Re: [PATCH 3/5] target/tricore: Fix RRPW_DEXTR

[Richard Henderson]

On 1/27/23 02:03, Bastian Koppelmann wrote:
> if we used const16 == 0 we would crash qemu with the error:
>
> ../tcg/tcg-op.c:196: tcg_gen_shri_i32: Assertion `arg2 >= 0 && arg2 < 32' failed
>
> This is a special case anyways as we can directly return cpu_gpr_d[r1]
> as this is the most significant word an nothing is shifted.
>
> Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
> ---
>   target/tricore/translate.c | 2 ++
>   1 file changed, 2 insertions(+)
>
> diff --git a/target/tricore/translate.c b/target/tricore/translate.c
> index 6149d4f5c0..62128c6aae 100644
> --- a/target/tricore/translate.c
> +++ b/target/tricore/translate.c
> @@ -8708,6 +8708,8 @@ static void decode_32Bit_opc(DisasContext *ctx)
>           const16 = MASK_OP_RRPW_POS(ctx->opcode);
>           if (r1 == r2) {
>               tcg_gen_rotli_tl(cpu_gpr_d[r3], cpu_gpr_d[r1], const16);
> +        } else if (const16 == 0) {
> +            tcg_gen_mov_tl(cpu_gpr_d[r3], cpu_gpr_d[r1]);
>           } else {
>               temp = tcg_temp_new();
>               tcg_gen_shli_tl(temp, cpu_gpr_d[r1], const16);

While correct, this entire operation is

   tcg_gen_extract2_tl(cpu_gpr_d[r3], cpu_gpr_d[r2], cpu_gpr_d[r1], 32 - 
const16);

which will take care of your two special cases as well.


r~