Subject:
|
The IR FIX -- was: Re: Is IR considered broken?
|
Newsgroups:
|
lugnet.robotics.rcx.legos
|
Date:
|
Mon, 26 Jul 1999 08:18:57 GMT
|
Viewed:
|
1301 times
|
| |
|
|
In lugnet.robotics.rcx.legos, Mark Falco writes:
> The precompiled tm-and-ir, kinda works for me. It is pretty good at
> receiving data, but the transmitted message often has garbage. When I
I had the same problem.
> 3. If it is broken is anyone actively working on fixing it? If not I'll
> take a stab at it.
> 4. If others have working code that uses IR, can you post it?
Here's the fix: enclode the old handler code in #ifdef TRASH,
use the new #else part. I'll say a few things about the motivation below.
(If you prefer to patch, download
http://www.informatik.hu-berlin.de/~mueller/rt/mindstorm/direct-ir.patch
it's based on LegOS 0.1.7 but should work w/ the latest snapshort as well.)
#ifdef TRASH
__asm__("
_dir_tx_handler:
; r6 saved by ROM
mov.w r0,@-r7
mov.w @_dir_tx_read,r0 ; dir_tx_read -> r0
mov.w @_dir_tx_end,r6
cmp.w r0,r6 ; transmission finished?
beq 0f
mov.b @r0+,r6l ; *(ptr++) -> r6
mov.b r6l,@0xdb:8 ; r6 -> transmit data register
bclr #0x7,@0xdc:8 ; clear transmit buffer empty
flag
bra 1f
;; Frank: here seems to be a mistake in the assember: it jumps past the
;; move instruction at label 1. Hence, dir_tx_read is never incremented
;; fix: use the c handler below
0:bclr #0x7,@0xda:8 ; disable transmission irq
1:mov.w r0,@_dir_tx_read ; store new pointer
mov.w @r7+,r0
rts
");
#else
void dir_tx_c(void) {
if (dir_tx_read<dir_tx_end) {
S_TDR = *(unsigned char *) dir_tx_read++;
S_SR &= ~SSR_TRANS_EMPTY;
}
else
S_CR &= ~SCR_TX_IRQ;
}
__asm__("
_dir_tx_handler:
; r6 saved by ROM
mov.w r0,@-r7 ; save all registers
mov.w r1,@-r7
mov.w r2,@-r7
mov.w r3,@-r7
mov.w r4,@-r7
mov.w r5,@-r7
jsr _dir_tx_c ; call scheduler
mov.w @r7+,r5 ; restore all registers
mov.w @r7+,r4
mov.w @r7+,r3
mov.w @r7+,r2
mov.w @r7+,r1
mov.w @r7+,r0
; r6 will be restored by ROM
rts ; return
");
#endif
OK, and here a few words:
1) the original assembler code is correct but the translation by the
H8 assembler is not. See the comment in the assembler code:
The branch is not executed correctly. I suspect a bug in the
calculation of relative branch offsets in the assembler
but that's just a wild guess.
2) the fix simly wraps code the save registers around a C-handler
with the same functionality. This works fine (no relative branches).
3) Notice 8N1 is the setting, and NOT 8O1 as seen in firmdl. Firmdl
actually uses odd parity. LegOS uses NO parity. If you use tm-and-ir
as a test, make sure you set 8N1 2400 baud, only then will you see
your "Hello".
|
|
Message is in Reply To:
 | | is IR considered broken?
|
| The precompiled tm-and-ir, kinda works for me. It is pretty good at receiving data, but the transmitted message often has garbage. When I compile it myself using the latest legOS snapshot, it always transmits garbage, others seem to have seen the (...) (25 years ago, 12-Jul-99, to lugnet.robotics.rcx.legos)
|
6 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
