| 03:39.48 | *** join/#elinux Ox83 (~Ox83@c-24-130-249-162.hsd1.ca.comcast.net) |
| 07:08.26 | *** join/#elinux lyakh (~lyakh@p57BD1E65.dip0.t-ipconnect.de) |
| 08:45.32 | *** join/#elinux hw_ (~hw@p578b3905.dip0.t-ipconnect.de) |
| 12:46.08 | *** join/#elinux gustavoz (~gustavoz@200.82.38.195) |
| 13:45.28 | *** join/#elinux flavioribeiro (~flavio@187.64.42.101) |
| 14:18.37 | *** join/#elinux GPSFan (~kenm@64.92.145.112) |
| 14:37.20 | *** join/#elinux pb___ (~pb@88-109-183-94.dynamic.dsl.as9105.com) |
| 15:10.51 | *** join/#elinux Tryum (~Tryum@e-corporation.info) |
| 15:23.20 | *** join/#elinux krushia (~krushia@h69-131-86-88.cntcnh.dsl.dynamic.tds.net) |
| 15:27.23 | *** join/#elinux thraxisp (~thraxisp@century.precidia.com) |
| 16:56.23 | *** join/#elinux 94SAAKAZO (~hw@p578b3905.dip0.t-ipconnect.de) |
| 17:04.10 | *** join/#elinux toi (~toi@d54C2A96D.access.telenet.be) |
| 17:35.07 | *** join/#elinux Ox83 (~Ox83@c-24-130-249-162.hsd1.ca.comcast.net) |
| 18:45.51 | *** join/#elinux krushia (~krushia@h69-131-86-88.cntcnh.dsl.dynamic.tds.net) |
| 20:58.15 | *** join/#elinux krushia (~krushia@h69-131-86-88.cntcnh.dsl.dynamic.tds.net) |
| 22:43.45 | *** join/#elinux tzanger (~tzanger@gromit.mixdown.ca) |
| 22:44.20 | tzanger | I'm playing aorund with some arm code (LPC2103) and I'm trying to understand why the compiler insists on my IRQ handler returning with "bx lr" |
| 22:44.28 | tzanger | I'm not using any thumb code anywhere |
| 22:44.52 | tzanger | I'm compiling with -Os but that shouldn't change the instruction set |
| 22:44.59 | roxfan | well, the compiler can't be sure of that |
| 22:45.01 | tzanger | (objdump is showing 32-bit wide insns) |
| 22:45.20 | roxfan | what if there is some dynamically-generated thumb code somewhare? |
| 22:45.23 | tzanger | roxfan: ok, but doen't bx lr start executing thumb instructions? |
| 22:45.27 | roxfan | no |
| 22:45.42 | roxfan | it sets mode depending on the low bit of lr |
| 22:45.44 | tzanger | i.e. if I hit a normal ARM instruction, won't it throw an illegal instruction exception? |
| 22:45.47 | tzanger | oh |
| 22:45.49 | tzanger | I didn't realize |
| 22:45.55 | tzanger | much ado over nothing :-) |
| 22:46.02 | roxfan | you're probably thinking of blx |
| 22:46.20 | roxfan | which does switch mode, though only in the " |
| 22:46.21 | tzanger | no, I was just assuming something incorrectly |
| 22:46.31 | roxfan | * in the "blx address" form |
| 22:47.12 | tzanger | also I'm looking at the startup code for various ARM implementations and one in particualr sets up fp with 0 before jumping to main() -- what's the purpose of that? |
| 22:48.42 | roxfan | who knows |
| 22:49.10 | roxfan | maybe the crt used requires it for some reason |
| 22:49.29 | tzanger | ok |
| 22:49.51 | tzanger | so bx looks at the low bit of the register it's going to use for an address to determine thumb or arm |
| 22:50.07 | roxfan | yep |
| 22:50.10 | tzanger | mov r0, #0 |
| 22:50.16 | tzanger | mov r1, #0 |
| 22:50.20 | tzanger | ldr r2, =main |
| 22:50.21 | tzanger | bx r2 |
| 22:50.30 | tzanger | would then work with both arm and thumb startup code |
| 22:50.54 | tzanger | and I imagine the compiler makes sure that all thumb functions start on a word boundary with the low bit set |
| 22:50.57 | tzanger | that's slick |
| 22:51.23 | roxfan | well, they still start on an even address |
| 22:51.35 | roxfan | but when you take its address, the compiler adds 1 |
| 22:52.14 | tzanger | oh yes wer're talking 16 bit boundaries not 8 bit, duh :-) |
| 22:52.17 | tzanger | it's been a bit of a day :-) |
| 22:59.25 | tzanger | hmm something's gotta be wrong |
| 22:59.45 | tzanger | the mapfile doesn't show .text any smaller with -mthumb-interwork |
| 22:59.53 | tzanger | oh wait, I'm probably not telling it to generate thumb code yet |
| 23:00.24 | roxfan | yeah, try just -mthumb |
| 23:00.48 | tzanger | :-) |
| 23:03.41 | tzanger | yeah, down to 2c0c from 3800 bytes |
| 23:03.42 | tzanger | took almost 4k out |
| 23:03.42 | tzanger | that's impressive :-) |
| 23:08.38 | *** join/#elinux thraxisp (~thraxisp@24.139.16.154) |
| 23:08.48 | *** join/#elinux krushia (~krushia@h69-131-86-88.cntcnh.dsl.dynamic.tds.net) |
| 23:18.02 | tzanger | roxfan: one last question, this time more about the compiler... I'm using what was built with openembedded for the gumstix boards. The compiler seems to be missing some low-level math functions. "__aeabi_uidiv" for example |
| 23:18.21 | tzanger | it's just an unsigned int division... is there something obvious I've done wrong? |
| 23:18.34 | tzanger | idiv's also mising as is idivmod |
| 23:19.20 | roxfan | uhh sounds like a mismatch between compiler and libs |
| 23:19.37 | roxfan | sorry i'm not very familiat with inner workings of gcc |
| 23:19.51 | tzanger | libs? you mean the compiler libs, right, as this is a standalone and all-in-one app (no third party libs) |
| 23:19.55 | tzanger | me either |
| 23:20.39 | roxfan | c runtime libs |
| 23:21.00 | roxfan | you still need division as processor doesn't support it directly |
| 23:23.06 | tzanger | right I understand why I need the support lib |
| 23:23.12 | tzanger | just not sure how to tell the compiler where it is |
| 23:37.51 | wmat | tzanger: youprobably just need to link in math functions |
| 23:38.14 | wmat | when you compile, add -lm to the command line |
| 23:39.01 | tzanger | I am not sure it's that but I can try |
| 23:40.35 | tzanger | it's in libgcc.a |
| 23:40.41 | tzanger | wonder why it can't get to that though |
| 23:41.38 | tzanger | if I manually link it in with -L/long/path/to/cross/compiler/libs -lgcc I now get an error saying it can't find 'raise' which __div0 wants |
| 23:41.41 | tzanger | I guess I could provide that |
| 23:43.33 | wmat | this all sounds familiar, but I can't dredge it up from my foggy memory |
| 23:45.11 | wmat | also, don't forget to place all library options after all source and object files on the command line, as lib inclusion is sensitive to placement on the command line |
| 23:46.57 | tzanger | yep |
| 23:47.00 | tzanger | I got that :-) |
| 23:47.07 | tzanger | and yeah it's ringing a very faint bell here too |
| 23:54.29 | wmat | did you use -nostartfiles -nostdlib? |
| 23:54.46 | *** join/#elinux dieter|work (~dieter@dslb-088-065-198-128.pools.arcor-ip.net) |
| 23:56.09 | wmat | try adding -lgcc |
| 23:56.11 | *** join/#elinux krushia (~krushia@h69-131-86-88.cntcnh.dsl.dynamic.tds.net) |