Please note:
The error I mentioned occures with optimzation -O2.
I now found out with -O3 everything works fine, no further
workaround required.
Volker.
mspgcc-users-requ...@lists.sourceforge.net wrote:
Send Mspgcc-users mailing list submissions to
mspgcc-users@lists.sourceforge.net
To subscribe or unsubscribe via the World Wide Web, visit
https://lists.sourceforge.net/lists/listinfo/mspgcc-users
or, via email, send a message with subject or body 'help' to
mspgcc-users-requ...@lists.sourceforge.net
You can reach the person managing the list at
mspgcc-users-ad...@lists.sourceforge.net
When replying, please edit your Subject line so it is more specific
than "Re: Contents of Mspgcc-users digest..."
Today's Topics:
1. BUG with float to int cast crashes msp430 + WORKAROUND (Volker Schindler)
--__--__--
Message: 1
Date: Sun, 28 May 2006 18:00:22 +0200
From: Volker Schindler <volker.schind...@aon.at>
To: mspgcc-users@lists.sourceforge.net
Subject: [Mspgcc-users] BUG with float to int cast crashes msp430 + WORKAROUND
Reply-To: mspgcc-users@lists.sourceforge.net
Hello,
I experienced problems with the floating point package.
If I use a cast from float to integer, the msp430 totally crashes.
This cast uses the __fixunssfsi subroutine, which in turn calles
the __fixsfsi routine. Code of this routine is included below.
The first command and the last command before ret indicate
that it needs 12 additional bytes on the stack. This additional
stack space however does not seem to be reserved properly.
In particular I had a look how the stack pointer is initialized an
I found out that the 12 up most bytes of the msp430f149 are
NOT used. The code generated by the compiler initializes r1
with the value 2548; the up most byte of the ram is 2559 instead,
so I would expect r1 to be initialized at 2560. (Note: the 149 has
2048Bytes Ram + 512bytes SFR).
Comparing these 12 bytes offset with the 12 bytes the __fixsfsi
needs additional stack space I became curios and tried some
tricks in order to get the stackpointer initialized at 2560.
Adding RESERVE_RAM(-12) does the job. I was a bit surprised
that the compiler allowed me to consume this neg. amount of bytes
above the stack, but the result was
(a) the sp get properly initialized at 2560
(b) the float to int cast no longer crashes the msp430 into nowhere land.
Greetings.
00003b26 <__fixsfsi>:
3b26: 31 80 0c 00 sub #12, r1 ;#0x000c
3b2a: 81 4e 08 00 mov r14, 8(r1) ;
3b2e: 81 4f 0a 00 mov r15, 10(r1) ;
3b32: 0e 41 mov r1, r14 ;
3b34: 0f 41 mov r1, r15 ;
3b36: 3f 52 add #8, r15 ;r2 As==11
3b38: b0 12 7a 3d call #15738 ;#0x3d7a
3b3c: 0e 43 clr r14 ;
3b3e: 6f 41 mov.b @r1, r15 ;
3b40: 6f 93 cmp.b #2, r15 ;r3 As==10
3b42: 3a 24 jz $+118 ;abs 0x3bb8
3b44: 0e 93 cmp #0, r14 ;r3 As==00
3b46: 33 20 jnz $+104 ;abs 0x3bae
3b48: 0e 43 clr r14 ;
3b4a: 6f 93 cmp.b #2, r15 ;r3 As==10
3b4c: 01 2c jc $+4 ;abs 0x3b50
3b4e: 1e 43 mov #1, r14 ;r3 As==01
3b50: 0e 93 cmp #0, r14 ;r3 As==00
3b52: 2d 20 jnz $+92 ;abs 0x3bae
3b54: 0e 43 clr r14 ;
3b56: 6f 92 cmp.b #4, r15 ;r2 As==10
3b58: 2d 24 jz $+92 ;abs 0x3bb4
3b5a: 0e 93 cmp #0, r14 ;r3 As==00
3b5c: 07 20 jnz $+16 ;abs 0x3b6c
3b5e: 1d 41 02 00 mov 2(r1), r13 ;
3b62: 0d 93 cmp #0, r13 ;r3 As==00
3b64: 24 38 jl $+74 ;abs 0x3bae
3b66: 3d 90 1f 00 cmp #31, r13 ;#0x001f
3b6a: 0b 38 jl $+24 ;abs 0x3b82
3b6c: c1 93 01 00 cmp.b #0, 1(r1) ;r3 As==00
3b70: 04 24 jz $+10 ;abs 0x3b7a
3b72: 0e 43 clr r14 ;
3b74: 3f 40 00 80 mov #-32768,r15 ;#0x8000
3b78: 21 3c jmp $+68 ;abs 0x3bbc
3b7a: 3e 43 mov #-1, r14 ;r3 As==11
3b7c: 3f 40 ff 7f mov #32767, r15 ;#0x7fff
3b80: 1d 3c jmp $+60 ;abs 0x3bbc
3b82: 3c 40 1e 00 mov #30, r12 ;#0x001e
3b86: 1e 41 04 00 mov 4(r1), r14 ;
3b8a: 1f 41 06 00 mov 6(r1), r15 ;
3b8e: 0c 8d sub r13, r12 ;
3b90: 0c 93 cmp #0, r12 ;r3 As==00
3b92: 05 24 jz $+12 ;abs 0x3b9e
3b94: 12 c3 clrc
3b96: 0f 10 rrc r15 ;
3b98: 0e 10 rrc r14 ;
3b9a: 1c 83 dec r12 ;
3b9c: fb 23 jnz $-8 ;abs 0x3b94
3b9e: c1 93 01 00 cmp.b #0, 1(r1) ;r3 As==00
3ba2: 0c 24 jz $+26 ;abs 0x3bbc
3ba4: 3e e3 inv r14 ;
3ba6: 3f e3 inv r15 ;
3ba8: 1e 53 inc r14 ;
3baa: 0f 63 adc r15 ;
3bac: 07 3c jmp $+16 ;abs 0x3bbc
3bae: 0e 43 clr r14 ;
3bb0: 0f 43 clr r15 ;
3bb2: 04 3c jmp $+10 ;abs 0x3bbc
3bb4: 1e 43 mov #1, r14 ;r3 As==01
3bb6: d1 3f jmp $-92 ;abs 0x3b5a
3bb8: 1e 43 mov #1, r14 ;r3 As==01
3bba: c4 3f jmp $-118 ;abs 0x3b44
3bbc: 31 50 0c 00 add #12, r1 ;#0x000c
3bc0: 30 41 ret
--__--__--
_______________________________________________
Mspgcc-users mailing list
Mspgcc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mspgcc-users
End of Mspgcc-users Digest
