[PATCH] x86: don't mistake ordinary immediates for SAE / rounding control

[H.J. Lu]

On Thu, Apr 14, 2022 at 7:12 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> The way SAE templates are constructed was always puzzling me (including
> the need for separate templates in the first place), and expressing the
> extzra attribute via Imm8 actually has a bad effect: Ordinary immediates
> would also be accepted, leading to an extra byte being added after the
> instruction (i.e. generating bad code). Before re-working this (in
> particular to accept proper Intel syntax there), fix the immediate issue
> by adding the so far missing check.
>
> --- a/gas/config/tc-i386.c
> +++ b/gas/config/tc-i386.c
> @@ -6285,6 +6285,11 @@ check_VecOperands (const insn_template *
>           return 1;
>         }
>      }
> +  else if (t->opcode_modifier.sae)
> +    {
> +       i.error = unsupported_syntax;
> +       return 1;
> +    }
>
>    /* Check the special Imm4 cases; must be the first operand.  */
>    if (t->cpu_flags.bitfield.cpuxop && t->operands == 5)
> --- a/gas/testsuite/gas/i386/inval-avx512f.l
> +++ b/gas/testsuite/gas/i386/inval-avx512f.l
> @@ -218,6 +218,8 @@
>  .*:313: Error: .*unsupported broadcast for `vcvtneps2bf16'
>  .*:316: Error: .*unsupported broadcast for `vcvtneps2bf16'
>  .*:319: Error: .*unsupported broadcast for `vcvtneps2bf16'
> +.*:321: Error: .*vaddps.*
> +.*:322: Error: .*vcmpss.*
>  GAS LISTING .*
>
>
> @@ -569,6 +571,7 @@ GAS LISTING .*
>  [      ]*318 \?\?\?\? 62F27E38[        ]+vcvtneps2bf16 \(%eax\)\{1to8\}, %xmm1
>  [      ]*318[  ]+7208
>  [      ]*319[  ]+vcvtneps2bf16 \(%eax\)\{1to16\}, %xmm1
> -[      ]*320[  ]+\?\?\?\? 8DB42600             \.p2align 4
> -[      ]*320[  ]+0000008D
> -[      ]*320[  ]+7600
> +[      ]*320[  ]*
> +[      ]*321[  ]+vaddps \$0xcc, %zmm0, %zmm0, %zmm0
> +[      ]*322[  ]+vcmpss \$0, \$0xcc, %xmm0, %xmm0, %k0
> +#pass
> --- a/gas/testsuite/gas/i386/inval-avx512f.s
> +++ b/gas/testsuite/gas/i386/inval-avx512f.s
> @@ -317,4 +317,8 @@ _start:
>         vcvtneps2bf16 (%eax){1to4}, %xmm1
>         vcvtneps2bf16 (%eax){1to8}, %xmm1
>         vcvtneps2bf16 (%eax){1to16}, %xmm1
> +
> +       vaddps $0xcc, %zmm0, %zmm0, %zmm0
> +       vcmpss $0, $0xcc, %xmm0, %xmm0, %k0
> +
>         .p2align 4
>

OK.

Thanks.

-- 
H.J.