[PATCH] [InstCombine] Attempt to eliminate redundant loads whose addresses are dependent on the result of a select instruction.

[Tilmann Scheller]

Hi David,

attached is a new patch which is now doing the transformation in GVN.

The bulk of the patch is still the same, I didn't find much logic within GVN which I could reuse, if there's any obvious way to simplify the patch please let me know :)

Nonetheless, the optimization is now in close distance to the other redundant load optimizations.

Regards,

Tilmann

-----Original Message-----
From: Tilmann Scheller [mailto:t.scheller at samsung.com] 
Sent: Thursday, September 04, 2014 8:23 AM
To: 'David Majnemer'
Cc: 'llvm-commits at cs.uiuc.edu'
Subject: RE: [PATCH] [InstCombine] Attempt to eliminate redundant loads whose addresses are dependent on the result of a select instruction.

Hi David,

fair enough, I'll try to figure out how to implement this in GVN then :)

Regards,

Tilmann

From: David Majnemer [mailto:david.majnemer at gmail.com] 
Sent: Thursday, September 04, 2014 12:15 AM
To: Tilmann Scheller
Cc: llvm-commits at cs.uiuc.edu
Subject: Re: [PATCH] [InstCombine] Attempt to eliminate redundant loads whose addresses are dependent on the result of a select instruction.

Hi Tilmann,

I don't think this a natural place to do this, I'm of the opinion that we should try to get GVN to do this sort of thing.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-GVN-Attempt-to-eliminate-redundant-loads-whose-addre.patch
Type: application/octet-stream
Size: 6522 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140904/621a9053/attachment.obj>