diff options
author | Stéphane Marchesin <[email protected]> | 2013-05-04 18:59:35 -0700 |
---|---|---|
committer | Stéphane Marchesin <[email protected]> | 2013-05-08 02:06:56 -0700 |
commit | fc24c7aedeff6639e9438dbe7b91ca3775511223 (patch) | |
tree | de0b578dd79afa5704a2951e397f62d79ddab188 /src/gallium/drivers/i915/i915_prim_vbuf.c | |
parent | 7f7c7fda83eced002a4dbff483cded2e31313acd (diff) |
i915g: Optimize batchbuffer sizes
Now that we don't throttle at every batchbuffer, we can shrink
the size of batchbuffers to achieve early flushing. This gives
a significant speed boost in a lot of games (on the order of
20%).
Diffstat (limited to 'src/gallium/drivers/i915/i915_prim_vbuf.c')
-rw-r--r-- | src/gallium/drivers/i915/i915_prim_vbuf.c | 8 |
1 files changed, 5 insertions, 3 deletions
diff --git a/src/gallium/drivers/i915/i915_prim_vbuf.c b/src/gallium/drivers/i915/i915_prim_vbuf.c index 4e57eeadea7..f9cd71be457 100644 --- a/src/gallium/drivers/i915/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915/i915_prim_vbuf.c @@ -704,12 +704,14 @@ i915_vbuf_render_create(struct i915_context *i915) i915_render->i915 = i915; - i915_render->base.max_vertex_buffer_bytes = 16*4096; + i915_render->base.max_vertex_buffer_bytes = 4*4096; /* NOTE: it must be such that state and vertices indices fit in a single - * batch buffer. + * batch buffer. 4096 is one batch buffer and 430 is the max amount of + * state in dwords. The result is the number of 16-bit indices which can + * fit in a single batch buffer. */ - i915_render->base.max_indices = 16*1024; + i915_render->base.max_indices = (4096 - 430 * 4) / 2; i915_render->base.get_vertex_info = i915_vbuf_render_get_vertex_info; i915_render->base.allocate_vertices = i915_vbuf_render_allocate_vertices; |