[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH v2 1/2] target-ppc: add vextu[bhw]lx instruction
From: |
Nikunj A Dadhania |
Subject: |
Re: [Qemu-devel] [PATCH v2 1/2] target-ppc: add vextu[bhw]lx instructions |
Date: |
Fri, 25 Nov 2016 12:43:37 +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 11/24/2016 12:32 PM, Nikunj A Dadhania wrote:
>> +#if defined(HOST_WORDS_BIGENDIAN)
>> +# if defined(CONFIG_INT128)
>> +# define VEXTULX_DO(name, size) \
>> + target_ulong glue(helper_, name)(target_ulong a, ppc_avr_t *b) \
>> + { \
>> + int index = (a & 0xf) * 8; \
>> + return int128_rshift(b->u128, index) & \
>> + MAKE_64BIT_MASK(0, size); \
>> + }
>> +# else
>> +# define VEXTULX_DO(name, size) \
>> + target_ulong glue(helper_, name)(target_ulong a, ppc_avr_t *b) \
>> + { \
>> + int index = (a & 0xf) * 8; \
>> + Int128 value = int128_make128(b->u64[LO_IDX], \
>> + b->u64[HI_IDX]); \
>> + return int128_rshift(value, index) & \
>> + MAKE_64BIT_MASK(0, size); \
>> + }
>> +# endif
>
> Why are these duplicated?
>
> Clearly the missed trick is that you should *never* check CONFIG_INT128 and
> always rely on Int128.
Because of this:
/* Altivec registers (128 bits) */
union ppc_avr_t {
[ SNIP ]
uint64_t u64[2];
int64_t s64[2];
#ifdef CONFIG_INT128
__uint128_t u128;
#endif
};
So I composed a Int128 using make128 with lo and hi to get a
Int128, and then call int128_rshift.
I think I can use "Int128 u128" in that union and that should do the
trick ! I will try that out.
Regards
Nikunj