On 2/17/21 5:02 AM, Richard Sandiford wrote:
> Xi Ruoyao via Gcc-patches <gcc-patches@gcc.gnu.org> writes:
>>>>>> I can't understand the comment either. To me it looks like it's
>>>>>> possible to
>>>>>> remove this "if (MSA_SUPPORTED_P (mode)) return 0;"
> I think the point is that the MSA loads and stores only have a 10-bit
> offset field instead of the usual 16-bit offset field and so the usual
> approaches to handling symbolic addresses won't work.
Ah.
>
>>>>>> CC Robert to get some help.
>>>>> Happy new lunar year folks.
>>>>>
>>>>> I found a newer email address of Robert. Hope it is still being used.
>>>>>
>>>>> Could someone update MAINTAINERS file by the way?
>>>> If you have an updated email address, I can reach out to Robert and see
>>>> if he wants his entry updated or removed.
>>> His latest reply in gcc mail lists used robert.sucha...@mips.com. But
>>> when I
>>> sent mail to it, the mail was just rejected with "access denied". Google
>>> told
>>> me Office 365 mail service (used by mips.com) rejects mail to deleted
>>> accounts
>>> with "access denied". So I'm not sure if this email address is invalid
>>> again,
>>> or
>>> Office 365 just dislikes me...
>> Hi Jeff,
>>
>> I think it's better to just fix the out-of-bound array access now by special
>> casing MAX_MACHINE_MODE, if we can't figure out if this entry should be
>> removed.
>> Either in MSA_SUPPORTED_P or in mips_symbol_insns.
>>
>> It's really "irrational" to leave such a obvious programming error in new
>> GCC 11
>> release... And I've built a Linux system (in Linux From Scratch way, X11 was
>> built and it runs correctly now) on the Loongson 3A4000 with patched
>> GCC-10.2.0,
>> and "-O3 -mmsa" in CFLAGS for most packages so the change should be OK.
> Yeah, agreed. I think the mips_symbol_insns patch is the right one,
> so I've pushed to it trunk. I think it's also worth backporting
> to release branches, but let me know how far back you've tested it.
Seems reasonable to me. THanks for lending a hand on this Richard.
jeff
