[llvm] r182499 - This is an update to a previous commit (r181216).

[Evan Cheng]

On May 23, 2013, at 6:39 AM, Rafael Espíndola <rafael.espindola at gmail.com> wrote:

> On 23 May 2013 01:42, Evan Cheng <evan.cheng at apple.com> wrote:
>> I see your point but I don't completely agree for the following reasons.
>> 
>> 1. It doesn't appear canonicalizing these expressions enable further optimization.
>> 2. Before Jean-Luc's first patch, they were not canonicalized to selects.
>> 3. No in tree targets would benefit from the canonicalization.
>> 
>> So to me, specializing the original transformation to match only a specific idiom seems reasonable to me.
>> 
>> If I had understood, please let me know.
> 
> I understand the argument for not cannonicalizing to select, but we
> should have a canonical form for this computation. From the points
> above it looks like the canonicalization should be
> 
> select C, B, 0 -> B * (uitofp i1 C)
> 
> When B is of floating point type.

Perhaps. However,

1. Nothing is currently producing these type of selects.
2. This the canonicalization out of select should not be a pre-requisite for this patch.

Evan

> 
>> Evan
> 
> Cheers,
> Rafael