Received-SPF: pass (google.com: domain of linux-kernel+bounces-177856-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Date: Mon, 13 May 2024 18:53:26 +0200
From: Andrew Jones <ajones@ventanamicro.com>
To: Xiao Wang <xiao.w.wang@intel.com>
Cc: paul.walmsley@sifive.com, palmer@dabbelt.com, aou@eecs.berkeley.edu, 
	luke.r.nels@gmail.com, xi.wang@gmail.com, bjorn@kernel.org, ast@kernel.org, 
	daniel@iogearbox.net, andrii@kernel.org, martin.lau@linux.dev, eddyz87@gmail.com, 
	song@kernel.org, yonghong.song@linux.dev, john.fastabend@gmail.com, 
	kpsingh@kernel.org, sdf@google.com, haoluo@google.com, jolsa@kernel.org, 
	linux-riscv@lists.infradead.org, linux-kernel@vger.kernel.org, bpf@vger.kernel.org, 
	pulehui@huawei.com, haicheng.li@intel.com, conor@kernel.org
Subject: Re: [PATCH v2] riscv, bpf: Optimize zextw insn with Zba extension
Message-ID: <20240513-5c6f04fb4a29963c63d09aa2@orel>
References: <20240511023436.3282285-1-xiao.w.wang@intel.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20240511023436.3282285-1-xiao.w.wang@intel.com>

On Sat, May 11, 2024 at 10:34:36AM GMT, Xiao Wang wrote:
> The Zba extension provides add.uw insn which can be used to implement
> zext.w with rs2 set as ZERO.
> 
> Signed-off-by: Xiao Wang <xiao.w.wang@intel.com>
> ---
> v2:
> * Add Zba description in the Kconfig. (Lehui)
> * Reword the Kconfig help message to make it clearer. (Conor)
> ---
>  arch/riscv/Kconfig       | 22 ++++++++++++++++++++++
>  arch/riscv/net/bpf_jit.h | 18 ++++++++++++++++++
>  2 files changed, 40 insertions(+)
> 
> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> index 6bec1bce6586..e262a8668b41 100644
> --- a/arch/riscv/Kconfig
> +++ b/arch/riscv/Kconfig
> @@ -586,6 +586,14 @@ config RISCV_ISA_V_PREEMPTIVE
>  	  preemption. Enabling this config will result in higher memory
>  	  consumption due to the allocation of per-task's kernel Vector context.
>  
> +config TOOLCHAIN_HAS_ZBA
> +	bool
> +	default y
> +	depends on !64BIT || $(cc-option,-mabi=lp64 -march=rv64ima_zba)
> +	depends on !32BIT || $(cc-option,-mabi=ilp32 -march=rv32ima_zba)
> +	depends on LLD_VERSION >= 150000 || LD_VERSION >= 23900
> +	depends on AS_HAS_OPTION_ARCH
> +
>  config TOOLCHAIN_HAS_ZBB
>  	bool
>  	default y
> @@ -601,6 +609,20 @@ config TOOLCHAIN_HAS_VECTOR_CRYPTO
>  	def_bool $(as-instr, .option arch$(comma) +v$(comma) +zvkb)
>  	depends on AS_HAS_OPTION_ARCH
>  
> +config RISCV_ISA_ZBA
> +	bool "Zba extension support for bit manipulation instructions"
> +	depends on TOOLCHAIN_HAS_ZBA

We handcraft the instruction, so why do we need toolchain support?

> +	depends on RISCV_ALTERNATIVE

Also, while riscv_has_extension_likely() will be accelerated with
RISCV_ALTERNATIVE, it's not required.

> +	default y
> +	help
> +	   Add support for enabling optimisations in the kernel when the Zba
> +	   extension is detected at boot.
> +
> +	   The Zba extension provides instructions to accelerate the generation
> +	   of addresses that index into arrays of basic data types.
> +
> +	   If you don't know what to do here, say Y.
> +
>  config RISCV_ISA_ZBB
>  	bool "Zbb extension support for bit manipulation instructions"
>  	depends on TOOLCHAIN_HAS_ZBB
> diff --git a/arch/riscv/net/bpf_jit.h b/arch/riscv/net/bpf_jit.h
> index f4b6b3b9edda..18a7885ba95e 100644
> --- a/arch/riscv/net/bpf_jit.h
> +++ b/arch/riscv/net/bpf_jit.h
> @@ -18,6 +18,11 @@ static inline bool rvc_enabled(void)
>  	return IS_ENABLED(CONFIG_RISCV_ISA_C);
>  }
>  
> +static inline bool rvzba_enabled(void)
> +{
> +	return IS_ENABLED(CONFIG_RISCV_ISA_ZBA) && riscv_has_extension_likely(RISCV_ISA_EXT_ZBA);
> +}
> +
>  static inline bool rvzbb_enabled(void)
>  {
>  	return IS_ENABLED(CONFIG_RISCV_ISA_ZBB) && riscv_has_extension_likely(RISCV_ISA_EXT_ZBB);
> @@ -937,6 +942,14 @@ static inline u16 rvc_sdsp(u32 imm9, u8 rs2)
>  	return rv_css_insn(0x7, imm, rs2, 0x2);
>  }
>  
> +/* RV64-only ZBA instructions. */
> +
> +static inline u32 rvzba_zextw(u8 rd, u8 rs1)
> +{
> +	/* add.uw rd, rs1, ZERO */
> +	return rv_r_insn(0x04, RV_REG_ZERO, rs1, 0, rd, 0x3b);
> +}
> +
>  #endif /* __riscv_xlen == 64 */
>  
>  /* Helper functions that emit RVC instructions when possible. */
> @@ -1159,6 +1172,11 @@ static inline void emit_zexth(u8 rd, u8 rs, struct rv_jit_context *ctx)
>  
>  static inline void emit_zextw(u8 rd, u8 rs, struct rv_jit_context *ctx)
>  {
> +	if (rvzba_enabled()) {
> +		emit(rvzba_zextw(rd, rs), ctx);
> +		return;
> +	}
> +
>  	emit_slli(rd, rs, 32, ctx);
>  	emit_srli(rd, rd, 32, ctx);
>  }
> -- 
> 2.25.1
>

Thanks,
drew