* Keith Whitwell <keithw@tungstengraphics.com>
*/
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+#include "t_context.h"
+#include "t_vertex.h"
+#define DBG 0
+/* Build and manage clipspace/ndc/window vertices.
+ */
-
-struct attr {
- int attrib;
- int vertoffset;
- int vertattrsize;
- int *inputptr;
- int inputstride;
- void (*insert)( const struct attr *a, char *v, const GLfloat *input );
- void (*extract)( const struct attr *a, GLfloat *output, const char *v );
- const GLfloat *vp;
-};
-
-
-static void insert_4f_viewport( const struct attr *a, char *v,
- const GLfloat *in )
-{
- GLfloat *out = (GLfloat *)v;
- const GLfloat * const vp = a->vp;
-
- out[0] = vp[0] * in[0] + vp[12];
- out[1] = vp[5] * in[1] + vp[13];
- out[2] = vp[10] * in[2] + vp[14];
- out[3] = in[3];
-}
-
-static void insert_3f_viewport( const struct attr *a, char *v,
- const GLfloat *in )
+static GLboolean match_fastpath( struct tnl_clipspace *vtx,
+ const struct tnl_clipspace_fastpath *fp)
{
- GLfloat *out = (GLfloat *)v;
- const GLfloat * const vp = a->vp;
-
- out[0] = vp[0] * in[0] + vp[12];
- out[1] = vp[5] * in[1] + vp[13];
- out[2] = vp[10] * in[2] + vp[14];
-}
+ GLuint j;
-static void insert_2f_viewport( const struct attr *a, char *v,
- const GLfloat *in )
-{
- GLfloat *out = (GLfloat *)v;
- const GLfloat * const vp = a->vp;
-
- out[0] = vp[0] * in[0] + vp[12];
- out[1] = vp[5] * in[1] + vp[13];
-}
+ if (vtx->attr_count != fp->attr_count)
+ return GL_FALSE;
+ for (j = 0; j < vtx->attr_count; j++)
+ if (vtx->attr[j].format != fp->attr[j].format ||
+ vtx->attr[j].inputsize != fp->attr[j].size ||
+ vtx->attr[j].vertoffset != fp->attr[j].offset)
+ return GL_FALSE;
+
+ if (fp->match_strides) {
+ if (vtx->vertex_size != fp->vertex_size)
+ return GL_FALSE;
-static void insert_4f( const struct attr *a, char *v, const GLfloat *in )
-{
- GLfloat *out = (GLfloat *)(v);
-
- out[0] = in[0];
- out[1] = in[1];
- out[2] = in[2];
- out[3] = in[3];
-}
-
-static void insert_3f_xyw( const struct attr *a, char *v, const GLfloat *in )
-{
- GLfloat *out = (GLfloat *)(v);
+ for (j = 0; j < vtx->attr_count; j++)
+ if (vtx->attr[j].inputstride != fp->attr[j].stride)
+ return GL_FALSE;
+ }
- out[0] = in[0];
- out[1] = in[1];
- out[2] = in[3];
+ return GL_TRUE;
}
-
-static void insert_3f( const struct attr *a, char *v, const GLfloat *in )
+static GLboolean search_fastpath_emit( struct tnl_clipspace *vtx )
{
- GLfloat *out = (GLfloat *)(v);
-
- out[0] = in[0];
- out[1] = in[1];
- out[2] = in[2];
-}
-
+ struct tnl_clipspace_fastpath *fp = vtx->fastpath;
-static void insert_2f( const struct attr *a, char *v, const GLfloat *in )
-{
- GLfloat *out = (GLfloat *)(v);
-
- out[0] = in[0];
- out[1] = in[1];
-}
+ for ( ; fp ; fp = fp->next) {
+ if (match_fastpath(vtx, fp)) {
+ vtx->emit = fp->func;
+ return GL_TRUE;
+ }
+ }
-static void insert_1f( const struct attr *a, char *v, const GLfloat *in )
-{
- GLfloat *out = (GLfloat *)(v);
-
- out[0] = in[0];
+ return GL_FALSE;
}
-static void insert_4ub_4f_rgba( const struct attr *a, char *v,
- const GLfloat *in )
+void _tnl_register_fastpath( struct tnl_clipspace *vtx,
+ GLboolean match_strides )
{
- UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]);
-}
+ struct tnl_clipspace_fastpath *fastpath = CALLOC_STRUCT(tnl_clipspace_fastpath);
+ GLuint i;
-static void insert_4ub_4f_bgra( const struct attr *a, char *v,
- const GLfloat *in )
-{
- UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]);
-}
+ fastpath->vertex_size = vtx->vertex_size;
+ fastpath->attr_count = vtx->attr_count;
+ fastpath->match_strides = match_strides;
+ fastpath->func = vtx->emit;
+ fastpath->attr = (struct tnl_attr_type *)
+ malloc(vtx->attr_count * sizeof(fastpath->attr[0]));
-static void insert_3ub_3f_rgb( const struct attr *a, char *v,
- const GLfloat *in )
-{
- UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
-}
+ for (i = 0; i < vtx->attr_count; i++) {
+ fastpath->attr[i].format = vtx->attr[i].format;
+ fastpath->attr[i].stride = vtx->attr[i].inputstride;
+ fastpath->attr[i].size = vtx->attr[i].inputsize;
+ fastpath->attr[i].offset = vtx->attr[i].vertoffset;
+ }
-static void insert_3ub_3f_bgr( const struct attr *a, char *v,
- const GLfloat *in )
-{
- UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
+ fastpath->next = vtx->fastpath;
+ vtx->fastpath = fastpath;
}
-static void insert_1ub_1f( const struct attr *a, char *v,
- const GLfloat *in )
-{
- UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
-}
/***********************************************************************
- * Functions to perform the reverse operations to the above, for
- * swrast translation and clip-interpolation.
- *
- * Currently always extracts a full 4 floats.
+ * Build codegen functions or return generic ones:
*/
-
-static void extract_4f_viewport( const struct attr *a, GLfloat *out,
- const char *v )
+static void choose_emit_func( GLcontext *ctx, GLuint count, GLubyte *dest)
{
- const GLfloat *in = (const GLfloat *)v;
- const GLfloat * const vp = a->vp;
-
- out[0] = (in[0] - vp[12]) / vp[0];
- out[1] = (in[1] - vp[13]) / vp[5];
- out[2] = (in[2] - vp[14]) / vp[10];
- out[3] = in[3];
-}
+ struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ struct tnl_clipspace_attr *a = vtx->attr;
+ const GLuint attr_count = vtx->attr_count;
+ GLuint j;
-static void extract_3f_viewport( const struct attr *a, GLfloat *out,
- const char *v )
-{
- const GLfloat *in = (const GLfloat *)v;
- const GLfloat * const vp = a->vp;
+ for (j = 0; j < attr_count; j++) {
+ GLvector4f *vptr = VB->AttribPtr[a[j].attrib];
+ a[j].inputstride = vptr->stride;
+ a[j].inputsize = vptr->size;
+ a[j].emit = a[j].insert[vptr->size - 1]; /* not always used */
+ }
+
+ vtx->emit = NULL;
- out[0] = (in[0] - vp[12]) / vp[0];
- out[1] = (in[1] - vp[13]) / vp[5];
- out[2] = (in[2] - vp[14]) / vp[10];
- out[3] = 1;
-}
+ /* Does this match an existing (hardwired, codegen or known-bad)
+ * fastpath?
+ */
+ if (search_fastpath_emit(vtx)) {
+ /* Use this result. If it is null, then it is already known
+ * that the current state will fail for codegen and there is no
+ * point trying again.
+ */
+ }
+ else if (vtx->codegen_emit) {
+ vtx->codegen_emit(ctx);
+ }
+ if (!vtx->emit) {
+ _tnl_generate_hardwired_emit(ctx);
+ }
-static void extract_2f_viewport( const struct attr *a, GLfloat *out,
- const char *v )
-{
- const GLfloat *in = (const GLfloat *)v;
- const GLfloat * const vp = a->vp;
-
- out[0] = (in[0] - vp[12]) / vp[0];
- out[1] = (in[1] - vp[13]) / vp[5];
- out[2] = 0;
- out[3] = 1;
+ /* Otherwise use the generic version:
+ */
+ if (!vtx->emit)
+ vtx->emit = _tnl_generic_emit;
+
+ vtx->emit( ctx, count, dest );
}
-static void extract_4f( const struct attr *a, GLfloat *out, const char *v )
-{
- const GLfloat *in = (const GLfloat *)v;
-
- out[0] = in[0];
- out[1] = in[1];
- out[2] = in[2];
- out[3] = in[3];
-}
-static void extract_3f_xyw( const struct attr *a, GLfloat *out, const char *v )
+static void choose_interp_func( GLcontext *ctx,
+ GLfloat t,
+ GLuint edst, GLuint eout, GLuint ein,
+ GLboolean force_boundary )
{
- const GLfloat *in = (const GLfloat *)v;
-
- out[0] = in[0];
- out[1] = in[1];
- out[2] = in[3];
- out[3] = 1;
-}
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ if (vtx->need_extras &&
+ (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) {
+ vtx->interp = _tnl_generic_interp_extras;
+ } else {
+ vtx->interp = _tnl_generic_interp;
+ }
-static void extract_3f( const struct attr *a, GLfloat *out, const char *v )
-{
- const GLfloat *in = (const GLfloat *)v;
-
- out[0] = in[0];
- out[1] = in[1];
- out[2] = in[2];
- out[3] = 1;
+ vtx->interp( ctx, t, edst, eout, ein, force_boundary );
}
-static void extract_2f( const struct attr *a, GLfloat *out, const char *v )
+static void choose_copy_pv_func( GLcontext *ctx, GLuint edst, GLuint esrc )
{
- const GLfloat *in = (const GLfloat *)v;
-
- out[0] = in[0];
- out[1] = in[1];
- out[2] = 0;
- out[3] = 1;
-}
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
-static void extract_1f( const struct attr *a, GLfloat *out, const char *v )
-{
- const GLfloat *in = (const GLfloat *)v;
-
- out[0] = in[0];
- out[1] = 0;
- out[2] = 0;
- out[3] = 1;
-}
+ if (vtx->need_extras &&
+ (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) {
+ vtx->copy_pv = _tnl_generic_copy_pv_extras;
+ } else {
+ vtx->copy_pv = _tnl_generic_copy_pv;
+ }
-static void extract_4ub_4f_rgba( const struct attr *a, GLfloat *out,
- const char *v )
-{
- UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]);
+ vtx->copy_pv( ctx, edst, esrc );
}
-static void extract_4ub_4f_bgra( const struct attr *a, GLfloat *out,
- const char *v )
-{
- UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]);
-}
-static void extract_3ub_3f_rgb( const struct attr *a, GLfloat *out,
- const char *v )
-{
- UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
- out[3] = 1;
-}
+/***********************************************************************
+ * Public entrypoints, mostly dispatch to the above:
+ */
-static void extract_3ub_3f_bgr( const struct attr *a, GLfloat *out,
- const char *v )
-{
- UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
- UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
- out[3] = 1;
-}
-static void extract_1ub_1f( const struct attr *a, GLfloat *out, const char *v )
+/* Interpolate between two vertices to produce a third:
+ */
+void _tnl_interp( GLcontext *ctx,
+ GLfloat t,
+ GLuint edst, GLuint eout, GLuint ein,
+ GLboolean force_boundary )
{
- UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
- out[1] = 0;
- out[2] = 0;
- out[3] = 1;
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ vtx->interp( ctx, t, edst, eout, ein, force_boundary );
}
-/***********************************************************************
- * Generic (non-codegen) functions for whole vertices or groups of
- * vertices
+/* Copy colors from one vertex to another:
*/
-
-static void generic_emit( GLcontext *ctx,
- GLuint start, GLuint end,
- void *dest,
- GLuint stride )
+void _tnl_copy_pv( GLcontext *ctx, GLuint edst, GLuint esrc )
{
- int i, j;
- char *v = (char *)dest;
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ vtx->copy_pv( ctx, edst, esrc );
+}
- vtx->vertex_buf = v - start * stride;
- vtx->vertex_stride = stride;
- end -= start;
+/* Extract a named attribute from a hardware vertex. Will have to
+ * reverse any viewport transformation, swizzling or other conversions
+ * which may have been applied:
+ */
+void _tnl_get_attr( GLcontext *ctx, const void *vin,
+ GLenum attr, GLfloat *dest )
+{
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ const struct tnl_clipspace_attr *a = vtx->attr;
+ const GLuint attr_count = vtx->attr_count;
+ GLuint j;
- for (j = 0; j < vtx->attr_count; j++) {
- GLvector4f *vptr = VB->AttrPtr[a[j].attrib];
- a[j].inputptr = STRIDE_4F(vptr->data, start * vptr->stride);
+ for (j = 0; j < attr_count; j++) {
+ if (a[j].attrib == attr) {
+ a[j].extract( &a[j], dest, (GLubyte *)vin + a[j].vertoffset );
+ return;
+ }
}
- for (i = 0 ; i < end ; i++, v += stride) {
- for (j = 0; j < vtx->attr_count; j++) {
- int *in = a[j].inputptr;
- (char *)a[j].inputptr += a[j].inputstride;
- a[j].out( &a[j], v + a[j].vertoffset, in );
- }
+ /* Else return the value from ctx->Current.
+ */
+ if (attr == _TNL_ATTRIB_POINTSIZE) {
+ /* If the hardware vertex doesn't have point size then use size from
+ * GLcontext. XXX this will be wrong if drawing attenuated points!
+ */
+ dest[0] = ctx->Point.Size;
+ }
+ else {
+ memcpy( dest, ctx->Current.Attrib[attr], 4*sizeof(GLfloat));
}
}
-static void generic_interp( GLcontext *ctx,
- GLfloat t,
- GLuint edst, GLuint eout, GLuint ein,
- GLboolean force_boundary )
+/* Complementary operation to the above.
+ */
+void _tnl_set_attr( GLcontext *ctx, void *vout,
+ GLenum attr, const GLfloat *src )
{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- char *vin = vtx->vertex_buf + ein * vtx->vertex_stride;
- char *vout = vtx->vertex_buf + eout * vtx->vertex_stride;
- char *vdst = vtx->vertex_buf + edst * vtx->vertex_stride;
- const struct attr *a = vtx->attr;
- int attr_count = vtx->attr_count;
- int j;
-
- for (j = 0; j < attr_count; j++) {
- GLfloat fin[4], fout[4], fdst[4];
-
- a[j].extract( &a[j], vin, fin );
- a[j].extract( &a[j], vout, fout );
-
- INTERP_F( t, fdst[3], fout[3], fin[3] );
- INTERP_F( t, fdst[2], fout[2], fin[2] );
- INTERP_F( t, fdst[1], fout[1], fin[1] );
- INTERP_F( t, fdst[0], fout[0], fin[0] );
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ const struct tnl_clipspace_attr *a = vtx->attr;
+ const GLuint attr_count = vtx->attr_count;
+ GLuint j;
- a[j].insert( &a[j], vdst, fdst );
+ for (j = 0; j < attr_count; j++) {
+ if (a[j].attrib == attr) {
+ a[j].insert[4-1]( &a[j], (GLubyte *)vout + a[j].vertoffset, src );
+ return;
+ }
}
}
-/* Extract color attributes from one vertex and insert them into
- * another. (Shortcircuit extract/insert with memcpy).
- */
-static void generic_copy_colors( GLcontext *ctx, GLuint edst, GLuint esrc )
+void *_tnl_get_vertex( GLcontext *ctx, GLuint nr )
{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- char *vsrc = vert_start + esrc * vert_stride;
- char *vdst = vert_start + edst * vert_stride;
- const struct attr *a = vtx->attr;
- int attr_count = vtx->attr_count;
- int j;
-
- for (j = 0; j < attr_count; j++) {
- if (a[j].attribute == VERT_ATTRIB_COLOR0 ||
- a[j].attribute == VERT_ATTRIB_COLOR1) {
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
- memcpy( vdst + a[j].vertoffset,
- vsrc + a[j].vertoffset,
- a[j].vertattrsize );
- }
+ return vtx->vertex_buf + nr * vtx->vertex_size;
}
-static void generic_get_attr( GLcontext *ctx, const char *vertex,
- GLenum attr, GLfloat *dest )
+void _tnl_invalidate_vertex_state( GLcontext *ctx, GLuint new_state )
{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- const struct attr *a = vtx->attr;
- int attr_count = vtx->attr_count;
- int j;
-
- for (j = 0; j < attr_count; j++) {
- if (a[j].attribute == attr) {
- a[j].extract( &a[j], vin, dest );
- return;
- }
+ if (new_state & (_DD_NEW_TRI_LIGHT_TWOSIDE|_DD_NEW_TRI_UNFILLED) ) {
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ vtx->new_inputs = ~0;
+ vtx->interp = choose_interp_func;
+ vtx->copy_pv = choose_copy_pv_func;
}
-
- /* Else return the value from ctx->Current
- */
- memcpy( dest, ctx->Current.Attrib[attr], 4*sizeof(GLfloat));
}
+static void invalidate_funcs( struct tnl_clipspace *vtx )
+{
+ vtx->emit = choose_emit_func;
+ vtx->interp = choose_interp_func;
+ vtx->copy_pv = choose_copy_pv_func;
+ vtx->new_inputs = ~0;
+}
+GLuint _tnl_install_attrs( GLcontext *ctx, const struct tnl_attr_map *map,
+ GLuint nr, const GLfloat *vp,
+ GLuint unpacked_size )
+{
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ GLuint offset = 0;
+ GLuint i, j;
+ assert(nr < _TNL_ATTRIB_MAX);
+ assert(nr == 0 || map[0].attrib == VERT_ATTRIB_POS);
+ vtx->new_inputs = ~0;
+ vtx->need_viewport = GL_FALSE;
+ if (vp) {
+ vtx->need_viewport = GL_TRUE;
+ }
+ for (j = 0, i = 0; i < nr; i++) {
+ const GLuint format = map[i].format;
+ if (format == EMIT_PAD) {
+ if (DBG)
+ printf("%d: pad %d, offset %d\n", i,
+ map[i].offset, offset);
-/***********************************************************************
- * Build codegen functions or return generic ones:
- */
+ offset += map[i].offset;
+ }
+ else {
+ GLuint tmpoffset;
+
+ if (unpacked_size)
+ tmpoffset = map[i].offset;
+ else
+ tmpoffset = offset;
+
+ if (vtx->attr_count != j ||
+ vtx->attr[j].attrib != map[i].attrib ||
+ vtx->attr[j].format != format ||
+ vtx->attr[j].vertoffset != tmpoffset) {
+ invalidate_funcs(vtx);
+
+ vtx->attr[j].attrib = map[i].attrib;
+ vtx->attr[j].format = format;
+ vtx->attr[j].vp = vp;
+ vtx->attr[j].insert = _tnl_format_info[format].insert;
+ vtx->attr[j].extract = _tnl_format_info[format].extract;
+ vtx->attr[j].vertattrsize = _tnl_format_info[format].attrsize;
+ vtx->attr[j].vertoffset = tmpoffset;
+ }
+
+
+ if (DBG)
+ printf("%d: %s, vp %p, offset %d\n", i,
+ _tnl_format_info[format].name, (void *)vp,
+ vtx->attr[j].vertoffset);
+
+ offset += _tnl_format_info[format].attrsize;
+ j++;
+ }
+ }
-static void choose_emit_func( GLcontext *ctx,
- GLuint start, GLuint end,
- void *dest,
- GLuint stride )
-{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- vtx->emit_func = generic_emit_func;
- vtx->emit_func( ctx, start, end, dest, stride );
-}
+ vtx->attr_count = j;
+ if (unpacked_size)
+ vtx->vertex_size = unpacked_size;
+ else
+ vtx->vertex_size = offset;
-static void choose_interp_func( GLcontext *ctx,
- GLfloat t,
- GLuint edst, GLuint eout, GLuint ein,
- GLboolean force_boundary )
-{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- vtx->interp_func = generic_interp_func;
- vtx->interp_func( ctx, t, edst, eout, ein, force_boundary );
+ assert(vtx->vertex_size <= vtx->max_vertex_size);
+ return vtx->vertex_size;
}
-static void choose_copy_color_func( GLcontext *ctx, GLuint edst, GLuint esrc )
-{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- vtx->copy_color_func = generic_copy_color_func;
- vtx->copy_color_func( ctx, edst, esrc );
-}
-
-/***********************************************************************
- * Public entrypoints, mostly dispatch to the above:
- */
-
-void _tnl_emit( GLcontext *ctx,
- GLuint start, GLuint end,
- void *dest,
- GLuint stride )
+void _tnl_invalidate_vertices( GLcontext *ctx, GLuint newinputs )
{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- vtx->emit_func( ctx, start, end, dest, stride );
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ vtx->new_inputs |= newinputs;
}
-/* Interpolate between two vertices to produce a third:
- */
-void _tnl_interp( GLcontext *ctx,
- GLfloat t,
- GLuint edst, GLuint eout, GLuint ein,
- GLboolean force_boundary )
-{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- vtx->interp_func( ctx, t, edst, eout, ein, force_boundary );
-}
-/* Copy colors from one vertex to another:
+/* This event has broader use beyond this file - will move elsewhere
+ * and probably invoke a driver callback.
*/
-void _tnl_copy_colors( GLcontext *ctx, GLuint edst, GLuint esrc )
+void _tnl_notify_pipeline_output_change( GLcontext *ctx )
{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- vtx->copy_color_func( ctx, edst, esrc );
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ invalidate_funcs(vtx);
}
-/* Extract a named attribute from a hardware vertex. Will have to
- * reverse any viewport transformation, swizzling or other conversions
- * which may have been applied:
- */
-void _tnl_get_attr( GLcontext *ctx, void *vertex, GLenum attrib,
- GLfloat *dest )
+static void adjust_input_ptrs( GLcontext *ctx, GLint diff)
{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- vtx->get_attr_func( ctx, vertex, attrib, dest );
+ struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ struct tnl_clipspace_attr *a = vtx->attr;
+ const GLuint count = vtx->attr_count;
+ GLuint j;
+
+ diff -= 1;
+ for (j=0; j<count; ++j) {
+ register GLvector4f *vptr = VB->AttribPtr[a->attrib];
+ (a++)->inputptr += diff*vptr->stride;
+ }
}
+static void update_input_ptrs( GLcontext *ctx, GLuint start )
+{
+ struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ struct tnl_clipspace_attr *a = vtx->attr;
+ const GLuint count = vtx->attr_count;
+ GLuint j;
+
+ for (j = 0; j < count; j++) {
+ GLvector4f *vptr = VB->AttribPtr[a[j].attrib];
+ if (vtx->emit != choose_emit_func) {
+ assert(a[j].inputstride == vptr->stride);
+ assert(a[j].inputsize == vptr->size);
+ }
-void _tnl_install_attrs( GLcontext *ctx, const struct dri_attr_map *map,
- GLuint nr, const GLfloat *vp )
-{
- struct dri_vertex_state *vtx = GET_VERTEX_STATE(ctx);
- int i;
-
- assert(nr < _TNL_ATTR_MAX);
-
- vtx->attr_count = nr;
- vtx->emit_func = choose_emit_func;
- vtx->interp_func = choose_interp_func;
- vtx->copy_color_func = choose_copy_color_func;
- vtx->get_attr_func = choose_get_attr_func;
-
- for (i = 0; i < nr; i++) {
- GLuint attrib = map[i].attrib;
- vtx->attr[i].attrib = map[i].attrib;
- vtx->attr[i].hw_format = map[i].hw_format;
- vtx->attr[i].vp = vp;
- vtx->attr[i].insert = attrib_info[attrib].insert;
- vtx->attr[i].extract = attrib_info[attrib].extract;
- vtx->attr[i].vertattrsize = attrib_info[attrib].attrsize;
- vtx->attr[i].vertoffset = offset;
- offset += attrib_info[attrib].attrsize;
+ a[j].inputptr = ((GLubyte *)vptr->data) + start * vptr->stride;
+ }
+
+ if (a->vp) {
+ vtx->vp_scale[0] = a->vp[MAT_SX];
+ vtx->vp_scale[1] = a->vp[MAT_SY];
+ vtx->vp_scale[2] = a->vp[MAT_SZ];
+ vtx->vp_scale[3] = 1.0;
+ vtx->vp_xlate[0] = a->vp[MAT_TX];
+ vtx->vp_xlate[1] = a->vp[MAT_TY];
+ vtx->vp_xlate[2] = a->vp[MAT_TZ];
+ vtx->vp_xlate[3] = 0.0;
}
}
+void _tnl_build_vertices( GLcontext *ctx,
+ GLuint start,
+ GLuint end,
+ GLuint newinputs )
+{
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ update_input_ptrs( ctx, start );
+ vtx->emit( ctx, end - start,
+ (GLubyte *)(vtx->vertex_buf +
+ start * vtx->vertex_size));
+}
-
-/* Populate a swrast SWvertex from an attrib-style vertex.
+/* Emit VB vertices start..end to dest. Note that VB vertex at
+ * postion start will be emitted to dest at position zero.
*/
-void _tnl_translate( GLcontext *ctx, const void *vertex, SWvertex *dest )
+void *_tnl_emit_vertices_to_buffer( GLcontext *ctx,
+ GLuint start,
+ GLuint end,
+ void *dest )
{
- _tnl_get_attr( ctx, vertex, VERT_ATTRIB_POS, dest.win );
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
- for (i = 0 ; i < ctx->Const.MaxTextureUnits ; i++)
- _tnl_get_attr( ctx, vertex, VERT_ATTRIB_TEX(i), dest.texcoord[i] );
-
- _tnl_get_attr( ctx, vertex, VERT_ATTRIB_COLOR0, tmp );
- UNCLAMPED_FLOAT_TO_CHAN_RGBA( dest.color, tmp );
-
- _tnl_get_attr( ctx, vertex, VERT_ATTRIB_COLOR1, tmp );
- UNCLAMPED_FLOAT_TO_CHAN_RGB( dest.specular, tmp );
+ update_input_ptrs(ctx, start);
+ /* Note: dest should not be adjusted for non-zero 'start' values:
+ */
+ vtx->emit( ctx, end - start, (GLubyte*) dest );
+ return (void *)((GLubyte *)dest + vtx->vertex_size * (end - start));
+}
- _tnl_get_attr( ctx, vertex, VERT_ATTRIB_FOG, tmp );
- dest.fog = tmp[0];
+/* Emit indexed VB vertices start..end to dest. Note that VB vertex at
+ * postion start will be emitted to dest at position zero.
+ */
- _tnl_get_attr( ctx, vertex, VERT_ATTRIB_INDEX, tmp );
- dest.index = (GLuint) tmp[0];
+void *_tnl_emit_indexed_vertices_to_buffer( GLcontext *ctx,
+ const GLuint *elts,
+ GLuint start,
+ GLuint end,
+ void *dest )
+{
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ GLuint oldIndex;
+ GLubyte *cdest = dest;
+
+ update_input_ptrs(ctx, oldIndex = elts[start++]);
+ vtx->emit( ctx, 1, cdest );
+ cdest += vtx->vertex_size;
+
+ for (; start < end; ++start) {
+ adjust_input_ptrs(ctx, elts[start] - oldIndex);
+ oldIndex = elts[start];
+ vtx->emit( ctx, 1, cdest);
+ cdest += vtx->vertex_size;
+ }
- _tnl_get_attr( ctx, vertex, VERT_ATTRIB_POINTSIZE, tmp );
- dest.pointSize = tmp[0];
+ return (void *) cdest;
}
-static void interp_extras( GLcontext *ctx,
- GLfloat t,
- GLuint dst, GLuint out, GLuint in,
- GLboolean force_boundary )
+void _tnl_init_vertices( GLcontext *ctx,
+ GLuint vb_size,
+ GLuint max_vertex_size )
{
- struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
- if (VB->ColorPtr[1]) {
- assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat));
+ _tnl_install_attrs( ctx, NULL, 0, NULL, 0 );
- INTERP_4F( t,
- GET_COLOR(VB->ColorPtr[1], dst),
- GET_COLOR(VB->ColorPtr[1], out),
- GET_COLOR(VB->ColorPtr[1], in) );
-
- if (VB->SecondaryColorPtr[1]) {
- INTERP_3F( t,
- GET_COLOR(VB->SecondaryColorPtr[1], dst),
- GET_COLOR(VB->SecondaryColorPtr[1], out),
- GET_COLOR(VB->SecondaryColorPtr[1], in) );
- }
+ vtx->need_extras = GL_TRUE;
+ if (max_vertex_size > vtx->max_vertex_size) {
+ _tnl_free_vertices( ctx );
+ vtx->max_vertex_size = max_vertex_size;
+ vtx->vertex_buf = (GLubyte *)_mesa_align_calloc(vb_size * max_vertex_size, 32 );
+ invalidate_funcs(vtx);
}
- if (VB->EdgeFlag) {
- VB->EdgeFlag[dst] = VB->EdgeFlag[out] || force_boundary;
+ switch(CHAN_TYPE) {
+ case GL_UNSIGNED_BYTE:
+ vtx->chan_scale[0] = 255.0;
+ vtx->chan_scale[1] = 255.0;
+ vtx->chan_scale[2] = 255.0;
+ vtx->chan_scale[3] = 255.0;
+ break;
+ case GL_UNSIGNED_SHORT:
+ vtx->chan_scale[0] = 65535.0;
+ vtx->chan_scale[1] = 65535.0;
+ vtx->chan_scale[2] = 65535.0;
+ vtx->chan_scale[3] = 65535.0;
+ break;
+ default:
+ vtx->chan_scale[0] = 1.0;
+ vtx->chan_scale[1] = 1.0;
+ vtx->chan_scale[2] = 1.0;
+ vtx->chan_scale[3] = 1.0;
+ break;
}
- generic_interp(ctx, t, dst, out, in, force_boundary);
+ vtx->identity[0] = 0.0;
+ vtx->identity[1] = 0.0;
+ vtx->identity[2] = 0.0;
+ vtx->identity[3] = 1.0;
+
+ vtx->codegen_emit = NULL;
+
+#ifdef USE_SSE_ASM
+ if (!_mesa_getenv("MESA_NO_CODEGEN"))
+ vtx->codegen_emit = _tnl_generate_sse_emit;
+#endif
}
-static void copy_pv_extras( GLcontext *ctx,
- GLuint dst, GLuint src )
+
+void _tnl_free_vertices( GLcontext *ctx )
{
- struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+ TNLcontext *tnl = TNL_CONTEXT(ctx);
+ if (tnl) {
+ struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+ struct tnl_clipspace_fastpath *fp, *tmp;
- if (VB->ColorPtr[1]) {
- COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst),
- GET_COLOR(VB->ColorPtr[1], src) );
+ if (vtx->vertex_buf) {
+ _mesa_align_free(vtx->vertex_buf);
+ vtx->vertex_buf = NULL;
+ }
- if (VB->SecondaryColorPtr[1]) {
- COPY_4FV( GET_COLOR(VB->SecondaryColorPtr[1], dst),
- GET_COLOR(VB->SecondaryColorPtr[1], src) );
+ for (fp = vtx->fastpath ; fp ; fp = tmp) {
+ tmp = fp->next;
+ FREE(fp->attr);
+
+ /* KW: At the moment, fp->func is constrained to be allocated by
+ * _mesa_exec_alloc(), as the hardwired fastpaths in
+ * t_vertex_generic.c are handled specially. It would be nice
+ * to unify them, but this probably won't change until this
+ * module gets another overhaul.
+ */
+ _mesa_exec_free((void *) fp->func);
+ FREE(fp);
}
- }
- generic_copy_colors(ctx, dst, src);
+ vtx->fastpath = NULL;
+ }
}
-
-#endif
projects / mesa.git / blobdiff