Since we are now consuming two ringbuffers at a time, we probably want a
pool larger than 4.. but we don't need each individual ringbuffer to be
so large, so offset the pool size increase by reducing rb size.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
pctx->flush = fd_context_flush;
for (i = 0; i < ARRAY_SIZE(ctx->rings); i++) {
- ctx->rings[i] = fd_ringbuffer_new(screen->pipe, 0x400000);
+ ctx->rings[i] = fd_ringbuffer_new(screen->pipe, 0x100000);
if (!ctx->rings[i])
goto fail;
}
* normally have to wait before resetting to the start of the next
* rb.
*/
- struct fd_ringbuffer *rings[4];
+ struct fd_ringbuffer *rings[8];
unsigned rings_idx;
/* normal draw/clear cmds: */