[llvm-commits] [PATCH/RFC] Loop Data Software Prefetching

[Hal Finkel]

----- Original Message -----
> From: "Nadav Rotem" <nrotem at apple.com>
> To: "Hal Finkel" <hfinkel at anl.gov>
> Cc: "Evan Cheng" <evan.cheng at apple.com>, "l >> Commit Messages and Patches for LLVM" <llvm-commits at cs.uiuc.edu>
> Sent: Thursday, January 31, 2013 12:55:52 AM
> Subject: Re: [llvm-commits] [PATCH/RFC] Loop Data Software Prefetching
> 
> 
> 
> 
> On Jan 29, 2013, at 3:22 PM, Hal Finkel < hfinkel at anl.gov > wrote:
> 
> 
> The problem is that this makes experimenting difficult; changing
> anything would require recompiling -- and it makes it impossible for
> the user to override the behavior.
> 
> 
> I am not against command line flags, I am just against committing
> them into LLVM. :) Command line flags are very useful for people who
> develop their own compiler stack and want to control some parameters
> (such as inlining thresholds). It is also useful for writing tests.
> But if the goal is to ease development and to experiment then you
> can add the command line flags and simply not commit them into LLVM.

To be honest, I think that this is kind of silly. Some target-provided parameters are fixed properties of the target, and so no user-level customization needs to be provided. Other things, however, like whether to use software prefetching, the prefetching distance, etc. are just heuristic inputs, and we need to provide the ability for the user to override them.

 -Hal

> 
> 
> If I add the command line flags, for example, to PPCTTI, then I'd
> need to use different flags in X86TTI, and connecting all of these
> flags with clang flags becomes messy.
> 
> What do you suggest?
> 
> 
> 
> 
>