[aarch64-port-dev ] RFR (M) 8228400: Remove built-in AArch64 simulator

[Dmitry Samersoff]

Hello Aleksey,

macroAssembler_aarch64.cpp:1414 call_VM_leaf_base

1. Do I understand correctly that we no longer use number_of_arguments
parameter?

Should we remove it and version of call_VM_leaf on l. 1430

2. Do we still need to stp/ldp rscratch1?

The rest looks good for me.

-Dmitry


On 26.07.2019 9:56, Aleksey Shipilev wrote:
> On 7/25/19 2:22 PM, Andrew Dinn wrote:
>> 1) File cpustate_aarch64.hpp exists primarily to declare a class
>> CPUState. This was needed to save/restore AArch64 register state on exit
>> from/re-entry into the simulator. I don't think anything else ought to
>> be using class CPUState or any of the other types it defines.
>>
>> Was there any good reason not simply to delete this file? (if so perhaps
>> whatever is keeping that file alive needs to be relocated to a home that
>> corresponds to the x86 file layout).
> 
> Right on, removed cpustate_aarch64.hpp.
> 
>> 2) File decode_aarch64.hpp contains almost entirely redundant stuff. I
>> believe the only code that is referenced from another file is the suite
>> of various pickbit* functions and their underlying mask* functions, the
>> client being code in file immediate_aarch64.cpp. All the enums are
>> redundant. So, I think this needs fixing by removing everything but the
>> pickbit* and mask* fns. It would probably be better to move these to
>> file immediate_aarch64.hpp and delete file decode_aarch64.hpp.
> 
> Right. I moved required definitions to immediate_aarch64.cpp and removed decode_aarch64.hpp.
> 
> New webrev:
>   http://cr.openjdk.java.net/~shade/8228400/webrev.03/
> 
> Testing: aarch64 cross-build; (gonna test tier1 once aarch64 box is free)
>