aarch64_opcode_table changes for BZ 19722

aarch64_opcode_table changes for BZ 19722

[Szabolcs Nagy]

Nick, I'm trying to understand your changes in

https://sourceware.org/bugzilla/show_bug.cgi?id=19722
https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=4bd13cde17a27c342b79b72bde9ef8e1b5373344

with the new verifier field some

 {a,b,c,d},

table entries were changed to

 {a,b,c,d,NULL},

others to

 c_INS(a,b,d),

the goal of those c_INS macros is not clear to me,
is it just to align the opcode table columns?

i'd like to fiddle with the opcode table too, so
let me know if these have some special semantics.

Re: aarch64_opcode_table changes for BZ 19722

[Nick Clifton]

Hi Szabolcs,

> with the new verifier field some
> 
>  {a,b,c,d},
> 
> table entries were changed to
> 
>  {a,b,c,d,NULL},
> 
> others to
> 
>  c_INS(a,b,d),
> 
> the goal of those c_INS macros is not clear to me,
> is it just to align the opcode table columns?

No - it was to make future changes to the table easier to do and
to make the verifier functions stand out.

My original idea was that I would use macros for every entry, 
since I thought that a lot of the fields we duplicated in every 
entry.  Of course this turned out to be incorrect and I soon
realised that I would need too many macros, and that the macros
would not make anything clearer.  So instead I decided to use
macros just for the common cases - where the fields are 
duplicated - and to use explicit initialisation for the 
"interesting" instructions.  This also meant that it would be
easy to find which instructions had verifiers and which did not.

Cheers
  Nick

Re: aarch64_opcode_table changes for BZ 19722

[Szabolcs Nagy]

On 29/04/16 12:28, Nick Clifton wrote:
>> with the new verifier field some
>>
>>  {a,b,c,d},
>>
>> table entries were changed to
>>
>>  {a,b,c,d,NULL},
>>
>> others to
>>
>>  c_INS(a,b,d),
>>
>> the goal of those c_INS macros is not clear to me,
>> is it just to align the opcode table columns?
> 
> No - it was to make future changes to the table easier to do and
> to make the verifier functions stand out.
> 

the opcodes/aarch64-gen tool now fails to build
because verify_ldpsw is undefined.

i guess aarch64-opc.o can be linked into the tool,
although it's a bit ugly since it's not used, just
referenced from the opcode table initializer..

Re: aarch64_opcode_table changes for BZ 19722

[Szabolcs Nagy]

On 29/04/16 16:03, Szabolcs Nagy wrote:
> the opcodes/aarch64-gen tool now fails to build
> because verify_ldpsw is undefined.
> 
> i guess aarch64-opc.o can be linked into the tool,

i spoke too soon, -opc depends on generated stuff.
i'll hack it around with dummy definitions.

> although it's a bit ugly since it's not used, just
> referenced from the opcode table initializer..
>