RISC-V: Adding T-Head multiply-accumulate instructions

This patch adds support for the T-Head MAC instructions.
The patch uses the T-Head specific decoder and translation.

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>
Message-Id: <20230131202013.2541053-8-christoph.muellner@vrull.eu>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

1 files changed, 83 insertions, 0 deletions

diff --git a/target/riscv/insn_trans/trans_xthead.c.inc b/target/riscv/insn_trans/trans_xthead.c.inc
index 089b51f468..31a4034927 100644
--- a/target/riscv/insn_trans/trans_xthead.c.inc
+++ b/target/riscv/insn_trans/trans_xthead.c.inc
@@ -46,6 +46,12 @@
     }                                            \
 } while (0)
 
+#define REQUIRE_XTHEADMAC(ctx) do {              \
+    if (!ctx->cfg_ptr->ext_xtheadmac) {          \
+        return false;                            \
+    }                                            \
+} while (0)
+
 #define REQUIRE_XTHEADSYNC(ctx) do {             \
     if (!ctx->cfg_ptr->ext_xtheadsync) {         \
         return false;                            \
@@ -299,6 +305,83 @@ static bool trans_th_mvnez(DisasContext *ctx, arg_th_mveqz *a)
     return gen_th_condmove(ctx, a, TCG_COND_NE);
 }
 
+/* XTheadMac */
+
+static bool gen_th_mac(DisasContext *ctx, arg_r *a,
+                       void (*accumulate_func)(TCGv, TCGv, TCGv),
+                       void (*extend_operand_func)(TCGv, TCGv))
+{
+    TCGv dest = dest_gpr(ctx, a->rd);
+    TCGv src0 = get_gpr(ctx, a->rd, EXT_NONE);
+    TCGv src1 = get_gpr(ctx, a->rs1, EXT_NONE);
+    TCGv src2 = get_gpr(ctx, a->rs2, EXT_NONE);
+    TCGv tmp = tcg_temp_new();
+
+    if (extend_operand_func) {
+        TCGv tmp2 = tcg_temp_new();
+        extend_operand_func(tmp, src1);
+        extend_operand_func(tmp2, src2);
+        tcg_gen_mul_tl(tmp, tmp, tmp2);
+        tcg_temp_free(tmp2);
+    } else {
+        tcg_gen_mul_tl(tmp, src1, src2);
+    }
+
+    accumulate_func(dest, src0, tmp);
+    gen_set_gpr(ctx, a->rd, dest);
+    tcg_temp_free(tmp);
+
+    return true;
+}
+
+/* th.mula: "rd = rd + rs1 * rs2" */
+static bool trans_th_mula(DisasContext *ctx, arg_th_mula *a)
+{
+    REQUIRE_XTHEADMAC(ctx);
+    return gen_th_mac(ctx, a, tcg_gen_add_tl, NULL);
+}
+
+/* th.mulah: "rd = sext.w(rd + sext.w(rs1[15:0]) * sext.w(rs2[15:0]))" */
+static bool trans_th_mulah(DisasContext *ctx, arg_th_mulah *a)
+{
+    REQUIRE_XTHEADMAC(ctx);
+    ctx->ol = MXL_RV32;
+    return gen_th_mac(ctx, a, tcg_gen_add_tl, tcg_gen_ext16s_tl);
+}
+
+/* th.mulaw: "rd = sext.w(rd + rs1 * rs2)" */
+static bool trans_th_mulaw(DisasContext *ctx, arg_th_mulaw *a)
+{
+    REQUIRE_XTHEADMAC(ctx);
+    REQUIRE_64BIT(ctx);
+    ctx->ol = MXL_RV32;
+    return gen_th_mac(ctx, a, tcg_gen_add_tl, NULL);
+}
+
+/* th.muls: "rd = rd - rs1 * rs2" */
+static bool trans_th_muls(DisasContext *ctx, arg_th_muls *a)
+{
+    REQUIRE_XTHEADMAC(ctx);
+    return gen_th_mac(ctx, a, tcg_gen_sub_tl, NULL);
+}
+
+/* th.mulsh: "rd = sext.w(rd - sext.w(rs1[15:0]) * sext.w(rs2[15:0]))" */
+static bool trans_th_mulsh(DisasContext *ctx, arg_th_mulsh *a)
+{
+    REQUIRE_XTHEADMAC(ctx);
+    ctx->ol = MXL_RV32;
+    return gen_th_mac(ctx, a, tcg_gen_sub_tl, tcg_gen_ext16s_tl);
+}
+
+/* th.mulsw: "rd = sext.w(rd - rs1 * rs2)" */
+static bool trans_th_mulsw(DisasContext *ctx, arg_th_mulsw *a)
+{
+    REQUIRE_XTHEADMAC(ctx);
+    REQUIRE_64BIT(ctx);
+    ctx->ol = MXL_RV32;
+    return gen_th_mac(ctx, a, tcg_gen_sub_tl, NULL);
+}
+
 /* XTheadSync */
 
 static bool trans_th_sfence_vmas(DisasContext *ctx, arg_th_sfence_vmas *a)