diff mbox series

[v5,10/21] target/loongarch: Add floating point arithmetic instruction translation

Message ID 1631624431-30658-11-git-send-email-gaosong@loongson.cn (mailing list archive)
State New, archived
Headers show
Series Add LoongArch linux-user emulation support | expand

Commit Message

gaosong Sept. 14, 2021, 1 p.m. UTC
This patch implement floating point arithmetic instruction translation.

This includes:
- F{ADD/SUB/MUL/DIV}.{S/D}
- F{MADD/MSUB/NMADD/NMSUB}.{S/D}
- F{MAX/MIN}.{S/D}
- F{MAXA/MINA}.{S/D}
- F{ABS/NEG}.{S/D}
- F{SQRT/RECIP/RSQRT}.{S/D}
- F{SCALEB/LOGB/COPYSIGN}.{S/D}
- FCLASS.{S/D}

Signed-off-by: Song Gao <gaosong@loongson.cn>
Signed-off-by: XiaoJuan Yang <yangxiaojuan@loongson.cn>
---
 target/loongarch/cpu.c                     |   1 +
 target/loongarch/fpu_helper.c              | 365 +++++++++++++++++++++++++++++
 target/loongarch/helper.h                  |  33 +++
 target/loongarch/insn_trans/trans_farith.c | 131 +++++++++++
 target/loongarch/insns.decode              |  56 +++++
 target/loongarch/internals.h               |   2 +
 target/loongarch/translate.c               |  11 +
 7 files changed, 599 insertions(+)
 create mode 100644 target/loongarch/fpu_helper.c
 create mode 100644 target/loongarch/insn_trans/trans_farith.c

Comments

Richard Henderson Sept. 14, 2021, 2:15 p.m. UTC | #1
On 9/14/21 6:00 AM, Song Gao wrote:
> +static void update_fcsr0_mask(CPULoongArchState *env, uintptr_t pc, int mask)
> +{
> +    int flags = get_float_exception_flags(&env->fp_status) & (~mask);
> +
> +    if (!flags) {
> +        SET_FP_CAUSE(env->fcsr0, flags);
> +        return;

If mask != 0, we may unintentionally leave fp_status.flags != 0, which will then be 
incorporated into the next fp operation.

> +    }
> +
> +    flags = ieee_ex_to_loongarch(flags);
> +    set_float_exception_flags(0, &env->fp_status);

I think this set should move above the if.

> +static bool trans_fmax_s(DisasContext *ctx, arg_fmt_fdfjfk *a)
> +{
> +    tcg_gen_umax_i64(cpu_fpr[a->fd], cpu_fpr[a->fj], cpu_fpr[a->fk]);

Um, this is an integer operation.  You want float32_maxnum.
Likewise for the other min/max operations.

> +static bool trans_fabs_s(DisasContext *ctx, arg_fmt_fdfj *a)
> +{
> +    tcg_gen_ext32u_tl(cpu_fpr[a->fj], cpu_fpr[a->fj]);

You shouldn't be modifying fj.

> +    tcg_gen_abs_i64(cpu_fpr[a->fd], cpu_fpr[a->fj]);

This is an integer operation.  You want

     tcg_gen_andi_i64(cpu_fpr[a->fd], cpu_fpr[a->fj],
                      MAKE_64BIT_MASK(0, 31));

> +static bool trans_fabs_d(DisasContext *ctx, arg_fmt_fdfj *a)
> +{
> +    tcg_gen_abs_i64(cpu_fpr[a->fd], cpu_fpr[a->fj]);

Similarly.

> +/*
> + * LoongArch requires NaN-boxing of narrower width floating point values.
> + * This applies when a 32-bit value is assigned to a 64-bit FP register.
> + */

Um, this comment does not reflect the language in the v1.00 manual.

The actual language says that the upper 32 bits are undefined ("can be any value").  We 
should state that here, and then that QEMU chooses to nanbox, because it is most likely to 
show guest bugs early.


r~
diff mbox series

Patch

diff --git a/target/loongarch/cpu.c b/target/loongarch/cpu.c
index b89bf51..487df05 100644
--- a/target/loongarch/cpu.c
+++ b/target/loongarch/cpu.c
@@ -143,6 +143,7 @@  static void loongarch_cpu_reset(DeviceState *dev)
     env->fcsr0_mask = 0x1f1f031f;
     env->fcsr0 = 0x0;
 
+    restore_fp_status(env);
     cs->exception_index = EXCP_NONE;
 }
 
diff --git a/target/loongarch/fpu_helper.c b/target/loongarch/fpu_helper.c
new file mode 100644
index 0000000..e7d3454
--- /dev/null
+++ b/target/loongarch/fpu_helper.c
@@ -0,0 +1,365 @@ 
+/*
+ * LoongArch float point emulation helpers for qemu
+ *
+ * Copyright (c) 2021 Loongson Technology Corporation Limited
+ *
+ * SPDX-License-Identifier: LGPL-2.1+
+ */
+
+#include "qemu/osdep.h"
+#include "cpu.h"
+#include "exec/helper-proto.h"
+#include "exec/exec-all.h"
+#include "exec/cpu_ldst.h"
+#include "fpu/softfloat.h"
+#include "internals.h"
+
+#define FLOAT_TO_INT32_OVERFLOW 0x7fffffff
+#define FLOAT_TO_INT64_OVERFLOW 0x7fffffffffffffffULL
+
+static inline uint64_t nanbox_s(float32 fp)
+{
+    return fp | MAKE_64BIT_MASK(32, 32);
+}
+
+/* convert loongarch rounding mode in fcsr0 to IEEE library */
+static const FloatRoundMode ieee_rm[4] = {
+    float_round_nearest_even,
+    float_round_to_zero,
+    float_round_up,
+    float_round_down
+};
+
+void restore_fp_status(CPULoongArchState *env)
+{
+    set_float_rounding_mode(ieee_rm[(env->fcsr0 >> FCSR0_RM) & 0x3],
+                            &env->fp_status);
+    set_flush_to_zero(0, &env->fp_status);
+}
+
+static int ieee_ex_to_loongarch(int xcpt)
+{
+    int ret = 0;
+    if (xcpt & float_flag_invalid) {
+        ret |= FP_INVALID;
+    }
+    if (xcpt & float_flag_overflow) {
+        ret |= FP_OVERFLOW;
+    }
+    if (xcpt & float_flag_underflow) {
+        ret |= FP_UNDERFLOW;
+    }
+    if (xcpt & float_flag_divbyzero) {
+        ret |= FP_DIV0;
+    }
+    if (xcpt & float_flag_inexact) {
+        ret |= FP_INEXACT;
+    }
+    return ret;
+}
+
+static void update_fcsr0_mask(CPULoongArchState *env, uintptr_t pc, int mask)
+{
+    int flags = get_float_exception_flags(&env->fp_status) & (~mask);
+
+    if (!flags) {
+        SET_FP_CAUSE(env->fcsr0, flags);
+        return;
+    }
+
+    flags = ieee_ex_to_loongarch(flags);
+    set_float_exception_flags(0, &env->fp_status);
+    SET_FP_CAUSE(env->fcsr0, flags);
+
+    if (GET_FP_ENABLES(env->fcsr0) & flags) {
+        do_raise_exception(env, EXCP_FPE, pc);
+    } else {
+        UPDATE_FP_FLAGS(env->fcsr0, flags);
+    }
+}
+
+static void update_fcsr0(CPULoongArchState *env, uintptr_t pc)
+{
+    update_fcsr0_mask(env, pc, 0);
+}
+
+uint64_t helper_fadd_s(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = nanbox_s(float32_add((uint32_t)fj, (uint32_t)fk, &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fadd_d(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = float64_add(fj, fk, &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fsub_s(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = nanbox_s(float32_sub((uint32_t)fj, (uint32_t)fk, &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fsub_d(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = float64_sub(fj, fk, &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fmul_s(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = nanbox_s(float32_mul((uint32_t)fj, (uint32_t)fk, &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fmul_d(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = float64_mul(fj, fk, &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fdiv_s(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = nanbox_s(float32_div((uint32_t)fj, (uint32_t)fk, &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fdiv_d(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = float64_div(fj, fk, &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fmaxa_s(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = nanbox_s(float32_maxnummag((uint32_t)fj,
+                                    (uint32_t)fk, &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fmaxa_d(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = float64_maxnummag(fj, fk, &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fmina_s(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = nanbox_s(float32_minnummag((uint32_t)fj,
+                                    (uint32_t)fk, &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fmina_d(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+
+    fd = float64_minnummag(fj, fk, &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fscaleb_s(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+    int32_t n = (int32_t)fk;
+
+    fd = nanbox_s(float32_scalbn((uint32_t)fj,
+                                 n >  0x200 ?  0x200 :
+                                 n < -0x200 ? -0x200 : n,
+                                 &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fscaleb_d(CPULoongArchState *env, uint64_t fj, uint64_t fk)
+{
+    uint64_t fd;
+    int64_t n = (int64_t)fk;
+
+    fd = float64_scalbn(fj,
+                        n >  0x1000 ?  0x1000 :
+                        n < -0x1000 ? -0x1000 : n,
+                        &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fsqrt_s(CPULoongArchState *env, uint64_t fj)
+{
+    uint64_t fd;
+
+    fd = nanbox_s(float32_sqrt((uint32_t)fj, &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fsqrt_d(CPULoongArchState *env, uint64_t fj)
+{
+    uint64_t fd;
+
+    fd = float64_sqrt(fj, &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_frecip_s(CPULoongArchState *env, uint64_t fj)
+{
+    uint64_t fd;
+
+    fd = nanbox_s(float32_div(float32_one, (uint32_t)fj, &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_frecip_d(CPULoongArchState *env, uint64_t fj)
+{
+    uint64_t fd;
+
+    fd = float64_div(float64_one, fj, &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_frsqrt_s(CPULoongArchState *env, uint64_t fj)
+{
+    uint64_t fd;
+    uint32_t fp;
+
+    fp = float32_sqrt((uint32_t)fj, &env->fp_status);
+    fd = nanbox_s(float32_div(float32_one, fp, &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_frsqrt_d(CPULoongArchState *env, uint64_t fj)
+{
+    uint64_t fp, fd;
+
+    fp = float64_sqrt(fj, &env->fp_status);
+    fd = float64_div(float64_one, fp, &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_flogb_s(CPULoongArchState *env, uint64_t fj)
+{
+    uint64_t fd;
+    uint32_t fp;
+    float_status *status = &env->fp_status;
+    FloatRoundMode old_mode = get_float_rounding_mode(status);
+
+    set_float_rounding_mode(float_round_down, status);
+    fp = float32_log2((uint32_t)fj, status);
+    fd = nanbox_s(float32_round_to_int(fp, status));
+    set_float_rounding_mode(old_mode, status);
+    update_fcsr0_mask(env, GETPC(), float_flag_inexact);
+    return fd;
+}
+
+uint64_t helper_flogb_d(CPULoongArchState *env, uint64_t fj)
+{
+    uint64_t fd;
+    float_status *status = &env->fp_status;
+    FloatRoundMode old_mode = get_float_rounding_mode(status);
+
+    set_float_rounding_mode(float_round_down, status);
+    fd = float64_log2(fj, status);
+    fd = float64_round_to_int(fd, status);
+    set_float_rounding_mode(old_mode, status);
+    update_fcsr0_mask(env, GETPC(), float_flag_inexact);
+    return fd;
+}
+
+uint64_t helper_fclass_s(CPULoongArchState *env, uint64_t fj)
+{
+    float32 f = fj;
+    bool sign = float32_is_neg(f);
+
+    if (float32_is_infinity(f)) {
+        return sign ? 1 << 0 : 1 << 7;
+    } else if (float32_is_zero(f)) {
+        return sign ? 1 << 3 : 1 << 4;
+    } else if (float32_is_zero_or_denormal(f)) {
+        return sign ? 1 << 2 : 1 << 5;
+    } else if (float32_is_any_nan(f)) {
+        float_status s = { }; /* for snan_bit_is_one */
+        return float32_is_quiet_nan(f, &s) ? 1 << 9 : 1 << 8;
+    } else {
+        return sign ? 1 << 1 : 1 << 6;
+    }
+}
+
+uint64_t helper_fclass_d(CPULoongArchState *env, uint64_t fj)
+{
+    float64 f = fj;
+    bool sign = float64_is_neg(f);
+
+    if (float64_is_infinity(f)) {
+        return sign ? 1 << 0 : 1 << 7;
+    } else if (float64_is_zero(f)) {
+        return sign ? 1 << 3 : 1 << 4;
+    } else if (float64_is_zero_or_denormal(f)) {
+        return sign ? 1 << 2 : 1 << 5;
+    } else if (float64_is_any_nan(f)) {
+        float_status s = { }; /* for snan_bit_is_one */
+        return float64_is_quiet_nan(f, &s) ? 1 << 9 : 1 << 8;
+    } else {
+        return sign ? 1 << 1 : 1 << 6;
+    }
+}
+
+uint64_t helper_fmuladd_s(CPULoongArchState *env, uint64_t fj,
+                          uint64_t fk, uint64_t fa, uint32_t flag)
+{
+    uint64_t fd;
+
+    fd = nanbox_s(float32_muladd((uint32_t)fj, (uint32_t)fk,
+                                 (uint32_t)fa, flag, &env->fp_status));
+    update_fcsr0(env, GETPC());
+    return fd;
+}
+
+uint64_t helper_fmuladd_d(CPULoongArchState *env, uint64_t fj,
+                          uint64_t fk, uint64_t fa, uint32_t flag)
+{
+    uint64_t fd;
+
+    fd = float64_muladd(fj, fk, fa, flag, &env->fp_status);
+    update_fcsr0(env, GETPC());
+    return fd;
+}
diff --git a/target/loongarch/helper.h b/target/loongarch/helper.h
index e4b4595..25a67fd 100644
--- a/target/loongarch/helper.h
+++ b/target/loongarch/helper.h
@@ -18,3 +18,36 @@  DEF_HELPER_3(asrtgt_d, void, env, tl, tl)
 DEF_HELPER_3(crc32, tl, tl, tl, tl)
 DEF_HELPER_3(crc32c, tl, tl, tl, tl)
 DEF_HELPER_2(cpucfg, tl, env, tl)
+
+/* Floating-point helper */
+DEF_HELPER_3(fadd_s, i64, env, i64, i64)
+DEF_HELPER_3(fadd_d, i64, env, i64, i64)
+DEF_HELPER_3(fsub_s, i64, env, i64, i64)
+DEF_HELPER_3(fsub_d, i64, env, i64, i64)
+DEF_HELPER_3(fmul_s, i64, env, i64, i64)
+DEF_HELPER_3(fmul_d, i64, env, i64, i64)
+DEF_HELPER_3(fdiv_s, i64, env, i64, i64)
+DEF_HELPER_3(fdiv_d, i64, env, i64, i64)
+DEF_HELPER_3(fmaxa_s, i64, env, i64, i64)
+DEF_HELPER_3(fmaxa_d, i64, env, i64, i64)
+DEF_HELPER_3(fmina_s, i64, env, i64, i64)
+DEF_HELPER_3(fmina_d, i64, env, i64, i64)
+
+DEF_HELPER_5(fmuladd_s, i64, env, i64, i64, i64, i32)
+DEF_HELPER_5(fmuladd_d, i64, env, i64, i64, i64, i32)
+
+DEF_HELPER_3(fscaleb_s, i64, env, i64, i64)
+DEF_HELPER_3(fscaleb_d, i64, env, i64, i64)
+
+DEF_HELPER_2(flogb_s, i64, env, i64)
+DEF_HELPER_2(flogb_d, i64, env, i64)
+
+DEF_HELPER_2(fsqrt_s, i64, env, i64)
+DEF_HELPER_2(fsqrt_d, i64, env, i64)
+DEF_HELPER_2(frsqrt_s, i64, env, i64)
+DEF_HELPER_2(frsqrt_d, i64, env, i64)
+DEF_HELPER_2(frecip_s, i64, env, i64)
+DEF_HELPER_2(frecip_d, i64, env, i64)
+
+DEF_HELPER_FLAGS_2(fclass_s, TCG_CALL_NO_RWG_SE, i64, env, i64)
+DEF_HELPER_FLAGS_2(fclass_d, TCG_CALL_NO_RWG_SE, i64, env, i64)
diff --git a/target/loongarch/insn_trans/trans_farith.c b/target/loongarch/insn_trans/trans_farith.c
new file mode 100644
index 0000000..005fa85
--- /dev/null
+++ b/target/loongarch/insn_trans/trans_farith.c
@@ -0,0 +1,131 @@ 
+/*
+ * LoongArch translate functions
+ *
+ * Copyright (c) 2021 Loongson Technology Corporation Limited
+ *
+ * SPDX-License-Identifier: LGPL-2.1+
+ */
+
+static bool gen_f3(DisasContext *ctx, arg_fmt_fdfjfk *a,
+                   void (*func)(TCGv, TCGv_env, TCGv, TCGv))
+{
+    func(cpu_fpr[a->fd], cpu_env, cpu_fpr[a->fj], cpu_fpr[a->fk]);
+    return true;
+}
+
+static bool gen_f2(DisasContext *ctx, arg_fmt_fdfj *a,
+                   void (*func)(TCGv, TCGv_env, TCGv))
+{
+    func(cpu_fpr[a->fd], cpu_env, cpu_fpr[a->fj]);
+    return true;
+}
+
+static bool gen_muladd(DisasContext *ctx, arg_fmt_fdfjfkfa *a,
+                       void (*func)(TCGv, TCGv_env, TCGv, TCGv, TCGv, TCGv_i32),
+                       int flag)
+{
+    TCGv_i32 tflag = tcg_constant_i32(flag);
+    func(cpu_fpr[a->fd], cpu_env, cpu_fpr[a->fj],
+         cpu_fpr[a->fk], cpu_fpr[a->fa], tflag);
+    return true;
+}
+
+static bool trans_fcopysign_s(DisasContext *ctx, arg_fmt_fdfjfk *a)
+{
+    tcg_gen_deposit_i64(cpu_fpr[a->fd], cpu_fpr[a->fk], cpu_fpr[a->fj], 0, 31);
+    return true;
+}
+
+static bool trans_fcopysign_d(DisasContext *ctx, arg_fmt_fdfjfk *a)
+{
+    tcg_gen_deposit_i64(cpu_fpr[a->fd], cpu_fpr[a->fk], cpu_fpr[a->fj], 0, 63);
+    return true;
+}
+
+static bool trans_fmax_s(DisasContext *ctx, arg_fmt_fdfjfk *a)
+{
+    tcg_gen_umax_i64(cpu_fpr[a->fd], cpu_fpr[a->fj], cpu_fpr[a->fk]);
+    gen_nanbox_s(cpu_fpr[a->fd], cpu_fpr[a->fd]);
+    return true;
+}
+
+static bool trans_fmax_d(DisasContext *ctx, arg_fmt_fdfjfk *a)
+{
+    tcg_gen_umax_i64(cpu_fpr[a->fd], cpu_fpr[a->fj], cpu_fpr[a->fk]);
+    return true;
+}
+
+static bool trans_fmin_s(DisasContext *ctx, arg_fmt_fdfjfk *a)
+{
+    tcg_gen_umin_i64(cpu_fpr[a->fd], cpu_fpr[a->fj], cpu_fpr[a->fk]);
+    gen_nanbox_s(cpu_fpr[a->fd], cpu_fpr[a->fd]);
+    return true;
+}
+
+static bool trans_fmin_d(DisasContext *ctx, arg_fmt_fdfjfk *a)
+{
+    tcg_gen_umin_i64(cpu_fpr[a->fd], cpu_fpr[a->fj], cpu_fpr[a->fk]);
+    return true;
+}
+
+static bool trans_fabs_s(DisasContext *ctx, arg_fmt_fdfj *a)
+{
+    tcg_gen_ext32u_tl(cpu_fpr[a->fj], cpu_fpr[a->fj]);
+    tcg_gen_abs_i64(cpu_fpr[a->fd], cpu_fpr[a->fj]);
+    gen_nanbox_s(cpu_fpr[a->fd], cpu_fpr[a->fd]);
+    return true;
+}
+
+static bool trans_fabs_d(DisasContext *ctx, arg_fmt_fdfj *a)
+{
+    tcg_gen_abs_i64(cpu_fpr[a->fd], cpu_fpr[a->fj]);
+    return true;
+}
+
+static bool trans_fneg_s(DisasContext *ctx, arg_fmt_fdfj *a)
+{
+    tcg_gen_xori_i64(cpu_fpr[a->fd], cpu_fpr[a->fj], 0x80000000);
+    gen_nanbox_s(cpu_fpr[a->fd], cpu_fpr[a->fd]);
+    return true;
+}
+
+static bool trans_fneg_d(DisasContext *ctx, arg_fmt_fdfj *a)
+{
+    tcg_gen_xori_i64(cpu_fpr[a->fd], cpu_fpr[a->fj], 0x8000000000000000LL);
+    return true;
+}
+
+TRANS(fadd_s, gen_f3, gen_helper_fadd_s)
+TRANS(fadd_d, gen_f3, gen_helper_fadd_d)
+TRANS(fsub_s, gen_f3, gen_helper_fsub_s)
+TRANS(fsub_d, gen_f3, gen_helper_fsub_d)
+TRANS(fmul_s, gen_f3, gen_helper_fmul_s)
+TRANS(fmul_d, gen_f3, gen_helper_fmul_d)
+TRANS(fdiv_s, gen_f3, gen_helper_fdiv_s)
+TRANS(fdiv_d, gen_f3, gen_helper_fdiv_d)
+TRANS(fmaxa_s, gen_f3, gen_helper_fmaxa_s)
+TRANS(fmaxa_d, gen_f3, gen_helper_fmaxa_d)
+TRANS(fmina_s, gen_f3, gen_helper_fmina_s)
+TRANS(fmina_d, gen_f3, gen_helper_fmina_d)
+TRANS(fscaleb_s, gen_f3, gen_helper_fscaleb_s)
+TRANS(fscaleb_d, gen_f3, gen_helper_fscaleb_d)
+TRANS(fsqrt_s, gen_f2, gen_helper_fsqrt_s)
+TRANS(fsqrt_d, gen_f2, gen_helper_fsqrt_d)
+TRANS(frecip_s, gen_f2, gen_helper_frecip_s)
+TRANS(frecip_d, gen_f2, gen_helper_frecip_d)
+TRANS(frsqrt_s, gen_f2, gen_helper_frsqrt_s)
+TRANS(frsqrt_d, gen_f2, gen_helper_frsqrt_d)
+TRANS(flogb_s, gen_f2, gen_helper_flogb_s)
+TRANS(flogb_d, gen_f2, gen_helper_flogb_d)
+TRANS(fclass_s, gen_f2, gen_helper_fclass_s)
+TRANS(fclass_d, gen_f2, gen_helper_fclass_d)
+TRANS(fmadd_s, gen_muladd, gen_helper_fmuladd_s, 0)
+TRANS(fmadd_d, gen_muladd, gen_helper_fmuladd_d, 0)
+TRANS(fmsub_s, gen_muladd, gen_helper_fmuladd_s, float_muladd_negate_c)
+TRANS(fmsub_d, gen_muladd, gen_helper_fmuladd_d, float_muladd_negate_c)
+TRANS(fnmadd_s, gen_muladd, gen_helper_fmuladd_s,
+      float_muladd_negate_product | float_muladd_negate_c)
+TRANS(fnmadd_d, gen_muladd, gen_helper_fmuladd_d,
+      float_muladd_negate_product | float_muladd_negate_c)
+TRANS(fnmsub_s, gen_muladd, gen_helper_fmuladd_s, float_muladd_negate_product)
+TRANS(fnmsub_d, gen_muladd, gen_helper_fmuladd_d, float_muladd_negate_product)
diff --git a/target/loongarch/insns.decode b/target/loongarch/insns.decode
index 66bc314..9e6a727 100644
--- a/target/loongarch/insns.decode
+++ b/target/loongarch/insns.decode
@@ -28,6 +28,10 @@ 
 %hint    0:5
 %whint   0:15
 %code    0:15
+%fd      0:5
+%fj      5:5
+%fk      10:5
+%fa      15:5
 
 #
 # Argument sets
@@ -49,6 +53,9 @@ 
 &fmt_whint          whint
 &fmt_rjrk           rj rk
 &fmt_code           code
+&fmt_fdfjfk         fd fj fk
+&fmt_fdfjfkfa       fd fj fk fa
+&fmt_fdfj           fd fj
 
 #
 # Formats
@@ -70,6 +77,9 @@ 
 @fmt_rdrjsi14        .... .... .............. ..... .....     &fmt_rdrjsi14       %rd %rj %si14
 @fmt_rjrk            .... ........ ..... ..... ..... .....    &fmt_rjrk           %rj %rk
 @fmt_code            .... ........ ..... ...............      &fmt_code           %code
+@fmt_fdfjfk          .... ........ ..... ..... ..... .....    &fmt_fdfjfk         %fd %fj %fk
+@fmt_fdfjfkfa        .... ........ ..... ..... ..... .....    &fmt_fdfjfkfa       %fd %fj %fk %fa
+@fmt_fdfj            .... ........ ..... ..... ..... .....    &fmt_fdfj           %fd %fj
 
 #
 # Fixed point arithmetic operation instruction
@@ -285,3 +295,49 @@  rdtimel_w        0000 00000000 00000 11000 ..... .....    @fmt_rdrj
 rdtimeh_w        0000 00000000 00000 11001 ..... .....    @fmt_rdrj
 rdtime_d         0000 00000000 00000 11010 ..... .....    @fmt_rdrj
 cpucfg           0000 00000000 00000 11011 ..... .....    @fmt_rdrj
+
+#
+# Floating point arithmetic operation instruction
+#
+fadd_s           0000 00010000 00001 ..... ..... .....    @fmt_fdfjfk
+fadd_d           0000 00010000 00010 ..... ..... .....    @fmt_fdfjfk
+fsub_s           0000 00010000 00101 ..... ..... .....    @fmt_fdfjfk
+fsub_d           0000 00010000 00110 ..... ..... .....    @fmt_fdfjfk
+fmul_s           0000 00010000 01001 ..... ..... .....    @fmt_fdfjfk
+fmul_d           0000 00010000 01010 ..... ..... .....    @fmt_fdfjfk
+fdiv_s           0000 00010000 01101 ..... ..... .....    @fmt_fdfjfk
+fdiv_d           0000 00010000 01110 ..... ..... .....    @fmt_fdfjfk
+fmadd_s          0000 10000001 ..... ..... ..... .....    @fmt_fdfjfkfa
+fmadd_d          0000 10000010 ..... ..... ..... .....    @fmt_fdfjfkfa
+fmsub_s          0000 10000101 ..... ..... ..... .....    @fmt_fdfjfkfa
+fmsub_d          0000 10000110 ..... ..... ..... .....    @fmt_fdfjfkfa
+fnmadd_s         0000 10001001 ..... ..... ..... .....    @fmt_fdfjfkfa
+fnmadd_d         0000 10001010 ..... ..... ..... .....    @fmt_fdfjfkfa
+fnmsub_s         0000 10001101 ..... ..... ..... .....    @fmt_fdfjfkfa
+fnmsub_d         0000 10001110 ..... ..... ..... .....    @fmt_fdfjfkfa
+fmax_s           0000 00010000 10001 ..... ..... .....    @fmt_fdfjfk
+fmax_d           0000 00010000 10010 ..... ..... .....    @fmt_fdfjfk
+fmin_s           0000 00010000 10101 ..... ..... .....    @fmt_fdfjfk
+fmin_d           0000 00010000 10110 ..... ..... .....    @fmt_fdfjfk
+fmaxa_s          0000 00010000 11001 ..... ..... .....    @fmt_fdfjfk
+fmaxa_d          0000 00010000 11010 ..... ..... .....    @fmt_fdfjfk
+fmina_s          0000 00010000 11101 ..... ..... .....    @fmt_fdfjfk
+fmina_d          0000 00010000 11110 ..... ..... .....    @fmt_fdfjfk
+fabs_s           0000 00010001 01000 00001 ..... .....    @fmt_fdfj
+fabs_d           0000 00010001 01000 00010 ..... .....    @fmt_fdfj
+fneg_s           0000 00010001 01000 00101 ..... .....    @fmt_fdfj
+fneg_d           0000 00010001 01000 00110 ..... .....    @fmt_fdfj
+fsqrt_s          0000 00010001 01000 10001 ..... .....    @fmt_fdfj
+fsqrt_d          0000 00010001 01000 10010 ..... .....    @fmt_fdfj
+frecip_s         0000 00010001 01000 10101 ..... .....    @fmt_fdfj
+frecip_d         0000 00010001 01000 10110 ..... .....    @fmt_fdfj
+frsqrt_s         0000 00010001 01000 11001 ..... .....    @fmt_fdfj
+frsqrt_d         0000 00010001 01000 11010 ..... .....    @fmt_fdfj
+fscaleb_s        0000 00010001 00001 ..... ..... .....    @fmt_fdfjfk
+fscaleb_d        0000 00010001 00010 ..... ..... .....    @fmt_fdfjfk
+flogb_s          0000 00010001 01000 01001 ..... .....    @fmt_fdfj
+flogb_d          0000 00010001 01000 01010 ..... .....    @fmt_fdfj
+fcopysign_s      0000 00010001 00101 ..... ..... .....    @fmt_fdfjfk
+fcopysign_d      0000 00010001 00110 ..... ..... .....    @fmt_fdfjfk
+fclass_s         0000 00010001 01000 01101 ..... .....    @fmt_fdfj
+fclass_d         0000 00010001 01000 01110 ..... .....    @fmt_fdfj
diff --git a/target/loongarch/internals.h b/target/loongarch/internals.h
index 1052cb6..096a4a7 100644
--- a/target/loongarch/internals.h
+++ b/target/loongarch/internals.h
@@ -19,4 +19,6 @@  void QEMU_NORETURN do_raise_exception(CPULoongArchState *env,
 
 const char *loongarch_exception_name(int32_t exception);
 
+void restore_fp_status(CPULoongArchState *env);
+
 #endif
diff --git a/target/loongarch/translate.c b/target/loongarch/translate.c
index d0e1daa..9befe8e 100644
--- a/target/loongarch/translate.c
+++ b/target/loongarch/translate.c
@@ -16,6 +16,7 @@ 
 #include "exec/translator.h"
 #include "exec/log.h"
 #include "qemu/qemu-print.h"
+#include "fpu/softfloat.h"
 #include "translate.h"
 #include "internals.h"
 
@@ -27,6 +28,15 @@  TCGv_i64 cpu_fpr[32];
 
 #define DISAS_STOP       DISAS_TARGET_0
 
+/*
+ * LoongArch requires NaN-boxing of narrower width floating point values.
+ * This applies when a 32-bit value is assigned to a 64-bit FP register.
+ */
+static void gen_nanbox_s(TCGv_i64 out, TCGv_i64 in)
+{
+    tcg_gen_ori_i64(out, in, MAKE_64BIT_MASK(32, 32));
+}
+
 void generate_exception(DisasContext *ctx, int excp)
 {
     tcg_gen_movi_tl(cpu_pc, ctx->base.pc_next);
@@ -176,6 +186,7 @@  static void gen_set_gpr(int reg_num, TCGv t, DisasExtend dst_ext)
 #include "insn_trans/trans_memory.c"
 #include "insn_trans/trans_atomic.c"
 #include "insn_trans/trans_extra.c"
+#include "insn_trans/trans_farith.c"
 
 static void loongarch_tr_translate_insn(DisasContextBase *dcbase, CPUState *cs)
 {