r250470 - Fix the subtarget features required by some x86 builtins.

Eric Christopher via cfe-commits cfe-commits at lists.llvm.org
Tue Oct 27 12:12:55 PDT 2015 

OK, so I fixed this last night in r251388. I decided to go with "," means
"and" and "|" means "or" and go from there. If we need a more complicated
syntax we'll go back to the drawing board :)

-eric

On Sun, Oct 18, 2015 at 9:57 AM Craig Topper <craig.topper at gmail.com> wrote:

> Interestingly, gcc doesn't get this right either
>
> *test.cpp:7:64:* *error: *‘*__builtin_ia32_xorpd256_mask*’ needs isa
> option -m32 -mavx512dq -mavx512vl
>
>
> I'm not even sure why it bothered to list -m32.
>
> If you then give it -mavx512dq without -mavx512vl it proceeds to throw
> internal compiler errors.
>
>
>
> *test.cpp:8:1:* *error: *unrecognizable insn:
>
>  }
>
> * ^*
>
> (insn 15 14 18 2 (set (reg:V4DF 89 [ D.22466 ])
>
>         (vec_merge:V4DF (xor:V4DF (reg:V4DF 92)
>
>                 (mem/c:V4DF (plus:DI (reg/f:DI 82 virtual-stack-vars)
>
>                         (const_int -96 [0xffffffffffffffa0])) [0 __A+0 S32
> A256]))
>
>             (reg:V4DF 93)
>
>             (subreg:QI (reg:SI 94) 0))) test.cpp:7 -1
>
>      (nil))
>
> *test.cpp:8:1:* *internal compiler error: *in extract_insn, at
> recog.c:2343
>
>
> *test.cpp:8:1:* *internal compiler error: *Abort trap: 6
>
> *g++:* *internal compiler error: *Abort trap: 6 (program cc1plus)
>
> On Fri, Oct 16, 2015 at 6:50 PM, Eric Christopher <echristo at gmail.com>
> wrote:
>
>> Sure, but if someone is planning on doing this for ARM there's some work
>> in the ARM TargetInfo that needs to be done :)
>>
>> -eric
>>
>> On Fri, Oct 16, 2015 at 3:24 PM Justin Bogner <mail at justinbogner.com>
>> wrote:
>>
>>> Looks like we'll need "and" for ARM's builtins as well - several require
>>> things like neon+fp16 or neon+vfp4.
>>>
>>> Eric Christopher <echristo at gmail.com> writes:
>>> > Right. My current direction is going to be:
>>> >
>>> > "foo+bar" we can represent as "and"
>>> > "foo,bar" we'll continue to represent as "or"
>>> >
>>> > this will lead to things like (possibly):
>>> >
>>> > "foo+bar, baz" - "either foo and bar, or baz."
>>> >
>>> > -eric
>>> >
>>> > On Fri, Oct 16, 2015 at 11:41 AM Justin Bogner <mail at justinbogner.com>
>>> > wrote:
>>> >
>>> >> Hm. Well, supporting both "and" and "or" combinations here sounds like
>>> >> an unpleasant path to go down. I guess the question is, which is
>>> >> weirder?
>>> >>
>>> >> - With fma, we have fma4 which is basically a synonym as far as
>>> >>   intrinsics go - some chips have 3-argument versions and some have 4,
>>> >>   but I don't think any of the intrinsics reflect that.
>>> >>
>>> >> - With avx512, we have combinations of features which introduce new
>>> >>   intrinsics, since the supported elements and supported vector types
>>> >>   are more or less orthogonal.
>>> >>
>>> >> I guess the fma situation is easier to special case if we want to
>>> avoid
>>> >> having to parse expressions here. Ie, we could have some kind of
>>> >> meta-feature that expands to "fma or fma4". Of course, the obvious
>>> name
>>> >> for that is "fma", which means fma3 today...
>>> >>
>>> >> Eric Christopher via cfe-commits <cfe-commits at lists.llvm.org> writes:
>>> >> > Lovely. Cc'ing Justin here. We'll need to figure out some change for
>>> >> these
>>> >> > builtins here if we need to have and as well as or.
>>> >> >
>>> >> > (Why on earth are these builtins subject to two separate features?
>>> *sigh*
>>> >> > It's ridiculous.)
>>> >> >
>>> >> > -eric
>>> >> >
>>> >> > On Thu, Oct 15, 2015 at 11:59 PM Craig Topper <
>>> craig.topper at gmail.com>
>>> >> > wrote:
>>> >> >
>>> >> >> Correct you avx512vl means it support 128 and 256-bit vectors.
>>> avx512bw
>>> >> >> means it supports byte and word elements. So you actually need
>>> both.
>>> >> >>
>>> >> >> On Thu, Oct 15, 2015 at 11:57 PM, Eric Christopher <
>>> echristo at gmail.com>
>>> >> >> wrote:
>>> >> >>
>>> >> >>> The comma separated list is currently represented as "one of
>>> these". I
>>> >> >>> couldn't parse your first sentence, for the avx512 ones are you
>>> saying
>>> >> that
>>> >> >>> it requires both and not just one of the options?
>>> >> >>>
>>> >> >>> -eric
>>> >> >>>
>>> >> >>> On Thu, Oct 15, 2015 at 11:55 PM Craig Topper <
>>> craig.topper at gmail.com>
>>> >> >>> wrote:
>>> >> >>>
>>> >> >>>> So for the AVX512 ones that list 2 features those features are
>>> both
>>> >> >>>> required, but for FMA you need either one of the features but not
>>> >> both.
>>> >> >>>> What is the comma separated list currently implemented as?
>>> >> >>>>
>>> >> >>>> On Thu, Oct 15, 2015 at 3:46 PM, Eric Christopher via
>>> cfe-commits <
>>> >> >>>> cfe-commits at lists.llvm.org> wrote:
>>> >> >>>>
>>> >> >>>>> Author: echristo
>>> >> >>>>> Date: Thu Oct 15 17:46:02 2015
>>> >> >>>>> New Revision: 250470
>>> >> >>>>>
>>> >> >>>>> URL: http://llvm.org/viewvc/llvm-project?rev=250470&view=rev
>>> >> >>>>> Log:
>>> >> >>>>> Fix the subtarget features required by some x86 builtins.
>>> >> >>>>>
>>> >> >>>>> Update the fma builtins to be fma/fma4 until some we can find
>>> some
>>> >> >>>>> documentation either way.
>>> >> >>>>>
>>> >> >>>>> Update a couple of the avx intrinsics because they were in the
>>> wrong
>>> >> >>>>> category.
>>> >> >>>>>
>>> >> >>>>> Modified:
>>> >> >>>>>     cfe/trunk/include/clang/Basic/BuiltinsX86.def
>>> >> >>>>>
>>> >> >>>>> Modified: cfe/trunk/include/clang/Basic/BuiltinsX86.def
>>> >> >>>>> URL:
>>> >> >>>>>
>>> >>
>>> http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/BuiltinsX86.def?rev=250470&r1=250469&r2=250470&view=diff
>>> >> >>>>>
>>> >> >>>>>
>>> >>
>>> ==============================================================================
>>> >> >>>>> --- cfe/trunk/include/clang/Basic/BuiltinsX86.def (original)
>>> >> >>>>> +++ cfe/trunk/include/clang/Basic/BuiltinsX86.def Thu Oct 15
>>> 17:46:02
>>> >> >>>>> 2015
>>> >> >>>>> @@ -718,38 +718,38 @@ TARGET_BUILTIN(__builtin_ia32_sha256msg1
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_sha256msg2, "V4iV4iV4i", "",
>>> "sha")
>>> >> >>>>>
>>> >> >>>>>  // FMA
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmaddps, "V4fV4fV4fV4f", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmaddpd, "V2dV2dV2dV2d", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmaddss, "V4fV4fV4fV4f", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmaddsd, "V2dV2dV2dV2d", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmsubps, "V4fV4fV4fV4f", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmsubpd, "V2dV2dV2dV2d", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmsubss, "V4fV4fV4fV4f", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmsubsd, "V2dV2dV2dV2d", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmaddps, "V4fV4fV4fV4f", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmaddpd, "V2dV2dV2dV2d", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmaddss, "V4fV4fV4fV4f", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmaddsd, "V2dV2dV2dV2d", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmsubps, "V4fV4fV4fV4f", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmsubpd, "V2dV2dV2dV2d", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmsubss, "V4fV4fV4fV4f", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmsubsd, "V2dV2dV2dV2d", "",
>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmaddsubps, "V4fV4fV4fV4f", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmaddsubpd, "V2dV2dV2dV2d", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmsubaddps, "V4fV4fV4fV4f", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmsubaddpd, "V2dV2dV2dV2d", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmaddps256, "V8fV8fV8fV8f", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmaddpd256, "V4dV4dV4dV4d", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmsubps256, "V8fV8fV8fV8f", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmsubpd256, "V4dV4dV4dV4d", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmaddps256, "V8fV8fV8fV8f", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmaddpd256, "V4dV4dV4dV4d", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmsubps256, "V8fV8fV8fV8f", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfnmsubpd256, "V4dV4dV4dV4d", "",
>>> >> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmaddsubps256, "V8fV8fV8fV8f",
>>> "",
>>> >> >>>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmaddsubpd256, "V4dV4dV4dV4d",
>>> "",
>>> >> >>>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmsubaddps256, "V8fV8fV8fV8f",
>>> "",
>>> >> >>>>> "fma")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_vfmsubaddpd256, "V4dV4dV4dV4d",
>>> "",
>>> >> >>>>> "fma")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmaddps, "V4fV4fV4fV4f", "",
>>> >> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmaddpd, "V2dV2dV2dV2d", "",
>>> >> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmaddss, "V4fV4fV4fV4f", "",
>>> >> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmaddsd, "V2dV2dV2dV2d", "",
>>> >> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmsubps, "V4fV4fV4fV4f", "",
>>> >> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmsubpd, "V2dV2dV2dV2d", "",
>>> >> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmsubss, "V4fV4fV4fV4f", "",
>>> >> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmsubsd, "V2dV2dV2dV2d", "",
>>> >> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmaddps, "V4fV4fV4fV4f", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmaddpd, "V2dV2dV2dV2d", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmaddss, "V4fV4fV4fV4f", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmaddsd, "V2dV2dV2dV2d", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmsubps, "V4fV4fV4fV4f", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmsubpd, "V2dV2dV2dV2d", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmsubss, "V4fV4fV4fV4f", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmsubsd, "V2dV2dV2dV2d", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmaddsubps, "V4fV4fV4fV4f", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmaddsubpd, "V2dV2dV2dV2d", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmsubaddps, "V4fV4fV4fV4f", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmsubaddpd, "V2dV2dV2dV2d", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmaddps256, "V8fV8fV8fV8f", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmaddpd256, "V4dV4dV4dV4d", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmsubps256, "V8fV8fV8fV8f", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmsubpd256, "V4dV4dV4dV4d", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmaddps256, "V8fV8fV8fV8f", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmaddpd256, "V4dV4dV4dV4d", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmsubps256, "V8fV8fV8fV8f", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfnmsubpd256, "V4dV4dV4dV4d", "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmaddsubps256, "V8fV8fV8fV8f",
>>> "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmaddsubpd256, "V4dV4dV4dV4d",
>>> "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmsubaddps256, "V8fV8fV8fV8f",
>>> "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_vfmsubaddpd256, "V4dV4dV4dV4d",
>>> "",
>>> >> >>>>> "fma,fma4")
>>> >> >>>>>
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_vfmaddpd128_mask,
>>> "V2dV2dV2dV2dUc",
>>> >> "",
>>> >> >>>>> "avx512vl")
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_vfmaddpd128_mask3,
>>> "V2dV2dV2dV2dUc",
>>> >> "",
>>> >> >>>>> "avx512vl")
>>> >> >>>>> @@ -952,8 +952,8 @@ TARGET_BUILTIN(__builtin_ia32_cmpps256_m
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_cmpps128_mask,   "UcV4fV4fIiUc",
>>> "",
>>> >> >>>>> "avx512vl")
>>> >> >>>>>
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_pcmpeqb512_mask,
>>> "LLiV64cV64cLLi", "",
>>> >> >>>>> "avx512bw")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_pcmpeqd512_mask, "sV16iV16is",
>>> "",
>>> >> >>>>> "avx512bw")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_pcmpeqq512_mask, "cV8LLiV8LLic",
>>> "",
>>> >> >>>>> "avx512bw")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_pcmpeqd512_mask, "sV16iV16is",
>>> "",
>>> >> >>>>> "avx512f")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_pcmpeqq512_mask, "cV8LLiV8LLic",
>>> "",
>>> >> >>>>> "avx512f")
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_pcmpeqw512_mask, "iV32sV32si",
>>> "",
>>> >> >>>>> "avx512bw")
>>> >> >>>>>
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_pcmpeqb256_mask, "iV32cV32ci",
>>> "",
>>> >> >>>>> "avx512vl,avx512bw")
>>> >> >>>>> @@ -966,8 +966,8 @@ TARGET_BUILTIN(__builtin_ia32_pcmpeqq128
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_pcmpeqw128_mask, "cV8sV8sc", "",
>>> >> >>>>> "avx512vl,avx512bw")
>>> >> >>>>>
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_pcmpgtb512_mask,
>>> "LLiV64cV64cLLi", "",
>>> >> >>>>> "avx512bw")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_pcmpgtd512_mask, "sV16iV16is",
>>> "",
>>> >> >>>>> "avx512bw")
>>> >> >>>>> -TARGET_BUILTIN(__builtin_ia32_pcmpgtq512_mask, "cV8LLiV8LLic",
>>> "",
>>> >> >>>>> "avx512bw")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_pcmpgtd512_mask, "sV16iV16is",
>>> "",
>>> >> >>>>> "avx512f")
>>> >> >>>>> +TARGET_BUILTIN(__builtin_ia32_pcmpgtq512_mask, "cV8LLiV8LLic",
>>> "",
>>> >> >>>>> "avx512f")
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_pcmpgtw512_mask, "iV32sV32si",
>>> "",
>>> >> >>>>> "avx512bw")
>>> >> >>>>>
>>> >> >>>>>  TARGET_BUILTIN(__builtin_ia32_pcmpgtb256_mask, "iV32cV32ci",
>>> "",
>>> >> >>>>> "avx512vl,avx512bw")
>>> >> >>>>>
>>> >> >>>>>
>>> >> >>>>> _______________________________________________
>>> >> >>>>> cfe-commits mailing list
>>> >> >>>>> cfe-commits at lists.llvm.org
>>> >> >>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
>>> >> >>>>>
>>> >> >>>>
>>> >> >>>>
>>> >> >>>>
>>> >> >>>> --
>>> >> >>>> ~Craig
>>> >> >>>>
>>> >> >>>
>>> >> >>
>>> >> >>
>>> >> >> --
>>> >> >> ~Craig
>>> >> >>
>>> >> > _______________________________________________
>>> >> > cfe-commits mailing list
>>> >> > cfe-commits at lists.llvm.org
>>> >> > http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
>>> >>
>>>
>>
>
>
> --
> ~Craig
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20151027/a09bd0a6/attachment-0001.html>

More information about the cfe-commits mailing list