Discussion:
8080 exerciser
(too old to reply)
IanB
2009-09-22 19:45:22 UTC
Permalink
Raw Message
Back in February I posted here asking anyone who had a CP/M 2.2 system
using an 8080 CPU to try out a short program for me. It was a port of
Frank Cringle's ZEXLAX (aka ZEXDOC and ZEXALL) Z80 exerciser that
tested an 8080 CPU instead.

The purpose was to get some numbers from genuine hardware so that I,
or anyone else, could check that a 8080 emulator was performing
correctly: exactly emulating the operation of an 8080.

Well, 7 months later I've just got my first response, and my web page
has been updated with the results provided by somebody (thanks again
Alexander) using a Russian KR580VM80A 8080 clone.

I would still like to get the results for a "genuine" Intel 8080 CPU,
or any of the other legitimate clones, so if anyone with the
appropriate hardware can spare half an hour or so to run the
application it would be much appreciated.

More details at :

http://www.sunhillow.eu/8080exerciser

NB. If you have written an emulator and, when testing it using the
exerciser on the web site, it mysteriously fails on the 8 bit aluop
tests and all the dcr tests then send me an email. Thanks to some
detective work by some Russian Vector-06c users I can give you one
place to start looking.

Ian
--
The From address is valid
Mr Emmanuel Roche, France
2009-09-22 20:18:08 UTC
Permalink
Raw Message
Hello, Ian!
I am sorry, but I am obliged to report that I was barely able to read,
with difficulty, your 2 Web pages. You must be younger than me. I have
absolutely no trouble with green chars on black background: this is my
favorite setting (it was funny to see that you enclosed pictures
containing... text!). However, I found it almost impossible to read
your black chars on dark blue background. Apparently, it is my
experience that Windows people expect text to appear with black
characters on a white background. At the time of the "IBM Clown", only
the "Mac" was doing that... It is funny how conventions change, over
time. But, if you want your Web pages to be read by others, I suggest
making them easier to read. I wonder how they appear with the LYNX
browser?

Yours Sincerely,
Mr. Emmanuel Roche, France
IanB
2009-09-22 20:42:18 UTC
Permalink
Raw Message
On Tue, 22 Sep 2009 13:18:08 -0700 (PDT), "Mr Emmanuel Roche, France"
Post by Mr Emmanuel Roche, France
However, I found it almost impossible to read
your black chars on dark blue background.
Strange. The web page specified black text on a light gray
background...

<body bgcolor="lightgray" text="#000000" link="#0000ff"
vlink="#800080" alink="#ff0000">

No idea why your browser couldn't cope with that. I'll change it
though.
Post by Mr Emmanuel Roche, France
(it was funny to see that you enclosed pictures
containing... text!).
They were screen shots from the 8080 emulator's terminal - safer than
typing them in again.


Ian
--
The From address is valid
Jason King
2009-09-23 01:34:41 UTC
Permalink
Raw Message
Post by IanB
Post by Mr Emmanuel Roche, France
However, I found it almost impossible to read
your black chars on dark blue background.
Strange. The web page specified black text on a light gray
background...
<body bgcolor="lightgray" text="#000000" link="#0000ff"
vlink="#800080" alink="#ff0000">
On Firefox 3.5.3 under Vista things appear as you'd expect.
My 49 yo eyes are probably close in mileage to M. Roche's.
M. Roche, you might want to report browser and OS to help diagnose the
problem.
If I had a real 8080 I'd test it. Closest I've got is memories of a
trip to the Intel museum where I saw one of those and a 4004.
Mr Emmanuel Roche, France
2009-09-23 06:42:53 UTC
Permalink
Raw Message
Hello, Jason and Ian!
Post by Jason King
Post by Mr Emmanuel Roche, France
However, I found it almost impossible to read
your black chars on dark blue background.
Strange.  The web page specified black text on a light gray
background...
Even stranger: Upon reading your answer (I always start a Windows
session by reading the comp.os.cpm Newsgroup), I had another look to
your Web pages... and they displayed as you say!
Post by Jason King
On Firefox 3.5.3 under Vista things appear as you'd expect.
My 49 yo eyes are probably close in mileage to M. Roche's.
M. Roche, you might want to report browser and OS to help diagnose the
problem.
Internet Explorer 6 under Windows 98SE.

I don't know what happened, yesterday.

I am busy, those days, so did not investigate SCELBAL. As usual as
that time, the source code you posted contains precious few comments.
Was there more information about the language and its implementation
in the books (I don't have them: just the DDJ articles)?

Question: Why are your dumps twice wider than usual? (Several people
complained because I published 78-characters wide articles...)

Yours Sincerely,
Mr. Emmanuel Roche, France
IanB
2009-09-23 07:33:24 UTC
Permalink
Raw Message
On Tue, 22 Sep 2009 23:42:53 -0700 (PDT), "Mr Emmanuel Roche, France"
Post by Mr Emmanuel Roche, France
Even stranger: Upon reading your answer (I always start a Windows
session by reading the comp.os.cpm Newsgroup), I had another look to
your Web pages... and they displayed as you say!
After your observation I changed the source page source from
bgcolor="lightgray" to bgcolor="#e0e0e0", which your browser seems to
accept?. The original version went through the W3 validator though,
so it should have displayed correctly in all compliant browsers.
Post by Mr Emmanuel Roche, France
I am busy, those days, so did not investigate SCELBAL. As usual as
that time, the source code you posted contains precious few comments.
Was there more information about the language and its implementation
in the books (I don't have them: just the DDJ articles)?
There are copious source comments (along with a lot of descriptive
text) in the book, but I didn't bother typing them in. It would have
taken at least ten times longer.

If you are interested then there is a web site that has the source +
comments, and, I think, one that has images of the pages from the
book. I haven't got the urls to hand at the moment, I'll see if I can
find them.
Post by Mr Emmanuel Roche, France
Question: Why are your dumps twice wider than usual?
They are exactly the same size as my other dumps, not unusual at all
:-)

I just wanted to provide a copy of the object code produced by the
assembler. I have found such things useful in the past, for
comparison purposes, when I have had to assemble source using a
different assembler.

Ian
--
The From address is valid
glen herrmannsfeldt
2009-09-23 07:51:29 UTC
Permalink
Raw Message
IanB <***@sunhillow.eu> wrote:
(snip)

< After your observation I changed the source page source from
< bgcolor="lightgray" to bgcolor="#e0e0e0", which your browser seems to
< accept?. The original version went through the W3 validator though,
< so it should have displayed correctly in all compliant browsers.

I haven't thought about this for a while, but I believe in 8 bit
(256 color) mode the browser will choose the color closest to
the one specified. It is possible that is dark blue when gray
was specified.

-- glen
Mr Emmanuel Roche, France
2009-09-23 10:12:06 UTC
Permalink
Raw Message
Hello, Ian!
Post by IanB
There are copious source comments (along with a lot of descriptive
text)  in the book, but I didn't bother typing them in.  It would have
taken at least ten times longer.
Hahaha! This is a way to spend those long Winter nights... This is how
I retyped all my texts. I always retype all the doc that I managed to
find, before recreating the source code.
Post by IanB
If you are interested then there is a web site that has the source +
comments, and, I think, one that has images of the pages from the
book.  I haven't got the urls to hand at the moment, I'll see if I can
find them.
I just noticed that, because I think there is not enough doc/links on
your SCELBAL Web page. I remember it, but a Newbie would not have this
background.
Post by IanB
Post by Mr Emmanuel Roche, France
Question: Why are your dumps twice wider than usual?
They are exactly the same size as my other dumps, not unusual at all
:-)
Really? In the CP/M world, 16 bytes has been the standard, since (at
least) PL/M, that is to say: 1972. I think that it was due to the
ASR-33 Teletype width (72 columns).
Post by IanB
I just wanted to provide a copy of the object code produced by the
assembler.  I have found such things useful in the past, for
comparison purposes, when I have had to assemble source using a
different assembler.
Even before CP/M, Gary Kildall devised such a thing. It is now known
as the "Intel HEX File Format". I have published a SAVEHEX.BAS and
LOADHEX.BAS showing, in addition, the bytes processed as a (standard)
dump on screen. I probably gave some background info (as I always do)
in the message. (ASM and MAC produce HEX files as output, so
everything would be "standard".)

Curiously, since I am a fan of the 8080, I never had the idea of
making an "exerciser". Me, the programming problem that I would like
to finish is how to write in as little as possible a program able to
display the 916 opcodes of the Z-80 CPU. Maybe this Winter?

Yours Sincerely,
Mr. Emmanuel Roche, France
Jim Higgins
2009-09-23 17:37:33 UTC
Permalink
Raw Message
Post by IanB
On Tue, 22 Sep 2009 23:42:53 -0700 (PDT), "Mr Emmanuel Roche, France"
Post by Mr Emmanuel Roche, France
Even stranger: Upon reading your answer (I always start a Windows
session by reading the comp.os.cpm Newsgroup), I had another look to
your Web pages... and they displayed as you say!
After your observation I changed the source page source from
bgcolor="lightgray" to bgcolor="#e0e0e0", which your browser seems to
accept?. The original version went through the W3 validator though,
so it should have displayed correctly in all compliant browsers.
The WC3 validator is broken when it comes to validating BGCOLOR
attributes. It will validate "asdfgh" as a named color and "#asdfgh"
as a hex value. But omit one or both of the quotes and it will
complain about missing quotes.
Axel Berger
2009-09-24 17:44:00 UTC
Permalink
Raw Message
Post by Jim Higgins
The WC3 validator is broken when it comes to validating BGCOLOR
attributes.
I don't know. I suppose one might make the case, that the actual colour
specified is part of the content and not part of the markup. W3C does
not check the content of pages for nonsense.
Jim Higgins
2009-09-27 13:59:44 UTC
Permalink
Raw Message
Post by Axel Berger
Post by Jim Higgins
The WC3 validator is broken when it comes to validating BGCOLOR
attributes.
I don't know. I suppose one might make the case, that the actual colour
specified is part of the content and not part of the markup. W3C does
not check the content of pages for nonsense.
I will concede your point when it comes to color names because the
HTML specification defines color data as being hex values and it is up
to the browser maker to cause the browser to internally substitute
proper hex values for whatever color names that browser maker chooses
to recognize. Thus bgcolor="asdfgh" is nonsense, but I can agree the
validator is not obliged to check for that.

But in the case of bgcolor="#ASDFGH" the specification states that the
portion after the # must consist of hex digits, so the validator
should not accept "#ASDFGH" as valid because those are not hex digits.
It accepts "#ASDFGH."
dott.Piergiorgio
2009-09-23 07:17:48 UTC
Permalink
Raw Message
Post by IanB
On Tue, 22 Sep 2009 13:18:08 -0700 (PDT), "Mr Emmanuel Roche, France"
Post by Mr Emmanuel Roche, France
However, I found it almost impossible to read
your black chars on dark blue background.
Strange. The web page specified black text on a light gray
background...
<body bgcolor="lightgray" text="#000000" link="#0000ff"
vlink="#800080" alink="#ff0000">
Actually, I have in bgcolor a triplet instead of a named colour, as
showed in my previous post....

Best regards from Italy,
Dott. Piergiorgio.
Jim Higgins
2009-09-23 17:28:13 UTC
Permalink
Raw Message
Post by IanB
On Tue, 22 Sep 2009 13:18:08 -0700 (PDT), "Mr Emmanuel Roche, France"
Post by Mr Emmanuel Roche, France
However, I found it almost impossible to read
your black chars on dark blue background.
Strange. The web page specified black text on a light gray
background...
<body bgcolor="lightgray" text="#000000" link="#0000ff"
vlink="#800080" alink="#ff0000">
If you're going to use color names instead of hex values, the above
BGCOLOR should be spelled "lightgrey," which gives you the color
equivalent to hex "D3D3D3."

"Lightgray" is invalid and IE6/7 renders it as dark blue. FF renders
it the same as "lightgrey." Too bad because I thought FF didn't
perpetuate this sort of thing.

A full list of HTML color names is here
http://www.w3schools.com/html/html_colornames.asp

Probably better in the long run to migrate to using CSS since
"bgcolor" for body and table tags has been deprecated.
IanB
2009-09-23 18:26:37 UTC
Permalink
Raw Message
On Wed, 23 Sep 2009 17:28:13 +0000, Jim Higgins
Post by Jim Higgins
"Lightgray" is invalid and IE6/7 renders it as dark blue. FF renders
it the same as "lightgrey."
Thanks for the explanation, and the information about WC3 not checking
the tag correctly. I assumed that as it passed the validator and
displayed correctly in IE8 that all was OK.... Oh well :-{
Post by Jim Higgins
Probably better in the long run to migrate to using CSS since
"bgcolor" for body and table tags has been deprecated.
Probably. CSS isn't something I've looked at, I don't create many web
pages (obviously!) and they are invariably pretty simple, but I
suppose I ought to get a bit more up to date.

Ian
--
The From address is valid
dott.Piergiorgio
2009-09-23 07:14:57 UTC
Permalink
Raw Message
Post by Mr Emmanuel Roche, France
Hello, Ian!
I am sorry, but I am obliged to report that I was barely able to read,
with difficulty, your 2 Web pages. You must be younger than me. I have
absolutely no trouble with green chars on black background: this is my
favorite setting (it was funny to see that you enclosed pictures
containing... text!). However, I found it almost impossible to read
your black chars on dark blue background.
I don't known what site you looked, but I see a medium gray background;
the HTML line in question is <body bgcolor="#e0e0e0" text="#000000"
link="#0000ff" vlink="#800080" alink="#ff0000">

(Apologies to Mr. "IanB" for the copy&paste of his code, but, at least
the colour triplets ends, at least in binary environment the, millennial
issue "de coloribus non est disputandus")

Best regards from Italy,
Dott. Piergiorgio.
glen herrmannsfeldt
2009-09-23 02:07:21 UTC
Permalink
Raw Message
IanB <***@sunhillow.eu> wrote:
(snip)

< Well, 7 months later I've just got my first response, and my web page
< has been updated with the results provided by somebody (thanks again
< Alexander) using a Russian KR580VM80A 8080 clone.

< I would still like to get the results for a "genuine" Intel 8080 CPU,
< or any of the other legitimate clones, so if anyone with the
< appropriate hardware can spare half an hour or so to run the
< application it would be much appreciated.

If I remember it right, the Russian 8080 is made from photographs
of the chip, and then masks generated from those.

Is there only one Intel 8080A mask step level, or more than one?

If more, then you would want to be sure to test each one,
though most often the new mask was just a shrink of the
previous one. I believe the Russians had to increase the
size and decrease the speed to make theirs work.

-- glen
A***@localhost.net
2009-09-23 02:31:31 UTC
Permalink
Raw Message
On Wed, 23 Sep 2009 02:07:21 +0000 (UTC), glen herrmannsfeldt
Post by glen herrmannsfeldt
(snip)
< Well, 7 months later I've just got my first response, and my web page
< has been updated with the results provided by somebody (thanks again
< Alexander) using a Russian KR580VM80A 8080 clone.
< I would still like to get the results for a "genuine" Intel 8080 CPU,
< or any of the other legitimate clones, so if anyone with the
< appropriate hardware can spare half an hour or so to run the
< application it would be much appreciated.
If I remember it right, the Russian 8080 is made from photographs
of the chip, and then masks generated from those.
Is there only one Intel 8080A mask step level, or more than one?
Not less than 3 and possibly more for intel. I don't think they ever
brought up to NMOSII. It's similar for other vendors.
Post by glen herrmannsfeldt
If more, then you would want to be sure to test each one,
though most often the new mask was just a shrink of the
previous one. I believe the Russians had to increase the
size and decrease the speed to make theirs work.
Even a shrink has effects that may be noted though most
are detectable as differences in signal timing on real parts
and not instruction execution.

That test should mostly pass on Z80 and family save for those
ops that affect the PV flag. Even then it's edge cases.

Allison
Post by glen herrmannsfeldt
-- glen
dott.Piergiorgio
2009-09-23 07:39:43 UTC
Permalink
Raw Message
Post by glen herrmannsfeldt
(snip)
< Well, 7 months later I've just got my first response, and my web page
< has been updated with the results provided by somebody (thanks again
< Alexander) using a Russian KR580VM80A 8080 clone.
< I would still like to get the results for a "genuine" Intel 8080 CPU,
< or any of the other legitimate clones, so if anyone with the
< appropriate hardware can spare half an hour or so to run the
< application it would be much appreciated.
If I remember it right, the Russian 8080 is made from photographs
of the chip, and then masks generated from those.
Is there only one Intel 8080A mask step level, or more than one?
I fear it's only an urban legend... It's not difficult to send someone
to do shopping in a electronics shop/radio shack and put the merchandise
in the next diplomatic pouch... ;)

we're talking of electronics chips whose are selled at retailer level,
after all: if the standard joe doe can enter an electronic shop and buy
some 8080s, why this can't be done by an Embassy worker, or a relative
of him, or even a symphatizer for the "People's Cause" (whose are much
more here in then W Euruope back then) ?

Perhaps part of the myth stems from the actual (and later) reaganian-era
embargo on cpu with more than the rather unusual and contrived 28 bit,
whose led some "free enterpriser" to actually design and produce a
28-bit CPU (perhaps I'm wrong on the actual # of bit)

Best regards from Italy,
Dott. Piergiorgio.
glen herrmannsfeldt
2009-09-23 07:55:40 UTC
Permalink
Raw Message
dott.Piergiorgio <***@kaigun.fastwebnet.it> wrote:
(snip)
Post by dott.Piergiorgio
I fear it's only an urban legend... It's not difficult to send someone
to do shopping in a electronics shop/radio shack and put the merchandise
in the next diplomatic pouch... ;)
we're talking of electronics chips whose are selled at retailer level,
after all: if the standard joe doe can enter an electronic shop and buy
some 8080s, why this can't be done by an Embassy worker, or a relative
of him, or even a symphatizer for the "People's Cause" (whose are much
more here in then W Euruope back then) ?
Most likely that is how they got the first one to do
the mask images from.

The next story is that the Russians didn't like the 0.1in
(2.54mm) pin spacing on US chip packages, so they adjusted it
to 2.5mm. It seems like such a small change, but over the
length of a 40 pin package it adds up.

-- glen
A***@localhost.net
2009-09-23 02:26:10 UTC
Permalink
Raw Message
Post by IanB
Back in February I posted here asking anyone who had a CP/M 2.2 system
using an 8080 CPU to try out a short program for me. It was a port of
Frank Cringle's ZEXLAX (aka ZEXDOC and ZEXALL) Z80 exerciser that
tested an 8080 CPU instead.
The purpose was to get some numbers from genuine hardware so that I,
or anyone else, could check that a 8080 emulator was performing
correctly: exactly emulating the operation of an 8080.
Well, 7 months later I've just got my first response, and my web page
has been updated with the results provided by somebody (thanks again
Alexander) using a Russian KR580VM80A 8080 clone.
I would still like to get the results for a "genuine" Intel 8080 CPU,
or any of the other legitimate clones, so if anyone with the
appropriate hardware can spare half an hour or so to run the
application it would be much appreciated.
http://www.sunhillow.eu/8080exerciser
NB. If you have written an emulator and, when testing it using the
exerciser on the web site, it mysteriously fails on the 8 bit aluop
tests and all the dcr tests then send me an email. Thanks to some
detective work by some Russian Vector-06c users I can give you one
place to start looking.
Ian
Site looks fine to me. Then again I dont use winders.

I'll have to try that on 8085. IF ever there is time. I have intel
es8080, 8080, 8080A, NEC 8080 (the infamous one)
and 8080AF. I'd have to put them in the MITS8800BT as
that one is bootable but takes too much room in the room.

It should pass on all but the NEC8080 as that had the flags
that behaved as intel early spec'ed them but never actually
implemented completely. Making it incompatable with real
parts.

Allison
IanB
2009-09-23 07:10:19 UTC
Permalink
Raw Message
Post by A***@localhost.net
I'll have to try that on 8085.
It will probably fail all the tests on an 8085. The unused flag bits
in an 8080 are defined values, and that is what the exerciser tests
for, but they are undefined in an 8085. Easy enough to create a
version that ignores the undefined bits though - the original Z80
exerciser had two versions for that reason.
Post by A***@localhost.net
IF ever there is time. I have intel
es8080, 8080, 8080A, NEC 8080 (the infamous one)
and 8080AF. I'd have to put them in the MITS8800BT as
that one is bootable but takes too much room in the room.
It would certainly be appreciated if you could run the exerciser on
such a diverse range of CPUs. There's absolutely no rush though.
Post by A***@localhost.net
It should pass on all but the NEC8080 as that had the flags
that behaved as intel early spec'ed them but never actually
implemented completely. Making it incompatable with real
parts.
I assume that is the difference in the state of the AC flag after
subtract and decrement? I wrote my emulator from early(ish) Intel
docs and this is the one difference I came across when testing against
the Russian CPU.

Ian
--
The From address is valid
MdntTrain
2009-10-05 01:09:36 UTC
Permalink
Raw Message
Post by IanB
Back in February I posted here asking anyone who had a CP/M 2.2 system
using an 8080 CPU to try out a short program for me.  It was a port of
Frank Cringle's ZEXLAX (aka ZEXDOC and ZEXALL) Z80 exerciser that
tested an 8080 CPU instead.
The purpose was to get some numbers from genuine hardware so that I,
or anyone else, could check that a 8080 emulator was performing
correctly: exactly emulating the operation of an 8080.
Well, 7 months later I've just got my first response, and my web page
has been updated with the results provided by somebody (thanks again
Alexander) using a Russian KR580VM80A 8080 clone.
I would still like to get the results for a "genuine" Intel 8080 CPU,
or any of the other legitimate clones, so if anyone with the
appropriate hardware can spare half an hour or so to run the
application it would be much appreciated.
http://www.sunhillow.eu/8080exerciser
NB.  If you have written an emulator and, when testing it using the
exerciser on the web site, it mysteriously fails on the 8 bit aluop
tests and all the dcr tests then send me an email.  Thanks to some
detective work by some Russian Vector-06c users I can give you one
place to start looking.
Ian
--
The From address is valid
The pre runs on my 8085... will the exer run ok? Seems to hang as
follows:

[Cimmeria CBIOS v2.2K2 64K]
FDC init OK
CP/M load OK
A>save 5 test1.com
A>test1
8080 Preliminary tests complete
CP/M load OK
A>mon

[Cim Monitor v2.1 64K]
FF00 %
00 <- if 00, upload OK [uploaded 8080 exerciser hex file]
FF00 %
IDE reset OK
Loading CBIOS ...

[Cimmeria CBIOS v2.2K2 64K]
FDC init OK
CP/M load OK
A>save 20 test2.com
A>test2
8080 instruction exerciser
dad <b,d,h,sp>................
IanB
2009-10-05 05:38:00 UTC
Permalink
Raw Message
Post by MdntTrain
The pre runs on my 8085... will the exer run ok?
Should do.
Post by MdntTrain
FDC init OK
CP/M load OK
A>save 20 test2.com
A>test2
8080 instruction exerciser
dad <b,d,h,sp>................
I would have thought that should work. Depending on the speed of your
system it could take a few minutes to run. I've only run it on an
emulator and have no appreciation of how long it will take on a real
CPU though.

NB The exerciser will probably produce a different set of CRC values
on the 8085 than it will on an 8080 - the unused flag register bits
are handled differently. Comparing your results against those of the
KR580VM80A will almost certainly show differences.

Thank's for trying the exerciser though. If it still doesn't work
then let me know and I'll try and work out why.

Regards
Ian

Ian
--
The From address is valid
MdntTrain
2009-10-05 15:03:53 UTC
Permalink
Raw Message
I would have thought that should work.  Depending on the speed of your
system it could take a few minutes to run.  I've only run it on an
emulator and have no appreciation of how long it will take on a real
CPU though.
NB The exerciser will probably produce a different set of CRC values
on the 8085 than it will on an 8080 - the unused flag register bits
are handled differently.  Comparing your results against those of the
KR580VM80A will almost certainly show differences.
Thank's for trying the exerciser though.  If it still doesn't work
then let me know and I'll try and work out why.
Regards
Ian
I let it run longer but it really takes too long. You need to find a
way to have this whole test take like 30 minutes. I let it run 90
minutes and only got this far:

[Cimmeria CBIOS v2.2K2 64K]
FDC init OK
CP/M load OK
A>save 20 test2.com
A>test2
8080 instruction exerciser
dad <b,d,h,sp>................ ERROR **** crc expected:00000000 found:
44331def
aluop nn...................... ERROR **** crc expected:00000000
found:f1be2031
aluop <b,c,d,e,h,l,m,a>.......

If you can get the time down, I can also try on a real, early ceramic
8080.

jS
IanB
2009-10-05 19:55:05 UTC
Permalink
Raw Message
Post by MdntTrain
I let it run longer but it really takes too long. You need to find a
way to have this whole test take like 30 minutes.
You could well be right; I'll have to have a closer look at the source
to see if I can water down the testing without losing the accuracy.

Should anyone want to do an overnight soak test though ... :-)
OK, thanks for getting that far - I can still use the results you
posted to see how those tests match up at least.

Ian
--
The From address is valid
IanB
2009-10-05 10:31:00 UTC
Permalink
Raw Message
Post by MdntTrain
The pre runs on my 8085... will the exer run ok? Seems to hang as
I should have done this before, so apologies to you and anyone else
who has tried the exerciser and thought it didn't work. I hadn't done
any timing tests on my emulator and didn't appreciate how long the
exerciser would take on normal hardware...

Extrapolated to a 2MHz 8080 the timings for the first 4 tests (the
rest should each take less than a minute) are

dad - 11 mins
aluop nn - 5 mins
aluop - 3 hours
daa - 10 mins

Anyone who may still be thinking of trying the exerciser for me should
obviously take these numbers into account :-(

Ian
--
The From address is valid
MdntTrain
2009-10-05 15:06:29 UTC
Permalink
Raw Message
Post by IanB
dad - 11 mins
aluop nn - 5 mins
aluop - 3 hours
daa - 10 mins
Btw, the DAD portion took much longer than 11 minutes @ 2mhz (but I
did not time it).

jS
IanB
2009-10-05 20:18:10 UTC
Permalink
Raw Message
Post by MdntTrain
did not time it).
Hmmm.

The 11 mins was just a guess based on some simple timing loops and the
back of an envelope.

However, I've just added a state counter to my 8080 emulator and
dividing the result for the DAD test by 2,000,000 comes up with 568
seconds (9.46 mins). I know the state counts are a bit different for
an 8085 (I'm going to add the counter to that emulator tomorrow) but I
wouldn't have thought it would be too far out?

Ian
--
The From address is valid
MdntTrain
2009-10-06 01:24:18 UTC
Permalink
Raw Message
Post by IanB
The 11 mins was just a guess based on some simple timing loops and the
back of an envelope.
However,  I've just added a state counter to my 8080 emulator and
dividing the result for the DAD test by 2,000,000 comes up with 568
seconds (9.46 mins).  I know the state counts are a bit different for
an 8085 (I'm going to add the counter to that emulator tomorrow) but I
wouldn't have thought it would be too far out?
Ian
No, you're right. I timed it this time and it was under 10 minutes.

I have an idea: just post a version of the test without that 3 hr
part.

jS
IanB
2009-10-06 13:49:49 UTC
Permalink
Raw Message
Post by MdntTrain
No, you're right. I timed it this time and it was under 10 minutes.
Thanks again for the feedback.
Post by MdntTrain
I have an idea: just post a version of the test without that 3 hr
part.
I didn't want to add too many files, I thought it might make things
even more confusing, so I've made two changes to the web site

- Added a note about how long the tests take, that there is a way of
reducing the time taken to 30 mins, and asking that anyone who wants
the shortened version to contact me.

- Added an 8085 specific version of the basic exerciser. This is
identical to the 8080 version but masks out the unused flag bits when
constructing the CRC values from the test results. I decided this was
really needed as it would otherwise be difficult telling whether a
test failed because of a cpu difference or because of the extra flag
bits.

Ian
--
The From address is valid
Loading...