- Contemporary messages sorted: [ by date ] [ by thread ] [ by subject ] [ by author ] [ by messages with attachments ]

From: Carlos Murillo <cmurillo_at_emtelsa.multi.net.co>

Date: Fri Sep 28 06:50:36 2001

These days, I usually prefer to fire up matlab, even in my older

DAQ & HPIL-specific, DOS-only pc, where I still have my 87/09/09 copy of

pc-matlab 3.13, and my copy of matlab-386 3.5 . If I need more

portability then I resort to (in order of preference) my HP71,

my HP67 or my HP41CV. But my holy grail for portable computations

would be a nice HP200LX w. extra mem. and pcmatlab 3.13; alas, I never

managed to find one I could afford. For a while I had a working HP100LX,

but then it died.

These are the pcmatlab 3.13 benchmarks on a Pentium 166; I think

it is a nice historical comparison:

bench

echo on, clc

% The MATLAB benchmarks. (May take 2-3 minutes)

%

% This demo file runs a set of 7 standard benchmarks:

%

% 1) N=50 Real matrix multiply

% 2) N=50 Real matrix inverse

% 3) N=25 Real eigenvalues

% 4) 4096-point complex FFT

% 5) LINPACK benchmark

% 6) 1000 iteration FOR loop

% 7) N=25 3-D mesh plot

%

% The benchmarks illuminate computer architectural issues that

% affect the speed of software like MATLAB on different machines.

%

% The time for your machine is measured and displayed versus times

% we've already taken from other standard machines.

% ------- Table of benchmark times -------

%

% 1) PC/XT (4.7MHz\8088\CGA)

% 2) PC/AT (6.0MHz\80286\EGA)

% 3) AT&T 6300 (8MHz\8086)

% 4) Compaq 80386 (16MHz\8MHz80287)

% 5) SUN-3/50 (15MHz with 68881)

% 6) SUN-3/260 (25MHz with FPA)

% 7) MicroVAX II (VMS\D_floating)

% 8) Macintosh (8MHz 68000)

% 9) Your Machine

% * inv eig fft LINPACK for mesh

times

times =

9.4500 20.9000 21.0000 12.3000 40.9000 18.0000 15.8000

11.1500 20.6000 20.3000 14.9000 46.9000 6.0000 9.9000

5.2000 11.6000 11.2000 7.1000 23.5000 8.7000 7.4000

5.1600 10.2900 10.1600 7.4700 24.1800 2.9700 3.7900

1.8400 3.6000 3.4000 3.5000 7.7000 2.0000 5.0000

0.7000 0.9000 1.1000 1.0000 1.7000 0.9000 2.0000

2.6000 3.1000 2.9000 2.9000 4.9000 3.2000 25.0000

252.0000 325.0000 306.0000 300.0000 926.0000 21.0000 21.0000

0.0600 0.0500 0.0500 0.0500 0.0684 0.1600 0.5500

pause % Strike any key to continue

echo off

clc

% ------- Table of speed ratios to PC/XT -------

%

% 1) PC/XT (4.7MHz\8088\CGA)

% 2) PC/AT (6.0MHz\80286\EGA)

% 3) AT&T 6300 (8MHz\8086)

% 4) Compaq 80386 (16MHz\8MHz80287)

% 5) SUN-3/50 (15MHz with 68881)

% 6) SUN-3/260 (25MHz with FPA)

% 7) MicroVAX II (VMS\D_floating)

% 8) Macintosh (8MHz 68000)

% 9) Your Machine

% * inv eig fft LINPACK for mesh

ratio

ratio =

1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000

0.8475 1.0146 1.0345 0.8255 0.8721 3.0000 1.5960

1.8173 1.8017 1.8750 1.7324 1.7404 2.0690 2.1351

1.8314 2.0311 2.0669 1.6466 1.6915 6.0606 4.1689

5.1359 5.8056 6.1765 3.5143 5.3117 9.0000 3.1600

13.5000 23.2222 19.0909 12.3000 24.0588 20.0000 7.9000

3.6346 6.7419 7.2414 4.2414 8.3469 5.6250 0.6320

0.0375 0.0643 0.0686 0.0410 0.0442 0.8571 0.7524

157.5000 418.0000 420.0000 246.0000 598.2518 112.5000 28.7273

pause % Strike any key to continue

echo off

clc

% Another popular number to look at is the number of KFlops obtained

% from Jack Dongarra's LINPACK benchmark (sometimes called the Argonne

% benchmark). This number is often advertised by computer manufacturers

% in national trade magazines, comparing their computer versus others

% in terms of floating point performance.

% The LINPACK benchmark compares the performance of different computer

% systems while solving 100'th order dense systems of linear equations in

% a Fortran environment. We're in a C environment, but this doesn't

% stop us from performing the equivalent calculation.

% For solving a system of 100 equations, approximately

nflops = 2/3*100^3 + 2*100^2

nflops =

6.8667e+005

% operations are performed. Using our times from above, we find the

% KFlop/second throughput for the various machines:

KFlops = round(nflops./times(:,5)'/1000)

KFlops =

Columns 1 through 6

17 15 29 28 89 404

Columns 7 through 9

140 1 10044

% PC-XT PC-AT AT&T 386 SUN3 SUN3F McrVAX Mac Yours

% Just for reference, the Cray X-MP achieves 33,000 KFlops!

? quit

2652666 flops.

At 03:38 PM 9/27/01 -0600, Dick wrote:

*>I'm still partial to having some of those functions, too, but with a computer
*

*>running Windows nearly everywhere I turn, the calculator that's built-in
*

there

*>seems to work fine. I gave all my TI calculators and printers away a
*

couple of

*>years back and really don't miss 'em, though I always liked 'em when I
*

used 'em.

*>They were harder to deal with where batteries were concerned, though, than
*

are

*>the old HP's, and I think the easy-to-modify battery case is the reason I've
*

*>ended up with the HP's.
*

--------------------------------------------------------------

Carlos E. Murillo-Sanchez carlos_murillo_at_nospammers.ieee.org

Received on Fri Sep 28 2001 - 06:50:36 BST

Date: Fri Sep 28 06:50:36 2001

These days, I usually prefer to fire up matlab, even in my older

DAQ & HPIL-specific, DOS-only pc, where I still have my 87/09/09 copy of

pc-matlab 3.13, and my copy of matlab-386 3.5 . If I need more

portability then I resort to (in order of preference) my HP71,

my HP67 or my HP41CV. But my holy grail for portable computations

would be a nice HP200LX w. extra mem. and pcmatlab 3.13; alas, I never

managed to find one I could afford. For a while I had a working HP100LX,

but then it died.

These are the pcmatlab 3.13 benchmarks on a Pentium 166; I think

it is a nice historical comparison:

bench

echo on, clc

% The MATLAB benchmarks. (May take 2-3 minutes)

%

% This demo file runs a set of 7 standard benchmarks:

%

% 1) N=50 Real matrix multiply

% 2) N=50 Real matrix inverse

% 3) N=25 Real eigenvalues

% 4) 4096-point complex FFT

% 5) LINPACK benchmark

% 6) 1000 iteration FOR loop

% 7) N=25 3-D mesh plot

%

% The benchmarks illuminate computer architectural issues that

% affect the speed of software like MATLAB on different machines.

%

% The time for your machine is measured and displayed versus times

% we've already taken from other standard machines.

% ------- Table of benchmark times -------

%

% 1) PC/XT (4.7MHz\8088\CGA)

% 2) PC/AT (6.0MHz\80286\EGA)

% 3) AT&T 6300 (8MHz\8086)

% 4) Compaq 80386 (16MHz\8MHz80287)

% 5) SUN-3/50 (15MHz with 68881)

% 6) SUN-3/260 (25MHz with FPA)

% 7) MicroVAX II (VMS\D_floating)

% 8) Macintosh (8MHz 68000)

% 9) Your Machine

% * inv eig fft LINPACK for mesh

times

times =

9.4500 20.9000 21.0000 12.3000 40.9000 18.0000 15.8000

11.1500 20.6000 20.3000 14.9000 46.9000 6.0000 9.9000

5.2000 11.6000 11.2000 7.1000 23.5000 8.7000 7.4000

5.1600 10.2900 10.1600 7.4700 24.1800 2.9700 3.7900

1.8400 3.6000 3.4000 3.5000 7.7000 2.0000 5.0000

0.7000 0.9000 1.1000 1.0000 1.7000 0.9000 2.0000

2.6000 3.1000 2.9000 2.9000 4.9000 3.2000 25.0000

252.0000 325.0000 306.0000 300.0000 926.0000 21.0000 21.0000

0.0600 0.0500 0.0500 0.0500 0.0684 0.1600 0.5500

pause % Strike any key to continue

echo off

clc

% ------- Table of speed ratios to PC/XT -------

%

% 1) PC/XT (4.7MHz\8088\CGA)

% 2) PC/AT (6.0MHz\80286\EGA)

% 3) AT&T 6300 (8MHz\8086)

% 4) Compaq 80386 (16MHz\8MHz80287)

% 5) SUN-3/50 (15MHz with 68881)

% 6) SUN-3/260 (25MHz with FPA)

% 7) MicroVAX II (VMS\D_floating)

% 8) Macintosh (8MHz 68000)

% 9) Your Machine

% * inv eig fft LINPACK for mesh

ratio

ratio =

1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000

0.8475 1.0146 1.0345 0.8255 0.8721 3.0000 1.5960

1.8173 1.8017 1.8750 1.7324 1.7404 2.0690 2.1351

1.8314 2.0311 2.0669 1.6466 1.6915 6.0606 4.1689

5.1359 5.8056 6.1765 3.5143 5.3117 9.0000 3.1600

13.5000 23.2222 19.0909 12.3000 24.0588 20.0000 7.9000

3.6346 6.7419 7.2414 4.2414 8.3469 5.6250 0.6320

0.0375 0.0643 0.0686 0.0410 0.0442 0.8571 0.7524

157.5000 418.0000 420.0000 246.0000 598.2518 112.5000 28.7273

pause % Strike any key to continue

echo off

clc

% Another popular number to look at is the number of KFlops obtained

% from Jack Dongarra's LINPACK benchmark (sometimes called the Argonne

% benchmark). This number is often advertised by computer manufacturers

% in national trade magazines, comparing their computer versus others

% in terms of floating point performance.

% The LINPACK benchmark compares the performance of different computer

% systems while solving 100'th order dense systems of linear equations in

% a Fortran environment. We're in a C environment, but this doesn't

% stop us from performing the equivalent calculation.

% For solving a system of 100 equations, approximately

nflops = 2/3*100^3 + 2*100^2

nflops =

6.8667e+005

% operations are performed. Using our times from above, we find the

% KFlop/second throughput for the various machines:

KFlops = round(nflops./times(:,5)'/1000)

KFlops =

Columns 1 through 6

17 15 29 28 89 404

Columns 7 through 9

140 1 10044

% PC-XT PC-AT AT&T 386 SUN3 SUN3F McrVAX Mac Yours

% Just for reference, the Cray X-MP achieves 33,000 KFlops!

? quit

2652666 flops.

At 03:38 PM 9/27/01 -0600, Dick wrote:

there

couple of

used 'em.

are

--------------------------------------------------------------

Carlos E. Murillo-Sanchez carlos_murillo_at_nospammers.ieee.org

Received on Fri Sep 28 2001 - 06:50:36 BST

*
This archive was generated by hypermail 2.3.0
: Fri Oct 10 2014 - 23:34:26 BST
*