AMD AOCC Performance On EPYC 7773X Milan-X Against GCC, Clang Compilers

[phoronix]

Phoronix: AMD AOCC Performance On EPYC 7773X Milan-X Against GCC, Clang Compilers

Last month with the AMD EPYC 7773X Linux benchmarks and Milan-X in the Azure cloud I showed the impressive capabilities of AMD's new Milan-X processors with 768MB of L3 cache per socket (1.5GB cache per 2P server!) for a range of workloads. All of that initial benchmarking as usual was done using the default GCC system compiler across all tested AMD/Intel processors. Of course, there also exists AMD's Optimizing C/C++ Compiler (AOCC) as a downstream of LLVM/Clang with various Zen optimization patches applied. Curious about the AOCC impact for Milan-X, here are some benchmarks looking at the EPYC 7773X 2P performance across AOCC, GCC, and LLVM Clang.

https://www.phoronix.com/vr.php?view=31029

[uid313]

So can AMD backport these improvements to the mainline LLVM/Clang?

[sdack]

Finally it is looking good for AOCC. Quite a few impressive results.

[lejeczek]

Finally it is looking good for AOCC. Quite a few impressive results.

Yes, that's right, it did not look that good in the past.
Hopefully that's not a fluke and @AMD they eventually, really began to realize that gamers-masturbators are not! their only customers.
Nicely phoronix keeps up the good work.
Thanks.

[CochainComplex]

Yes, that's right, it did not look that good in the past.
Hopefully that's not a fluke and @AMD they eventually, really began to realize that gamers-masturbators are not! their only customers.
Nicely phoronix keeps up the good work.
Thanks.

AOCC helps also the gamer masturbators. Ever compiled mesa and the kernel with AOCC? Helps also on a 3600XT

[CochainComplex]

So can AMD backport these improvements to the mainline LLVM/Clang?

Im asking myself if this is per se possible or if they have altered clang so much that its not possible to patch upstream clang without breaking all the other platform support.

[wpupkin]

Correct me if I'm wrong, but:
1. according to GCC manual you should path same compiler flags to LDFLAGS because this is LINK time optimization;
2. especially meson build system need additional attention, because it wants maximum control over compiler flags, it has <language>_flags option for this and it definitely will build each source file separately then link everything together (but ok, let's assume it will use *FLAGS from environment). Also it will force -O3 when --buildtype release used, so if something was tested with such test specifically compiled with -O2 and --buildtype release -- surprise, it was tested with -O3;
3. GCC LTO and Clang LTO are two different things, only common they have is the name "LTO". IDK about Clang LTO, maybe it's safe to path -flto here and there in CFLAGS, but with GCC you'll likely get misoptimized binary.

For example LeelaChessZero strange dip.
dickvsstick.com again?

[Tawheed]

Gcc is so slow its garbage

[skeevy420]

AOCC helps also the gamer masturbators. Ever compiled mesa and the kernel with AOCC? Helps also on a 3600XT

Stay out of my day planner.

[onlyLinuxLuvUBack]

Gcc is so slow its garbage

Michael maybe round2 article could be vlang+v's db server vs zig+sqlite3-amag c src+sqlite3-in-mem-db-mode+static link ? memory db query battle ?