Hacker News Reader

Comment by codr7 | Hacker News Reader

I've written a lot of interpreters in C/++ lately, and spent a lot of time profiling and optimizing them.

I can't remember -O3 ever being faster than -O2, it's usually slightly slower for me.

For interpreters, that's highly likely to be true, especially if they are mainly spending time in switch tables or executing basic opcodes.

Care to explain? -O3 generates larger code than -O2?

Yes, -O3 tends to include a lot of features that increase code size, like aggressive loop unrolling. If you are jumping around a large amount of code, -O3 generally performs more poorly than -O2, but if you are running a tight loop (like HPC code), -O3 is better.

In the past, at a time when I worked on a very performance sensitive codebase that was also limited in scope, we compiled with -Osize and did all the loop optimizations we wanted manually (and with pragmas). That produced faster code than -O2 or -O3.

Regarding unrolling, -O3 contains -funroll-and-jam but not -funroll-loops. You may want one or the other, maybe both, depending on circumstances. I don't see much benefit from the available pragmas on HPC-type code unless for OpenMP, and "omp simd" isn't necessary to get vectorization in the places I've seen people say it is. Mileage always varies somewhat, of course. (Before second-guessing anything, use -fopt-info.)

Modern x86 CPUs have micro instr caches to store small loops (about 50 instr) and medium loops (~2k instr). Also, the bottleneck is usually the instruction decoding (Alder Lake made huge changes on that, so this might change).

In other words, loop unrolling is, more often than not, harmful.

It’s a shame that Osize can sometimes produce truly awful code. There are a few optimisations in there that trade a byte for a massive slowdown.

You asked for minimum size, and that's what you got. I'd say that's working as it should.

A more granular control over optimisation would be good, however.

Probably just some tweaks to O2 would be enough, after all people are selecting Os over O2 because they see better performance, and that should not be happening.

You can enable/disable individual optimizations. How much more granular do you need?

Surely a profile-guided build should be able to only apply -Os to those functions where it doesn't cause a lot of problems.

In the application I referred to, PGO was also used. However, that only applies -Os to cold code, and if what you're doing is very branchy, it can help even in the hot path.

Python, specifically, is supposed to get most benefit from -fno-semantic-interposition

What about when you factor in PGO? (-fprofile-generate & -fprofile-use)

This is the rub. the compiler is flying completely blind without PGO.

I turned on PGO for the D compiler recently and it compiled it's tests 15% faster. Boom.

Haven't really been down that rabbit hole yet, thanks for reminding me.

Reply on news.ycombinator.com