<div dir="ltr">This seems very much over-the-top. It would be better to either find the specific bug or else just allocate the BOs we use for states as snooped. See also the anv_gem_set_caching call in genX_query.c. </div> <div class="gmail_quote"><div dir="ltr">On Fri, Dec 7, 2018 at 6:06 PM Rafael Antognolli <<a href="mailto:rafael.antognolli@intel.com">rafael.antognolli@intel.com</a>> wrote: </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">TODO: This is just flushing the entire dynamic states on every execbuf. Maybe it's too much. However, in theory we should be already flushing the states as needed, but I think we didn't hit any bug due to the coherence implied by userptr. --- src/intel/vulkan/anv_batch_chain.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/src/intel/vulkan/anv_batch_chain.c b/src/intel/vulkan/anv_batch_chain.c index 65df28ccb91..99009679435 100644 --- a/src/intel/vulkan/anv_batch_chain.c +++ b/src/intel/vulkan/anv_batch_chain.c @@ -1366,6 +1366,10 @@ anv_reloc_list_add_dep(struct anv_cmd_buffer *cmd_buffer, anv_block_pool_foreach_bo(bo_list, iter, bo) { _mesa_set_add(relocs->deps, bo); + if (!cmd_buffer->device->info.has_llc) { + for (uint32_t i = 0; i < bo->size; i += CACHELINE_SIZE) + __builtin_ia32_clflush(bo->map + i); + } } } -- 2.17.1 _______________________________________________ mesa-dev mailing list <a href="mailto:mesa-dev@lists.freedesktop.org" target="_blank">mesa-dev@lists.freedesktop.org</a> <a href="https://lists.freedesktop.org/mailman/listinfo/mesa-dev" rel="noreferrer" target="_blank">https://lists.freedesktop.org/mailman/listinfo/mesa-dev</a> </blockquote></div>