<html><head><style type='text/css'>p { margin: 0; }</style></head><body><div style='font-family: Times New Roman; font-size: 12pt; color: #000000'>My patch below:<br>>====================================<<br>diff -r 2b680924a73f make/lib/Awt2dLibraries.gmk<br>--- a/make/lib/Awt2dLibraries.gmk Wed Sep 16 18:34:38 2015 +0300<br>+++ b/make/lib/Awt2dLibraries.gmk Thu Oct 01 17:06:38 2015 +0300<br>@@ -243,7 +243,7 @@<br> EXCLUDES := $(LIBAWT_EXCLUDES), \<br> EXCLUDE_FILES := $(LIBAWT_EXFILES), \<br> OPTIMIZATION := LOW, \<br>- CFLAGS := $(CFLAGS_JDKLIB) $(LIBAWT_CFLAGS), \<br>+ CFLAGS := -fgcse-after-reload -ftree-vectorize $(CFLAGS_JDKLIB) $(LIBAWT_CFLAGS), \<br> DISABLED_WARNINGS_gcc := sign-compare unused-result maybe-uninitialized \<br> format-nonliteral parentheses, \<br>>====================================<<br><br>The next commands will show gcc commandline<br>touch jdk/src/java.desktop/share/native/libawt/java2d/loops/AnyInt.c<br>make images LOG=debug<br><br>----- bourges.laurent@gmail.com wrote:
<br>> <p dir="ltr">Sergey,</p>
<p dir="ltr">Thanks for having tests.</p>
<p dir="ltr">What is the units of your your results ? <br>>
I guess it is on the time axis: slower values are better.</p>
<p dir="ltr">How did you hack the gcc options in the openjdk build scripts ?<br>>
I could try on my local build too.</p>
<p dir="ltr">Bye,<br>>
Laurent</p>
<p dir="ltr">Le 1 oct. 2015 17:39, "Sergey Bylokhov" <<a href="mailto:sergey.bylokhov@oracle.com" target="_blank">sergey.bylokhov@oracle.com</a>> a écrit :<br>>
><br>>
> I got results below after I made a hack and added -march=native to the libawt library:<br>>
><br>>
> "EllipseFill.fillEllipse 1400"<br>>
> 8u60_RE: 6,540<br>>
> 9_dev: 8,457<br>>
> 9_hack: 6,276<br>>
><br>>
> So we have a window for tweaking.<br>>
><br>>
> ----- <a href="mailto:sergey.bylokhov@oracle.com" target="_blank">sergey.bylokhov@oracle.com</a> wrote:<br>>
><br>>
> > Hello,<br>>
> > I built both version of jdk8 and jdk9 on my local system, and compares<br>>
> > output of preprocessor and generated assemblers, both are identical.<br>>
> > But jdk8u60 from RE actually 20% faster than my version of jdk8. It<br>>
> > seems that the difference is in the compiler and/or in some<br>>
> > compiler(gcc) options used by default.<br>>
> ><br>>
> > ----- <a href="mailto:james.graham@oracle.com" target="_blank">james.graham@oracle.com</a> wrote:<br>>
> ><br>>
> > > As far as why the software loops are slower...<br>>
> > ><br>>
> > > Did any command line options change for compiling IntArgbPre.c?<br>>
> > Touch<br>>
> > ><br>>
> > > the file and rebuild and verify if the compiler options are the same<br>>
> ><br>>
> > > (and that both builds use the same compiler)...<br>>
> > ><br>>
> > > ...jim<br>>
</p>
</div></body></html>