On: ErLLVM: LLVM backend for high performance Erlang

[kalleboo]

> Performance/$ is still going to rise

But is performance/core also going to keep rising? Or do we need smart compilers/languages to make use of all the cheap cores?

[lucian1900 → kalleboo]

We need a correct, expressive language that easily lets you makes use of all those cheap cores. Sort of like Erlang.

[duaneb → kalleboo]

I mean, this conversation has been around for easily 10 years. We live in a world where we already have to do this—look at all the progress in the last decade on the prevalence of futures, async i/o, and channels/queues/whatever. Memory bandwidth and latency has proven to be far more of a performance drag that has improved far slower—the likely areas that will lead to performance problems will be in areas other than single-core clock speed.

[eternalban → duaneb]

Well, that -- the Memory bottleneck -- is precisely why we need a rethink of platforms and tool chains.

Languages will provide the semantics (in context of memory hierarchy) of data locality, and compilers that optimize for that.

[nop edit]

[drkrab → eternalban]

As it turns out, Erlang is already providing much better memory locality than your average language. Each process runs in a contiguous memory area (heap+stack) which fits perfectly with a many-core world.

[duaneb → eternalban]

> is precisely why we need a rethink of platforms and tool chains.

My point was, people are already thinking. It's not an easy problem and there's no obvious way forward.