[Ffmpeg-devel] [PATCH] Vorbis decoder

[Rich Felker]

On Thu, May 12, 2005 at 09:10:11AM -0600, Mike Melanson wrote:
> Rich Felker wrote:
> >I don't believe float is ever faster than a _proper_ int
> 
> 	The issue, per my understanding, is that these perceptual audio 
> decoders need to work with numbers that are larger than 32 bits in 
> precision. You either use a float/double and use the CPU's FPU 
> instructions or you break the 64-bit number up into 2 32-bit numbers and 
> do a little more work to perform, e.g., a multiplication (vs. a single 
> FMUL instruction for the floating point data type).

i find this hard to believe.

> 	CPUs without FPU benefit from pure integer implementation since it 
> 	is faster than having to do FP ops in software with integers.
> 
> > implementation, except maybe on broken intel crap (i.e. P4).
> 
> 	What on Earth makes you call the P4 "broken", smart guy? I remember 
> that whole Pentium 1 FDIV bug fiasco, but I can not find similar bugs 
> for P4.

this has been discussed extensively by many mplayer (and lavc)
developers. it sacrifices all the performance issues that are actually
useful in the real world in order to benchmark high on stupid dsp
tests, which is pointless because you use a real dsp if you want a
dsp. essentially the p4 is slow and crippled both as a dsp and as a
general-purpose cpu. key flaws:

- slow memory io and cache
- slow bit arithmetic
- slow conditionals/deep pipeline

it's no secret that a p4 performs video encoding/decoding at the same
speed as an athlon that's half the price and barely over half the
clock.

rich