Re: [PATCH 15/46] tcg/optimize: Use fold_masks_zsa in fold_extract

[Pierrick Bouvier]

On 12/10/24 07:23, Richard Henderson wrote:
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>   tcg/optimize.c | 9 ++++-----
>   1 file changed, 4 insertions(+), 5 deletions(-)
>
> diff --git a/tcg/optimize.c b/tcg/optimize.c
> index 5f3a824a07..955151f4cf 100644
> --- a/tcg/optimize.c
> +++ b/tcg/optimize.c
> @@ -1716,7 +1716,7 @@ static bool fold_eqv(OptContext *ctx, TCGOp *op)
>   
>   static bool fold_extract(OptContext *ctx, TCGOp *op)
>   {
> -    uint64_t z_mask_old, z_mask;
> +    uint64_t z_mask_old, z_mask, s_mask, a_mask = -1;
>       int pos = op->args[2];
>       int len = op->args[3];
>   
> @@ -1731,12 +1731,11 @@ static bool fold_extract(OptContext *ctx, TCGOp *op)
>       z_mask_old = arg_info(op->args[1])->z_mask;
>       z_mask = extract64(z_mask_old, pos, len);
>       if (pos == 0) {
> -        ctx->a_mask = z_mask_old ^ z_mask;
> +        a_mask = z_mask_old ^ z_mask;
>       }
> -    ctx->z_mask = z_mask;
> -    ctx->s_mask = smask_from_zmask(z_mask);
> +    s_mask = smask_from_zmask(z_mask);
>   
> -    return fold_masks(ctx, op);
> +    return fold_masks_zsa(ctx, op, z_mask, s_mask, a_mask);
>   }

In this case, we could have a fold_masks_za function.

Sometimes, we put special values for one flag, or sometimes we call 
smask_from_zmask, or we rely on the implicit behaviour of the 
fold_masks_* variant, so this makes it a bit hard to follow.

Either we should go fully explicit (only fold_masks, with the three 
masks, and every call site has to mention them), or specialized 
functions for every situation we need to support. But a mix in the 
middle feels a bit awkward.

>   
>   static bool fold_extract2(OptContext *ctx, TCGOp *op)

Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>