RFR(S): 8187964: [s390][ppc]: Intrinsify Math.multiplyHigh(long, long)

[Schmidt, Lutz]

Andrew,
unsigned multiplyHigh would definitely help [s390]. Think of just one instruction instead of seven. You could even get the full length product at the same cost.
Regards, Lutz

 

On 07.10.2017, 11:09, "Andrew Haley" <aph at redhat.com> wrote:

    One thought about this: we might generate better code on these
    machines if we had an unsigned multiplyHigh intrinsic and did the
    adjustment for signed arithmetic in Java code, where the sign
    adjustment could be scheduled separately by C2.
    
    OK, I'll get on with implementing unsignedMultiplyHigh!
    
    -- 
    Andrew Haley
    Java Platform Lead Engineer
    Red Hat UK Ltd. <https://www.redhat.com>
    EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671