Re: where's the bottleneck?

On 4/12/08, Carl Karsten <[email protected]> wrote:
>  model name      : AMD Turion(tm) 64 X2 Mobile Technology TL-60
>  cache size      : 512 KB
>  bogomips        : 4004.46
>  [transcode] V: IA32/AMD64 accel | sse3 sse2 sse 3dnowext 3dnow mmxext mmx
> cmove asm
>  real    3m59.267s

>  model name      : Intel(R) Pentium(R) M processor 1.73GHz
>  cache size      : 2048 KB
>  [transcode] V: IA32/AMD64 accel | sse2 (sse2 sse mmxext mmx asm C)
>  [transcode] V: IA32/AMD64 accel | using sse memcpy
>  real    4m18.462s

Here are two possibilities: less cache and (more significantly) not
using an accelerated memcpy. You could try playing with the --accel
option explicitly or, for a definite answer, build transcode with
execution profiling enabled on each and use gprof on it...

    M