[PATCH] [TTI] Add a new hook to TargetTransformInfo to query the target if a load of a constant should be converted to simply the constant itself.
Hal Finkel
hfinkel at anl.gov
Mon Jan 27 17:19:12 PST 2014
----- Original Message -----
> From: "Juergen Ributzka" <juergen at apple.com>
> To: "Hal J. Finkel" <hfinkel at anl.gov>
> Cc: "Chandler Carruth" <chandlerc at google.com>, "LLVM Commits" <llvm-commits at cs.uiuc.edu>
> Sent: Monday, January 27, 2014 7:11:29 PM
> Subject: Re: [PATCH] [TTI] Add a new hook to TargetTransformInfo to query the target if a load of a constant should
> be converted to simply the constant itself.
>
>
> Here is the updated patch.
Looks good, thanks! Removing the TTI dependence seems like a nice cleanup.
-Hal
> -Juergen
>
>
>
>
>
>
>
>
> On Jan 27, 2014, at 4:32 PM, Hal Finkel < hfinkel at anl.gov > wrote:
>
>
>
> ----- Original Message -----
>
>
> From: "Chandler Carruth" < chandlerc at google.com >
> To: "Juergen Ributzka" < juergen at apple.com >
> Cc: "LLVM Commits" < llvm-commits at cs.uiuc.edu >
> Sent: Monday, January 27, 2014 6:26:27 PM
> Subject: Re: [PATCH] [TTI] Add a new hook to TargetTransformInfo to
> query the target if a load of a constant should
> be converted to simply the constant itself.
>
>
>
> Why is this going into the TargetTransformInfo which is for use by
> IR-level analysis routines when its only caller is in the
> SelectionDAG? The SelectionDAG can query TargetLowering directly.
>
>
> I agree, this should be in TargetLowering. Otherwise, this seems
> reasonable to me.
>
> -Hal
>
>
>
>
>
> On Mon, Jan 27, 2014 at 4:20 PM, Juergen Ributzka < juergen at apple.com
>
>
> wrote:
>
>
> Hi Hal
>
> this is the new target hook we talked about before. I also added Evan
> and Nadav, because I had to modify the ARM and X86 backend.
>
> Before this patch we used getIntImmCost to determine if a load of a
> constant should be converted to just a constant, but the threshold
> for this was set to an arbitrary value. This value works well for
> the two targets (ARM and X86) that implement this target-hook, but
> it isn't target-independent at all.
>
> Now targets have the possibility to decide directly if this
> optimization should be performed. The default value is set to false
> to preserve the current behavior.
>
> Please review.
>
> Thanks
>
> -Juergen
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>
> --
> Hal Finkel
> Assistant Computational Scientist
> Leadership Computing Facility
> Argonne National Laboratory
>
--
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory
More information about the llvm-commits
mailing list