On Wed, 31 Jul 2013 08:57:08 +0800, David Crayford wrote:
>Do you still get an 0C9 with FLOAT(IEEE)? I suspect not.
>
M&C says:
0C9
Fixed-point-divide exception. The reason code is 9.
... in agreement with the CEE3209S message text. The generated
machine code is:
000009 | * I = 42 / I;
000112 5800 D0A0 000009 | L r0,I(,r13,160)
000116 4160 002A 000009 | LA r6,42
00011A 8E60 0020 000009 | SRDA r6,32
00011E 1D60 000009 | DR r6,r0
000120 1807 000009 | LR r0,r7
000122 5000 D0A0 000009 | ST r0,I(,r13,160)
... nary a FLOP in sight. (Just an otiose LR.) I'm not even going to
try IEEE. (The source code is a couple plies back; be my guest.)
(You mean IEEE *isn't* the default!? I guess I'll put it in my Makefile.)
/usr/include/signal.h says:
...
/* Flags returned in the si_code member of a siginfo_t */
...
#define FPE_INTDIV 31 /* SIGFPE caused by an integer *
* divide by zero */
#define FPE_INTOVF 32 /* SIGFPE caused by an integer *
* overflow */
... It appears that SIGFPE (Floating point exception) is being used as
a catch-all for any arithmetic exception. Except fixed-point overflow,
which is ignored, notwithstanding the definition of FPE_INTOVF.
>On 31/07/2013, at 8:53 AM, Paul Gilmartin wrote:
>>
>> CEE3209S The system detected a fixed-point divide exception (System
>> Completion Code=0C9).
>> From entry point main at compile unit offset +00000096 at entry
>> offset +00000096 at address 47F089BE.
>> [1] + Done(136) ( gmake overflow 2>&1 | tee log ) && ./overflow
>> 50332526 Floating point exception ./overflow
>>
>> "Floating point"? Oh, well.
-- gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
