<html> <head> <base href="https://bugs.freedesktop.org/" /> </head> <body> <div> <a class="bz_bug_link bz_status_NEW " title="NEW - [HSW Regression] Synmark2_v6 Multithread performance case FPS reduced by 36%" href="https://bugs.freedesktop.org/show_bug.cgi?id=91788#c4">Comment # 4</a> on <a class="bz_bug_link bz_status_NEW " title="NEW - [HSW Regression] Synmark2_v6 Multithread performance case FPS reduced by 36%" href="https://bugs.freedesktop.org/show_bug.cgi?id=91788">bug 91788</a> from <a class="email" href="mailto:currojerez@riseup.net" title="Francisco Jerez <currojerez@riseup.net>"> Francisco Jerez</a> <pre>Yeah, apparently the regression is caused by the additional validation done for each image unit when _NEW_TEXTURE is flagged. I would expect validation for a single disabled image unit to be cheap, little more than a couple of pointer dereferences and a branch, but still it seems to add up to a measurable overhead due to the huge number of image units we expose. The attached patch removes the _NEW_TEXTURE-dependent state (actually just the _Valid field) from the image unit struct so this per-image unit validation is no longer necessary. Instead the driver is required to call a function to calculate the same value when needed, what may make per-image state updates in the driver a little more expensive but is likely to be a net win because the overall complexity is reduced since now only image units which are actually bound to a shader need to be validated.</pre> </div> <hr> You are receiving this mail because: <ul> <li>You are the QA Contact for the bug.</li> </ul> </body> </html>