Hi Luca,
I've commited your other fixes. Good catch of the fence_signalled
negated logic.
I have some doubts about this patch though. It is impossible to satisfy
both DONTBLOCK and UNSYNCHRONIZED flags simultaneously, so the result of
specifying both is implementation dependent. Why should UNSYNCHRONIZED
be favored over DONTBLOCK?
Jose
On Tue, 2010-02-23 at 08:55 -0800, Luca Barbieri wrote:
> Depends on previous signalled value patch.
>
> If the reqiest is unsynchronized, we must not do anything, so do that
> check first.
>
> Fixes a bug if DONTBLOCK | UNSYNCHRONIZED.
>
> ---
> .../auxiliary/pipebuffer/pb_buffer_fenced.c | 34 +++++++++----------
> 1 files changed, 16 insertions(+), 18 deletions(-)
>
> diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
> b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
> index d97f749..46e490b 100644
> --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
> +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
> @@ -688,27 +688,25 @@ fenced_buffer_map(struct pb_buffer *buf,
> /*
> * Serialize writes.
> */
> - while((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) ||
> - ((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ) &&
> - (flags & PIPE_BUFFER_USAGE_CPU_WRITE))) {
> + if (flags & PIPE_BUFFER_USAGE_UNSYNCHRONIZED) {
> + while((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) ||
> + ((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ) &&
> + (flags & PIPE_BUFFER_USAGE_CPU_WRITE))) {
>
> - /*
> - * Don't wait for the GPU to finish accessing it, if blocking is
> forbidden.
> - */
> - if((flags & PIPE_BUFFER_USAGE_DONTBLOCK) &&
> - ops->fence_signalled(ops, fenced_buf->fence, 0) != 0) {
> - goto done;
> - }
> + /*
> + * Don't wait for the GPU to finish accessing it, if blocking is
> forbidden.
> + */
> + if((flags & PIPE_BUFFER_USAGE_DONTBLOCK) &&
> + ops->fence_signalled(ops, fenced_buf->fence, 0) != 0) {
> + goto done;
> + }
>
> - if (flags & PIPE_BUFFER_USAGE_UNSYNCHRONIZED) {
> - break;
> + /*
> + * Wait for the GPU to finish accessing. This will release and
> re-acquire
> + * the mutex, so all copies of mutable state must be discarded.
> + */
> + fenced_buffer_finish_locked(fenced_mgr, fenced_buf);
> }
> -
> - /*
> - * Wait for the GPU to finish accessing. This will release and
> re-acquire
> - * the mutex, so all copies of mutable state must be discarded.
> - */
> - fenced_buffer_finish_locked(fenced_mgr, fenced_buf);
> }
>
> if(fenced_buf->buffer) {
------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev
