This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] [ARM] neon-testgen.ml typo


On 12 November 2015 at 23:18, Christophe Lyon
<christophe.lyon@linaro.org> wrote:
> On 6 November 2015 at 21:29, Christophe Lyon <christophe.lyon@linaro.org> wrote:
>> On 4 November 2015 at 13:16, Ramana Radhakrishnan
>> <ramana.gcc@googlemail.com> wrote:
>>> On Fri, Oct 30, 2015 at 2:42 PM, Christophe Lyon
>>> <christophe.lyon@linaro.org> wrote:
>>>> On 30 October 2015 at 15:33, Ramana Radhakrishnan
>>>> <ramana.radhakrishnan@foss.arm.com> wrote:
>>>>>
>>>>>
>>>>> On 29/10/15 17:23, Jim Wilson wrote:
>>>>>> I noticed a comment typo in this file while using grep to look for
>>>>>> other stuff.  The typo is easy to fix.
>>>>>>
>>>>>> I tried running neon-testgen.ml to verify, but it is apparently no
>>>>>> longer valid ocaml, as it doesn't work with the ocamlc 4.01.0 I have
>>>>>> on Ubuntu 14.04.  I get a syntax error.  Someone who knows ocaml will
>>>>>> have to fix this.  Meanwhile, the patch to fix the typo should still
>>>>>> be OK, as this is a separate problem.
>>>>>>
>>>>>> Jim
>>>>>>
>>>>>
>>>>> This is OK.
>>>>>
>>>>> I'd really like neon-testgen.ml and the tests in gcc.target/arm/neon to be removed if all the intrinsics are now tested from Christophe's work in getting his advsimd tests integrated. Where are we on that ?
>>>>>
>>>>
>>>> The tests I added cover all ARMv7 intrinsics. I have converted all my tests.
>>>
>>> Good and thank you for doing that.
>>>
>>>>
>>>> There were a few additions to support some aarch64 specific intrinsics.
>>>>
>>>> However, most of the tests in gcc.target/arm/neon contain scan-asm
>>>> directives which mine don't.
>>>> My tests do check functionality (they are executable, comparing their
>>>> results against expected values).
>>>
>>> I don't think the scan-asm in those tests gives you anything useful at
>>> O0 with undefined behaviour in the testcases. In any case for the more
>>> esoteric intrinsics (i.e. the ones that do saturation etc..) having
>>> the run time tests is a better test. I do not see this as being useful
>>> any more in terms of the testing coverage this provides.
>>
>> Ideally we'd need to add more tests for the new armv8 intrinsics....
>>
>>> A patch to remove neon-testgen.ml and the tests in gcc.target/arm/neon
>>> is pre-approved.
>>>
>> OK I'l give it a look, but that will after e/o stage1 probably.
>>
>
> I have just removed them all, as r230274.
>
> ChangeLog is attached.
>

Hmm. I hadn't noticed that the crypto intrinsics tests where generated by
neon-testgen.ml, I thought they were hand-written.
The tests I added do not cover the crypto intrinsics, so I'm going
to revert r230274 and restore all the tests generated by neon-testgen.ml
until we have better coverage in advsimd-intrinsics.

Sorry for the oversight.

Christophe.


> Christophe.
>
>>>
>>> regards
>>> Ramana
>>>
>>>
>>>
>>>>
>>>> Christophe.
>>>>
>>>>> regards
>>>>> Ramana


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]