Megalextoria
Retro computing and gaming, sci-fi books, tv and movies and other geeky stuff.

Home » Digital Archaeology » Computer Arcana » Commodore » Commodore 8-bit » DHRYSTONE results for Aztec C65 Commodore 64 and Apple IIe now available
Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend 
Switch to threaded view of this topic Create a new topic Submit Reply
DHRYSTONE results for Aztec C65 Commodore 64 and Apple IIe now available [message #210291] Sun, 15 September 2013 07:22 Go to next message
Bill Buckels is currently offline  Bill Buckels
Messages: 1418
Registered: November 2012
Karma: 0
Senior Member
Verifiable working programs including diskimages and source code now
available:

DHRYSTONE 1.1 for Aztec C65 C II Vers. 1.05h Commodore 64 cross-compiler:

http://www.aztecmuseum.ca/extras/DHRYC64.zip

DHRYSTONE 1.1 for Apple IIe ProDOS 8 Aztec C65 3.2b cross-compiler:

http://www.aztecmuseum.ca/extras/DHRY.zip

The results are as follows:

Commodore 64: 41 DHRYSTONES/SECOND
Apple IIe: 39 DHRYSTONES/SECOND

The DHRYSTONE Version 1.1 that I have used was written in 1984 one year
after C II Vers. 1.05h was released and last updated the same year that
Aztec C65 3.2b was released.

1. So how does this compare to other Commodore 64 C Compilers?

cc65 cross-compiler - 45 DHRYSTONES/SECOND
Power C native mode compiler - 36 DHRYSTONES/SECOND

The Power C 2.8 stats are right out of the DHRYSTONE source code. For cc65
stats:

http://www.cc65.org/mailarchive/2009-07/6616.html

Note: According to DHRYSTONE's author (Rick Richardson), benchmarkers must
turn-off or bypass optimizers which perform more than peephole optimization.
So cc65 stats have been downgraded by the 5 DHRYSTONES/SECOND from 50 to 45
as "verified" in the published stats (see link above) in the cc65 mailing
list to meet DHRYSTONE specifications. To paraphrase Rick, "salesmen like
the higher measure".

I am sketchy on what version of DHRYSTONE was used to obtain the cc65 stats
as well, but I am not overly concerned since cc65's DHRY results are only
marginally faster, and keeping in mind that Aztec C supports both single and
double precision floating point math with a math library included, and other
features like inline assembly this doesn't matter much to me.

AFAIK cc65 never did a comparison to Aztec C65's C64 version despite the
fact that IMO it is more than trivial for any C programmer to do so. This is
a 30 year old compiler, and it is well understood and well documented, comes
with clear examples, runs "right-out of the box", etc. Some excuses were
made but it is a poor workman who blames his tools:

http://www.cc65.org/mailarchive/2009-10/7171.html

I don't stumble through documentation either; I take time to read it. I must
be getting old, because I feel like saying "If it was good enough for
Grandpa, it's good enough for me". As for memcpy, I guess I could have used
the Aztec C equivalent but didn't need it. I used an appropriate version of
DHRY...

I am in the process of porting the older CII compiler to Aztec CG65 3.2c,
and when done I would expect almost equivalent DHRY on the C64 with either
compiler.

2. So how does this compare to other Apple IIe C compilers?

I am speculating at this point because I have no stats for cc65 for the
Apple II (just for the C64 sort-of) but if my theory is correct, and the
same ratio of 37 Apple II to 41 C64 applies to cc65, then cc65 should
come-in at 40 DHRYSTONES/SEC on the Apple II as opposed to Aztec C65 3.2b's
39 DHRYSTONES/SEC. I'll need to leave it to the cc65 people to sort that one
out. Without verifiable results for cc65, it's hard to tell.

I am comparing Apples to Bananas, so I am going to have to stick with my
theory and regard Aztec C 3.2b and cc65 as almost equivalent DHRYSTONE on
the Apple IIe unless and until verifiable results can be produced making an
honest effort to definitely prove otherwise.

With Apple IIe Aztec C65's support for a Unix like shell and command line
programs and file I/O for both DOS 3.3 and ProDOS 8, inline assembly,
floating point, ProDOS SYS programs and no need of a launcher or some other
contrivance, clearly documented overlay support etc. there seems no
compelling reason not to use Aztec C65 on the Apple IIe unless religious
fervor or some other reason is taken into account.

Best Regards,

Bill Buckels
Re: DHRYSTONE results for Aztec C65 Commodore 64 and Apple IIe now available [message #210296 is a reply to message #210291] Tue, 17 September 2013 10:57 Go to previous messageGo to next message
Bill Buckels is currently offline  Bill Buckels
Messages: 1418
Registered: November 2012
Karma: 0
Senior Member
"Bill Buckels" <bbuckels@mts.net> wrote:

> there seems no compelling reason not to use Aztec C65 on the Apple IIe
> unless religious fervor or some other reason is taken into account.

I should add that Aztec C65 C II Vers. 1.05h Commodore 64 cross-compiler
which is the compiler distributed from the Aztec C65 website (Aztec64) runs
fine in Windows XP, and ran fine in DOSBox until Windows 7 when it became
broken. I suspect that it probably runs fine in Linux under DOSEmu but I
have not taken time to do tests.

In Windows 7 Aztec C65 C II Vers. 1.05h runs fine in VirtualBOX and FreeDOS,
but the VirtualBOX extensions for drive sharing seem to be unconfigurable
for Windows, and FTP is reported to be needed to port the compiler to
VirtualBOX which seems to be one pile of trouble. In my opinion VirtualBOX
seems to cater more to the Linux crowd so it doesn't surprise me that
support for Windows drive sharing seems not to work anymore.

Never mind that. I am in the process of attempting to port the Aztec C65 C
II Vers. 1.05h C64 libraries to the CG65 3.2c compiler and will report my
progress as I get further along. The 3.2c compiler, like the 3.2b compiler
works fine under DOSBox for Windows 7, and also in DOSEmu under Linux
(tested in Ubuntu).

That's all for now.

Bill
Re: DHRYSTONE results for Aztec C65 Commodore 64 and Apple IIe now available [message #210308 is a reply to message #210291] Sat, 21 September 2013 08:35 Go to previous messageGo to next message
&lt;address_is is currently offline  &lt;address_is
Messages: 19
Registered: January 2013
Karma: 0
Junior Member
"Bill Buckels" <bbuckels@mts.net> wrote:
> Verifiable working programs including diskimages and source code now
> available:
>
> DHRYSTONE 1.1 for Aztec C65 C II Vers. 1.05h Commodore 64 cross-compiler:
>
> http://www.aztecmuseum.ca/extras/DHRYC64.zip
>
> DHRYSTONE 1.1 for Apple IIe ProDOS 8 Aztec C65 3.2b cross-compiler:
>
> http://www.aztecmuseum.ca/extras/DHRY.zip
>
> The results are as follows:
>
> Commodore 64: 41 DHRYSTONES/SECOND
> Apple IIe: 39 DHRYSTONES/SECOND
>
> The DHRYSTONE Version 1.1 that I have used was written in 1984 one year
> after C II Vers. 1.05h was released and last updated the same year that
> Aztec C65 3.2b was released.
>
> 1. So how does this compare to other Commodore 64 C Compilers?
>
> cc65 cross-compiler - 45 DHRYSTONES/SECOND
> Power C native mode compiler - 36 DHRYSTONES/SECOND
>
> The Power C 2.8 stats are right out of the DHRYSTONE source code. For cc65
> stats:
>
> http://www.cc65.org/mailarchive/2009-07/6616.html
>
> Note: According to DHRYSTONE's author (Rick Richardson), benchmarkers must
> turn-off or bypass optimizers which perform more than peephole optimization.
> So cc65 stats have been downgraded by the 5 DHRYSTONES/SECOND from 50 to 45
> as "verified" in the published stats (see link above) in the cc65 mailing
> list to meet DHRYSTONE specifications. To paraphrase Rick, "salesmen like
> the higher measure".
>
> I am sketchy on what version of DHRYSTONE was used to obtain the cc65 stats
> as well, but I am not overly concerned since cc65's DHRY results are only
> marginally faster, and keeping in mind that Aztec C supports both single and
> double precision floating point math with a math library included, and other
> features like inline assembly this doesn't matter much to me.
>
> AFAIK cc65 never did a comparison to Aztec C65's C64 version despite the
> fact that IMO it is more than trivial for any C programmer to do so. This is
> a 30 year old compiler, and it is well understood and well documented, comes
> with clear examples, runs "right-out of the box", etc. Some excuses were
> made but it is a poor workman who blames his tools:
>
> http://www.cc65.org/mailarchive/2009-10/7171.html

I believe you could post the above to the cc65 mailing list too, as a
follow-up to the mentioned thread.

--
SD!
Re: DHRYSTONE results for Aztec C65 Commodore 64 and Apple IIe now available [message #210309 is a reply to message #210308] Sat, 21 September 2013 19:57 Go to previous messageGo to next message
Bill Buckels is currently offline  Bill Buckels
Messages: 1418
Registered: November 2012
Karma: 0
Senior Member
<address_is@invalid.invalid> wrote:

> I believe you could post the above to the cc65 mailing list too, as a
> follow-up to the mentioned thread.

These results have been posted on csa2 and here. They will be posted on the
Aztec C website with other Aztec C DHRY results for other processors and
platforms.

I believe that is adequate for someone with my age and experience, and for
something as old and experienced as Aztec C.

Regards,

Bill
Re: DHRYSTONE results for Aztec C65 Commodore 64 and Apple IIe now available [message #210314 is a reply to message #210308] Sun, 22 September 2013 16:22 Go to previous messageGo to next message
Groepaz is currently offline  Groepaz
Messages: 640
Registered: December 2011
Karma: 0
Senior Member
address_is@invalid.invalid wrote:

>> I am sketchy on what version of DHRYSTONE was used to obtain the cc65
>> stats as well

a variant of the 1.1 source, slightly touched up to compile cleanly in a non
K&R environment. basically exactly the same that was used here too, minus
some small changes that wouldnt affect the result of the tests.

> I believe you could post the above to the cc65 mailing list too, as a
> follow-up to the mentioned thread.

maybe also run the program through a recent cc65 for some surprises:

using -Osir (all optimizations) 73 dhrystones, 6723 bytes
using -O (only basic peephole) 65 dhrystones, 6224 bytes
no optimizations 46 dhrystones, 6754 bytes

vs
48 dhrystones, 22587 bytes

http://hitmen.c02.at/temp/testsuite/dhrystone.zip (to compile use cl65 -
DNO_OLD_FUNC_DECL [-Osir] -o test.prg dstone.c)

so basically aztec c is about as fast as cc65 (provided you disable even the
dumbest optimizations), but produces a binary that is like 3 times as
big.... It'd be interesting how these figures change when you enable all
possible optimizations in aztec c :)

--

http://www.hitmen-console.org http://magicdisk.untergrund.net
http://www.pokefinder.org http://ftp.pokefinder.org

I should warn you that underneath these clothes I'm wearing boxer shorts and
I know how to use them.
<Robert Orben>
Re: DHRYSTONE results for Aztec C65 Commodore 64 and Apple IIe now available [message #210315 is a reply to message #210314] Sun, 22 September 2013 21:00 Go to previous message
Bill Buckels is currently offline  Bill Buckels
Messages: 1418
Registered: November 2012
Karma: 0
Senior Member
"Groepaz" <groepaz@gmx.net> wrote:
> a variant of the 1.1 source, slightly touched up to compile cleanly in a
> non K&R environment. basically exactly the same that was used here too,
> minus some small changes that wouldnt affect the result of the tests.

> http://hitmen.c02.at/temp/testsuite/dhrystone.zip

Thanks.

> so basically aztec c is about as fast as cc65 (provided you disable even
> the dumbest optimizations), but produces a binary that is like 3 times as
> big.... It'd be interesting how these figures change when you enable all
> possible optimizations in aztec c :)

All kidding aside, writing an optimizer for Aztec C65 would be more than a
Winter Project.... I think it would be more interesting to port the older
Aztec CII compiler libraries to the Aztec CG65 3.2c cross-compiler.

Bill
  Switch to threaded view of this topic Create a new topic Submit Reply
Previous Topic: CBM 128D as a terminal in CP/M mode?
Next Topic: commodore 64/128
Goto Forum:
  

-=] Back to Top [=-
[ Syndicate this forum (XML) ] [ RSS ] [ PDF ]

Current Time: Fri Mar 29 02:50:47 EDT 2024

Total time taken to generate the page: 0.03305 seconds