Whats the fastest cpu (per $) regarding golang compiler performance?

[slinso]

Hi all,

I'm interested in the compiler performance of different cpus. How much impact has the number of cores? Or is higher frequency with a few cores better regarding compile time / $.

Maybe someone already owns a new ryzen cpu. Is it worth buying a highend Intel cpu (i7 or xeon) for 1000$ or is the highend AMD for 500$ maybe even better for go.

My C++ projects would benefit the most from more cores. But that's something different because we are talking about 10 minutes or more compile time.

I think using Dave Cheneys compiler benchmarks (benchkube) and (benchjuju)  should provide solid base numbers for the performance. Of course smaller projects could behave differently. But if a small project compiles in 1s or 2s doesn't matter that much to the development cycle.

I have not moved the codebase to a ramdisk, because i would like to see numbers from normal dev setups.

go version go1.8.1 linux/amd64

cat /proc/cpuinfo | head | grep "model name"
model name      : Intel(R) Core(TM) i7-6700T CPU @ 2.80GHz

RAM: 16GB

Disk: Samsung 850 Evo

avg. runtime:

benchjuju: 23.7s

benchkube: 21.3s 

[Marvin Stenger]

Something you have to consider, this will probably make it into go 1.9:
https://go-review.googlesource.com/?polygerrit=0#/c/40693/

[Jesper Louis Andersen]

My intuitive hunch is that memory speed is going to affect compilation speeds a lot as well, so that might be factored into the question as well. Adding cores has no value unless you can keep them well fed with input they can compile.

--
You received this message because you are subscribed to the Google Groups "golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[slinso]

A lot of factors will affect compilation speeds. And scaling won't be linear by adding more cores, I know, but how well does the compiler already uses the available cores. How well does the compiler scale? The mentioned CL above will provide some nice performance increasements for multicore systems.

Without digging too deep I just wanted to know how the compilation speeds changes on different cpus.