<html><head><meta http-equiv="Content-Type" content="text/html charset=us-ascii"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><br class=""><div><blockquote type="cite" class=""><div class="">On Mar 19, 2015, at 4:46 PM, Eric Christopher <<a href="mailto:echristo@gmail.com" class="">echristo@gmail.com</a>> wrote:</div><br class="Apple-interchange-newline"><div class=""><div dir="ltr" class=""><br class=""><br class=""><div class="gmail_quote">On Thu, Mar 19, 2015 at 4:33 PM Matthias Braun <<a href="mailto:matze@braunis.de" class="">matze@braunis.de</a>> wrote:<br class=""><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi t.p.northover, qcolombet,<br class="">
<br class="">
Splitting this into an own commit to get some visibility.<br class="">
<br class="">
After doing some extensive benchmarking wiht llvm test-suite, I see no correctness or compiletime changes. I see some very slight benchmark runtime improvements (which may or may not be noise). The only noticeable regression is in a benchmark were literally 6 lines of assembly changed for the better, so that must be cache issues.<br class="">
<br class="">
REPOSITORY<br class="">
rL LLVM<br class="">
<br class="">
<a href="http://reviews.llvm.org/D8474" target="_blank" class="">http://reviews.llvm.org/D8474</a><br class="">
<br class="">
Files:<br class="">
lib/Target/AArch64/<u class=""></u>AArch64Subtarget.h<br class="">
<br class="">
Index: lib/Target/AArch64/<u class=""></u>AArch64Subtarget.h<br class="">
==============================<u class=""></u>==============================<u class=""></u>=======<br class="">
--- lib/Target/AArch64/<u class=""></u>AArch64Subtarget.h<br class="">
+++ lib/Target/AArch64/<u class=""></u>AArch64Subtarget.h<br class="">
@@ -91,6 +91,7 @@<br class="">
bool enablePostMachineScheduler() const override {<br class="">
return isCortexA53() || isCortexA57();<br class="">
}<br class="">
+ bool enableSubRegLiveness() const override { return true; }<br class="">
<br class=""></blockquote><div class=""><br class=""></div><div class="">Drive by feedback:</div><div class=""><br class=""></div><div class="">Most of AArch64 is option controlled for this sort of thing. Did you want to add an option with init(true) for this?</div><div class=""><br class=""></div><div class="">Personally no preference, but thought I'd bring it up.</div></div></div></div></blockquote><div><br class=""></div></div>In this case there is already the global -enable-subreg-liveness=false switch available, so I think there is no need for an additional aarch64 option.<div class=""><br class=""></div><div class="">- Matthias</div></body></html>