[OpenJDK Rasterizer] AWT & gcc 4.8 optimization options

Sergey Bylokhov Sergey.Bylokhov at oracle.com
Wed Oct 21 19:04:16 UTC 2015


The bug was filed:
https://bugs.openjdk.java.net/browse/JDK-8140266

I will try a few steps before push the current solution:
- change the code, so this option will not be necessary.
- check other compilers, it seems gcc 4.9.2 will be default for jdk9.

On 16.10.15 10:00, Laurent Bourgès wrote:
> Sergey,
>
> Could you please fill a new bug related to the libawt compilation on linux ?
>
> As it concerns awt and not only Marlin, I would prefer to let you or
> anybody else handle this issue.
>
> To sum up:
> We observed performance loss between jdk8 and jdk9 on Maskfill (c
> software loops).
>
> I provided the EllipseFill test (jmh) and you provided a patch to gcc
> compiler options:
> CFLAGS := -fgcse-after-reload -ftree-vectorize $(CFLAGS_JDKLIB)
> $(LIBAWT_CFLAGS), \
>
> See discussion:
> http://mail.openjdk.java.net/pipermail/graphics-rasterizer-dev/2015-October/000332.html
>
> I confirmed it improved performance ~10% on my machine i7-4800 (amd64
> linux) that is comparable to jdk8 performance.
>
> Regards,
> Laurent
>


-- 
Best regards, Sergey.


More information about the graphics-rasterizer-dev mailing list