Most of AVX-512 intrinsics were not supported in 3.8 version and added between 3.8 and 3.9. In this case, when a BE intrinsic was introduced between releases, it can be removed, right?
Anyway, we’d better keep the BE intrinsic inside than maintain it in the autoupgrade. > Also should we change the AutoUpgrade code for the cmp intrinsics in the > backend to also use zero instead of undef? This pass was initially designed in order to keep backward compatibility and not for optimizations, right? - Elena From: Craig Topper [mailto:craig.top...@gmail.com] Sent: Tuesday, July 05, 2016 09:39 To: Demikhovsky, Elena <elena.demikhov...@intel.com> Cc: Breger, Igor <igor.bre...@intel.com>; Eric Christopher via cfe-commits <cfe-commits@lists.llvm.org>; LLVM Commits (llvm-comm...@lists.llvm.org) <llvm-comm...@lists.llvm.org> Subject: Re: r274110 - [AVX512] Zero extend cmp intrinsic return value. Unfortunately the policy is such that the current version of LLVM must be able to read a bitcode or IR text file generated by any released version of LLVM/clang from 3.0 till now. This is why the AutoUpgrade.cpp file exists and it keeps growing and its only going to get worse. Hopefully with the discussion about the version number for the release after 3.9 we end up with a time based compatibility policy. This will allow us to remove the auto upgrade code over time, but it will take many years before we can remove auto upgrade being added now. On Monday, July 4, 2016, Demikhovsky, Elena <elena.demikhov...@intel.com<mailto:elena.demikhov...@intel.com>> wrote: I think that AVX-512 intrinsic handled in Clang should be removed from AutoUpgrade. If we decide to replace a call to llvm-backend intrinsic with generating LLVM IR directly from clang, we can remove llvm-backend intrinsic at all. Otherwise we’ll need to duplicate support in FE and BE for all AVX-512 intrinsics. - Elena From: Craig Topper [mailto:craig.top...@gmail.com<javascript:_e(%7B%7D,'cvml','craig.top...@gmail.com');>] Sent: Monday, July 04, 2016 21:53 To: Breger, Igor <igor.bre...@intel.com<javascript:_e(%7B%7D,'cvml','igor.bre...@intel.com');>> Cc: Eric Christopher via cfe-commits <cfe-commits@lists.llvm.org<javascript:_e(%7B%7D,'cvml','cfe-commits@lists.llvm.org');>>; Demikhovsky, Elena <elena.demikhov...@intel.com<javascript:_e(%7B%7D,'cvml','elena.demikhov...@intel.com');>> Subject: Re: r274110 - [AVX512] Zero extend cmp intrinsic return value. We have to keep the autoupgrade code due to the requirement that we need to be able to load bitcode files from previous versions of llvm. Though the intrinsic autograde code is starting to become large. ~Craig On Mon, Jul 4, 2016 at 2:55 AM, Breger, Igor <igor.bre...@intel.com<javascript:_e(%7B%7D,'cvml','igor.bre...@intel.com');>> wrote: Thanks for working on this! Regards, Igor From: Craig Topper [mailto:craig.top...@gmail.com<javascript:_e(%7B%7D,'cvml','craig.top...@gmail.com');>] Sent: Monday, July 04, 2016 10:20 To: Breger, Igor Cc: Eric Christopher via cfe-commits; Demikhovsky, Elena; Ouriel, Boaz Subject: Re: r274110 - [AVX512] Zero extend cmp intrinsic return value. I've modified the indices we use for the zero vector in r274484. This gives better codegen as it now looks like a concat to the backend. So now we just end up emitting unnecessary kshiftlw/kshiftrw pairs instead of converting to a wider vector op and back. ~Craig On Sun, Jul 3, 2016 at 1:32 PM, Craig Topper <craig.top...@gmail.com<javascript:_e(%7B%7D,'cvml','craig.top...@gmail.com');>> wrote: Also should we change the AutoUpgrade code for the cmp intrinsics in the backend to also use zero instead of undef? ~Craig On Sun, Jul 3, 2016 at 12:11 AM, Breger, Igor <igor.bre...@intel.com<javascript:_e(%7B%7D,'cvml','igor.bre...@intel.com');>> wrote: Hello Craig, Thanks a lot for pointing it out to me. I familiar with this problem, we are planning to improve CodeGen to handle with this case in near future. Regards, Igor From: Craig Topper [mailto:craig.top...@gmail.com<javascript:_e(%7B%7D,'cvml','craig.top...@gmail.com');>] Sent: Saturday, July 02, 2016 08:46 To: Breger, Igor; Eric Christopher via cfe-commits Subject: Re: r274110 - [AVX512] Zero extend cmp intrinsic return value. This change codgens to something really awful now. Can you take a look? .section __TEXT,__text,regular,pure_instructions .section __TEXT,__literal8,8byte_literals .p2align 3 LCPI0_0: .quad -1 .section __TEXT,__const .p2align 6 LCPI0_1: .quad 0 .quad 1 .quad 2 .quad 3 .quad 8 .quad 8 .quad 8 .quad 8 .section __TEXT,__text,regular,pure_instructions .globl _test_mm_cmpeq_epu32_mask .p2align 4, 0x90 _test_mm_cmpeq_epu32_mask: vpcmpeqd %xmm1, %xmm0, %k1 vpbroadcastq LCPI0_0(%rip), %zmm0 {%k1} {z} vpxord %zmm1, %zmm1, %zmm1 vmovdqa64 LCPI0_1(%rip), %zmm2 vpermt2q %zmm1, %zmm2, %zmm0 vpsllq $63, %zmm0, %zmm0 vptestmq %zmm0, %zmm0, %k0 kmovw %k0, %eax retq ~Craig On Wed, Jun 29, 2016 at 1:14 AM, Igor Breger via cfe-commits <cfe-commits@lists.llvm.org<javascript:_e(%7B%7D,'cvml','cfe-commits@lists.llvm.org');>> wrote: Author: ibreger Date: Wed Jun 29 03:14:17 2016 New Revision: 274110 URL: http://llvm.org/viewvc/llvm-project?rev=274110&view=rev Log: [AVX512] Zero extend cmp intrinsic return value. Differential Revision: http://reviews.llvm.org/D21746 Modified: cfe/trunk/lib/CodeGen/CGBuiltin.cpp cfe/trunk/test/CodeGen/avx512vl-builtins.c Modified: cfe/trunk/lib/CodeGen/CGBuiltin.cpp URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGBuiltin.cpp?rev=274110&r1=274109&r2=274110&view=diff ============================================================================== --- cfe/trunk/lib/CodeGen/CGBuiltin.cpp (original) +++ cfe/trunk/lib/CodeGen/CGBuiltin.cpp Wed Jun 29 03:14:17 2016 @@ -6460,8 +6460,8 @@ static Value *EmitX86MaskedCompare(CodeG Indices[i] = i; for (unsigned i = NumElts; i != 8; ++i) Indices[i] = NumElts; - Cmp = CGF.Builder.CreateShuffleVector(Cmp, UndefValue::get(Cmp->getType()), - Indices); + Cmp = CGF.Builder.CreateShuffleVector( + Cmp, llvm::Constant::getNullValue(Cmp->getType()), Indices); } return CGF.Builder.CreateBitCast(Cmp, IntegerType::get(CGF.getLLVMContext(), Modified: cfe/trunk/test/CodeGen/avx512vl-builtins.c URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/avx512vl-builtins.c?rev=274110&r1=274109&r2=274110&view=diff ============================================================================== --- cfe/trunk/test/CodeGen/avx512vl-builtins.c (original) +++ cfe/trunk/test/CodeGen/avx512vl-builtins.c Wed Jun 29 03:14:17 2016 @@ -8,6 +8,7 @@ __mmask8 test_mm_cmpeq_epu32_mask(__m128i __a, __m128i __b) { // CHECK-LABEL: @test_mm_cmpeq_epu32_mask // CHECK: icmp eq <4 x i32> %{{.*}}, %{{.*}} + // CHECK: shufflevector <4 x i1> %{{.*}}, <4 x i1> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 4, i32 4, i32 4> return (__mmask8)_mm_cmpeq_epu32_mask(__a, __b); } _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org<javascript:_e(%7B%7D,'cvml','cfe-commits@lists.llvm.org');> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits --------------------------------------------------------------------- Intel Israel (74) Limited This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies. --------------------------------------------------------------------- Intel Israel (74) Limited This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies. --------------------------------------------------------------------- Intel Israel (74) Limited This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies. -- ~Craig --------------------------------------------------------------------- Intel Israel (74) Limited This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
_______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
