[PATCH][x86] Enable broadwell march to llvm/clang

Robert Khasanov rob.khasanov at gmail.com
Fri Sep 19 03:40:32 PDT 2014


Thank you for review!
Committed rev. 218116, 218117, 218118.

2014-09-18 21:13 GMT+04:00 Eric Christopher <echristo at gmail.com>:

> Cool, a comment to that effect in the code would probably be nice.
>
> Otherwise LGTM.
>
> -eric
>
> On Thu, Sep 18, 2014 at 6:14 AM, Robert Khasanov <rob.khasanov at gmail.com>
> wrote:
>
>> Difference between two sets of builtins that _addcarryx can be used only
>> if ADX feature available. Whereas _addcarry and _subborrow can be used w/o
>> ADX CPUID.
>> That's why I included "#ifdef __ADX__" line inside adxintrin.h, put
>> _addcarryx intrinsics under #ifdef and other outside #ifdef, finally
>> divided into two patches.
>>
>>
>> 2014-09-17 21:34 GMT+04:00 Eric Christopher <echristo at gmail.com>:
>>
>>>
>>> +#include <adxintrin.h>
>>> +
>>>
>>> #ifdef __ADX__ ?
>>>
>>> And any reason the two sets of builtins aren't in the same patch or were
>>> you planning on one large commit for this?
>>>
>>> -eric
>>>
>>>
>>> On Wed, Sep 17, 2014 at 10:30 AM, Reid Kleckner <rnk at google.com> wrote:
>>>
>>>> Clang patches LGTM with some tweaks.
>>>>
>>>> +// RUN: %clang_cc1 -triple x86_64-unknown-unknown -target-feature +adx
>>>> -emit-llvm -o - %s | FileCheck %s
>>>> +
>>>> +// Don't include mm_malloc.h, it's system specific.
>>>> +#define __MM_MALLOC_H
>>>>
>>>> You can use the -ffreestanding flag to avoid this quirk, I believe.
>>>>
>>>> On Wed, Sep 17, 2014 at 1:31 AM, Robert Khasanov <
>>>> rob.khasanov at gmail.com> wrote:
>>>>
>>>>> Ping
>>>>> Could anyone review clang patches?
>>>>>
>>>>> 2014-08-25 15:58 GMT+04:00 Robert Khasanov <rob.khasanov at gmail.com>:
>>>>>
>>>>>> Ping
>>>>>>
>>>>>>
>>>>>>
>>>>>> 2014-08-21 13:53 GMT+04:00 Robert Khasanov <rob.khasanov at gmail.com>:
>>>>>>
>>>>>>> Thanks for review, Nadav!
>>>>>>>
>>>>>>> Committed rev216161, 216162, 216163, 216164.
>>>>>>>
>>>>>>> Could anyone review clang patches?
>>>>>>>
>>>>>>>
>>>>>>> 2014-08-20 20:41 GMT+04:00 Nadav Rotem <nrotem at apple.com>:
>>>>>>>
>>>>>>> The LLVM patches LGTM.  Thanks Robert!
>>>>>>>>
>>>>>>>>
>>>>>>>> > On Aug 20, 2014, at 8:55 AM, Robert Khasanov <
>>>>>>>> rob.khasanov at gmail.com> wrote:
>>>>>>>> >
>>>>>>>> > Hi All,
>>>>>>>> >
>>>>>>>> > These patches enable Broadwell subtarget:
>>>>>>>> > Broadwell ISA includes Haswell ISA + ADX + RDSEED + SMAP
>>>>>>>> > New instructions were already enabled. These patches define
>>>>>>>> broadwell cpu, SMAP Feature and add intrinsics to LLVM/clang
>>>>>>>> >
>>>>>>>> <001_llvm_broadwell_target.patch><002_llvm_adox_adcx.patch><003_llvm_smap.patch><004_llvm_adc_sbb.patch><005_clang_broadwell_target.patch><006_clang_adx.patch><007_clang_adc_sbb.patch>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140919/307871c4/attachment.html>


More information about the llvm-commits mailing list