Ralph Smeets wrote:
> On the PC, it's not only the ATRAC algorithm that's running, but
> also writing a packet of data received by the soundcard to memory, and
writing
> this data back to disc.
I just did an experiment. On my 350 Mhz Pentium II, I can write data to my
old hard disk at 2.5 MB/sec, using 5% CPU time, so writing an ATRAC data
stream (say 40KB/sec) would use ~ .1% of available CPU. I then ran Windows
Sound Recorder (hint: probably not mega efficient) at 44.1kHz 16 bit stereo,
requiring less than 1% CPU time. An 11.9 mS time frame gives me ~ 512
samples per channel, which is 256 * 9 FFT butterflies* per frame. Guessing
40 instructions per butterfly gives 256*9*40 operation per frame per
channel, or 256*9*40*84 ops per channel per second, which is 15 Mips. You
have to add bit allocation and so forth, so double this to get ~ 30 Mips. My
CPU can do 350 Mips.
simon
* where a "butterfly" is a' = a * sin(w) + b * cos(w); b' = a * cos(w) + b *
sin(w); and the sines and coses are obtained from a lookup table
-----------------------------------------------------------------
To stop getting this list send a message containing just the word
"unsubscribe" to [EMAIL PROTECTED]
