[erlang-questions] Erlang/OTP 13A and Darwin
Wed Mar 25 12:42:09 CET 2009
I assume what they are getting at is an emulator data representation issue. Too large integers overflow the fixnum format (28 bits on a 32-bit host, I believe), which means falling back to a less efficient representation and more complex arithmetic. 64-bit words mean you can represent about 60 bits or so as "immediates", so operations on 32-bit quantities, in particular, go faster and can be represented more compactly.
This should, at a guess, be handled fairly well by just an appropriate port.
From: Camille Troillard <>
To: Per Gustafsson <>
Sent: Wednesday, March 25, 2009 11:59:00 AM
Subject: Re: [erlang-questions] Erlang/OTP 13A and Darwin
Thanks Ulf and Per,
On Wed, Mar 25, 2009 at 11:42 AM, Per Gustafsson <> wrote:
A type of calculation that becomes much more efficient on 64-bit machines are 32-bit fixed integer calculations,
That is very interesting, would that mean that we could have faster arithmetic in Erlang without too much effort? I was looking for better math performance. What about 32 bits builds? Are there some possible (planned?) optimisations regarding this domain?
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the erlang-questions