Merge branch 'master' of ssh://git.freedesktop.org/git/mesa/mesa into r600_state_predict
[mesa.git] / src / mesa / drivers / dri / i965 / brw_wm_surface_state.c
index 40d6c38f4fa91a6bcd99f03a79f54da8e410d228..096f74394eb6db3365eb5d939f7b60d1895886e6 100644 (file)
@@ -38,7 +38,7 @@
 #include "intel_mipmap_tree.h"
 #include "intel_batchbuffer.h"
 #include "intel_tex.h"
-
+#include "intel_fbo.h"
 
 #include "brw_context.h"
 #include "brw_state.h"
@@ -176,17 +176,6 @@ static GLuint translate_tex_format( GLuint mesa_format, GLenum internal_format,
    }
 }
 
-struct brw_wm_surface_key {
-   GLenum target, depthmode;
-   dri_bo *bo;
-   GLint format, internal_format;
-   GLint first_level, last_level;
-   GLint width, height, depth;
-   GLint pitch, cpp;
-   uint32_t tiling;
-   GLuint offset;
-};
-
 static void
 brw_set_surface_tiling(struct brw_surface_state *surf, uint32_t tiling)
 {
@@ -208,7 +197,7 @@ brw_set_surface_tiling(struct brw_surface_state *surf, uint32_t tiling)
 
 static dri_bo *
 brw_create_texture_surface( struct brw_context *brw,
-                           struct brw_wm_surface_key *key )
+                           struct brw_surface_key *key )
 {
    struct brw_surface_state surf;
    dri_bo *bo;
@@ -263,7 +252,7 @@ brw_create_texture_surface( struct brw_context *brw,
       surf.ss0.cube_neg_z = 1;
    }
 
-   bo = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+   bo = brw_upload_cache(&brw->surface_cache, BRW_SS_SURFACE,
                         key, sizeof(*key),
                         &key->bo, key->bo ? 1 : 0,
                         &surf, sizeof(surf),
@@ -287,7 +276,7 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
    struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current;
    struct intel_texture_object *intelObj = intel_texture_object(tObj);
    struct gl_texture_image *firstImage = tObj->Image[0][intelObj->firstLevel];
-   struct brw_wm_surface_key key;
+   struct brw_surface_key key;
    const GLuint surf = SURF_INDEX_TEXTURE(unit);
 
    memset(&key, 0, sizeof(key));
@@ -316,10 +305,11 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
    key.tiling = intelObj->mt->region->tiling;
 
    dri_bo_unreference(brw->wm.surf_bo[surf]);
-   brw->wm.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-                                         &key, sizeof(key),
-                                         &key.bo, key.bo ? 1 : 0,
-                                         NULL);
+   brw->wm.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
+                                            &key, sizeof(key),
+                                            &key.bo, key.bo ? 1 : 0,
+                                            NULL);
    if (brw->wm.surf_bo[surf] == NULL) {
       brw->wm.surf_bo[surf] = brw_create_texture_surface(brw, &key);
    }
@@ -328,12 +318,12 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
 
 
 /**
- * Create the constant buffer surface.  Fragment shader constanst will be
+ * Create the constant buffer surface.  Vertex/fragment shader constants will be
  * read from this buffer with Data Port Read instructions/messages.
  */
-static dri_bo *
+dri_bo *
 brw_create_constant_surface( struct brw_context *brw,
-                             struct brw_wm_surface_key *key )
+                             struct brw_surface_key *key )
 {
    const GLint w = key->width - 1;
    struct brw_surface_state surf;
@@ -345,8 +335,6 @@ brw_create_constant_surface( struct brw_context *brw,
    surf.ss0.surface_type = BRW_SURFACE_BUFFER;
    surf.ss0.surface_format = BRW_SURFACEFORMAT_R32G32B32A32_FLOAT;
 
-   /* This is ok for all textures with channel width 8bit or less:
-    */
    assert(key->bo);
    if (key->bo)
       surf.ss1.base_addr = key->bo->offset; /* reloc */
@@ -356,10 +344,10 @@ brw_create_constant_surface( struct brw_context *brw,
    surf.ss2.width = w & 0x7f;            /* bits 6:0 of size or width */
    surf.ss2.height = (w >> 7) & 0x1fff;  /* bits 19:7 of size or width */
    surf.ss3.depth = (w >> 20) & 0x7f;    /* bits 26:20 of size or width */
-   surf.ss3.pitch = (key->pitch * key->cpp) - 1;
-   brw_set_surface_tiling(&surf, key->tiling);
+   surf.ss3.pitch = (key->pitch * key->cpp) - 1; /* ignored?? */
+   brw_set_surface_tiling(&surf, key->tiling); /* tiling now allowed */
  
-   bo = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+   bo = brw_upload_cache(&brw->surface_cache, BRW_SS_SURFACE,
                         key, sizeof(*key),
                         &key->bo, key->bo ? 1 : 0,
                         &surf, sizeof(surf),
@@ -377,39 +365,70 @@ brw_create_constant_surface( struct brw_context *brw,
    return bo;
 }
 
+/* Creates a new WM constant buffer reflecting the current fragment program's
+ * constants, if needed by the fragment program.
+ *
+ * Otherwise, constants go through the CURBEs using the brw_constant_buffer
+ * state atom.
+ */
+static drm_intel_bo *
+brw_wm_update_constant_buffer(struct brw_context *brw)
+{
+   struct intel_context *intel = &brw->intel;
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   const struct gl_program_parameter_list *params = fp->program.Base.Parameters;
+   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   drm_intel_bo *const_buffer;
+
+   /* BRW_NEW_FRAGMENT_PROGRAM */
+   if (!fp->use_const_buffer)
+      return NULL;
+
+   const_buffer = drm_intel_bo_alloc(intel->bufmgr, "fp_const_buffer",
+                                    size, 64);
+
+   /* _NEW_PROGRAM_CONSTANTS */
+   dri_bo_subdata(const_buffer, 0, size, params->ParameterValues);
+
+   return const_buffer;
+}
 
 /**
- * Update the surface state for a constant buffer.
+ * Update the surface state for a WM constant buffer.
  * The constant buffer will be (re)allocated here if needed.
  */
-static dri_bo *
-brw_update_constant_surface( GLcontext *ctx,
-                             GLuint surf,
-                             dri_bo *const_buffer,
-                             const struct gl_program_parameter_list *params)
+static void
+brw_update_wm_constant_surface( GLcontext *ctx,
+                                GLuint surf)
 {
    struct brw_context *brw = brw_context(ctx);
-   struct brw_wm_surface_key key;
-   struct intel_context *intel = &brw->intel;
-   const int size = params->NumParameters * 4 * sizeof(GLfloat);
-
-   /* free old const buffer if too small */
-   if (const_buffer && const_buffer->size < size) {
-      dri_bo_unreference(const_buffer);
-      const_buffer = NULL;
-   }
+   struct brw_surface_key key;
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   const struct gl_program_parameter_list *params =
+      fp->program.Base.Parameters;
+
+   /* If we're in this state update atom, we need to update WM constants, so
+    * free the old buffer and create a new one for the new contents.
+    */
+   dri_bo_unreference(fp->const_buffer);
+   fp->const_buffer = brw_wm_update_constant_buffer(brw);
 
-   /* alloc new buffer if needed */
-   if (!const_buffer) {
-      const_buffer =
-         drm_intel_bo_alloc(intel->bufmgr, "vp/fp_const_buffer", size, 64);
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (fp->const_buffer == 0) {
+      drm_intel_bo_unreference(brw->wm.surf_bo[surf]);
+      brw->wm.surf_bo[surf] = NULL;
+      return;
    }
 
    memset(&key, 0, sizeof(key));
 
    key.format = MESA_FORMAT_RGBA_FLOAT32;
    key.internal_format = GL_RGBA;
-   key.bo = const_buffer;
+   key.bo = fp->const_buffer;
    key.depthmode = GL_NONE;
    key.pitch = params->NumParameters;
    key.width = params->NumParameters;
@@ -424,17 +443,59 @@ brw_update_constant_surface( GLcontext *ctx,
    */
 
    dri_bo_unreference(brw->wm.surf_bo[surf]);
-   brw->wm.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
+   brw->wm.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
                                             &key, sizeof(key),
                                             &key.bo, key.bo ? 1 : 0,
                                             NULL);
    if (brw->wm.surf_bo[surf] == NULL) {
       brw->wm.surf_bo[surf] = brw_create_constant_surface(brw, &key);
    }
+   brw->state.dirty.brw |= BRW_NEW_WM_SURFACES;
+}
 
-   return const_buffer;
+/**
+ * Updates surface / buffer for fragment shader constant buffer, if
+ * one is required.
+ *
+ * This consumes the state updates for the constant buffer, and produces
+ * BRW_NEW_WM_SURFACES to get picked up by brw_prepare_wm_surfaces for
+ * inclusion in the binding table.
+ */
+static void prepare_wm_constant_surface(struct brw_context *brw )
+{
+   GLcontext *ctx = &brw->intel.ctx;
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   GLuint surf = SURF_INDEX_FRAG_CONST_BUFFER;
+
+   drm_intel_bo_unreference(fp->const_buffer);
+   fp->const_buffer = brw_wm_update_constant_buffer(brw);
+
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (fp->const_buffer == 0) {
+      if (brw->wm.surf_bo[surf] != NULL) {
+        drm_intel_bo_unreference(brw->wm.surf_bo[surf]);
+        brw->wm.surf_bo[surf] = NULL;
+        brw->state.dirty.brw |= BRW_NEW_WM_SURFACES;
+      }
+      return;
+   }
+
+   brw_update_wm_constant_surface(ctx, surf);
 }
 
+const struct brw_tracked_state brw_wm_constant_surface = {
+   .dirty = {
+      .mesa = (_NEW_PROGRAM_CONSTANTS),
+      .brw = (BRW_NEW_FRAGMENT_PROGRAM),
+      .cache = 0
+   },
+   .prepare = prepare_wm_constant_surface,
+};
+
 
 /**
  * Sets up a surface state structure to point at the given region.
@@ -442,18 +503,22 @@ brw_update_constant_surface( GLcontext *ctx,
  * usable for further buffers when doing ARB_draw_buffer support.
  */
 static void
-brw_update_region_surface(struct brw_context *brw, struct intel_region *region,
-                         unsigned int unit, GLboolean cached)
+brw_update_renderbuffer_surface(struct brw_context *brw,
+                               struct gl_renderbuffer *rb,
+                               unsigned int unit)
 {
    GLcontext *ctx = &brw->intel.ctx;
    dri_bo *region_bo = NULL;
+   struct intel_renderbuffer *irb = intel_renderbuffer(rb);
+   struct intel_region *region = irb ? irb->region : NULL;
    struct {
       unsigned int surface_type;
       unsigned int surface_format;
-      unsigned int width, height, cpp;
+      unsigned int width, height, pitch, cpp;
       GLubyte color_mask[4];
       GLboolean color_blend;
       uint32_t tiling;
+      uint32_t draw_offset;
    } key;
 
    memset(&key, 0, sizeof(key));
@@ -462,14 +527,29 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region,
       region_bo = region->buffer;
 
       key.surface_type = BRW_SURFACE_2D;
-      if (region->cpp == 4)
+      switch (irb->texformat->MesaFormat) {
+      case MESA_FORMAT_ARGB8888:
         key.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM;
-      else
+        break;
+      case MESA_FORMAT_RGB565:
         key.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM;
+        break;
+      case MESA_FORMAT_ARGB1555:
+        key.surface_format = BRW_SURFACEFORMAT_B5G5R5A1_UNORM;
+        break;
+      case MESA_FORMAT_ARGB4444:
+        key.surface_format = BRW_SURFACEFORMAT_B4G4R4A4_UNORM;
+        break;
+      default:
+        _mesa_problem(ctx, "Bad renderbuffer format: %d\n",
+                      irb->texformat->MesaFormat);
+      }
       key.tiling = region->tiling;
-      key.width = region->pitch; /* XXX: not really! */
+      key.width = region->width;
       key.height = region->height;
+      key.pitch = region->pitch;
       key.cpp = region->cpp;
+      key.draw_offset = region->draw_offset; /* cur 3d or cube face offset */
    } else {
       key.surface_type = BRW_SURFACE_NULL;
       key.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM;
@@ -477,6 +557,7 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region,
       key.width = 1;
       key.height = 1;
       key.cpp = 4;
+      key.draw_offset = 0;
    }
    memcpy(key.color_mask, ctx->Color.ColorMask,
          sizeof(key.color_mask));
@@ -484,12 +565,11 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region,
                      ctx->Color.BlendEnabled);
 
    dri_bo_unreference(brw->wm.surf_bo[unit]);
-   brw->wm.surf_bo[unit] = NULL;
-   if (cached) 
-       brw->wm.surf_bo[unit] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-              &key, sizeof(key),
-              &region_bo, 1,
-              NULL);
+   brw->wm.surf_bo[unit] = brw_search_cache(&brw->surface_cache,
+                                           BRW_SS_SURFACE,
+                                           &key, sizeof(key),
+                                           &region_bo, 1,
+                                           NULL);
 
    if (brw->wm.surf_bo[unit] == NULL) {
       struct brw_surface_state surf;
@@ -498,13 +578,34 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region,
 
       surf.ss0.surface_format = key.surface_format;
       surf.ss0.surface_type = key.surface_type;
+      if (key.tiling == I915_TILING_NONE) {
+        surf.ss1.base_addr = key.draw_offset;
+      } else {
+        uint32_t tile_offset = key.draw_offset % 4096;
+
+        surf.ss1.base_addr = key.draw_offset - tile_offset;
+
+        assert(BRW_IS_G4X(brw) || tile_offset == 0);
+        if (BRW_IS_G4X(brw)) {
+           if (key.tiling == I915_TILING_X) {
+              /* Note that the low bits of these fields are missing, so
+               * there's the possibility of getting in trouble.
+               */
+              surf.ss5.x_offset = (tile_offset % 512) / key.cpp / 4;
+              surf.ss5.y_offset = tile_offset / 512 / 2;
+           } else {
+              surf.ss5.x_offset = (tile_offset % 128) / key.cpp / 4;
+              surf.ss5.y_offset = tile_offset / 128 / 2;
+           }
+        }
+      }
       if (region_bo != NULL)
-        surf.ss1.base_addr = region_bo->offset; /* reloc */
+        surf.ss1.base_addr += region_bo->offset; /* reloc */
 
       surf.ss2.width = key.width - 1;
       surf.ss2.height = key.height - 1;
       brw_set_surface_tiling(&surf, key.tiling);
-      surf.ss3.pitch = (key.width * key.cpp) - 1;
+      surf.ss3.pitch = (key.pitch * key.cpp) - 1;
 
       /* _NEW_COLOR */
       surf.ss0.color_blend = key.color_blend;
@@ -514,8 +615,9 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region,
       surf.ss0.writedisable_alpha = !key.color_mask[3];
 
       /* Key size will never match key size for textures, so we're safe. */
-      brw->wm.surf_bo[unit] = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
-                                             &key, sizeof(key),
+      brw->wm.surf_bo[unit] = brw_upload_cache(&brw->surface_cache,
+                                               BRW_SS_SURFACE,
+                                               &key, sizeof(key),
                                               &region_bo, 1,
                                               &surf, sizeof(surf),
                                               NULL, NULL);
@@ -524,12 +626,12 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region,
          * them both.  We might be able to figure out from other state
          * a more restrictive relocation to emit.
          */
-        dri_bo_emit_reloc(brw->wm.surf_bo[unit],
-                          I915_GEM_DOMAIN_RENDER,
-                          I915_GEM_DOMAIN_RENDER,
-                          0,
-                          offsetof(struct brw_surface_state, ss1),
-                          region_bo);
+        drm_intel_bo_emit_reloc(brw->wm.surf_bo[unit],
+                                offsetof(struct brw_surface_state, ss1),
+                                region_bo,
+                                surf.ss1.base_addr,
+                                I915_GEM_DOMAIN_RENDER,
+                                I915_GEM_DOMAIN_RENDER);
       }
    }
 }
@@ -544,7 +646,9 @@ brw_wm_get_binding_table(struct brw_context *brw)
 {
    dri_bo *bind_bo;
 
-   bind_bo = brw_search_cache(&brw->cache, BRW_SS_SURF_BIND,
+   assert(brw->wm.nr_surfaces <= BRW_WM_MAX_SURF);
+
+   bind_bo = brw_search_cache(&brw->surface_cache, BRW_SS_SURF_BIND,
                              NULL, 0,
                              brw->wm.surf_bo, brw->wm.nr_surfaces,
                              NULL);
@@ -560,7 +664,7 @@ brw_wm_get_binding_table(struct brw_context *brw)
          else
             data[i] = 0;
 
-      bind_bo = brw_upload_cache( &brw->cache, BRW_SS_SURF_BIND,
+      bind_bo = brw_upload_cache( &brw->surface_cache, BRW_SS_SURF_BIND,
                                  NULL, 0,
                                  brw->wm.surf_bo, brw->wm.nr_surfaces,
                                  data, data_size,
@@ -590,46 +694,23 @@ static void prepare_wm_surfaces(struct brw_context *brw )
    GLuint i;
    int old_nr_surfaces;
 
+   /* _NEW_BUFFERS */
    /* Update surfaces for drawing buffers */
-   if (brw->state.nr_color_regions > 1) {
-      for (i = 0; i < brw->state.nr_color_regions; i++) {
-         brw_update_region_surface(brw, brw->state.color_regions[i], i,
-                                  GL_FALSE);
+   if (ctx->DrawBuffer->_NumColorDrawBuffers >= 1) {
+      for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
+         brw_update_renderbuffer_surface(brw,
+                                        ctx->DrawBuffer->_ColorDrawBuffers[i],
+                                        i);
       }
    } else {
-      brw_update_region_surface(brw, brw->state.color_regions[0], 0, GL_TRUE);
+      brw_update_renderbuffer_surface(brw, NULL, 0);
    }
 
    old_nr_surfaces = brw->wm.nr_surfaces;
    brw->wm.nr_surfaces = MAX_DRAW_BUFFERS;
 
-   /* Update surface / buffer for vertex shader constant buffer */
-   {
-      const GLuint surf = SURF_INDEX_VERT_CONST_BUFFER + 1;
-      struct brw_vertex_program *vp =
-         (struct brw_vertex_program *) brw->vertex_program;
-      vp->const_buffer =
-         brw_update_constant_surface(ctx,
-                                     SURF_INDEX_VERT_CONST_BUFFER,
-                                     vp->const_buffer,
-                                     vp->program.Base.Parameters);
-
-      brw->wm.nr_surfaces = surf + 1;
-   }
-
-   /* Update surface / buffer for fragment shader constant buffer */
-   {
-      const GLuint surf = SURF_INDEX_FRAG_CONST_BUFFER + 1;
-      struct brw_fragment_program *fp =
-         (struct brw_fragment_program *) brw->fragment_program;
-      fp->const_buffer =
-         brw_update_constant_surface(ctx,
-                                     SURF_INDEX_FRAG_CONST_BUFFER,
-                                     fp->const_buffer,
-                                     fp->program.Base.Parameters);
-
-      brw->wm.nr_surfaces = surf + 1;
-   }
+   if (brw->wm.surf_bo[SURF_INDEX_FRAG_CONST_BUFFER] != NULL)
+       brw->wm.nr_surfaces = SURF_INDEX_FRAG_CONST_BUFFER + 1;
 
    /* Update surfaces for textures */
    for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
@@ -659,14 +740,16 @@ static void prepare_wm_surfaces(struct brw_context *brw )
    brw->wm.bind_bo = brw_wm_get_binding_table(brw);
 
    if (brw->wm.nr_surfaces != old_nr_surfaces)
-      brw->state.dirty.brw |= BRW_NEW_NR_SURFACES;
+      brw->state.dirty.brw |= BRW_NEW_NR_WM_SURFACES;
 }
 
-
 const struct brw_tracked_state brw_wm_surfaces = {
    .dirty = {
-      .mesa = _NEW_COLOR | _NEW_TEXTURE | _NEW_BUFFERS,
-      .brw = BRW_NEW_CONTEXT,
+      .mesa = (_NEW_COLOR |
+               _NEW_TEXTURE |
+               _NEW_BUFFERS),
+      .brw = (BRW_NEW_CONTEXT |
+             BRW_NEW_WM_SURFACES),
       .cache = 0
    },
    .prepare = prepare_wm_surfaces,