[PATCH v3 04/14] RISC-V: Adding XTheadBb ISA extension

[Christoph Muellner <christoph.muellner@vrull.eu>]

From: Christoph Müllner <christoph.muellner@vrull.eu>

This patch adds support for the XTheadBb ISA extension.
The patch uses the T-Head specific decoder and translation.

Co-developed-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Co-developed-by: LIU Zhiwei <zhiwei_liu@linux.alibaba.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Christoph Müllner <christoph.muellner@vrull.eu>
---
Changes in v2:
- Add ISA_EXT_DATA_ENTRY()
- Split XtheadB* extension into individual commits
- Make implementation compatible with RV32.
- Use single decoder for XThead extensions

 target/riscv/cpu.c                         |   2 +
 target/riscv/cpu.h                         |   1 +
 target/riscv/insn_trans/trans_xthead.c.inc | 124 +++++++++++++++++++++
 target/riscv/translate.c                   |   4 +-
 target/riscv/xthead.decode                 |  20 ++++
 5 files changed, 149 insertions(+), 2 deletions(-)

diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index 4b46130c5b..b995470dd6 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -110,6 +110,7 @@ static const struct isa_ext_data isa_edata_arr[] = {
     ISA_EXT_DATA_ENTRY(svnapot, true, PRIV_VERSION_1_12_0, ext_svnapot),
     ISA_EXT_DATA_ENTRY(svpbmt, true, PRIV_VERSION_1_12_0, ext_svpbmt),
     ISA_EXT_DATA_ENTRY(xtheadba, true, PRIV_VERSION_1_11_0, ext_xtheadba),
+    ISA_EXT_DATA_ENTRY(xtheadbb, true, PRIV_VERSION_1_11_0, ext_xtheadbb),
     ISA_EXT_DATA_ENTRY(xtheadcmo, true, PRIV_VERSION_1_11_0, ext_xtheadcmo),
     ISA_EXT_DATA_ENTRY(xtheadsync, true, PRIV_VERSION_1_11_0, ext_xtheadsync),
     ISA_EXT_DATA_ENTRY(xventanacondops, true, PRIV_VERSION_1_12_0, ext_XVentanaCondOps),
@@ -1075,6 +1076,7 @@ static Property riscv_cpu_extensions[] = {
 
     /* Vendor-specific custom extensions */
     DEFINE_PROP_BOOL("xtheadba", RISCVCPU, cfg.ext_xtheadba, false),
+    DEFINE_PROP_BOOL("xtheadbb", RISCVCPU, cfg.ext_xtheadbb, false),
     DEFINE_PROP_BOOL("xtheadcmo", RISCVCPU, cfg.ext_xtheadcmo, false),
     DEFINE_PROP_BOOL("xtheadsync", RISCVCPU, cfg.ext_xtheadsync, false),
     DEFINE_PROP_BOOL("xventanacondops", RISCVCPU, cfg.ext_XVentanaCondOps, false),
diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index d3191bf27b..ff92705010 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -474,6 +474,7 @@ struct RISCVCPUConfig {
 
     /* Vendor-specific custom extensions */
     bool ext_xtheadba;
+    bool ext_xtheadbb;
     bool ext_xtheadcmo;
     bool ext_xtheadsync;
     bool ext_XVentanaCondOps;
diff --git a/target/riscv/insn_trans/trans_xthead.c.inc b/target/riscv/insn_trans/trans_xthead.c.inc
index a7da156869..ea6cd6e305 100644
--- a/target/riscv/insn_trans/trans_xthead.c.inc
+++ b/target/riscv/insn_trans/trans_xthead.c.inc
@@ -22,6 +22,12 @@
     }                                            \
 } while (0)
 
+#define REQUIRE_XTHEADBB(ctx) do {               \
+    if (!ctx->cfg_ptr->ext_xtheadbb) {           \
+        return false;                            \
+    }                                            \
+} while (0)
+
 #define REQUIRE_XTHEADCMO(ctx) do {              \
     if (!ctx->cfg_ptr->ext_xtheadcmo) {          \
         return false;                            \
@@ -67,6 +73,124 @@ GEN_TRANS_TH_ADDSL(1)
 GEN_TRANS_TH_ADDSL(2)
 GEN_TRANS_TH_ADDSL(3)
 
+/* XTheadBb */
+
+/* th.srri is an alternate encoding for rori (from Zbb) */
+static bool trans_th_srri(DisasContext *ctx, arg_th_srri * a)
+{
+    REQUIRE_XTHEADBB(ctx);
+    return gen_shift_imm_fn_per_ol(ctx, a, EXT_NONE,
+                                   tcg_gen_rotri_tl, gen_roriw, NULL);
+}
+
+/* th.srriw is an alternate encoding for roriw (from Zbb) */
+static bool trans_th_srriw(DisasContext *ctx, arg_th_srriw *a)
+{
+    REQUIRE_XTHEADBB(ctx);
+    REQUIRE_64BIT(ctx);
+    ctx->ol = MXL_RV32;
+    return gen_shift_imm_fn(ctx, a, EXT_NONE, gen_roriw, NULL);
+}
+
+/* th.ext and th.extu perform signed/unsigned bitfield extraction */
+static bool gen_th_bfextract(DisasContext *ctx, arg_th_bfext *a,
+                             void (*f)(TCGv, TCGv, unsigned int, unsigned int))
+{
+    TCGv dest = dest_gpr(ctx, a->rd);
+    TCGv source = get_gpr(ctx, a->rs1, EXT_ZERO);
+
+    if (a->lsb <= a->msb) {
+        f(dest, source, a->lsb, a->msb - a->lsb + 1);
+        gen_set_gpr(ctx, a->rd, dest);
+    }
+    return true;
+}
+
+static bool trans_th_ext(DisasContext *ctx, arg_th_ext *a)
+{
+    REQUIRE_XTHEADBB(ctx);
+    return gen_th_bfextract(ctx, a, tcg_gen_sextract_tl);
+}
+
+static bool trans_th_extu(DisasContext *ctx, arg_th_extu *a)
+{
+    REQUIRE_XTHEADBB(ctx);
+    return gen_th_bfextract(ctx, a, tcg_gen_extract_tl);
+}
+
+/* th.ff0: find first zero (clz on an inverted input) */
+static bool gen_th_ff0(DisasContext *ctx, arg_th_ff0 *a, DisasExtend ext)
+{
+    TCGv dest = dest_gpr(ctx, a->rd);
+    TCGv src1 = get_gpr(ctx, a->rs1, ext);
+
+    int olen = get_olen(ctx);
+    TCGv t = tcg_temp_new();
+
+    tcg_gen_not_tl(t, src1);
+    if (olen != TARGET_LONG_BITS) {
+        if (olen == 32) {
+            gen_clzw(dest, t);
+        } else {
+            g_assert_not_reached();
+        }
+    } else {
+        gen_clz(dest, t);
+    }
+
+    tcg_temp_free(t);
+    gen_set_gpr(ctx, a->rd, dest);
+
+    return true;
+}
+
+static bool trans_th_ff0(DisasContext *ctx, arg_th_ff0 *a)
+{
+    REQUIRE_XTHEADBB(ctx);
+    return gen_th_ff0(ctx, a, EXT_NONE);
+}
+
+/* th.ff1 is an alternate encoding for clz (from Zbb) */
+static bool trans_th_ff1(DisasContext *ctx, arg_th_ff1 *a)
+{
+    REQUIRE_XTHEADBB(ctx);
+    return gen_unary_per_ol(ctx, a, EXT_NONE, gen_clz, gen_clzw);
+}
+
+static void gen_th_revw(TCGv ret, TCGv arg1)
+{
+    tcg_gen_bswap32_tl(ret, arg1, TCG_BSWAP_OS);
+}
+
+/* th.rev is an alternate encoding for the RV64 rev8 (from Zbb) */
+static bool trans_th_rev(DisasContext *ctx, arg_th_rev *a)
+{
+    REQUIRE_XTHEADBB(ctx);
+
+    return gen_unary_per_ol(ctx, a, EXT_NONE, tcg_gen_bswap_tl, gen_th_revw);
+}
+
+/* th.revw is a sign-extended byte-swap of the lower word */
+static bool trans_th_revw(DisasContext *ctx, arg_th_revw *a)
+{
+    REQUIRE_XTHEADBB(ctx);
+    REQUIRE_64BIT(ctx);
+    return gen_unary(ctx, a, EXT_NONE, gen_th_revw);
+}
+
+/* th.tstnbz is equivalent to an orc.b (from Zbb) with inverted result */
+static void gen_th_tstnbz(TCGv ret, TCGv source1)
+{
+    gen_orc_b(ret, source1);
+    tcg_gen_not_tl(ret, ret);
+}
+
+static bool trans_th_tstnbz(DisasContext *ctx, arg_th_tstnbz *a)
+{
+    REQUIRE_XTHEADBB(ctx);
+    return gen_unary(ctx, a, EXT_ZERO, gen_th_tstnbz);
+}
+
 /* XTheadCmo */
 
 static inline int priv_level(DisasContext *ctx)
diff --git a/target/riscv/translate.c b/target/riscv/translate.c
index 68baf84807..3bae961be0 100644
--- a/target/riscv/translate.c
+++ b/target/riscv/translate.c
@@ -130,8 +130,8 @@ static bool always_true_p(DisasContext *ctx  __attribute__((__unused__)))
 
 static bool has_xthead_p(DisasContext *ctx  __attribute__((__unused__)))
 {
-    return ctx->cfg_ptr->ext_xtheadba || ctx->cfg_ptr->ext_xtheadcmo ||
-           ctx->cfg_ptr->ext_xtheadsync;
+    return ctx->cfg_ptr->ext_xtheadba || ctx->cfg_ptr->ext_xtheadbb ||
+           ctx->cfg_ptr->ext_xtheadcmo || ctx->cfg_ptr->ext_xtheadsync;
 }
 
 #define MATERIALISE_EXT_PREDICATE(ext)  \
diff --git a/target/riscv/xthead.decode b/target/riscv/xthead.decode
index b149f13018..8cd140891b 100644
--- a/target/riscv/xthead.decode
+++ b/target/riscv/xthead.decode
@@ -13,14 +13,23 @@
 %rd        7:5
 %rs1       15:5
 %rs2       20:5
+%sh5       20:5
+%sh6       20:6
 
 # Argument sets
 &r         rd rs1 rs2                               !extern
+&r2        rd rs1                                   !extern
+&shift     shamt rs1 rd                             !extern
+&th_bfext  msb lsb rs1 rd
 
 # Formats
 @sfence_vm  ....... ..... .....   ... ..... ....... %rs1
 @rs2_s      ....... ..... ..... ... ..... .......   %rs2 %rs1
 @r          ....... ..... .....  ... ..... .......  &r %rs2 %rs1 %rd
+@r2         ....... ..... .....  ... ..... .......  &r2 %rs1 %rd
+@th_bfext   msb:6  lsb:6  .....  ... ..... .......  &th_bfext %rs1 %rd
+@sh5        ....... ..... .....  ... ..... .......  &shift  shamt=%sh5      %rs1 %rd
+@sh6        ...... ...... .....  ... ..... .......  &shift shamt=%sh6 %rs1 %rd
 
 # XTheadBa
 # Instead of defining a new encoding, we simply use the decoder to
@@ -38,6 +47,17 @@ th_addsl1        0000001 ..... ..... 001 ..... 0001011 @r
 th_addsl2        0000010 ..... ..... 001 ..... 0001011 @r
 th_addsl3        0000011 ..... ..... 001 ..... 0001011 @r
 
+# XTheadBb
+th_ext           ...... ...... ..... 010 ..... 0001011 @th_bfext
+th_extu          ...... ...... ..... 011 ..... 0001011 @th_bfext
+th_ff0           1000010 00000 ..... 001 ..... 0001011 @r2
+th_ff1           1000011 00000 ..... 001 ..... 0001011 @r2
+th_srri          000100 ...... ..... 001 ..... 0001011 @sh6
+th_srriw         0001010 ..... ..... 001 ..... 0001011 @sh5
+th_rev           1000001 00000 ..... 001 ..... 0001011 @r2
+th_revw          1001000 00000 ..... 001 ..... 0001011 @r2
+th_tstnbz        1000000 00000 ..... 001 ..... 0001011 @r2
+
 # XTheadCmo
 th_dcache_call   0000000 00001 00000 000 00000 0001011
 th_dcache_ciall  0000000 00011 00000 000 00000 0001011
-- 
2.39.0