From: "Zane H. Healy" Subject: SIMH is Int or FP important Newsgroups: alt.sys.pdp10 Organization: Aracnet User-Agent: tin/1.4.4-20000803 ("Vet for the Insane") (UNIX) (Linux/2.2.19 (i686)) Lines: 15 Message-ID: X-Complaints-To: abuse@usenetserver.com X-Abuse-Info: Please be sure to forward a copy of ALL headers X-Abuse-Info: Otherwise we will be unable to process your complaint properly. NNTP-Posting-Date: Wed, 03 Oct 2001 21:08:24 EDT Date: Thu, 04 Oct 2001 01:08:24 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.tele.dk!small.news.tele.dk!64.152.100.70!sjcppf01.usenetserver.com!e420r-sjo4.usenetserver.com!usenetserver.com!sjcpnn01.usenetserver.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7020 I'm trying to decide between a Pentium III 1Ghz or a Pentium IV 1.5Ghz, and am wondering what would provide the best performance for programs such as SIMH. Are emulators Interger or FP intensive? These two chips are the only two choices, other than waiting a few months and getting a Pentium IV 1.8Ghz. The problem with waiting being the 1.8Ghz system will have PC133 SDRAM instead of RDRAM. The other key factor in the decision is that the 1.0Ghz or 1.8Ghz system would be free, while the 1.5Ghz would cost me almost $400. Based on that I'm also considering waiting and getting the 1.8Ghz system, but I don't like the thought of being stuck with SDRAM in a Pentium IV system. Zane ###### From: bugs@netcom.com (Mark Hittinger) Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Date: 4 Oct 2001 01:18:23 GMT Organization: MindSpring Enterprises Lines: 27 Message-ID: <9pgdcv$22a$1@slb7.atl.mindspring.net> References: NNTP-Posting-Host: c7.ae.21.58 X-Newsreader: NN version 6.5.0 CURRENT #121 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed00.sul.t-online.de!t-online.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!netnews.com!xfer02.netnews.com!newsfeed2.earthlink.net!newsfeed.earthlink.net!news.mindspring.net!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7010 "Zane H. Healy" writes: >I'm trying to decide between a Pentium III 1Ghz or a Pentium IV 1.5Ghz, and >am wondering what would provide the best performance for programs such as >SIMH. Are emulators Interger or FP intensive? The emulators are very integer intensive. I think both choices would yield incredible PDP-10 performance compared to the actual machines :-) There are some cooling issues with the P IV part that you might want to look into. The P IV part has a fallback mode when it gets too hot - it slows itself down - kinda cheezy. Check into compiler issues related to generating P IV optimized code vs regular Pentium optimized code. With both boxes you'll need to invest in extra fans and focus on cooling. It would be nice to get the part with the largest instruction cache. We are actually starting to see motherboards with memory interleaving showing up - its about time. Later Mark Hittinger Earthlink bugs@netcom.com ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important References: Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. Date: 03 Oct 2001 19:09:07 -0700 Message-ID: Lines: 14 User-Agent: Gnus/5.0807 (Gnus v5.8.7) Emacs/20.7 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 3 Oct 2001 19:10:26 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newscore.univie.ac.at!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!news-hog.berkeley.edu!ucberkeley!enews.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.sys.pdp10:7015 "Zane H. Healy" writes: > I'm trying to decide between a Pentium III 1Ghz or a Pentium IV 1.5Ghz, and > am wondering what would provide the best performance for programs such as > SIMH. Are emulators Interger or FP intensive? Integer. Any particular reason why you're not considering a 1.4 GHz Athlon? I think that's a better value for the money. On most tasks it is faster than the Pentium IV 1.5 GHz. If you get an Athlon, I'd suggest getting a motherboard with the AMD 760 chipset, using DDR SDRAM. ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important References: <9pgdcv$22a$1@slb7.atl.mindspring.net> Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. Date: 03 Oct 2001 19:12:33 -0700 Message-ID: Lines: 13 User-Agent: Gnus/5.0807 (Gnus v5.8.7) Emacs/20.7 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 3 Oct 2001 19:13:51 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!news-hog.berkeley.edu!ucberkeley!enews.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.sys.pdp10:7016 bugs@netcom.com (Mark Hittinger) writes: > We are actually starting to see motherboards with memory interleaving showing > up - its about time. Not really necessary with SDRAM (or DDR SDRAM), because these divide the internal memory into separate banks and overlap access. Thus they can support interleave with only one memory module. External interleave over a common bus does not offer any advantage over this. External interleave of two banks using independent busses can be a win, if the bandwidth from the CPU (FSB) to the memory control is substantially higher than the memory bandwidth. None of the currently available x86 processors have a high enough FSB bandwidth to take advantage of this. ###### From: "Zane H. Healy" Subject: Re: SIMH is Int or FP important Newsgroups: alt.sys.pdp10 References: Organization: Aracnet User-Agent: tin/1.4.4-20000803 ("Vet for the Insane") (UNIX) (Linux/2.2.19 (i686)) Lines: 17 Message-ID: X-Complaints-To: abuse@usenetserver.com X-Abuse-Info: Please be sure to forward a copy of ALL headers X-Abuse-Info: Otherwise we will be unable to process your complaint properly. NNTP-Posting-Date: Wed, 03 Oct 2001 22:45:05 EDT Date: Thu, 04 Oct 2001 02:45:05 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!HSNX.atgi.net!sjcppf01.usenetserver.com!e420r-sjo4.usenetserver.com!usenetserver.com!sjcpnn01.usenetserver.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7018 Eric Smith wrote: > Any particular reason why you're not considering a 1.4 GHz Athlon? I think > that's a better value for the money. On most tasks it is faster than > the Pentium IV 1.5 GHz. > If you get an Athlon, I'd suggest getting a motherboard with the AMD 760 > chipset, using DDR SDRAM. Well, for one thing I don't do clone chips. The other reason, touched upon in the third paragraph, is that the 1.0Ghz or 1.8Ghz system would be free, the 1.5Ghz would cost me about $400. I think it's boiling down to what would offer me better Integer performance, a Pentium III 1Ghz, or a Pentium IV 1.8Ghz with SDRAM (i845 chipset). It's looking like I should wait a couple months for the Pentium IV 1.8Ghz. Zane ###### Message-ID: <3BBBD30D.5E9BA94C@ne.mediaone.net> From: Monty Brandenberg Reply-To: mcbinc@ne.mediaone.net Organization: MCB, Inc. X-Mailer: Mozilla 4.77 [en] (X11; U; Linux 2.2.19-7.0.1smp i686) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important References: Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 11 Date: Thu, 04 Oct 2001 03:09:16 GMT NNTP-Posting-Host: 65.96.128.35 X-Complaints-To: abuse@mediaone.net X-Trace: typhoon.ne.mediaone.net 1002164956 65.96.128.35 (Wed, 03 Oct 2001 23:09:16 EDT) NNTP-Posting-Date: Wed, 03 Oct 2001 23:09:16 EDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!feed2.news.rcn.net!rcn!chnws02.mediaone.net!chnws06.ne.mediaone.net!24.128.8.70!typhoon.ne.mediaone.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7007 "Zane H. Healy" wrote: > > Well, for one thing I don't do clone chips. What is a simulator but a clone? -- Monty Brandenberg, Software Consultant MCB, Inc. mcbinc@world.std.com P.O. Box 426188 mcbinc@ne.mediaone.net Cambridge, MA 02142-0021 617.864.6907 ###### Reply-To: "Doug Quebbeman" From: "Doug Quebbeman" Newsgroups: alt.sys.pdp10 References: Subject: Re: SIMH is Int or FP important Date: Thu, 4 Oct 2001 08:06:58 -0400 Lines: 71 Organization: Full Circle Systems MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0027_01C14CAB.8A6380E0" X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2919.6700 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6700 NNTP-Posting-Host: lou-ts10-46.iglou.com X-Original-NNTP-Posting-Host: lou-ts10-46.iglou.com Message-ID: <3bbc50d3_3@news.iglou.com> X-Trace: news.iglou.com 1002197203 lou-ts10-46.iglou.com (4 Oct 2001 08:06:43 -0400) X-Authenticated-User: dougq X-Original-NNTP-Posting-Host: 204.255.236.153 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newspeer.radix.net!uunet!ash.uu.net!news.iglou.com Xref: chonsp.franklin.ch alt.sys.pdp10:7001 This is a multi-part message in MIME format. ------=_NextPart_000_0027_01C14CAB.8A6380E0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable "Zane H. Healy" wrote in message = news:cEOu7.3794$T%4.278435@sjcpnn01.usenetserver.com... > I'm trying to decide between a Pentium III 1Ghz or a Pentium IV = 1.5Ghz, and > am wondering what would provide the best performance for programs such = as > SIMH. Are emulators Interger or FP intensive? =20 Based on numbers from the CDC-simulator-work-in-progress, they will be = about the same speed, with the P-IV a little slower. However, we've discussed = this before. You're expecting P-IV code-generating compilers to eliminate much of the = speed differential, while I'm skeptical. I'd go with the 1.0GHz P-III... -dq ------=_NextPart_000_0027_01C14CAB.8A6380E0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
"Zane H. Healy" <healyzh@shell1.aracnet.com= >=20 wrote in message news:cEOu7.= 3794$T%4.278435@sjcpnn01.usenetserver.com...
> I'm trying to decide between a Pentium III 1Ghz or a Pentium = IV=20 1.5Ghz, and
> am wondering what would provide the best performance = for=20 programs such as
> SIMH.  Are emulators Interger or FP=20 intensive? 
Based on numbers from the CDC-simulator-work-in-progress, they will = be=20 about
the same speed, with the P-IV a little slower. However, we've = discussed=20 this before.
You're expecting P-IV code-generating compilers to eliminate much = of the=20 speed
differential, while I'm skeptical.
 
I'd go with the 1.0GHz P-III...
 
-dq
 
------=_NextPart_000_0027_01C14CAB.8A6380E0-- ###### Reply-To: "Doug Quebbeman" From: "Doug Quebbeman" Newsgroups: alt.sys.pdp10 References: Subject: Re: SIMH is Int or FP important Date: Thu, 4 Oct 2001 08:12:32 -0400 Lines: 32 Organization: Full Circle Systems X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2919.6700 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6700 NNTP-Posting-Host: lou-ts10-46.iglou.com X-Original-NNTP-Posting-Host: lou-ts10-46.iglou.com Message-ID: <3bbc5223_1@news.iglou.com> X-Trace: news.iglou.com 1002197539 lou-ts10-46.iglou.com (4 Oct 2001 08:12:19 -0400) X-Authenticated-User: dougq X-Original-NNTP-Posting-Host: 204.255.236.153 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed00.sul.t-online.de!t-online.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!news-out.visi.com!hermes.visi.com!uunet!ash.uu.net!news.iglou.com Xref: chonsp.franklin.ch alt.sys.pdp10:7000 "Zane H. Healy" wrote in message news:R2Qu7.4786$T%4.369074@sjcpnn01.usenetserver.com... > Eric Smith wrote: > > Any particular reason why you're not considering a 1.4 GHz Athlon? I think > > that's a better value for the money. On most tasks it is faster than > > the Pentium IV 1.5 GHz. > > > If you get an Athlon, I'd suggest getting a motherboard with the AMD 760 > > chipset, using DDR SDRAM. > > Well, for one thing I don't do clone chips. The other reason, touched upon > in the third paragraph, is that the 1.0Ghz or 1.8Ghz system would be free, > the 1.5Ghz would cost me about $400. The Athlon isn't a clone chip, it's an entirely unique CPU architecture that includes the IA32 instruction set. Since the Intel Itanium (IA-64) will execute IA32 instructions in a special microcoded emulation mode, doesn't that make Intel's next chip a (superset-based) clone of its own Pentium line? Games seem to be the ultimate CPU benchmark, and most of the gamers I know are very pleased with the Athlons they're running... Regards, -dq ###### From: Bob Supnik Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Date: Thu, 04 Oct 2001 13:01:56 -0400 Organization: Posted via Supernews, http://www.supernews.com Message-ID: References: X-Newsreader: Forte Agent 1.8/32.548 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Complaints-To: newsabuse@supernews.com Lines: 34 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed00.sul.t-online.de!t-online.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!feed1.uncensored-news.com!propagator-la!news-in-la.newsfeeds.com!feed.textport.net!sn-xit-04!sn-post-01!supernews.com!corp.supernews.com!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7021 SIMH does essentially no floating point. The displayed "elapsed time in cycles" is kept in double precision, to avoid using a 64b integer in 32b only simulators; that's it. All simulated floating point is done with integer calculations, because none of the simulated systems used IEEE format. The Pentium IV's longer pipeline offsets its higher clock rate; for most integer applications, a 1.4Mhz P4 benchmarks like a 1.0Mhz P3. Aside from integer computation rate, about the only other factor in SIMH performance is the efficiency of them memory hierarchy; larger first and second level caches help, as done higher performance memory. /Bob On Thu, 04 Oct 2001 01:08:24 GMT, "Zane H. Healy" wrote: >I'm trying to decide between a Pentium III 1Ghz or a Pentium IV 1.5Ghz, and >am wondering what would provide the best performance for programs such as >SIMH. Are emulators Interger or FP intensive? > >These two chips are the only two choices, other than waiting a few months >and getting a Pentium IV 1.8Ghz. The problem with waiting being the 1.8Ghz >system will have PC133 SDRAM instead of RDRAM. > >The other key factor in the decision is that the 1.0Ghz or 1.8Ghz system >would be free, while the 1.5Ghz would cost me almost $400. Based on that >I'm also considering waiting and getting the 1.8Ghz system, but I don't like >the thought of being stuck with SDRAM in a Pentium IV system. > > Zane > ###### Path: chonsp.franklin.ch!not-for-mail From: Neil Franklin Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Date: 04 Oct 2001 22:17:23 +0200 Organization: My own Private Self Lines: 41 Message-ID: <6ulmirrwh8.fsf@chonsp.franklin.ch> References: NNTP-Posting-Host: chonsp.franklin.ch X-Trace: chonsp.franklin.ch 1002226643 445 10.0.3.2 (4 Oct 2001 20:17:23 GMT) X-Complaints-To: news@chonsp.franklin.ch NNTP-Posting-Date: 4 Oct 2001 20:17:23 GMT X-Newsreader: Gnus v5.7/Emacs 20.4 Xref: chonsp.franklin.ch alt.sys.pdp10:7023 "Zane H. Healy" writes: > I'm trying to decide between a Pentium III 1Ghz or a Pentium IV 1.5Ghz, and > am wondering what would provide the best performance for programs such as > SIMH. Are emulators Interger or FP intensive? Totally Integer. May not even use FP at all. And if they do use it, then only for emulationg FP instructions. So only if you have enough of them in your code you run under the emulator. > These two chips are the only two choices, other than waiting a few months > and getting a Pentium IV 1.8Ghz. The problem with waiting being the 1.8Ghz > system will have PC133 SDRAM instead of RDRAM. > I'm also considering waiting and getting the 1.8Ghz system, but I don't like > the thought of being stuck with SDRAM in a Pentium IV system. Funny. I would call that an advantage. Today I nearly allways select product on the manufacturers reputation. And RDRAM (made by Rambus,Inc) is for me a total K.O. criterium, since their attempts to to sue their competitors off the market, after their own technology failled to grab customers on its (lack of) technological merits. So that makes it SDRAM. And PC133 is only about 5% slower than the DDR stuff. Dito Pentium from Intel is an negative point, given their attempts fo stop competition by constantly changing pinouts and so devaluating otherwise re-usable motherboards. And AMDs work good. Other thing against 1.8GHz: energy usage, heat production, cooling problems. A 1GHz machine will be less trouble in that respect. Neil "using 350MHz AMD K6-2 since 3 years and totally satisfied" Franklin -- Neil Franklin, neil@franklin.ch.remove http://neil.franklin.ch/ Hacker, Unix Guru, El Eng HTL/BSc, Sysadmin, Archer, Roleplayer - Intellectual Property is Intellectual Robbery ###### From: "Zane H. Healy" Subject: Re: SIMH is Int or FP important Newsgroups: alt.sys.pdp10 References: <3bbc50d3_3@news.iglou.com> Organization: Aracnet User-Agent: tin/1.4.4-20000803 ("Vet for the Insane") (UNIX) (Linux/2.2.19 (i686)) Lines: 21 Message-ID: X-Complaints-To: abuse@usenetserver.com X-Abuse-Info: Please be sure to forward a copy of ALL headers X-Abuse-Info: Otherwise we will be unable to process your complaint properly. NNTP-Posting-Date: Thu, 04 Oct 2001 15:36:40 EDT Date: Thu, 04 Oct 2001 19:36:40 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!HSNX.atgi.net!sjcppf01.usenetserver.com!e420r-sjo4.usenetserver.com!usenetserver.com!sjcpnn01.usenetserver.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7034 Doug Quebbeman wrote: > Based on numbers from the CDC-simulator-work-in-progress, they will be about > the same speed, with the P-IV a little slower. However, we've discussed this before. > You're expecting P-IV code-generating compilers to eliminate much of the speed > differential, while I'm skeptical. > I'd go with the 1.0GHz P-III... Now I remember who I was talking about this with. I was trying to figure that out yesterday, as I seemed to remember it had something to do with emulators. Remember your Pentium III is a dual 1Ghz system and the Pentium IV 1.7Ghz is a single. All three of the potential systems I'm talking about are single processor systems. And yes, I do expect a Pentium IV optimized program to run a lot better than the Pentium III. BTW, if anyone else is trying to decide, I just found an interesting site that I need to read through doing a compare against a 1Ghz Pentium III, 1.2Ghz Athlon, and a 1.7Ghz Pentium IV. http://www.linuxhardware.org/features/01/07/13/145249.shtml Zane ###### Sender: phr2001-nospam@ruckus.brouhaha.com From: Paul Rubin Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important References: Date: 14 Oct 2001 23:13:51 -0700 Message-ID: <7xofn91lcw.fsf@ruckus.brouhaha.com> Organization: Nightsong/Fort GNOX Lines: 25 User-Agent: Gnus/5.0807 (Gnus v5.8.7) Emacs/20.7 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 14 Oct 2001 23:17:12 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!enews.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.sys.pdp10:7129 1. The simulators likely don't use any FP at all, since the PDP10 and VAX don't use IEEE floats, so exact simulation would have to be done with integers. If the simulation is approximated with IEEE floats and format conversion, FP performance still doesn't matter because the IEEE FP arithmetic on a modern CPU is orders of magnitude faster than a PDP-10. 2. All the systems you mentioned will run PDP-10 code faster than any real PDP-10 ever built could. In a very real sense, making the simulation run too much faster than the real thing makes the simulation unfaithful. I'm assuming you don't have actual production PDP-10 code that you want to run for some practical purpose, but rather you want to recreate the feeling and atmosphere of the PDP-10 environment. A lot of the programming decisions etc. of PDP-10 programs were based on the speed of the real PDP-10. If you were a vintage aircraft buff and were building a working replica of a WW1 biplane, would you make it fly at supersonic speed? I won't go as far as saying simulators should normally be artificially slowed down to simulate the speed of real 10's, but I question the purpose of making them faster and faster, once they're a few times faster than real 10's, which they already are. 3. Given the choices of systems you mentioned, I'd definitely pick the free one without worrying too much about minor speed differences between it and the others. ###### From: "Zane H. Healy" Subject: Re: SIMH is Int or FP important Newsgroups: alt.sys.pdp10 References: <7xofn91lcw.fsf@ruckus.brouhaha.com> Organization: Aracnet User-Agent: tin/1.4.4-20000803 ("Vet for the Insane") (UNIX) (Linux/2.2.19 (i686)) Lines: 29 Message-ID: X-Complaints-To: abuse@usenetserver.com X-Abuse-Info: Please be sure to forward a copy of ALL headers X-Abuse-Info: Otherwise we will be unable to process your complaint properly. NNTP-Posting-Date: Mon, 15 Oct 2001 14:55:09 EDT Date: Mon, 15 Oct 2001 18:55:09 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!enews.sgi.com!newshub2.rdc1.sfba.home.com!news.home.com!sjcppf01.usenetserver.com!e420r-sjo4.usenetserver.com!usenetserver.com!sjcpnn01.usenetserver.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7130 Paul Rubin wrote: > biplane, would you make it fly at supersonic speed? I won't go as far > as saying simulators should normally be artificially slowed down to > simulate the speed of real 10's, but I question the purpose of making > them faster and faster, once they're a few times faster than real 10's, > which they already are. Actually most emulators provide the ability to run at the original speed. Of course most are used for playing games. What I'd like to see is the ability for SIMH to use less CPU cycles when the computer it's simulating isn't doing anything (this ability had existed for TS10 I believe). In my case I'm wanting max speed because I'd like to be able to run at least 3 copies of the emulator at once, and be able to do other stuff. Right now I've got a 500Mhz Celeron on a MicroATX board that is totally dedicated to running two copies of SIMH so I've got TOPS-10 and TOPS-20 running at the same time. This gives me about 2.5x the speed of a KS10 for each. > 3. Given the choices of systems you mentioned, I'd definitely pick the > free one without worrying too much about minor speed differences between > it and the others. There were actually two of us trying to decide what to get. I approached it from the standpoint of running Linux and SIMH, the woman I work with is a hard core gamer and approached it from that standpoint. We both ordered the same system, the 1Ghz Pentium III. I should know in a couple weeks how well it works out. Zane ###### Message-ID: <3BCB43CA.E8811687@jetnet.ab.ca> From: Ben Franchuk X-Mailer: Mozilla 4.77 [en] (X11; U; Linux 2.2.19 i586) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> <9qh9vr02n08@drn.newsguy.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 14 Date: Mon, 15 Oct 2001 14:15:06 -0600 NNTP-Posting-Host: 207.153.6.61 X-Trace: newsfeed.slurp.net 1003242097 207.153.6.61 (Tue, 16 Oct 2001 09:21:37 CDT) NNTP-Posting-Date: Tue, 16 Oct 2001 09:21:37 CDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.slurp.net!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7132 Tim Shoppa wrote: > There are *some* hooks into kernel scheduling but they were experimental > last time I used them (and they probably change every couple of years, > so whatever I knew about them under the 0.98 kernel patches is completely > inapplicable now I'm sure) that could make per-process limits like this. > Look in your kernel sources at the kernel/sched.c module. > > Is running at niceness +20 workable on your system? (It may not be, > depending on what the machine is used for otherwise!) One other option is to increase the frequency of the time slices under linux. While this will not help with wasted cycles it could improve I/O response as you can sample more often. Ben Franchuk. ###### From: Arthur Krewat Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Lines: 38 Message-ID: <3BCB4787.3A0A681E@bartek.dontspamme.net> References: <7xofn91lcw.fsf@ruckus.brouhaha.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.78 [en] (X11; U; SunOS 5.8 i86pc) X-Accept-Language: en Date: Mon, 15 Oct 2001 20:32:49 GMT NNTP-Posting-Host: 24.186.100.134 X-Trace: news02.optonline.net 1003177969 24.186.100.134 (Mon, 15 Oct 2001 16:32:49 EDT) NNTP-Posting-Date: Mon, 15 Oct 2001 16:32:49 EDT Organization: Optimum Online Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeeds.belnet.be!news.belnet.be!feed2.onemain.com!feed1.onemain.com!cyclone2.usenetserver.com!usenetserver.com!news01.optonline.net!news02.optonline.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7145 "Zane H. Healy" wrote: > > Paul Rubin wrote: > > biplane, would you make it fly at supersonic speed? I won't go as far > > as saying simulators should normally be artificially slowed down to > > simulate the speed of real 10's, but I question the purpose of making > > them faster and faster, once they're a few times faster than real 10's, > > which they already are. > > Actually most emulators provide the ability to run at the original speed. > Of course most are used for playing games. What I'd like to see is the > ability for SIMH to use less CPU cycles when the computer it's simulating > isn't doing anything (this ability had existed for TS10 I believe). There was a hack that allowed detection of the idle job loop in TOPS-10 7.03. The conditional is still there, but I believe something needs to be altered to compile it (something about a function that changed names). I'd very much like to get this working with SIMH, and I might even take a look at it if these damn paying customers would get off my back! :) I would need help to find the TOPS-20 equivelants of the idle job. Or, just take any AOJ/SOJ instructions that do nothing but jump to themselves and instead of emulating each cycle, just zeroing out the AC and sleeping. To be consistent, you would have to figure the approximate emulation speed into the equation to figure out how long to sleep. > There were actually two of us trying to decide what to get. I approached it > from the standpoint of running Linux and SIMH, the woman I work with is a > hard core gamer and approached it from that standpoint. We both ordered the > same system, the 1Ghz Pentium III. I should know in a couple weeks how well > it works out. I'd very much like to hear about this. Right now, I run a dual 533Mhz Celeron, and running SIMH at 100% on one cpu, the machine is still very snappy (AGP ATI card, 384megs of ram, Abit BP6 motherboard). I may go back to 433's and overclock them, since the memory access would be faster, but ... aak ###### From: "Zane H. Healy" Subject: Re: SIMH is Int or FP important Newsgroups: alt.sys.pdp10 References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> Organization: Aracnet User-Agent: tin/1.4.4-20000803 ("Vet for the Insane") (UNIX) (Linux/2.2.19 (i686)) Lines: 22 Message-ID: X-Complaints-To: abuse@usenetserver.com X-Abuse-Info: Please be sure to forward a copy of ALL headers X-Abuse-Info: Otherwise we will be unable to process your complaint properly. NNTP-Posting-Date: Mon, 15 Oct 2001 18:30:00 EDT Date: Mon, 15 Oct 2001 22:30:00 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!kanja.arnes.si!news-hub.siol.net!cyclone.bc.net!sjcppf01.usenetserver.com!e420r-sjo4.usenetserver.com!usenetserver.com!sjcpnn01.usenetserver.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7158 Arthur Krewat wrote: >> same system, the 1Ghz Pentium III. I should know in a couple weeks how well >> it works out. > I'd very much like to hear about this. Right now, I run a dual 533Mhz Celeron, > and running SIMH at 100% on one cpu, the machine is still very snappy (AGP > ATI card, 384megs of ram, Abit BP6 motherboard). I may go back to 433's and > overclock them, since the memory access would be faster, but ... My chief concern is that the computer I'm getting is a Name Brand computer (actually with x86 systems that's a first for me). As a result I'm wondering how bad the parts in it will be (at least it's got a Matrox graphics card). Somehow I've nightmares of getting it to work well with Linux. Out of curiousity, do you know of any way under Linux to limit a process to x% of the CPU. So for example a copy of SIMH gets a maximum of 10% of the CPU time? I started thinking and this is probably better than trying to get SIMH to 'idle'. The only problem being, I don't see an obvious way to do this under UNIX. Zane ###### From: Arthur Krewat Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Lines: 42 Message-ID: <3BCB9EAE.7694CA37@bartek.dontspamme.net> References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.78 [en] (X11; U; SunOS 5.8 i86pc) X-Accept-Language: en Date: Tue, 16 Oct 2001 02:47:58 GMT NNTP-Posting-Host: 24.186.100.134 X-Trace: news02.optonline.net 1003200478 24.186.100.134 (Mon, 15 Oct 2001 22:47:58 EDT) NNTP-Posting-Date: Mon, 15 Oct 2001 22:47:58 EDT Organization: Optimum Online Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!cyclone2.usenetserver.com!usenetserver.com!news01.optonline.net!news02.optonline.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7147 "Zane H. Healy" wrote: > > Arthur Krewat wrote: > >> same system, the 1Ghz Pentium III. I should know in a couple weeks how well > >> it works out. > > > I'd very much like to hear about this. Right now, I run a dual 533Mhz Celeron, > > and running SIMH at 100% on one cpu, the machine is still very snappy (AGP > > ATI card, 384megs of ram, Abit BP6 motherboard). I may go back to 433's and > > overclock them, since the memory access would be faster, but ... > > My chief concern is that the computer I'm getting is a Name Brand computer > (actually with x86 systems that's a first for me). As a result I'm > wondering how bad the parts in it will be (at least it's got a Matrox > graphics card). Somehow I've nightmares of getting it to work well with > Linux. True, I've built all of my own x86 UNIX boxes and always was sure to pick the correct parts that were on the Solaris HCL. Although, Linux does support a heck of a lot more devices than Solaris. As long as the thing has an IDE or Adaptec SCSI controller, and a 3com (or almost any) network card, I can't imagine it giving you any trouble. Unless, of course, you need sound :) > Out of curiousity, do you know of any way under Linux to limit a process to > x% of the CPU. So for example a copy of SIMH gets a maximum of 10% of the > CPU time? I started thinking and this is probably better than trying to get > SIMH to 'idle'. The only problem being, I don't see an obvious way to do > this under UNIX. None that I know of, except "nice" which only puts you at a lower priority and doesn't guarantee specific size timeslices. Somehow, I think it would be better to alter SIMH and TS10 to allow a certain amount of (configurable) sleep time for every instruction, not to mention the aforementioned optimizations of AOJ/SOJ instructions that loop to themselves. Maybe a feature to set the cycle time (in milliseconds) per instruction, so you can really see what a KS10/KL10 was like. Or, even slow down to a PDP-6 with all the extra instructions of a KS10 :) I'd like to slow it down to KA10 speed and see what it was like when I was 14 - I vaguely remember the experience. They upgraded 2 months later to KS10's so I never worked on the KA10 much. aak ###### Reply-To: "Douglas H. Quebbeman" From: "Douglas H. Quebbeman" Newsgroups: alt.sys.pdp10 References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> Subject: Re: SIMH is Int or FP important Date: Tue, 16 Oct 2001 07:43:35 -0400 Lines: 48 Organization: Full Circle Systems X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.50.4522.1200 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4522.1200 NNTP-Posting-Host: 204.250.0.238 X-Original-NNTP-Posting-Host: 204.250.0.238 Message-ID: <3bcc1d68$1_3@news.iglou.com> X-Trace: news.iglou.com 1003232616 204.250.0.238 (16 Oct 2001 07:43:36 -0400) X-Authenticated-User: dougq X-Original-NNTP-Posting-Host: 204.250.0.238 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed00.sul.t-online.de!newsfeed01.sul.t-online.de!t-online.de!lnewspeer00.lnd.ops.eu.uu.net!emea.uu.net!zur.uu.net!ash.uu.net!news.iglou.com Xref: chonsp.franklin.ch alt.sys.pdp10:7135 "Zane H. Healy" wrote in message news:IrJy7.112912$Rb.4059038@sjcpnn01.usenetserver.com... > Arthur Krewat wrote: > >> same system, the 1Ghz Pentium III. I should know in a couple weeks how well > >> it works out. > > > I'd very much like to hear about this. Right now, I run a dual 533Mhz Celeron, > > and running SIMH at 100% on one cpu, the machine is still very snappy (AGP > > ATI card, 384megs of ram, Abit BP6 motherboard). I may go back to 433's and > > overclock them, since the memory access would be faster, but ... > > My chief concern is that the computer I'm getting is a Name Brand computer > (actually with x86 systems that's a first for me). As a result I'm > wondering how bad the parts in it will be (at least it's got a Matrox > graphics card). Somehow I've nightmares of getting it to work well with > Linux. While I haven't yet tried installed Linux on our D*lls, I don't anticipate you'll have any significant trouble. Then again, that could depend more on the distro you choose as it would the platform. I'm a Slackware fan, and it's a fairly seamless installation if you choose a complete install. > Out of curiousity, do you know of any way under Linux to limit a process to > x% of the CPU. So for example a copy of SIMH gets a maximum of 10% of the > CPU time? I started thinking and this is probably better than trying to get > SIMH to 'idle'. The only problem being, I don't see an obvious way to do > this under UNIX. Isn't there some appropriate place in the SIMH code to stick a yield() ? -dq -- Surgically Excise the Pig-Latin from my e-mail address in order to reply... No Tourbots ###### From: Tim Shoppa Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Date: 16 Oct 2001 05:42:35 -0700 Organization: Trailing-Edge Technologies Lines: 44 Message-ID: <9qh9vr02n08@drn.newsguy.com> References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> NNTP-Posting-Host: p-388.newsdawg.com X-Newsreader: Direct Read News 2.90 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!enews.sgi.com!pln-w!spln!dex!extra.newsguy.com!newsp.newsguy.com!drn Xref: chonsp.franklin.ch alt.sys.pdp10:7139 In article , "Zane says... > >Arthur Krewat wrote: >>> same system, the 1Ghz Pentium III. I should know in a couple weeks how well >>> it works out. > >>I'd very much like to hear about this. Right now, I run a dual 533Mhz Celeron, >> and running SIMH at 100% on one cpu, the machine is still very snappy (AGP >> ATI card, 384megs of ram, Abit BP6 motherboard). I may go back to 433's and >> overclock them, since the memory access would be faster, but ... > >My chief concern is that the computer I'm getting is a Name Brand computer >(actually with x86 systems that's a first for me). As a result I'm >wondering how bad the parts in it will be (at least it's got a Matrox >graphics card). Somehow I've nightmares of getting it to work well with >Linux. It'll probably be OK. If it has on-motherboard Ethernet or Sound, you may discover that these don't work out-of-the-box with most Linux distributions. (Although if you're willing to do your own modprobes and/or patch your kernel, you can often get them working months before the mainstream Linux distributions catch up.) Some Name Brand motherboards aren't very well supported by lmsensors (if that's important to you - one site I've set up has 30 Linux boxes and we monitor CPU temperatures on one central machine. Almost as impressive on tours as the "spectral sorter" we used with puched cards when I was an undergrad!) >Out of curiousity, do you know of any way under Linux to limit a process to >x% of the CPU. So for example a copy of SIMH gets a maximum of 10% of the >CPU time? I started thinking and this is probably better than trying to get >SIMH to 'idle'. The only problem being, I don't see an obvious way to do >this under UNIX. There are *some* hooks into kernel scheduling but they were experimental last time I used them (and they probably change every couple of years, so whatever I knew about them under the 0.98 kernel patches is completely inapplicable now I'm sure) that could make per-process limits like this. Look in your kernel sources at the kernel/sched.c module. Is running at niceness +20 workable on your system? (It may not be, depending on what the machine is used for otherwise!) Tim. ###### Sender: phr2001-nospam@ruckus.brouhaha.com From: Paul Rubin Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> Date: 16 Oct 2001 06:53:20 -0700 Message-ID: <7xu1wzk7xr.fsf@ruckus.brouhaha.com> Organization: Nightsong/Fort GNOX Lines: 13 User-Agent: Gnus/5.0807 (Gnus v5.8.7) Emacs/20.7 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 16 Oct 2001 06:56:55 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!enews.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.sys.pdp10:7151 "Zane H. Healy" writes: > My chief concern is that the computer I'm getting is a Name Brand computer > (actually with x86 systems that's a first for me). This isn't such a great idea. Name Brand Vendors try to differentiate themselves from other Name Brand Vendors (and from no-name vendors) by putting Special Unique Crap in their box which they then try to convince you is Good For Something. They include enough Windoze DLL's to hold the thing together if you're running that Evil OS From Redmond, but if you're running Linux you end up with Compatibility Problems. It's better to get generic, no-name stuff that stays with standard interfaces. That's what usually works with the least fuss. ###### From: Arthur Krewat Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Lines: 25 Message-ID: <3BCC4E55.A41A6EA6@bartek.dontspamme.net> References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> <9qh9vr02n08@drn.newsguy.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.78 [en] (X11; U; SunOS 5.8 i86pc) X-Accept-Language: en Date: Tue, 16 Oct 2001 15:17:49 GMT NNTP-Posting-Host: 24.186.100.134 X-Trace: news02.optonline.net 1003245469 24.186.100.134 (Tue, 16 Oct 2001 11:17:49 EDT) NNTP-Posting-Date: Tue, 16 Oct 2001 11:17:49 EDT Organization: Optimum Online Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!cyclone2.usenetserver.com!usenetserver.com!news01.optonline.net!news02.optonline.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7146 Tim Shoppa wrote: > > >Out of curiousity, do you know of any way under Linux to limit a process to > >x% of the CPU. So for example a copy of SIMH gets a maximum of 10% of the > >CPU time? I started thinking and this is probably better than trying to get > >SIMH to 'idle'. The only problem being, I don't see an obvious way to do > >this under UNIX. > > There are *some* hooks into kernel scheduling but they were experimental > last time I used them (and they probably change every couple of years, > so whatever I knew about them under the 0.98 kernel patches is completely > inapplicable now I'm sure) that could make per-process limits like this. > Look in your kernel sources at the kernel/sched.c module. > > Is running at niceness +20 workable on your system? (It may not be, > depending on what the machine is used for otherwise!) I still think that there should be an option in the simulators to time every instruction at a certain cycle period. You could then setup the simulator to almost exactly match the original cycle time of a KS/KL/KI/KA and see what it was really like to run at those speeds. Those of us who like to MONGEN the monitor in record time can ignore it :) aak ###### From: "Zane H. Healy" Subject: Re: SIMH is Int or FP important Newsgroups: alt.sys.pdp10 References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> <9qh9vr02n08@drn.newsguy.com> Organization: Aracnet User-Agent: tin/1.4.4-20000803 ("Vet for the Insane") (UNIX) (Linux/2.2.19 (i686)) Lines: 36 Message-ID: X-Complaints-To: abuse@usenetserver.com X-Abuse-Info: Please be sure to forward a copy of ALL headers X-Abuse-Info: Otherwise we will be unable to process your complaint properly. NNTP-Posting-Date: Tue, 16 Oct 2001 14:32:19 EDT Date: Tue, 16 Oct 2001 18:32:19 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeeds.belnet.be!news.belnet.be!cyclone.bc.net!sjcppf01.usenetserver.com!e420r-sjo4.usenetserver.com!usenetserver.com!sjcpnn01.usenetserver.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7156 Tim Shoppa wrote: > It'll probably be OK. If it has on-motherboard Ethernet or Sound, you > may discover that these don't work out-of-the-box with most Linux distributions. > (Although if you're willing to do your own modprobes and/or patch your > kernel, you can often get them working months before the mainstream The onboard Ethernet and Sound are my big concern, I'd just as soon disable them if possible. However, some of these motherboards that are out now are such junk that you don't seem to be able to. As for messing with the Kernel, no problem, I've been doing that since 0.12, and I prefer to run my own kernel configuration rather than what someone else thinks I should be running. Though the systems are getting fast enough and have enough RAM, I've not bothered to tweak the last couple installs I did (mainly lack of time). > There are *some* hooks into kernel scheduling but they were experimental > last time I used them (and they probably change every couple of years, > so whatever I knew about them under the 0.98 kernel patches is completely > inapplicable now I'm sure) that could make per-process limits like this. > Look in your kernel sources at the kernel/sched.c module. The kernel is a lot different from what it was in the 0.98 days, for one thing there is a LOT more source code :^) I'll have to take a look for this, sounds scary as I prefer not to dig into the source of the kernel if I can help it, but I might find a pointer there. > Is running at niceness +20 workable on your system? (It may not be, > depending on what the machine is used for otherwise!) I'm wanting to us the system to run 3 or more copies of SIMH, plus use it as a workstation. Nicing it might be an option, but I'd really rather limit the percentage of CPU time that a specific process can use. This is just another example of why UNIX sucks. Give me VMS! Zane ###### From: "Zane H. Healy" Subject: Re: SIMH is Int or FP important Newsgroups: alt.sys.pdp10 References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> <7xu1wzk7xr.fsf@ruckus.brouhaha.com> Organization: Aracnet User-Agent: tin/1.4.4-20000803 ("Vet for the Insane") (UNIX) (Linux/2.2.19 (i686)) Lines: 23 Message-ID: <27%y7.151249$Rb.5048089@sjcpnn01.usenetserver.com> X-Complaints-To: abuse@usenetserver.com X-Abuse-Info: Please be sure to forward a copy of ALL headers X-Abuse-Info: Otherwise we will be unable to process your complaint properly. NNTP-Posting-Date: Tue, 16 Oct 2001 14:36:46 EDT Date: Tue, 16 Oct 2001 18:36:46 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!cyclone.bc.net!sjcppf01.usenetserver.com!e420r-sjo4.usenetserver.com!usenetserver.com!sjcpnn01.usenetserver.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7155 Paul Rubin wrote: > "Zane H. Healy" writes: >> My chief concern is that the computer I'm getting is a Name Brand computer >> (actually with x86 systems that's a first for me). > This isn't such a great idea. Name Brand Vendors try to differentiate > themselves from other Name Brand Vendors (and from no-name vendors) by > putting Special Unique Crap in their box which they then try to > convince you is Good For Something. They include enough Windoze DLL's > to hold the thing together if you're running that Evil OS From Redmond, > but if you're running Linux you end up with Compatibility Problems. > It's better to get generic, no-name stuff that stays with standard > interfaces. That's what usually works with the least fuss. Remember, the key thing here is I'm not paying for this system. There is *NO WAY* I'd buy a name brand PC. If I was actually paying for this, I'd be going to the local 'hole in the wall' PC shop, and getting specific parts. Who knows, I might find myself replacing the motherboard with a decent Asus board, and toss in a few pieces I've got laying around. At which point it should be a usable system. Zane ###### From: Tim Shoppa Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Date: 16 Oct 2001 12:37:22 -0700 Organization: Trailing-Edge Technologies Lines: 61 Message-ID: <9qi29i01lo3@drn.newsguy.com> References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> <9qh9vr02n08@drn.newsguy.com> NNTP-Posting-Host: p-436.newsdawg.com X-Newsreader: Direct Read News 2.90 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!12483852!news.imp.ch!sunqbc.risq.qc.ca!cpk-news-hub1.bbnplanet.com!cambridge1-snf1.gtei.net!news.gtei.net!bos-service1.ext.raytheon.com!cyclone.swbell.net!pln-e!spln!dex!extra.newsguy.com!newsp.newsguy.com!drn Xref: chonsp.franklin.ch alt.sys.pdp10:7137 In article , "Zane says... > >Tim Shoppa wrote: >> It'll probably be OK. If it has on-motherboard Ethernet or Sound, you >>may discover that these don't work out-of-the-box with most Linux distributions. >> (Although if you're willing to do your own modprobes and/or patch your >> kernel, you can often get them working months before the mainstream > >The onboard Ethernet and Sound are my big concern, I'd just as soon disable >them if possible. However, some of these motherboards that are out now are >such junk that you don't seem to be able to. I've never seen a motherboard *that* bad. In any event, with lspci and modprobe settings you'd probably be able to prevent Linux from using the broken stuff. (I'm sure someone will come up with a counterexample!) Configuration-wise, I'm most familiar with the recent Pentium III Dell boxes with Intel 82801AA chipsets and I'm relatively happy with them. On-board SCSI tends to be Adaptec-compatible and on-board Ethernet tends to be 3Com compatible. The sound (AC97-derived) wasn't recognized by 'sndconfig' under Redhat 7.0, but it did work if you set it up by hand (and besides, Redhat 7.1 and 7.2 sndconfig do recognize it.) We slap two 4-screen Appian video boards into it (for a total of 9 screens including the on-motherboard Intel-based graphics), put another Ethernet card in, and it all works nicely under Linux. Tim. (Often the problem isn't that Linux cannot use the "new on motherboard" stuff, but that it doesn't recognize that it As for messing with the >Kernel, no problem, I've been doing that since 0.12, and I prefer to run my >own kernel configuration rather than what someone else thinks I should be >running. Though the systems are getting fast enough and have enough RAM, >I've not bothered to tweak the last couple installs I did (mainly lack of >time). > >> There are *some* hooks into kernel scheduling but they were experimental >> last time I used them (and they probably change every couple of years, >> so whatever I knew about them under the 0.98 kernel patches is completely >> inapplicable now I'm sure) that could make per-process limits like this. >> Look in your kernel sources at the kernel/sched.c module. > >The kernel is a lot different from what it was in the 0.98 days, for one >thing there is a LOT more source code :^) I'll have to take a look for >this, sounds scary as I prefer not to dig into the source of the kernel if I >can help it, but I might find a pointer there. > >> Is running at niceness +20 workable on your system? (It may not be, >> depending on what the machine is used for otherwise!) > >I'm wanting to us the system to run 3 or more copies of SIMH, plus use it as >a workstation. Nicing it might be an option, but I'd really rather limit >the percentage of CPU time that a specific process can use. This is just >another example of why UNIX sucks. Give me VMS! > > Zane > > ###### From: wilson@dbit.com (John Wilson) Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Date: 16 Oct 2001 17:50:00 -0400 Organization: D Bit, Troy, NY Lines: 40 Message-ID: <9qia28$jv9$1@dbit.dbit.com> References: <3BCB4787.3A0A681E@bartek.dontspamme.net> <7xu1wzk7xr.fsf@ruckus.brouhaha.com> NNTP-Posting-Host: dbit.com X-Original-NNTP-Posting-Host: dbit.com X-Trace: 16 Oct 2001 18:02:29 -0400, dbit.com XPident: wilson X-Original-NNTP-Posting-Host: 216.153.192.72 XPident: Unknown Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed.nextra.ch!news.nextra.ch!nextra.at!newsfeed2.skycache.com!newsfeed1.cidera.com!Cidera!news.wizvax.net!dbit.com!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7133 In article <7xu1wzk7xr.fsf@ruckus.brouhaha.com>, Paul Rubin wrote: >This isn't such a great idea. Name Brand Vendors try to differentiate >themselves from other Name Brand Vendors (and from no-name vendors) by >putting Special Unique Crap in their box which they then try to >convince you is Good For Something. They include enough Windoze DLL's >to hold the thing together if you're running that Evil OS From Redmond, >but if you're running Linux you end up with Compatibility Problems. > >It's better to get generic, no-name stuff that stays with standard >interfaces. That's what usually works with the least fuss. Hear hear. I've been bitten a bunch of times by big-name stuff that isn't quite compatible with the industry standards. With the no-name crap, standards compliance is all they have going for them so they tend to do a really good job. Of course, it still pays to be picky about which no-name crap you buy, some of it really is crap but a lot of it is way better than the big-name stuff. Also speaking of big names, is it just me or is Red Hat not very stable? Things usually seem to go better with Slackware, it's been around longer and just seems happier all around. Re Linux scheduling, last time I looked into this was for SMP systems (different situation though, we wanted to ensure that Ersatz-11's mainline thread *did* get 100% of one CPU so that the I/O threads could duke it out on the other(s) w/o affecting emulated CPU performance), and there was a set of scheduler patches floating around (for 2.2.12 IIRC) which enabled you to guarantee various things, but they had apparently been vetoed by Linus so they were never going to go into the standard distrib. That's the funny thing about supposedly open-source software, it's only open if you talk the powers that be into merging your changes into the central source tree, otherwise you're just making local hacks which will go nowhere. I don't know if that means he's completely opposed to his scheduler getting dorked up with special-purpose constraints, or whether it was just that one particular package that irritated him. In any case, if you can have some fun with sched_yield() every so often it might help. John Wilson D Bit ###### From: Tim Shoppa Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Date: 16 Oct 2001 19:40:29 -0700 Organization: Trailing-Edge Technologies Lines: 24 Message-ID: <9qir2t01137@drn.newsguy.com> References: <3BCB4787.3A0A681E@bartek.dontspamme.net> <7xu1wzk7xr.fsf@ruckus.brouhaha.com> <9qia28$jv9$1@dbit.dbit.com> NNTP-Posting-Host: p-932.newsdawg.com X-Newsreader: Direct Read News 2.90 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeeds.belnet.be!news.belnet.be!opentransit.net!cpk-news-hub1.bbnplanet.com!news.gtei.net!nntp.abs.net!pln-e!spln!dex!extra.newsguy.com!newsp.newsguy.com!drn Xref: chonsp.franklin.ch alt.sys.pdp10:7136 In article <9qia28$jv9$1@dbit.dbit.com>, wilson@dbit.com says... >Also speaking of big names, is it just me or is Red Hat not very stable? >Things usually seem to go better with Slackware, it's been around longer >and just seems happier all around. Redhat seriously confused a whole bunch of folks in the 6.0-6.2 time frame with their funny kernel mods that needed a specially-Redhat-patched version of gcc to build right. The situation has improved greatly since 7.0 came out, though they still make some choices about gcc patches that not everyone agrees with. The Redhat folks have done an amazing amount of work to make a distribution that will install a working Linux onto some of the most oddball hardware you've ever imagined without needing anything like a kernel build. The Slackware stuff comes without all the warts needed to do this. If you want to go one step further, go to www.linuxfromscratch.org and build your own system purely from sources. It takes a couple of hours on a modernish system, and the glibc build is a pretty heavy-duty workout of your CPU. (I've seen a couple of name-brand boxes go into thermal shutdown doing a glibc build!) Tim. ###### Sender: prep@k9 Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> <9qh9vr02n08@drn.newsguy.com> From: Paul Repacholi Message-ID: <87bsj6b9n5.fsf@prep.synonet.com> Lines: 16 User-Agent: Gnus/5.0808 (Gnus v5.8.8) Emacs/20.7 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Organization: iQnet Cache-Post-Path: osb-serv.angelina.pe!unknown@p066.perth01.dial.usertools.net X-Cache: nntpcache 2.3.3 (see http://www.nntpcache.org/) Cache-Post-Path: news.satix.net!unknown@news-01.core.usertools.net X-Cache: nntpcache 2.3.3 (see http://www.nntpcache.org/) NNTP-Posting-Host: news.satix.net X-Trace: ozemail.com.au 1003335877 203.132.96.3 (Thu, 18 Oct 2001 02:24:37 EST) NNTP-Posting-Date: Thu, 18 Oct 2001 02:24:37 EST Distribution: world Date: 17 Oct 2001 22:53:34 +0800 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!enews.sgi.com!newshub2.rdc1.sfba.home.com!news.home.com!sjc1.nntp.concentric.net!newsfeed.concentric.net!newsfeed.ozemail.com.au!ozemail.com.au!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7143 "Zane H. Healy" writes: > The kernel is a lot different from what it was in the 0.98 days, for > one thing there is a LOT more source code :^) I'll have to take a > look for this, sounds scary as I prefer not to dig into the source > of the kernel if I can help it, but I might find a pointer there. If you look on /., there is a bit on the Linux census. Follow that, and they have a pointer to an archive of many of the linux kernels. -- Paul Repacholi 1 Crescent Rd., +61 (08) 9257-1001 Kalamunda. West Australia 6076 Raw, Cooked or Well-done, it's all half baked. EPIC, The Architecture of the future, always has been, always will be. ###### From: jmfbahciv@aol.com Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Date: Fri, 26 Oct 01 07:49:20 GMT Organization: UltraNet Communications, Inc. Lines: 22 Message-ID: <9rberh$f4h$3@bob.news.rcn.net> References: <3BCB9EAE.7694CA37@bartek.dontspamme.net> <9r9tbm$tki$1@dbit.dbit.com> X-Trace: UmFuZG9tSVbMSIwwMDgVr1+BI1j7bOu0gZfGCwRO02FJylyvE/GmsLhiMGDSA0UJ X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 26 Oct 2001 10:45:05 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newscore.univie.ac.at!194.74.65.73.MISMATCH!btnet-peer0!btnet!newsfeed.mathworks.com!news.voicenet.com!feed2.news.rcn.net!feed1.news.rcn.net!rcn!207-172-216-110 Xref: chonsp.franklin.ch alt.sys.pdp10:7280 In article <9r9tbm$tki$1@dbit.dbit.com>, wilson@dbit.com (John Wilson) wrote: >In article , >Eric Smith wrote: >>My HP-35 and HP-41 calculator microcode-level simulators use SIGITIMER >>to wake up every .1 seconds, and execute 35 and 65 microinstructions, >>respectively, so that they run at the exact (nominal) timing of the real >>calculators. (Exact because all microinstructions take a fixed amount >>of time on the real hardware.) >Most people don't want to run slowly just for old >time's sake .... That certainly isn't keeping in the spirit of TOPS-10. /BAH Subtract a hundred and four for e-mail. ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> <3BCB9EAE.7694CA37@bartek.dontspamme.net> Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. Date: 24 Oct 2001 11:08:43 -0700 Message-ID: Lines: 17 User-Agent: Gnus/5.0807 (Gnus v5.8.7) Emacs/20.7 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 24 Oct 2001 11:13:49 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!enews.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.sys.pdp10:7289 Arthur Krewat writes: > Somehow, I think it would be better to alter SIMH and TS10 to allow a certain > amount of (configurable) sleep time for every instruction, Unless you want it to run *very* slowly, you don't want to sleep after every instruction. Do it after every 1000 instructions, or perhaps even 10,000. The overhead for a system call is huge compared to the time to simulate one PDP-10 instruction. My HP-35 and HP-41 calculator microcode-level simulators use SIGITIMER to wake up every .1 seconds, and execute 35 and 65 microinstructions, respectively, so that they run at the exact (nominal) timing of the real calculators. (Exact because all microinstructions take a fixed amount of time on the real hardware.) Even on a Pentium 90, those simulators make no noticable contribution to the load average. ###### From: Arthur Krewat Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Organization: Kilonet.net Lines: 16 Message-ID: <3BD70C8D.6D685ABB@bartek.dontspamme.net> References: <7xofn91lcw.fsf@ruckus.brouhaha.com> <3BCB4787.3A0A681E@bartek.dontspamme.net> <3BCB9EAE.7694CA37@bartek.dontspamme.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.78 [en] (X11; U; SunOS 5.8 i86pc) X-Accept-Language: en Date: Wed, 24 Oct 2001 18:52:51 GMT NNTP-Posting-Host: 24.186.100.134 X-Trace: news02.optonline.net 1003949571 24.186.100.134 (Wed, 24 Oct 2001 14:52:51 EDT) NNTP-Posting-Date: Wed, 24 Oct 2001 14:52:51 EDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.tesion.net!news.belwue.de!fu-berlin.de!cpk-news-hub1.bbnplanet.com!news.gtei.net!newsfeed1.cidera.com!Cidera!cyclone2.usenetserver.com!usenetserver.com!news01.optonline.net!news02.optonline.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7284 Eric Smith wrote: > > Arthur Krewat writes: > > Somehow, I think it would be better to alter SIMH and TS10 to allow a certain > > amount of (configurable) sleep time for every instruction, > > Unless you want it to run *very* slowly, you don't want to sleep after > every instruction. Do it after every 1000 instructions, or perhaps even > 10,000. The overhead for a system call is huge compared to the time to > simulate one PDP-10 instruction. Yeah, I was looking for a timer with a small resolution, but didn't find anything too usefull and cross-platform. Your idea makes more sense, especially with the system call load. aak ###### From: wilson@dbit.com (John Wilson) Newsgroups: alt.sys.pdp10 Subject: Re: SIMH is Int or FP important Date: 25 Oct 2001 16:40:22 -0400 Organization: D Bit, Troy, NY Lines: 48 Message-ID: <9r9tbm$tki$1@dbit.dbit.com> References: <3BCB9EAE.7694CA37@bartek.dontspamme.net> NNTP-Posting-Host: dbit.com X-Original-NNTP-Posting-Host: dbit.com X-Trace: 25 Oct 2001 16:47:00 -0400, dbit.com XPident: wilson X-Original-NNTP-Posting-Host: 216.153.192.72 XPident: Unknown Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeeds.belnet.be!news.belnet.be!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!netnews.com!xfer02.netnews.com!newsfeed1.cidera.com!Cidera!news.wizvax.net!dbit.com!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:7276 In article , Eric Smith wrote: >My HP-35 and HP-41 calculator microcode-level simulators use SIGITIMER >to wake up every .1 seconds, and execute 35 and 65 microinstructions, >respectively, so that they run at the exact (nominal) timing of the real >calculators. (Exact because all microinstructions take a fixed amount >of time on the real hardware.) This gets the average execution rate but it's very jittery, and you can have trouble with latency (even slow machines can answer interrupts very fast with good coding). On Linux/x86, setitimer()'s resolution is pretty course (10 msec) so it's tough to get fine-grained precision. I don't know a 100% perfect answer to this problem. E11 has a "SET THROTTLE" command which allows inserting an X-microsecond delay on every Yth emulated instruction fetch, which is still pretty kludgey (you have to tweak X and Y until you get the speed you want), but usually people can tune up the behavior they want. Most people don't want to run slowly just for old time's sake (we all remember what it felt like to work on dog-slow machines all day, and we're glad we don't have to any more), they do it because they have software timing loops (or other race conditions), so it just needs to be at least slow enough to avoid that problem, w/o being too slow to be useful. Sometimes people really do depend on precise timing that would have to be emulated dead-on, but it's rare. Luckily DEC made a wide range of different CPUs so people who depended too much on the behavior of one model already got bitten decades ago, and rooted dependences on CPU timing out of programs they cared about. I remember reading ages ago about a high-speed add-on CPU card for the Apple ][, which would slow down to normal speed for the next N milliseconds after each I/O access was made. Cute thought, if only it were really that simple... This all seems too familiar, on our old RSTS system some of the real-time games were too difficult when run on an unloaded system, so people would often log in on two terminals, type 10 GOTO 10 / RUN on one of them, and play the game on the other. Yuck! That was gross even then (but RSTS's timer granularity is 1 second, so it's hard to do small delays). I used to think it might be handy to build a trivial Unibus device with a bunch of no-op ports which delay various amounts before hitting SSYN (like, one each for 500 nsec, 1 usec, 2 usec, 4 usec), so that you could whip up short but fairly precise delays with a few TSTs. I wonder if emulation will become important enough that we'll start seeing user-mode instructions that do something like that on modern CPUs... As long as they're interruptable (a single memory cycle that takes 4 usec would be very bad news these days), why not. John Wilson D Bit