Re: [Qemu-devel] [PATCH v4 3/9] target/mips: Split mips instruction handling

[Aleksandar Markovic <amarkovic@wavecomp.com>]

Hi, Craig,

> From: Craig Janeczek <jancraig@amazon.com>
> Sent: Thursday, August 30, 2018 9:30 PM
> To: qemu-devel@nongnu.org
> Cc: Aleksandar Markovic; aurelien@aurel32.net; Craig Janeczek
> Subject: [PATCH v4 3/9] target/mips: Split mips instruction handling
> 
> Splits the instruction handling switch statement from the original
> legacy code.
> 
> Signed-off-by: Craig Janeczek <jancraig@amazon.com>
> ---
>  v1
>     - NA
>  v2
>     - NA
>  v3
>     - NA
>  v4
>     - Initial patch
> 
>  target/mips/mips-defs.h |  1 +
>  target/mips/translate.c | 28 +++++++++++++++++++++++++++-
>  2 files changed, 28 insertions(+), 1 deletion(-)
> 
> diff --git a/target/mips/mips-defs.h b/target/mips/mips-defs.h
> index d239069975..5a409757f0 100644
> --- a/target/mips/mips-defs.h
> +++ b/target/mips/mips-defs.h
> @@ -50,6 +50,7 @@
>  #define   ASE_SMARTMIPS 0x00400000
>  #define   ASE_MICROMIPS 0x00800000
>  #define   ASE_MSA       0x01000000
> +#define   ASE_MXU       0x02000000
> 
>  /* Chip specific instructions. */
>  #define                INSN_LOONGSON2E  0x20000000
> diff --git a/target/mips/translate.c b/target/mips/translate.c
> index a598f45558..53d896ebf9 100644
> --- a/target/mips/translate.c
> +++ b/target/mips/translate.c
> @@ -17855,6 +17855,28 @@ static void decode_opc_special(CPUMIPSState *env, DisasContext *ctx)
>      }
>  }
> 
> +static void decode_opc_special2_mxu(CPUMIPSState *env, DisasContext *ctx)
> +{
> +    int rs, rt, rd;
> +    uint32_t op1;
> +
> +    rs = (ctx->opcode >> 21) & 0x1f;
> +    rt = (ctx->opcode >> 16) & 0x1f;
> +    rd = (ctx->opcode >> 11) & 0x1f;
> +
> +    op1 = MASK_SPECIAL2(ctx->opcode);
> +
> +    switch (op1) {
> +    case OPC_MUL:
> +        gen_arith(ctx, op1, rd, rs, rt);
> +        break;
> +    default:            /* Invalid */
> +        MIPS_INVAL("special2_mxu");
> +        generate_exception_end(ctx, EXCP_RI);
> +        break;
> +    }
> +}
> +

This (case OPC_MUL) just looks very odd to me. Why would OPC_MUL somehow be supposed to be included here? Is there any documentation to support this? For example of other kind: OPC_MADD is not included in this switch, but there is an OPC_MADD equivalent in MXU. At the same time, there is an OPC_MUL equivalent in MXU too.

This looks to me as a very unclear opcode organization. Too bad the MXU documentation that you linked to doesn't have opcode specifications. Xburst base set documentation would be very helpful, but there is no such doc to my knowledge.

Sincerely,
Aleksandar

>  static void decode_opc_special2_legacy(CPUMIPSState *env, DisasContext *ctx)
>  {
>      int rs, rt, rd;
> @@ -19836,7 +19858,11 @@ static void decode_opc(CPUMIPSState *env, DisasContext *ctx)
>          decode_opc_special(env, ctx);
>          break;
>      case OPC_SPECIAL2:
> -        decode_opc_special2_legacy(env, ctx);
> +        if (ctx->insn_flags & ASE_MXU) {
> +            decode_opc_special2_mxu(env, ctx);
> +        } else {
> +            decode_opc_special2_legacy(env, ctx);
> +        }
>          break;
>      case OPC_SPECIAL3:
>          decode_opc_special3(env, ctx);
> --
> 2.18.0
>