On Fri, 2023-06-02 at 23:04 +0800, Weiwei Li wrote:
> 
> 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 <rbradf...@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.

Hi, thank you for sharing this! Sorry for the delay in response. I had
a chance to test your version and found a small issue with the .Q
implementation when I tested it.

Here is the fix:

diff --git a/target/riscv/insn_trans/trans_rvzacas.c.inc
b/target/riscv/insn_trans/trans_rvzacas.c.inc
index c063079c6a..e556f48da8 100644
--- a/target/riscv/insn_trans/trans_rvzacas.c.inc
+++ b/target/riscv/insn_trans/trans_rvzacas.c.inc
@@ -129,11 +129,12 @@ static bool trans_amocas_q(DisasContext *ctx,
arg_amocas_q *a)
     TCGv_i128 src2 = tcg_temp_new_i128();
     TCGv_i64 src2l = get_gpr(ctx, a->rs2, EXT_NONE);
     TCGv_i64 src2h = get_gpr(ctx, a->rs2 == 0 ? 0 : a->rs2 + 1,
EXT_NONE);
-    TCGv_i64 destl = get_gpr(ctx, a->rs2, EXT_NONE);
-    TCGv_i64 desth = get_gpr(ctx, a->rs2 == 0 ? 0 : a->rs2 + 1,
EXT_NONE);
+    TCGv_i64 destl = get_gpr(ctx, a->rd, EXT_NONE);
+    TCGv_i64 desth = get_gpr(ctx, a->rd == 0 ? 0 : a->rd + 1,
EXT_NONE);
     
     tcg_gen_concat_i64_i128(src2, src2l, src2h);
     tcg_gen_concat_i64_i128(dest, destl, desth);
+    decode_save_opc(ctx);
     tcg_gen_atomic_cmpxchg_i128(dest, src1, dest, src2, ctx->mem_idx,
                                 (MO_ALIGN | MO_TEUO));

Since you already have an implementation ready i'm happy for your
version to be the one included.

> 
> 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.

Thank you - I can see from your version how this would work.

> 
> > +
> > +#if TARGET_LONG_BITS == 32
> We commonly use TARGET_RISCV32 instead of TARGET_LONG_BITS in similar
> cases.

Noted - thank you. Richard in another thread also pointed out my
version didn't handle dynamic xlen correctly.


> > +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.

Noted!

> > +    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.

Thanks!

> > +    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
> 

Thank you for the review - this was my first TCG patch and i've learnt
a lot from that.

Cheers,

Rob

> > +#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"
> 


Reply via email to