Re: [PATCH 0/2] LoongArch: Add optimized functions.

[Richard Henderson <richard.henderson@linaro.org>]

On 9/28/22 09:42, Xi Ruoyao wrote:
>> There is nothing in string-maskoff.h that the compiler should not be able to produce
>> itself from the generic version.  Having a brief look, the compiler simply needs to be
>> improved to unify two current AND patterns (which is an existing bug) and add the
>> additional case for bstrins.d.
> 
> Added GCC LoongArch port maintainer into Cc:.
> 
> It's actually more complicated.  Without the inline assembly in
> repeat_bytes(), the compiler does not hoist the 4-instruction 64-bit
> immediate load sequence out of a loop for "some reason I don't know
> yet".

Oh that's interesting.  I suspect that adding a REG_EQUAL note for each (or perhaps just 
the last) insn emitted by loongarch_move_integer would fix that.

>> Similarly, there is nothing in string-fzi.h that should not be gotten from longlong.h;
>> your only changes are to use __builtin_clz, which longlong.h exports as count_trailing_zeros.
> 
> No, it does not.  By default longlong.h uses a table driven approach for
> count_{trailing,leading}_zeros.  I can add __loongarch__ (or
> __loongarch_lp64) into longlong.h though.

That would best, as that header is used many more places.

>  IIUC I need to submit the change to GCC, then Glibc merges longlong.h from GCC, right?

I believe that's correct.


r~