Pentium Pro cooks fast!

Date	1997-04-19 18:26
From	Robin Whittle
Subject	Pentium Pro cooks fast!
	Further to the discussion of CPUs a while ago, I just fired up my new Pentium Pro 180 MHz (a 150 with attention to cooling - it runs perfectly at 180 but not at all at 200 . . .) I can't compare its performance to a Pentium, but it runs rings around the 486 100 MHz which I have been using for over a year. It is between 5.6 and 6 times the speed! My biggest piece has 17 complex sound generators (8 sine waves, each with filtered random k rate noise for AM and FM, then summed and distorted) which go into 17 binaural processes. It goes for 8 minutes 40 seconds. Time to cook at full quality (44.1 kHz, ksmps = 3): 486 100 MHz 20Meg RAM 36 hours PPro 180 MHz 64 Meg RAM 6 hours! This is using the same binary (compiled with DJGPP). The PPro did the same time under MSDOS as in a DOS box inside Win95. The actual figures are only about 2 minutes from 36 and 6 hours shown above. Another less complex piece showed a speed factor of 5.6 : 1 Since Csound can't reasonobly be made to run with threads on multiple CPUs, the next stop is a DEC Alpha, or some horrendously expensive workstation from SGI. Time for me to stop worrying about CPU speed for a while and cook up some more music! Cooling fans for these things are noisy, and I have organised the cabinet, power supply and CPU fans to be very quiet. The hard disk is in its own thick aluminium case with its own cooling system. Consequently the computer is physically very quiet. I should have my Zefiro digital audio ISA bus card soon (www.zefiro.com). Dreaming now of a small squadron of Alphas in the back room, on the LAN, with chunky power supplies and fans making all the noise they like. The latest Power PC chip seems to be quite savage with floating point too. - Robin . Robin Whittle . . http://www.ozemail.com.au/~firstpr firstpr@ozemail.com.au . . 11 Miller St. Heidelberg Heights 3081 Melbourne Australia . . Ph +61-3-9459-2889 Fax +61-3-9458-1736 . . Consumer advocacy in telecommunications, especially privacy . . . . First Principles - Research and expression - music, . . music industry, telecommunications . . human factors in technology adoption. . . . Real World Interfaces - Hardware and software, especially . . for music . Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa07084; 20 Apr 97 18:34 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id aa28862; 20 Apr 97 18:33 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Sun, 20 Apr 1997 18:33:09 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (SAA29643); Sun, 20 Apr 1997 18:28:56 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Sun, 20 Apr 97 18:28:44 +0100 Received: from server3.syd.mail.ozemail.net [203.108.7.41] by hermes via ESMTP (SAA29632); Sun, 20 Apr 1997 18:28:42 +0100 Received: from oznet07.ozemail.com.au (oznet07.ozemail.com.au [203.2.192.122]) by server3.syd.mail.ozemail.net (8.8.4/8.6.12) with ESMTP id DAA18751; Mon, 21 Apr 1997 03:27:19 +1000 (EST) Received: from dialup01.interdomain.net.au (dialup01.interdomain.net.au [203.17.167.131]) by oznet07.ozemail.com.au (8.8.4/8.6.12) with SMTP id DAA08573; Mon, 21 Apr 1997 03:27:16 +1000 (EST) Message-Id: <199704201727.DAA08573@oznet07.ozemail.com.au> Comments: Authenticated sender is From: Robin Whittle To: csound@maths.ex.ac.uk Date: Sun, 20 Apr 1997 03:26:18 +1000 Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7BIT Subject: Re: 3D Sound Opcode? Cc: Peter Kearton Priority: normal X-Mailer: Pegasus Mail for Windows (v2.23) Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk Pete Kearton wrote: > Pablo Sotuyo Blanco and Robin Whittle have been discussing > binaural opcode, but I was under the impression that HRTF opcode > based on MIT's compact data is included in the latest postings of > CSound, although it does not work on the PC platform yet. Am I wrong > in this ? There is a Head Related Transfer Function opcode in the latest release, but I have not tried it. HRTFs present serious and probably insurmountable problems when interpolating between them for angles other than those which have been sampled. Also, I think they would be very CPU intensive. My ugen does more than just the HRTF function of replicating a sound source a fixed distance from the head. I want to review my code before making it available, and I think it will take me about two weeks before I can do that. I will post to the list when I have it ready. I hope to make available MSDOS and Linux binaries of the latest Csound version, with the binaural ugen and my other ugens for those who are not into recompiling Csound. - Robin . Robin Whittle . . http://www.ozemail.com.au/~firstpr firstpr@ozemail.com.au . . 11 Miller St. Heidelberg Heights 3081 Melbourne Australia . . Ph +61-3-9459-2889 Fax +61-3-9458-1736 . . Consumer advocacy in telecommunications, especially privacy . . . . First Principles - Research and expression - music, . . music industry, telecommunications . . human factors in technology adoption. . . . Real World Interfaces - Hardware and software, especially . . for music . Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa08654; 21 Apr 97 5:58 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id am09908; 21 Apr 97 5:57 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Sun, 20 Apr 1997 21:10:35 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (VAA03781); Sun, 20 Apr 1997 21:07:55 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Sun, 20 Apr 97 21:07:47 +0100 Received: from mailrelay1.cc.columbia.edu [128.59.35.143] by hermes via ESMTP (VAA03775); Sun, 20 Apr 1997 21:07:45 +0100 Received: from dogma.columbia.edu (dialup-8-2.cc.columbia.edu [128.59.36.190]) by mailrelay1.cc.columbia.edu (8.8.5/8.8.5) with SMTP id QAA13804 for ; Sun, 20 Apr 1997 16:07:45 -0400 (EDT) Message-Id: <335A7718.2AA2@columbia.edu> Date: Sun, 20 Apr 1997 16:05:44 -0400 From: Gregory Boduch Reply-To: gb141@columbia.edu Organization: http://www.juilliard.edu http://www.columbia.edu X-Mailer: Mozilla 3.01Gold (Win95; I) Mime-Version: 1.0 To: csound@maths.ex.ac.uk Subject: Re: Pentium Pro cooks fast! References: <199704201727.DAA08598@oznet07.ozemail.com.au> Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk Content-type: multipart/mixed; boundary="------------287216BBB57" --------------287216BBB57 Content-type: text/plain; charset="us-ascii" Content-transfer-encoding: 7bit Robin Whittle wrote: > > > I can't compare its performance to a Pentium, but it runs rings > around the 486 100 MHz which I have been using for over a year. > It is between 5.6 and 6 times the speed! > As a side note, there seems to be a problem with the Pentium Pro cpu. A Linux benchmark called Oscillates, which simply tries to send as many wavetable-lookup oscillators as it can to the dsp, returns a score of 42 oscillators on a P200 w/ 32Meg, 36 oscilators on a P166 MMX, and only 28 on a PPro 200 w/ 32Meg and 27 on a dual PPro 180 w/ 32Meg. Any thoughts? The code for the benchmark by Dave Topper is included. GB --------------287216BBB57 Content-Disposition: inline; filename="oscillates.c" Content-type: text/plain; charset="us-ascii"; name="oscillates.c" Content-transfer-encoding: 7bit #include #include #include #include #include #define BUFSIZE 8192 #define SR_RATE 44100 #define NBUFS 4 #define WTABLESIZE 1000 #define FREQ 278.0 #define NOSCILS 1000 void makewave(short wtable[]) { int i; for(i = 0; i < WTABLESIZE; i++) { wtable[i] = sin((double) (2PI (float)i/WTABLESIZE)) * 32767.0; } return; } short oscil(float si, short wave[], float phs) { int i; i = phs; phs += si; if(phs >= WTABLESIZE) phs -= WTABLESIZE; return(wave[i]); } main () { int len,format,stereo,speed,cycle,frag_size; int audio_fd; int arg; short waveform[WTABLESIZE]; short sampbuff[NBUFS][BUFSIZE]; int i,j,k,bufno; int active; float si[NOSCILS],phase[NOSCILS]; / Open the audio port / if ((audio_fd = open("/dev/dsp", O_WRONLY, 0)) == -1) { perror("/dev/dsp"); exit(1); } / Set the fragment size / arg = 0x777f000d; if (ioctl(audio_fd, SNDCTL_DSP_SETFRAGMENT, &arg)==-1) { perror("incorrect fragment size"); exit(1); } / Put the audio port in "sync" mode / ioctl(audio_fd, SNDCTL_DSP_SYNC, 0); / Set the format for output samples / format = AFMT_S16_LE; if (ioctl(audio_fd, SNDCTL_DSP_SETFMT, &format)==-1) { perror("SNDCTL_DSP_SETFMT"); exit(1); } / Set stereo / mono (1/0) / stereo = 0; if (ioctl(audio_fd, SNDCTL_DSP_STEREO, &stereo)==-1) { perror("SNDCTL_DSP_STEREO"); exit(1); } / Set sample rate / speed = SR_RATE; if (ioctl(audio_fd, SNDCTL_DSP_SPEED, &speed)==-1) { / Fatal error / perror("SNDCTL_DSP_SPEED"); exit(1); } / Get the size of the audio buffer / / Note that the driver computes this optimally / if (ioctl(audio_fd, SNDCTL_DSP_GETBLKSIZE, &frag_size) == -1) { exit(1); } printf("Audio buffer: %d\n",frag_size); / make a wavetable for the oscillator / makewave(waveform); / load buffers and write them to the audio port / for (i = 0; i < NOSCILS; i++) { si[i] = (FREQ + (i5)) * WTABLESIZE/SR_RATE; phase[i] = 0.0; } bufno = 0; active = 1; for (i = 0; i < 99999; i++) { /* fill a buffer with sound / for (k = 0; k < active; k++) { if (k == 0) { for (j = 0; j < BUFSIZE; j++) { sampbuff[bufno][j] = oscil(si[k],waveform,&phase[k]) 1.0/active; } } else { for (j = 0; j < BUFSIZE; j++) { sampbuff[bufno][j] += oscil(si[k],waveform,&phase[k]) * 1.0/active; } } } if ((len = write(audio_fd, sampbuff[bufno], 2BUFSIZE)) == -1) { perror("audio write"); exit(1); } / shift to the next buffer / if (++bufno > 3) { if (++active > NOSCILS) active -= 1; printf("%d oscillators, daddy-oh!\n",active); bufno = 0; } } / Close the audio port / close(audio_fd); } --------------287216BBB57-- Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa08731; 21 Apr 97 6:02 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id ac10278; 21 Apr 97 6:01 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Sun, 20 Apr 1997 23:20:23 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (XAA06802); Sun, 20 Apr 1997 23:12:35 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Sun, 20 Apr 97 23:12:28 +0100 Received: from root@big.fishnet.net [205.216.133.3] by hermes via ESMTP (XAA06799); Sun, 20 Apr 1997 23:12:22 +0100 Received: from sinewave (port082.vta.fishnet.net [205.216.133.231]) by big.fishnet.net (8.7.5/8.6.9) with SMTP id OAA06328 for ; Sun, 20 Apr 1997 14:16:19 -0700 Message-Id: <335A8745.1954C581@rcsreg.com> Date: Sun, 20 Apr 1997 14:14:45 -0700 From: Toby X-Mailer: Mozilla 3.01Gold (X11; I; Linux 2.0.24 i486) Mime-Version: 1.0 To: csound@maths.ex.ac.uk Subject: Re: Pentium Pro cooks fast! References: <199704201727.DAA08598@oznet07.ozemail.com.au> <335A7718.2AA2@columbia.edu> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk > Linux benchmark called Oscillates, which simply tries to send as many > wavetable-lookup oscillators as it can to the dsp, returns a score of 42 > oscillators on a P200 w/ 32Meg, 36 oscilators on a P166 MMX, and only > 28 on a PPro 200 w/ 32Meg and 27 on a dual PPro 180 w/ 32Meg. > > Any thoughts? Well, I WAS* happy with my 486DX2/80 VLB, but it only goes to 7 oscilators with 'oscillates' before interruptions are heard. You just cost me about $500 (If I had it, anyway). My condolences to those with the dual PPro's. Jeesh Toby -There otta be a law- Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa08826; 21 Apr 97 6:22 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id ab10667; 21 Apr 97 6:22 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Mon, 21 Apr 1997 00:59:03 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (AAA09368); Mon, 21 Apr 1997 00:57:23 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Mon, 21 Apr 97 00:57:18 +0100 Received: from server3.syd.mail.ozemail.net [203.108.7.41] by hermes via ESMTP (AAA09360); Mon, 21 Apr 1997 00:57:15 +0100 From: ggerrard@ozemail.com.au Received: from oznet02.ozemail.com.au (oznet02.ozemail.com.au [203.2.192.124]) by server3.syd.mail.ozemail.net (8.8.4/8.6.12) with ESMTP id JAA25622 for ; Mon, 21 Apr 1997 09:57:17 +1000 (EST) Received: from [203.108.202.61] (slmel13p45.ozemail.com.au [203.108.202.61]) by oznet02.ozemail.com.au (8.8.4/8.6.12) with SMTP id JAA17651; Mon, 21 Apr 1997 09:57:08 +1000 (EST) Message-Id: <199704202357.JAA17651@oznet02.ozemail.com.au> Subject: Re: Pentium Pro cooks fast! Date: Mon, 21 Apr 97 09:56:24 +1000 X-Mailer: Claris Emailer 2.0, March 15, 1997 To: Robin Whittle , Csound Mailing List Mime-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk Robin Whittle >Further to the discussion of CPUs a while ago, I just fired up my new >Pentium Pro 180 MHz (a 150 with attention to cooling - it runs perfectly >at 180 but not at all at 200 . . .) > >I can't compare its performance to a Pentium, but it runs rings >around the 486 100 MHz which I have been using for over a year. > >It is between 5.6 and 6 times the speed! > >My biggest piece has 17 complex sound generators (8 sine waves, each >with filtered random k rate noise for AM and FM, then summed and >distorted) which go into 17 binaural processes. It goes for 8 >minutes 40 seconds. > >Time to cook at full quality (44.1 kHz, ksmps = 3): > > 486 100 MHz 20Meg RAM 36 hours > > PPro 180 MHz 64 Meg RAM 6 hours! >..[snip] Now, i don't wanna start a war or anything... But it may be interesting to run one of these tests on a Mac to see how long it takes. If you could send me the .sco and .orc files I'll do it while I peacefully slumber. For such large amounts of data, I would expect disk speed to be a factor, not just raw cpu speed. Graeme Gerrard ggerrard@ozemail.com.auDELETE_THIS Resonant Multimedia "Back off man, I'm a scientist" - Dan Ackroyd ph +61 3 9525 7869 Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa08844; 21 Apr 97 6:22 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id ad10667; 21 Apr 97 6:22 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Mon, 21 Apr 1997 01:55:23 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (BAA10561); Mon, 21 Apr 1997 01:52:52 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Mon, 21 Apr 97 01:52:47 +0100 Received: from relay-11.mail.demon.net [194.217.242.137] by hermes via SMTP (BAA10558); Mon, 21 Apr 1997 01:52:46 +0100 Received: from rwentk.demon.co.uk ([158.152.68.129]) by relay-10.mail.demon.net id aa1023304; 21 Apr 97 0:55 BST Message-Id: <3.0.32.19970420224718.006ad9e8@sdps.demon.co.uk> X-Sender: rwentk@sdps.demon.co.uk X-Mailer: Windows Eudora Pro Version 3.0 (32) Date: Mon, 21 Apr 1997 00:54:47 +0100 To: csound@maths.ex.ac.uk From: Richard Wentk Subject: Re: Pentium Pro cooks fast! Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk At 16:05 20/04/97 -0400, you wrote: >Robin Whittle wrote: >> >> >> I can't compare its performance to a Pentium, but it runs rings >> around the 486 100 MHz which I have been using for over a year. >> It is between 5.6 and 6 times the speed! >> > >As a side note, there seems to be a problem with the Pentium Pro cpu. A >Linux benchmark called Oscillates, which simply tries to send as many >wavetable-lookup oscillators as it can to the dsp, returns a score of 42 >oscillators on a P200 w/ 32Meg, 36 oscilators on a P166 MMX, and only >28 on a PPro 200 w/ 32Meg and 27 on a dual PPro 180 w/ 32Meg. > >Any thoughts? The code for the benchmark by Dave Topper is included. As I understand it this comes down to cache management and the efficiency of the branch prediction. And we already know that in a Windows environment the PPro is worse at legacy 16-bit code than a Pentium. I once came across a case where someone was convinced a Pentium was slower than a 486, because it ran their own home-brew bench mark slower. It turned out that this rather special test was tripping up the Pentium's branch prediction scheme. This is usually more efficient when dealing with 'real' code, but incurs a bigger time penalty than the 486 when it gets it wrong. As a result the 486 was actually faster in this one particular accidentally contrived situation. Of course in real life, with real apps, this would never (well - hardly ever) happen. Aside from this - were these machines identical in other ways? Particularly, was the memory the same type and speed? SDRAM, EDO and standard RAM all make a difference. Some PCs run with 60nS and others 70nS RAM, and real speed freaks can get 50nS. Also - how closely can you map Oscillates to a Csound run? Csound seems like a more general and wide-ranging workout for a processor. I'd guess that the only way to see how well a processor does is to try it out. There's only a 'problem' if it's a problem for real apps. In any case, I'd be interested to see how well the K6 does with this benchmark just as a rough comparison. Caching, branch prediction and instruction decoding are all supposed to be improved over the Pentium and PPro. I'll be getting a K6 in the next month or three, and it will be fun to compare Robin's timings with that and my current P133. R. Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa08850; 21 Apr 97 6:22 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id ae10667; 21 Apr 97 6:22 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Mon, 21 Apr 1997 02:03:43 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (CAA10844); Mon, 21 Apr 1997 02:01:40 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Mon, 21 Apr 97 02:01:35 +0100 Received: from mail2.digital.com [204.123.2.56] by hermes via SMTP (CAA10839); Mon, 21 Apr 1997 02:01:33 +0100 Received: from matriarch.shl.dec.com by mail2.digital.com (5.65 EXP 4/12/95 for V3.2/1.0/WV) id AA18293; Sun, 20 Apr 1997 17:56:15 -0700 Received: from gobbledoc.shl.dec.com by matriarch.shl.dec.com; (5.65v3.2/1.1.8.2/16Aug96-1226PM) id AA04465; Mon, 21 Apr 1997 10:55:16 +1000 Received: by gobbledoc.shl.dec.com with SMTP (Microsoft Exchange Server Internet Mail Connector Version 4.0.994.63) id <01BC4E42.40540EF0@gobbledoc.shl.dec.com>; Mon, 21 Apr 1997 10:53:31 +1000 Message-Id: From: Greg Sullivan To: 'Csound Mailing List' Subject: RE: Pentium Pro cooks fast! Date: Mon, 21 Apr 1997 10:53:30 +1000 X-Mailer: Microsoft Exchange Server Internet Mail Connector Version 4.0.994.63 Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk Robin Whittle wrote: > >Dreaming now of a small squadron of Alphas in the back room, on the >LAN, with chunky power supplies and fans making all the noise they >like. Don't be selfish - put them on the Internet where we can all use them. Greg. > Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa08898; 21 Apr 97 6:23 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id ab10742; 21 Apr 97 6:22 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Mon, 21 Apr 1997 00:50:58 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (AAA09089); Mon, 21 Apr 1997 00:47:42 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Mon, 21 Apr 97 00:47:36 +0100 Received: from mail2.digital.com [204.123.2.56] by hermes via SMTP (AAA09084); Mon, 21 Apr 1997 00:47:35 +0100 Received: from matriarch.shl.dec.com by mail2.digital.com (5.65 EXP 4/12/95 for V3.2/1.0/WV) id AA19535; Sun, 20 Apr 1997 16:40:18 -0700 Received: from gobbledoc.shl.dec.com by matriarch.shl.dec.com; (5.65v3.2/1.1.8.2/16Aug96-1226PM) id AA04279; Mon, 21 Apr 1997 09:39:20 +1000 Received: by gobbledoc.shl.dec.com with SMTP (Microsoft Exchange Server Internet Mail Connector Version 4.0.994.63) id <01BC4E37.A4BC6870@gobbledoc.shl.dec.com>; Mon, 21 Apr 1997 09:37:35 +1000 Message-Id: From: Greg Sullivan To: 'csound' Subject: RE: Pentium Pro cooks fast! Date: Mon, 21 Apr 1997 09:37:33 +1000 X-Mailer: Microsoft Exchange Server Internet Mail Connector Version 4.0.994.63 Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk It's conceivable that some code in this benchmark consists of 16-bit instructions. Such instructions run slower on a Pentium Pro, I think, than a Pentium of the same clock rate. This is the reason that the Pro isn't really recommended for running Windows 95, because Windows 95 retains a significant amount of old 16-bit code. That doesn't mean that the Pro won't run Windows 95 fast, it just means that it's cheaper to get the same performance with a fast Pentium. This is just an idea - I know nothing about this benchmark program. Greg. >---------- >From: Gregory Boduch[SMTP:gb141@columbia.edu] >Sent: Monday, April 21, 1997 6:05 AM >To: csound@noether.ex.ac.uk >Subject: Re: Pentium Pro cooks fast! > ><> >Robin Whittle wrote: >> >> >> I can't compare its performance to a Pentium, but it runs rings >> around the 486 100 MHz which I have been using for over a year. >> It is between 5.6 and 6 times the speed! >> > >As a side note, there seems to be a problem with the Pentium Pro cpu. A >Linux benchmark called Oscillates, which simply tries to send as many >wavetable-lookup oscillators as it can to the dsp, returns a score of 42 >oscillators on a P200 w/ 32Meg, 36 oscilators on a P166 MMX, and only >28 on a PPro 200 w/ 32Meg and 27 on a dual PPro 180 w/ 32Meg. > >Any thoughts? The code for the benchmark by Dave Topper is included. > >GB > > Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa09245; 21 Apr 97 6:53 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id aa11615; 21 Apr 97 6:52 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Mon, 21 Apr 1997 06:52:46 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (GAA17497); Mon, 21 Apr 1997 06:50:01 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Mon, 21 Apr 97 06:49:55 +0100 Received: from whale.gu.net [194.93.190.4] by hermes via ESMTP (GAA17494); Mon, 21 Apr 1997 06:49:51 +0100 Received: from creator.gu.kiev.ua (creator.gu.kiev.ua [194.93.190.3]) by whale.gu.kiev.ua (8.8.5/8.7.3) with ESMTP id IAA114972; Mon, 21 Apr 1997 08:49:35 +0300 Received: from frya.zgik.zaporizhzhe.ua (frya.zgik.zaporizhzhe.ua [193.124.62.254]) by creator.gu.kiev.ua with ESMTP id IAA14778; Mon, 21 Apr 1997 08:46:30 +0300 (EET DST) Received: from mlt.zaporizhzhe.ua by frya.zgik.zaporizhzhe.ua with ESMTP id IAA23676; (8.7.5/vak+evp/1.8e) Mon, 21 Apr 1997 08:38:44 +0300 (EET DST) Received: from snt.mlt.zaporizhzhe.ua by mlt.zaporizhzhe.ua with ESMTP id IAA28995; (8.7.5/vak+evp/1.8e) Mon, 21 Apr 1997 08:37:43 +0300 (EET DST) Received: from snt.vadims by snt.mlt.zaporizhzhe.ua with SMTP id HAA03135; (8.7.5/zah/1.4b) Mon, 21 Apr 1997 07:34:30 +0300 (EET DST) Message-Id: <335AEE8E.18B5@snt.mlt.zaporizhzhe.ua> Date: Mon, 21 Apr 1997 05:35:26 +0100 From: "Vadim V. Sytnikov" Reply-To: vadims@snt.mlt.zaporizhzhe.ua Organization: Gamos JSC, Moscow, Russia X-Mailer: Mozilla 3.0 (Win95; U) Mime-Version: 1.0 To: gb141@columbia.edu Cc: CSound Mailing List Subject: CPU Benchmarks References: <199704201727.DAA08598@oznet07.ozemail.com.au> <335A7718.2AA2@columbia.edu> Content-Type: text/plain; charset=koi8-r Content-Transfer-Encoding: 7bit Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk Gregory Boduch wrote: > As a side note, there seems to be a problem with the Pentium Pro cpu. A > Linux benchmark called Oscillates, which simply tries to send as many > wavetable-lookup oscillators as it can to the dsp, returns a score of 42 > oscillators on a P200 w/ 32Meg, 36 oscilators on a P166 MMX, and only > 28 on a PPro 200 w/ 32Meg and 27 on a dual PPro 180 w/ 32Meg. > If you REALLY want to compare several Intel CPUs with some benchmark, you should do the following: 1) install some 32-bit Microsoft OS (I'm very sorry, but you CANNOT neglect this -- see below), 2) use Intel Reference C Compiler (aka Proton) to compile your benchmark with proper "target optimizations" flags for each of your CPUs. You cannot use non-MS or non-32bit OS since Proton does not support them. You should not use compiler other then Proton since in such a case chances are your compiler will generate more "blended" (Intel's term) code then necessary -- i.e. insufficiently tuned. It doesn't mean that Proton is the very best: it simply means that it should be used for benchmarking. I use Symantec C for the work but Proton for benchmarks and selective optimizing. Next problem -- good CPU benchmark must NOT contain any I/O calls or library function calls WITHIN the test loop. Oscillates obviously does not follow this guideline. P6 is by no means worse -- but it is MUCH more sensitive to all this. And in general, we should compare computers, not CPUs. Regards, Vadim. Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa09298; 21 Apr 97 7:12 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id aa11990; 21 Apr 97 7:12 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Mon, 21 Apr 1997 07:10:44 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (HAA17875); Mon, 21 Apr 1997 07:07:18 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Mon, 21 Apr 97 07:07:13 +0100 Received: from mailrelay1.cc.columbia.edu [128.59.35.143] by hermes via ESMTP (HAA17865); Mon, 21 Apr 1997 07:07:12 +0100 Received: from dogma.columbia.edu (dialup-9-9.cc.columbia.edu [128.59.36.213]) by mailrelay1.cc.columbia.edu (8.8.5/8.8.5) with SMTP id CAA27279 for ; Mon, 21 Apr 1997 02:07:15 -0400 (EDT) Message-Id: <335B0395.6CFA@columbia.edu> Date: Mon, 21 Apr 1997 02:05:10 -0400 From: Gregory Boduch Reply-To: gb141@columbia.edu Organization: http://www.juilliard.edu http://www.columbia.edu X-Mailer: Mozilla 3.01Gold (Win95; I) Mime-Version: 1.0 To: csound@maths.ex.ac.uk Subject: Re: CPU Benchmarks References: <199704201727.DAA08598@oznet07.ozemail.com.au> <335A7718.2AA2@columbia.edu> <335AEE8E.18B5@snt.mlt.zaporizhzhe.ua> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk Vadim V. Sytnikov wrote: > > And in general, we should compare computers, not CPUs. > As far as I know, the Oscillates benchmark was written only to compare real-time audio performance of systems under Linux with OSS, and not to benchmark the CPU performance over all. GB Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa09896; 21 Apr 97 9:59 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id aa16120; 21 Apr 97 9:58 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Mon, 21 Apr 1997 09:57:40 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (JAA23549); Mon, 21 Apr 1997 09:45:47 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Mon, 21 Apr 97 09:45:38 +0100 Received: from agora.stm.it [194.20.43.1] by hermes via ESMTP (JAA23530); Mon, 21 Apr 1997 09:45:32 +0100 Received: from x (ppp04_03.stm.it [194.20.251.131]) by agora.stm.it (8.8.5/8.6.6) with SMTP id KAA13598 for ; Mon, 21 Apr 1997 10:45:25 +0200 (ITADST) Message-Id: <335B290C.6103@agora.stm.it> Date: Mon, 21 Apr 1997 10:45:00 +0200 From: Gabriel Maldonado X-Mailer: Mozilla 3.01 (Win95; I) Mime-Version: 1.0 To: Csound Mailing List Subject: Re: Pentium Pro cooks fast! Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk Hi everybody, I tested my computer performance in realtime with the following orc-sco: ------------------------------------------------------------------- ; orc for realtime test sr = 40000 ; also experiment with other sr and kr rates kr = 400 ksmps = 100 nchnls = 1 instr 1 inum notnum icps cpsmidi iamp ampmidi 4000 kmp linenr iamp, 0, .05, .03 ; use relase < 7 secs a1 oscil kmp, icps, 1 ; try also with oscili out a1 endin -------------------------------------------------------------------- ; sco for realtime test f1 0 128 10 1 0 1 0 1 0 1 ; audio function f0 3600 ; allows realtime midi playing for 1 hour e --------------------------------------------------------------- I called my realtime version of csound with the following switches: csound.exe \ -e -O -P99 -p4 -b100 -B100 -odac \ miditest.orc miditest.sco note that the -O flag (non-standard) = suppresses all console text outputs -e flag (non-standard) = enables non-standard sr -p4 (non-standard) = 4 hardware buffers -b100 and -B100 = length of each buffer so total buffering space is about 10042=800 samples or 400/40000 = 1/50 of second theoric latency delay at sr = 40000hz (actually latency delay is a bit grater than 1/50 of sec in win95) -P99 = enables midi sustain pedal holding down the sustain pedal and playing notes once a time with my master-keyboard, I achieved the following realtime performace before incurring in dropouts: at sr=40000 and kr=400 I achieved 50 oscili or 90 oscil polyphony at sr=32000 and kr=320 I achieved 64 oscili or 126 oscil polyphony at sr=16000 and kr=160 I achieved 144 oscili or 280 oscil polyphony with a greater buffer length performance are a little better and are almost equal using standard sr (i.e. 44100 etc.) but latency delay is bigger and don't allow a good keyboard-playing feeling. Note that wavetable length is only of 128 samples so it should be enterely contained in cache memory. My computer is a Pentium 133 with 32M of non-EDO ram and 512k of cache. Could you do this experiment with other PC configurations and other platforms? I'm especially curious about the realtime performance of Pentiums with EDO ram and Pentium-Pros with this orc-sco combination. thanks -- Gabriel Maldonado mailto:g.maldonado@agora.stm.it http://www.agora.stm.it/G.Maldonado/home2.htm Received: from stork2.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa10426; 21 Apr 97 11:04 BST Received: from goggins.bath.ac.uk by stork.maths.Bath.AC.UK id aa18122; 21 Apr 97 11:04 BST Received: from hermes.ex.ac.uk by goggins.bath.ac.uk with SMTP (PP); Mon, 21 Apr 1997 11:02:30 +0100 Received: from noether [144.173.8.10] by hermes via SMTP (KAA27223); Mon, 21 Apr 1997 10:40:20 +0100 Received: from hermes.ex.ac.uk by maths.exeter.ac.uk; Mon, 21 Apr 97 10:40:10 +0100 Received: from pp@goggins.bath.ac.uk [138.38.32.13] by hermes via ESMTP (KAA27208); Mon, 21 Apr 1997 10:40:09 +0100 From: jpff@maths.bath.ac.uk Message-Id: <199704210940.KAA27208@hermes> Received: from maths.Bath.AC.UK (actually host omphalos.maths.bath.ac.uk) by goggins.bath.ac.uk with SMTP (PP); Mon, 21 Apr 1997 10:39:41 +0100 Date: Mon, 21 Apr 97 10:39:43 BST Subject: Re: HRTFs on PC To: csound@maths.ex.ac.uk Sender: owner-csound-outgoing@maths.ex.ac.uk Precedence: bulk Message written at 19 Apr 1997 21:33:57 +0100 --- Copy of mail to PKEARTON@psy.surrey.AC.UK --- In-reply-to: <30681E27E2@psy.surrey.ac.uk> (message from Peter Kearton on Tue, 8 Apr 1997 12:38:21 GMT) I think that I have it fixed but not yet on the released system. I have been away for a couple of weeks -- great fun at SEAMUS -- and have only just returned. I have a list of things to do to csound. Richard Dobson has a list of fixes for me (including improved realtime performance), and I have Robin's opcodes inserted, as well as the new soundin (called diskin in my version). Michael Clarke has sent me a new opcode as well, and there are a host of minor things to do. So gang, do you want a quick release of an interim system, or do you want to wait until I have done it all? I am way for half next week as well, and teaching starts Monday. .....and also this message may stand as a holding message for all of you who mailed me since Easter. I had 12Mb of new mail when I got home, and I am about half way through it. ==John