svga/svga_state_vs.c
+++ b/src/gallium/drivers/svga/svga_state_vs.c
@@ -211,7 +211,7 @@ static int update_zero_stride( struct svga_context *svga,
          mapped_buffer = pipe_buffer_map_range(svga->pipe.screen, 
                                                vbuffer->buffer,
                                                vel->src_offset,
-                                               util_format_get_size(vel->src_format),
+                                               util_format_get_blocksize(vel->src_format),
                                                PIPE_BUFFER_USAGE_CPU_READ);
          translate->set_buffer(translate, vel->vertex_buffer_index,
                                mapped_buffer,
diff --git a/src/gallium/drivers/trace/tr_rbug.c b/src/gallium/drivers/trace/tr_rbug.c
index af1d7f3224..c31b1d8698 100644
--- a/src/gallium/drivers/trace/tr_rbug.c
+++ b/src/gallium/drivers/trace/tr_rbug.c
@@ -26,6 +26,7 @@
  **************************************************************************/
 
 
+#include "util/u_format.h"
 #include "util/u_string.h"
 #include "util/u_memory.h"
 #include "util/u_simple_list.h"
@@ -203,9 +204,9 @@ trace_rbug_texture_info(struct trace_rbug *tr_rbug, struct rbug_header *header,
                                &t->width0, 1,
                                &t->height0, 1,
                                &t->depth0, 1,
-                               pf_get_blockwidth(t->format),
-                               pf_get_blockheight(t->format),
-                               pf_get_blocksize(t->format),
+                               util_format_get_blockwidth(t->format),
+                               util_format_get_blockheight(t->format),
+                               util_format_get_blocksize(t->format),
                                t->last_level,
                                t->nr_samples,
                                t->tex_usage,
@@ -254,11 +255,11 @@ trace_rbug_texture_read(struct trace_rbug *tr_rbug, struct rbug_header *header,
 
    rbug_send_texture_read_reply(tr_rbug->con, serial,
                                 t->texture->format,
-                                pf_get_blockwidth(t->texture->format),
-                                pf_get_blockheight(t->texture->format),
-                                pf_get_blocksize(t->texture->format),
+                                util_format_get_blockwidth(t->texture->format),
+                                util_format_get_blockheight(t->texture->format),
+                                util_format_get_blocksize(t->texture->format),
                                 (uint8_t*)map,
-                                t->stride * pf_get_nblocksy(t->texture->format, t->height),
+                                t->stride * util_format_get_nblocksy(t->texture->format, t->height),
                                 t->stride,
                                 NULL);
 
diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c
index f69f7da000..ac20a47af1 100644
--- a/src/gallium/drivers/trace/tr_screen.c
+++ b/src/gallium/drivers/trace/tr_screen.c
@@ -25,6 +25,7 @@
  *
  **************************************************************************/
 
+#include "util/u_format.h"
 #include "util/u_memory.h"
 #include "util/u_simple_list.h"
 
@@ -425,7 +426,7 @@ trace_screen_transfer_unmap(struct pipe_screen *_screen,
    struct pipe_transfer *transfer = tr_trans->transfer;
 
    if(tr_trans->map) {
-      size_t size = pf_get_nblocksy(transfer->texture->format, transfer->width) * transfer->stride;
+      size_t size = util_format_get_nblocksy(transfer->texture->format, transfer->width) * transfer->stride;
 
       trace_dump_call_begin("pipe_screen", "transfer_write");
 
diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h
index d334114d62..6bfff1cc59 100644
--- a/src/gallium/include/pipe/p_format.h
+++ b/src/gallium/include/pipe/p_format.h
@@ -174,140 +174,6 @@ enum pipe_format {
  */
 extern const char *pf_name( enum pipe_format format );
 
-/**
- * Return bits for a particular component.
- * \param comp  component index, starting at 0
- */
-static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp )
-{
-   uint size;
-
-   if (pf_swizzle_x(format) == comp) {
-      size = pf_size_x(format);
-   }
-   else if (pf_swizzle_y(format) == comp) {
-      size = pf_size_y(format);
-   }
-   else if (pf_swizzle_z(format) == comp) {
-      size = pf_size_z(format);
-   }
-   else if (pf_swizzle_w(format) == comp) {
-      size = pf_size_w(format);
-   }
-   else {
-      size = 0;
-   }
-   if (pf_layout( format ) == PIPE_FORMAT_LAYOUT_RGBAZS)
-      return size << pf_exp2( format );
-   return size << (pf_mixed_scale8( format ) * 3);
-}
-
-
-/**
- * Return total bits needed for the pixel format per block.
- */
-static INLINE uint pf_get_blocksizebits( enum pipe_format format )
-{
-   switch (pf_layout(format)) {
-   case PIPE_FORMAT_LAYOUT_RGBAZS:
-   case PIPE_FORMAT_LAYOUT_MIXED:
-      return
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_0 ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_1 ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_B ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) +
-         pf_get_component_bits( format, PIPE_FORMAT_COMP_S );
-   case PIPE_FORMAT_LAYOUT_YCBCR:
-      assert( format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV );
-      return 32;
-   case PIPE_FORMAT_LAYOUT_DXT:
-        switch(format) {
-        case PIPE_FORMAT_DXT1_RGBA:
-        case PIPE_FORMAT_DXT1_RGB:
-        case PIPE_FORMAT_DXT1_SRGBA:
-        case PIPE_FORMAT_DXT1_SRGB:
-           return 64;
-        case PIPE_FORMAT_DXT3_RGBA:
-        case PIPE_FORMAT_DXT5_RGBA:
-        case PIPE_FORMAT_DXT3_SRGBA:
-        case PIPE_FORMAT_DXT5_SRGBA:
-           return 128;
-        default:
-           assert( 0 );
-           return 0;
-        }
-
-   default:
-      assert( 0 );
-      return 0;
-   }
-}
-
-/**
- * Return bytes per element for the given format.
- */
-static INLINE uint pf_get_blocksize( enum pipe_format format )
-{
-   assert(pf_get_blocksizebits(format) % 8 == 0);
-   return pf_get_blocksizebits(format) / 8;
-}
-
-static INLINE uint pf_get_blockwidth( enum pipe_format format )
-{
-   switch (pf_layout(format)) {
-   case PIPE_FORMAT_LAYOUT_YCBCR:
-      return 2;
-   case PIPE_FORMAT_LAYOUT_DXT:
-      return 4;
-   default:
-      return 1;
-   }
-}
-
-static INLINE uint pf_get_blockheight( enum pipe_format format )
-{
-   switch (pf_layout(format)) {
-   case PIPE_FORMAT_LAYOUT_DXT:
-      return 4;
-   default:
-      return 1;
-   }
-}
-
-static INLINE unsigned
-pf_get_nblocksx(enum pipe_format format, unsigned x)
-{
-   unsigned blockwidth = pf_get_blockwidth(format);
-   return (x + blockwidth - 1) / blockwidth;
-}
-
-static INLINE unsigned
-pf_get_nblocksy(enum pipe_format format, unsigned y)
-{
-   unsigned blockheight = pf_get_blockheight(format);
-   return (y + blockheight - 1) / blockheight;
-}
-
-static INLINE unsigned
-pf_get_nblocks(enum pipe_format format, unsigned width, unsigned height)
-{
-   return pf_get_nblocksx(format, width) * pf_get_nblocksy(format, height);
-}
-
-static INLINE size_t
-pf_get_stride(enum pipe_format format, unsigned width)
-{
-   return pf_get_nblocksx(format, width) * pf_get_blocksize(format);
-}
-
-static INLINE size_t
-pf_get_2d_size(enum pipe_format format, size_t stride, unsigned height)
-{
-   return pf_get_nblocksy(format, height) * stride;
-}
 
 enum pipe_video_chroma_format
 {
diff --git a/src/gallium/state_trackers/python/st_sample.c b/src/gallium/state_trackers/python/st_sample.c
index 97ca2afc54..9637741421 100644
--- a/src/gallium/state_trackers/python/st_sample.c
+++ b/src/gallium/state_trackers/python/st_sample.c
@@ -30,6 +30,7 @@
 #include "pipe/p_format.h"
 #include "pipe/p_state.h"
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 #include "util/u_tile.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
@@ -461,7 +462,7 @@ st_sample_dxt_pixel_block(enum pipe_format format,
          for(ch = 0; ch < 4; ++ch)
             rgba[y*rgba_stride + x*4 + ch] = (float)(data[i].rgba[y*4*4 + x*4 + ch])/255.0f;
    
-   memcpy(raw, data[i].raw, pf_get_blocksize(format));
+   memcpy(raw, data[i].raw, util_format_get_blocksize(format));
 }
 
 
@@ -473,7 +474,7 @@ st_sample_generic_pixel_block(enum pipe_format format,
 {
    unsigned i;
    unsigned x, y, ch;
-   int blocksize = pf_get_blocksize(format);
+   int blocksize = util_format_get_blocksize(format);
    
    for(i = 0; i < blocksize; ++i)
       raw[i] = (uint8_t)st_random();
@@ -548,11 +549,11 @@ st_sample_surface(struct st_surface *surface, float *rgba)
    if (raw) {
       enum pipe_format format = texture->format;
       uint x, y;
-      int nblocksx = pf_get_nblocksx(format, width);
-      int nblocksy = pf_get_nblocksy(format, height);
-      int blockwidth = pf_get_blockwidth(format);
-      int blockheight = pf_get_blockheight(format);
-      int blocksize = pf_get_blocksize(format);
+      int nblocksx = util_format_get_nblocksx(format, width);
+      int nblocksy = util_format_get_nblocksy(format, height);
+      int blockwidth = util_format_get_blockwidth(format);
+      int blockheight = util_format_get_blockheight(format);
+      int blocksize = util_format_get_blocksize(format);
 
 
       for (y = 0; y < nblocksy; ++y) {
diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c
index b8535a4217..a3294e877a 100644
--- a/src/gallium/state_trackers/python/st_softpipe_winsys.c
+++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c
@@ -169,8 +169,8 @@ st_softpipe_surface_buffer_create(struct pipe_winsys *winsys,
    const unsigned alignment = 64;
    unsigned nblocksy;
 
-   nblocksy = pf_get_nblocksy(format, height);
-   *stride = align(pf_get_stride(format, width), alignment);
+   nblocksy = util_format_get_nblocksy(format, height);
+   *stride = align(util_format_get_stride(format, width), alignment);
 
    return winsys->buffer_create(winsys, alignment,
                                 usage,
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
index 6fd402cee4..7106a06492 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
@@ -247,7 +247,7 @@ nouveau_drm_handle_from_pt(struct drm_api *api, struct pipe_screen *pscreen,
 		return false;
 
 	*handle = mt->bo->handle;
-	*stride = pf_get_stride(mt->base.format, mt->base.width0);
+	*stride = util_format_get_stride(mt->base.format, mt->base.width0);
 	return true;
 }
 
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
index 76acc99ad7..dfecb8a728 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
@@ -35,6 +35,7 @@
 #include "radeon_bo_gem.h"
 #include "softpipe/sp_texture.h"
 #include "r300_context.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include <X11/Xutil.h>
 
@@ -121,8 +122,8 @@ static struct pipe_buffer *radeon_surface_buffer_create(struct pipe_winsys *ws,
     const unsigned alignment = 64;
     unsigned nblocksy, size;
 
-    nblocksy = pf_get_nblocksy(format, height);
-    *stride = align(pf_get_stride(format, width), alignment);
+    nblocksy = util_format_get_nblocksy(format, height);
+    *stride = align(util_format_get_stride(format, width), alignment);
     size = *stride * nblocksy;
 
     return radeon_buffer_create(ws, 64, usage, size);
diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c
index a36a9e52a6..6ee3ede38c 100644
--- a/src/gallium/winsys/egl_xlib/sw_winsys.c
+++ b/src/gallium/winsys/egl_xlib/sw_winsys.c
@@ -163,8 +163,8 @@ surface_buffer_create(struct pipe_winsys *winsys,
    const unsigned alignment = 64;
    unsigned nblocksy;
 
-   nblocksy = pf_get_nblocksy(format, height);
-   *stride = align(pf_get_stride(format, width), alignment);
+   nblocksy = util_format_get_nblocksy(format, height);
+   *stride = align(util_format_get_stride(format, width), alignment);
 
    return winsys->buffer_create(winsys, alignment,
                                 usage,
diff --git a/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c b/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c
index 3fc44925a8..f15bcd37b5 100644
--- a/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c
+++ b/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c
@@ -141,8 +141,8 @@ static struct pipe_buffer* xsp_surface_buffer_create
    const unsigned int ALIGNMENT = 1;
    unsigned nblocksy;
 
-   nblocksy = pf_get_nblocksy(format, height);
-   *stride = align(pf_get_stride(format, width), ALIGNMENT);
+   nblocksy = util_format_get_nblocksy(format, height);
+   *stride = align(util_format_get_stride(format, width), ALIGNMENT);
 
    return pws->buffer_create(pws, ALIGNMENT, usage,
                              *stride * nblocksy);
diff --git a/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c b/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c
index 7122365027..7d076be3a3 100644
--- a/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c
+++ b/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c
@@ -137,8 +137,8 @@ gdi_llvmpipe_displaytarget_create(struct llvmpipe_winsys *winsys,
    gdt->width = width;
    gdt->height = height;
 
-   bpp = util_format_get_bits(format);
-   cpp = util_format_get_size(format);
+   bpp = util_format_get_blocksizebits(format);
+   cpp = util_format_get_blocksize(format);
    
    gdt->stride = align(width * cpp, alignment);
    gdt->size = gdt->stride * height;
diff --git a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c
index 40f612e893..2ad794c3f0 100644
--- a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c
+++ b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c
@@ -163,8 +163,8 @@ gdi_softpipe_surface_buffer_create(struct pipe_winsys *winsys,
    const unsigned alignment = 64;
    unsigned nblocksy;
 
-   nblocksy = pf_get_nblocksy(format, height);
-   *stride = align(pf_get_stride(format, width), alignment);
+   nblocksy = util_format_get_nblocksy(format, height);
+   *stride = align(util_format_get_stride(format, width), alignment);
 
    return winsys->buffer_create(winsys, alignment,
                                 usage,
@@ -271,10 +271,10 @@ gdi_softpipe_present(struct pipe_screen *screen,
 
     memset(&bmi, 0, sizeof(BITMAPINFO));
     bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER);
-    bmi.bmiHeader.biWidth = texture->stride[surface->level] / util_format_get_size(surface->format);
+    bmi.bmiHeader.biWidth = texture->stride[surface->level] / util_format_get_blocksize(surface->format);
     bmi.bmiHeader.biHeight= -(long)surface->height;
     bmi.bmiHeader.biPlanes = 1;
-    bmi.bmiHeader.biBitCount = util_format_get_bits(surface->format);
+    bmi.bmiHeader.biBitCount = util_format_get_blocksizebits(surface->format);
     bmi.bmiHeader.biCompression = BI_RGB;
     bmi.bmiHeader.biSizeImage = 0;
     bmi.bmiHeader.biXPelsPerMeter = 0;
diff --git a/src/gallium/winsys/xlib/xlib_cell.c b/src/gallium/winsys/xlib/xlib_cell.c
index cd838e7322..47ae0519a4 100644
--- a/src/gallium/winsys/xlib/xlib_cell.c
+++ b/src/gallium/winsys/xlib/xlib_cell.c
@@ -289,8 +289,8 @@ xm_surface_buffer_create(struct pipe_winsys *winsys,
    const unsigned alignment = 64;
    unsigned nblocksy;
 
-   nblocksy = pf_get_nblocksy(format, height);
-   *stride = align(pf_get_stride(format, width), alignment);
+   nblocksy = util_format_get_nblocksy(format, height);
+   *stride = align(util_format_get_stride(format, width), alignment);
 
    return winsys->buffer_create(winsys, alignment,
                                 usage,
diff --git a/src/gallium/winsys/xlib/xlib_llvmpipe.c b/src/gallium/winsys/xlib/xlib_llvmpipe.c
index d96311e452..2a434b5fd2 100644
--- a/src/gallium/winsys/xlib/xlib_llvmpipe.c
+++ b/src/gallium/winsys/xlib/xlib_llvmpipe.c
@@ -262,10 +262,10 @@ xm_llvmpipe_display(struct xmesa_buffer *xm_buffer,
    {
       if (xm_dt->tempImage == NULL)
       {
-         assert(pf_get_blockwidth(xm_dt->format) == 1);
-         assert(pf_get_blockheight(xm_dt->format) == 1);
+         assert(util_format_get_blockwidth(xm_dt->format) == 1);
+         assert(util_format_get_blockheight(xm_dt->format) == 1);
          alloc_shm_ximage(xm_dt, xm_buffer,
-                          xm_dt->stride / pf_get_blocksize(xm_dt->format),
+                          xm_dt->stride / util_format_get_blocksize(xm_dt->format),
                           xm_dt->height);
       }
 
@@ -331,8 +331,8 @@ xm_displaytarget_create(struct llvmpipe_winsys *winsys,
    xm_dt->width = width;
    xm_dt->height = height;
 
-   nblocksy = pf_get_nblocksy(format, height);
-   xm_dt->stride = align(pf_get_stride(format, width), alignment);
+   nblocksy = util_format_get_nblocksy(format, height);
+   xm_dt->stride = align(util_format_get_stride(format, width), alignment);
    size = xm_dt->stride * nblocksy;
 
 #ifdef USE_XSHM
diff --git a/src/gallium/winsys/xlib/xlib_softpipe.c b/src/gallium/winsys/xlib/xlib_softpipe.c
index 48cee0a41d..f7c0099584 100644
--- a/src/gallium/winsys/xlib/xlib_softpipe.c
+++ b/src/gallium/winsys/xlib/xlib_softpipe.c
@@ -255,10 +255,10 @@ xlib_softpipe_display_surface(struct xmesa_buffer *b,
    {
       if (xm_buf->tempImage == NULL) 
       {
-         assert(pf_get_blockwidth(surf->texture->format) == 1);
-         assert(pf_get_blockheight(surf->texture->format) == 1);
+         assert(util_format_get_blockwidth(surf->texture->format) == 1);
+         assert(util_format_get_blockheight(surf->texture->format) == 1);
          alloc_shm_ximage(xm_buf, b, spt->stride[surf->level] /
-                          pf_get_blocksize(surf->texture->format), surf->height);
+                          util_format_get_blocksize(surf->texture->format), surf->height);
       }
 
       ximage = xm_buf->tempImage;
@@ -363,8 +363,8 @@ xm_surface_buffer_create(struct pipe_winsys *winsys,
    const unsigned alignment = 64;
    unsigned nblocksy, size;
 
-   nblocksy = pf_get_nblocksy(format, height);
-   *stride = align(pf_get_stride(format, width), alignment);
+   nblocksy = util_format_get_nblocksy(format, height);
+   *stride = align(util_format_get_stride(format, width), alignment);
    size = *stride * nblocksy;
 
 #ifdef USE_XSHM
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
index c61c77479e..7c664267d4 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.c
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -857,8 +857,8 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
 					   usage, dstx, dsty,
 					   width, height);
 
-   assert(pf_get_blockwidth(ptDraw->texture->format) == 1);
-   assert(pf_get_blockheight(ptDraw->texture->format) == 1);
+   assert(util_format_get_blockwidth(ptDraw->texture->format) == 1);
+   assert(util_format_get_blockheight(ptDraw->texture->format) == 1);
 
    /* map the stencil buffer */
    drawMap = screen->transfer_map(screen, ptDraw);
diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c
index 78aed07bf9..45ce34a85f 100644
--- a/src/mesa/state_tracker/st_cb_fbo.c
+++ b/src/mesa/state_tracker/st_cb_fbo.c
@@ -105,8 +105,8 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
 
       assert(strb->format != PIPE_FORMAT_NONE);
       
-      strb->stride = pf_get_stride(strb->format, width);
-      size = pf_get_2d_size(strb->format, strb->stride, height);
+      strb->stride = util_format_get_stride(strb->format, width);
+      size = util_format_get_2d_size(strb->format, strb->stride, height);
       
       strb->data = _mesa_malloc(size);
       
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index 000e6eb2a5..6e1ecb1c50 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -833,7 +833,7 @@ decompress_with_blit(GLcontext * ctx, GLenum target, GLint level,
    /* copy/pack data into user buffer */
    if (st_equal_formats(stImage->pt->format, format, type)) {
       /* memcpy */
-      const uint bytesPerRow = width * util_format_get_size(stImage->pt->format);
+      const uint bytesPerRow = width * util_format_get_blocksize(stImage->pt->format);
       ubyte *map = screen->transfer_map(screen, tex_xfer);
       GLuint row;
       for (row = 0; row < height; row++) {
@@ -915,7 +915,7 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
                                             PIPE_TRANSFER_READ, 0, 0,
                                             stImage->base.Width,
                                             stImage->base.Height);
-      texImage->RowStride = stImage->transfer->stride / pf_get_blocksize(stImage->pt->format);
+      texImage->RowStride = stImage->transfer->stride / util_format_get_blocksize(stImage->pt->format);
    }
    else {
       /* Otherwise, the image should actually be stored in
@@ -1178,7 +1178,7 @@ st_CompressedTexSubImage2D(GLcontext *ctx, GLenum target, GLint level,
                                             xoffset, yoffset,
                                             width, height);
       
-      srcBlockStride = pf_get_stride(pformat, width);
+      srcBlockStride = util_format_get_stride(pformat, width);
       dstBlockStride = stImage->transfer->stride;
    } else {
       assert(stImage->pt);
@@ -1192,16 +1192,16 @@ st_CompressedTexSubImage2D(GLcontext *ctx, GLenum target, GLint level,
       return;
    }
 
-   assert(xoffset % pf_get_blockwidth(pformat) == 0);
-   assert(yoffset % pf_get_blockheight(pformat) == 0);
-   assert(width % pf_get_blockwidth(pformat) == 0);
-   assert(height % pf_get_blockheight(pformat) == 0);
+   assert(xoffset % util_format_get_blockwidth(pformat) == 0);
+   assert(yoffset % util_format_get_blockheight(pformat) == 0);
+   assert(width % util_format_get_blockwidth(pformat) == 0);
+   assert(height % util_format_get_blockheight(pformat) == 0);
 
-   for (y = 0; y < height; y += pf_get_blockheight(pformat)) {
+   for (y = 0; y < height; y += util_format_get_blockheight(pformat)) {
       /* don't need to adjust for xoffset and yoffset as st_texture_image_map does that */
-      const char *src = (const char*)data + srcBlockStride * pf_get_nblocksy(pformat, y);
-      char *dst = (char*)texImage->Data + dstBlockStride * pf_get_nblocksy(pformat, y);
-      memcpy(dst, src, pf_get_stride(pformat, width));
+      const char *src = (const char*)data + srcBlockStride * util_format_get_nblocksy(pformat, y);
+      char *dst = (char*)texImage->Data + dstBlockStride * util_format_get_nblocksy(pformat, y);
+      memcpy(dst, src, util_format_get_stride(pformat, width));
    }
 
    if (stImage->pt) {
@@ -1691,10 +1691,10 @@ copy_image_data_to_texture(struct st_context *st,
                             dstLevel,
                             stImage->base.Data,
                             stImage->base.RowStride * 
-                            pf_get_blocksize(stObj->pt->format),
+                            util_format_get_blocksize(stObj->pt->format),
                             stImage->base.RowStride *
                             stImage->base.Height *
-                            pf_get_blocksize(stObj->pt->format));
+                            util_format_get_blocksize(stObj->pt->format));
       _mesa_align_free(stImage->base.Data);
       stImage->base.Data = NULL;
    }
diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c
index 7700551830..5a433dd7b9 100644
--- a/src/mesa/state_tracker/st_gen_mipmap.c
+++ b/src/mesa/state_tracker/st_gen_mipmap.c
@@ -37,6 +37,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
+#include "util/u_format.h"
 #include "util/u_gen_mipmap.h"
 #include "util/u_math.h"
 
@@ -146,8 +147,8 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target,
       srcData = (ubyte *) screen->transfer_map(screen, srcTrans);
       dstData = (ubyte *) screen->transfer_map(screen, dstTrans);
 
-      srcStride = srcTrans->stride / pf_get_blocksize(srcTrans->texture->format);
-      dstStride = dstTrans->stride / pf_get_blocksize(dstTrans->texture->format);
+      srcStride = srcTrans->stride / util_format_get_blocksize(srcTrans->texture->format);
+      dstStride = dstTrans->stride / util_format_get_blocksize(dstTrans->texture->format);
 
       _mesa_generate_mipmap_level(target, datatype, comps,
                                   0 /*border*/,
-- 
cgit v1.2.3


From 6ba83cd63f84a4d74dd679d62662d59533fd1bdb Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Wed, 16 Dec 2009 06:07:39 +0100
Subject: r300g: add texture compare modes

---
 src/gallium/drivers/r300/r300_context.h    |   2 +
 src/gallium/drivers/r300/r300_emit.c       |  17 +++--
 src/gallium/drivers/r300/r300_fs.c         | 102 +++++++++++++++++++++++------
 src/gallium/drivers/r300/r300_fs.h         |  32 ++++++---
 src/gallium/drivers/r300/r300_state.c      |  23 ++++++-
 src/gallium/drivers/r300/r300_tgsi_to_rc.c |   9 ++-
 6 files changed, 148 insertions(+), 37 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 6bd2766730..232530b7dc 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -91,6 +91,8 @@ struct r300_rs_block {
 };
 
 struct r300_sampler_state {
+    struct pipe_sampler_state state;
+
     uint32_t filter0;      /* R300_TX_FILTER0: 0x4400 */
     uint32_t filter1;      /* R300_TX_FILTER1: 0x4440 */
     uint32_t border_color; /* R300_TX_BORDER_COLOR: 0x45c0 */
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index d7b6511d6d..199ce3a945 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -158,6 +158,13 @@ static const float * get_shader_constant(
                     vec[1] = 1.0 / tex->height0;
                     break;
 
+                /* Texture compare-fail value. */
+                /* XXX Since Gallium doesn't support GL_ARB_shadow_ambient,
+                 * this is always (0,0,0,0). */
+                case RC_STATE_SHADOW_AMBIENT:
+                    vec[3] = 0;
+                    break;
+
                 default:
                     debug_printf("r300: Implementation error: "
                         "Unknown RC_CONSTANT type %d\n", constant->u.State[0]);
@@ -1030,18 +1037,20 @@ validate:
 
     if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) {
         if (r300screen->caps->is_r500) {
-            r500_emit_fragment_program_code(r300, &r300->fs->code);
+            r500_emit_fragment_program_code(r300, &r300->fs->shader->code);
         } else {
-            r300_emit_fragment_program_code(r300, &r300->fs->code);
+            r300_emit_fragment_program_code(r300, &r300->fs->shader->code);
         }
         r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER;
     }
 
     if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER_CONSTANTS) {
         if (r300screen->caps->is_r500) {
-            r500_emit_fs_constant_buffer(r300, &r300->fs->code.constants);
+            r500_emit_fs_constant_buffer(r300,
+                                         &r300->fs->shader->code.constants);
         } else {
-            r300_emit_fs_constant_buffer(r300, &r300->fs->code.constants);
+            r300_emit_fs_constant_buffer(r300,
+                                         &r300->fs->shader->code.constants);
         }
         r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER_CONSTANTS;
     }
diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c
index 79b01bb4dc..4e1b61ca40 100644
--- a/src/gallium/drivers/r300/r300_fs.c
+++ b/src/gallium/drivers/r300/r300_fs.c
@@ -22,6 +22,9 @@
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
 #include "tgsi/tgsi_dump.h"
 
 #include "r300_context.h"
@@ -33,8 +36,8 @@
 #include "radeon_compiler.h"
 
 /* Convert info about FS input semantics to r300_shader_semantics. */
-static void r300_shader_read_fs_inputs(struct tgsi_shader_info* info,
-                                       struct r300_shader_semantics* fs_inputs)
+void r300_shader_read_fs_inputs(struct tgsi_shader_info* info,
+                                struct r300_shader_semantics* fs_inputs)
 {
     int i;
     unsigned index;
@@ -66,7 +69,6 @@ static void r300_shader_read_fs_inputs(struct tgsi_shader_info* info,
     }
 }
 
-
 static void find_output_registers(struct r300_fragment_program_compiler * compiler,
                                   struct r300_fragment_shader * fs)
 {
@@ -95,7 +97,7 @@ static void allocate_hardware_inputs(
     void * mydata)
 {
     struct r300_shader_semantics* inputs =
-        &((struct r300_fragment_shader*)c->UserData)->inputs;
+        (struct r300_shader_semantics*)c->UserData;
     int i, reg = 0;
 
     /* Allocate input registers. */
@@ -114,31 +116,45 @@ static void allocate_hardware_inputs(
     }
 }
 
-void r300_translate_fragment_shader(struct r300_context* r300,
-                                    struct r300_fragment_shader* fs)
+static void get_compare_state(
+    struct r300_context* r300,
+    struct r300_fragment_program_external_state* state,
+    unsigned shadow_samplers)
+{
+    memset(state, 0, sizeof(*state));
+
+    for (int i = 0; i < r300->sampler_count; i++) {
+        struct r300_sampler_state* s = r300->sampler_states[i];
+
+        if (s && s->state.compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+            /* XXX Gallium doesn't provide us with any information regarding
+             * this mode, so we are screwed. I'm setting 0 = LUMINANCE. */
+            state->unit[i].depth_texture_mode = 0;
+
+            /* Fortunately, no need to translate this. */
+            state->unit[i].texture_compare_func = s->state.compare_func;
+        }
+    }
+}
+
+static void r300_translate_fragment_shader(
+    struct r300_context* r300,
+    struct r300_fragment_shader_code* shader)
 {
+    struct r300_fragment_shader* fs = r300->fs;
     struct r300_fragment_program_compiler compiler;
     struct tgsi_to_rc ttr;
 
-    /* Initialize. */
-    r300_shader_read_fs_inputs(&fs->info, &fs->inputs);
-
     /* Setup the compiler. */
     memset(&compiler, 0, sizeof(compiler));
     rc_init(&compiler.Base);
     compiler.Base.Debug = DBG_ON(r300, DBG_FP);
 
-    compiler.code = &fs->code;
+    compiler.code = &shader->code;
+    compiler.state = shader->compare_state;
     compiler.is_r500 = r300_screen(r300->context.screen)->caps->is_r500;
     compiler.AllocateHwInputs = &allocate_hardware_inputs;
-    compiler.UserData = fs;
-
-    /* XXX: Program compilation depends on texture compare modes,
-     * which are sampler state. Therefore, programs need to be recompiled
-     * depending on this state as in the classic Mesa driver.
-     *
-     * This is not yet handled correctly.
-     */
+    compiler.UserData = &fs->inputs;
 
     find_output_registers(&compiler, fs);
 
@@ -153,6 +169,8 @@ void r300_translate_fragment_shader(struct r300_context* r300,
 
     r300_tgsi_to_rc(&ttr, fs->state.tokens);
 
+    fs->shadow_samplers = compiler.Base.Program.ShadowSamplers;
+
     /* Invoke the compiler */
     r3xx_compile_fragment_program(&compiler);
     if (compiler.Base.Error) {
@@ -164,5 +182,51 @@ void r300_translate_fragment_shader(struct r300_context* r300,
 
     /* And, finally... */
     rc_destroy(&compiler.Base);
-    fs->translated = TRUE;
+}
+
+boolean r300_pick_fragment_shader(struct r300_context* r300)
+{
+    struct r300_fragment_shader* fs = r300->fs;
+    struct r300_fragment_program_external_state state;
+    struct r300_fragment_shader_code* ptr;
+
+    if (!fs->first) {
+        /* Build the fragment shader for the first time. */
+        fs->first = fs->shader = CALLOC_STRUCT(r300_fragment_shader_code);
+
+        /* BTW shadow samplers will be known after the first translation,
+         * therefore we set ~0, which means it should look at all sampler
+         * states. This choice doesn't have any impact on the correctness. */
+        get_compare_state(r300, &fs->shader->compare_state, ~0);
+        r300_translate_fragment_shader(r300, fs->shader);
+        return TRUE;
+
+    } else if (fs->shadow_samplers) {
+        get_compare_state(r300, &state, fs->shadow_samplers);
+
+        /* Check if the currently-bound shader has been compiled
+         * with the texture-compare state we need. */
+        if (memcmp(&fs->shader->compare_state, &state, sizeof(state)) != 0) {
+            /* Search for the right shader. */
+            ptr = fs->first;
+            while (ptr) {
+                if (memcmp(&ptr->compare_state, &state, sizeof(state)) == 0) {
+                    fs->shader = ptr;
+                    return TRUE;
+                }
+                ptr = ptr->next;
+            }
+
+            /* Not found, gotta compile a new one. */
+            ptr = CALLOC_STRUCT(r300_fragment_shader_code);
+            ptr->next = fs->first;
+            fs->first = fs->shader = ptr;
+
+            ptr->compare_state = state;
+            r300_translate_fragment_shader(r300, ptr);
+            return TRUE;
+        }
+    }
+
+    return FALSE;
 }
diff --git a/src/gallium/drivers/r300/r300_fs.h b/src/gallium/drivers/r300/r300_fs.h
index 630e2d0c8a..40ce874353 100644
--- a/src/gallium/drivers/r300/r300_fs.h
+++ b/src/gallium/drivers/r300/r300_fs.h
@@ -30,6 +30,13 @@
 #include "radeon_code.h"
 #include "r300_shader_semantics.h"
 
+struct r300_fragment_shader_code {
+    struct r300_fragment_program_external_state compare_state;
+    struct rX00_fragment_program_code code;
+
+    struct r300_fragment_shader_code* next;
+};
+
 struct r300_fragment_shader {
     /* Parent class */
     struct pipe_shader_state state;
@@ -37,21 +44,28 @@ struct r300_fragment_shader {
     struct tgsi_shader_info info;
     struct r300_shader_semantics inputs;
 
-    /* Has this shader been translated yet? */
-    boolean translated;
+    /* Bits 0-15: TRUE if it's a shadow sampler, FALSE otherwise. */
+    unsigned shadow_samplers;
 
-    /* Compiled code */
-    struct rX00_fragment_program_code code;
+    /* Currently-bound fragment shader. */
+    struct r300_fragment_shader_code* shader;
+
+    /* List of the same shaders compiled with different texture-compare
+     * states. */
+    struct r300_fragment_shader_code* first;
 };
 
+void r300_shader_read_fs_inputs(struct tgsi_shader_info* info,
+                                struct r300_shader_semantics* fs_inputs);
 
-void r300_translate_fragment_shader(struct r300_context* r300,
-                                    struct r300_fragment_shader* fs);
+/* Return TRUE if the shader was switched and should be re-emitted. */
+boolean r300_pick_fragment_shader(struct r300_context* r300);
 
-static inline boolean r300_fragment_shader_writes_depth(struct r300_fragment_shader *fs)
+static INLINE boolean r300_fragment_shader_writes_depth(struct r300_fragment_shader *fs)
 {
     if (!fs)
-	return FALSE;
-    return (fs->code.writes_depth) ? TRUE : FALSE;
+        return FALSE;
+    return (fs->shader->code.writes_depth) ? TRUE : FALSE;
 }
+
 #endif /* R300_FS_H */
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 5f332b2e0c..8bcd6c5060 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -355,6 +355,7 @@ static void* r300_create_fs_state(struct pipe_context* pipe,
     fs->state.tokens = tgsi_dup_tokens(shader->tokens);
 
     tgsi_scan_shader(shader->tokens, &fs->info);
+    r300_shader_read_fs_inputs(&fs->info, &fs->inputs);
 
     return (void*)fs;
 }
@@ -368,11 +369,10 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader)
     if (fs == NULL) {
         r300->fs = NULL;
         return;
-    } else if (!fs->translated) {
-        r300_translate_fragment_shader(r300, fs);
     }
 
     r300->fs = fs;
+    r300_pick_fragment_shader(r300);
 
     r300->dirty_state |= R300_NEW_FRAGMENT_SHADER | R300_NEW_FRAGMENT_SHADER_CONSTANTS;
 }
@@ -381,7 +381,14 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader)
 static void r300_delete_fs_state(struct pipe_context* pipe, void* shader)
 {
     struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader;
-    rc_constants_destroy(&fs->code.constants);
+    struct r300_fragment_shader_code *tmp, *ptr = fs->first;
+
+    while (ptr) {
+        tmp = ptr;
+        ptr = ptr->next;
+        rc_constants_destroy(&tmp->code.constants);
+        FREE(tmp);
+    }
     FREE((void*)fs->state.tokens);
     FREE(shader);
 }
@@ -547,6 +554,8 @@ static void*
     int lod_bias;
     union util_color uc;
 
+    sampler->state = *state;
+
     sampler->filter0 |=
         (r300_translate_wrap(state->wrap_s) << R300_TX_WRAP_S_SHIFT) |
         (r300_translate_wrap(state->wrap_t) << R300_TX_WRAP_T_SHIFT) |
@@ -597,6 +606,14 @@ static void r300_bind_sampler_states(struct pipe_context* pipe,
     }
 
     r300->sampler_count = count;
+
+    /* Pick a fragment shader based on the texture compare state. */
+    if (r300->fs && (r300->dirty_state & R300_ANY_NEW_SAMPLERS)) {
+        if (r300_pick_fragment_shader(r300)) {
+            r300->dirty_state |= R300_NEW_FRAGMENT_SHADER |
+                                 R300_NEW_FRAGMENT_SHADER_CONSTANTS;
+        }
+    }
 }
 
 static void r300_lacks_vertex_textures(struct pipe_context* pipe,
diff --git a/src/gallium/drivers/r300/r300_tgsi_to_rc.c b/src/gallium/drivers/r300/r300_tgsi_to_rc.c
index 9fb2de2403..096cdb20bb 100644
--- a/src/gallium/drivers/r300/r300_tgsi_to_rc.c
+++ b/src/gallium/drivers/r300/r300_tgsi_to_rc.c
@@ -212,7 +212,8 @@ static void transform_srcreg(
     dst->Negate = src->Register.Negate ? RC_MASK_XYZW : 0;
 }
 
-static void transform_texture(struct rc_instruction * dst, struct tgsi_instruction_texture src)
+static void transform_texture(struct rc_instruction * dst, struct tgsi_instruction_texture src,
+                              uint32_t *shadowSamplers)
 {
     switch(src.Texture) {
         case TGSI_TEXTURE_1D:
@@ -233,14 +234,17 @@ static void transform_texture(struct rc_instruction * dst, struct tgsi_instructi
         case TGSI_TEXTURE_SHADOW1D:
             dst->U.I.TexSrcTarget = RC_TEXTURE_1D;
             dst->U.I.TexShadow = 1;
+            *shadowSamplers |= 1 << dst->U.I.TexSrcUnit;
             break;
         case TGSI_TEXTURE_SHADOW2D:
             dst->U.I.TexSrcTarget = RC_TEXTURE_2D;
             dst->U.I.TexShadow = 1;
+            *shadowSamplers |= 1 << dst->U.I.TexSrcUnit;
             break;
         case TGSI_TEXTURE_SHADOWRECT:
             dst->U.I.TexSrcTarget = RC_TEXTURE_RECT;
             dst->U.I.TexShadow = 1;
+            *shadowSamplers |= 1 << dst->U.I.TexSrcUnit;
             break;
     }
 }
@@ -269,7 +273,8 @@ static void transform_instruction(struct tgsi_to_rc * ttr, struct tgsi_full_inst
 
     /* Texturing. */
     if (src->Instruction.Texture)
-       transform_texture(dst, src->Texture);
+        transform_texture(dst, src->Texture,
+                          &ttr->compiler->Program.ShadowSamplers);
 }
 
 static void handle_immediate(struct tgsi_to_rc * ttr, struct tgsi_full_immediate * imm)
-- 
cgit v1.2.3


From be6fb5c08350138ac85c1bf727553aaa9f4583e6 Mon Sep 17 00:00:00 2001
From: Vinson Lee <vlee@vmware.com>
Date: Sat, 2 Jan 2010 20:17:43 -0800
Subject: r300g: Use C-style comments.

---
 src/gallium/drivers/r300/r300_emit.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index eeb97a2d37..b44c7bdbb3 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -871,7 +871,7 @@ validate:
             goto validate;
         }
     } else {
-        // debug_printf("No VBO while emitting dirty state!\n");
+        /* debug_printf("No VBO while emitting dirty state!\n"); */
     }
     if (!r300->winsys->validate(r300->winsys)) {
         r300->context.flush(&r300->context, 0, NULL);
@@ -999,7 +999,7 @@ validate:
     */
 
     /* Finally, emit the VBO. */
-    //r300_emit_vertex_buffer(r300);
+    /* r300_emit_vertex_buffer(r300); */
 
     r300->dirty_hw++;
 }
-- 
cgit v1.2.3


From e78648db0a91269ed83238bfa3ced5d4d7dae2e0 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Thu, 24 Dec 2009 17:40:53 +0100
Subject: r300g: fix outputting depth in the FS

---
 src/gallium/drivers/r300/r300_emit.c            | 17 +++++++++++++++++
 src/gallium/drivers/r300/r300_state_invariant.c |  4 +---
 2 files changed, 18 insertions(+), 3 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 1dc9216a7b..596ec21bc0 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -283,6 +283,22 @@ void r300_emit_fs_constant_buffer(struct r300_context* r300,
     END_CS;
 }
 
+static void r300_emit_fragment_depth_config(struct r300_context* r300,
+                                            struct r300_fragment_shader* fs)
+{
+    CS_LOCALS(r300);
+
+    BEGIN_CS(4);
+    if (r300_fragment_shader_writes_depth(fs)) {
+        OUT_CS_REG(R300_FG_DEPTH_SRC, R300_FG_DEPTH_SRC_SHADER);
+        OUT_CS_REG(R300_US_W_FMT, R300_W_FMT_W24 | R300_W_SRC_US);
+    } else {
+        OUT_CS_REG(R300_FG_DEPTH_SRC, R300_FG_DEPTH_SRC_SCAN);
+        OUT_CS_REG(R300_US_W_FMT, R300_W_FMT_W0 | R300_W_SRC_US);
+    }
+    END_CS;
+}
+
 void r500_emit_fragment_program_code(struct r300_context* r300,
                                      struct rX00_fragment_program_code* generic_code)
 {
@@ -1036,6 +1052,7 @@ validate:
     }
 
     if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) {
+        r300_emit_fragment_depth_config(r300, r300->fs);
         if (r300screen->caps->is_r500) {
             r500_emit_fragment_program_code(r300, &r300->fs->shader->code);
         } else {
diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c
index 64dc14ea76..f25f3ca217 100644
--- a/src/gallium/drivers/r300/r300_state_invariant.c
+++ b/src/gallium/drivers/r300/r300_state_invariant.c
@@ -43,7 +43,7 @@ void r300_emit_invariant_state(struct r300_context* r300)
     struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps;
     CS_LOCALS(r300);
 
-    BEGIN_CS(20 + (caps->has_tcl ? 2: 0));
+    BEGIN_CS(16 + (caps->has_tcl ? 2: 0));
 
     /*** Graphics Backend (GB) ***/
     /* Various GB enables */
@@ -66,8 +66,6 @@ void r300_emit_invariant_state(struct r300_context* r300)
     OUT_CS_REG(R300_FG_FOG_COLOR_R, 0x0);
     OUT_CS_REG(R300_FG_FOG_COLOR_G, 0x0);
     OUT_CS_REG(R300_FG_FOG_COLOR_B, 0x0);
-    OUT_CS_REG(R300_FG_DEPTH_SRC, 0x0);
-    OUT_CS_REG(R300_US_W_FMT, 0x0);
 
     /*** VAP ***/
     /* Sign/normalize control */
-- 
cgit v1.2.3


From bf60eb3fec844a7c3793aba0c70da56b74a17344 Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Thu, 24 Dec 2009 03:10:33 +0100
Subject: r300g: add WPOS

---
 src/gallium/drivers/r300/r300_emit.c             | 22 ++++++++++
 src/gallium/drivers/r300/r300_fs.c               | 21 +++++++++
 src/gallium/drivers/r300/r300_shader_semantics.h |  2 +
 src/gallium/drivers/r300/r300_state.c            |  6 +++
 src/gallium/drivers/r300/r300_state_derived.c    | 20 +++++++++
 src/gallium/drivers/r300/r300_vs.c               | 56 ++++++++++++++++++++++--
 6 files changed, 124 insertions(+), 3 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 596ec21bc0..30a1000c29 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -165,6 +165,28 @@ static const float * get_shader_constant(
                     vec[3] = 0;
                     break;
 
+                case RC_STATE_R300_VIEWPORT_SCALE:
+                    if (r300->rs_state->enable_vte) {
+                        vec[0] = r300->viewport_state->xscale;
+                        vec[1] = r300->viewport_state->yscale;
+                        vec[2] = r300->viewport_state->zscale;
+                    } else {
+                        vec[0] = 1;
+                        vec[1] = 1;
+                        vec[2] = 1;
+                    }
+                    break;
+
+                case RC_STATE_R300_VIEWPORT_OFFSET:
+                    if (r300->rs_state->enable_vte) {
+                        vec[0] = r300->viewport_state->xoffset;
+                        vec[1] = r300->viewport_state->yoffset;
+                        vec[2] = r300->viewport_state->zoffset;
+                    } else {
+                        /* Zeros. */
+                    }
+                    break;
+
                 default:
                     debug_printf("r300: Implementation error: "
                         "Unknown RC_CONSTANT type %d\n", constant->u.State[0]);
diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c
index 4e1b61ca40..60ea9c171d 100644
--- a/src/gallium/drivers/r300/r300_fs.c
+++ b/src/gallium/drivers/r300/r300_fs.c
@@ -63,6 +63,11 @@ void r300_shader_read_fs_inputs(struct tgsi_shader_info* info,
                 fs_inputs->fog = i;
                 break;
 
+            case TGSI_SEMANTIC_POSITION:
+                assert(index == 0);
+                fs_inputs->wpos = i;
+                break;
+
             default:
                 assert(0);
         }
@@ -114,6 +119,9 @@ static void allocate_hardware_inputs(
     if (inputs->fog != ATTR_UNUSED) {
         allocate(mydata, inputs->fog, reg++);
     }
+    if (inputs->wpos != ATTR_UNUSED) {
+        allocate(mydata, inputs->wpos, reg++);
+    }
 }
 
 static void get_compare_state(
@@ -144,6 +152,7 @@ static void r300_translate_fragment_shader(
     struct r300_fragment_shader* fs = r300->fs;
     struct r300_fragment_program_compiler compiler;
     struct tgsi_to_rc ttr;
+    int wpos = fs->inputs.wpos;
 
     /* Setup the compiler. */
     memset(&compiler, 0, sizeof(compiler));
@@ -171,6 +180,18 @@ static void r300_translate_fragment_shader(
 
     fs->shadow_samplers = compiler.Base.Program.ShadowSamplers;
 
+    /**
+     * Transform the program to support WPOS.
+     *
+     * Introduce a small fragment at the start of the program that will be
+     * the only code that directly reads the WPOS input.
+     * All other code pieces that reference that input will be rewritten
+     * to read from a newly allocated temporary. */
+    if (wpos != ATTR_UNUSED) {
+        /* Moving the input to some other reg is not really necessary. */
+        rc_transform_fragment_wpos(&compiler.Base, wpos, wpos, TRUE);
+    }
+
     /* Invoke the compiler */
     r3xx_compile_fragment_program(&compiler);
     if (compiler.Base.Error) {
diff --git a/src/gallium/drivers/r300/r300_shader_semantics.h b/src/gallium/drivers/r300/r300_shader_semantics.h
index 85184e2cfd..6796841b29 100644
--- a/src/gallium/drivers/r300/r300_shader_semantics.h
+++ b/src/gallium/drivers/r300/r300_shader_semantics.h
@@ -40,6 +40,7 @@ struct r300_shader_semantics {
     int bcolor[ATTR_COLOR_COUNT];
     int generic[ATTR_GENERIC_COUNT];
     int fog;
+    int wpos;
 };
 
 static INLINE void r300_shader_semantics_reset(
@@ -50,6 +51,7 @@ static INLINE void r300_shader_semantics_reset(
     info->pos = ATTR_UNUSED;
     info->psize = ATTR_UNUSED;
     info->fog = ATTR_UNUSED;
+    info->wpos = ATTR_UNUSED;
 
     for (i = 0; i < ATTR_COLOR_COUNT; i++) {
         info->color[i] = ATTR_UNUSED;
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 872a393321..ad8e210dcd 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -719,6 +719,9 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
     r300->dirty_state |= R300_NEW_RS_BLOCK;
     r300->dirty_state |= R300_NEW_SCISSOR;
     r300->dirty_state |= R300_NEW_VIEWPORT;
+    if (r300->fs && r300->fs->inputs.wpos != ATTR_UNUSED) {
+        r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
+    }
 }
 
 /* Free rasterizer state. */
@@ -897,6 +900,9 @@ static void r300_set_viewport_state(struct pipe_context* pipe,
     }
 
     r300->dirty_state |= R300_NEW_VIEWPORT;
+    if (r300->fs && r300->fs->inputs.wpos != ATTR_UNUSED) {
+        r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
+    }
 }
 
 static void r300_set_vertex_buffers(struct pipe_context* pipe,
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index 727ae7ade6..242716fccc 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -410,6 +410,26 @@ static void r300_update_rs_block(struct r300_context* r300,
         }
     }
 
+    /* Rasterize WPOS. */
+    if (vs_outputs->wpos != ATTR_UNUSED) {
+        /* Always rasterize if it's written by the VS,
+         * otherwise it locks up. */
+        rX00_rs_tex(rs, tex_count, tex_count, FALSE);
+
+        /* Write it to the FS input register if it's used by the FS. */
+        if (fs_inputs->wpos != ATTR_UNUSED) {
+            rX00_rs_tex_write(rs, tex_count, fp_offset);
+            fp_offset++;
+        }
+        tex_count++;
+    } else {
+        /* Skip the FS input register, leave it uninitialized. */
+        /* If we try to set it to (0,0,0,1), it will lock up. */
+        if (fs_inputs->wpos != ATTR_UNUSED) {
+            fp_offset++;
+        }
+    }
+
     /* Rasterize at least one color, or bad things happen. */
     if (col_count == 0 && tex_count == 0) {
         rX00_rs_col(rs, 0, 0, TRUE);
diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c
index c4ed0d712f..57531fc554 100644
--- a/src/gallium/drivers/r300/r300_vs.c
+++ b/src/gallium/drivers/r300/r300_vs.c
@@ -33,6 +33,8 @@
 
 #include "radeon_compiler.h"
 
+#include "util/u_math.h"
+
 /* Convert info about VS output semantics into r300_shader_semantics. */
 static void r300_shader_read_vs_outputs(
     struct tgsi_shader_info* info,
@@ -144,6 +146,13 @@ static void r300_shader_vap_output_fmt(
         gen_count++;
     }
 
+    /* WPOS. */
+    if (vs_outputs->wpos != ATTR_UNUSED) {
+        hwfmt[1] |= (R300_INPUT_CNTL_TC0 << gen_count);
+        hwfmt[3] |= (4 << (3 * gen_count));
+        gen_count++;
+    }
+
     /* XXX magic */
     assert(gen_count <= 8);
 }
@@ -195,6 +204,13 @@ static void r300_stream_locations_notcl(
         gen_count++;
     }
 
+    /* WPOS. */
+    if (vs_outputs->wpos != ATTR_UNUSED) {
+        assert(tabi < 16);
+        stream_loc[tabi++] = 6 + gen_count;
+        gen_count++;
+    }
+
     /* XXX magic */
     assert(gen_count <= 8);
 
@@ -246,6 +262,33 @@ static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c)
     if (outputs->fog != ATTR_UNUSED) {
         c->code->outputs[outputs->fog] = reg++;
     }
+
+    /* WPOS. */
+    if (outputs->wpos != ATTR_UNUSED) {
+        c->code->outputs[outputs->wpos] = reg++;
+    }
+}
+
+static void r300_insert_wpos(struct r300_vertex_program_compiler* c,
+                             struct r300_shader_semantics* outputs)
+{
+    int i, lastOutput = 0;
+
+    /* Find the max output index. */
+    lastOutput = MAX2(lastOutput, outputs->psize);
+    for (i = 0; i < ATTR_COLOR_COUNT; i++) {
+        lastOutput = MAX2(lastOutput, outputs->color[i]);
+        lastOutput = MAX2(lastOutput, outputs->bcolor[i]);
+    }
+    for (i = 0; i < ATTR_GENERIC_COUNT; i++) {
+        lastOutput = MAX2(lastOutput, outputs->generic[i]);
+    }
+    lastOutput = MAX2(lastOutput, outputs->fog);
+
+    /* Set WPOS after the last output. */
+    lastOutput++;
+    rc_copy_output(&c->Base, 0, lastOutput); /* out[lastOutput] = out[0]; */
+    outputs->wpos = lastOutput;
 }
 
 void r300_translate_vertex_shader(struct r300_context* r300,
@@ -253,11 +296,10 @@ void r300_translate_vertex_shader(struct r300_context* r300,
 {
     struct r300_vertex_program_compiler compiler;
     struct tgsi_to_rc ttr;
+    boolean use_wpos = TRUE;
 
     /* Initialize. */
     r300_shader_read_vs_outputs(&vs->info, &vs->outputs);
-    r300_shader_vap_output_fmt(&vs->outputs, vs->hwfmt);
-    r300_stream_locations_notcl(&vs->outputs, vs->stream_loc_notcl);
 
     /* Setup the compiler */
     rc_init(&compiler.Base);
@@ -277,9 +319,17 @@ void r300_translate_vertex_shader(struct r300_context* r300,
 
     r300_tgsi_to_rc(&ttr, vs->state.tokens);
 
-    compiler.RequiredOutputs = ~(~0 << vs->info.num_outputs);
+    compiler.RequiredOutputs = ~(~0 << (vs->info.num_outputs+use_wpos));
     compiler.SetHwInputOutput = &set_vertex_inputs_outputs;
 
+    /* Insert the WPOS output. */
+    if (use_wpos) {
+        r300_insert_wpos(&compiler, &vs->outputs);
+    }
+
+    r300_shader_vap_output_fmt(&vs->outputs, vs->hwfmt);
+    r300_stream_locations_notcl(&vs->outputs, vs->stream_loc_notcl);
+
     /* Invoke the compiler */
     r3xx_compile_vertex_program(&compiler);
     if (compiler.Base.Error) {
-- 
cgit v1.2.3


From 1a4f242be60237fb1f1acf346b1e641167bc6cee Mon Sep 17 00:00:00 2001
From: Marek Olšák <maraeo@gmail.com>
Date: Thu, 24 Dec 2009 14:23:57 +0100
Subject: r300g: if no DS buffer is set, disable reading from and writing to it

It fixes another "Bad CS" issue.
---
 src/gallium/drivers/r300/r300_emit.c  | 11 +++++++++--
 src/gallium/drivers/r300/r300_state.c |  1 +
 2 files changed, 10 insertions(+), 2 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 30a1000c29..f8bfa714fe 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -121,8 +121,15 @@ void r300_emit_dsa_state(struct r300_context* r300,
     }*/
 
     OUT_CS_REG_SEQ(R300_ZB_CNTL, 3);
-    OUT_CS(dsa->z_buffer_control);
-    OUT_CS(dsa->z_stencil_control);
+
+    if (r300->framebuffer_state.zsbuf) {
+        OUT_CS(dsa->z_buffer_control);
+        OUT_CS(dsa->z_stencil_control);
+    } else {
+        OUT_CS(0);
+        OUT_CS(0);
+    }
+
     OUT_CS(dsa->stencil_ref_mask);
     OUT_CS_REG(R300_ZB_ZTOP, r300->ztop_state.z_buffer_top);
 
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index affd0b4591..a145a7f18a 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -522,6 +522,7 @@ static void
     }
     r300->dirty_state |= R300_NEW_FRAMEBUFFERS;
     r300->dirty_state |= R300_NEW_BLEND;
+    r300->dirty_state |= R300_NEW_DSA;
 }
 
 /* Create fragment shader state. */
-- 
cgit v1.2.3


From d496399156a7d9a683999e41f126232f89375f2b Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 10 Jan 2010 09:35:55 -0800
Subject: r300g: Start using atoms.

No benefits yet.
---
 src/gallium/drivers/r300/r300_blit.c    |  2 +-
 src/gallium/drivers/r300/r300_context.c | 15 +++++++++++++++
 src/gallium/drivers/r300/r300_context.h | 14 ++++++++++++--
 src/gallium/drivers/r300/r300_emit.c    | 14 +++++++++-----
 src/gallium/drivers/r300/r300_emit.h    |  2 +-
 src/gallium/drivers/r300/r300_state.c   |  7 ++++---
 6 files changed, 42 insertions(+), 12 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_blit.c b/src/gallium/drivers/r300/r300_blit.c
index ffe066d536..6c6e0567a4 100644
--- a/src/gallium/drivers/r300/r300_blit.c
+++ b/src/gallium/drivers/r300/r300_blit.c
@@ -27,7 +27,7 @@
 
 static void r300_blitter_save_states(struct r300_context* r300)
 {
-    util_blitter_save_blend(r300->blitter, r300->blend_state);
+    util_blitter_save_blend(r300->blitter, r300->blend_state.state);
     util_blitter_save_depth_stencil_alpha(r300->blitter, r300->dsa_state);
     util_blitter_save_rasterizer(r300->blitter, r300->rs_state);
     util_blitter_save_fragment_shader(r300->blitter, r300->fs);
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index d5c2d63d39..2cdc946e90 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -30,6 +30,7 @@
 
 #include "r300_blit.h"
 #include "r300_context.h"
+#include "r300_emit.h"
 #include "r300_flush.h"
 #include "r300_query.h"
 #include "r300_render.h"
@@ -107,6 +108,18 @@ static void r300_flush_cb(void *data)
     cs_context_copy->context.flush(&cs_context_copy->context, 0, NULL);
 }
 
+#define R300_INIT_ATOM(name) \
+    r300->name##_state.state = NULL; \
+    r300->name##_state.emit = r300_emit_##name##_state; \
+    r300->name##_state.dirty = FALSE; \
+    insert_at_tail(&r300->atom_list, &r300->name##_state);
+
+static void r300_setup_atoms(struct r300_context* r300)
+{
+    make_empty_list(&r300->atom_list);
+    R300_INIT_ATOM(blend);
+}
+
 struct pipe_context* r300_create_context(struct pipe_screen* screen,
                                          struct radeon_winsys* radeon_winsys)
 {
@@ -166,6 +179,8 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
             PIPE_BUFFER_USAGE_VERTEX, 4096);
     make_empty_list(&r300->query_list);
 
+    r300_setup_atoms(r300);
+
     r300_init_flush_functions(r300);
 
     r300_init_query_functions(r300);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 232530b7dc..41582504fd 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -30,9 +30,18 @@
 #include "pipe/p_context.h"
 #include "pipe/p_inlines.h"
 
+struct r300_context;
+
 struct r300_fragment_shader;
 struct r300_vertex_shader;
 
+struct r300_atom {
+    struct r300_atom *prev, *next;
+    void* state;
+    void (*emit)(struct r300_context*, void*);
+    boolean dirty;
+};
+
 struct r300_blend_state {
     uint32_t blend_control;       /* R300_RB3D_CBLEND: 0x4e04 */
     uint32_t alpha_blend_control; /* R300_RB3D_ABLEND: 0x4e08 */
@@ -135,7 +144,6 @@ struct r300_ztop_state {
     uint32_t z_buffer_top;      /* R300_ZB_ZTOP: 0x4f14 */
 };
 
-#define R300_NEW_BLEND           0x00000001
 #define R300_NEW_BLEND_COLOR     0x00000002
 #define R300_NEW_CLIP            0x00000004
 #define R300_NEW_DSA             0x00000008
@@ -273,8 +281,10 @@ struct r300_context {
     struct r300_vertex_info* vertex_info;
 
     /* Various CSO state objects. */
+    /* Beginning of atom list. */
+    struct r300_atom atom_list;
     /* Blend state. */
-    struct r300_blend_state* blend_state;
+    struct r300_atom blend_state;
     /* Blend color state. */
     struct r300_blend_color_state* blend_color_state;
     /* User clip planes. */
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index f8bfa714fe..8c9c7e9d03 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -25,6 +25,7 @@
 
 #include "util/u_format.h"
 #include "util/u_math.h"
+#include "util/u_simple_list.h"
 
 #include "r300_context.h"
 #include "r300_cs.h"
@@ -36,9 +37,9 @@
 #include "r300_texture.h"
 #include "r300_vs.h"
 
-void r300_emit_blend_state(struct r300_context* r300,
-                           struct r300_blend_state* blend)
+void r300_emit_blend_state(struct r300_context* r300, void* state)
 {
+    struct r300_blend_state* blend = (struct r300_blend_state*)state;
     CS_LOCALS(r300);
     BEGIN_CS(8);
     OUT_CS_REG_SEQ(R300_RB3D_CBLEND, 3);
@@ -978,6 +979,7 @@ void r300_emit_dirty_state(struct r300_context* r300)
 {
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
     struct r300_texture* tex;
+    struct r300_atom* atom;
     int i, dirty_tex = 0;
     boolean invalid = FALSE;
 
@@ -1060,9 +1062,11 @@ validate:
         r300->dirty_state &= ~R300_NEW_QUERY;
     }
 
-    if (r300->dirty_state & R300_NEW_BLEND) {
-        r300_emit_blend_state(r300, r300->blend_state);
-        r300->dirty_state &= ~R300_NEW_BLEND;
+    foreach(atom, &r300->atom_list) {
+        if (atom->dirty) {
+            atom->emit(r300, atom->state);
+            atom->dirty = FALSE;
+        }
     }
 
     if (r300->dirty_state & R300_NEW_BLEND_COLOR) {
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 3797d3d332..34356438e4 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -32,7 +32,7 @@ struct r300_vertex_program_code;
 void r300_emit_aos(struct r300_context* r300, unsigned offset);
 
 void r300_emit_blend_state(struct r300_context* r300,
-                           struct r300_blend_state* blend);
+                           void* blend);
 
 void r300_emit_blend_color_state(struct r300_context* r300,
                                  struct r300_blend_color_state* bc);
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index a145a7f18a..db8aca6c9f 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -317,8 +317,8 @@ static void r300_bind_blend_state(struct pipe_context* pipe,
 {
     struct r300_context* r300 = r300_context(pipe);
 
-    r300->blend_state = (struct r300_blend_state*)state;
-    r300->dirty_state |= R300_NEW_BLEND;
+    r300->blend_state.state = state;
+    r300->blend_state.dirty = TRUE;
 }
 
 /* Free blend state. */
@@ -521,8 +521,9 @@ static void
         r300->dirty_state |= R300_NEW_SCISSOR;
     }
     r300->dirty_state |= R300_NEW_FRAMEBUFFERS;
-    r300->dirty_state |= R300_NEW_BLEND;
     r300->dirty_state |= R300_NEW_DSA;
+
+    r300->blend_state.dirty = TRUE;
 }
 
 /* Create fragment shader state. */
-- 
cgit v1.2.3


From 9d3db601c782805113e60ee7f6976184f2786427 Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 10 Jan 2010 10:26:15 -0800
Subject: r300g: Atomize blend color.

---
 src/gallium/drivers/r300/r300_context.c |  5 +++--
 src/gallium/drivers/r300/r300_context.h |  3 +--
 src/gallium/drivers/r300/r300_emit.c    |  9 ++-------
 src/gallium/drivers/r300/r300_emit.h    |  6 ++----
 src/gallium/drivers/r300/r300_state.c   | 10 ++++++----
 5 files changed, 14 insertions(+), 19 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 2cdc946e90..d16889de34 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -70,7 +70,7 @@ static void r300_destroy_context(struct pipe_context* context)
         FREE(query);
     }
 
-    FREE(r300->blend_color_state);
+    FREE(r300->blend_color_state.state);
     FREE(r300->rs_block);
     FREE(r300->scissor_state);
     FREE(r300->vertex_info);
@@ -118,6 +118,7 @@ static void r300_setup_atoms(struct r300_context* r300)
 {
     make_empty_list(&r300->atom_list);
     R300_INIT_ATOM(blend);
+    R300_INIT_ATOM(blend_color);
 }
 
 struct pipe_context* r300_create_context(struct pipe_screen* screen,
@@ -168,7 +169,7 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     r300->shader_hash_table = util_hash_table_create(r300_shader_key_hash,
         r300_shader_key_compare);
 
-    r300->blend_color_state = CALLOC_STRUCT(r300_blend_color_state);
+    r300->blend_color_state.state = CALLOC_STRUCT(r300_blend_color_state);
     r300->rs_block = CALLOC_STRUCT(r300_rs_block);
     r300->scissor_state = CALLOC_STRUCT(r300_scissor_state);
     r300->vertex_info = CALLOC_STRUCT(r300_vertex_info);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 41582504fd..c916a860f6 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -144,7 +144,6 @@ struct r300_ztop_state {
     uint32_t z_buffer_top;      /* R300_ZB_ZTOP: 0x4f14 */
 };
 
-#define R300_NEW_BLEND_COLOR     0x00000002
 #define R300_NEW_CLIP            0x00000004
 #define R300_NEW_DSA             0x00000008
 #define R300_NEW_FRAMEBUFFERS    0x00000010
@@ -286,7 +285,7 @@ struct r300_context {
     /* Blend state. */
     struct r300_atom blend_state;
     /* Blend color state. */
-    struct r300_blend_color_state* blend_color_state;
+    struct r300_atom blend_color_state;
     /* User clip planes. */
     struct pipe_clip_state clip_state;
     /* Shader constants. */
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 8c9c7e9d03..5ae9c2a9bd 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -58,9 +58,9 @@ void r300_emit_blend_state(struct r300_context* r300, void* state)
     END_CS;
 }
 
-void r300_emit_blend_color_state(struct r300_context* r300,
-                                 struct r300_blend_color_state* bc)
+void r300_emit_blend_color_state(struct r300_context* r300, void* state)
 {
+    struct r300_blend_color_state* bc = (struct r300_blend_color_state*)state;
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
     CS_LOCALS(r300);
 
@@ -1069,11 +1069,6 @@ validate:
         }
     }
 
-    if (r300->dirty_state & R300_NEW_BLEND_COLOR) {
-        r300_emit_blend_color_state(r300, r300->blend_color_state);
-        r300->dirty_state &= ~R300_NEW_BLEND_COLOR;
-    }
-
     if (r300->dirty_state & R300_NEW_CLIP) {
         r300_emit_clip_state(r300, &r300->clip_state);
         r300->dirty_state &= ~R300_NEW_CLIP;
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 34356438e4..005a9d50b0 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -31,11 +31,9 @@ struct r300_vertex_program_code;
 
 void r300_emit_aos(struct r300_context* r300, unsigned offset);
 
-void r300_emit_blend_state(struct r300_context* r300,
-                           void* blend);
+void r300_emit_blend_state(struct r300_context* r300, void* state);
 
-void r300_emit_blend_color_state(struct r300_context* r300,
-                                 struct r300_blend_color_state* bc);
+void r300_emit_blend_color_state(struct r300_context* r300, void* state);
 
 void r300_emit_clip_state(struct r300_context* r300,
                           struct pipe_clip_state* clip);
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index db8aca6c9f..35d698b820 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -340,20 +340,22 @@ static void r300_set_blend_color(struct pipe_context* pipe,
                                  const struct pipe_blend_color* color)
 {
     struct r300_context* r300 = r300_context(pipe);
+    struct r300_blend_color_state* state =
+        (struct r300_blend_color_state*)r300->blend_color_state.state;
     union util_color uc;
 
     util_pack_color(color->color, PIPE_FORMAT_A8R8G8B8_UNORM, &uc);
-    r300->blend_color_state->blend_color = uc.ui;
+    state->blend_color = uc.ui;
 
     /* XXX if FP16 blending is enabled, we should use the FP16 format */
-    r300->blend_color_state->blend_color_red_alpha =
+    state->blend_color_red_alpha =
         float_to_fixed10(color->color[0]) |
         (float_to_fixed10(color->color[3]) << 16);
-    r300->blend_color_state->blend_color_green_blue =
+    state->blend_color_green_blue =
         float_to_fixed10(color->color[2]) |
         (float_to_fixed10(color->color[1]) << 16);
 
-    r300->dirty_state |= R300_NEW_BLEND_COLOR;
+    r300->blend_color_state.dirty = TRUE;
 }
 
 static void r300_set_clip_state(struct pipe_context* pipe,
-- 
cgit v1.2.3


From 249374b85346663417fc09e4c1ad3c6fb908067b Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 10 Jan 2010 10:36:57 -0800
Subject: r300g: Atomize UCP.

Meh.
---
 src/gallium/drivers/r300/r300_context.c | 3 +++
 src/gallium/drivers/r300/r300_context.h | 3 +--
 src/gallium/drivers/r300/r300_emit.c    | 9 ++-------
 src/gallium/drivers/r300/r300_emit.h    | 3 +--
 src/gallium/drivers/r300/r300_state.c   | 4 ++--
 5 files changed, 9 insertions(+), 13 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index d16889de34..489f701e46 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -71,6 +71,7 @@ static void r300_destroy_context(struct pipe_context* context)
     }
 
     FREE(r300->blend_color_state.state);
+    FREE(r300->clip_state.state);
     FREE(r300->rs_block);
     FREE(r300->scissor_state);
     FREE(r300->vertex_info);
@@ -119,6 +120,7 @@ static void r300_setup_atoms(struct r300_context* r300)
     make_empty_list(&r300->atom_list);
     R300_INIT_ATOM(blend);
     R300_INIT_ATOM(blend_color);
+    R300_INIT_ATOM(clip);
 }
 
 struct pipe_context* r300_create_context(struct pipe_screen* screen,
@@ -170,6 +172,7 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
         r300_shader_key_compare);
 
     r300->blend_color_state.state = CALLOC_STRUCT(r300_blend_color_state);
+    r300->clip_state.state = CALLOC_STRUCT(pipe_clip_state);
     r300->rs_block = CALLOC_STRUCT(r300_rs_block);
     r300->scissor_state = CALLOC_STRUCT(r300_scissor_state);
     r300->vertex_info = CALLOC_STRUCT(r300_vertex_info);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index c916a860f6..3a54659df4 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -144,7 +144,6 @@ struct r300_ztop_state {
     uint32_t z_buffer_top;      /* R300_ZB_ZTOP: 0x4f14 */
 };
 
-#define R300_NEW_CLIP            0x00000004
 #define R300_NEW_DSA             0x00000008
 #define R300_NEW_FRAMEBUFFERS    0x00000010
 #define R300_NEW_FRAGMENT_SHADER 0x00000020
@@ -287,7 +286,7 @@ struct r300_context {
     /* Blend color state. */
     struct r300_atom blend_color_state;
     /* User clip planes. */
-    struct pipe_clip_state clip_state;
+    struct r300_atom clip_state;
     /* Shader constants. */
     struct r300_constant_buffer shader_constants[PIPE_SHADER_TYPES];
     /* Depth, stencil, and alpha state. */
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 5ae9c2a9bd..1517eed923 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -77,9 +77,9 @@ void r300_emit_blend_color_state(struct r300_context* r300, void* state)
     }
 }
 
-void r300_emit_clip_state(struct r300_context* r300,
-                          struct pipe_clip_state* clip)
+void r300_emit_clip_state(struct r300_context* r300, void* state)
 {
+    struct pipe_clip_state* clip = (struct pipe_clip_state*)state;
     int i;
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
     CS_LOCALS(r300);
@@ -1069,11 +1069,6 @@ validate:
         }
     }
 
-    if (r300->dirty_state & R300_NEW_CLIP) {
-        r300_emit_clip_state(r300, &r300->clip_state);
-        r300->dirty_state &= ~R300_NEW_CLIP;
-    }
-
     if (r300->dirty_state & R300_NEW_DSA) {
         r300_emit_dsa_state(r300, r300->dsa_state);
         r300->dirty_state &= ~R300_NEW_DSA;
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 005a9d50b0..81b5f735ef 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -35,8 +35,7 @@ void r300_emit_blend_state(struct r300_context* r300, void* state);
 
 void r300_emit_blend_color_state(struct r300_context* r300, void* state);
 
-void r300_emit_clip_state(struct r300_context* r300,
-                          struct pipe_clip_state* clip);
+void r300_emit_clip_state(struct r300_context* r300, void* state);
 
 void r300_emit_dsa_state(struct r300_context* r300,
                          struct r300_dsa_state* dsa);
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 35d698b820..732292fdaf 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -364,8 +364,8 @@ static void r300_set_clip_state(struct pipe_context* pipe,
     struct r300_context* r300 = r300_context(pipe);
 
     if (r300_screen(pipe->screen)->caps->has_tcl) {
-        r300->clip_state = *state;
-        r300->dirty_state |= R300_NEW_CLIP;
+        memcpy(r300->clip_state.state, state, sizeof(struct pipe_clip_state));
+        r300->clip_state.dirty = TRUE;
     } else {
         draw_flush(r300->draw);
         draw_set_clip_state(r300->draw, state);
-- 
cgit v1.2.3


From 07ea7e6c80ef2bcb69ab12af69d27f7e118bc15a Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 10 Jan 2010 11:08:14 -0800
Subject: r300g: Atomize DSA.

Also a bit of ztop.
---
 src/gallium/drivers/r300/r300_blit.c          |  2 +-
 src/gallium/drivers/r300/r300_context.c       |  1 +
 src/gallium/drivers/r300/r300_context.h       |  3 +--
 src/gallium/drivers/r300/r300_emit.c          | 13 ++----------
 src/gallium/drivers/r300/r300_emit.h          |  3 +--
 src/gallium/drivers/r300/r300_state.c         |  6 +++---
 src/gallium/drivers/r300/r300_state_derived.c | 29 +++++++++++++++------------
 7 files changed, 25 insertions(+), 32 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_blit.c b/src/gallium/drivers/r300/r300_blit.c
index 6c6e0567a4..b2fec4a50b 100644
--- a/src/gallium/drivers/r300/r300_blit.c
+++ b/src/gallium/drivers/r300/r300_blit.c
@@ -28,7 +28,7 @@
 static void r300_blitter_save_states(struct r300_context* r300)
 {
     util_blitter_save_blend(r300->blitter, r300->blend_state.state);
-    util_blitter_save_depth_stencil_alpha(r300->blitter, r300->dsa_state);
+    util_blitter_save_depth_stencil_alpha(r300->blitter, r300->dsa_state.state);
     util_blitter_save_rasterizer(r300->blitter, r300->rs_state);
     util_blitter_save_fragment_shader(r300->blitter, r300->fs);
     util_blitter_save_vertex_shader(r300->blitter, r300->vs);
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 489f701e46..81a14c02b2 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -121,6 +121,7 @@ static void r300_setup_atoms(struct r300_context* r300)
     R300_INIT_ATOM(blend);
     R300_INIT_ATOM(blend_color);
     R300_INIT_ATOM(clip);
+    R300_INIT_ATOM(dsa);
 }
 
 struct pipe_context* r300_create_context(struct pipe_screen* screen,
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 3a54659df4..1445ba5699 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -144,7 +144,6 @@ struct r300_ztop_state {
     uint32_t z_buffer_top;      /* R300_ZB_ZTOP: 0x4f14 */
 };
 
-#define R300_NEW_DSA             0x00000008
 #define R300_NEW_FRAMEBUFFERS    0x00000010
 #define R300_NEW_FRAGMENT_SHADER 0x00000020
 #define R300_NEW_FRAGMENT_SHADER_CONSTANTS    0x00000040
@@ -290,7 +289,7 @@ struct r300_context {
     /* Shader constants. */
     struct r300_constant_buffer shader_constants[PIPE_SHADER_TYPES];
     /* Depth, stencil, and alpha state. */
-    struct r300_dsa_state* dsa_state;
+    struct r300_atom dsa_state;
     /* Fragment shader. */
     struct r300_fragment_shader* fs;
     /* Framebuffer state. We currently don't need our own version of this. */
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 1517eed923..852ac98863 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -107,9 +107,9 @@ void r300_emit_clip_state(struct r300_context* r300, void* state)
 
 }
 
-void r300_emit_dsa_state(struct r300_context* r300,
-                           struct r300_dsa_state* dsa)
+void r300_emit_dsa_state(struct r300_context* r300, void* state)
 {
+    struct r300_dsa_state* dsa = (struct r300_dsa_state*)state;
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
     CS_LOCALS(r300);
 
@@ -983,10 +983,6 @@ void r300_emit_dirty_state(struct r300_context* r300)
     int i, dirty_tex = 0;
     boolean invalid = FALSE;
 
-    if (!(r300->dirty_state)) {
-        return;
-    }
-
     /* Check size of CS. */
     /* Make sure we have at least 8*1024 spare dwords. */
     /* XXX It would be nice to know the number of dwords we really need to
@@ -1069,11 +1065,6 @@ validate:
         }
     }
 
-    if (r300->dirty_state & R300_NEW_DSA) {
-        r300_emit_dsa_state(r300, r300->dsa_state);
-        r300->dirty_state &= ~R300_NEW_DSA;
-    }
-
     if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) {
         r300_emit_fragment_depth_config(r300, r300->fs);
         if (r300screen->caps->is_r500) {
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 81b5f735ef..7cea50e627 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -37,8 +37,7 @@ void r300_emit_blend_color_state(struct r300_context* r300, void* state);
 
 void r300_emit_clip_state(struct r300_context* r300, void* state);
 
-void r300_emit_dsa_state(struct r300_context* r300,
-                         struct r300_dsa_state* dsa);
+void r300_emit_dsa_state(struct r300_context* r300, void* state);
 
 void r300_emit_fragment_program_code(struct r300_context* r300,
                                      struct rX00_fragment_program_code* generic_code);
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 732292fdaf..288b2149b5 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -463,8 +463,8 @@ static void r300_bind_dsa_state(struct pipe_context* pipe,
 {
     struct r300_context* r300 = r300_context(pipe);
 
-    r300->dsa_state = (struct r300_dsa_state*)state;
-    r300->dirty_state |= R300_NEW_DSA;
+    r300->dsa_state.state = state;
+    r300->dsa_state.dirty = TRUE;
 }
 
 /* Free DSA state. */
@@ -523,9 +523,9 @@ static void
         r300->dirty_state |= R300_NEW_SCISSOR;
     }
     r300->dirty_state |= R300_NEW_FRAMEBUFFERS;
-    r300->dirty_state |= R300_NEW_DSA;
 
     r300->blend_state.dirty = TRUE;
+    r300->dsa_state.dirty = TRUE;
 }
 
 /* Create fragment shader state. */
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index 22660a52d9..0f6ff02d9c 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -508,7 +508,7 @@ static boolean r300_dsa_alpha_test_enabled(struct r300_dsa_state* dsa)
 
 static void r300_update_ztop(struct r300_context* r300)
 {
-    r300->ztop_state.z_buffer_top = R300_ZTOP_ENABLE;
+    uint32_t ztop = r300->ztop_state.z_buffer_top;
 
     /* This is important enough that I felt it warranted a comment.
      *
@@ -534,14 +534,20 @@ static void r300_update_ztop(struct r300_context* r300)
      */
 
     /* ZS writes */
-    if (r300_dsa_writes_depth_stencil(r300->dsa_state) &&
-           (r300_dsa_alpha_test_enabled(r300->dsa_state) ||   /* (1) */
-            r300->fs->info.uses_kill)) {                      /* (2) */
-        r300->ztop_state.z_buffer_top = R300_ZTOP_DISABLE;
-    } else if (r300_fragment_shader_writes_depth(r300->fs)) { /* (5) */
-        r300->ztop_state.z_buffer_top = R300_ZTOP_DISABLE;
-    } else if (r300->query_current) {                         /* (6) */
-        r300->ztop_state.z_buffer_top = R300_ZTOP_DISABLE;
+    if (r300_dsa_writes_depth_stencil(r300->dsa_state.state) &&
+           (r300_dsa_alpha_test_enabled(r300->dsa_state.state) ||/* (1) */
+            r300->fs->info.uses_kill)) {                         /* (2) */
+        ztop = R300_ZTOP_DISABLE;
+    } else if (r300_fragment_shader_writes_depth(r300->fs)) {    /* (5) */
+        ztop = R300_ZTOP_DISABLE;
+    } else if (r300->query_current) {                            /* (6) */
+        ztop = R300_ZTOP_DISABLE;
+    } else {
+        ztop = R300_ZTOP_ENABLE;
+    }
+
+    if (r300->ztop_state.z_buffer_top != ztop) {
+        r300->ztop_state.z_buffer_top = ztop;
     }
 }
 
@@ -553,8 +559,5 @@ void r300_update_derived_state(struct r300_context* r300)
         r300_update_derived_shader_state(r300);
     }
 
-    if (r300->dirty_state &
-            (R300_NEW_DSA | R300_NEW_FRAGMENT_SHADER | R300_NEW_QUERY)) {
-        r300_update_ztop(r300);
-    }
+    r300_update_ztop(r300);
 }
-- 
cgit v1.2.3


From 8a2c961798b4ab1f1095f14d814242422020d4f9 Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 10 Jan 2010 11:49:25 -0800
Subject: r300g: Atomize ZTOP.

Also do state-change checks. ZTOP's too important to not check.
---
 src/gallium/drivers/r300/r300_context.c       |  7 +++++--
 src/gallium/drivers/r300/r300_context.h       |  2 +-
 src/gallium/drivers/r300/r300_emit.c          | 13 +++++++++++--
 src/gallium/drivers/r300/r300_emit.h          |  2 ++
 src/gallium/drivers/r300/r300_state_derived.c |  9 ++++++---
 5 files changed, 25 insertions(+), 8 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 81a14c02b2..abd5d26874 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -76,6 +76,7 @@ static void r300_destroy_context(struct pipe_context* context)
     FREE(r300->scissor_state);
     FREE(r300->vertex_info);
     FREE(r300->viewport_state);
+    FREE(r300->ztop_state.state);
     FREE(r300);
 }
 
@@ -118,6 +119,7 @@ static void r300_flush_cb(void *data)
 static void r300_setup_atoms(struct r300_context* r300)
 {
     make_empty_list(&r300->atom_list);
+    R300_INIT_ATOM(ztop);
     R300_INIT_ATOM(blend);
     R300_INIT_ATOM(blend_color);
     R300_INIT_ATOM(clip);
@@ -172,20 +174,21 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     r300->shader_hash_table = util_hash_table_create(r300_shader_key_hash,
         r300_shader_key_compare);
 
+    r300_setup_atoms(r300);
+
     r300->blend_color_state.state = CALLOC_STRUCT(r300_blend_color_state);
     r300->clip_state.state = CALLOC_STRUCT(pipe_clip_state);
     r300->rs_block = CALLOC_STRUCT(r300_rs_block);
     r300->scissor_state = CALLOC_STRUCT(r300_scissor_state);
     r300->vertex_info = CALLOC_STRUCT(r300_vertex_info);
     r300->viewport_state = CALLOC_STRUCT(r300_viewport_state);
+    r300->ztop_state.state = CALLOC_STRUCT(r300_ztop_state);
 
     /* Open up the OQ BO. */
     r300->oqbo = screen->buffer_create(screen, 4096,
             PIPE_BUFFER_USAGE_VERTEX, 4096);
     make_empty_list(&r300->query_list);
 
-    r300_setup_atoms(r300);
-
     r300_init_flush_functions(r300);
 
     r300_init_query_functions(r300);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 1445ba5699..56c49f7efa 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -311,7 +311,7 @@ struct r300_context {
     /* Viewport state. */
     struct r300_viewport_state* viewport_state;
     /* ZTOP state. */
-    struct r300_ztop_state ztop_state;
+    struct r300_atom ztop_state;
 
     /* Vertex buffers for Gallium. */
     struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 852ac98863..27fb9aa9ba 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -113,7 +113,7 @@ void r300_emit_dsa_state(struct r300_context* r300, void* state)
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
     CS_LOCALS(r300);
 
-    BEGIN_CS(r300screen->caps->is_r500 ? 10 : 8);
+    BEGIN_CS(r300screen->caps->is_r500 ? 8 : 6);
     OUT_CS_REG(R300_FG_ALPHA_FUNC, dsa->alpha_function);
 
     /* not needed since we use the 8bit alpha ref */
@@ -132,7 +132,6 @@ void r300_emit_dsa_state(struct r300_context* r300, void* state)
     }
 
     OUT_CS(dsa->stencil_ref_mask);
-    OUT_CS_REG(R300_ZB_ZTOP, r300->ztop_state.z_buffer_top);
 
     /* XXX it seems r3xx doesn't support STENCILREFMASK_BF */
     if (r300screen->caps->is_r500) {
@@ -956,6 +955,16 @@ void r300_emit_texture_count(struct r300_context* r300)
 
 }
 
+void r300_emit_ztop_state(struct r300_context* r300, void* state)
+{
+    struct r300_ztop_state* ztop = (struct r300_ztop_state*)state;
+    CS_LOCALS(r300);
+
+    BEGIN_CS(2);
+    OUT_CS_REG(R300_ZB_ZTOP, ztop->z_buffer_top);
+    END_CS;
+}
+
 void r300_flush_textures(struct r300_context* r300)
 {
     CS_LOCALS(r300);
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 7cea50e627..0ccae0031b 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -90,6 +90,8 @@ void r300_emit_viewport_state(struct r300_context* r300,
 
 void r300_emit_texture_count(struct r300_context* r300);
 
+void r300_emit_ztop_state(struct r300_context* r300, void* state);
+
 void r300_flush_textures(struct r300_context* r300);
 
 /* Emit all dirty state. */
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index 0f6ff02d9c..a4029fca49 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -508,7 +508,9 @@ static boolean r300_dsa_alpha_test_enabled(struct r300_dsa_state* dsa)
 
 static void r300_update_ztop(struct r300_context* r300)
 {
-    uint32_t ztop = r300->ztop_state.z_buffer_top;
+    struct r300_ztop_state* ztop_state =
+        (struct r300_ztop_state*)r300->ztop_state.state;
+    uint32_t ztop = ztop_state->z_buffer_top;
 
     /* This is important enough that I felt it warranted a comment.
      *
@@ -546,8 +548,9 @@ static void r300_update_ztop(struct r300_context* r300)
         ztop = R300_ZTOP_ENABLE;
     }
 
-    if (r300->ztop_state.z_buffer_top != ztop) {
-        r300->ztop_state.z_buffer_top = ztop;
+    if (ztop_state->z_buffer_top != ztop) {
+        ztop_state->z_buffer_top = ztop;
+        r300->ztop_state.dirty = TRUE;
     }
 }
 
-- 
cgit v1.2.3


From 46fafdd455bc1f3ee05c076b3c8c541ecd4132dc Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 10 Jan 2010 14:29:32 -0800
Subject: r300g: Atomize rasterizer.

I want to stab things now.
---
 src/gallium/drivers/r300/r300_blit.c          |  2 +-
 src/gallium/drivers/r300/r300_context.c       |  1 +
 src/gallium/drivers/r300/r300_context.h       |  3 +--
 src/gallium/drivers/r300/r300_emit.c          | 36 ++++++++++++++-------------
 src/gallium/drivers/r300/r300_emit.h          |  2 +-
 src/gallium/drivers/r300/r300_render.c        |  7 +++---
 src/gallium/drivers/r300/r300_state.c         | 13 +++++-----
 src/gallium/drivers/r300/r300_state_derived.c |  2 +-
 8 files changed, 34 insertions(+), 32 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_blit.c b/src/gallium/drivers/r300/r300_blit.c
index b2fec4a50b..c14414fff6 100644
--- a/src/gallium/drivers/r300/r300_blit.c
+++ b/src/gallium/drivers/r300/r300_blit.c
@@ -29,7 +29,7 @@ static void r300_blitter_save_states(struct r300_context* r300)
 {
     util_blitter_save_blend(r300->blitter, r300->blend_state.state);
     util_blitter_save_depth_stencil_alpha(r300->blitter, r300->dsa_state.state);
-    util_blitter_save_rasterizer(r300->blitter, r300->rs_state);
+    util_blitter_save_rasterizer(r300->blitter, r300->rs_state.state);
     util_blitter_save_fragment_shader(r300->blitter, r300->fs);
     util_blitter_save_vertex_shader(r300->blitter, r300->vs);
 }
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index abd5d26874..9319b5ecfc 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -124,6 +124,7 @@ static void r300_setup_atoms(struct r300_context* r300)
     R300_INIT_ATOM(blend_color);
     R300_INIT_ATOM(clip);
     R300_INIT_ATOM(dsa);
+    R300_INIT_ATOM(rs);
 }
 
 struct pipe_context* r300_create_context(struct pipe_screen* screen,
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 56c49f7efa..e8c56bbf87 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -147,7 +147,6 @@ struct r300_ztop_state {
 #define R300_NEW_FRAMEBUFFERS    0x00000010
 #define R300_NEW_FRAGMENT_SHADER 0x00000020
 #define R300_NEW_FRAGMENT_SHADER_CONSTANTS    0x00000040
-#define R300_NEW_RASTERIZER      0x00000080
 #define R300_NEW_RS_BLOCK        0x00000100
 #define R300_NEW_SAMPLER         0x00000200
 #define R300_ANY_NEW_SAMPLERS    0x0001fe00
@@ -295,7 +294,7 @@ struct r300_context {
     /* Framebuffer state. We currently don't need our own version of this. */
     struct pipe_framebuffer_state framebuffer_state;
     /* Rasterizer state. */
-    struct r300_rs_state* rs_state;
+    struct r300_atom rs_state;
     /* RS block state. */
     struct r300_rs_block* rs_block;
     /* Sampler states. */
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 27fb9aa9ba..60e4a109d4 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -145,6 +145,9 @@ static const float * get_shader_constant(
     struct rc_constant * constant,
     struct r300_constant_buffer * externals)
 {
+    struct r300_viewport_state* viewport =
+        (struct r300_viewport_state*)r300->viewport_state;
+    boolean vte_enabled = viewport->vte_control & ~R300_VTX_W0_FMT;
     static float vec[4] = { 0.0, 0.0, 0.0, 1.0 };
     struct pipe_texture *tex;
 
@@ -167,16 +170,17 @@ static const float * get_shader_constant(
 
                 /* Texture compare-fail value. */
                 /* XXX Since Gallium doesn't support GL_ARB_shadow_ambient,
-                 * this is always (0,0,0,0). */
+                 * this is always (0,0,0,0), right? */
                 case RC_STATE_SHADOW_AMBIENT:
                     vec[3] = 0;
                     break;
 
                 case RC_STATE_R300_VIEWPORT_SCALE:
-                    if (r300->rs_state->enable_vte) {
-                        vec[0] = r300->viewport_state->xscale;
-                        vec[1] = r300->viewport_state->yscale;
-                        vec[2] = r300->viewport_state->zscale;
+                /* XXX argfl stop crossing state */
+                    if (vte_enabled) {
+                        vec[0] = viewport->xscale;
+                        vec[1] = viewport->yscale;
+                        vec[2] = viewport->zscale;
                     } else {
                         vec[0] = 1;
                         vec[1] = 1;
@@ -185,10 +189,10 @@ static const float * get_shader_constant(
                     break;
 
                 case RC_STATE_R300_VIEWPORT_OFFSET:
-                    if (r300->rs_state->enable_vte) {
-                        vec[0] = r300->viewport_state->xoffset;
-                        vec[1] = r300->viewport_state->yoffset;
-                        vec[2] = r300->viewport_state->zoffset;
+                    if (vte_enabled) {
+                        vec[0] = viewport->xoffset;
+                        vec[1] = viewport->yoffset;
+                        vec[2] = viewport->zoffset;
                     } else {
                         /* Zeros. */
                     }
@@ -576,8 +580,9 @@ void r300_emit_query_end(struct r300_context* r300)
         r300_emit_query_finish(r300, query);
 }
 
-void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs)
+void r300_emit_rs_state(struct r300_context* r300, void* state)
 {
+    struct r300_rs_state* rs = (struct r300_rs_state*)state;
     CS_LOCALS(r300);
 
     BEGIN_CS(22);
@@ -655,7 +660,8 @@ static void r300_emit_scissor_regs(struct r300_context* r300,
 void r300_emit_scissor_state(struct r300_context* r300,
                              struct r300_scissor_state* scissor)
 {
-    if (r300->rs_state->rs.scissor) {
+    /* XXX argfl! */
+    if (((struct r300_rs_state*)r300->rs_state.state)->rs.scissor) {
         r300_emit_scissor_regs(r300, &scissor->scissor);
     } else {
         r300_emit_scissor_regs(r300, &scissor->framebuffer);
@@ -926,7 +932,8 @@ void r300_emit_viewport_state(struct r300_context* r300,
     OUT_CS_32F(viewport->zscale);
     OUT_CS_32F(viewport->zoffset);
 
-    if (r300->rs_state->enable_vte) {
+    /* XXX words fail me. */
+    if (((struct r300_rs_state*)r300->rs_state.state)->enable_vte) {
         OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control);
     } else {
         OUT_CS_REG(R300_VAP_VTE_CNTL, 0);
@@ -1100,11 +1107,6 @@ validate:
         r300->dirty_state &= ~R300_NEW_FRAMEBUFFERS;
     }
 
-    if (r300->dirty_state & R300_NEW_RASTERIZER) {
-        r300_emit_rs_state(r300, r300->rs_state);
-        r300->dirty_state &= ~R300_NEW_RASTERIZER;
-    }
-
     if (r300->dirty_state & R300_NEW_RS_BLOCK) {
         r300_emit_rs_block_state(r300, r300->rs_block);
         r300->dirty_state &= ~R300_NEW_RS_BLOCK;
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 0ccae0031b..a6539b218b 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -59,7 +59,7 @@ void r300_emit_query_begin(struct r300_context* r300,
 
 void r300_emit_query_end(struct r300_context* r300);
 
-void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs);
+void r300_emit_rs_state(struct r300_context* r300, void* state);
 
 void r300_emit_rs_block_state(struct r300_context* r300,
                               struct r300_rs_block* rs);
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index a4ac9ad9a7..7098f66f0c 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -71,14 +71,15 @@ uint32_t r300_translate_primitive(unsigned prim)
 
 static boolean r300_nothing_to_draw(struct r300_context *r300)
 {
-    return r300->rs_state->rs.scissor &&
+    return ((struct r300_rs_state*)r300->rs_state.state)->rs.scissor &&
            r300->scissor_state->scissor.empty_area;
 }
 
 static uint32_t r300_provoking_vertex_fixes(struct r300_context *r300,
                                             unsigned mode)
 {
-    uint32_t color_control = r300->rs_state->color_control;
+    struct r300_rs_state* rs = (struct r300_rs_state*)r300->rs_state.state;
+    uint32_t color_control = rs->color_control;
 
     /* By default (see r300_state.c:r300_create_rs_state) color_control is
      * initialized to provoking the first vertex.
@@ -98,7 +99,7 @@ static uint32_t r300_provoking_vertex_fixes(struct r300_context *r300,
      * ~ C.
      */
 
-    if (r300->rs_state->rs.flatshade_first) {
+    if (rs->rs.flatshade_first) {
         switch (mode) {
             case PIPE_PRIM_TRIANGLE_FAN:
                 color_control |= R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_SECOND;
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 288b2149b5..8740a082b9 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -519,7 +519,8 @@ static void
                           r300_screen(r300->context.screen)->caps->is_r500);
 
     /* Don't rely on the order of states being set for the first time. */
-    if (!r300->rs_state || !r300->rs_state->rs.scissor) {
+    /* XXX ( >&) */
+    if (!r300->rs_state.state) {
         r300->dirty_state |= R300_NEW_SCISSOR;
     }
     r300->dirty_state |= R300_NEW_FRAMEBUFFERS;
@@ -721,9 +722,10 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
         draw_set_rasterizer_state(r300->draw, &rs->rs);
     }
 
-    r300->rs_state = rs;
+    r300->rs_state.state = rs;
+    r300->rs_state.dirty = TRUE;
+
     /* XXX Clean these up when we move to atom emits */
-    r300->dirty_state |= R300_NEW_RASTERIZER;
     r300->dirty_state |= R300_NEW_RS_BLOCK;
     r300->dirty_state |= R300_NEW_SCISSOR;
     r300->dirty_state |= R300_NEW_VIEWPORT;
@@ -868,10 +870,7 @@ static void r300_set_scissor_state(struct pipe_context* pipe,
     r300_set_scissor_regs(state, &r300->scissor_state->scissor,
                           r300_screen(r300->context.screen)->caps->is_r500);
 
-    /* Don't rely on the order of states being set for the first time. */
-    if (!r300->rs_state || r300->rs_state->rs.scissor) {
-        r300->dirty_state |= R300_NEW_SCISSOR;
-    }
+    r300->dirty_state |= R300_NEW_SCISSOR;
 }
 
 static void r300_set_viewport_state(struct pipe_context* pipe,
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index a4029fca49..e82aa07a62 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -139,7 +139,7 @@ static void r300_vertex_psc(struct r300_context* r300)
 
     /* If TCL is bypassed, map vertex streams to equivalent VS output
      * locations. */
-    if (r300->rs_state->enable_vte) {
+    if (((struct r300_rs_state*)r300->rs_state.state)->enable_vte) {
         stream_tab = identity;
     } else {
         stream_tab = r300->vs->stream_loc_notcl;
-- 
cgit v1.2.3


From e7d760ff0974aa6eb53fc43ec8f796b4e2410365 Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 10 Jan 2010 17:51:40 -0800
Subject: r300g: Move ROPCNTL to the top of the emit order.

According to the docs, this decreases stalls, and indeed we get a tiny
bit more glxgears from it.
---
 src/gallium/drivers/r300/r300_emit.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 60e4a109d4..cfeb99a0ad 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -42,6 +42,7 @@ void r300_emit_blend_state(struct r300_context* r300, void* state)
     struct r300_blend_state* blend = (struct r300_blend_state*)state;
     CS_LOCALS(r300);
     BEGIN_CS(8);
+    OUT_CS_REG(R300_RB3D_ROPCNTL, blend->rop);
     OUT_CS_REG_SEQ(R300_RB3D_CBLEND, 3);
     if (r300->framebuffer_state.nr_cbufs) {
         OUT_CS(blend->blend_control);
@@ -53,7 +54,6 @@ void r300_emit_blend_state(struct r300_context* r300, void* state)
         OUT_CS(0);
         /* XXX also disable fastfill here once it's supported */
     }
-    OUT_CS_REG(R300_RB3D_ROPCNTL, blend->rop);
     OUT_CS_REG(R300_RB3D_DITHER_CTL, blend->dither);
     END_CS;
 }
-- 
cgit v1.2.3


From 47f59cfc8eba6574c9ca0ae8799e8fbd9b393fb7 Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 10 Jan 2010 18:07:28 -0800
Subject: r300g: Atomize viewport.

Goddammit, some of these hax are really annoying.
---
 src/gallium/drivers/r300/r300_context.c |  5 +++--
 src/gallium/drivers/r300/r300_context.h |  3 +--
 src/gallium/drivers/r300/r300_emit.c    | 13 ++++---------
 src/gallium/drivers/r300/r300_emit.h    |  3 +--
 src/gallium/drivers/r300/r300_state.c   | 32 +++++++++++++++++---------------
 5 files changed, 26 insertions(+), 30 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 9319b5ecfc..f003451083 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -75,7 +75,7 @@ static void r300_destroy_context(struct pipe_context* context)
     FREE(r300->rs_block);
     FREE(r300->scissor_state);
     FREE(r300->vertex_info);
-    FREE(r300->viewport_state);
+    FREE(r300->viewport_state.state);
     FREE(r300->ztop_state.state);
     FREE(r300);
 }
@@ -125,6 +125,7 @@ static void r300_setup_atoms(struct r300_context* r300)
     R300_INIT_ATOM(clip);
     R300_INIT_ATOM(dsa);
     R300_INIT_ATOM(rs);
+    R300_INIT_ATOM(viewport);
 }
 
 struct pipe_context* r300_create_context(struct pipe_screen* screen,
@@ -182,7 +183,7 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     r300->rs_block = CALLOC_STRUCT(r300_rs_block);
     r300->scissor_state = CALLOC_STRUCT(r300_scissor_state);
     r300->vertex_info = CALLOC_STRUCT(r300_vertex_info);
-    r300->viewport_state = CALLOC_STRUCT(r300_viewport_state);
+    r300->viewport_state.state = CALLOC_STRUCT(r300_viewport_state);
     r300->ztop_state.state = CALLOC_STRUCT(r300_ztop_state);
 
     /* Open up the OQ BO. */
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index e8c56bbf87..70b381fe74 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -156,7 +156,6 @@ struct r300_ztop_state {
 #define R300_NEW_VERTEX_FORMAT   0x04000000
 #define R300_NEW_VERTEX_SHADER   0x08000000
 #define R300_NEW_VERTEX_SHADER_CONSTANTS    0x10000000
-#define R300_NEW_VIEWPORT        0x20000000
 #define R300_NEW_QUERY           0x40000000
 #define R300_NEW_KITCHEN_SINK    0x7fffffff
 
@@ -308,7 +307,7 @@ struct r300_context {
     /* Vertex shader. */
     struct r300_vertex_shader* vs;
     /* Viewport state. */
-    struct r300_viewport_state* viewport_state;
+    struct r300_atom viewport_state;
     /* ZTOP state. */
     struct r300_atom ztop_state;
 
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index cfeb99a0ad..aa21682e4b 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -146,7 +146,7 @@ static const float * get_shader_constant(
     struct r300_constant_buffer * externals)
 {
     struct r300_viewport_state* viewport =
-        (struct r300_viewport_state*)r300->viewport_state;
+        (struct r300_viewport_state*)r300->viewport_state.state;
     boolean vte_enabled = viewport->vte_control & ~R300_VTX_W0_FMT;
     static float vec[4] = { 0.0, 0.0, 0.0, 1.0 };
     struct pipe_texture *tex;
@@ -918,9 +918,9 @@ void r300_emit_vs_constant_buffer(struct r300_context* r300,
     END_CS;
 }
 
-void r300_emit_viewport_state(struct r300_context* r300,
-                              struct r300_viewport_state* viewport)
+void r300_emit_viewport_state(struct r300_context* r300, void* state)
 {
+    struct r300_viewport_state* viewport = (struct r300_viewport_state*)state;
     CS_LOCALS(r300);
 
     BEGIN_CS(9);
@@ -932,7 +932,7 @@ void r300_emit_viewport_state(struct r300_context* r300,
     OUT_CS_32F(viewport->zscale);
     OUT_CS_32F(viewport->zoffset);
 
-    /* XXX words fail me. */
+    /* XXX words still fail me. */
     if (((struct r300_rs_state*)r300->rs_state.state)->enable_vte) {
         OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control);
     } else {
@@ -1138,11 +1138,6 @@ validate:
         r300->dirty_state &= ~(R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES);
     }
 
-    if (r300->dirty_state & R300_NEW_VIEWPORT) {
-        r300_emit_viewport_state(r300, r300->viewport_state);
-        r300->dirty_state &= ~R300_NEW_VIEWPORT;
-    }
-
     if (dirty_tex) {
         r300_flush_textures(r300);
     }
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index a6539b218b..1a76de0da6 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -85,8 +85,7 @@ void r300_emit_vs_constant_buffer(struct r300_context* r300,
 void r300_emit_vertex_shader(struct r300_context* r300,
                              struct r300_vertex_shader* vs);
 
-void r300_emit_viewport_state(struct r300_context* r300,
-                              struct r300_viewport_state* viewport);
+void r300_emit_viewport_state(struct r300_context* r300, void* state);
 
 void r300_emit_texture_count(struct r300_context* r300);
 
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 8740a082b9..4c0b307ea8 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -724,11 +724,11 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
 
     r300->rs_state.state = rs;
     r300->rs_state.dirty = TRUE;
+    r300->viewport_state.dirty = TRUE; /* XXX */
 
     /* XXX Clean these up when we move to atom emits */
     r300->dirty_state |= R300_NEW_RS_BLOCK;
     r300->dirty_state |= R300_NEW_SCISSOR;
-    r300->dirty_state |= R300_NEW_VIEWPORT;
     if (r300->fs && r300->fs->inputs.wpos != ATTR_UNUSED) {
         r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
     }
@@ -877,36 +877,38 @@ static void r300_set_viewport_state(struct pipe_context* pipe,
                                     const struct pipe_viewport_state* state)
 {
     struct r300_context* r300 = r300_context(pipe);
+    struct r300_viewport_state* viewport =
+        (struct r300_viewport_state*)r300->viewport_state.state;
 
     /* Do the transform in HW. */
-    r300->viewport_state->vte_control = R300_VTX_W0_FMT;
+    viewport->vte_control = R300_VTX_W0_FMT;
 
     if (state->scale[0] != 1.0f) {
-        r300->viewport_state->xscale = state->scale[0];
-        r300->viewport_state->vte_control |= R300_VPORT_X_SCALE_ENA;
+        viewport->xscale = state->scale[0];
+        viewport->vte_control |= R300_VPORT_X_SCALE_ENA;
     }
     if (state->scale[1] != 1.0f) {
-        r300->viewport_state->yscale = state->scale[1];
-        r300->viewport_state->vte_control |= R300_VPORT_Y_SCALE_ENA;
+        viewport->yscale = state->scale[1];
+        viewport->vte_control |= R300_VPORT_Y_SCALE_ENA;
     }
     if (state->scale[2] != 1.0f) {
-        r300->viewport_state->zscale = state->scale[2];
-        r300->viewport_state->vte_control |= R300_VPORT_Z_SCALE_ENA;
+        viewport->zscale = state->scale[2];
+        viewport->vte_control |= R300_VPORT_Z_SCALE_ENA;
     }
     if (state->translate[0] != 0.0f) {
-        r300->viewport_state->xoffset = state->translate[0];
-        r300->viewport_state->vte_control |= R300_VPORT_X_OFFSET_ENA;
+        viewport->xoffset = state->translate[0];
+        viewport->vte_control |= R300_VPORT_X_OFFSET_ENA;
     }
     if (state->translate[1] != 0.0f) {
-        r300->viewport_state->yoffset = state->translate[1];
-        r300->viewport_state->vte_control |= R300_VPORT_Y_OFFSET_ENA;
+        viewport->yoffset = state->translate[1];
+        viewport->vte_control |= R300_VPORT_Y_OFFSET_ENA;
     }
     if (state->translate[2] != 0.0f) {
-        r300->viewport_state->zoffset = state->translate[2];
-        r300->viewport_state->vte_control |= R300_VPORT_Z_OFFSET_ENA;
+        viewport->zoffset = state->translate[2];
+        viewport->vte_control |= R300_VPORT_Z_OFFSET_ENA;
     }
 
-    r300->dirty_state |= R300_NEW_VIEWPORT;
+    r300->viewport_state.dirty = TRUE;
     if (r300->fs && r300->fs->inputs.wpos != ATTR_UNUSED) {
         r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
     }
-- 
cgit v1.2.3


From 8ca491386d0fb9e675e7dfbdd05bc09af74d75d3 Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 10 Jan 2010 22:32:58 -0800
Subject: r300g: Move TCL bypass switch to main context.

---
 src/gallium/drivers/r300/r300_context.h       |  7 ++---
 src/gallium/drivers/r300/r300_emit.c          | 44 ++++++++++++---------------
 src/gallium/drivers/r300/r300_state.c         |  7 +++--
 src/gallium/drivers/r300/r300_state_derived.c |  9 +++---
 4 files changed, 31 insertions(+), 36 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 70b381fe74..05ea7ad0f9 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -71,11 +71,6 @@ struct r300_rs_state {
     /* Draw-specific rasterizer state */
     struct pipe_rasterizer_state rs;
 
-    /* Whether or not to enable the VTE. This is referenced at the very
-     * last moment during emission of VTE state, to decide whether or not
-     * the VTE should be used for transformation. */
-    boolean enable_vte;
-
     uint32_t vap_control_status;    /* R300_VAP_CNTL_STATUS: 0x2140 */
     uint32_t point_size;            /* R300_GA_POINT_SIZE: 0x421c */
     uint32_t point_minmax;          /* R300_GA_POINT_MINMAX: 0x4230 */
@@ -322,6 +317,8 @@ struct r300_context {
     uint32_t dirty_state;
     /* Flag indicating whether or not the HW is dirty. */
     uint32_t dirty_hw;
+    /* Whether the TCL engine should be in bypass mode. */
+    boolean tcl_bypass;
 
     /** Combination of DBG_xxx flags */
     unsigned debug;
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index aa21682e4b..86d4000d08 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -147,7 +147,6 @@ static const float * get_shader_constant(
 {
     struct r300_viewport_state* viewport =
         (struct r300_viewport_state*)r300->viewport_state.state;
-    boolean vte_enabled = viewport->vte_control & ~R300_VTX_W0_FMT;
     static float vec[4] = { 0.0, 0.0, 0.0, 1.0 };
     struct pipe_texture *tex;
 
@@ -176,25 +175,22 @@ static const float * get_shader_constant(
                     break;
 
                 case RC_STATE_R300_VIEWPORT_SCALE:
-                /* XXX argfl stop crossing state */
-                    if (vte_enabled) {
-                        vec[0] = viewport->xscale;
-                        vec[1] = viewport->yscale;
-                        vec[2] = viewport->zscale;
-                    } else {
+                    if (r300->tcl_bypass) {
                         vec[0] = 1;
                         vec[1] = 1;
                         vec[2] = 1;
+                    } else {
+                        vec[0] = viewport->xscale;
+                        vec[1] = viewport->yscale;
+                        vec[2] = viewport->zscale;
                     }
                     break;
 
                 case RC_STATE_R300_VIEWPORT_OFFSET:
-                    if (vte_enabled) {
+                    if (!r300->tcl_bypass) {
                         vec[0] = viewport->xoffset;
                         vec[1] = viewport->yoffset;
                         vec[2] = viewport->zoffset;
-                    } else {
-                        /* Zeros. */
                     }
                     break;
 
@@ -923,22 +919,22 @@ void r300_emit_viewport_state(struct r300_context* r300, void* state)
     struct r300_viewport_state* viewport = (struct r300_viewport_state*)state;
     CS_LOCALS(r300);
 
-    BEGIN_CS(9);
-    OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6);
-    OUT_CS_32F(viewport->xscale);
-    OUT_CS_32F(viewport->xoffset);
-    OUT_CS_32F(viewport->yscale);
-    OUT_CS_32F(viewport->yoffset);
-    OUT_CS_32F(viewport->zscale);
-    OUT_CS_32F(viewport->zoffset);
-
-    /* XXX words still fail me. */
-    if (((struct r300_rs_state*)r300->rs_state.state)->enable_vte) {
-        OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control);
-    } else {
+    if (r300->tcl_bypass) {
+        BEGIN_CS(2);
         OUT_CS_REG(R300_VAP_VTE_CNTL, 0);
+        END_CS;
+    } else {
+        BEGIN_CS(9);
+        OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6);
+        OUT_CS_32F(viewport->xscale);
+        OUT_CS_32F(viewport->xoffset);
+        OUT_CS_32F(viewport->yscale);
+        OUT_CS_32F(viewport->yoffset);
+        OUT_CS_32F(viewport->zscale);
+        OUT_CS_32F(viewport->zoffset);
+        OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control);
+        END_CS;
     }
-    END_CS;
 }
 
 void r300_emit_texture_count(struct r300_context* r300)
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 4c0b307ea8..da1f40c8ef 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -605,8 +605,6 @@ static void* r300_create_rs_state(struct pipe_context* pipe,
     /* Copy rasterizer state for Draw. */
     rs->rs = *state;
 
-    rs->enable_vte = !state->bypass_vs_clip_and_viewport;
-
 #ifdef PIPE_ARCH_LITTLE_ENDIAN
     rs->vap_control_status = R300_VC_NO_SWAP;
 #else
@@ -722,9 +720,12 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
         draw_set_rasterizer_state(r300->draw, &rs->rs);
     }
 
+    r300->tcl_bypass = rs->rs.bypass_vs_clip_and_viewport;
+
     r300->rs_state.state = rs;
     r300->rs_state.dirty = TRUE;
-    r300->viewport_state.dirty = TRUE; /* XXX */
+    /* XXX Why is this still needed, dammit!? */
+    r300->viewport_state.dirty = TRUE;
 
     /* XXX Clean these up when we move to atom emits */
     r300->dirty_state |= R300_NEW_RS_BLOCK;
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index 4396978fd1..192846411b 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -139,10 +139,10 @@ static void r300_vertex_psc(struct r300_context* r300)
 
     /* If TCL is bypassed, map vertex streams to equivalent VS output
      * locations. */
-    if (((struct r300_rs_state*)r300->rs_state.state)->enable_vte) {
-        stream_tab = identity;
-    } else {
+    if (r300->tcl_bypass) {
         stream_tab = r300->vs->stream_loc_notcl;
+    } else {
+        stream_tab = identity;
     }
 
     /* Vertex shaders have no semantics on their inputs,
@@ -556,9 +556,10 @@ static void r300_update_ztop(struct r300_context* r300)
 
 void r300_update_derived_state(struct r300_context* r300)
 {
+    /* XXX */
     if (r300->dirty_state &
         (R300_NEW_FRAGMENT_SHADER | R300_NEW_VERTEX_SHADER |
-         R300_NEW_VERTEX_FORMAT)) {
+         R300_NEW_VERTEX_FORMAT) || r300->rs_state.dirty) {
         r300_update_derived_shader_state(r300);
     }
 
-- 
cgit v1.2.3


From 1e9ffb63401b7bdaf05c53e07b588128566b437d Mon Sep 17 00:00:00 2001
From: Corbin Simpson <MostAwesomeDude@gmail.com>
Date: Sun, 10 Jan 2010 23:02:48 -0800
Subject: r300g: Atomize scissors.

Argfl. Some of this code is so questionable.
---
 src/gallium/drivers/r300/r300_context.c |  5 +++--
 src/gallium/drivers/r300/r300_context.h |  3 +--
 src/gallium/drivers/r300/r300_emit.c    |  9 ++-------
 src/gallium/drivers/r300/r300_emit.h    |  3 +--
 src/gallium/drivers/r300/r300_render.c  |  2 +-
 src/gallium/drivers/r300/r300_state.c   | 26 ++++++++++++++------------
 6 files changed, 22 insertions(+), 26 deletions(-)

(limited to 'src/gallium/drivers/r300/r300_emit.c')

diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index f003451083..af95bbe789 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -73,7 +73,7 @@ static void r300_destroy_context(struct pipe_context* context)
     FREE(r300->blend_color_state.state);
     FREE(r300->clip_state.state);
     FREE(r300->rs_block);
-    FREE(r300->scissor_state);
+    FREE(r300->scissor_state.state);
     FREE(r300->vertex_info);
     FREE(r300->viewport_state.state);
     FREE(r300->ztop_state.state);
@@ -125,6 +125,7 @@ static void r300_setup_atoms(struct r300_context* r300)
     R300_INIT_ATOM(clip);
     R300_INIT_ATOM(dsa);
     R300_INIT_ATOM(rs);
+    R300_INIT_ATOM(scissor);
     R300_INIT_ATOM(viewport);
 }
 
@@ -181,7 +182,7 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     r300->blend_color_state.state = CALLOC_STRUCT(r300_blend_color_state);
     r300->clip_state.state = CALLOC_STRUCT(pipe_clip_state);
     r300->rs_block = CALLOC_STRUCT(r300_rs_block);
-    r300->scissor_state = CALLOC_STRUCT(r300_scissor_state);
+    r300->scissor_state.state = CALLOC_STRUCT(r300_scissor_state);
     r300->vertex_info = CALLOC_STRUCT(r300_vertex_info);
     r300->viewport_state.state = CALLOC_STRUCT(r300_viewport_state);
     r300->ztop_state.state = CALLOC_STRUCT(r300_ztop_state);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 05ea7ad0f9..70a04d6f44 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -145,7 +145,6 @@ struct r300_ztop_state {
 #define R300_NEW_RS_BLOCK        0x00000100
 #define R300_NEW_SAMPLER         0x00000200
 #define R300_ANY_NEW_SAMPLERS    0x0001fe00
-#define R300_NEW_SCISSOR         0x00020000
 #define R300_NEW_TEXTURE         0x00040000
 #define R300_ANY_NEW_TEXTURES    0x03fc0000
 #define R300_NEW_VERTEX_FORMAT   0x04000000
@@ -295,7 +294,7 @@ struct r300_context {
     struct r300_sampler_state* sampler_states[8];
     int sampler_count;
     /* Scissor state. */
-    struct r300_scissor_state* scissor_state;
+    struct r300_atom scissor_state;
     /* Texture states. */
     struct r300_texture* textures[8];
     int texture_count;
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 86d4000d08..0e5533c790 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -653,9 +653,9 @@ static void r300_emit_scissor_regs(struct r300_context* r300,
     END_CS;
 }
 
-void r300_emit_scissor_state(struct r300_context* r300,
-                             struct r300_scissor_state* scissor)
+void r300_emit_scissor_state(struct r300_context* r300, void* state)
 {
+    struct r300_scissor_state* scissor = (struct r300_scissor_state*)state;
     /* XXX argfl! */
     if (((struct r300_rs_state*)r300->rs_state.state)->rs.scissor) {
         r300_emit_scissor_regs(r300, &scissor->scissor);
@@ -1108,11 +1108,6 @@ validate:
         r300->dirty_state &= ~R300_NEW_RS_BLOCK;
     }
 
-    if (r300->dirty_state & R300_NEW_SCISSOR) {
-        r300_emit_scissor_state(r300, r300->scissor_state);
-        r300->dirty_state &= ~R300_NEW_SCISSOR;
-    }
-
     /* Samplers and textures are tracked separately but emitted together. */
     if (r300->dirty_state &
             (R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES)) {
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 1a76de0da6..05a6bfeae8 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -64,8 +64,7 @@ void r300_emit_rs_state(struct r300_context* r300, void* state);
 void r300_emit_rs_block_state(struct r300_context* r300,
                               struct r300_rs_block* rs);
 
-void r300_emit_scissor_state(struct r300_context* r300,
-                             struct r300_scissor_state* scissor);
+void r300_emit_scissor_state(struct r300_context* r300, void* state);
 
 void r300_emit_texture(struct r300_context* r300,
                        struct r300_sampler_state* sampler,
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index 7098f66f0c..2bd3acea41 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -72,7 +72,7 @@ uint32_t r300_translate_primitive(unsigned prim)
 static boolean r300_nothing_to_draw(struct r300_context *r300)
 {
     return ((struct r300_rs_state*)r300->rs_state.state)->rs.scissor &&
-           r300->scissor_state->scissor.empty_area;
+           ((struct r300_scissor_state*)r300->scissor_state.state)->scissor.empty_area;
 }
 
 static uint32_t r300_provoking_vertex_fixes(struct r300_context *r300,
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index da1f40c8ef..86773adc8d 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -504,7 +504,9 @@ static void
                                const struct pipe_framebuffer_state* state)
 {
     struct r300_context* r300 = r300_context(pipe);
-    struct pipe_scissor_state scissor;
+    struct r300_scissor_state* scissor =
+        (struct r300_scissor_state*)r300->scissor_state.state;
+    struct pipe_scissor_state pscissor;
 
     if (r300->draw) {
         draw_flush(r300->draw);
@@ -512,21 +514,19 @@ static void
 
     r300->framebuffer_state = *state;
 
-    scissor.minx = scissor.miny = 0;
-    scissor.maxx = state->width;
-    scissor.maxy = state->height;
-    r300_set_scissor_regs(&scissor, &r300->scissor_state->framebuffer,
+    /* XXX Arg. This is silly. */
+    pscissor.minx = pscissor.miny = 0;
+    pscissor.maxx = state->width;
+    pscissor.maxy = state->height;
+    r300_set_scissor_regs(&pscissor, &scissor->framebuffer,
                           r300_screen(r300->context.screen)->caps->is_r500);
 
     /* Don't rely on the order of states being set for the first time. */
-    /* XXX ( >&) */
-    if (!r300->rs_state.state) {
-        r300->dirty_state |= R300_NEW_SCISSOR;
-    }
     r300->dirty_state |= R300_NEW_FRAMEBUFFERS;
 
     r300->blend_state.dirty = TRUE;
     r300->dsa_state.dirty = TRUE;
+    r300->scissor_state.dirty = TRUE;
 }
 
 /* Create fragment shader state. */
@@ -725,11 +725,11 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
     r300->rs_state.state = rs;
     r300->rs_state.dirty = TRUE;
     /* XXX Why is this still needed, dammit!? */
+    r300->scissor_state.dirty = TRUE;
     r300->viewport_state.dirty = TRUE;
 
     /* XXX Clean these up when we move to atom emits */
     r300->dirty_state |= R300_NEW_RS_BLOCK;
-    r300->dirty_state |= R300_NEW_SCISSOR;
     if (r300->fs && r300->fs->inputs.wpos != ATTR_UNUSED) {
         r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
     }
@@ -867,11 +867,13 @@ static void r300_set_scissor_state(struct pipe_context* pipe,
                                    const struct pipe_scissor_state* state)
 {
     struct r300_context* r300 = r300_context(pipe);
+    struct r300_scissor_state* scissor =
+        (struct r300_scissor_state*)r300->scissor_state.state;
 
-    r300_set_scissor_regs(state, &r300->scissor_state->scissor,
+    r300_set_scissor_regs(state, &scissor->scissor,
                           r300_screen(r300->context.screen)->caps->is_r500);
 
-    r300->dirty_state |= R300_NEW_SCISSOR;
+    r300->scissor_state.dirty = TRUE;
 }
 
 static void r300_set_viewport_state(struct pipe_context* pipe,
-- 
cgit v1.2.3