Request for review(XXS): 6946892 c1 shouldn't sign-extend to upper 32bits on x64
igor.veresov at oracle.com
Tue Apr 27 08:46:57 PDT 2010
Thank you, Tom!
I've tested with SPECjvm98 and SPECjvm2008, they work ok. I'll keep an
eye on related things that may bubble up.
On 4/26/10 12:16 PM, Tom Rodriguez wrote:
> This looks good. Have you tested the 64 bit client much after this change? The sign extension here could be hiding some missing sign extension at the LIR level.
> On Apr 23, 2010, at 9:13 PM, Igor Veresov wrote:
>> C1 does sign-extension when it loads ints and shorts from memory to 64-bit registers. This causes problems for C2 because it relies on the fact the int passed in a 64-bit register is zero-extended.
>> Webrev: http://cr.openjdk.java.net/~iveresov/6946892/webrev.00/
More information about the hotspot-compiler-dev