Yes, oops. Shouldn't fix any invalidation issues. Could fix unbounded memory growth due to switching back and forth between cp and 3d without setting any new images. Please fold this into your series. <div class="gmail_quote">On May 19, 2016 4:13 AM, "Samuel Pitoiset" <<a href="mailto:samuel.pitoiset@gmail.com">samuel.pitoiset@gmail.com</a>> wrote: <blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Oops? Your patch is based on your gl43 branch with my images series but this is not yet upstream. ;) Anyway, the idea sounds good to me and might explain some invalidation issues, perhaps? I'll have look later.<div class="elided-text"> On 05/19/2016 03:28 AM, Ilia Mirkin wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> Otherwise we can end up in a situation where that bin just grows and grows. Signed-off-by: Ilia Mirkin <<a href="mailto:imirkin@alum.mit.edu" target="_blank">imirkin@alum.mit.edu</a>> --- src/gallium/drivers/nouveau/nvc0/nvc0_tex.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/src/gallium/drivers/nouveau/nvc0/nvc0_tex.c b/src/gallium/drivers/nouveau/nvc0/nvc0_tex.c index cc5ea5e..2523c20 100644 --- a/src/gallium/drivers/nouveau/nvc0/nvc0_tex.c +++ b/src/gallium/drivers/nouveau/nvc0/nvc0_tex.c @@ -963,6 +963,11 @@ nvc0_validate_suf(struct nvc0_context *nvc0, int s) struct nouveau_pushbuf *push = nvc0->base.pushbuf; struct nvc0_screen *screen = nvc0->screen; + if (s == 5) + nouveau_bufctx_reset(nvc0->bufctx_cp, NVC0_BIND_CP_SUF); + else + nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_SUF); + for (int i = 0; i < NVC0_MAX_IMAGES; ++i) { struct pipe_image_view *view = &nvc0->images[s][i]; int width, height, depth; </blockquote> -- </div> -Samuel </blockquote></div>