Re: [PATCH v2 065/101] target/arm: Implement SME2 FCLAMP, SCLAMP, UCLAMP

["Alex Bennée" <alex.bennee@linaro.org>]

Richard Henderson <richard.henderson@linaro.org> writes:

> On 6/21/25 16:50, Richard Henderson wrote:
>> +static bool trans_FCLAMP(DisasContext *s, arg_zzz_en *a)
>> +{
>> +    static gen_helper_gvec_3_ptr * const fn[] = {
>> +        gen_helper_sme2_bfclamp,
>> +        gen_helper_sme2_fclamp_h,
>> +        gen_helper_sme2_fclamp_s,
>> +        gen_helper_sme2_fclamp_d,
>> +    };
>> +    TCGv_ptr fpst;
>> +    int vl;
>> +
>> +    /* This insn uses MO_8 to encode BFloat16. */
>> +    if (a->esz == MO_8
>> +        ? dc_isar_feature(aa64_sme2_b16b16, s)
>> +        : dc_isar_feature(aa64_sme2, s)) {
>
> Missing !'s.

It gets a lot further into the run but we then fail on:

  Program received signal SIGILL, Illegal instruction.
  0x00007ffff65d6690 in kai_kernel_imatmul_clamp_f32_f32p2vlx1_f32p2vlx1b_2vlx2vl_sme2_mopa ()
  (gdb) x/5i $pc
  => 0x7ffff65d6690 <kai_kernel_imatmul_clamp_f32_f32p2vlx1_f32p2vlx1b_2vlx2vl_sme2_mopa+496>:    fclamp  {z0.s-z3.s}, z26.s, z24.s
     0x7ffff65d6694 <kai_kernel_imatmul_clamp_f32_f32p2vlx1_f32p2vlx1b_2vlx2vl_sme2_mopa+500>:    fclamp  {z8.s-z11.s}, z26.s, z24.s
     0x7ffff65d6698 <kai_kernel_imatmul_clamp_f32_f32p2vlx1_f32p2vlx1b_2vlx2vl_sme2_mopa+504>:    st1w    {z0.s, z8.s}, pn8, [x26]
     0x7ffff65d669c <kai_kernel_imatmul_clamp_f32_f32p2vlx1_f32p2vlx1b_2vlx2vl_sme2_mopa+508>:    add     x26, x26, x23
     0x7ffff65d66a0 <kai_kernel_imatmul_clamp_f32_f32p2vlx1_f32p2vlx1b_2vlx2vl_sme2_mopa+512>:
      b.eq        0x7ffff65d66bc <kai_kernel_imatmul_clamp_f32_f32p2vlx1_f32p2vlx1b_2vlx2vl_sme2_mopa+540>  // b.none
  (gdb) x/1w $pc
  0x7ffff65d6690 <kai_kernel_imatmul_clamp_f32_f32p2vlx1_f32p2vlx1b_2vlx2vl_sme2_mopa+496>:       0xc1b8cb40
  (gdb) 

>
>
> r~

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro