On Sep 14, 2011, at 1:39 PM, Landon Fuller wrote:
>
> On Sep 14, 2011, at 4:21 PM, Tom Rodriguez wrote:
>
>>>
>>>
>>>> http://cr.openjdk.java.net/~never/7089790_bsd_vs_linux
>>>
>>> src/os_cpu/bsd_x86/vm/os_bsd_x86.cpp:
>>>
>>> ! // HACK: si_code does not work on linux 2.2.12-20!!!
>>> ! // HACK: si_code does not work on bsd 2.2.12-20!!!
>>
>> I don't think I want to fix things like that. They just make the code diverge.
>>
>>>
>>> I doubt there is a bsd 2.2.12 kernel :-) Seriously, I suppose si_code does work on BSD?
>
> The s/linux/bsd/ issue aside, si_code is not set on SIGFPE for <= 10.6.
>
> I believe this was fixed in 10.7, but the Apple guys can confirm (tracked by rdar://4428508 ?).
>
>>> src/os_cpu/bsd_x86/vm/bsd_x86_32.s:
>>> src/os_cpu/bsd_x86/vm/bsd_x86_64.s:
>>>
>>> + #ifdef __APPLE__
>>> + .align 4
>>> + #else
>>> .align 16
>>> + #endif
>>>
>>> Maybe we should use a ALIGN macro instead.
>>
>> Maybe. I'm not sure I even understand the purpose of the change. It seems pretty spurious.
>
> IIRC, Apple's assembler .align directive takes a power of two -- i.e., .align 4^2
Ok. Why isn't it just using p2align which appears to have exactly the same meaning? And why is the second one needed since the next function starts with p2align 4? Why would BSD want 16^2 alignment anyway?
diff -r 9b3694fcfe94 src/os_cpu/bsd_x86/vm/bsd_x86_32.s
--- a/src/os_cpu/bsd_x86/vm/bsd_x86_32.s
+++ b/src/os_cpu/bsd_x86/vm/bsd_x86_32.s
@@ -56,23 +56,13 @@
# Support for void os::Solaris::init_thread_fpu_state() in os_solaris_i486.cpp
# Set fpu to 53 bit precision. This happens too early to use a stub.
# ported from solaris_x86_32.s
-#ifdef __APPLE__
- .align 4
-#else
- .align 16
-#endif
+ .p2align 4,,15
SYMBOL(fixcw):
pushl $0x27f
fldcw 0(%esp)
popl %eax
ret
-#ifdef __APPLE__
- .align 4
-#else
- .align 16
-#endif
-
.globl SYMBOL(SafeFetch32), SYMBOL(Fetch32PFI), SYMBOL(Fetch32Resume)
.globl SYMBOL(SafeFetchN)
## TODO: avoid exposing Fetch32PFI and Fetch32Resume.
tom
>
> -landonf