9 Dec
2005
9 Dec
'05
7:35 a.m.
Subject: Re: Legacy apps in Windows/OS X was Re: Old MS-DOS & Win Software
From: "Chuck Guzis" <cclist at sydex.com>
Date: Thu, 08 Dec 2005 21:59:46 -0800
To: cctalk at classiccmp.org
On 12/8/2005 at 9:49 PM Allison wrote:
I think you're thinking of the NEC V20/ uPD70108 and the V30/uPD70116 CPUs.
Plug-compatible with the 8088 and 8086, but with a couple of instructions
added to switch modes into 8080 emulation. Overall, emulation was pretty
good, but should have been Z80 instead of 8080 and there were a couple of
small bugs in the emulation that weren't fixed.
The V series were 1982 introduction. I was with NEC on the launch. By then
>
>Subject: Re: Legacy apps in Windows/OS X was Re: Old MS-DOS & Win
Software
From:
woodelf <bfranchuk at jetnet.ab.ca>
Date: Thu, 08 Dec 2005 19:30:35 -0700
To: General Discussion: On-Topic and Off-Topic Posts
<cctalk at
classiccmp.org>
What I do find odd that the 8086 came out first then a few years later
the 8088.
It still think the 8086/8086 could have had two instruction sets --
8080 emulation and
a better 16 bit design.
Not that I know of. I remember being able to get 8088s before the 86s
and that was back in 78 or early 79. Most vendors used the V20/V30, however, because they
were low-power CMOS.
They also implemented some of the 80186 entensions, like ENTER and LEAVE
and PUSHA/POPA and multi-bit immediate count shifts like SHR AX,5. If
you've got an old 8088/8086 box and need to run some code compiled to use
the 186/286 extensions, the V20/V30 is a very neat drop-in solution.
Not initially, The first generation of Vseries were NMOS. I have samples
of both.
The other differences were internal effective address calculation was handled
differntly from 8088 which shaved a few cycles of the execution. For the
same clock V series were between 5-10% faster. Also the V20/V30 had 8080
emulation.
Allison
9 Dec
9 Dec
12:30 p.m.
Allison wrote:
The other differences were internal effective address
calculation was handled
differntly from 8088 which shaved a few cycles of the execution. For the
same clock V series were between 5-10% faster. Also the V20/V30 had 8080
emulation.
I had forgotton about the V20's. Still a good way to get a few more % of
speed.
I don't think anybody used the 8080 emulation unless you had CP/M software
you needed to run. MSDOS was the only big OS being pushed as I think
Pascal and CP/M86 had been discontinued by then.
Allison
.
10 Dec
10 Dec
5:14 a.m.
woodelf wrote:
I had forgotton about the V20's. Still a good way
to get a few more % of
speed.
My memory may be clouded by university-grade levels of S&D&R&R but I
distinctly remember selling a kit with a V20 and 8087 to XT-class PC
owners in the shop I used to work in, *way* back in the early 90s.
These machines were usable but getting pretty low-end, the CPUs were
generally socketed (we'd^H^H^H^HI would - the other guy wouldn't do them
'cos of his eyesight - desolder soldered in ones and fit a socket and
the kit, all for an hour of workshop time) and it really was "bolt-on
power". Just the V20 alone made a noticable difference to some
programs, and the 8087 made maths stuff a hell of a lot quicker.
Gordon.
10:58 p.m.
Gordon JC Pearce wrote:
power". Just the V20 alone made a noticable
difference to some
programs, and the 8087 made maths stuff a hell of a lot quicker.
Agreed, although you had to be using something that had measureable
performance. Word processing didn't get any faster obviously, but running for
example FractINT was simply amazing (8087 makes FractINT actually usable on 8088).
I was always a little disappointed that the 8087 didn't speed up integer math
any (FDIV was at least 193 cycles, whereas DIV is never more than 162)... I
could have really used that.
--
Jim Leonard (trixter at oldskool.org) http://www.oldskool.org/
Want to help an ambitious games project? http://www.mobygames.com/
Or check out some trippy MindCandy at http://www.mindcandydvd.com/
12 Dec
12 Dec
5:31 p.m.
On 12/10/05, Jim Leonard <trixter at oldskool.org> wrote:
I was always a little disappointed that the 8087
didn't speed up integer
math
any (FDIV was at least 193 cycles, whereas DIV is never more than 162)...
I
could have really used that.
It speeds up 32 bit and 64 bit integer math.....
Even on newer processors it was floating point was useful for many non-FP
things. Desipte intel's documentation to the contrary, the fastest block
transfer loop on the (pre-MMX) Pentium was done using FILDQ-FISTPQ.
// GCC w/ AT&T Assembly warning!!!
inline void copy64(int *p1,int *p2, int n)
{
asm("
test $1,%2
jz 0f
dec %2
movl (%0,%2,4),%%ebx
movl %%ebx,(%1,%2,4)
0:
shrl %2
dec %2
jl 2f
1:
fildq (%0,%2,8)
fistpq (%1,%2,8)
dec %2
jge 1b
2:
"
:
: "S" (p1), "D" (p2), "c" (n)
: "ebx");
}
6:40 p.m.
Eric J Korpela wrote:
On 8088, not 32-bit; see my previous message for timings. 64-bit, yes.
I was always a
little disappointed that the 8087 didn't speed up integer
math
any (FDIV was at least 193 cycles, whereas DIV is never more than 162)...
I
could have really used that.
It speeds up 32 bit and 64 bit integer math..... Even on newer processors it was floating point was
useful for many non-FP
things. Desipte intel's documentation to the contrary, the fastest block
transfer loop on the (pre-MMX) Pentium was done using FILDQ-FISTPQ.
Ah yes, I almost crapped a brick when I found that out!
--
Jim Leonard (trixter at oldskool.org) http://www.oldskool.org/
Want to help an ambitious games project? http://www.mobygames.com/
Or check out some trippy MindCandy at http://www.mindcandydvd.com/
9 Dec
9 Dec
12:54 p.m.
On 12/9/2005 at 7:35 AM Allison wrote:
The other differences were internal effective address
calculation was
handled
differntly from 8088 which shaved a few cycles of the execution. For the
same clock V series were between 5-10% faster. Also the V20/V30 had 8080
emulation.
Hmmm, thought I just said that. One of the more highly-touted performance
improvements in the V20/V30 was the implemenation of dual internal data
busses. In practice, I don't know if it made a huge difference in speed,
but NEC claimed up to a 30% improvement for som operations..
FWIW, my reference manual for the 70108/70116 is dated August 1985 and
identifies both as "Low Power CMOS"; no mention otherwise of any other fab
technology. NEC did later offer the -H series variants (V20H, V30H, etc.)
which featured fully static operation and lower power consumption than the
V20/V30--and you could get them in clock speeds of up to 16 MHz.
I've got a reasonably thick folder in my files titled "Natick" with a bunch
of correspondence with the NEC folks there on the whole subject. The V40,
which was sort of an 80188 in CMOS (PGA, QFP and PLCC pacakges) also
claimed to emulate the 8080 instruction set, but I never got hold of one to
try it out (or I didn't care to).
I thought it interesting that the 8080 emulation of the V20/V30 emulated
the Intel 8080A and not tne NEC 8080A instruction set.
Cheers,
Chuck
1:25 p.m.
Chuck Guzis wrote:
Hmmm, thought I just said that. One of the more
highly-touted performance
improvements in the V20/V30 was the implemenation of dual internal data
busses. In practice, I don't know if it made a huge difference in speed,
but NEC claimed up to a 30% improvement for som operations..
That was a long time ago ... the XT clone wars. Back then a M$ mouse $100.
No need to gouge the public with windows /85 :)
I thought it interesting that the 8080 emulation of the
V20/V30 emulated
the Intel 8080A and not tne NEC 8080A instruction set.
So what instructions did the NEC 8080A have different?
Cheers,
Chuck
.
1:48 p.m.
On 12/9/2005 at 11:25 AM woodelf wrote:
So what instructions did the NEC 8080A have different?
The NEC 8080A kept an extra flag around in the status register, the
"subtract" flag. Again, this ties in with that awful DAA instruction.
Suppose you have the following code:
mvi a,22h
sbi 13h
daa
On the Intel 8080a, thre result's 15h, which is clearly wrong. However, on
the NEC, it's correctly rendered as 09h, since the NEC tracks the use of
the subtract, rather than an add. The Z80 does this also.
However, this isn't a fatal error in the Intel part; you simply have to do
your BCD subtracts using 10's complement addition. So, 10's complement of
13h is 87h and
mvi a,22h
adi 87h
daa
Correctly yields 09h. The 8086 has the DAS instruction to adjust BCD
subtracts, but the 8080 lacks it.
Cheers,
Chuck
6870
days inactive
6873
days old
8 comments
6 participants
participants (6)
-
ajp166＠bellatlantic.net -
bfranchuk＠jetnet.ab.ca -
cclist＠sydex.com -
gordonjcp＠gjcp.net -
korpela＠ssl.berkeley.edu -
trixter＠oldskool.org