Re: [Qemu-devel] [PATCH 1/8] target/ppc: Optimize emulation of lvsl and lvsr instructions

From: Stefan Brankovic <stefan.brankovic@rt-rk.com>
To: Richard Henderson <richard.henderson@linaro.org>, qemu-devel@nongnu.org
Cc: david@gibson.dropbear.id.au
Subject: Re: [Qemu-devel] [PATCH 1/8] target/ppc: Optimize emulation of lvsl and lvsr instructions
Date: Mon, 17 Jun 2019 13:31:00 +0200	[thread overview]
Message-ID: <8e37914f-dfbf-c45f-f153-18e029858e50@rt-rk.com> (raw)
In-Reply-To: <0d30e79a-12c8-7fa5-237c-25bd00682d2b@linaro.org>

On 6.6.19. 18:46, Richard Henderson wrote:
> On 6/6/19 5:15 AM, Stefan Brankovic wrote:
>> +    tcg_gen_addi_i64(result, sh, 7);
>> +    for (i = 7; i >= 1; i--) {
>> +        tcg_gen_shli_i64(tmp, sh, i * 8);
>> +        tcg_gen_or_i64(result, result, tmp);
>> +        tcg_gen_addi_i64(sh, sh, 1);
>> +    }
> Better to replicate sh into the 8 positions and then use one add.
>
>      tcg_gen_muli_i64(sh, sh, 0x0101010101010101ull);
>      tcg_gen_addi_i64(hi_result, sh, 0x0001020304050607ull);
>      tcg_gen_addi_i64(lo_result, sh, 0x08090a0b0c0d0e0full);
>
> and
>
>      tcg_gen_subfi_i64(hi_result, 0x1011121314151617ull, sh);
>      tcg_gen_subfi_i64(lo_result, 0x18191a1b1c1d1e1full, sh);
>
> for lvsr.
>
I think you are right, this is definitely better way of implementing it. 
I will adopt your approach in v2.

Kind Regards,

Stefan

> r~