diff options
| author | Brian Paul <brianp@vmware.com> | 2015-10-15 12:33:00 -0600 |
|---|---|---|
| committer | Brian Paul <brianp@vmware.com> | 2015-10-20 12:52:40 -0600 |
| commit | f6d4e20d10d2316b70b73676e97b2c1e5cf7634a (patch) | |
| tree | 6d17b5d15172830b7885174c95df2e2af3d584e2 | |
| parent | d11fefa96165836ffeed531a74319a64aa98a696 (diff) | |
| download | external_mesa3d-f6d4e20d10d2316b70b73676e97b2c1e5cf7634a.zip external_mesa3d-f6d4e20d10d2316b70b73676e97b2c1e5cf7634a.tar.gz external_mesa3d-f6d4e20d10d2316b70b73676e97b2c1e5cf7634a.tar.bz2 | |
vbo: reduce number of vertex buffer mappings for vertex attributes
Whenever we got a glColor, glNormal, glTexCoord, etc. call outside a
glBegin/End pair, we'd immediately map a vertex buffer to begin
accumulating vertex data. In some cases, such as with display lists,
this led to excessive vertex buffer mapping. For example, if we have
a display list such as:
glNewList(42, GL_COMPILE);
glBegin(prim);
glVertex2f();
...
glVertex2f();
glEnd();
glEndList();
Then did:
glColor3f();
glCallList(42);
We'd map a vertex buffer as soon as we saw glColor3f but we'd never
actually write anything to it. Note that the vertex position data
was put into a vertex buffer during display list compilation.
With this change, we delay mapping the vertex buffer until we actually
have a vertex to write to it (triggered by a glVertex() call). In the
above case, we no longer map a vertex buffer when setting the color and
calling the list.
For drivers such as VMware's, reducing buffer mappings gives improved
performance.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
| -rw-r--r-- | src/mesa/vbo/vbo_exec_api.c | 18 |
1 files changed, 13 insertions, 5 deletions
diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index 7ae08fe..789869a 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -446,10 +446,6 @@ do { \ \ assert(sz == 1 || sz == 2); \ \ - if (unlikely(!(ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT))) { \ - vbo_exec_begin_vertices(ctx); \ - } \ - \ /* check if attribute size or type is changing */ \ if (unlikely(exec->vtx.active_sz[A] != N * sz) || \ unlikely(exec->vtx.attrtype[A] != T)) { \ @@ -470,6 +466,15 @@ do { \ /* This is a glVertex call */ \ GLuint i; \ \ + if (unlikely((ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) == 0)) { \ + vbo_exec_begin_vertices(ctx); \ + } \ + \ + if (unlikely(!exec->vtx.buffer_ptr)) { \ + vbo_exec_vtx_map(exec); \ + } \ + assert(exec->vtx.buffer_ptr); \ + \ /* copy 32-bit words */ \ for (i = 0; i < exec->vtx.vertex_size; i++) \ exec->vtx.buffer_ptr[i] = exec->vtx.vertex[i]; \ @@ -482,7 +487,10 @@ do { \ \ if (++exec->vtx.vert_count >= exec->vtx.max_vert) \ vbo_exec_vtx_wrap( exec ); \ - } \ + } else { \ + /* we now have accumulated per-vertex attributes */ \ + ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT; \ + } \ } while (0) #define ERROR(err) _mesa_error( ctx, err, __func__ ) |