MathGroup Archive 2006

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Does Mathematica run on HP-UX (Itanium) ?

  • To: mathgroup at
  • Subject: [mg64914] Re: Does Mathematica run on HP-UX (Itanium) ?
  • From: "Dave (from the UK)" <see-my-signature at>
  • Date: Tue, 7 Mar 2006 06:12:17 -0500 (EST)
  • References: <dtu2ph$f03$>
  • Reply-to: Mar-2006 at
  • Sender: owner-wri-mathgroup at

Dave (from the UK) wrote:
> According to the Wolfram site
> Mathematica 5.2 runs on HP-UX with the PA-RISC processor.
> But if you had an Itanium box (specifically zx2000) running HP-UX would 
> it run? I'm led to believe HP-UX on 64-bit Itanium has a binary 
> compatabiltiy layer which allows binaries for PA-RISC to run. See:
> where it states:
> "Additionally, moving to a new architecture usually creates numerous 
> compatibility challenges and often forces customers to leave behind 
> application software that they value. HP has addressed this issue nicely 
> with HP-UX 11i v2 for Itanium processor-based systems, which offers near 
> functional equality and binary compatibility to HP-UX 11i v1 for PA-RISC."
> If anyone does run Mathematica on the HP-UX operating system on the 
> Itanium processor, what is the performance like? I have no idea how much 
> this binary compatability impacts performance. If the Itanium emulates 
> all the PC-RISC instructions, I suspect it runs slower than on the older 
> PA-RISC chip, despite the much higher peformance of Itanium.
> I know Mathematica runs on Linux on Itanium, but that is not an option, 
> for other reasons, so I am just interested if it runs on HP-UX with the 
> Itanium processor

I have now found the answer to this, so I thought I'd post it.

Someone else has tested it and found that yet is does run on HP-UX 11i 
on Itanium, but the performance is pretty poor.

For the *very* limited testing that has been done, it would appear one 
would not want to a buy an Itanium machine with HP-UX to run Mathematica.

A very slightly modified version of the benchmark at

was used. The only mods were to remove the notebook bits, so it run on 
the command line. Testing on a couple of Sun showed negligible 
difference between the modified and unmodified benchmark, so you can 
take those scores as being reasonable representative.

Scores were

1) zx200O Itanium running Linux.

Version   = 5.2 for Linux IA-64 (June 21, 2005)
Times     = {3.4043, 3.15332, 4.02442, 12.6035, 7.50391, 2.97754, 
12.0498, 24.375, 47.1485, 23.7686, 65.9424, 8.95899, 15.2725, 17.9746, 
Total Kernel Time = 275.352,  Total Time = 275.4
Benchmark = 1.86003

And on the same machine running HP-UX 11i, rather than Linux. In this 
case, HP-UX is emulating a PA-RISC processor.

Version   = 5.2 for HP-UX PA-RISC (64 bit) (June 21, 2005)
Times     = {43.2, 33.15, 48.42, 57.3, 97.97, 31.08, 51.22, 550.63, 
277.15, 101.79, 181.01, 37.56, 68.52, 107.57, 104.59}
Total Kernel Time = 1791.16,  Total Time = 1815.
Benchmark = 0.265283

As you can see, it is a factor of 1.86003/0.265283= 7 x slower when run 
on an emulation mode.

Dave K

Minefield Consultant and Solitaire Expert (MCSE).

Please note my email address changes periodically to avoid spam.
It is always of the form: month-year@domain. Hitting reply will work
for a couple of months only. Later set it manually.

  • Prev by Date: Re: Compile Fourier (2)
  • Next by Date: Re: ndsolve & compile
  • Previous by thread: Re: Listable functions with two brackets f[][] (SubValues)
  • Next by thread: Re: Listable functions with two brackets f[][] (SubValues)