[Vm-dev] Cog Primitive Performance

Ben Coman btc at openinworld.com
Tue Apr 18 06:46:45 UTC 2017


On Tue, Apr 18, 2017 at 7:45 AM, Eliot Miranda <eliot.miranda at gmail.com>
wrote:

>
> Hi All,
>
>     Clément and I have been looking at primitive performance in the
> context of string hash.  We'll say more about string hash in a later
> message (and there is nothing to be alarmed about here).  But it is worth
> describing the performance measurements of the various kinds of primitive
> dispatch in Cog.  The following measurements are collected from a 64-bit
> Squeak trunk image with a 64-bit Cog Spur VM with four different
> implementations of hashMultiply, three of them primitive.  The times are
> collected on a 2.3 GHz Intel Core i7 MacMini.
>
> The doit below compares the time taken to go round a loop containing 10
> unrolled copies of an operation 10 million times, for a total of 100
> million invocations of the operation.  The operations are
> - doing nothing
> - sending yourself
> - invoking the SmallInteger>>hashMultiply method
> - invoking a version of hashMultiply written as a normal interpreter
> primitive (more on this below)
> - invoking a version of hashMultiply written as a C function called from
> machine code
> - invoking a version of hashMultiply written as a machine code primitive
>

Where does lowcode fit into this picture?

cheers -ben
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/vm-dev/attachments/20170418/42dca3f2/attachment-0001.html>


More information about the Vm-dev mailing list