Re: [PATCH 2/3] newlib: riscv: Optimize memchr() and memrchr()

On 5/7/25 8:19 PM, Kito Cheng wrote:
> Hi Eric:
> 
> Here is a local GCC patch for zilsd code gen, could you try that?
> I believe this should just fix that on the compiler side...
> let me know if further issue you found or you don't know how to build gcc
> Plan to push this to upstream GCC later if there is no further issue :)
> 
> https://github.com/kito-cheng/gcc/tree/kitoc/zilsd-codegen

I had a hell of a time getting the cross compiler built, but it's at
least partially working. The assembler doesn't recognize Zilsd or Zclsd.
I think it still might be the binutils version (I had to build that
beforehand for the compiler).

In any case, here's the results I get. Without Zilsd:

$ riscv32-unknown-elf-gcc -march=rv32ic -mabi=ilp32 -O3 -S test.c
$ cat test.s
(output truncated)
main:
        addi    sp,sp,-32
        li      a4,5
        li      a5,7
        lui     a0,%hi(.LC0)
        addi    a1,sp,8
        mv      a2,a4
        mv      a3,a5
        addi    a0,a0,%lo(.LC0)
        sw      ra,28(sp)
        sw      a4,8(sp)
        sw      a5,12(sp)
        call    printf
        lw      ra,28(sp)
        li      a0,0
        addi    sp,sp,32
        jr      ra
        .size   main, .-main
        .ident  "GCC: (GNU) 16.0.0 20250430 (experimental)"
        .section        .note.GNU-stack,"",@progbits

With Zilsd:

$ riscv32-unknown-elf-gcc -march=rv32ic_zclsd_zilsd -mabi=ilp32 -O3 -S test.c
$ cat test.s
(output truncated)
main:
        addi    sp,sp,-32
        li      a4,5
        li      a5,7
        lui     a0,%hi(.LC0)
        addi    a1,sp,8
        mv      a2,a4
        mv      a3,a5
        addi    a0,a0,%lo(.LC0)
        sw      ra,28(sp)
        sd      a4,8(sp)
        call    printf
        lw      ra,28(sp)
        li      a0,0
        addi    sp,sp,32
        jr      ra
        .size   main, .-main
        .ident  "GCC: (GNU) 16.0.0 20250430 (experimental)"
        .section        .note.GNU-stack,"",@progbits

So GCC is at least generating the correct assembly. If Clang hasn't been
updated since I last built it, maybe you could convince them to somehow
incorporate your changes?