Re: [Qemu-devel] [PATCH 2/2] target-sh4: implement ftrv instruction

[Aurelien Jarno]

On Fri, Nov 21, 2008 at 07:55:42PM +0000, Mans Rullgard wrote:
> This partially implements the ftrv instruction.  It gives correct
> results under normal conditions.  FPU exceptions are not correctly
> generated.
> 
> Signed-off-by: Mans Rullgard <address@hidden>
> ---
>  target-sh4/translate.c |   40 ++++++++++++++++++++++++++++++++++++++++
>  1 files changed, 40 insertions(+), 0 deletions(-)
> 
> diff --git a/target-sh4/translate.c b/target-sh4/translate.c
> index 2d6dfe6..9c2da41 100644
> --- a/target-sh4/translate.c
> +++ b/target-sh4/translate.c
> @@ -1694,6 +1694,46 @@ static void _decode_opc(DisasContext * ctx)
>           tcg_temp_free_i64(fp);
>       }
>       return;
> +    case 0xf0fd: /* ftrv XMTRX,FVn */
> +        if ((ctx->opcode & 0x300) != 0x100)
> +            break;
> +        if (!(ctx->fpscr & FPSCR_PR)) {
> +            TCGv fv[4];
> +            TCGv fp;
> +            int n = B11_8 & 0xc;
> +            int i;
> +
> +            /*
> +             * NOTE: The values calculated here are likely to differ
> +             * from hardware since the hardware sacrifices accuracy
> +             * for speed in this instruction.
> +             *
> +             * FIXME: FPU exceptions are not correctly raised here.
> +             */
> +
> +            fp = tcg_temp_new();
> +
> +            for (i = 0; i < 4; i++) {
> +                fv[i] = tcg_temp_new();
> +                tcg_gen_mov_i32(fv[i], cpu_fregs[n+i]);
> +            }
> +
> +            for (i = 0; i < 4; i++) {
> +                gen_helper_fmul_FT(cpu_fregs[n+i], cpu_fregs[16+i], fv[0]);
> +                gen_helper_fmul_FT(fp,             cpu_fregs[20+i], fv[1]);
> +                gen_helper_fadd_FT(cpu_fregs[n+i], cpu_fregs[n+i],  fp);
> +                gen_helper_fmul_FT(fp,             cpu_fregs[24+i], fv[2]);
> +                gen_helper_fadd_FT(cpu_fregs[n+i], cpu_fregs[n+i],  fp);
> +                gen_helper_fmul_FT(fp,             cpu_fregs[28+i], fv[3]);
> +                gen_helper_fadd_FT(cpu_fregs[n+i], cpu_fregs[n+i],  fp);
> +            }
> +
> +            for (i = 0; i < 4; i++)
> +                tcg_temp_free(fv[i]);
> +            tcg_temp_free(fp);
> +            return;
> +        }
> +        break;
>      }
>  
>      fprintf(stderr, "unknown instruction 0x%04x at pc 0x%08x\n",

While this code probably produce the correct result, I am not sure it is
the best way to do it. It calls 28 times an helper, and uses 5 temp
variable. IMHO the ftrv should be implemented in an helper, called from
translate.c.

-- 
Aurelien Jarno                          GPG: 1024D/F1BCDB73
address@hidden                 http://www.aurel32.net