Re: [PATCH] RISC-V: Don't unconditionally add m, a, f, d in arch-canonicalize

[Palmer Dabbelt <palmer@dabbelt.com>]

On Wed, 25 May 2022 07:00:11 PDT (-0700), gcc-patches@gcc.gnu.org wrote:
> Committed, Thanks for fixing my stupid bug :P

IMO this is a good candidate for a backport.

> On Wed, May 25, 2022 at 9:26 PM Simon Cook <simon.cook@embecosm.com> wrote:
>>
>> This solves an issue where rv32i, etc. are canonicalized to rv32imafd
>> since the g->i addition of 'm', 'a', 'f', 'd' is not actually gated by
>> whether the input was rv32g/rv64g.
>>
>> gcc/ChangeLog:
>>
>>         * config/riscv/arch-canonicalize: Only add mafd extension if
>>         base was rv32/rv64g.
>> ---
>>  gcc/config/riscv/arch-canonicalize | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/gcc/config/riscv/arch-canonicalize
>> b/gcc/config/riscv/arch-canonicalize
>> index 71b2232b29e..fd7651ac491 100755
>> --- a/gcc/config/riscv/arch-canonicalize
>> +++ b/gcc/config/riscv/arch-canonicalize
>> @@ -73,8 +73,8 @@ def arch_canonicalize(arch, isa_spec):
>>    std_exts = []
>>    if arch[:5] in ['rv32e', 'rv32i', 'rv32g', 'rv64i', 'rv64g']:
>>      new_arch = arch[:5].replace("g", "i")
>> -    std_exts = ['m', 'a', 'f', 'd']
>>      if arch[:5] in ['rv32g', 'rv64g']:
>> +      std_exts = ['m', 'a', 'f', 'd']
>>        if not is_isa_spec_2p2:
>>          extra_long_ext = ['zicsr', 'zifencei']
>>    else:
>> --
>> 2.32.1