[PATCH] Add new indexed load/store intrinsics.

Hao Liu Hao.Liu at arm.com
Sun Apr 26 18:47:59 PDT 2015

In http://reviews.llvm.org/D9194#161690, @delena wrote:

> We are interested in these intrinsics anyway:
> <4 x double> @llvm.masked.indexed.load.v4f64 (double* <ptr>, <4 x i32> <index>, i32 <alignment>, <4 x i1>%mask, <4 x double>%paththru)
> Presence of index means that the sequential load is not a solution.
>  If you are not going to implement this interface now, I’ll do this later.
> - Elena
>   From: Renato Golin [mailto:renato.golin at linaro.org] Sent: Thursday, April 23, 2015 10:29 To: reviews+http://reviews.llvm.org/D9194+public+a2f9c543cd08afc3@reviews.llvm.org Cc: Ahmed Bougacha; Hao.Liu at arm.com; llvm-commits at cs.uiuc.edu; qcolombet at apple.com; Demikhovsky, Elena Subject: Re: [PATCH] Add new indexed load/store intrinsics.

Yes, it's still useful in your situation. I just think the interleaved load/store doesn't need this intrinsic.




More information about the llvm-commits mailing list