[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [DotGNU] this week - Inlining methods tum's way

From: Thong (Tum) Nguyen
Subject: RE: [DotGNU] this week - Inlining methods tum's way
Date: Sat, 12 Jun 2004 00:59:48 +1200

> -----Original Message-----
> From: Gopal V [mailto:address@hidden
> Sent: Saturday, 12 June 2004 00:40
> To: address@hidden
> Subject: [DotGNU] this week - Inlining methods tum's way
> Hi,
> This has been a really fast week for
> and it has got much faster nearer to the weekend :)
> The pnetmark floating point benchmarks jumped by
> a factor of 16 (yes, 1600%) between Monday and Friday.
> On my machine it went from a piddling 57 to a colossal
> 864 .
> One of the programs I'm working on (which essentially
> is a lot of Sin/Cos transforms on an array) went up
> for about 8 times faster for the inner loops . So yes,
> expect that CSVorbis Ogg player to play smoothly
> even when you're compiling..
> But amidst all this rejoicing, a simple tragedy
> wiped out the rest of tum's hacks for NULL check
> elimination - I don't know if it was lightning that
> hit his hard disk, but we can only imagine .
> Code like that was never meant to be .

On the positive side of things, it will give me a chance to rewrite it (for
the 3rd time) with some new ideas I've had in the meantime.  It'll be faster
to do because I've already learnt from all my previous attempts the best and
worst ways to implement it :-).

> In other news, Klaus Triechel has been hacking away
> at ml-pnet (and in turn causing me to hack away at
> cscc).

reply via email to

[Prev in Thread] Current Thread [Next in Thread]