From: frank.chang@sifive.com To: qemu-riscv@nongnu.org Cc: Frank Chang <frank.chang@sifive.com>, Bin Meng <bin.meng@windriver.com>, Richard Henderson <richard.henderson@linaro.org>, qemu-devel@nongnu.org, Chih-Min Chao <chihmin.chao@sifive.com>, Palmer Dabbelt <palmer@dabbelt.com>, Alistair Francis <alistair.francis@wdc.com>, Kito Cheng <kito.cheng@sifive.com> Subject: [PATCH v4 3/8] target/riscv: zfh: half-precision convert and move Date: Wed, 20 Oct 2021 11:06:46 +0800 [thread overview] Message-ID: <20211020030653.213565-4-frank.chang@sifive.com> (raw) In-Reply-To: <20211020030653.213565-1-frank.chang@sifive.com> From: Kito Cheng <kito.cheng@sifive.com> Signed-off-by: Kito Cheng <kito.cheng@sifive.com> Signed-off-by: Chih-Min Chao <chihmin.chao@sifive.com> Signed-off-by: Frank Chang <frank.chang@sifive.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Acked-by: Alistair Francis <alistair.francis@wdc.com> --- target/riscv/fpu_helper.c | 67 +++++ target/riscv/helper.h | 12 + target/riscv/insn32.decode | 19 ++ target/riscv/insn_trans/trans_rvzfh.c.inc | 288 ++++++++++++++++++++++ target/riscv/translate.c | 10 + 5 files changed, 396 insertions(+) diff --git a/target/riscv/fpu_helper.c b/target/riscv/fpu_helper.c index 20bb89ad14f..2ed9b03193c 100644 --- a/target/riscv/fpu_helper.c +++ b/target/riscv/fpu_helper.c @@ -460,3 +460,70 @@ uint64_t helper_fsqrt_h(CPURISCVState *env, uint64_t rs1) float16 frs1 = check_nanbox_h(rs1); return nanbox_h(float16_sqrt(frs1, &env->fp_status)); } + +target_ulong helper_fcvt_w_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return float16_to_int32(frs1, &env->fp_status); +} + +target_ulong helper_fcvt_wu_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return (int32_t)float16_to_uint32(frs1, &env->fp_status); +} + +target_ulong helper_fcvt_l_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return float16_to_int64(frs1, &env->fp_status); +} + +target_ulong helper_fcvt_lu_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return float16_to_uint64(frs1, &env->fp_status); +} + +uint64_t helper_fcvt_h_w(CPURISCVState *env, target_ulong rs1) +{ + return nanbox_h(int32_to_float16((int32_t)rs1, &env->fp_status)); +} + +uint64_t helper_fcvt_h_wu(CPURISCVState *env, target_ulong rs1) +{ + return nanbox_h(uint32_to_float16((uint32_t)rs1, &env->fp_status)); +} + +uint64_t helper_fcvt_h_l(CPURISCVState *env, target_ulong rs1) +{ + return nanbox_h(int64_to_float16(rs1, &env->fp_status)); +} + +uint64_t helper_fcvt_h_lu(CPURISCVState *env, target_ulong rs1) +{ + return nanbox_h(uint64_to_float16(rs1, &env->fp_status)); +} + +uint64_t helper_fcvt_h_s(CPURISCVState *env, uint64_t rs1) +{ + float32 frs1 = check_nanbox_s(rs1); + return nanbox_h(float32_to_float16(frs1, true, &env->fp_status)); +} + +uint64_t helper_fcvt_s_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return nanbox_s(float16_to_float32(frs1, true, &env->fp_status)); +} + +uint64_t helper_fcvt_h_d(CPURISCVState *env, uint64_t rs1) +{ + return nanbox_h(float64_to_float16(rs1, true, &env->fp_status)); +} + +uint64_t helper_fcvt_d_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return float16_to_float64(frs1, true, &env->fp_status); +} diff --git a/target/riscv/helper.h b/target/riscv/helper.h index c6c0323fafc..b50672d1684 100644 --- a/target/riscv/helper.h +++ b/target/riscv/helper.h @@ -74,6 +74,18 @@ DEF_HELPER_FLAGS_3(fdiv_h, TCG_CALL_NO_RWG, i64, env, i64, i64) DEF_HELPER_FLAGS_3(fmin_h, TCG_CALL_NO_RWG, i64, env, i64, i64) DEF_HELPER_FLAGS_3(fmax_h, TCG_CALL_NO_RWG, i64, env, i64, i64) DEF_HELPER_FLAGS_2(fsqrt_h, TCG_CALL_NO_RWG, i64, env, i64) +DEF_HELPER_FLAGS_2(fcvt_s_h, TCG_CALL_NO_RWG, i64, env, i64) +DEF_HELPER_FLAGS_2(fcvt_h_s, TCG_CALL_NO_RWG, i64, env, i64) +DEF_HELPER_FLAGS_2(fcvt_d_h, TCG_CALL_NO_RWG, i64, env, i64) +DEF_HELPER_FLAGS_2(fcvt_h_d, TCG_CALL_NO_RWG, i64, env, i64) +DEF_HELPER_FLAGS_2(fcvt_w_h, TCG_CALL_NO_RWG, tl, env, i64) +DEF_HELPER_FLAGS_2(fcvt_wu_h, TCG_CALL_NO_RWG, tl, env, i64) +DEF_HELPER_FLAGS_2(fcvt_l_h, TCG_CALL_NO_RWG, tl, env, i64) +DEF_HELPER_FLAGS_2(fcvt_lu_h, TCG_CALL_NO_RWG, tl, env, i64) +DEF_HELPER_FLAGS_2(fcvt_h_w, TCG_CALL_NO_RWG, i64, env, tl) +DEF_HELPER_FLAGS_2(fcvt_h_wu, TCG_CALL_NO_RWG, i64, env, tl) +DEF_HELPER_FLAGS_2(fcvt_h_l, TCG_CALL_NO_RWG, i64, env, tl) +DEF_HELPER_FLAGS_2(fcvt_h_lu, TCG_CALL_NO_RWG, i64, env, tl) /* Special functions */ DEF_HELPER_2(csrr, tl, env, int) diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode index 66c231a3010..ba40f3e7f89 100644 --- a/target/riscv/insn32.decode +++ b/target/riscv/insn32.decode @@ -739,5 +739,24 @@ fsub_h 0000110 ..... ..... ... ..... 1010011 @r_rm fmul_h 0001010 ..... ..... ... ..... 1010011 @r_rm fdiv_h 0001110 ..... ..... ... ..... 1010011 @r_rm fsqrt_h 0101110 00000 ..... ... ..... 1010011 @r2_rm +fsgnj_h 0010010 ..... ..... 000 ..... 1010011 @r +fsgnjn_h 0010010 ..... ..... 001 ..... 1010011 @r +fsgnjx_h 0010010 ..... ..... 010 ..... 1010011 @r fmin_h 0010110 ..... ..... 000 ..... 1010011 @r fmax_h 0010110 ..... ..... 001 ..... 1010011 @r +fcvt_h_s 0100010 00000 ..... ... ..... 1010011 @r2_rm +fcvt_s_h 0100000 00010 ..... ... ..... 1010011 @r2_rm +fcvt_h_d 0100010 00001 ..... ... ..... 1010011 @r2_rm +fcvt_d_h 0100001 00010 ..... ... ..... 1010011 @r2_rm +fcvt_w_h 1100010 00000 ..... ... ..... 1010011 @r2_rm +fcvt_wu_h 1100010 00001 ..... ... ..... 1010011 @r2_rm +fmv_x_h 1110010 00000 ..... 000 ..... 1010011 @r2 +fcvt_h_w 1101010 00000 ..... ... ..... 1010011 @r2_rm +fcvt_h_wu 1101010 00001 ..... ... ..... 1010011 @r2_rm +fmv_h_x 1111010 00000 ..... 000 ..... 1010011 @r2 + +# *** RV64 Zfh Extension (in addition to RV32 Zfh) *** +fcvt_l_h 1100010 00010 ..... ... ..... 1010011 @r2_rm +fcvt_lu_h 1100010 00011 ..... ... ..... 1010011 @r2_rm +fcvt_h_l 1101010 00010 ..... ... ..... 1010011 @r2_rm +fcvt_h_lu 1101010 00011 ..... ... ..... 1010011 @r2_rm diff --git a/target/riscv/insn_trans/trans_rvzfh.c.inc b/target/riscv/insn_trans/trans_rvzfh.c.inc index 9764d76f8bc..d1250257666 100644 --- a/target/riscv/insn_trans/trans_rvzfh.c.inc +++ b/target/riscv/insn_trans/trans_rvzfh.c.inc @@ -171,6 +171,93 @@ static bool trans_fsqrt_h(DisasContext *ctx, arg_fsqrt_h *a) return true; } +static bool trans_fsgnj_h(DisasContext *ctx, arg_fsgnj_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + if (a->rs1 == a->rs2) { /* FMOV */ + gen_check_nanbox_h(cpu_fpr[a->rd], cpu_fpr[a->rs1]); + } else { + TCGv_i64 rs1 = tcg_temp_new_i64(); + TCGv_i64 rs2 = tcg_temp_new_i64(); + + gen_check_nanbox_h(rs1, cpu_fpr[a->rs1]); + gen_check_nanbox_h(rs2, cpu_fpr[a->rs2]); + + /* This formulation retains the nanboxing of rs2. */ + tcg_gen_deposit_i64(cpu_fpr[a->rd], rs2, rs1, 0, 15); + tcg_temp_free_i64(rs1); + tcg_temp_free_i64(rs2); + } + + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fsgnjn_h(DisasContext *ctx, arg_fsgnjn_h *a) +{ + TCGv_i64 rs1, rs2, mask; + + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + rs1 = tcg_temp_new_i64(); + gen_check_nanbox_h(rs1, cpu_fpr[a->rs1]); + + if (a->rs1 == a->rs2) { /* FNEG */ + tcg_gen_xori_i64(cpu_fpr[a->rd], rs1, MAKE_64BIT_MASK(15, 1)); + } else { + rs2 = tcg_temp_new_i64(); + gen_check_nanbox_h(rs2, cpu_fpr[a->rs2]); + + /* + * Replace bit 15 in rs1 with inverse in rs2. + * This formulation retains the nanboxing of rs1. + */ + mask = tcg_const_i64(~MAKE_64BIT_MASK(15, 1)); + tcg_gen_not_i64(rs2, rs2); + tcg_gen_andc_i64(rs2, rs2, mask); + tcg_gen_and_i64(rs1, mask, rs1); + tcg_gen_or_i64(cpu_fpr[a->rd], rs1, rs2); + + tcg_temp_free_i64(mask); + tcg_temp_free_i64(rs2); + } + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fsgnjx_h(DisasContext *ctx, arg_fsgnjx_h *a) +{ + TCGv_i64 rs1, rs2; + + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + rs1 = tcg_temp_new_i64(); + gen_check_nanbox_s(rs1, cpu_fpr[a->rs1]); + + if (a->rs1 == a->rs2) { /* FABS */ + tcg_gen_andi_i64(cpu_fpr[a->rd], rs1, ~MAKE_64BIT_MASK(15, 1)); + } else { + rs2 = tcg_temp_new_i64(); + gen_check_nanbox_s(rs2, cpu_fpr[a->rs2]); + + /* + * Xor bit 15 in rs1 with that in rs2. + * This formulation retains the nanboxing of rs1. + */ + tcg_gen_andi_i64(rs2, rs2, MAKE_64BIT_MASK(15, 1)); + tcg_gen_xor_i64(cpu_fpr[a->rd], rs1, rs2); + + tcg_temp_free_i64(rs2); + } + + mark_fs_dirty(ctx); + return true; +} + static bool trans_fmin_h(DisasContext *ctx, arg_fmin_h *a) { REQUIRE_FPU; @@ -192,3 +279,204 @@ static bool trans_fmax_h(DisasContext *ctx, arg_fmax_h *a) mark_fs_dirty(ctx); return true; } + +static bool trans_fcvt_s_h(DisasContext *ctx, arg_fcvt_s_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_s_h(cpu_fpr[a->rd], cpu_env, cpu_fpr[a->rs1]); + + mark_fs_dirty(ctx); + + return true; +} + +static bool trans_fcvt_d_h(DisasContext *ctx, arg_fcvt_d_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + REQUIRE_EXT(ctx, RVD); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_d_h(cpu_fpr[a->rd], cpu_env, cpu_fpr[a->rs1]); + + mark_fs_dirty(ctx); + + + return true; +} + +static bool trans_fcvt_h_s(DisasContext *ctx, arg_fcvt_h_s *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_s(cpu_fpr[a->rd], cpu_env, cpu_fpr[a->rs1]); + + mark_fs_dirty(ctx); + + return true; +} + +static bool trans_fcvt_h_d(DisasContext *ctx, arg_fcvt_h_d *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + REQUIRE_EXT(ctx, RVD); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_d(cpu_fpr[a->rd], cpu_env, cpu_fpr[a->rs1]); + + mark_fs_dirty(ctx); + + return true; +} + +static bool trans_fcvt_w_h(DisasContext *ctx, arg_fcvt_w_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv dest = dest_gpr(ctx, a->rd); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_w_h(dest, cpu_env, cpu_fpr[a->rs1]); + gen_set_gpr(ctx, a->rd, dest); + return true; +} + +static bool trans_fcvt_wu_h(DisasContext *ctx, arg_fcvt_wu_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv dest = dest_gpr(ctx, a->rd); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_wu_h(dest, cpu_env, cpu_fpr[a->rs1]); + gen_set_gpr(ctx, a->rd, dest); + return true; +} + +static bool trans_fcvt_h_w(DisasContext *ctx, arg_fcvt_h_w *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv t0 = get_gpr(ctx, a->rs1, EXT_SIGN); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_w(cpu_fpr[a->rd], cpu_env, t0); + + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fcvt_h_wu(DisasContext *ctx, arg_fcvt_h_wu *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv t0 = get_gpr(ctx, a->rs1, EXT_SIGN); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_wu(cpu_fpr[a->rd], cpu_env, t0); + + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fmv_x_h(DisasContext *ctx, arg_fmv_x_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv dest = dest_gpr(ctx, a->rd); + +#if defined(TARGET_RISCV64) + /* 16 bits -> 64 bits */ + tcg_gen_ext16s_tl(dest, cpu_fpr[a->rs1]); +#else + /* 16 bits -> 32 bits */ + tcg_gen_extrl_i64_i32(dest, cpu_fpr[a->rs1]); + tcg_gen_ext16s_tl(dest, dest); +#endif + + gen_set_gpr(ctx, a->rd, dest); + return true; +} + +static bool trans_fmv_h_x(DisasContext *ctx, arg_fmv_h_x *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv t0 = get_gpr(ctx, a->rs1, EXT_ZERO); + + tcg_gen_extu_tl_i64(cpu_fpr[a->rd], t0); + gen_nanbox_h(cpu_fpr[a->rd], cpu_fpr[a->rd]); + + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fcvt_l_h(DisasContext *ctx, arg_fcvt_l_h *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv dest = dest_gpr(ctx, a->rd); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_l_h(dest, cpu_env, cpu_fpr[a->rs1]); + gen_set_gpr(ctx, a->rd, dest); + return true; +} + +static bool trans_fcvt_lu_h(DisasContext *ctx, arg_fcvt_lu_h *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv dest = dest_gpr(ctx, a->rd); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_lu_h(dest, cpu_env, cpu_fpr[a->rs1]); + gen_set_gpr(ctx, a->rd, dest); + return true; +} + +static bool trans_fcvt_h_l(DisasContext *ctx, arg_fcvt_h_l *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv t0 = get_gpr(ctx, a->rs1, EXT_SIGN); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_l(cpu_fpr[a->rd], cpu_env, t0); + + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fcvt_h_lu(DisasContext *ctx, arg_fcvt_h_lu *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv t0 = get_gpr(ctx, a->rs1, EXT_SIGN); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_lu(cpu_fpr[a->rd], cpu_env, t0); + + mark_fs_dirty(ctx); + return true; +} diff --git a/target/riscv/translate.c b/target/riscv/translate.c index 75048149f5a..442ef42f441 100644 --- a/target/riscv/translate.c +++ b/target/riscv/translate.c @@ -132,6 +132,16 @@ static void gen_nanbox_h(TCGv_i64 out, TCGv_i64 in) * * Here, the result is always nan-boxed, even the canonical nan. */ +static void gen_check_nanbox_h(TCGv_i64 out, TCGv_i64 in) +{ + TCGv_i64 t_max = tcg_const_i64(0xffffffffffff0000ull); + TCGv_i64 t_nan = tcg_const_i64(0xffffffffffff7e00ull); + + tcg_gen_movcond_i64(TCG_COND_GEU, out, in, t_max, in, t_nan); + tcg_temp_free_i64(t_max); + tcg_temp_free_i64(t_nan); +} + static void gen_check_nanbox_s(TCGv_i64 out, TCGv_i64 in) { TCGv_i64 t_max = tcg_constant_i64(0xffffffff00000000ull); -- 2.25.1
WARNING: multiple messages have this Message-ID (diff)
From: frank.chang@sifive.com To: qemu-riscv@nongnu.org Cc: Kito Cheng <kito.cheng@sifive.com>, Chih-Min Chao <chihmin.chao@sifive.com>, Frank Chang <frank.chang@sifive.com>, Richard Henderson <richard.henderson@linaro.org>, Alistair Francis <alistair.francis@wdc.com>, Palmer Dabbelt <palmer@dabbelt.com>, Bin Meng <bin.meng@windriver.com>, qemu-devel@nongnu.org Subject: [PATCH v4 3/8] target/riscv: zfh: half-precision convert and move Date: Wed, 20 Oct 2021 11:06:46 +0800 [thread overview] Message-ID: <20211020030653.213565-4-frank.chang@sifive.com> (raw) In-Reply-To: <20211020030653.213565-1-frank.chang@sifive.com> From: Kito Cheng <kito.cheng@sifive.com> Signed-off-by: Kito Cheng <kito.cheng@sifive.com> Signed-off-by: Chih-Min Chao <chihmin.chao@sifive.com> Signed-off-by: Frank Chang <frank.chang@sifive.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Acked-by: Alistair Francis <alistair.francis@wdc.com> --- target/riscv/fpu_helper.c | 67 +++++ target/riscv/helper.h | 12 + target/riscv/insn32.decode | 19 ++ target/riscv/insn_trans/trans_rvzfh.c.inc | 288 ++++++++++++++++++++++ target/riscv/translate.c | 10 + 5 files changed, 396 insertions(+) diff --git a/target/riscv/fpu_helper.c b/target/riscv/fpu_helper.c index 20bb89ad14f..2ed9b03193c 100644 --- a/target/riscv/fpu_helper.c +++ b/target/riscv/fpu_helper.c @@ -460,3 +460,70 @@ uint64_t helper_fsqrt_h(CPURISCVState *env, uint64_t rs1) float16 frs1 = check_nanbox_h(rs1); return nanbox_h(float16_sqrt(frs1, &env->fp_status)); } + +target_ulong helper_fcvt_w_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return float16_to_int32(frs1, &env->fp_status); +} + +target_ulong helper_fcvt_wu_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return (int32_t)float16_to_uint32(frs1, &env->fp_status); +} + +target_ulong helper_fcvt_l_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return float16_to_int64(frs1, &env->fp_status); +} + +target_ulong helper_fcvt_lu_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return float16_to_uint64(frs1, &env->fp_status); +} + +uint64_t helper_fcvt_h_w(CPURISCVState *env, target_ulong rs1) +{ + return nanbox_h(int32_to_float16((int32_t)rs1, &env->fp_status)); +} + +uint64_t helper_fcvt_h_wu(CPURISCVState *env, target_ulong rs1) +{ + return nanbox_h(uint32_to_float16((uint32_t)rs1, &env->fp_status)); +} + +uint64_t helper_fcvt_h_l(CPURISCVState *env, target_ulong rs1) +{ + return nanbox_h(int64_to_float16(rs1, &env->fp_status)); +} + +uint64_t helper_fcvt_h_lu(CPURISCVState *env, target_ulong rs1) +{ + return nanbox_h(uint64_to_float16(rs1, &env->fp_status)); +} + +uint64_t helper_fcvt_h_s(CPURISCVState *env, uint64_t rs1) +{ + float32 frs1 = check_nanbox_s(rs1); + return nanbox_h(float32_to_float16(frs1, true, &env->fp_status)); +} + +uint64_t helper_fcvt_s_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return nanbox_s(float16_to_float32(frs1, true, &env->fp_status)); +} + +uint64_t helper_fcvt_h_d(CPURISCVState *env, uint64_t rs1) +{ + return nanbox_h(float64_to_float16(rs1, true, &env->fp_status)); +} + +uint64_t helper_fcvt_d_h(CPURISCVState *env, uint64_t rs1) +{ + float16 frs1 = check_nanbox_h(rs1); + return float16_to_float64(frs1, true, &env->fp_status); +} diff --git a/target/riscv/helper.h b/target/riscv/helper.h index c6c0323fafc..b50672d1684 100644 --- a/target/riscv/helper.h +++ b/target/riscv/helper.h @@ -74,6 +74,18 @@ DEF_HELPER_FLAGS_3(fdiv_h, TCG_CALL_NO_RWG, i64, env, i64, i64) DEF_HELPER_FLAGS_3(fmin_h, TCG_CALL_NO_RWG, i64, env, i64, i64) DEF_HELPER_FLAGS_3(fmax_h, TCG_CALL_NO_RWG, i64, env, i64, i64) DEF_HELPER_FLAGS_2(fsqrt_h, TCG_CALL_NO_RWG, i64, env, i64) +DEF_HELPER_FLAGS_2(fcvt_s_h, TCG_CALL_NO_RWG, i64, env, i64) +DEF_HELPER_FLAGS_2(fcvt_h_s, TCG_CALL_NO_RWG, i64, env, i64) +DEF_HELPER_FLAGS_2(fcvt_d_h, TCG_CALL_NO_RWG, i64, env, i64) +DEF_HELPER_FLAGS_2(fcvt_h_d, TCG_CALL_NO_RWG, i64, env, i64) +DEF_HELPER_FLAGS_2(fcvt_w_h, TCG_CALL_NO_RWG, tl, env, i64) +DEF_HELPER_FLAGS_2(fcvt_wu_h, TCG_CALL_NO_RWG, tl, env, i64) +DEF_HELPER_FLAGS_2(fcvt_l_h, TCG_CALL_NO_RWG, tl, env, i64) +DEF_HELPER_FLAGS_2(fcvt_lu_h, TCG_CALL_NO_RWG, tl, env, i64) +DEF_HELPER_FLAGS_2(fcvt_h_w, TCG_CALL_NO_RWG, i64, env, tl) +DEF_HELPER_FLAGS_2(fcvt_h_wu, TCG_CALL_NO_RWG, i64, env, tl) +DEF_HELPER_FLAGS_2(fcvt_h_l, TCG_CALL_NO_RWG, i64, env, tl) +DEF_HELPER_FLAGS_2(fcvt_h_lu, TCG_CALL_NO_RWG, i64, env, tl) /* Special functions */ DEF_HELPER_2(csrr, tl, env, int) diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode index 66c231a3010..ba40f3e7f89 100644 --- a/target/riscv/insn32.decode +++ b/target/riscv/insn32.decode @@ -739,5 +739,24 @@ fsub_h 0000110 ..... ..... ... ..... 1010011 @r_rm fmul_h 0001010 ..... ..... ... ..... 1010011 @r_rm fdiv_h 0001110 ..... ..... ... ..... 1010011 @r_rm fsqrt_h 0101110 00000 ..... ... ..... 1010011 @r2_rm +fsgnj_h 0010010 ..... ..... 000 ..... 1010011 @r +fsgnjn_h 0010010 ..... ..... 001 ..... 1010011 @r +fsgnjx_h 0010010 ..... ..... 010 ..... 1010011 @r fmin_h 0010110 ..... ..... 000 ..... 1010011 @r fmax_h 0010110 ..... ..... 001 ..... 1010011 @r +fcvt_h_s 0100010 00000 ..... ... ..... 1010011 @r2_rm +fcvt_s_h 0100000 00010 ..... ... ..... 1010011 @r2_rm +fcvt_h_d 0100010 00001 ..... ... ..... 1010011 @r2_rm +fcvt_d_h 0100001 00010 ..... ... ..... 1010011 @r2_rm +fcvt_w_h 1100010 00000 ..... ... ..... 1010011 @r2_rm +fcvt_wu_h 1100010 00001 ..... ... ..... 1010011 @r2_rm +fmv_x_h 1110010 00000 ..... 000 ..... 1010011 @r2 +fcvt_h_w 1101010 00000 ..... ... ..... 1010011 @r2_rm +fcvt_h_wu 1101010 00001 ..... ... ..... 1010011 @r2_rm +fmv_h_x 1111010 00000 ..... 000 ..... 1010011 @r2 + +# *** RV64 Zfh Extension (in addition to RV32 Zfh) *** +fcvt_l_h 1100010 00010 ..... ... ..... 1010011 @r2_rm +fcvt_lu_h 1100010 00011 ..... ... ..... 1010011 @r2_rm +fcvt_h_l 1101010 00010 ..... ... ..... 1010011 @r2_rm +fcvt_h_lu 1101010 00011 ..... ... ..... 1010011 @r2_rm diff --git a/target/riscv/insn_trans/trans_rvzfh.c.inc b/target/riscv/insn_trans/trans_rvzfh.c.inc index 9764d76f8bc..d1250257666 100644 --- a/target/riscv/insn_trans/trans_rvzfh.c.inc +++ b/target/riscv/insn_trans/trans_rvzfh.c.inc @@ -171,6 +171,93 @@ static bool trans_fsqrt_h(DisasContext *ctx, arg_fsqrt_h *a) return true; } +static bool trans_fsgnj_h(DisasContext *ctx, arg_fsgnj_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + if (a->rs1 == a->rs2) { /* FMOV */ + gen_check_nanbox_h(cpu_fpr[a->rd], cpu_fpr[a->rs1]); + } else { + TCGv_i64 rs1 = tcg_temp_new_i64(); + TCGv_i64 rs2 = tcg_temp_new_i64(); + + gen_check_nanbox_h(rs1, cpu_fpr[a->rs1]); + gen_check_nanbox_h(rs2, cpu_fpr[a->rs2]); + + /* This formulation retains the nanboxing of rs2. */ + tcg_gen_deposit_i64(cpu_fpr[a->rd], rs2, rs1, 0, 15); + tcg_temp_free_i64(rs1); + tcg_temp_free_i64(rs2); + } + + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fsgnjn_h(DisasContext *ctx, arg_fsgnjn_h *a) +{ + TCGv_i64 rs1, rs2, mask; + + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + rs1 = tcg_temp_new_i64(); + gen_check_nanbox_h(rs1, cpu_fpr[a->rs1]); + + if (a->rs1 == a->rs2) { /* FNEG */ + tcg_gen_xori_i64(cpu_fpr[a->rd], rs1, MAKE_64BIT_MASK(15, 1)); + } else { + rs2 = tcg_temp_new_i64(); + gen_check_nanbox_h(rs2, cpu_fpr[a->rs2]); + + /* + * Replace bit 15 in rs1 with inverse in rs2. + * This formulation retains the nanboxing of rs1. + */ + mask = tcg_const_i64(~MAKE_64BIT_MASK(15, 1)); + tcg_gen_not_i64(rs2, rs2); + tcg_gen_andc_i64(rs2, rs2, mask); + tcg_gen_and_i64(rs1, mask, rs1); + tcg_gen_or_i64(cpu_fpr[a->rd], rs1, rs2); + + tcg_temp_free_i64(mask); + tcg_temp_free_i64(rs2); + } + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fsgnjx_h(DisasContext *ctx, arg_fsgnjx_h *a) +{ + TCGv_i64 rs1, rs2; + + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + rs1 = tcg_temp_new_i64(); + gen_check_nanbox_s(rs1, cpu_fpr[a->rs1]); + + if (a->rs1 == a->rs2) { /* FABS */ + tcg_gen_andi_i64(cpu_fpr[a->rd], rs1, ~MAKE_64BIT_MASK(15, 1)); + } else { + rs2 = tcg_temp_new_i64(); + gen_check_nanbox_s(rs2, cpu_fpr[a->rs2]); + + /* + * Xor bit 15 in rs1 with that in rs2. + * This formulation retains the nanboxing of rs1. + */ + tcg_gen_andi_i64(rs2, rs2, MAKE_64BIT_MASK(15, 1)); + tcg_gen_xor_i64(cpu_fpr[a->rd], rs1, rs2); + + tcg_temp_free_i64(rs2); + } + + mark_fs_dirty(ctx); + return true; +} + static bool trans_fmin_h(DisasContext *ctx, arg_fmin_h *a) { REQUIRE_FPU; @@ -192,3 +279,204 @@ static bool trans_fmax_h(DisasContext *ctx, arg_fmax_h *a) mark_fs_dirty(ctx); return true; } + +static bool trans_fcvt_s_h(DisasContext *ctx, arg_fcvt_s_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_s_h(cpu_fpr[a->rd], cpu_env, cpu_fpr[a->rs1]); + + mark_fs_dirty(ctx); + + return true; +} + +static bool trans_fcvt_d_h(DisasContext *ctx, arg_fcvt_d_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + REQUIRE_EXT(ctx, RVD); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_d_h(cpu_fpr[a->rd], cpu_env, cpu_fpr[a->rs1]); + + mark_fs_dirty(ctx); + + + return true; +} + +static bool trans_fcvt_h_s(DisasContext *ctx, arg_fcvt_h_s *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_s(cpu_fpr[a->rd], cpu_env, cpu_fpr[a->rs1]); + + mark_fs_dirty(ctx); + + return true; +} + +static bool trans_fcvt_h_d(DisasContext *ctx, arg_fcvt_h_d *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + REQUIRE_EXT(ctx, RVD); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_d(cpu_fpr[a->rd], cpu_env, cpu_fpr[a->rs1]); + + mark_fs_dirty(ctx); + + return true; +} + +static bool trans_fcvt_w_h(DisasContext *ctx, arg_fcvt_w_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv dest = dest_gpr(ctx, a->rd); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_w_h(dest, cpu_env, cpu_fpr[a->rs1]); + gen_set_gpr(ctx, a->rd, dest); + return true; +} + +static bool trans_fcvt_wu_h(DisasContext *ctx, arg_fcvt_wu_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv dest = dest_gpr(ctx, a->rd); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_wu_h(dest, cpu_env, cpu_fpr[a->rs1]); + gen_set_gpr(ctx, a->rd, dest); + return true; +} + +static bool trans_fcvt_h_w(DisasContext *ctx, arg_fcvt_h_w *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv t0 = get_gpr(ctx, a->rs1, EXT_SIGN); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_w(cpu_fpr[a->rd], cpu_env, t0); + + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fcvt_h_wu(DisasContext *ctx, arg_fcvt_h_wu *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv t0 = get_gpr(ctx, a->rs1, EXT_SIGN); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_wu(cpu_fpr[a->rd], cpu_env, t0); + + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fmv_x_h(DisasContext *ctx, arg_fmv_x_h *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv dest = dest_gpr(ctx, a->rd); + +#if defined(TARGET_RISCV64) + /* 16 bits -> 64 bits */ + tcg_gen_ext16s_tl(dest, cpu_fpr[a->rs1]); +#else + /* 16 bits -> 32 bits */ + tcg_gen_extrl_i64_i32(dest, cpu_fpr[a->rs1]); + tcg_gen_ext16s_tl(dest, dest); +#endif + + gen_set_gpr(ctx, a->rd, dest); + return true; +} + +static bool trans_fmv_h_x(DisasContext *ctx, arg_fmv_h_x *a) +{ + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv t0 = get_gpr(ctx, a->rs1, EXT_ZERO); + + tcg_gen_extu_tl_i64(cpu_fpr[a->rd], t0); + gen_nanbox_h(cpu_fpr[a->rd], cpu_fpr[a->rd]); + + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fcvt_l_h(DisasContext *ctx, arg_fcvt_l_h *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv dest = dest_gpr(ctx, a->rd); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_l_h(dest, cpu_env, cpu_fpr[a->rs1]); + gen_set_gpr(ctx, a->rd, dest); + return true; +} + +static bool trans_fcvt_lu_h(DisasContext *ctx, arg_fcvt_lu_h *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv dest = dest_gpr(ctx, a->rd); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_lu_h(dest, cpu_env, cpu_fpr[a->rs1]); + gen_set_gpr(ctx, a->rd, dest); + return true; +} + +static bool trans_fcvt_h_l(DisasContext *ctx, arg_fcvt_h_l *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv t0 = get_gpr(ctx, a->rs1, EXT_SIGN); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_l(cpu_fpr[a->rd], cpu_env, t0); + + mark_fs_dirty(ctx); + return true; +} + +static bool trans_fcvt_h_lu(DisasContext *ctx, arg_fcvt_h_lu *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_FPU; + REQUIRE_ZFH(ctx); + + TCGv t0 = get_gpr(ctx, a->rs1, EXT_SIGN); + + gen_set_rm(ctx, a->rm); + gen_helper_fcvt_h_lu(cpu_fpr[a->rd], cpu_env, t0); + + mark_fs_dirty(ctx); + return true; +} diff --git a/target/riscv/translate.c b/target/riscv/translate.c index 75048149f5a..442ef42f441 100644 --- a/target/riscv/translate.c +++ b/target/riscv/translate.c @@ -132,6 +132,16 @@ static void gen_nanbox_h(TCGv_i64 out, TCGv_i64 in) * * Here, the result is always nan-boxed, even the canonical nan. */ +static void gen_check_nanbox_h(TCGv_i64 out, TCGv_i64 in) +{ + TCGv_i64 t_max = tcg_const_i64(0xffffffffffff0000ull); + TCGv_i64 t_nan = tcg_const_i64(0xffffffffffff7e00ull); + + tcg_gen_movcond_i64(TCG_COND_GEU, out, in, t_max, in, t_nan); + tcg_temp_free_i64(t_max); + tcg_temp_free_i64(t_nan); +} + static void gen_check_nanbox_s(TCGv_i64 out, TCGv_i64 in) { TCGv_i64 t_max = tcg_constant_i64(0xffffffff00000000ull); -- 2.25.1
next prev parent reply other threads:[~2021-10-20 3:14 UTC|newest] Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-10-20 3:06 [PATCH v4 0/8] target/riscv: support Zfh, Zfhmin extension v0.1 frank.chang 2021-10-20 3:06 ` [PATCH v4 1/8] target/riscv: zfh: half-precision load and store frank.chang 2021-10-20 3:06 ` frank.chang 2021-10-20 3:06 ` [PATCH v4 2/8] target/riscv: zfh: half-precision computational frank.chang 2021-10-20 3:06 ` frank.chang 2021-10-20 3:06 ` frank.chang [this message] 2021-10-20 3:06 ` [PATCH v4 3/8] target/riscv: zfh: half-precision convert and move frank.chang 2021-10-20 3:06 ` [PATCH v4 4/8] target/riscv: zfh: half-precision floating-point compare frank.chang 2021-10-20 3:06 ` frank.chang 2021-10-20 3:06 ` [PATCH v4 5/8] target/riscv: zfh: half-precision floating-point classify frank.chang 2021-10-20 3:06 ` frank.chang 2021-10-20 3:06 ` [PATCH v4 6/8] target/riscv: zfh: add Zfh cpu property frank.chang 2021-10-20 3:06 ` frank.chang 2021-10-20 23:25 ` Alistair Francis 2021-10-20 23:25 ` Alistair Francis 2021-10-20 3:06 ` [PATCH v4 7/8] target/riscv: zfh: implement zfhmin extension frank.chang 2021-10-20 3:06 ` frank.chang 2021-10-20 3:06 ` [PATCH v4 8/8] target/riscv: zfh: add Zfhmin cpu property frank.chang 2021-10-20 3:06 ` frank.chang 2021-10-20 23:25 ` Alistair Francis 2021-10-20 23:25 ` Alistair Francis 2021-10-20 23:26 ` Alistair Francis 2021-10-20 23:26 ` Alistair Francis
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=20211020030653.213565-4-frank.chang@sifive.com \ --to=frank.chang@sifive.com \ --cc=alistair.francis@wdc.com \ --cc=bin.meng@windriver.com \ --cc=chihmin.chao@sifive.com \ --cc=kito.cheng@sifive.com \ --cc=palmer@dabbelt.com \ --cc=qemu-devel@nongnu.org \ --cc=qemu-riscv@nongnu.org \ --cc=richard.henderson@linaro.org \ /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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.