vm stop question
dmitrij.pochepko at bell-sw.com
Mon Dec 3 13:23:05 UTC 2018
I see. Would it make sense to add new diagnostic flag like:
"SaveRegistersOnStop", which can be enabled for debug purposes then,
since it's not critical?
On 30/11/2018 11:13 PM, dean.long at oracle.com wrote:
> Saving all the registers probably isn't critical, because stop() isn't
> expect to return, like warning(), but would help crash analysis.
> Saving the registers in a stub rather than inline would reduce waste
> in the code cache.
> On 11/30/18 8:22 AM, Dmitrij Pochepko wrote:
>> Hi all,
>> I look into MacroAssembler::stop method and see, that all platforms
>> except ppc are saving registers on stack. Is it a bug on ppc or
>> is there a reason to save registers on stack there?
>> This code is inlined in multiple places and wasting code cache. I'd
>> like to remove or reshape it to be less bulky on aarch64.
More information about the hotspot-compiler-dev