X-Git-Url: https://git.cworth.org/git?a=blobdiff_plain;f=src%2Fexa%2Fopannotate_i965_prepare_composite.mdwn;h=75bda49fa7db68ba04c1c272f4c0c8e421fcfe7c;hb=58e046dcdfeea6374c732d1bb5630674f3286423;hp=7ff167ec66cfb486cc529457d5878ba24b59d7f9;hpb=899df8c7fc2a62d35189acbbcacf22a12416ef69;p=cworth.org diff --git a/src/exa/opannotate_i965_prepare_composite.mdwn b/src/exa/opannotate_i965_prepare_composite.mdwn index 7ff167e..75bda49 100644 --- a/src/exa/opannotate_i965_prepare_composite.mdwn +++ b/src/exa/opannotate_i965_prepare_composite.mdwn @@ -1,11 +1,11 @@ -[[meta title="Using opannotate to make sense of profiles"]] +[[!meta title="Using opannotate to make sense of profiles"]] -[[tag exa performance xorg]] +[[!tag exa performance xorg]] After I recently posted some surprising [[profiles|mozilla_i965_profiles]], I received useful feedback from Michel Dänzer, Adam Jackson, and Eric Anholt. There was general -agreement that the i965_prepare_composite function is generally stupid +agreement that the `i965_prepare_composite` function is generally stupid about acting synchronously in order to reuse a single state buffer, and that this shouldn't be too hard to optimize. Options include using a ring of buffers and synchronizing only when wrapping around and also