qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH 6/6] target/mips: Convert Loongson [D]MULT[U].G opcodes to de


From: Richard Henderson
Subject: Re: [PATCH 6/6] target/mips: Convert Loongson [D]MULT[U].G opcodes to decodetree
Date: Thu, 21 Jan 2021 10:06:14 -1000
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Thunderbird/68.10.0

On 1/12/21 11:55 AM, Philippe Mathieu-Daudé wrote:
> Convert the following opcodes to decodetree:
> 
> - MULT.G - multiply 32-bit signed integers
> - MULTU.G - multiply 32-bit unsigned integers
> - DMULT.G - multiply 64-bit signed integers
> - DMULTU.G - multiply 64-bit unsigned integers
> 
> Now that all opcodes from the extension have been converted, we
> can remove completely gen_loongson_integer() and its 2 calls in
> decode_opc_special2_legacy() and decode_opc_special3_legacy().
> 
> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
> ---
>  target/mips/godson2.decode    |  5 ++
>  target/mips/loong-ext.decode  |  5 ++
>  target/mips/loong_translate.c | 58 ++++++++++++++++++++++
>  target/mips/translate.c       | 92 +----------------------------------
>  4 files changed, 70 insertions(+), 90 deletions(-)
> 
> diff --git a/target/mips/godson2.decode b/target/mips/godson2.decode
> index 805452fa975..cf12d9072ec 100644
> --- a/target/mips/godson2.decode
> +++ b/target/mips/godson2.decode
> @@ -13,6 +13,11 @@
>  
>  @rs_rt_rd       ...... rs:5  rt:5  rd:5  ..... ......   &muldiv
>  
> +MULT.G          011111 ..... ..... ..... 00000 011000   @rs_rt_rd
> +MULTU.G         011111 ..... ..... ..... 00000 011001   @rs_rt_rd
> +DMULT.G         011111 ..... ..... ..... 00000 011100   @rs_rt_rd
> +DMULTU.G        011111 ..... ..... ..... 00000 011101   @rs_rt_rd
> +
>  DIV.G           011111 ..... ..... ..... 00000 011010   @rs_rt_rd
>  DIVU.G          011111 ..... ..... ..... 00000 011011   @rs_rt_rd
>  DDIV.G          011111 ..... ..... ..... 00000 011110   @rs_rt_rd
> diff --git a/target/mips/loong-ext.decode b/target/mips/loong-ext.decode
> index b0715894ee1..2281afaad95 100644
> --- a/target/mips/loong-ext.decode
> +++ b/target/mips/loong-ext.decode
> @@ -14,6 +14,11 @@
>  
>  @rs_rt_rd       ...... rs:5  rt:5  rd:5  ..... ......   &muldiv
>  
> +MULT.G          011100 ..... ..... ..... 00000 010000   @rs_rt_rd
> +DMULT.G         011100 ..... ..... ..... 00000 010001   @rs_rt_rd
> +MULTU.G         011100 ..... ..... ..... 00000 010010   @rs_rt_rd
> +DMULTU.G        011100 ..... ..... ..... 00000 010011   @rs_rt_rd
> +
>  DIV.G           011100 ..... ..... ..... 00000 010100   @rs_rt_rd
>  DDIV.G          011100 ..... ..... ..... 00000 010101   @rs_rt_rd
>  DIVU.G          011100 ..... ..... ..... 00000 010110   @rs_rt_rd
> diff --git a/target/mips/loong_translate.c b/target/mips/loong_translate.c
> index 50609ce4178..2af94535921 100644
> --- a/target/mips/loong_translate.c
> +++ b/target/mips/loong_translate.c
> @@ -263,6 +263,64 @@ static bool trans_DMODU_G(DisasContext *s, arg_muldiv *a)
>      return gen_lext_MODU_G(s, a->rt, a->rs, a->rd, true);
>  }
>  
> +static bool gen_lext_MULT_G(DisasContext *s, int rd, int rs, int rt,
> +                            bool is_double, bool is_unsigned)
> +{
> +    TCGv t0, t1;
> +
> +    if (is_double) {
> +        if (TARGET_LONG_BITS != 64) {
> +            return false;
> +        }
> +        check_mips_64(s);
> +    }
> +
> +    if (rd == 0) {
> +        /* Treat as NOP. */
> +        return true;
> +    }
> +
> +    t0 = tcg_temp_new();
> +    t1 = tcg_temp_new();
> +
> +    gen_load_gpr(t0, rs);
> +    gen_load_gpr(t1, rt);
> +
> +    if (is_unsigned && !is_double) {
> +        tcg_gen_ext32u_tl(t0, t0);
> +        tcg_gen_ext32u_tl(t1, t1);
> +    }

While this is a faithful conversion of the existing code, these extensions make
no difference to the result.  They are redundant with

> +    tcg_gen_mul_tl(cpu_gpr[rd], t0, t1);
> +    if (!is_double) {
> +        tcg_gen_ext32s_tl(cpu_gpr[rd], cpu_gpr[rd]);

this one, which discards any bit that might have been set by the input bits
that are cleared.

There is no actual difference between MULT.G and MULTU.G, or DMULT.G and
DMULTU.G, because they don't record the most significant bits of the infinite
result in any way.

> +static bool trans_MULT_G(DisasContext *s, arg_muldiv *a)
> +{
> +    return gen_lext_MULT_G(s, a->rt, a->rs, a->rd, false, false);
> +}
> +
> +static bool trans_MULTU_G(DisasContext *s, arg_muldiv *a)
> +{
> +    return gen_lext_MULT_G(s, a->rt, a->rs, a->rd, false, true);
> +}
> +
> +static bool trans_DMULT_G(DisasContext *s, arg_muldiv *a)
> +{
> +    return gen_lext_MULT_G(s, a->rt, a->rs, a->rd, true, false);
> +}
> +
> +static bool trans_DMULTU_G(DisasContext *s, arg_muldiv *a)
> +{
> +    return gen_lext_MULT_G(s, a->rt, a->rs, a->rd, true, true);
> +}

So... if you want to clean this up afterward, or before is up to you.


r~



reply via email to

[Prev in Thread] Current Thread [Next in Thread]