Merge branch 'gallium-edgeflags'

Conflicts:
	src/mesa/state_tracker/st_draw.c

7 files changed, 130 insertions, 112 deletions

diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index 09baff875bc..46c8cbb3098 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -106,7 +106,18 @@ find_translated_vp(struct st_context *st,
    /* Nothing in our key yet.  This will change:
     */
    memset(&key, 0, sizeof key);
-   key.dummy = 0;
+
+   /* When this is true, we will add an extra input to the vertex
+    * shader translation (for edgeflags), an extra output with
+    * edgeflag semantics, and extend the vertex shader to pass through
+    * the input to the output.  We'll need to use similar logic to set
+    * up the extra vertex_element input for edgeflags.
+    * _NEW_POLYGON, ST_NEW_EDGEFLAGS_DATA
+    */
+   key.passthrough_edgeflags = (st->vertdata_edgeflags && (
+                                st->ctx->Polygon.FrontMode != GL_FILL ||
+                                st->ctx->Polygon.BackMode != GL_FILL));
+
 
    /* Do we need to throw away old translations after a change in the
     * GL program string?
@@ -218,8 +229,8 @@ update_vp( struct st_context *st )
 const struct st_tracked_state st_update_vp = {
    "st_update_vp",					/* name */
    {							/* dirty */
-      0,						/* mesa */
-      ST_NEW_VERTEX_PROGRAM                             /* st */
+      _NEW_POLYGON,					/* mesa */
+      ST_NEW_VERTEX_PROGRAM | ST_NEW_EDGEFLAGS_DATA	/* st */
    },
    update_vp					/* update */
 };
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
index b7607286583..831909a3f8f 100644
--- a/src/mesa/state_tracker/st_context.h
+++ b/src/mesa/state_tracker/st_context.h
@@ -55,6 +55,7 @@ struct bitmap_cache;
 #define ST_NEW_FRAGMENT_PROGRAM        0x2
 #define ST_NEW_VERTEX_PROGRAM          0x4
 #define ST_NEW_FRAMEBUFFER             0x8
+#define ST_NEW_EDGEFLAGS_DATA          0x10
 
 
 struct st_state_flags {
@@ -120,6 +121,7 @@ struct st_context
    struct st_state_flags dirty;
 
    GLboolean missing_textures;
+   GLboolean vertdata_edgeflags;
 
    /** Mapping from VERT_RESULT_x to post-transformed vertex slot */
    const GLuint *vertex_result_to_slot;
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
index ec8092d7e9e..e54f21be600 100644
--- a/src/mesa/state_tracker/st_draw.c
+++ b/src/mesa/state_tracker/st_draw.c
@@ -217,59 +217,7 @@ st_pipe_vertex_format(GLenum type, GLuint size, GLenum format,
 }
 
 
-/*
- * If edge flags are needed, setup an bitvector of flags and call
- * pipe->set_edgeflags().
- * XXX memleak: need to free the returned pointer at some point
- */
-static void *
-setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count,
-                const struct gl_client_array *array)
-{
-   struct pipe_context *pipe = ctx->st->pipe;
-
-   if ((primMode == GL_TRIANGLES ||
-        primMode == GL_QUADS ||
-        primMode == GL_POLYGON) &&
-       (ctx->Polygon.FrontMode != GL_FILL ||
-        ctx->Polygon.BackMode != GL_FILL)) {
-      /* need edge flags */
-      GLint i;
-      unsigned *vec;
-      struct st_buffer_object *stobj = st_buffer_object(array->BufferObj);
-      ubyte *map;
-
-      if (!stobj || stobj->Base.Name == 0) {
-         /* edge flags are not in a VBO */
-         return NULL;
-      }
-
-      vec = (unsigned *) _mesa_calloc(sizeof(unsigned) * ((count + 31) / 32));
-      if (!vec)
-         return NULL;
-
-      map = pipe_buffer_map(pipe->screen, stobj->buffer, PIPE_BUFFER_USAGE_CPU_READ);
-      map = ADD_POINTERS(map, array->Ptr);
-
-      for (i = 0; i < count; i++) {
-         if (*((float *) map))
-            vec[i/32] |= 1 << (i % 32);
-
-         map += array->StrideB;
-      }
 
-      pipe_buffer_unmap(pipe->screen, stobj->buffer);
-
-      pipe->set_edgeflags(pipe, vec);
-
-      return vec;
-   }
-   else {
-      /* edge flags not needed */
-      pipe->set_edgeflags(pipe, NULL);
-      return NULL;
-   }
-}
 
 
 /**
@@ -279,6 +227,7 @@ setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count,
  */
 static GLboolean
 is_interleaved_arrays(const struct st_vertex_program *vp,
+                      const struct st_vp_varient *vpv,
                       const struct gl_client_array **arrays,
                       GLboolean *userSpace)
 {
@@ -288,7 +237,7 @@ is_interleaved_arrays(const struct st_vertex_program *vp,
    GLuint num_client_arrays = 0;
    const GLubyte *client_addr = NULL;
 
-   for (attr = 0; attr < vp->num_inputs; attr++) {
+   for (attr = 0; attr < vpv->num_inputs; attr++) {
       const GLuint mesaAttr = vp->index_to_input[attr];
       const struct gl_buffer_object *bufObj = arrays[mesaAttr]->BufferObj;
       const GLsizei stride = arrays[mesaAttr]->StrideB; /* in bytes */
@@ -321,7 +270,7 @@ is_interleaved_arrays(const struct st_vertex_program *vp,
       }
    }
 
-   *userSpace = (num_client_arrays == vp->num_inputs);
+   *userSpace = (num_client_arrays == vpv->num_inputs);
    /* printf("user space: %d (%d %d)\n", (int) *userSpace,num_client_arrays,vp->num_inputs); */
 
    return GL_TRUE;
@@ -333,15 +282,16 @@ is_interleaved_arrays(const struct st_vertex_program *vp,
  */
 static void
 get_arrays_bounds(const struct st_vertex_program *vp,
-                       const struct gl_client_array **arrays,
-                       GLuint max_index,
-                       const GLubyte **low, const GLubyte **high)
+                  const struct st_vp_varient *vpv,
+                  const struct gl_client_array **arrays,
+                  GLuint max_index,
+                  const GLubyte **low, const GLubyte **high)
 {
    const GLubyte *low_addr = NULL;
    const GLubyte *high_addr = NULL;
    GLuint attr;
 
-   for (attr = 0; attr < vp->num_inputs; attr++) {
+   for (attr = 0; attr < vpv->num_inputs; attr++) {
       const GLuint mesaAttr = vp->index_to_input[attr];
       const GLint stride = arrays[mesaAttr]->StrideB;
       const GLubyte *start = arrays[mesaAttr]->Ptr;
@@ -373,6 +323,7 @@ get_arrays_bounds(const struct st_vertex_program *vp,
 static void
 setup_interleaved_attribs(GLcontext *ctx,
                           const struct st_vertex_program *vp,
+                          const struct st_vp_varient *vpv,
                           const struct gl_client_array **arrays,
                           GLuint max_index,
                           GLboolean userSpace,
@@ -383,7 +334,7 @@ setup_interleaved_attribs(GLcontext *ctx,
    GLuint attr;
    const GLubyte *offset0 = NULL;
 
-   for (attr = 0; attr < vp->num_inputs; attr++) {
+   for (attr = 0; attr < vpv->num_inputs; attr++) {
       const GLuint mesaAttr = vp->index_to_input[attr];
       struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
       struct st_buffer_object *stobj = st_buffer_object(bufobj);
@@ -394,7 +345,7 @@ setup_interleaved_attribs(GLcontext *ctx,
       if (attr == 0) {
          const GLubyte *low, *high;
 
-         get_arrays_bounds(vp, arrays, max_index, &low, &high);
+         get_arrays_bounds(vp, vpv, arrays, max_index, &low, &high);
          /*printf("buffer range: %p %p  %d\n", low, high, high-low);*/
 
          offset0 = low;
@@ -435,6 +386,7 @@ setup_interleaved_attribs(GLcontext *ctx,
 static void
 setup_non_interleaved_attribs(GLcontext *ctx,
                               const struct st_vertex_program *vp,
+                              const struct st_vp_varient *vpv,
                               const struct gl_client_array **arrays,
                               GLuint max_index,
                               GLboolean *userSpace,
@@ -444,7 +396,7 @@ setup_non_interleaved_attribs(GLcontext *ctx,
    struct pipe_context *pipe = ctx->st->pipe;
    GLuint attr;
 
-   for (attr = 0; attr < vp->num_inputs; attr++) {
+   for (attr = 0; attr < vpv->num_inputs; attr++) {
       const GLuint mesaAttr = vp->index_to_input[attr];
       struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
       GLsizei stride = arrays[mesaAttr]->StrideB;
@@ -569,12 +521,14 @@ st_draw_vbo(GLcontext *ctx,
 {
    struct pipe_context *pipe = ctx->st->pipe;
    const struct st_vertex_program *vp;
+   const struct st_vp_varient *vpv;
    const struct pipe_shader_state *vs;
    struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS];
    GLuint attr;
    struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS];
    unsigned num_vbuffers, num_velements;
    GLboolean userSpace = GL_FALSE;
+   GLboolean vertDataEdgeFlags;
 
    /* Gallium probably doesn't want this in some cases. */
    if (!index_bounds_valid)
@@ -584,11 +538,19 @@ st_draw_vbo(GLcontext *ctx,
    /* sanity check for pointer arithmetic below */
    assert(sizeof(arrays[0]->Ptr[0]) == 1);
 
+   vertDataEdgeFlags = arrays[VERT_ATTRIB_EDGEFLAG]->BufferObj &&
+                       arrays[VERT_ATTRIB_EDGEFLAG]->BufferObj->Name;
+   if (vertDataEdgeFlags != ctx->st->vertdata_edgeflags) {
+      ctx->st->vertdata_edgeflags = vertDataEdgeFlags;
+      ctx->st->dirty.st |= ST_NEW_EDGEFLAGS_DATA;
+   }
+
    st_validate_state(ctx->st);
 
    /* must get these after state validation! */
    vp = ctx->st->vp;
-   vs = &ctx->st->vp_varient->state;
+   vpv = ctx->st->vp_varient;
+   vs = &vpv->state;
 
 #if 0
    if (MESA_VERBOSE & VERBOSE_GLSL) {
@@ -601,21 +563,21 @@ st_draw_vbo(GLcontext *ctx,
    /*
     * Setup the vbuffer[] and velements[] arrays.
     */
-   if (is_interleaved_arrays(vp, arrays, &userSpace)) {
+   if (is_interleaved_arrays(vp, vpv, arrays, &userSpace)) {
       /*printf("Draw interleaved\n");*/
-      setup_interleaved_attribs(ctx, vp, arrays, max_index, userSpace,
+      setup_interleaved_attribs(ctx, vp, vpv, arrays, max_index, userSpace,
                                 vbuffer, velements);
       num_vbuffers = 1;
-      num_velements = vp->num_inputs;
+      num_velements = vpv->num_inputs;
       if (num_velements == 0)
          num_vbuffers = 0;
    }
    else {
       /*printf("Draw non-interleaved\n");*/
-      setup_non_interleaved_attribs(ctx, vp, arrays, max_index,
+      setup_non_interleaved_attribs(ctx, vp, vpv, arrays, max_index,
                                     &userSpace, vbuffer, velements);
-      num_vbuffers = vp->num_inputs;
-      num_velements = vp->num_inputs;
+      num_vbuffers = vpv->num_inputs;
+      num_velements = vpv->num_inputs;
    }
 
 #if 0
@@ -687,10 +649,6 @@ st_draw_vbo(GLcontext *ctx,
           * through to driver & draw module.  These interfaces still
           * need a bit of work...
           */
-         setup_edgeflags(ctx, prims[i].mode,
-                         prims[i].start + indexOffset, prims[i].count,
-                         arrays[VERT_ATTRIB_EDGEFLAG]);
-
          prim = translate_prim( ctx, prims[i].mode );
 
          pipe->draw_range_elements(pipe, indexBuf, indexSize,
@@ -701,10 +659,6 @@ st_draw_vbo(GLcontext *ctx,
       }
       else {
          for (i = 0; i < nr_prims; i++) {
-            setup_edgeflags(ctx, prims[i].mode,
-                            prims[i].start + indexOffset, prims[i].count,
-                            arrays[VERT_ATTRIB_EDGEFLAG]);
-
             prim = translate_prim( ctx, prims[i].mode );
             
             pipe->draw_elements(pipe, indexBuf, indexSize,
@@ -721,10 +675,6 @@ st_draw_vbo(GLcontext *ctx,
       GLuint prim;
 
       for (i = 0; i < nr_prims; i++) {
-         setup_edgeflags(ctx, prims[i].mode,
-                         prims[i].start, prims[i].count,
-                         arrays[VERT_ATTRIB_EDGEFLAG]);
-
          prim = translate_prim( ctx, prims[i].mode );
 
          pipe->draw_arrays(pipe, prim, prims[i].start, prims[i].count);
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
index 1611d53e2fe..5c9be46a776 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.c
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -718,6 +718,16 @@ emit_face_var( struct st_translate *t,
    t->inputs[t->inputMapping[FRAG_ATTRIB_FACE]] = ureg_src(face_temp);
 }
 
+static void
+emit_edgeflags( struct st_translate *t,
+                 const struct gl_program *program )
+{
+   struct ureg_program *ureg = t->ureg;
+   struct ureg_dst edge_dst = t->outputs[t->outputMapping[VERT_RESULT_EDGE]];
+   struct ureg_src edge_src = t->inputs[t->inputMapping[VERT_ATTRIB_EDGEFLAG]];
+
+   ureg_MOV( ureg, edge_dst, edge_src );
+}
 
 /**
  * Translate Mesa program to TGSI format.
@@ -738,10 +748,11 @@ emit_face_var( struct st_translate *t,
  *
  * \return  array of translated tokens, caller's responsibility to free
  */
-const struct tgsi_token *
+enum pipe_error
 st_translate_mesa_program(
    GLcontext *ctx,
    uint procType,
+   struct ureg_program *ureg,
    const struct gl_program *program,
    GLuint numInputs,
    const GLuint inputMapping[],
@@ -751,11 +762,10 @@ st_translate_mesa_program(
    GLuint numOutputs,
    const GLuint outputMapping[],
    const ubyte outputSemanticName[],
-   const ubyte outputSemanticIndex[] )
+   const ubyte outputSemanticIndex[],
+   boolean passthrough_edgeflags )
 {
    struct st_translate translate, *t;
-   struct ureg_program *ureg;
-   const struct tgsi_token *tokens = NULL;
    unsigned i;
 
    t = &translate;
@@ -764,11 +774,7 @@ st_translate_mesa_program(
    t->procType = procType;
    t->inputMapping = inputMapping;
    t->outputMapping = outputMapping;
-   t->ureg = ureg_create( procType );
-   if (t->ureg == NULL)
-      return NULL;
-
-   ureg = t->ureg;
+   t->ureg = ureg;
 
    /*_mesa_print_program(program);*/
 
@@ -828,6 +834,8 @@ st_translate_mesa_program(
                                            outputSemanticName[i],
                                            outputSemanticIndex[i] );
       }
+      if (passthrough_edgeflags)
+         emit_edgeflags( t, program );
    }
 
    /* Declare address register.
@@ -899,8 +907,7 @@ st_translate_mesa_program(
                         t->insn[t->labels[i].branch_target] );
    }
 
-   tokens = ureg_get_tokens( ureg, NULL );
-   ureg_destroy( ureg );
+   return PIPE_OK;
 
 out:
    FREE(t->insn);
@@ -909,17 +916,9 @@ out:
 
    if (t->error) {
       debug_printf("%s: translate error flag set\n", __FUNCTION__);
-      FREE((void *)tokens);
-      tokens = NULL;
-   }
-
-   if (!tokens) {
-      debug_printf("%s: failed to translate Mesa program:\n", __FUNCTION__);
-      _mesa_print_program(program);
-      debug_assert(0);
    }
 
-   return tokens;
+   return PIPE_ERROR_OUT_OF_MEMORY;
 }
 
 
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h
index dc0362fe790..e3c5bd1d94d 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.h
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h
@@ -30,6 +30,7 @@
 #define ST_MESA_TO_TGSI_H
 
 #include "main/mtypes.h"
+#include "tgsi/tgsi_ureg.h"
 
 
 #if defined __cplusplus
@@ -39,10 +40,11 @@ extern "C" {
 struct tgsi_token;
 struct gl_program;
 
-const struct tgsi_token *
+enum pipe_error
 st_translate_mesa_program(
    GLcontext *ctx,
    uint procType,
+   struct ureg_program *ureg,
    const struct gl_program *program,
    GLuint numInputs,
    const GLuint inputMapping[],
@@ -52,7 +54,8 @@ st_translate_mesa_program(
    GLuint numOutputs,
    const GLuint outputMapping[],
    const ubyte outputSemanticName[],
-   const ubyte outputSemanticIndex[] );
+   const ubyte outputSemanticIndex[],
+   boolean passthrough_edgeflags );
 
 void
 st_free_tokens(const struct tgsi_token *tokens);
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index d91ed6806dc..6a869fae904 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -115,6 +115,9 @@ st_prepare_vertex_program(struct st_context *st,
          stvp->num_inputs++;
       }
    }
+   /* bit of a hack, presetup potentially unused edgeflag input */
+   stvp->input_to_index[VERT_ATTRIB_EDGEFLAG] = stvp->num_inputs;
+   stvp->index_to_input[stvp->num_inputs] = VERT_ATTRIB_EDGEFLAG;
 
    /* Compute mapping of vertex program outputs to slots.
     */
@@ -184,6 +187,10 @@ st_prepare_vertex_program(struct st_context *st,
          }
       }
    }
+   /* similar hack to above, presetup potentially unused edgeflag output */
+   stvp->result_to_output[VERT_RESULT_EDGE] = stvp->num_outputs;
+   stvp->output_semantic_name[stvp->num_outputs] = TGSI_SEMANTIC_EDGEFLAG;
+   stvp->output_semantic_index[stvp->num_outputs] = 0;
 }
 
 
@@ -194,22 +201,47 @@ st_translate_vertex_program(struct st_context *st,
 {
    struct st_vp_varient *vpv = CALLOC_STRUCT(st_vp_varient);
    struct pipe_context *pipe = st->pipe;
+   struct ureg_program *ureg;
+   enum pipe_error error;
+   unsigned num_outputs;
+
+   ureg = ureg_create( TGSI_PROCESSOR_VERTEX );
+   if (ureg == NULL)
+      return NULL;
+
+   vpv->num_inputs = stvp->num_inputs;
+   num_outputs = stvp->num_outputs;
+   if (key->passthrough_edgeflags) {
+      vpv->num_inputs++;
+      num_outputs++;
+   }
 
-   vpv->state.tokens = 
+   error = 
       st_translate_mesa_program(st->ctx,
                                 TGSI_PROCESSOR_VERTEX,
+                                ureg,
                                 &stvp->Base.Base,
                                 /* inputs */
-                                stvp->num_inputs,
+                                vpv->num_inputs,
                                 stvp->input_to_index,
                                 NULL, /* input semantic name */
                                 NULL, /* input semantic index */
                                 NULL,
                                 /* outputs */
-                                stvp->num_outputs,
+                                num_outputs,
                                 stvp->result_to_output,
                                 stvp->output_semantic_name,
-                                stvp->output_semantic_index );
+                                stvp->output_semantic_index,
+                                key->passthrough_edgeflags );
+
+   if (error)
+      goto fail;
+
+   vpv->state.tokens = ureg_get_tokens( ureg, NULL );
+   if (!vpv->state.tokens)
+      goto fail;
+
+   ureg_destroy( ureg );
 
    vpv->driver_shader = pipe->create_vs_state(pipe, &vpv->state);
 
@@ -224,6 +256,14 @@ st_translate_vertex_program(struct st_context *st,
    }
 
    return vpv;
+
+fail:
+   debug_printf("%s: failed to translate Mesa program:\n", __FUNCTION__);
+   _mesa_print_program(&stvp->Base.Base);
+   debug_assert(0);
+
+   ureg_destroy( ureg );
+   return NULL;
 }
 
 
@@ -244,7 +284,9 @@ st_translate_fragment_program(struct st_context *st,
    GLuint defaultInputMapping[FRAG_ATTRIB_MAX];
    GLuint interpMode[16];  /* XXX size? */
    GLuint attr;
+   enum pipe_error error;
    const GLbitfield inputsRead = stfp->Base.Base.InputsRead;
+   struct ureg_program *ureg;
    GLuint vslot = 0;
 
    uint fs_num_inputs = 0;
@@ -379,9 +421,15 @@ st_translate_fragment_program(struct st_context *st,
    if (!inputMapping)
       inputMapping = defaultInputMapping;
 
-   stfp->state.tokens = 
+   ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
+   if (ureg == NULL)
+      return;
+
+
+   error = 
       st_translate_mesa_program(st->ctx,
                                 TGSI_PROCESSOR_FRAGMENT,
+                                ureg,
                                 &stfp->Base.Base,
                                 /* inputs */
                                 fs_num_inputs,
@@ -393,8 +441,10 @@ st_translate_fragment_program(struct st_context *st,
                                 fs_num_outputs,
                                 outputMapping,
                                 fs_output_semantic_name,
-                                fs_output_semantic_index );
+                                fs_output_semantic_index, FALSE );
 
+   stfp->state.tokens = ureg_get_tokens( ureg, NULL );
+   ureg_destroy( ureg );
    stfp->driver_shader = pipe->create_fs_state(pipe, &stfp->state);
 
    if ((ST_DEBUG & DEBUG_TGSI) && (ST_DEBUG & DEBUG_MESA)) {
diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h
index 88aadbd7510..6b9a9226df5 100644
--- a/src/mesa/state_tracker/st_program.h
+++ b/src/mesa/state_tracker/st_program.h
@@ -73,7 +73,7 @@ struct st_fragment_program
 
 struct st_vp_varient_key
 {
-   char dummy;                  /* currently unused */
+   boolean passthrough_edgeflags;
 };
 
 
@@ -100,6 +100,9 @@ struct st_vp_varient
 
    /** Next in linked list */
    struct st_vp_varient *next;  
+
+   /** similar to that in st_vertex_program, but with information about edgeflags too */
+   GLuint num_inputs;
 };