RFR (XS) 8215353: x86_32 build failures after JDK-8214751 (X86: Support for VNNI Instructions)

[Aleksey Shipilev]

Thanks!

This does not all-too-trivial, but I guess there is no sense it trying to push it through jdk-submit
(Oracle does not build and test x86_32?). I think ad-hoc tier1 testing on x86_32 is enough for this.
Agreed?

-Aleksey

On 12/13/18 2:56 PM, Tobias Hartmann wrote:
> Hi Aleksey,
> 
> looks good to me.
> 
> Best regards,
> Tobias
> 
> On 13.12.18 14:47, Aleksey Shipilev wrote:
>> Bug:
>>   https://bugs.openjdk.java.net/browse/JDK-8215353
>>
>> mulI_rReg and friends are not available in x86_32.ad, there are mulI_eReg and friends. I think
>> 32-bit code should use those.
>>
>> Fix:
>>
>> diff -r 9a73a4e4011f -r 6108789077bb src/hotspot/cpu/x86/x86_32.ad
>> --- a/src/hotspot/cpu/x86/x86_32.ad     Thu Dec 13 15:11:25 2018 +0530
>> +++ b/src/hotspot/cpu/x86/x86_32.ad     Thu Dec 13 14:10:00 2018 +0100
>> @@ -7760,9 +7760,9 @@
>>    match(Set dst (MulAddS2I (Binary dst src1) (Binary src2 src3)));
>>    effect(KILL cr, KILL src2);
>>
>> -  expand %{ mulI_rReg(dst, src1, cr);
>> -           mulI_rReg(src2, src3, cr);
>> -           addI_rReg(dst, src2, cr); %}
>> +  expand %{ mulI_eReg(dst, src1, cr);
>> +           mulI_eReg(src2, src3, cr);
>> +           addI_eReg(dst, src2, cr); %}
>>  %}
>>
>>  // Multiply Register Int to Long
>>
>>
>> Testing: x86_32: build, compiler/loopopts/superword/Vec_MulAddS2I.java, hotspot/tier1
>>
>> Thanks,
>> -Aleksey
>>


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20181213/c14830e4/signature.asc>