Mesa (master): mesa/st: don't reuse vertex buffers for bitmap, clear quads
Keith Whitwell
keithw at kemper.freedesktop.org
Fri Oct 2 18:03:28 UTC 2009
Module: Mesa
Branch: master
Commit: 918199fb0f5d84121e0ac5821168cd0e886b22e9
URL: http://cgit.freedesktop.org/mesa/mesa/commit/?id=918199fb0f5d84121e0ac5821168cd0e886b22e9
Author: Keith Whitwell <keithw at vmware.com>
Date: Fri Oct 2 15:36:47 2009 +0100
mesa/st: don't reuse vertex buffers for bitmap, clear quads
Currently using max_slots > 1 will cause synchronous rendering
if the driver flushes its command buffers between one bitmap and the
next.
Need to improve buffer_write to allow NO_WAIT (as well as no_flush)
updates to buffers where we know there is no conflict with previous
data.
---
src/mesa/state_tracker/st_cb_bitmap.c | 13 ++++++++++++-
src/mesa/state_tracker/st_cb_clear.c | 13 ++++++++++++-
2 files changed, 24 insertions(+), 2 deletions(-)
diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c
index 902fb38..a22fa68 100644
--- a/src/mesa/state_tracker/st_cb_bitmap.c
+++ b/src/mesa/state_tracker/st_cb_bitmap.c
@@ -330,7 +330,18 @@ setup_bitmap_vertex_data(struct st_context *st,
const GLfloat clip_y0 = (GLfloat)(y0 / fb_height * 2.0 - 1.0);
const GLfloat clip_x1 = (GLfloat)(x1 / fb_width * 2.0 - 1.0);
const GLfloat clip_y1 = (GLfloat)(y1 / fb_height * 2.0 - 1.0);
- const GLuint max_slots = 4096 / sizeof(st->bitmap.vertices);
+
+ /* XXX: Need to improve buffer_write to allow NO_WAIT (as well as
+ * no_flush) updates to buffers where we know there is no conflict
+ * with previous data. Currently using max_slots > 1 will cause
+ * synchronous rendering if the driver flushes its command buffers
+ * between one bitmap and the next. Our flush hook below isn't
+ * sufficient to catch this as the driver doesn't tell us when it
+ * flushes its own command buffers. Until this gets fixed, pay the
+ * price of allocating a new buffer for each bitmap cache-flush to
+ * avoid synchronous rendering.
+ */
+ const GLuint max_slots = 1; /* 4096 / sizeof(st->bitmap.vertices); */
GLuint i;
if (st->bitmap.vbuf_slot >= max_slots) {
diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c
index 8a8c99f..3651072 100644
--- a/src/mesa/state_tracker/st_cb_clear.c
+++ b/src/mesa/state_tracker/st_cb_clear.c
@@ -116,7 +116,18 @@ draw_quad(GLcontext *ctx,
{
struct st_context *st = ctx->st;
struct pipe_context *pipe = st->pipe;
- const GLuint max_slots = 1024 / sizeof(st->clear.vertices);
+
+ /* XXX: Need to improve buffer_write to allow NO_WAIT (as well as
+ * no_flush) updates to buffers where we know there is no conflict
+ * with previous data. Currently using max_slots > 1 will cause
+ * synchronous rendering if the driver flushes its command buffers
+ * between one bitmap and the next. Our flush hook below isn't
+ * sufficient to catch this as the driver doesn't tell us when it
+ * flushes its own command buffers. Until this gets fixed, pay the
+ * price of allocating a new buffer for each bitmap cache-flush to
+ * avoid synchronous rendering.
+ */
+ const GLuint max_slots = 1; /* 1024 / sizeof(st->clear.vertices); */
GLuint i;
if (st->clear.vbuf_slot >= max_slots) {
More information about the mesa-commit
mailing list