Re: [PATCH] match.pd: rewrite select to branchless expression

[Jeff Law <jeffreyalaw@gmail.com>]

On 11/8/22 13:15, Andrew Pinski via Gcc-patches wrote:
> On Tue, Nov 8, 2022 at 12:02 PM Michael Collison <collison@rivosinc.com> wrote:
>> This patches transforms (cond (and (x , 0x1) == 0), y, (z op y)) into
>> (-(and (x , 0x1)) & z ) op y, where op is a '^' or a '|'. It also
>> transforms (cond (and (x , 0x1) != 0), (z op y), y ) into (-(and (x ,
>> 0x1)) & z ) op y.
>>
>> Matching this patterns allows GCC to generate branchless code for one of
>> the functions in coremark.
>>
>> Bootstrapped and tested on x86 and RISC-V. Okay?
> This seems like a (much) reduced (simplified?) version of
> https://gcc.gnu.org/pipermail/gcc-patches/2021-November/584411.html .
> I have not had time for the last year to go through the comments on
> that patch and resubmit it though.
> It seems like you are aiming for one specific case in coremarks rather
> than a more generic fix too.

I'm fairly confident it was developed independently.  Michael did this 
transformation for LLVM and reached out to me a month or two ago for 
suggestions on the GCC implementation.


My recollection is I suggested phi-opt or match.pd with a slight 
preference for phi-opt as I wasn't offhand sure if we'd have a form 
suitable for match.pd.


THe pattern is just a conditional xor/ior with all is said and done.  
While the inspiration comes from coremark, I don't think it's supposed 
to be specific to coremark.


jeff