#mageec Log for 2013-07-23

--- Log opened Tue Jul 23 00:00:43 2013
-!- munaafghumran [~mg0950@cpc10-aztw24-2-0-cust177.aztw.cable.virginmedia.com] has quit [Ping timeout: 248 seconds]01:38
-!- munaafghumran [~mg0950@cpc10-aztw24-2-0-cust177.aztw.cable.virginmedia.com] has joined #mageec08:17
munaafghumranjeremy: am I just taking out the printf and fprintf's in MiBench programs, or all types of IO?11:06
jamespallisterhi munaafghumran, I would have thought all offchip IO needs to be removed11:09
@jeremybennettmunaafghumran: Anything that assumes a substantive OS.11:09
@jeremybennettjamespallister's comment is also good.11:09
munaafghumranwouldn't offchip IO be board dependent? and if so could someone let me know the exact board model so I can check what's available on it11:25
AWhetterWe're trying to make the benchmarks work on as many boards as possible though, right?11:26
jamespallisterYes, in most cases its board dependent11:28
jamespallistersomewhat standardised over ARM Cortex-M platforms11:29
jamespallistermost of the troubles will be with the standard library requiring host-OS features, such as file systems and terminal IO11:29
munaafghumranyeah, and a fair few of mibench uses dataset files and therefore fscanf. WCET all are fine. I am not really too sure what to do about this sorry! I mean do I just ignore the ones that need datasets?11:34
AWhetterMaybe you could just copy the data files into the source files and replicate the behaviour somehow? But I guess it depends how big they are because the boards don't have loads of RAM.11:35
jamespallisterit depends how beneficial and how much time you think it would take11:35
AWhetterDepending on what it's using the files for11:36
jamespallisterfor the mibench benchmarks I used, i did what AWhetter was suggesting11:36
jamespallistere.g. dijkstra I copy and pasted the data files in11:36
jamespallisterSame for the encryption benchmarks - I just made it operate on an initialised buffer instead of streaming from a file11:37
munaafghumranright okay, that makes more sense11:37
AWhetterI've been making all of the benchmarks self verify today. I'm up to the last one and it's running fine in gdb, but running it as a standalone program fails16:46
@simoncookin what way is it failing, and can you get information out somehow about what is wrong16:55
AWhetterSometimes. The program requires lots of loops so if I go to deep then it's printing out too much16:57
AWhetterI know that compiling it on 32 bit works and 64 doesn't16:58
AWhetterFrom the top level loop I can see that it's producing different output the first iteration and different for the rest, whereas 32 bit does not16:58
jamespallisterwhich program is this16:59
AWhetterfrom the benchmarks16:59
AWhetterI have a feeling that gdb is maybe giving me zeroed memory where as outside it isn't17:00
AWhetterbecause it's doing a lot of bitwise operations, and the algorithm looks like its meant for 32 bit17:00
jamespallisterin sha.h, LONG is defined as unsigned long17:07
jamespallisterin a 64bit compile, I'm guessing this is 64 bits as opposed to 32?17:07
AWhetterunsigned long is 64 bits in both I believe17:07
AWhetteroh got it17:08
AWhetterIt was some data not being initialised17:08
AWhetterMaybe not. gdb is failing now17:11
AWhetterI didn't restart gdb :P It does work.17:14
@simoncookoh what, to load a different binary17:14
AWhetteryep >.<17:14
@simoncookdoesnt file/load reload the file17:14
AWhetterI thought it did but maybe not17:15
AWhetteroh I didn't use file either17:16
AWhetterIt's been a long day...17:16
@simoncookblame the temperature, heat isnt good for productivity im sure17:17
AWhetterBeing awoken at 3:30am because of a storm doesn't help either17:19
@simoncookI would imagine so yes17:20
@simoncookFYI for all:  http://mageec.org/irclogs17:21
-!- munaafghumran [~mg0950@cpc10-aztw24-2-0-cust177.aztw.cable.virginmedia.com] has quit [Ping timeout: 256 seconds]18:36
--- Log closed Wed Jul 24 00:00:49 2013

Generated by irclog2html.py 2.12.1 by Marius Gedminas - find it at mg.pov.lt!