GCC assembly help
» > >
»
» View previous topic :: View next topic  
As some of you know, GCC for Prizm has been delayed a few days because of a bug in the assembly of some RAM initialization code. For some reason, the file crt0.s (the source) is generating interrupt errors when compiled with GCC, but not when compiled with the fx-9860 SDK and Hitachi's compiler. We've been looking at it for some time and it's not immediately apparent what the difference is. Any help would be appreciated.

The files, crt0.s, default.g3a, and DFAULTstock.G3A (the working file) are available here
∂²Ψ -(2m(V(x)-E)Ψ
----- = -------------
∂x² ℏ²Ψ
What details would you be able to share about what the RAM initialization code is supposed to do, exactly, and what you suspect it's not doing?
Kerm from what I can gather the error is somewhere between the linker setting the right addresses in for _bbss, _ebss, _bdata, and _edata and the asm in crt0.s copying that data to the right locations.

A quick look at the dis-assembly it almost seems that gcc is setting those entries to their destinations rather than their locations in the g3a, which leads me to believe it is a but in the prizm.lkr that Tari made. What is odd about that is the ld script the fx-9860 uses does the same thing.

"Always code as if the person who will maintain your code is a maniac serial killer that knows where you live" -Unknown

"If you've done something right no one will know that you've done anything at all" -Futurama

"Have a nice day, or not, the choice is yours." Tom Steiner

<Michael_V> or create a Borg collective and call it The 83+
<Michael_V> Lower your slide cases and prepare to be silent linked. Memory clears are futile.
/me grabs the files and attempts to have a look

EDIT:
Looked at it some..

Disassembling a g3a
Objdump is quite useful for this task:
  • $ dd bs=1 if=default.g3a of=default.bin skip=28672
  • Truncate last four bytes of default.bin (I used XVI32 to do this)
  • $ sh3-unknown-elf-objdump -b binary -m sh -D default.bin > default.S


So, a problem:

Code:
   0:   2f e6          mov.l   r14,@-r15
   2:   4f 22          sts.l   pr,@-r15
   4:   2f 46          mov.l   r4,@-r15
   6:   b0 0d          bsr   0x24

  24:   e6 00          mov   #0,r6
  26:   a0 02          bra   0x2e

  2e:   d5 0e          mov.l   0x68,r5   ! 8100004
  30:   32 52          cmp/hs   r5,r2

That's execution flow on entry from Jonimus' default.g3a. The cmp/hs instruction at the end there is comparing against r2, which hasn't been modified at all on entry (unless I'm missing some information). The GCC SH ABI specifies R2 as scratch, so I doubt there's anything useful going on there, and it could be some sort of terrible bug.

Aside from that, I'll need to have a look at crt0 some more- it does seem that Jonimus' earlier assertion seems at least somewhat correct.
タリ ·· Ask questions the smart way ·· digipres.club
Excellent point, and nice catch! Do we understand what that code is supposed to do, exactly?
Never mind, I'm dumb. BRA has a delay slot, so we're actually loading r2 with 0x8100004 before entering the loop there. I'll paste my whole disassembly here, with a few annotations:

Code:

default.bin:     file format binary


Disassembly of section .data:

00000000 <.data>:
! dumping to stack frame
   0:   2f e6          mov.l   r14,@-r15
   2:   4f 22          sts.l   pr,@-r15
   4:   2f 46          mov.l   r4,@-r15
   
   6:   b0 0d          bsr   0x24
   
   8:   6e 53          mov   r5,r14
   a:   e6 01          mov   #1,r6
   c:   e4 00          mov   #0,r4
   e:   b0 21          bsr   0x54
  10:   65 63          mov   r6,r5
  12:   d7 03          mov.l   0x20,r7   ! 300080
  14:   65 ed          extu.w   r14,r5
  16:   64 f6          mov.l   @r15+,r4
  18:   4f 26          lds.l   @r15+,pr
  1a:   47 2b          jmp   @r7
  1c:   6e f6          mov.l   @r15+,r14
  1e:   00 00          .word 0x0000
  20:   00 30          .word 0x0030
  22:   00 80          .word 0x0080
 
! subroutine 1
  24:   e6 00          mov   #0,r6
  26:   a0 02          bra   0x2e
  28:   d2 0e          mov.l   0x64,r2   ! 8100004
  2a:   22 62          mov.l   r6,@r2
  2c:   72 04          add   #4,r2
 
  2e:   d5 0e          mov.l   0x68,r5   ! 8100004
  30:   32 52          cmp/hs   r5,r2
  32:   8b fa          bf   0x2a        ! jump if r2 < r5
  34:   d6 0d          mov.l   0x6c,r6   ! 8100004
  36:   a0 02          bra   0x3e
  38:   d5 0e          mov.l   0x74,r5   ! 300084
  3a:   26 22          mov.l   r2,@r6
  3c:   76 04          add   #4,r6
  3e:   d4 0c          mov.l   0x70,r4   ! 8100004
  40:   36 42          cmp/hs   r4,r6
  42:   8b fa          bf   0x3a
  44:   d1 02          mov.l   0x50,r1   ! 8100000
  46:   d5 07          mov.l   0x64,r5   ! 8100004
  48:   34 58          sub   r5,r4
  4a:   74 04          add   #4,r4
  4c:   00 0b          rts   
  4e:   21 42          mov.l   r4,@r1
  50:   08 10          .word 0x0810
  52:   00 00          .word 0x0000
  54:   d2 02          mov.l   0x60,r2   ! 80020070
  56:   e0 29          mov   #41,r0
  58:   42 2b          jmp   @r2
  5a:   00 09          nop   
  5c:   00 00          .word 0x0000
  5e:   00 00          .word 0x0000
  60:   80 02          mov.b   r0,@(2,r0)
  62:   00 70          .word 0x0070
  64:   08 10          .word 0x0810
  66:   00 04          mov.b   r0,@(r0,r0)
  68:   08 10          .word 0x0810
  6a:   00 04          mov.b   r0,@(r0,r0)
  6c:   08 10          .word 0x0810
  6e:   00 04          mov.b   r0,@(r0,r0)
  70:   08 10          .word 0x0810
  72:   00 04          mov.b   r0,@(r0,r0)
  74:   00 30          .word 0x0030
  76:   00 84          mov.b   r8,@(r0,r0)
   ...
  80:   00 0b          rts   
  82:   00 09          nop   
タリ ·· Ask questions the smart way ·· digipres.club
Success at last from Tari and Jonimus' hard effort! Cf. Cemetech Advances Casio Prizm Coding
  
Register to Join the Conversation
Have your own thoughts to add to this or any other topic? Want to ask a question, offer a suggestion, share your own programs and projects, upload a file to the file archives, get help with calculator and computer programming, or simply chat with like-minded coders and tech and calculator enthusiasts via the site-wide AJAX SAX widget? Registration for a free Cemetech account only takes a minute.

» Go to Registration page
GCC assembly help
»
» View previous topic :: View next topic  
Page 1 of 1
» All times are UTC - 5 Hours
 
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum

 

Advertisement