RFR(M): 8207343: Automate vtable/itable stub size calculation

Schmidt, Lutz lutz.schmidt at sap.com
Thu Aug 2 12:02:43 UTC 2018


Hi Zhongwei,

thank you for testing aarch64. Given my lack of expertise, I am surprised there are only those two issues.
Ad 1.: fixed. The declaration just a few lines above was forgotten to adapt. 
Ad 2.: adapted. I pushed the estimate to 152 bytes. I expected this value would require some adjustment.

Please find a new webrev with the changes at http://cr.openjdk.java.net/~lucy/webrevs/8207343.01

Best Regards,
Lutz

On 02.08.18, 04:45, "Zhongwei Yao" <Zhongwei.Yao at arm.com> wrote:

    Hi, Lutz,
    
    I have tested it on aarch64 by running jtreg tests. And find two tiny issues in vtableStubs_aarch64.cpp's VtableStubs::create_itable_stub function:
    1. typecheckSize on line 212 is not defined.
    2. estimate on line 211 is not large enough, I get 148 in gc/g1/TestFromCardCacheIndex.java case. Here is the assertion failure from that case:
        assert(slop_delta >= 0) failed: itable #3: Code size estimate (140) for lookup_interface_method too small, required: 148
    
    However, I don't have tested the modification in vtableStubs_arm.cpp due to I don't have an arm32 environment at hand.
    
    --
    Best regards,
    Zhongwei
    
    ________________________________________
    From: hotspot-compiler-dev <hotspot-compiler-dev-bounces at openjdk.java.net> on behalf of Schmidt, Lutz <lutz.schmidt at sap.com>
    Sent: Monday, July 30, 2018 3:57:05 PM
    To: hotspot-compiler-dev at openjdk.java.net
    Subject: RFR(M): 8207343: Automate vtable/itable stub size calculation
    
    Dear all,
    
    may I please request reviews for this change:
    
    Bug:    https://bugs.openjdk.java.net/browse/JDK-8207343
    Webrev: http://cr.openjdk.java.net/~lucy/webrevs/8207343.00/
    
    With this change, I try to get rid of the a-priory size guessing for vtable and itable stubs. Please refer to the bug description for all the details. I didn't want to duplicate that text.
    
    ARM and AARCH64 help requested!
    The edits in vtableStubs_aarch64.cpp and vtableStubs_arm.cpp are made blindfolded. I am neither an ARM expert nor do I have build or test hardware available. I would be very grateful if one of the ARM gurus could please fill in for me.
    
    Thank you!
    Lutz
    
    
    



More information about the hotspot-compiler-dev mailing list