Version française
Home     About     Download     Resources     Contact us    

This site is updated infrequently. For up-to-date information, please visit the new OCaml website at

Browse thread
[Caml-list] Hyperthreading
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: -- (:)
From: David McClain <dmcclain1@m...>
Subject: [Caml-list] Hyperthreading
> How do you measure "CPU utilization"?  Different tools have different

I just used the simple minded task manager graphs. I don't have those
details just yet, but I do have the Intel VTune tools to do a serious look
at what's happening inside the CPU.

Another possibility occured to me, but it would mimmick the low CPU
utilization we get. But on the newer processors, instruction ordering may
have an impact on the speed of execution. I have some other 3rd party tools
here that appear to be using enormous amounts of CPU cycles for apparently
simple computations. But these tools are intended to run on a wide variety
of Pentium platforms, and they were definitely not especially compiled for
the P4.

> My general feeling (although I have no proof of that) is that the
> memory usage patterns of Caml programs aren't radically different from
> those of more mainstream programs.

You may well be correct about this... but then, how many programs take care
to tune the pattern of memory use? In my own numerical code, I know that
enormous floating-point arrays are being consumed and produced at prodigious
rates. The need to have better data locality never showed itself before on
those older processors. But clearly, speeding up the CPU by a factor of 10
to reach an overall throughput improvement of only a factor of 2, points to
inefficiencies that I now ought to address.

But ulimately, it would be much nicer if some hardware guru's could get
better educated on the nature of high level languages, and stop thinking
that everything is written in hand-crafted C code. We need some paradigm
shifts in the way processors attack high-level programs.


- DM

To unsubscribe, mail Archives:
Bug reports: FAQ:
Beginner's list: