<html> <head> <base href="https://bugs.freedesktop.org/" /> </head> <body> <div> <a class="bz_bug_link bz_status_ASSIGNED " title="ASSIGNED - Illegal instruction _mesa_x86_64_transform_points4_general" href="https://bugs.freedesktop.org/show_bug.cgi?id=27512#c12">Comment # 12</a> on <a class="bz_bug_link bz_status_ASSIGNED " title="ASSIGNED - Illegal instruction _mesa_x86_64_transform_points4_general" href="https://bugs.freedesktop.org/show_bug.cgi?id=27512">bug 27512</a> from <a class="email" href="mailto:sroland@vmware.com" title="Roland Scheidegger <sroland@vmware.com>"> Roland Scheidegger</a> <pre>(In reply to Patrick Baggett from <a href="show_bug.cgi?id=27512#c10">comment #10</a>) > Created <a href="attachment.cgi?id=120821" name="attach_120821" title="attachment-30621-0.html">attachment 120821</a> <a href="attachment.cgi?id=120821&action=edit" title="attachment-30621-0.html">[details]</a> > attachment-30621-0.html > > Given that there is a _mesa_3dnow_transform_points4_2d in the x86-64 asm > (using MMX/3DNow! is deprecated in x86-64), it appears that this code was > copy-pasted. I wrote a quick patch to change prefetch[w] to prefetcht1, > which is more or less the equivalent in SSE. However, I'm not actually sure > those prefetches really benefit the code since they appear to be monotonic > addresses and hinting only 16 bytes ahead (a cache line is almost always at > least 32 bytes) -- maybe that sort of testing is for another day. I'd agree that it's dubious that "modern" cpus would benefit - as you said addresses are monotonic and certainly hw prefetchers should handle that pretty well. Though you could argue someone might still use some cpus with terrible prefetchers, and the prefetch instructions should not hurt (at least not much) on modern cpus neither...</pre> </div> <hr> You are receiving this mail because: <ul> <li>You are the assignee for the bug.</li> </ul> </body> </html>