On Mon, Jan 16, 2012 at 08:15:50PM -0800, Zsolt Vasvari wrote:
> Or let's take a networking example (since the communactions between a
> CPU and the GPU is a network)
> Which one is faster?
> Sending a 1MB file in ONE CHUNK over a 56kbps dial-up line or sending
> the same file, ONE BIT AT A TIME, using TCP/IP over high speed
> internet? My guess is the former...
Ummm, where do you get this (mistaken) idea that IP only transmits
one bit (or even one byte) at a time? Please post your reference
(specifically, the RFC that makes this claim).
Second, a 56 kb/s dial-up modem, even with the speed increase gained by
adding error control[1], you're still getting far, FAR less throughput
than even a relatively slow HSI link (or even a much-slower DSL link, at
around 512 kb/s...or is that 512 kB/s? sorry, I don't do DSL...too
fscking slow compared to the HSI option here, where I get up to 4 MB/s,
limited by my laptop, not the network bandwidth).
And with a dial-up modem, you are most definitely NOT sending a 1 MB file
in a single chunk. You are sending a lot of much smaller packets (the
size of which depends on exactly which file transfer protocol you are
referring to), but you are definitely not sending it in one big chunk.
But then, with TCP/IP, you also are not sending one fscking bit at a
time, either.
Want the math? See [1] below.
Later,
--jim
[1] No, that was not a joke. Adding error control, whether it's CCITT
(now ITU-T) V.42 or MNP3/4 does in fact, with a relatively error-free
line, ADD throughput. Adding error control changes the number of
bits/character from 10 to 8 by stripping start/stop bits as part of
its async-to-sync conversion. Here's the math:
For V.32bis (14.4 kb/s)...you can use the same math for higher
dial-up speeds to calculate the end-to-end throughput:
Without error control:
T(max) == (14400 bps / 10 bits/char) == 1440 cps maximum
With error control (V.42/LAPM or MNP3/4):
T(max) = (14400 bps / 8 bits/char) * 128/135 * 62/63 == 1679.58 cps
where: 128/135 is the FCS + framing overhead
62/63 is the bit stuffing overhead (typical average 1 bit
stuffed for every 62 bits of data)
--
THE SCORE: ME: 2 CANCER: 0
73 DE N5IAL (/4) | "This 'telephone' has too many
spooky1...@gmail.com | shortcomings to be seriously considered
< Running FreeBSD 7.0 > | as a means of communication. The device
ICBM / Hurricane: | is inherently of no value to us."
30.44406N 86.59909W | (Western Union internal memo, 1876)
Android Apps Listing at http://www.jstrack.org/barcodes.html
--
You received this message because you are subscribed to the Google
Groups "Android Developers" group.
To post to this group, send email to android-developers@googlegroups.com
To unsubscribe from this group, send email to
android-developers+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/android-developers?hl=en
