Re: [Qemu-devel] [PATCH] tci: Add implementation for INDEX_op_ld16u_i64

[Richard Henderson]

On 4/10/19 9:48 AM, Stefan Weil wrote:
> @@ -127,6 +127,12 @@ static void tci_write_reg8(tcg_target_ulong *regs, 
> TCGReg index, uint8_t value)
>      tci_write_reg(regs, index, value);
>  }
>  
> +static void
> +tci_write_reg16(tcg_target_ulong *regs, TCGReg index, uint16_t value)
> +{
> +    tci_write_reg(regs, index, value);
> +}
> +
>  static void
>  tci_write_reg32(tcg_target_ulong *regs, TCGReg index, uint32_t value)
>  {
> @@ -854,7 +862,14 @@ uintptr_t tcg_qemu_tb_exec(CPUArchState *env, uint8_t 
> *tb_ptr)
>              tci_write_reg8(regs, t0, *(uint8_t *)(t1 + t2));
>              break;
>          case INDEX_op_ld8s_i64:
> +            TODO();
> +            break;
>          case INDEX_op_ld16u_i64:
> +            t0 = *tb_ptr++;
> +            t1 = tci_read_r(regs, &tb_ptr);
> +            t2 = tci_read_s32(&tb_ptr);
> +            tci_write_reg16(regs, t0, *(uint16_t *)(t1 + t2));
> +            break;

Looks ok, I guess, although the introduction of tci_write_reg16 seems redundant
with the uint16_t value that is loaded.

Why not use tci_write_reg64, since that is the size of the register you're
modifying?  Let the zero-extension explicit in the name of the opcode to be
reflected in the zero-extension implied by the passing of a uint16_t value to a
uint64_t argument.


r~