<html> <head> <base href="https://bugs.freedesktop.org/"> </head> <body> <div> <a class="bz_bug_link bz_status_NEW " title="NEW - POSTAL 2 poor performance at certain times, RadeonSI driver" href="https://bugs.freedesktop.org/show_bug.cgi?id=97340#c1">Comment # 1</a> on <a class="bz_bug_link bz_status_NEW " title="NEW - POSTAL 2 poor performance at certain times, RadeonSI driver" href="https://bugs.freedesktop.org/show_bug.cgi?id=97340">bug 97340</a> from <a class="email" href="mailto:daniel@constexpr.org" title="Daniel Scharrer <daniel@constexpr.org>"> Daniel Scharrer</a> <pre>Created <a href="attachment.cgi?id=125916" name="attach_125916" title="GALLIUM_HUD graphs for the scene">attachment 125916</a> <a href="attachment.cgi?id=125916&action=edit" title="GALLIUM_HUD graphs for the scene">[details]</a> GALLIUM_HUD graphs for the scene I can reproduce this on VI. According to GALLIUM_HUD, the frame drops correspond to shader compilations. apitrace profiling shows the slowest calls are glDrawRangeElements, many of which take over 80ms (!). The calls don't use any user shaders so the compilations must be for internal shaders generated by Mesa to implement fixed-function functionality. Trace: <a href="http://constexpr.org/tmp/POSTAL2-radeonsi.2.trace.xz">http://constexpr.org/tmp/POSTAL2-radeonsi.2.trace.xz</a> (73 MiB) Not sure if there is a good way to fix this (besides making the shader compiler faster). Maybe there could be a lighting super-shader that can be used while more optimized variants are compiled. Kernel: 4.7.0-gentoo Mesa: git-a73116e LLVM: r278309 GPU: R9 380X CPU: FX 8350</pre> </div> <hr> You are receiving this mail because: <ul> <li>You are the assignee for the bug.</li> </ul> </body> </html>