Re: Q: float arrays

Thorsten Ohl (ohl@crunch.ikp.physik.th-darmstadt.de)
Wed, 30 Oct 1996 15:52:28 +0100

Date: Wed, 30 Oct 1996 15:52:28 +0100
Message-Id: <9610301452.AA34276@crunch>
From: Thorsten Ohl <ohl@crunch.ikp.physik.th-darmstadt.de>
To: caml-list@margaux.inria.fr
Subject: Re: Q: float arrays
In-Reply-To: <199610301340.OAA26778@pauillac.inria.fr>
<199610301340.OAA26778@pauillac.inria.fr>

>>>>> "Xavier" == Xavier Leroy <Xavier.Leroy@inria.fr> writes:

Xavier> Polymorphism and higher-order functions don't mix well with
Xavier> high performance. If you need Fortran-like performance, there
Xavier> are cases where you must write Fortran-style code.

I know that I'm asking too much, but wouldn't it be nice it the
compiler did it for me? In the example at hand,

let f2 = map (function (p,p') -> (p*.p'))

the compiler could inline the multiplication automagically, iff it
still had access to the definition of the map function.

A trivial example is code like the following:

let exponentiate make_unit eps norm plus minus times scale x =
let u = make_unit x in
let rec sumup s n x1 xn =
if norm (xn) <= eps *. norm (minus s u) then
s
else
sumup (plus s xn) (n +. 1.0) x1 (scale (1.0 /. n) (times x1 xn))
in
(* usually, times x u == x, but u can be a projector which
will speed up things for special cases. *)
sumup u 2.0 x (times x u)

let exp_ =
exponentiate (fun _ -> 1.0) 1e-14 abs_float ( +. ) ( -. ) ( *. ) ( *. )

let exp_matrix =
exponentiate unit_like 1e-14 infinity_norm add subtract multiply scale_matrix

Here partial application really shines and it could shine even
brighter if there was no speed penalty ...

Cheers,
-Thorsten

-- 
Thorsten Ohl, Physics Department, TH Darmstadt --- PGP: AF 38 FF CE 03 8A 2E A7
http://crunch.ikp.physik.th-darmstadt.de/~ohl/ -------- 8F 2A C1 86 8C 06 32 6B