RFR: 8266950: Remove vestigial support for non-strict floating-point execution [v2]

[David Holmes]

Correction ...

On 22/05/2021 12:15 pm, David Holmes wrote:
> Hi Vladimir,
> 
> On 22/05/2021 4:19 am, Vladimir Ivanov wrote:
>> On Tue, 18 May 2021 04:26:00 GMT, David Holmes <dholmes at openjdk.org> 
>> wrote:
>>
>>>> As part of JEP 306, the vestiges of HotSpot support for non-strict 
>>>> floating-point execution can be removed. All methods implicitly have 
>>>> strictfp semantics so the explicit checks for is_strict() can be 
>>>> replaced by true and the code reformulated accordingly.
>>>>
>>>> There are still some names that include "strict" that could 
>>>> potentially be renamed to remove it, but the fact we have to have 
>>>> strict fp semantics is still important on some platforms, so the 
>>>> names help reinforce that IMO.
>>>>
>>>> Testing: tiers 1-3
>>>>
>>>> Thanks,
>>>> David
>>>
>>> David Holmes has updated the pull request incrementally with one 
>>> additional commit since the last revision:
>>>
>>>    lir_div_strictfp and lir_mul_strictfp
>>
>> There are some suspicious failures on linux-x86 in pre-submit testing 
>> results:
>> - compiler/c1/Test6855215.java
>> - compiler/intrinsics/string/TestStringLatin1IndexOfChar.java
>>
>> The tests explicitly specify `-XX:UseSSE=0`, so it may be related to 
>> the patch. Anybody interested in linux-x86 want to take a look? @shade 
>> @DamonFool
> 
> I'll take a look at the patch again because it is supposed to involve no 
> functional changes.

That isn't actually true. We are now forcing 32-bit x86 to always have 
strict semantics, where previously they could use non-strict.

David

> Thanks,
> David
> 
>> -------------
>>
>> PR: https://git.openjdk.java.net/jdk/pull/3991
>>