Rebase

Apply requested changes (don't force gold, don't use LTO/PGO)

Rebuilt gtkd, required changes. If Tilix succeeds I'll push the gtkd diff and update this, and it should be good to go.

Sorry for that. Should be good to go now, @JoshStrobl.

Working on this. Having issues building gtkd with new LDC, need time to look into it. Will hopefully be done by today.

Move apache-2.0 license entry

Remove unused patches

Remove explicit dependency on llvm

Remove argument that disables docs building

Hi! This should be merged as part of the LLVM upgrade. Could you update this to version 2.35? Would be appreciated.

Fix licenses

Rebuild with old licenses

I applied the licenses to the wrong project :')

Move setup to build, change python3-devel to pkgconfig(python3)

Remove cmark replaces, move setup to build

Move setup to build

Fix license and change setup to build

Remove unused patch

Don't render HTML

Move /usr/bin/cmark to /usr/bin/cmark-python

abi_symbols and abi_symbols32 are marked as binary, as there were so many symbol changes that arcanist was refusing to upload the files.

List of rebuilds:

• blender
• graph-tool
• rust
• beignet
• python-llvmlite
• dolphin-emu
• ldc
• qt-creator
• lldb
• mesalib

In D9641#157883, @Jacalz wrote:

Oh gosh, sorry @serebit. I did look though the changes, but it looks like I missed that one. Will rely om ctrl + f in the future :I

In D9641#157880, @Jacalz wrote:

@joebonrichie Mentioned earlier in the thread that it might be worth adding --enable-fma`along with AVX2. Any luck with that?

I created a patch for using AVX2. All the reverse-dependent benchmarks I've used *are linked against the haswell lib*, but they're getting no performance benefit from it. I've tested gmic, gromacs, hugin, and octave, through Phoronix (each of which do use system FFTW).

Yeah, I'm getting no benefit from Clear's flags over the ones we're currently using. AVX2 is still a pretty substantial improvement, though, so I'll push a patch adding Haswell AVX2 builds.

I'm doing some test passes with fastmath and the other additional Clear flags. It seems that the difference between Opt and Gen with smaller FFTs is nonexistent, but with larger FFTs like 4096x4096, there is a slight benefit that's outside of variance. I'll continue testing.

Here are the flags I was running with.
All three: -g2 -pipe -fPIC -Wformat -Wformat-security -D_FORTIFY_SOURCE=2 -fstack-protector-strong --param=ssp-buffer-size=32 -fasynchronous-unwind-tables -ftree-vectorize -feliminate-unused-debug-types -Wall -Wno-error -Wp,-D_REENTRANT
AVX2: -mtune=haswell -march=haswell -O3
Generic: -mtune=generic -march=x86-64 -O2
Optimized: -mtune=generic -march=x86-64 -O3 -flto

Tested a bit with O3 and AVX2. AVX2 helps a *lot*, O3 doesn't seem to help much if at all. Here's the table.

Still an issue in Firefox 80.0.1, seemingly. Tested with http://www.fileformat.info/info/emoji/browsertest.htm.

This can be closed as of R2083:ea9a6c4dcd3cdbd32dd06f140e6290976dfee122.

This seems to require 32-bit builds of Qt5, which we don't ship as far as I can tell.