Real Benchmarks are quite difficult to do & require tools like Intels VTune
( for intel proc. of course )
TinyBenchmarks focuses on 2 things - testing speed of send of a single
message ( benchFib ) & measuring bytecodes / sec using calculation of a
"bytecode heavy benchmark ( IMO it was counting number of primes )
the 26 benchFib takes ~ 5000 us with Pharo 8 on my laptop same pure C on
msvc 2019 optimized code 32 bit ( taking argument in a register instead of
getting it from stack takes ~1500 us )
same C code is 2 times slower compiled in x64 - I can post the C code if
anyone is interested
Today Smalltalk & other dynamic langs reaches pritty close C speed.
I am in business optimizing my VM since decades & JITed languages have the
big advantage that they can measure out the underlying hardware & emitting
quite different code .
different Assembler.
Frank
-----Ursprüngliche Nachricht-----
Von: Squeak-dev [mailto:
[hidden email]] Im
Auftrag von Yoshiki Ohshima
Gesendet: Freitag, 26. Juli 2019 21:53
An: The general-purpose Squeak developers list
Betreff: Re: [squeak-dev] Dorados (was: Efficient Implementation of
theSmalltalk-80 System: what was the hardware?)
Ah, speaking of which, I remember adapting The green book benchmarks to
Squeak while ago. Ill try to find it (it might be on SqueakMap, possibly.)
-- Yoshiki
> On Jul 26, 2019, at 11:48 AM, tim Rowledge <
[hidden email]> wrote:
>
>
>
>> On 2019-07-26, at 8:38 AM, Bert Freudenberg <
[hidden email]> wrote:
>>
>> Somebody help me with the Dorado math here interpreting Squeak on
JavaScript in a browser on my *phone*.
>
> I don't think the tinyBenchmarks is really good enough here; run the
proper old benchmark suite.
>
> tim
> --
> tim Rowledge;
[hidden email];
http://www.rowledge.org/tim Useful
> Latin Phrases:- Fac me cocleario vomere! = Gag me with a spoon!
>
>
>
Locked
