fpetrogalli (Francesco Petrogalli)
User

Projects

User does not belong to any projects.

User Details

User Since
Oct 12 2016, 4:50 AM (110 w, 23 h)

Recent Activity

Tue, Nov 20

fpetrogalli accepted D53927: [AArch64] Enable libm vectorized functions via SLEEF.
Tue, Nov 20, 10:05 AM
fpetrogalli added a comment to D53927: [AArch64] Enable libm vectorized functions via SLEEF.

I am sorry for having kept you waiting for the patch. I was not fully aware of the approval process upstream (didn’t do much contributions upstream myself), and before being able to approve I had to understand the process.

Tue, Nov 20, 7:36 AM

Mon, Nov 19

fpetrogalli added a comment to D54412: [RFC] Re-implementing -fveclib with OpenMP.

IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.

Mon, Nov 19, 2:47 PM
fpetrogalli added a comment to D54412: [RFC] Re-implementing -fveclib with OpenMP.

I still would like to see the ability to differentiate quasi-intrinsic veclib functions versus user-written vector functions and enable veclib w/o enabling user-written vector functions, even if -fopenmp-simd becomes default. Some of our ICC customers take perf data with and w/o openmp simd (and proprietary simd pragma) and they expect SVML to be used for both cases. I don't have much preferences in how to accomplish it. For example, if we'd like to keep the header file contents totally standard based, we could use something like a command line flag -fclang-vector-intrinsic=file1,file2,..,fileN to indicate that "declare simd" in those header file is "quasi-intrinsic". We could silently do this for the veclibs shipped with clang (e.g., check if "declare simd" is coming from clang include directory) ---- I'm fine with this ----, but that's a bit unfair to the 3rd party veclibs.

Mon, Nov 19, 2:44 PM
fpetrogalli added a comment to D54412: [RFC] Re-implementing -fveclib with OpenMP.

I would avoid having a clang-specific pragma. By just using the OpenMP one, we provide the library vendors with a mechanism that is based on a standard and therefore can be used by any compiler.

Mon, Nov 19, 7:18 AM

Fri, Nov 16

fpetrogalli added inline comments to D54412: [RFC] Re-implementing -fveclib with OpenMP.
Fri, Nov 16, 1:14 PM
fpetrogalli added inline comments to D54412: [RFC] Re-implementing -fveclib with OpenMP.
Fri, Nov 16, 11:24 AM
fpetrogalli updated the diff for D54412: [RFC] Re-implementing -fveclib with OpenMP.
Fri, Nov 16, 11:24 AM
fpetrogalli added a comment to D53927: [AArch64] Enable libm vectorized functions via SLEEF.

Hi @steleman , thank you for your patience.

Fri, Nov 16, 9:18 AM

Sun, Nov 11

fpetrogalli created D54412: [RFC] Re-implementing -fveclib with OpenMP.
Sun, Nov 11, 7:31 PM

Fri, Nov 9

fpetrogalli added a comment to D53927: [AArch64] Enable libm vectorized functions via SLEEF.

sorry for the delay in getting back to you. I have a couple of observations, for you and @shibatch.

Fri, Nov 9, 8:56 PM

Fri, Nov 2

fpetrogalli added a comment to D53927: [AArch64] Enable libm vectorized functions via SLEEF.

GLOBAL _ZGVnN2v_atan
GLOBAL _ZGVnN4v_atanf_u35
GLOBAL _ZGVnN4v_atanf
GLOBAL _ZGVnN2v_atan_u35
GLOBAL _ZGVnN2v_atan_u35
GLOBAL _ZGVnN2v_atan
GLOBAL _ZGVnN4v_atanf
GLOBAL _ZGVnN4v_atanf_u35

Fri, Nov 2, 8:08 AM

Thu, Nov 1

fpetrogalli updated subscribers of D53927: [AArch64] Enable libm vectorized functions via SLEEF.

OK, I like the idea of having a generalized way of expressing the libm <-> SLEEF bindings.

Thu, Nov 1, 6:25 PM

Wed, Oct 31

fpetrogalli added a comment to D53927: [AArch64] Enable libm vectorized functions via SLEEF.

Apologies for the confidentiality notice at the end my last message. It wasn't supposed to be there, please ignore it. Nothing of what I have written is confidential.

Wed, Oct 31, 11:40 AM
fpetrogalli added a comment to D53927: [AArch64] Enable libm vectorized functions via SLEEF.

Hi @steleman , thank you for working on this!

Wed, Oct 31, 9:19 AM

Tue, Oct 30

fpetrogalli added a comment to D53613: RFC: Explicit Vector Length Intrinsics and Attributes.

thank you for sending this out. I was wondering whether we really want to separate the concepts of vector lane predication and dynamic vector length.

Tue, Oct 30, 3:21 PM

Sep 26 2018

fpetrogalli created D52579: Vector Clone Tests for the loop vectorizer.
Sep 26 2018, 5:09 PM

Mar 20 2018

fpetrogalli added a comment to D22792: VecClone Pass.

Hi Matt - overall the patch looks good, I just have a couple of comments.

Mar 20 2018, 6:19 AM

Jan 19 2018

fpetrogalli added a comment to D40575: LoopVectorize support for simd functions.

Hi Matt, this is very nice. I have a couple of comments to add to Hal's one.

Jan 19 2018, 2:20 AM

Nov 28 2017

fpetrogalli added inline comments to D22792: VecClone Pass.
Nov 28 2017, 12:41 AM

Nov 6 2017

fpetrogalli added inline comments to D22792: VecClone Pass.
Nov 6 2017, 2:25 PM
fpetrogalli added a comment to D22792: VecClone Pass.

thank you for updating the patch.

Nov 6 2017, 3:33 AM

Oct 23 2017

fpetrogalli added a comment to D22792: VecClone Pass.

Hello - gentle ping on this patch. I think it is an important addition to the compiler. Do we need to reactivate the discussion in the mailing list?

Oct 23 2017, 8:52 AM

Sep 1 2017

fpetrogalli added inline comments to D22792: VecClone Pass.
Sep 1 2017, 3:46 AM
fpetrogalli added a comment to D22792: VecClone Pass.

Hello again,

Sep 1 2017, 3:34 AM
fpetrogalli added inline comments to D22792: VecClone Pass.
Sep 1 2017, 3:15 AM

Aug 29 2017

fpetrogalli accepted D37064: [InstCombine] Fold insert sequence if first ins has multiple users..

This LGTM now.

Aug 29 2017, 3:41 AM

Aug 24 2017

fpetrogalli added inline comments to D37064: [InstCombine] Fold insert sequence if first ins has multiple users..
Aug 24 2017, 2:25 AM

Apr 7 2017

fpetrogalli added a comment to D30739: [OpenMP] "declare simd" for AArch64 Advanced SIMD..

thank you for reviewing this. If you don't mind, I will wait applying the changes you requested, Unfortunately I am not ready yet to send out an update as I am still working on the vector ABI. I will ping you when ready.

Apr 7 2017, 1:23 AM

Mar 28 2017

fpetrogalli added a comment to D31417: [OpenMP] Add support for omp simd pragmas without runtime.

Hi Graham,

Mar 28 2017, 4:43 AM

Mar 23 2017

fpetrogalli planned changes to D30739: [OpenMP] "declare simd" for AArch64 Advanced SIMD..

Dear all,

Mar 23 2017, 10:29 AM

Mar 21 2017

fpetrogalli added a comment to D30739: [OpenMP] "declare simd" for AArch64 Advanced SIMD..

Thanks, I do not have commit rights, can anyone commit this?

Mar 21 2017, 6:23 AM
fpetrogalli added a comment to D30739: [OpenMP] "declare simd" for AArch64 Advanced SIMD..

In principal looks good to me although I'm not really familiar with this part. Does that work for you if you have the declare simd in a header file and the implementation in another file? On x86_64 I currently get:

Mar 21 2017, 6:13 AM

Mar 15 2017

fpetrogalli added inline comments to D30739: [OpenMP] "declare simd" for AArch64 Advanced SIMD..
Mar 15 2017, 6:46 AM
fpetrogalli added a comment to D30739: [OpenMP] "declare simd" for AArch64 Advanced SIMD..

Gentle ping.

Mar 15 2017, 6:25 AM
fpetrogalli updated subscribers of D30739: [OpenMP] "declare simd" for AArch64 Advanced SIMD..
Mar 15 2017, 6:24 AM

Mar 14 2017

fpetrogalli added a comment to D30897: [WIP] New AsmPrinterHandler that is able to interleave source code.

Hi Roger,

Mar 14 2017, 2:09 AM

Mar 8 2017

fpetrogalli updated the diff for D30739: [OpenMP] "declare simd" for AArch64 Advanced SIMD..
  • fixed formatting;
  • added two tests that were missing.
Mar 8 2017, 9:32 AM
fpetrogalli updated the summary of D30739: [OpenMP] "declare simd" for AArch64 Advanced SIMD..
Mar 8 2017, 8:19 AM
fpetrogalli created D30739: [OpenMP] "declare simd" for AArch64 Advanced SIMD..
Mar 8 2017, 8:15 AM

Jan 23 2017

fpetrogalli added a comment to D27249: [LoopVectorize] Use OpenMP vector routines..

All, as we discussed, we agree to follow the GCC VectorABI and be GCC VectorABI compatible, right? It could be happened, I am looking at an old patch. Thanks.

Jan 23 2017, 7:10 AM
fpetrogalli updated the diff for D27249: [LoopVectorize] Use OpenMP vector routines..

This is just a rebase with a recent llvm codebase.

Jan 23 2017, 6:50 AM
fpetrogalli updated the diff for D27250: [OpenMP] TargetLibraryInfo from "declare simd"..

I have updated the patch to use the name mangling scheme we agreed - compatible with the GCC name mangling scheme.

Jan 23 2017, 6:49 AM

Dec 12 2016

fpetrogalli updated the diff for D27250: [OpenMP] TargetLibraryInfo from "declare simd"..

I have updated the name mangling of the associated vector function.

Dec 12 2016, 5:39 AM
fpetrogalli updated the diff for D27249: [LoopVectorize] Use OpenMP vector routines..
  1. rebase
  2. Fix bug in copy constructor of the TargetLibraryInfoImpl class - I have missed copying out the multimap needed in the algorithm
  3. Fix bug in the TLII::mangle method (now it returns a std::string instead of a StringRef)
Dec 12 2016, 5:32 AM

Dec 9 2016

fpetrogalli added a comment to D27249: [LoopVectorize] Use OpenMP vector routines..

I have removed all reviewers, this is not intended for code review.

Dec 9 2016, 7:13 AM
fpetrogalli added a comment to D27250: [OpenMP] TargetLibraryInfo from "declare simd"..

I have removed all reviewer, this is not intended for code review.

Dec 9 2016, 7:12 AM
fpetrogalli removed reviewers for D27250: [OpenMP] TargetLibraryInfo from "declare simd".: ABataev, hfinkel, cfe-commits.
Dec 9 2016, 7:11 AM
fpetrogalli removed reviewers for D27249: [LoopVectorize] Use OpenMP vector routines.: mzolotukhin, hfinkel, llvm-commits.
Dec 9 2016, 6:56 AM

Dec 5 2016

fpetrogalli added a comment to D27249: [LoopVectorize] Use OpenMP vector routines..

Just a high level question: why can't this extend the same insert mechanism we already use for the other vectorised functions, instead of creating a new one for OpenMP?

Dec 5 2016, 2:49 AM

Nov 30 2016

fpetrogalli added a comment to D27250: [OpenMP] TargetLibraryInfo from "declare simd"..

You did just some minor changes. The main questions are still unanswered.

Nov 30 2016, 7:41 AM
fpetrogalli added a comment to D27250: [OpenMP] TargetLibraryInfo from "declare simd"..
Nov 30 2016, 7:41 AM
fpetrogalli added a comment to D27250: [OpenMP] TargetLibraryInfo from "declare simd"..

I updated the comments related to code formatting.

Nov 30 2016, 7:05 AM
fpetrogalli updated the diff for D27249: [LoopVectorize] Use OpenMP vector routines..

Rebase plus clang format.

Nov 30 2016, 7:01 AM
fpetrogalli updated the diff for D27250: [OpenMP] TargetLibraryInfo from "declare simd"..

Rebase plus apply clang format.

Nov 30 2016, 6:57 AM
fpetrogalli added a comment to D27256: [ARM] Add ARM contributions agreement..

If that's there already, why add a new one on the root of the project?

Nov 30 2016, 4:57 AM
fpetrogalli added a comment to D27256: [ARM] Add ARM contributions agreement..

Er, will we need one of those per company? Odd...

Nov 30 2016, 4:33 AM
fpetrogalli retitled D27256: [ARM] Add ARM contributions agreement. from to [ARM] Add ARM contributions agreement..
Nov 30 2016, 4:10 AM
fpetrogalli added reviewers for D27249: [LoopVectorize] Use OpenMP vector routines.: hfinkel, mzolotukhin.
Nov 30 2016, 3:39 AM
fpetrogalli updated subscribers of D27249: [LoopVectorize] Use OpenMP vector routines..
Nov 30 2016, 2:32 AM
fpetrogalli added reviewers for D27250: [OpenMP] TargetLibraryInfo from "declare simd".: ABataev, hfinkel.
Nov 30 2016, 2:30 AM
fpetrogalli retitled D27250: [OpenMP] TargetLibraryInfo from "declare simd". from to [OpenMP] TargetLibraryInfo from "declare simd"..
Nov 30 2016, 2:21 AM
fpetrogalli added a reviewer for D27249: [LoopVectorize] Use OpenMP vector routines.: llvm-commits.
Nov 30 2016, 2:15 AM
fpetrogalli retitled D27249: [LoopVectorize] Use OpenMP vector routines. from to [LoopVectorize] Use OpenMP vector routines..
Nov 30 2016, 1:53 AM

Oct 12 2016

fpetrogalli added a comment to D24951: Import/adapt the SLEEF vector math-function library as an LLVM runtime.

The library name has been updated to be vmath. The function names now all look like:

__llvm_<func>_<type>[_<abi>]

So, for example, for sin the library contains on my x86_64 system:

__llvm_sin_f64
__llvm_sin_u1_f64
__llvm_sin_f32
__llvm_sin_u1_f32
__llvm_sin_u1_v2f64_sse2
__llvm_sin_v2f64_sse2
__llvm_sin_u1_v4f32_sse2
__llvm_sin_v4f32_sse2
__llvm_sin_u1_v4f64_avx
__llvm_sin_v4f64_avx
__llvm_sin_u1_v8f32_avx
__llvm_sin_v8f32_avx
__llvm_sin_u1_v4f64_avx2
__llvm_sin_v4f64_avx2
__llvm_sin_u1_v8f32_avx2
__llvm_sin_v8f32_avx2
Oct 12 2016, 4:55 AM