Re: sqrt via SSE2 registers

[Tim Prince <tprince at computer dot org>]

On Monday 18 February 2002 13:06, Tom Womack wrote:

> There's then the issue, though it's probably more one for the next glibc
> release after gcc-3.1 appears, of whether a sin() implementation using SSE2
> code and a suitable rational-function approximation could get adequate
> results in less than the 190-or-so cycles that fsin takes: I'm pretty sure
> it's possible, even given that the necessary two divides can't take less
> than 70 ticks and that one might want a table-lookup for argument
> reduction.
>
> Annoyingly, I suspect Intel have done this already though not under a
> GPL-compatible license: 

The scheme with Intel C is not so good; if you wish to employ the sqrt 
instructions in line, while referring to their SSE library for the others, 
you must #include a table which specifies which functions are to link to the 
library.  AFAIK it does not work with C++.  In principle, you could link with 
gcc to the Intel math libraries, both the parallel versions in libsvml, and 
the single copy versions, by writing in the Intel mangled function name.