On 3/4/19 7:13 AM, Mateja Marjanovic wrote:
> + } else {
> + TCGv_i64 t0 = tcg_temp_new();
> + TCGv_i64 t1 = tcg_temp_new();
> + TCGv_i64 t2 = tcg_temp_new();
> + uint64_t mask0 = (1ULL << 32) - 1;
> + uint64_t mask1 = mask0 << 32;
> +
> + tcg_gen_movi_i64(t1, 0);
> + tcg_gen_andi_i64(t0, cpu_gpr[rt], mask1);
> + tcg_gen_or_i64(t1, t0, t1);
> + tcg_gen_andi_i64(t0, cpu_mmr[rt], mask0);
> + tcg_gen_shri_i64(t0, t0, 64);
This line should abort, since the shift count is out of range.
Are you testing with --enable-debug-tcg?
> + tcg_gen_or_i64(t1, t0, t1);
> + tcg_gen_mov_i64(t2, t1);
But again, this whole thing is two deposit operations:
tcg_gen_mov_i64(t0, cpu_gpr[rt]);
tcg_gen_deposit_i64(cpu_gpr[rd], cpu_gpr[rt], cpu_mmr[rt], 0, 32);
tcg_gen_deposit_i64(cpu_mmr[rd], cpu_mmr[rt], t0, 0, 32);
r~
