[llvm-dev] EuroLLVM Numerics info

Tue Apr 9 14:34:01 PDT 2019

On 2nd look, this really is a sqrt-specific optimization:
https://reviews.llvm.org/rL357943

I think we still want the more general fmul+fdiv canonicalization, but that
transform has some risk of exposing different missing optimizations, so
I'll wait a bit before trying that.

On Fri, Apr 5, 2019 at 9:20 AM Venkataramanan Kumar <
venkataramanan.kumar.llvm at gmail.com> wrote:

> Hi Sanjay,
>
> Thanks,  yes I was also thinking of looking at re association pass to see
> if it can catch few more cases.
>
> regards,
> Venkat.
>
> On Fri, 5 Apr 2019 at 19:48, Sanjay Patel <spatel at rotateright.com> wrote:
>
>> [adding back the mailing list]
>>
>> Thanks for the example!
>>
>> define double @_Z12sqrt_squaredd(double) {
>> %2 = tail call fast double @llvm.sqrt.f64(double %0)
>> %3 = fdiv fast double 1.000000e+00, %2
>> %4 = fmul fast double %3, %3
>> ret double %4
>> }
>>
>> So at first look, this has nothing to do with sqrt specifically. We are
>> missing a basic factorization / re-association for fmul and fdiv:
>>       // (L1 / L2) * (R1 / R2) --> (L1 * R1) / (L2 * R2)
>>
>> If we do that, then existing transforms should kick in to reduce the sqrt
>> example. I'll try to fix this in instcombine soon. If you have more
>> complicated examples where we miss this, that would suggest that we need to
>> make an enhancement to the "reassociate" pass too.
>>
>> On Fri, Apr 5, 2019 at 12:26 AM Venkataramanan Kumar <
>> venkataramanan.kumar.llvm at gmail.com> wrote:
>>
>>> Hi Sanjay,
>>>
>>> On Thu, 4 Apr 2019 at 21:21, Sanjay Patel <spatel at rotateright.com>
>>> wrote:
>>>
>>>> 1. Do you have a larger code example that shows the missed sqrt/div
>>>> optimization?
>>>>
>>>> We optimize the example you provided already in IR:
>>>> #include <math.h>
>>>> float sqrt_squared(float x) {
>>>>   return 1.0f/sqrtf(x) * 1.0f/sqrtf(x);
>>>> }
>>>>
>>>> Something like this
>>> #include <math.h>
>>> double sqrt_squared(double x) {
>>> double y = 1.0/sqrt(x) ;
>>> double z = y * y ;
>>> return z;
>>> }
>>> REF: https://godbolt.org/z/2lhnJq
>>>
>>>
>>>> $ clang -O1 -ffast-math -emit-llvm  sqrt.c -S -o -
>>>> define float @sqrt_squared(float) local_unnamed_addr #0 {
>>>>   %2 = fdiv fast float 1.000000e+00, %0
>>>>   ret float %2
>>>> }
>>>>
>>>> 2. Yes, transforms like the above should work with vector types. If
>>>> not, please file a bug.
>>>>
>>> Oh! thanks let me check.
>>>
>>> regards,
>>> Venkat.
>>>
>>>>
>>>> On Wed, Apr 3, 2019 at 11:47 PM Venkataramanan Kumar via llvm-dev <
>>>> llvm-dev at lists.llvm.org> wrote:
>>>>
>>>>> Hi Micheal,
>>>>>
>>>>> Thanks for the blog post.  Just like to point out few things that I
>>>>> thought is related to FP Numerics.
>>>>> LLVM could do some additional transformation with "sqrt" and
>>>>> "division" under fast math on X86 like 1/sqrt(x)* 1/sqrt(x) to 1/x.   These
>>>>> are long latency instructions and could get benefit if enabled under unsafe
>>>>> math.
>>>>>
>>>>> Also are we considering doing such FP transforms on vector floating
>>>>> point types?
>>>>>
>>>>> regards,
>>>>> Venkat.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On Mon, 1 Apr 2019 at 19:56, Cameron McInally via llvm-dev <
>>>>> llvm-dev at lists.llvm.org> wrote:
>>>>>
>>>>>> Hey Michael,
>>>>>>
>>>>>> Thank you for working on this!
>>>>>>
>>>>>> I'd like to touch on a topic mentioned in the blog post. The
>>>>>> constrained intrinsics work is at a road block on how to proceed with the
>>>>>> constrained  implementation in the backends, i.e. D55506. Reviews/ideas in
>>>>>> this area would be greatly appreciated (attn: target code owners).
>>>>>>
>>>>>> Thanks,
>>>>>> Cameron
>>>>>>
>>>>>>
>>>>>> On Thu, Mar 28, 2019 at 5:54 PM Michael Berg via llvm-dev <
>>>>>> llvm-dev at lists.llvm.org> wrote:
>>>>>>
>>>>>>> All:  There will be a BoF talk at the EuroLLVM conference regarding
>>>>>>> Numerics (FMF and module flags which control fp behavior and optimization).
>>>>>>>
>>>>>>> Even if you are not going to be in attendance, please reply to this
>>>>>>> thread as we are collecting open issues and ideas for future direction in
>>>>>>> all layers of LLVM for which optimizations are controlled by numerics flags.
>>>>>>> Please read over the numerics blog if you like for reference
>>>>>>> material:
>>>>>>>
>>>>>>>
>>>>>>> https://urldefense.proofpoint.com/v2/url?u=http-3A__blog.llvm.org_2019_03_llvm-2Dnumerics-2Dblog.html&d=DwIGaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=O_4M49EtSpZ_-BQYeigzGv0P4__noMcSu2RYEjS1vKs&m=CHHCGUAnxC91rtvrnhhoN0jXGXgXv8EQb8C4XiUEoxY&s=xzUyP83nqSY1h9GrDdHch4INrIFIU-Ycij4S4Tmxdko&e=
>>>>>>>
>>>>>>> Regards,
>>>>>>> Michael
>>>>>>> _______________________________________________
>>>>>>> LLVM Developers mailing list
>>>>>>> llvm-dev at lists.llvm.org
>>>>>>>
>>>>>>> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.llvm.org_cgi-2Dbin_mailman_listinfo_llvm-2Ddev&d=DwIGaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=O_4M49EtSpZ_-BQYeigzGv0P4__noMcSu2RYEjS1vKs&m=CHHCGUAnxC91rtvrnhhoN0jXGXgXv8EQb8C4XiUEoxY&s=5CVfTGSHNKAmAzx4dc0HQzDTXMucnDgx_NPimota5m8&e=
>>>>>>>
>>>>>> _______________________________________________
>>>>>> LLVM Developers mailing list
>>>>>> llvm-dev at lists.llvm.org
>>>>>> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>>>>>
>>>>> _______________________________________________
>>>>> LLVM Developers mailing list
>>>>> llvm-dev at lists.llvm.org
>>>>> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>>>>
>>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20190409/5035ed63/attachment.html>