RFR [XS] 8041957: -XX:UseAVX=0 cause assert(UseAVX) failed

[Christian Thalinger]

On Apr 29, 2014, at 8:18 AM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:

> On 4/28/14 5:45 PM, Christian Thalinger wrote:
>> FlagSetting didn’t work?
> 
> It only works for bool flags:
> 
> FlagSetting(bool& fl, bool newValue)

Ah, right.  Use have to use UIntFlagSetting instead.

> 
> Error: Could not find a match for FlagSetting::FlagSetting(int, int) needed in VM_Version_StubGenerator::generate_get_cpu_info().
> 
> Vladimir
> 
>> 
>> On Apr 28, 2014, at 12:58 PM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:
>> 
>>> I updated changes:
>>> 
>>> http://cr.openjdk.java.net/~kvn/8041957/webrev
>>> 
>>> Thanks,
>>> Vladimir
>>> 
>>> On 4/25/14 7:12 PM, Vladimir Kozlov wrote:
>>>> On 4/25/14 6:37 PM, Christian Thalinger wrote:
>>>>> This is tricky.  Changing the assert means that we cannot catch the
>>>>> case where we were asked to not use AVX but we do.  Maybe we should
>>>>> enable AVX instruction temporarily by using:
>>>>> 
>>>>> FlagSetting avx(UseAVX, 99);
>>>>> 
>>>>> in generate_get_cpu_info() to be able to generate the instructions.
>>>> 
>>>> I will try it next week.
>>>> 
>>>> Thanks,
>>>> Vladimir
>>>> 
>>>>> 
>>>>> On Apr 25, 2014, at 9:13 AM, Vladimir Kozlov
>>>>> <vladimir.kozlov at oracle.com> wrote:
>>>>> 
>>>>>> http://cr.openjdk.java.net/~kvn/8041957/webrev
>>>>>> https://bugs.openjdk.java.net/browse/JDK-8041957
>>>>>> 
>>>>>> Use VM_Version::supports_avx() instead of UseAVX in vmovdqu
>>>>>> instruction asserts.
>>>>>> 
>>>>>> Thanks,
>>>>>> Vladimir
>>>>> 
>> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20140429/dd602f91/attachment-0001.html>