From: Weiwei Li <liweiwei@iscas.ac.cn>
To: Rob Bradford <rbradford@rivosinc.com>, qemu-devel@nongnu.org
Cc: qemu-riscv@nongnu.org, Palmer Dabbelt <palmer@dabbelt.com>,
Alistair Francis <alistair.francis@wdc.com>,
Bin Meng <bin.meng@windriver.com>,
Weiwei Li <liweiwei@iscas.ac.cn>,
Daniel Henrique Barboza <dbarboza@ventanamicro.com>,
Liu Zhiwei <zhiwei_liu@linux.alibaba.com>
Subject: Re: [PATCH 1/2] target/riscv: Add Zacas ISA extension support
Date: Fri, 2 Jun 2023 23:04:30 +0800 [thread overview]
Message-ID: <6e6e4cdb-fb24-286c-f493-0cce73f38615@iscas.ac.cn> (raw)
In-Reply-To: <20230602121638.36342-2-rbradford@rivosinc.com>
On 2023/6/2 20:16, Rob Bradford wrote:
> This commit adds support for the the amocas.{w,d,q} instructions behind
> a new property to enable that instruction.
>
> Signed-off-by: Rob Bradford <rbradford@rivosinc.com>
> ---
I also implemented an initial version for this extension without any
tests a few days ago.
You can find it in https://github.com/plctlab/plct-qemu/tree/plct-zacas-dev.
Hope it can help you.
> target/riscv/cpu.c | 4 +
> target/riscv/cpu.h | 1 +
> target/riscv/insn32.decode | 5 +
> target/riscv/insn_trans/trans_rvzacas.c.inc | 146 ++++++++++++++++++++
> target/riscv/translate.c | 1 +
> 5 files changed, 157 insertions(+)
> create mode 100644 target/riscv/insn_trans/trans_rvzacas.c.inc
> diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
> index db0875fb43..e99833eb4a 100644
> --- a/target/riscv/cpu.c
> +++ b/target/riscv/cpu.c
> @@ -82,6 +82,7 @@ static const struct isa_ext_data isa_edata_arr[] = {
> ISA_EXT_DATA_ENTRY(zicsr, PRIV_VERSION_1_10_0, ext_icsr),
> ISA_EXT_DATA_ENTRY(zifencei, PRIV_VERSION_1_10_0, ext_ifencei),
> ISA_EXT_DATA_ENTRY(zihintpause, PRIV_VERSION_1_10_0, ext_zihintpause),
> + ISA_EXT_DATA_ENTRY(zacas, PRIV_VERSION_1_12_0, ext_zacas),
> ISA_EXT_DATA_ENTRY(zawrs, PRIV_VERSION_1_12_0, ext_zawrs),
> ISA_EXT_DATA_ENTRY(zfh, PRIV_VERSION_1_11_0, ext_zfh),
> ISA_EXT_DATA_ENTRY(zfhmin, PRIV_VERSION_1_11_0, ext_zfhmin),
> @@ -1604,6 +1605,9 @@ static Property riscv_cpu_extensions[] = {
> DEFINE_PROP_BOOL("x-zvfh", RISCVCPU, cfg.ext_zvfh, false),
> DEFINE_PROP_BOOL("x-zvfhmin", RISCVCPU, cfg.ext_zvfhmin, false),
>
> + /* Atomic CAS (Zacas) */
> + DEFINE_PROP_BOOL("x-zacas", RISCVCPU, cfg.ext_zacas, false),
> +
> DEFINE_PROP_END_OF_LIST(),
> };
>
> diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
> index de7e43126a..ac4d9e8e9c 100644
> --- a/target/riscv/cpu.h
> +++ b/target/riscv/cpu.h
> @@ -436,6 +436,7 @@ struct RISCVCPUConfig {
> bool ext_smaia;
> bool ext_ssaia;
> bool ext_sscofpmf;
> + bool ext_zacas;
> bool rvv_ta_all_1s;
> bool rvv_ma_all_1s;
>
> diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode
> index 73d5d1b045..97d17ee520 100644
> --- a/target/riscv/insn32.decode
> +++ b/target/riscv/insn32.decode
> @@ -255,6 +255,11 @@ amomax_d 10100 . . ..... ..... 011 ..... 0101111 @atom_st
> amominu_d 11000 . . ..... ..... 011 ..... 0101111 @atom_st
> amomaxu_d 11100 . . ..... ..... 011 ..... 0101111 @atom_st
>
> +# *** Zacas Extension
> +amocas_w 00101 . . ..... ..... 010 ..... 0101111 @atom_st
> +amocas_d 00101 . . ..... ..... 011 ..... 0101111 @atom_st
> +amocas_q 00101 . . ..... ..... 100 ..... 0101111 @atom_st
> +
> # *** RV32F Standard Extension ***
> flw ............ ..... 010 ..... 0000111 @i
> fsw ....... ..... ..... 010 ..... 0100111 @s
> diff --git a/target/riscv/insn_trans/trans_rvzacas.c.inc b/target/riscv/insn_trans/trans_rvzacas.c.inc
> new file mode 100644
> index 0000000000..3f1b58ee8a
> --- /dev/null
> +++ b/target/riscv/insn_trans/trans_rvzacas.c.inc
> @@ -0,0 +1,146 @@
> +/*
> + * RISC-V translation routines for Zacas extension
> + *
> + * Copyright (c) 2023 Rivos Inc
> + *
> + * This program is free software; you can redistribute it and/or modify it
> + * under the terms and conditions of the GNU General Public License,
> + * version 2 or later, as published by the Free Software Foundation.
> + *
> + * This program is distributed in the hope it will be useful, but WITHOUT
> + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
> + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
> + * more details.
> + *
> + * You should have received a copy of the GNU General Public License along with
> + * this program. If not, see <http://www.gnu.org/licenses/>.
> + */
> +
> +#define REQUIRE_ZACAS(ctx) do { \
> + if (!ctx->cfg_ptr->ext_zacas) { \
> + return false; \
> + } \
> +} while (0)
> +
> +static bool amocas_tl(DisasContext *ctx, arg_atomic *a, MemOp mop)
> +{
> + TCGv retv = dest_gpr(ctx, a->rd);
> + TCGv addr = get_address(ctx, a->rs1, 0);
> + TCGv newv = get_gpr(ctx, a->rs2, EXT_ZERO);
> + TCGv cmpv = get_gpr(ctx, a->rd, EXT_ZERO);
> +
> + decode_save_opc(ctx);
> + tcg_gen_atomic_cmpxchg_tl(retv, addr, cmpv, newv, ctx->mem_idx, mop);
> + gen_set_gpr(ctx, a->rd, retv);
> +
> + return true;
> +}
> +
> +#if TARGET_LONG_BITS == 32
> +static bool trans_amocas_w(DisasContext *ctx, arg_amocas_w *a)
> +{
> + REQUIRE_ZACAS(ctx);
> + return amocas_tl(ctx, a, MO_TESL | MO_ALIGN);
> +}
> +#else
> +static bool trans_amocas_w(DisasContext *ctx, arg_amocas_w *a)
> +{
> + REQUIRE_64BIT(ctx);
> + REQUIRE_ZACAS(ctx);
> + ctx->ol = MXL_RV32;
> + return amocas_tl(ctx, a, MO_TESL | MO_ALIGN);
> +}
> +#endif
This seems unnecessary to use different trans function for RV32 and RV64.
tcg_gen_atomic_cmpxchg_tl itself will do the actually cmp and xchg
operation based on size in memop
instead of the target long size.
> +
> +#if TARGET_LONG_BITS == 32
We commonly use TARGET_RISCV32 instead of TARGET_LONG_BITS in similar
cases.
> +static bool trans_amocas_d(DisasContext *ctx, arg_amocas_w *a)
> +{
> + TCGv_i64 retv, newv, cmpv;
> + TCGv_i32 cmpv_l, cmpv_h, newv_l, newv_h;
> + TCGv addr;
> +
> + REQUIRE_ZACAS(ctx);
> +
> + if (a->rd % 2 == 1 || a->rs2 % 2 == 1) {
> + return false;
> + }
> +
> + addr = get_address(ctx, a->rs1, 0);
> + cmpv_l = get_gpr(ctx, a->rd, 0);
It's better to use EXT_NONE instead of 0 in this and following cases.
> + cmpv_h = a->rd == 0 ? get_gpr(ctx, 0, 0) : get_gpr(ctx, a->rd + 1, 0);
We can use ctx->zero for get_gpr(ctx, 0, 0) directly here.
> + cmpv = tcg_temp_new_i64();
> + tcg_gen_concat_i32_i64(cmpv, cmpv_l, cmpv_h);
> + newv_l = get_gpr(ctx, a->rs2, 0);
> + newv_h = a->rs2 == 0 ? get_gpr(ctx, 0, 0) : get_gpr(ctx, a->rs2 + 1, 0);
> + newv = tcg_temp_new_i64();
> + tcg_gen_concat_i32_i64(newv, newv_l, newv_h);
> + retv = tcg_temp_new_i64();
> +
> + decode_save_opc(ctx);
> + tcg_gen_atomic_cmpxchg_i64(retv, addr, cmpv, newv, ctx->mem_idx,
> + MO_TESQ | MO_ALIGN);
> +
> + if (a->rd != 0) {
> + TCGv_i32 retv_l = tcg_temp_new_i32();
> + TCGv_i32 retv_h = tcg_temp_new_i32();
> + tcg_gen_extr_i64_i32(retv_l, retv_h, retv);
> + gen_set_gpr(ctx, a->rd, retv_l);
> + gen_set_gpr(ctx, a->rd + 1, retv_h);
> + }
> +
> + return true;
> +}
> +#else
> +static bool trans_amocas_d(DisasContext *ctx, arg_amocas_w *a)
> +{
> + REQUIRE_64BIT(ctx);
This seems not correct. amocas_d can work on RV32, so I think it can
also work on RV64 with xl = 32
> + REQUIRE_ZACAS(ctx);
> + return amocas_tl(ctx, a, MO_TESQ | MO_ALIGN);
> +}
> +#endif
> +
> +#if TARGET_LONG_BITS == 32
> +static bool trans_amocas_q(DisasContext *ctx, arg_amocas_w *a)
> +{
> + return false;
> +}
This case is unnecessary.
We can just add REQUIRE_64BIT(ctx) in following trans_amocas_q
> +#else
> +static bool trans_amocas_q(DisasContext *ctx, arg_amocas_w *a)
> +{
> + TCGv_i128 retv, newv, cmpv;
> + TCGv_i64 cmpv_l, cmpv_h, newv_l, newv_h;
> + TCGv addr;
> +
> + REQUIRE_64BIT(ctx);
> + REQUIRE_ZACAS(ctx);
> +
> + if (a->rd % 2 == 1 || a->rs2 % 2 == 1) {
> + return false;
> + }
> +
> + addr = get_address(ctx, a->rs1, 0);
> + cmpv_l = get_gpr(ctx, a->rd, 0);
> + cmpv_h = a->rd == 0 ? get_gpr(ctx, 0, 0) : get_gpr(ctx, a->rd + 1, 0);
> + cmpv = tcg_temp_new_i128();
> + tcg_gen_concat_i64_i128(cmpv, cmpv_l, cmpv_h);
> + newv_l = get_gpr(ctx, a->rs2, 0);
> + newv_h = a->rs2 == 0 ? get_gpr(ctx, 0, 0) : get_gpr(ctx, a->rs2 + 1, 0);
> + newv = tcg_temp_new_i128();
> + tcg_gen_concat_i64_i128(newv, newv_l, newv_h);
> + retv = tcg_temp_new_i128();
> +
> + decode_save_opc(ctx);
> + tcg_gen_atomic_cmpxchg_i128(retv, addr, cmpv, newv, ctx->mem_idx,
> + MO_TEUO | MO_ALIGN);
> +
> + if (a->rd != 0) {
> + TCGv_i64 retv_l = tcg_temp_new_i64();
> + TCGv_i64 retv_h = tcg_temp_new_i64();
> + tcg_gen_extr_i128_i64(retv_l, retv_h, retv);
> + gen_set_gpr(ctx, a->rd, retv_l);
> + gen_set_gpr(ctx, a->rd + 1, retv_h);
> + }
> +
> + return true;
> +}
> +#endif
> diff --git a/target/riscv/translate.c b/target/riscv/translate.c
> index 928da0d3f0..55438f5ebf 100644
> --- a/target/riscv/translate.c
> +++ b/target/riscv/translate.c
> @@ -1074,6 +1074,7 @@ static uint32_t opcode_at(DisasContextBase *dcbase, target_ulong pc)
> #include "insn_trans/trans_rvv.c.inc"
> #include "insn_trans/trans_rvb.c.inc"
> #include "insn_trans/trans_rvzicond.c.inc"
> +#include "insn_trans/trans_rvzacas.c.inc"
> #include "insn_trans/trans_rvzawrs.c.inc"
> #include "insn_trans/trans_rvzicbo.c.inc"
> #include "insn_trans/trans_rvzfh.c.inc"
next prev parent reply other threads:[~2023-06-02 15:05 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-06-02 12:16 RFC: target/riscv: Add support for Zacas extension Rob Bradford
2023-06-02 12:16 ` [PATCH 1/2] target/riscv: Add Zacas ISA extension support Rob Bradford
2023-06-02 15:04 ` Weiwei Li [this message]
2023-06-20 15:22 ` Rob Bradford
2023-06-02 15:08 ` Weiwei Li
2023-06-03 3:28 ` Richard Henderson
2023-06-02 12:16 ` [PATCH 2/2] disas/riscv: Add support for amocas.{w,d,q} instructions Rob Bradford
2023-06-02 12:44 ` [PATCH 2/2] disas/riscv: Add support for amocas.{w, d, q} instructions Rob Bradford
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=6e6e4cdb-fb24-286c-f493-0cce73f38615@iscas.ac.cn \
--to=liweiwei@iscas.ac.cn \
--cc=alistair.francis@wdc.com \
--cc=bin.meng@windriver.com \
--cc=dbarboza@ventanamicro.com \
--cc=palmer@dabbelt.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-riscv@nongnu.org \
--cc=rbradford@rivosinc.com \
--cc=zhiwei_liu@linux.alibaba.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).