GCC 6 Is On Track To Be Officially Released In About One Month

[phoronix]

Phoronix: GCC 6 Is On Track To Be Officially Released In About One Month

Longtime GCC developer Richard Biener shared a status update today concerning the state of the GNU Compiler Collection 6...

http://www.phoronix.com/scan.php?page=news_item&px=GCC-6-Track-Next-Month

[hansg]

Is there any word on whether they will be supporting any of the experimental work on modules done by clang and visual studio? This is one area of language development I'm really looking forward to...

[juno]

GCC 6 comes with HSA, right?

[mlau]

GCC 6 comes with HSA, right?

Yes it does.

[curaga]

- The size of LTO object files was reduced by about 11% (measured by compiling Firefox 46.0). - Link-time parallelization (enabled using -flto=n) was significantly improved by decreasing the size of streamed data when partitioning programs. The size of streamed IL while compiling Firefox 46.0 was reduced by 66%.

Can't wait for these two. Webkit has become such a pig, with gcc 5.2 I can barely build it at -flto={1,2,3}. All three use nearly all RAM and swap I have, 4 and above fail miserably. It's curious 1 doesn't use any less than 3. I have 8GB RAM and 1GB swap, and a build comes so close there's less than 100mb swap left, with RAM full.

[Guest]

and a build comes so close there's less than 100mb swap left, with RAM full.

You can create a 8 gb file, format it as a swap partition and mount it. It's not a solution for low ram, but works as a workaround.

[carewolf]

Can't wait for these two. Webkit has become such a pig, with gcc 5.2 I can barely build it at -flto={1,2,3}. All three use nearly all RAM and swap I have, 4 and above fail miserably. It's curious 1 doesn't use any less than 3. I have 8GB RAM and 1GB swap, and a build comes so close there's less than 100mb swap left, with RAM full.

You should have 16Gbyte of memory for linking Webkit without LTO, and last time I tried an LTO build of WebKit it used over 30Gbyte of memory and ended in an internal compiler error. It really isn't designed for being linked that way.

[curaga]

You can create a 8 gb file, format it as a swap partition and mount it.

Sorry, I do not want any more swap storms. If webkit became more bloated, I'd have to go help gcc reduce its RAM usage, heh.

You should have 16Gbyte of memory for linking Webkit without LTO, and last time I tried an LTO build of WebKit it used over 30Gbyte of memory and ended in an internal compiler error.

Why do you have such high figures? I guess you mean the gtk port, but there shouldn't be that huge differences. I can do a non-LTO link of webkitfltk with less than 1GB RAM use, and as mentioned, a LTO link uses just under 9GB. I do have to use -flto-compression=9, perhaps you can try that?

It really isn't designed for being linked that way.

I disagree, Apple themselves use LTO with clang for their builds, and in my own tests, LTO dropped the browser's size by 14.5%.

[carewolf]

Sorry, I do not want any more swap storms. If webkit became more bloated, I'd have to go help gcc reduce its RAM usage, heh. Why do you have such high figures?.

I link with debug symbols, that is the real memory eater. And no I don't mean the GTK-port. I was/am the maintainer of QtWebKit, though that is now officially unmaintained. A normal debug link or even worse release-build with debug symbols uses around 8Gbyte of memory. Pure release builds use less, but we only do those on 32-bit platforms where there isn't enough memory for a release-build with debug symbols.

Of course you can use much less if you don't do a link-time dead-core analysis and compression, or split WebKit into multiple libraries.