qemu-ppc
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-ppc] [Qemu-devel] [PATCH 02/13] target-ppc: add mask_u128 rout


From: Nikunj A Dadhania
Subject: Re: [Qemu-ppc] [Qemu-devel] [PATCH 02/13] target-ppc: add mask_u128 routine
Date: Tue, 06 Dec 2016 10:49:26 +0530
User-agent: Notmuch/0.21 (https://notmuchmail.org) Emacs/25.0.94.1 (x86_64-redhat-linux-gnu)

Richard Henderson <address@hidden> writes:

> On 12/05/2016 03:25 AM, Nikunj A Dadhania wrote:
>> +#if defined(CONFIG_INT128)
>> +FUNC_MASK(mask_u128, Int128, 128, Int128, ~((__uint128_t)0));
>> +#else
>> +static inline Int128 mask_u128(int start, int end)
>> +{
>> +    Int128 r = {0};
>> +    if (start > 63) {
>> +        r.hi = 0;
>> +        r.lo = mask_u64(start - 64, end - 64);
>> +    } else if (end < 64) {
>> +        r.hi = mask_u64(start, end);
>> +        r.lo = 0;
>> +    } else {
>> +        r.hi = mask_u64(start, 63);
>> +        r.lo = mask_u64(0, end - 64);
>> +    }
>> +    return r;
>> +}
>>  #endif
>
> First, I would really really like you to stop adding *any* ifdefs on
> CONFIG_INT128.  All that's going to do is make sure that there's code that is
> almost never tested, since x86_64 (and other 64-bit hosts) does support 
> int128.

I did test both the cases above by flipping the switch of CONFIG_INT128.
Initially was planning to do this in int128.h, but the bit numbering is
different and wont be usable for other architecture.

> Second, you're not using the Int128 interface correctly.  Better would be
>
> static inline Int128 mask_u128(int start, int end)
> {
>     uint64_t hi, lo;
>     if (start > 63) {
>         hi = 0;
>         lo = mask_u64(start - 64, end - 64);
>     } else if (end < 64) {
>         hi = mask_u64(start, end);
>         lo = 0;
>     } else {
>         hi = mask_u64(start, 63);
>         lo = mask_u64(0, end - 64);
>     }
>     return make_int128(lo, hi);
> }

Sure will use this.

Regards
Nikunj




reply via email to

[Prev in Thread] Current Thread [Next in Thread]