Re: [Qemu-devel] [PATCH 7/7] target/m68k: add fmovem

[Richard Henderson]

On 06/25/2017 12:21 PM, Laurent Vivier wrote:
> +uint32_t HELPER(fmovem)(CPUM68KState *env, uint32_t addr, uint32_t ext)
> +{
> +    uintptr_t ra = GETPC();
> +    int mode = (ext >> 11) & 0x3;
> +    int is_load = ((ext & 0x2000) == 0);
> +    uint16_t mask;
> +    int incr, i;
> +
> +    if (m68k_feature(env, M68K_FEATURE_FPU)) {
> +        incr = 12;
> +    } else {
> +        incr = 8;
> +    }
> +
> +    if ((mode & 0x1) == 1) {
> +        /* Dynamic register list */
> +        int reg = extract32(ext, 4, 3);
> +        mask = env->dregs[reg] & 0x00FF;
> +    } else {
> +        /* Static register list */
> +        mask = ext & 0x00FF;
> +    }
> +
> +    if (!is_load && (mode & 2) == 0) {

There's too much decoding being done in the helper.

I think it would be better to split this into 6 helpers:

  fmovem{d,x}_st_predec
  fmovem{d,x}_st_postdec
  fmovem{d,x}_ld_postdec

It should be easy to share code for these via inline functions.

Pass in the mask as an argument.  This will either be a constant loaded via 
tcg_const_i32 or the zero-extended contents of the register.


r~