Date: Mon, 15 Jun 2015 14:32:01 -0700 From: Adrian Chadd <adrian@freebsd.org> To: Stanislav Sedov <stas@freebsd.org> Cc: "freebsd-mips@freebsd.org" <freebsd-mips@freebsd.org> Subject: Re: [rfc] fix umul_ppmm() in our libgcc Message-ID: <CAJ-VmokZJvFuLaCGOG34cCE7kCnMbR0E5zmAia045uxcg3tq5g@mail.gmail.com> In-Reply-To: <7895DD8F-6165-4EF9-AAE7-098C601470D8@freebsd.org> References: <CAJ-Vmonsmu5dGtWKhJchzZfSsVUjtwEVSEij4u6kwk64E0ZOXw@mail.gmail.com> <7895DD8F-6165-4EF9-AAE7-098C601470D8@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
The error is unsupported constraints. '=h' (hi register) is
unsupported as of some GCC > 4.2.1.
-a
On 15 June 2015 at 10:15, Stanislav Sedov <stas@freebsd.org> wrote:
>
>> On Jun 13, 2015, at 3:41 PM, Adrian Chadd <adrian@freebsd.org> wrote:
>>
>> Hi,
>>
>> our libgcc in -base is very .. old. It chokes when compiling for
>> mips32. This patch seems to do the right thing.
>>
>> Does anyone have any positive/negative feedback?
>>
>> Thanks,
>>
>>
>> -adrian
>>
>>
>> Index: contrib/gcc/longlong.h
>> ===================================================================
>> --- contrib/gcc/longlong.h (revision 284090)
>> +++ contrib/gcc/longlong.h (working copy)
>> @@ -584,11 +584,11 @@
>>
>> #if defined (__mips__) && W_TYPE_SIZE == 32
>> #define umul_ppmm(w1, w0, u, v) \
>> - __asm__ ("multu %2,%3" \
>> - : "=l" ((USItype) (w0)), \
>> - "=h" ((USItype) (w1)) \
>> - : "d" ((USItype) (u)), \
>> - "d" ((USItype) (v)))
>> + do { \
>> + UDItype __ll = (UDItype)(u) * (v); \
>> + w1 = __ll >> 32; \
>> + w0 = __ll; \
>> + } while (0)
>> #define UMUL_TIME 10
>> #define UDIV_TIME 100
>> #endif /* __mips__ */
>
> This looks right to me in terms of functionality, but what is wrong
> with the original code (multu)? That one seems correct as well.
>
> What error do you see when compiling it?
>
> --
> ST4096-RIPE
>
>
>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-VmokZJvFuLaCGOG34cCE7kCnMbR0E5zmAia045uxcg3tq5g>