Golang SSA performance numbers

https://groups.google.com/forum/#!topic/golang-dev/m1r8kSle30Y

by signa11 • 10 years ago

140 26 10 years ago

groups.google.com

The last thing I'm curious about is compilation time, especially since that's taken a hit or two recently. I know it's easily a net win with numbers like that, but one of my favorite parts of Go is the quick build times, and thus quick iteration times.

Also, since Go is self-compiled, does that mean we'll have to wait until 1.7.1 for the speed improvements to hit the compiler? :-P

Spiritus2• 10 years ago

I've seen the numbers ~10% slower floating around.

UniQP3• 10 years ago

[citation needed]

UniQP3• 10 years ago

I'm sure, you can also self-compile the tip. Would be really interesting if this was done for benchmarks.

bpineau• 10 years ago

The linked post gives an estimation :

"Tip compiler (with SSA internal checks off) is about 7% slower than go1.6 to compile net/http (go test -a -c -gcflags=-d=ssa/check/off net/http)"

... though we may wonder wether the mesured compiler itself was compiled with or without SSA (with 7% being in the ballpark of the expected SSA gains).

davecheney4• 10 years ago

The compiler always compiles itself as part of the bootstrap process. The final compiler is built with itself and thus uses its own SSA code generation.

pjmlp2• 10 years ago

No worries, it will never reach the C and C++ build times which are often measured in minutes and hours.

UniQP3• 10 years ago

Compile time is an issue of C++ but not of C.

davecheney4• 10 years ago

Also, since Go is self-compiled, does that mean we'll have to wait until 1.7.1 for the speed improvements to hit the compiler? :-P

No. The compiler compiles itself during the bootstrap phase.

thethimble• 10 years ago

Does the old compiler compile the new compiler after which the new compiler can compile itself? Bootstrapping is incredible.

wtbob• 10 years ago

one of my favorite parts of Go is the quick build times

One of my favourite parts of Go was the quick build times. It used to feel like 'go build' was as fast as 'python'; sadly, that's no longer the case.

jerf• 10 years ago

Despite the attractive nuisance of "go build", if you're in a fast save-compile-test cycle, you really want "go install". Both do pretty much the same thing (ignoring where the result goes), but install keeps the intermediate package compilations around and does the minimum, whereas build seems to discard them. Put "-v" on both of them and you can see the differences.

dilap• 10 years ago

Yeah, it got about twice as slow going 1.4 => 1.5, which makes it just slow enough to be annoying.

(It's still much faster than something like C++ though, and feels about the same to me as the (snappy) Objective-C.)

andrewflnr• 10 years ago

Can anyone say why the Mandelbrot benchmark, out of all of them, was slower with SSA? Any chance it was just a fluke?

davecheney4• 10 years ago

The inner loop went from 15 instructions to 16 because the register optimiser is not finished yet.

rurban• 10 years ago

Thanks. That's what I asked on the list, but HN was faster :)

vessenes3• 10 years ago

I just posted some results there on our own codebase.

Server load time dropped from 25.5s or so to 22.2s; a really nice win on a large and varied codebase for these improvements. I can't imagine not wanting them.

Golang SSA performance numbers

Related Stories