[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <c33984f0-57dc-4833-a96d-164425fc7578@huaweicloud.com>
Date: Mon, 15 Jan 2024 20:26:43 +0800
From: Pu Lehui <pulehui@...weicloud.com>
To: Björn Töpel <bjorn@...nel.org>, bpf@...r.kernel.org,
linux-riscv@...ts.infradead.org, netdev@...r.kernel.org
Cc: Alexei Starovoitov <ast@...nel.org>,
Daniel Borkmann <daniel@...earbox.net>, Andrii Nakryiko <andrii@...nel.org>,
Martin KaFai Lau <martin.lau@...ux.dev>, Song Liu <song@...nel.org>,
Yonghong Song <yhs@...com>, John Fastabend <john.fastabend@...il.com>,
KP Singh <kpsingh@...nel.org>, Stanislav Fomichev <sdf@...gle.com>,
Hao Luo <haoluo@...gle.com>, Jiri Olsa <jolsa@...nel.org>,
Palmer Dabbelt <palmer@...belt.com>, Conor Dooley <conor@...nel.org>,
Luke Nelson <luke.r.nels@...il.com>, Pu Lehui <pulehui@...wei.com>
Subject: Re: [PATCH bpf-next v2 6/6] riscv, bpf: Optimize bswap insns with Zbb
support
On 2023/9/28 19:08, Björn Töpel wrote:
> Pu Lehui <pulehui@...weicloud.com> writes:
>
>> From: Pu Lehui <pulehui@...wei.com>
>>
>> Optimize bswap instructions by rev8 Zbb instruction conbined with srli
>> instruction. And Optimize 16-bit zero-extension with Zbb support.
>>
>> Signed-off-by: Pu Lehui <pulehui@...wei.com>
>> ---
>> arch/riscv/net/bpf_jit.h | 67 +++++++++++++++++++++++++++++++++
>> arch/riscv/net/bpf_jit_comp64.c | 50 +-----------------------
>> 2 files changed, 69 insertions(+), 48 deletions(-)
>>
>> diff --git a/arch/riscv/net/bpf_jit.h b/arch/riscv/net/bpf_jit.h
>> index 944bdd6e4..a04eed672 100644
>> --- a/arch/riscv/net/bpf_jit.h
>> +++ b/arch/riscv/net/bpf_jit.h
>> @@ -1135,12 +1135,79 @@ static inline void emit_sextw(u8 rd, u8 rs, struct rv_jit_context *ctx)
>> emit_addiw(rd, rs, 0, ctx);
>> }
>>
>> +static inline void emit_zexth(u8 rd, u8 rs, struct rv_jit_context *ctx)
>> +{
>> + if (rvzbb_enabled()) {
>> + emit(rvzbb_zexth(rd, rs), ctx);
>> + } else {
>> + emit_slli(rd, rs, 48, ctx);
>> + emit_srli(rd, rd, 48, ctx);
>> + }
>> +}
>> +
>
> Prefer early-exit.
>
>> static inline void emit_zextw(u8 rd, u8 rs, struct rv_jit_context *ctx)
>> {
>> emit_slli(rd, rs, 32, ctx);
>> emit_srli(rd, rd, 32, ctx);
>> }
>>
>> +static inline void emit_bswap(u8 rd, s32 imm, struct rv_jit_context *ctx)
>> +{
>> + if (rvzbb_enabled()) {
>> + int bits = 64 - imm;
>> +
>> + emit(rvzbb_rev8(rd, rd), ctx);
>> + if (bits)
>> + emit_srli(rd, rd, bits, ctx);
>> + } else {
>> + emit_li(RV_REG_T2, 0, ctx);
>> +
>> + emit_andi(RV_REG_T1, rd, 0xff, ctx);
>> + emit_add(RV_REG_T2, RV_REG_T2, RV_REG_T1, ctx);
>> + emit_slli(RV_REG_T2, RV_REG_T2, 8, ctx);
>> + emit_srli(rd, rd, 8, ctx);
>> + if (imm == 16)
>> + goto out_be;
>> +
>> + emit_andi(RV_REG_T1, rd, 0xff, ctx);
>> + emit_add(RV_REG_T2, RV_REG_T2, RV_REG_T1, ctx);
>> + emit_slli(RV_REG_T2, RV_REG_T2, 8, ctx);
>> + emit_srli(rd, rd, 8, ctx);
>> +
>> + emit_andi(RV_REG_T1, rd, 0xff, ctx);
>> + emit_add(RV_REG_T2, RV_REG_T2, RV_REG_T1, ctx);
>> + emit_slli(RV_REG_T2, RV_REG_T2, 8, ctx);
>> + emit_srli(rd, rd, 8, ctx);
>> + if (imm == 32)
>> + goto out_be;
>> +
>> + emit_andi(RV_REG_T1, rd, 0xff, ctx);
>> + emit_add(RV_REG_T2, RV_REG_T2, RV_REG_T1, ctx);
>> + emit_slli(RV_REG_T2, RV_REG_T2, 8, ctx);
>> + emit_srli(rd, rd, 8, ctx);
>> +
>> + emit_andi(RV_REG_T1, rd, 0xff, ctx);
>> + emit_add(RV_REG_T2, RV_REG_T2, RV_REG_T1, ctx);
>> + emit_slli(RV_REG_T2, RV_REG_T2, 8, ctx);
>> + emit_srli(rd, rd, 8, ctx);
>> +
>> + emit_andi(RV_REG_T1, rd, 0xff, ctx);
>> + emit_add(RV_REG_T2, RV_REG_T2, RV_REG_T1, ctx);
>> + emit_slli(RV_REG_T2, RV_REG_T2, 8, ctx);
>> + emit_srli(rd, rd, 8, ctx);
>> +
>> + emit_andi(RV_REG_T1, rd, 0xff, ctx);
>> + emit_add(RV_REG_T2, RV_REG_T2, RV_REG_T1, ctx);
>> + emit_slli(RV_REG_T2, RV_REG_T2, 8, ctx);
>> + emit_srli(rd, rd, 8, ctx);
>> +out_be:
>> + emit_andi(RV_REG_T1, rd, 0xff, ctx);
>> + emit_add(RV_REG_T2, RV_REG_T2, RV_REG_T1, ctx);
>> +
>> + emit_mv(rd, RV_REG_T2, ctx);
>> + }
>> +}
>
> Definitely early-exit for this one!
>
> This function really show-cases why ZBB is nice! ;-)
>
> I'll take the next rev of series for a test!
>
Okay, the relevant modifications will be presented in v3 and will be
sent soon.
>
> Björn
Powered by blists - more mailing lists