cell: asst fixes to get driver building/running again.

Note that SPU vertex transformation is disabled at this time.

src/gallium/drivers/cell/ppu/cell_clear.c

@@ -48,6 +48,7 @@ void
 cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps,
                    unsigned clearValue)
 {
+   struct pipe_screen *screen = pipe->screen;
    struct cell_context *cell = cell_context(pipe);
    uint surfIndex;
 
@@ -56,7 +57,8 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps,
 
 
    if (!cell->cbuf_map[0])
-      cell->cbuf_map[0] = pipe_surface_map(ps);
+      cell->cbuf_map[0] = screen->surface_map(screen, ps,
+                                              PIPE_BUFFER_USAGE_GPU_WRITE);
 
    if (ps == cell->framebuffer.zsbuf) {
       surfIndex = 1;

src/gallium/drivers/cell/ppu/cell_context.c

@@ -73,11 +73,13 @@ cell_draw_create(struct cell_context *cell)
 {
    struct draw_context *draw = draw_create();
 
+#if 0 /* broken */
    if (getenv("GALLIUM_CELL_VS")) {
       /* plug in SPU-based vertex transformation code */
       draw->shader_queue_flush = cell_vertex_shader_queue_flush;
       draw->driver_private = cell;
    }
+#endif
 
    return draw;
 }
@@ -108,6 +110,8 @@ cell_create_context(struct pipe_screen *screen,
 
    cell->pipe.draw_arrays = cell_draw_arrays;
    cell->pipe.draw_elements = cell_draw_elements;
+   cell->pipe.draw_range_elements = cell_draw_range_elements;
+   cell->pipe.set_edgeflags = cell_set_edgeflags;
 
    cell->pipe.clear = cell_clear_surface;
    cell->pipe.flush = cell_flush;

src/gallium/drivers/cell/ppu/cell_draw_arrays.c

@@ -59,7 +59,8 @@ cell_map_constant_buffers(struct cell_context *sp)
    }
 
    draw_set_mapped_constant_buffer(sp->draw,
-                                   sp->mapped_constants[PIPE_SHADER_VERTEX]);
+                                   sp->mapped_constants[PIPE_SHADER_VERTEX],
+                                   sp->constants[PIPE_SHADER_VERTEX].size);
 }
 
 static void
@@ -92,10 +93,12 @@ cell_draw_arrays(struct pipe_context *pipe, unsigned mode,
  * XXX should the element buffer be specified/bound with a separate function?
  */
 boolean
-cell_draw_elements(struct pipe_context *pipe,
-                       struct pipe_buffer *indexBuffer,
-                       unsigned indexSize,
-                       unsigned mode, unsigned start, unsigned count)
+cell_draw_range_elements(struct pipe_context *pipe,
+                         struct pipe_buffer *indexBuffer,
+                         unsigned indexSize,
+                         unsigned min_index,
+                         unsigned max_index,
+                         unsigned mode, unsigned start, unsigned count)
 {
    struct cell_context *sp = cell_context(pipe);
    struct draw_context *draw = sp->draw;
@@ -152,3 +155,25 @@ cell_draw_elements(struct pipe_context *pipe,
 
    return TRUE;
 }
+
+
+boolean
+cell_draw_elements(struct pipe_context *pipe,
+                   struct pipe_buffer *indexBuffer,
+                   unsigned indexSize,
+                   unsigned mode, unsigned start, unsigned count)
+{
+   return cell_draw_range_elements( pipe, indexBuffer,
+                                    indexSize,
+                                    0, 0xffffffff,
+                                    mode, start, count );
+}
+
+
+
+void
+cell_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags)
+{
+   struct cell_context *cell = cell_context(pipe);
+   draw_set_edgeflags(cell->draw, edgeflags);
+}

src/gallium/drivers/cell/ppu/cell_draw_arrays.h

@@ -29,14 +29,26 @@
 #define CELL_DRAW_ARRAYS_H
 
 
-boolean cell_draw_arrays(struct pipe_context *pipe, unsigned mode,
-                         unsigned start, unsigned count);
+extern boolean
+cell_draw_arrays(struct pipe_context *pipe, unsigned mode,
+                 unsigned start, unsigned count);
 
-boolean cell_draw_elements(struct pipe_context *pipe,
-                           struct pipe_buffer *indexBuffer,
-                           unsigned indexSize,
-                           unsigned mode, unsigned start, unsigned count);
+extern boolean
+cell_draw_elements(struct pipe_context *pipe,
+                   struct pipe_buffer *indexBuffer,
+                   unsigned indexSize,
+                   unsigned mode, unsigned start, unsigned count);
 
+extern boolean
+cell_draw_range_elements(struct pipe_context *pipe,
+                         struct pipe_buffer *indexBuffer,
+                         unsigned indexSize,
+                         unsigned min_index,
+                         unsigned max_index,
+                         unsigned mode, unsigned start, unsigned count);
+
+extern void
+cell_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags);
 
 
 #endif /* CELL_DRAW_ARRAYS_H */

src/gallium/drivers/cell/ppu/cell_pipe_state.c

@@ -294,6 +294,8 @@ cell_set_framebuffer_state(struct pipe_context *pipe,
       struct pipe_surface *csurf = fb->cbufs[0];
       struct pipe_surface *zsurf = fb->zsbuf;
       uint i;
+      uint flags = (PIPE_BUFFER_USAGE_GPU_WRITE |
+                    PIPE_BUFFER_USAGE_GPU_READ);
 
       /* unmap old surfaces */
       for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
@@ -313,10 +315,10 @@ cell_set_framebuffer_state(struct pipe_context *pipe,
 
       /* map new surfaces */
       if (csurf)
-         cell->cbuf_map[0] = pipe_surface_map(csurf);
+         cell->cbuf_map[0] = pipe_surface_map(csurf, flags);
 
       if (zsurf)
-         cell->zsbuf_map = pipe_surface_map(zsurf);
+         cell->zsbuf_map = pipe_surface_map(zsurf, flags);
 
       cell->dirty |= CELL_NEW_FRAMEBUFFER;
    }

src/gallium/drivers/cell/ppu/cell_state_emit.c

@@ -162,13 +162,13 @@ cell_emit_state(struct cell_context *cell)
       const struct draw_context *const draw = cell->draw;
       struct cell_shader_info info;
 
-      info.num_outputs = draw->num_vs_outputs;
-      info.declarations = (uintptr_t) draw->machine.Declarations;
-      info.num_declarations = draw->machine.NumDeclarations;
-      info.instructions = (uintptr_t) draw->machine.Instructions;
-      info.num_instructions = draw->machine.NumInstructions;
-      info.immediates = (uintptr_t) draw->machine.Imms;
-      info.num_immediates = draw->machine.ImmLimit / 4;
+      info.num_outputs = draw_num_vs_outputs(draw);
+      info.declarations = (uintptr_t) draw->vs.machine.Declarations;
+      info.num_declarations = draw->vs.machine.NumDeclarations;
+      info.instructions = (uintptr_t) draw->vs.machine.Instructions;
+      info.num_instructions = draw->vs.machine.NumInstructions;
+      info.immediates = (uintptr_t) draw->vs.machine.Imms;
+      info.num_immediates = draw->vs.machine.ImmLimit / 4;
 
       emit_state_cmd(cell, CELL_CMD_STATE_BIND_VS,
 		     & info, sizeof(info));

src/gallium/drivers/cell/ppu/cell_state_shader.c

@@ -41,7 +41,7 @@
 static INLINE struct cell_fragment_shader_state *
 cell_fragment_shader_state(void *shader)
 {
-   return (struct pipe_shader_state *) shader;
+   return (struct cell_fragment_shader_state *) shader;
 }
 
 
@@ -49,7 +49,7 @@ cell_fragment_shader_state(void *shader)
 static INLINE struct cell_vertex_shader_state *
 cell_vertex_shader_state(void *shader)
 {
-   return (struct pipe_shader_state *) shader;
+   return (struct cell_vertex_shader_state *) shader;
 }
 
 

src/gallium/drivers/cell/ppu/cell_surface.c

@@ -26,11 +26,12 @@
  **************************************************************************/
 
 #include "pipe/p_defines.h"
-#include "util/u_memory.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_winsys.h"
-#include "util/p_tile.h"
+#include "util/u_memory.h"
 #include "util/u_rect.h"
+#include "util/u_tile.h"
+
 #include "cell_context.h"
 #include "cell_surface.h"
 
@@ -46,12 +47,12 @@ cell_surface_copy(struct pipe_context *pipe,
 {
    assert( dst->cpp == src->cpp );
 
-   pipe_copy_rect(pipe_surface_map(dst),
+   pipe_copy_rect(pipe_surface_map(dst, PIPE_BUFFER_USAGE_CPU_WRITE),
                   &dst->block,
                   dst->stride,
                   dstx, dsty,
                   width, height,
-                  pipe_surface_map(src),
+                  pipe_surface_map(src, PIPE_BUFFER_USAGE_CPU_READ),
                   do_flip ? -src->stride : src->stride,
                   srcx, do_flip ? height - 1 - srcy : srcy);
 
@@ -81,7 +82,7 @@ cell_surface_fill(struct pipe_context *pipe,
                   unsigned width, unsigned height, unsigned value)
 {
    unsigned i, j;
-   void *dst_map = pipe_surface_map(dst);
+   void *dst_map = pipe_surface_map(dst, PIPE_BUFFER_USAGE_CPU_WRITE);
 
    assert(dst->stride > 0);
 

src/gallium/drivers/cell/ppu/cell_texture.c

@@ -33,8 +33,9 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
-#include "util/u_memory.h"
 #include "pipe/p_winsys.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
 
 #include "cell_context.h"
 #include "cell_state.h"
@@ -68,11 +69,13 @@ cell_texture_layout(struct cell_texture * spt)
       pt->nblocksx[level] = pf_get_nblocksx(&pt->block, width);  
       pt->nblocksy[level] = pf_get_nblocksy(&pt->block, height);  
 
+      spt->stride[level] = pt->nblocksx[level] * pt->block.size;
+
       spt->level_offset[level] = spt->buffer_size;
 
       spt->buffer_size += (pt->nblocksy[level] *
 			  ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) *
-			  pt->nblocksx[level] * pt->block.size;
+                           pt->nblocksx[level] * pt->block.size);
 
       width  = minify(width);
       height = minify(height);
@@ -147,7 +150,8 @@ cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture,
 static struct pipe_surface *
 cell_get_tex_surface_screen(struct pipe_screen *screen,
                             struct pipe_texture *pt,
-                            unsigned face, unsigned level, unsigned zslice)
+                            unsigned face, unsigned level, unsigned zslice,
+                            unsigned usage)
 {
    struct pipe_winsys *ws = screen->winsys;
    struct cell_texture *spt = cell_texture(pt);
@@ -166,6 +170,10 @@ cell_get_tex_surface_screen(struct pipe_screen *screen,
       ps->nblocksy = pt->nblocksy[level];
       ps->stride = spt->stride[level];
       ps->offset = spt->level_offset[level];
+      ps->usage = usage;
+
+      /* XXX may need to override usage flags (see sp_texture.c) */
+
 
       if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) {
 	 ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) *
@@ -228,10 +236,11 @@ cell_tile_texture(struct cell_context *cell,
    assert(w % TILE_SIZE == 0);
    assert(h % TILE_SIZE == 0);
 
-   surf = screen->get_tex_surface(screen, &texture->base, face, level, zslice);
+   surf = screen->get_tex_surface(screen, &texture->base, face, level, zslice,
+                                  PIPE_BUFFER_USAGE_CPU_WRITE);
    ASSERT(surf);
 
-   src = (const uint *) pipe_surface_map(surf);
+   src = (const uint *) pipe_surface_map(surf, PIPE_BUFFER_USAGE_CPU_WRITE);
 
    if (texture->tiled_data) {
       align_free(texture->tiled_data);
@@ -246,11 +255,12 @@ cell_tile_texture(struct cell_context *cell,
 }
 
 
-
 void
 cell_update_texture_mapping(struct cell_context *cell)
 {
+#if 0
    uint face = 0, level = 0, zslice = 0;
+#endif
    uint i;
 
    for (i = 0; i < CELL_MAX_SAMPLERS; i++) {
@@ -275,10 +285,52 @@ cell_update_texture_mapping(struct cell_context *cell)
 }
 
 
+static void *
+cell_surface_map( struct pipe_screen *screen,
+                  struct pipe_surface *surface,
+                  unsigned flags )
+{
+   ubyte *map;
+
+   if (flags & ~surface->usage) {
+      assert(0);
+      return NULL;
+   }
+
+   map = screen->winsys->buffer_map( screen->winsys, surface->buffer, flags );
+   if (map == NULL)
+      return NULL;
+
+   /* May want to different things here depending on read/write nature
+    * of the map:
+    */
+   if (surface->texture &&
+       (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) 
+   {
+      /* Do something to notify sharing contexts of a texture change.
+       * In softpipe, that would mean flushing the texture cache.
+       */
+#if 0
+      cell_screen(screen)->timestamp++;
+#endif
+   }
+   
+   return map + surface->offset;
+}
+
+
+static void
+cell_surface_unmap(struct pipe_screen *screen,
+                   struct pipe_surface *surface)
+{
+   screen->winsys->buffer_unmap( screen->winsys, surface->buffer );
+}
+
+
 void
 cell_init_texture_functions(struct cell_context *cell)
 {
-   cell->pipe.texture_update = cell_texture_update;
+   /*cell->pipe.texture_update = cell_texture_update;*/
 }
 
 void
@@ -287,4 +339,7 @@ cell_init_screen_texture_funcs(struct pipe_screen *screen)
    screen->texture_create = cell_texture_create_screen;
    screen->texture_release = cell_texture_release_screen;
    screen->get_tex_surface = cell_get_tex_surface_screen;
+
+   screen->surface_map = cell_surface_map;
+   screen->surface_unmap = cell_surface_unmap;
 }

src/gallium/drivers/cell/ppu/cell_texture.h

@@ -41,6 +41,7 @@ struct cell_texture
    struct pipe_texture base;
 
    unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS];
+   unsigned long stride[PIPE_MAX_TEXTURE_LEVELS];
 
    /* The data is held here:
     */

src/gallium/drivers/cell/ppu/cell_vbuf.c

@@ -37,6 +37,7 @@
 #include "cell_spu.h"
 #include "cell_vbuf.h"
 #include "draw/draw_vbuf.h"
+#include "util/u_memory.h"
 
 
 /** Allow vertex data to be inlined after RENDER command */

src/gallium/drivers/cell/ppu/cell_vertex_fetch.c

@@ -260,6 +260,7 @@ emit_fetch(struct spe_function *p,
 
 void cell_update_vertex_fetch(struct draw_context *draw)
 {
+#if 0
    struct cell_context *const cell =
        (struct cell_context *) draw->driver_private;
    struct spe_function *p = &cell->attrib_fetch;
@@ -337,4 +338,7 @@ void cell_update_vertex_fetch(struct draw_context *draw)
 	     cell->attrib_fetch_offsets[function_index[i]];
       }
    }
+#else
+   assert(0);
+#endif
 }

src/gallium/drivers/cell/ppu/cell_vertex_shader.c

@@ -32,6 +32,7 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_context.h"
 #include "pipe/p_winsys.h"
+#include "util/u_math.h"
 
 #include "cell_context.h"
 #include "cell_draw_arrays.h"
@@ -50,6 +51,7 @@
 void
 cell_vertex_shader_queue_flush(struct draw_context *draw)
 {
+#if 0
    struct cell_context *const cell =
        (struct cell_context *) draw->driver_private;
    struct cell_command_vs *const vs = &cell_global.command[0].vs;
@@ -138,4 +140,7 @@ cell_vertex_shader_queue_flush(struct draw_context *draw)
 
    draw->vs.post_nr = draw->vs.queue_nr;
    draw->vs.queue_nr = 0;
+#else
+   assert(0);
+#endif
 }

src/gallium/drivers/cell/spu/spu_exec.h

@@ -99,7 +99,7 @@ struct spu_exec_machine
     * 1  address
     */
    struct spu_exec_vector       Temps[TGSI_EXEC_NUM_TEMPS 
-				      + TGSI_EXEC_NUM_ADDRS + 1]
+                                      + TGSI_EXEC_NUM_TEMP_EXTRAS + 1]
        ALIGN16_ATTRIB;
 
    struct spu_exec_vector       *Addrs;

src/gallium/drivers/cell/spu/spu_render.c

@@ -35,7 +35,7 @@
 #include "spu_tri.h"
 #include "spu_tile.h"
 #include "cell/common.h"
-
+#include "util/u_memory.h"
 
 
 /**

src/gallium/drivers/cell/spu/spu_tri.c

@@ -32,6 +32,7 @@
 #include <transpose_matrix4x4.h>
 #include "pipe/p_compiler.h"
 #include "pipe/p_format.h"
+#include "util/u_math.h"
 #include "spu_colorpack.h"
 #include "spu_main.h"
 #include "spu_texture.h"

src/gallium/drivers/cell/spu/spu_util.c

@@ -1,4 +1,6 @@
+
 #include "pipe/p_shader_tokens.h"
+#include "pipe/p_debug.h"
 #include "tgsi/tgsi_parse.h"
 //#include "tgsi_build.h"
 #include "tgsi/tgsi_util.h"

src/gallium/drivers/cell/spu/spu_vertex_shader.c

@@ -36,13 +36,35 @@
 
 #include "pipe/p_state.h"
 #include "pipe/p_shader_tokens.h"
-#include "spu_vertex_shader.h"
-#include "spu_exec.h"
+#include "util/u_math.h"
 #include "draw/draw_private.h"
 #include "draw/draw_context.h"
 #include "cell/common.h"
+#include "spu_vertex_shader.h"
+#include "spu_exec.h"
 #include "spu_main.h"
 
+
+#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float))
+
+
+#define CLIP_RIGHT_BIT 0x01
+#define CLIP_LEFT_BIT 0x02
+#define CLIP_TOP_BIT 0x04
+#define CLIP_BOTTOM_BIT 0x08
+#define CLIP_FAR_BIT 0x10
+#define CLIP_NEAR_BIT 0x20
+
+
+static INLINE float
+dot4(const float *a, const float *b)
+{
+   return (a[0]*b[0] +
+           a[1]*b[1] +
+           a[2]*b[2] +
+           a[3]*b[3]);
+}
+
 static INLINE unsigned
 compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr)
 {