| From: | mef |
| Date: | 5 Sep 2000 at 22:47:56 |
| Subject: | RE: AMIOPEN: has anyone run SPEC benchmark using VP? |
Apropo benchmarking. I located the following article on Microsoft's website
that states its "common executable format" (Microsoft's equivalent to Tao's
VP) is 3-20% slower than when compiling the same benchmarks directly to the
native processor. Further, translating code compiled to CEF to its native
format is up to 20% larger when compared to compiling directly to the native
processor using their visual C/C++ compiler. Not sure what benchmarks
Microsoft used.
Unfortunately, I cannot locate any documentation that specifies Microsoft's
CEF format in order to compare it to Tao's VP.
The link is:
http://www.microsoft.com/windows/embedded/ce/
developer/applications/appdevelopment/cef2.asp
So I am getting more curious how Tao's VP compares in terms of execution
speed and code footprint. Hopefully it is in the same ballpark as
Microsoft's technology, and ideally it is better for the same benchmarks.
Marc
>-----Original Message-----
>From: Marc Fiuczynski
>Sent: Friday, September 01, 2000 11:41 AM
>To: open@amiga.com
>Subject: RE: AMIOPEN: has anyone run SPEC benchmark using VP?
>
>
>Matthew (and everyone else interested in benchmarking)...
>
>It just occurred to me that another way to evaluate how well the VP
>compiler/translator does versus compiling natively is the Dhrystone 2.1
>benchmark. (Note to the cynics: the Dhrystone benchmark isn't
>really used
>anymore to compare different processors, because it has a small memory
>footprint and does silly things like strcpy/strcmp. However,
>people in the
>embedded industry still use it, because there currently isn't
>anything much
>better.)
>
>Go to any of your favorite search engines and you'll find
>references to its
>source code all over the place. The porting effort primarily
>would be to
>add timing support (presumably using the Amiga SDKs
>microtimer), so that it
>can measure Dhrystone MIPS. In fact, most versions of the Dhrystone
>benchmark already have support for the old Amiga. :)
>
>The end comparison would be to see how well the Dhrystone
>benchmark runs on
>the Amiga SDK versus natively on your linux box.
>
>Cheers,
>Marc
>
>
>>-----Original Message-----
>>From: Marc Fiuczynski
>>Sent: Thursday, August 31, 2000 4:31 PM
>>To: open@amiga.com
>>Subject: RE: AMIOPEN: has anyone run SPEC benchmark using VP?
>>
>>
>>The URL is www.spec.org. Unfortunately, this particular
>>benchmark isn't
>>free. In fact is costs anywhere from $300 - $600. It is,
>however, the
>>benchmark that all processor manufacturers use to determine a relative
>>performance rating. Luckily, my University has a copy of this
>>benchmark.
>>I'll get to it when I have more time.
>>
>>Cheers,
>>Marc
>>
>>>-----Original Message-----
>>>From: Matthew J Fletcher [mailto:amimjf@connectfree.co.uk]
>>>Sent: Thursday, August 31, 2000 1:33 PM
>>>To: open@amiga.com
>>>Subject: RE: AMIOPEN: has anyone run SPEC benchmark using VP?
>>>
>>>
>>>Hi,.
>>>
>>>
>>>> >Nothing or noone prevents you from porting any benchmarks, anyway.
>>>>
>>>> I am a lazy boy. Can someone else do it please. :) No,
>>>honestly, I'll get
>>>> to this sometime in October or November (after I finish
>>writing up my
>>>> dissertation). However, if someone feels the urge, they
>>>should go for it.
>>>
>>>URL`s ??
>>>
>>>If its nothing to involved (like asm), or threads.
>>>
>>>regards,
>>>
>>>---
>>>Matthew J Fletcher amimjf@connectfree.co.uk
>>>Serck Controls Ltd NPD Firmware
>>>www.amimjf.connectfree.co.ukICQ amimjf 44193496
>>>Please Use PGP !!! Key Available On Request
>>>
>>>
>>>Subscribe/Unsubscribe: open-request@amiga.com
>>>Amiga FAQ: http://www.amiga.com/faq.html
>>>
>>Subscribe/Unsubscribe: open-request@amiga.com
>>Amiga FAQ: http://www.amiga.com/faq.html
>>
>Subscribe/Unsubscribe: open-request@amiga.com
>Amiga FAQ: http://www.amiga.com/faq.html
>
Subscribe/Unsubscribe: open-request@amiga.com
Amiga FAQ: http://www.amiga.com/faq.html