[PATCH] [AArch64 NEON] Fix wrong imm used in BIC instruction.

[Ana Pazos]

Hi Kevin,

The confusion is because you tried to do some code cleanup while addressing
a real issue (inverted immediate).

I think the bitconvert node was not needed. Might have been a copy-paste
error from other pattern when I wrote it.

I tried your change on the miscompare failures I see in spec 2006 perl tests
with neon enabled and it worked.

So  LGTM.

Thanks,
Ana.

-----Original Message-----
From: llvm-commits-bounces at cs.uiuc.edu
[mailto:llvm-commits-bounces at cs.uiuc.edu] On Behalf Of Kevin Qin
Sent: Monday, January 06, 2014 5:57 PM
To: kevinqindev at gmail.com
Cc: llvm-commits at cs.uiuc.edu
Subject: Re: [PATCH] [AArch64 NEON] Fix wrong imm used in BIC instruction.


  Hi Tim,

  Thanks for your review. This part of codes was implemented by Ana long
time ago. I believe there is no such bitconvert node in DAG which holds same
value type in its operand and result. I guess for pattern matcher, its a
dumb node and just skip it to match next. So from functionality perspective,
it's OK to insert this kind of bitconvert everythere and any times. But to
keep code readable and more efficiency, I suggest don't use this dumb node.
So I removed it in this patch.  Do you agree?

http://llvm-reviews.chandlerc.com/D2500
_______________________________________________
llvm-commits mailing list
llvm-commits at cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits