[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH v8 64/78] target/riscv: rvv-1.0: narrowing floating-point/int
From: |
Alistair Francis |
Subject: |
Re: [PATCH v8 64/78] target/riscv: rvv-1.0: narrowing floating-point/integer type-convert |
Date: |
Tue, 26 Oct 2021 16:32:46 +1000 |
On Fri, Oct 15, 2021 at 6:41 PM <frank.chang@sifive.com> wrote:
>
> From: Frank Chang <frank.chang@sifive.com>
>
> Signed-off-by: Frank Chang <frank.chang@sifive.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Alistair
> ---
> target/riscv/helper.h | 22 ++++-----
> target/riscv/insn32.decode | 15 ++++---
> target/riscv/insn_trans/trans_rvv.c.inc | 59 +++++++++++++++++++++----
> target/riscv/vector_helper.c | 45 ++++++++++---------
> 4 files changed, 97 insertions(+), 44 deletions(-)
>
> diff --git a/target/riscv/helper.h b/target/riscv/helper.h
> index 606bf72d5cb..1a0d817f0f5 100644
> --- a/target/riscv/helper.h
> +++ b/target/riscv/helper.h
> @@ -938,16 +938,18 @@ DEF_HELPER_5(vfwcvt_f_x_v_w, void, ptr, ptr, ptr, env,
> i32)
> DEF_HELPER_5(vfwcvt_f_f_v_h, void, ptr, ptr, ptr, env, i32)
> DEF_HELPER_5(vfwcvt_f_f_v_w, void, ptr, ptr, ptr, env, i32)
>
> -DEF_HELPER_5(vfncvt_xu_f_v_h, void, ptr, ptr, ptr, env, i32)
> -DEF_HELPER_5(vfncvt_xu_f_v_w, void, ptr, ptr, ptr, env, i32)
> -DEF_HELPER_5(vfncvt_x_f_v_h, void, ptr, ptr, ptr, env, i32)
> -DEF_HELPER_5(vfncvt_x_f_v_w, void, ptr, ptr, ptr, env, i32)
> -DEF_HELPER_5(vfncvt_f_xu_v_h, void, ptr, ptr, ptr, env, i32)
> -DEF_HELPER_5(vfncvt_f_xu_v_w, void, ptr, ptr, ptr, env, i32)
> -DEF_HELPER_5(vfncvt_f_x_v_h, void, ptr, ptr, ptr, env, i32)
> -DEF_HELPER_5(vfncvt_f_x_v_w, void, ptr, ptr, ptr, env, i32)
> -DEF_HELPER_5(vfncvt_f_f_v_h, void, ptr, ptr, ptr, env, i32)
> -DEF_HELPER_5(vfncvt_f_f_v_w, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_xu_f_w_b, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_xu_f_w_h, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_xu_f_w_w, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_x_f_w_b, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_x_f_w_h, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_x_f_w_w, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_f_xu_w_h, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_f_xu_w_w, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_f_x_w_h, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_f_x_w_w, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_f_f_w_h, void, ptr, ptr, ptr, env, i32)
> +DEF_HELPER_5(vfncvt_f_f_w_w, void, ptr, ptr, ptr, env, i32)
>
> DEF_HELPER_6(vredsum_vs_b, void, ptr, ptr, ptr, ptr, env, i32)
> DEF_HELPER_6(vredsum_vs_h, void, ptr, ptr, ptr, ptr, env, i32)
> diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode
> index 664d0fb3716..c4fdc76a269 100644
> --- a/target/riscv/insn32.decode
> +++ b/target/riscv/insn32.decode
> @@ -600,11 +600,16 @@ vfwcvt_f_x_v 010010 . ..... 01011 001 .....
> 1010111 @r2_vm
> vfwcvt_f_f_v 010010 . ..... 01100 001 ..... 1010111 @r2_vm
> vfwcvt_rtz_xu_f_v 010010 . ..... 01110 001 ..... 1010111 @r2_vm
> vfwcvt_rtz_x_f_v 010010 . ..... 01111 001 ..... 1010111 @r2_vm
> -vfncvt_xu_f_v 100010 . ..... 10000 001 ..... 1010111 @r2_vm
> -vfncvt_x_f_v 100010 . ..... 10001 001 ..... 1010111 @r2_vm
> -vfncvt_f_xu_v 100010 . ..... 10010 001 ..... 1010111 @r2_vm
> -vfncvt_f_x_v 100010 . ..... 10011 001 ..... 1010111 @r2_vm
> -vfncvt_f_f_v 100010 . ..... 10100 001 ..... 1010111 @r2_vm
> +
> +vfncvt_xu_f_w 010010 . ..... 10000 001 ..... 1010111 @r2_vm
> +vfncvt_x_f_w 010010 . ..... 10001 001 ..... 1010111 @r2_vm
> +vfncvt_f_xu_w 010010 . ..... 10010 001 ..... 1010111 @r2_vm
> +vfncvt_f_x_w 010010 . ..... 10011 001 ..... 1010111 @r2_vm
> +vfncvt_f_f_w 010010 . ..... 10100 001 ..... 1010111 @r2_vm
> +vfncvt_rod_f_f_w 010010 . ..... 10101 001 ..... 1010111 @r2_vm
> +vfncvt_rtz_xu_f_w 010010 . ..... 10110 001 ..... 1010111 @r2_vm
> +vfncvt_rtz_x_f_w 010010 . ..... 10111 001 ..... 1010111 @r2_vm
> +
> vredsum_vs 000000 . ..... ..... 010 ..... 1010111 @r_vm
> vredand_vs 000001 . ..... ..... 010 ..... 1010111 @r_vm
> vredor_vs 000010 . ..... ..... 010 ..... 1010111 @r_vm
> diff --git a/target/riscv/insn_trans/trans_rvv.c.inc
> b/target/riscv/insn_trans/trans_rvv.c.inc
> index 63d33b22068..92a23b52e49 100644
> --- a/target/riscv/insn_trans/trans_rvv.c.inc
> +++ b/target/riscv/insn_trans/trans_rvv.c.inc
> @@ -2622,17 +2622,17 @@ static bool opfv_narrow_check(DisasContext *s,
> arg_rmr *a)
> vext_check_sd(s, a->rd, a->rs2, a->vm);
> }
>
> -#define GEN_OPFV_NARROW_TRANS(NAME) \
> +#define GEN_OPFV_NARROW_TRANS(NAME, HELPER, FRM) \
> static bool trans_##NAME(DisasContext *s, arg_rmr *a) \
> { \
> if (opfv_narrow_check(s, a)) { \
> uint32_t data = 0; \
> static gen_helper_gvec_3_ptr * const fns[2] = { \
> - gen_helper_##NAME##_h, \
> - gen_helper_##NAME##_w, \
> + gen_helper_##HELPER##_h, \
> + gen_helper_##HELPER##_w, \
> }; \
> TCGLabel *over = gen_new_label(); \
> - gen_set_rm(s, RISCV_FRM_DYN); \
> + gen_set_rm(s, FRM); \
> tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_vl, 0, over); \
> \
> data = FIELD_DP32(data, VDATA, VM, a->vm); \
> @@ -2648,11 +2648,52 @@ static bool trans_##NAME(DisasContext *s, arg_rmr *a)
> \
> return false; \
> }
>
> -GEN_OPFV_NARROW_TRANS(vfncvt_xu_f_v)
> -GEN_OPFV_NARROW_TRANS(vfncvt_x_f_v)
> -GEN_OPFV_NARROW_TRANS(vfncvt_f_xu_v)
> -GEN_OPFV_NARROW_TRANS(vfncvt_f_x_v)
> -GEN_OPFV_NARROW_TRANS(vfncvt_f_f_v)
> +GEN_OPFV_NARROW_TRANS(vfncvt_f_xu_w, vfncvt_f_xu_w, RISCV_FRM_DYN)
> +GEN_OPFV_NARROW_TRANS(vfncvt_f_x_w, vfncvt_f_x_w, RISCV_FRM_DYN)
> +GEN_OPFV_NARROW_TRANS(vfncvt_f_f_w, vfncvt_f_f_w, RISCV_FRM_DYN)
> +/* Reuse the helper function from vfncvt.f.f.w */
> +GEN_OPFV_NARROW_TRANS(vfncvt_rod_f_f_w, vfncvt_f_f_w, RISCV_FRM_ROD)
> +
> +static bool opxfv_narrow_check(DisasContext *s, arg_rmr *a)
> +{
> + return require_rvv(s) &&
> + require_scale_rvf(s) &&
> + vext_check_isa_ill(s) &&
> + /* OPFV narrowing instructions ignore vs1 check */
> + vext_check_sd(s, a->rd, a->rs2, a->vm);
> +}
> +
> +#define GEN_OPXFV_NARROW_TRANS(NAME, HELPER, FRM) \
> +static bool trans_##NAME(DisasContext *s, arg_rmr *a) \
> +{ \
> + if (opxfv_narrow_check(s, a)) { \
> + uint32_t data = 0; \
> + static gen_helper_gvec_3_ptr * const fns[3] = { \
> + gen_helper_##HELPER##_b, \
> + gen_helper_##HELPER##_h, \
> + gen_helper_##HELPER##_w, \
> + }; \
> + TCGLabel *over = gen_new_label(); \
> + gen_set_rm(s, FRM); \
> + tcg_gen_brcondi_tl(TCG_COND_EQ, cpu_vl, 0, over); \
> + \
> + data = FIELD_DP32(data, VDATA, VM, a->vm); \
> + tcg_gen_gvec_3_ptr(vreg_ofs(s, a->rd), vreg_ofs(s, 0), \
> + vreg_ofs(s, a->rs2), cpu_env, \
> + s->vlen / 8, s->vlen / 8, data, \
> + fns[s->sew]); \
> + mark_vs_dirty(s); \
> + gen_set_label(over); \
> + return true; \
> + } \
> + return false; \
> +}
> +
> +GEN_OPXFV_NARROW_TRANS(vfncvt_xu_f_w, vfncvt_xu_f_w, RISCV_FRM_DYN)
> +GEN_OPXFV_NARROW_TRANS(vfncvt_x_f_w, vfncvt_x_f_w, RISCV_FRM_DYN)
> +/* Reuse the helper functions from vfncvt.xu.f.w and vfncvt.x.f.w */
> +GEN_OPXFV_NARROW_TRANS(vfncvt_rtz_xu_f_w, vfncvt_xu_f_w, RISCV_FRM_RTZ)
> +GEN_OPXFV_NARROW_TRANS(vfncvt_rtz_x_f_w, vfncvt_x_f_w, RISCV_FRM_RTZ)
>
> /*
> *** Vector Reduction Operations
> diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
> index dda8e2ab5a2..37b2451bd05 100644
> --- a/target/riscv/vector_helper.c
> +++ b/target/riscv/vector_helper.c
> @@ -3820,31 +3820,36 @@ GEN_VEXT_V_ENV(vfwcvt_f_f_v_w, 4, 8)
>
> /* Narrowing Floating-Point/Integer Type-Convert Instructions */
> /* (TD, T2, TX2) */
> +#define NOP_UU_B uint8_t, uint16_t, uint32_t
> #define NOP_UU_H uint16_t, uint32_t, uint32_t
> #define NOP_UU_W uint32_t, uint64_t, uint64_t
> /* vfncvt.xu.f.v vd, vs2, vm # Convert float to unsigned integer. */
> -RVVCALL(OPFVV1, vfncvt_xu_f_v_h, NOP_UU_H, H2, H4, float32_to_uint16)
> -RVVCALL(OPFVV1, vfncvt_xu_f_v_w, NOP_UU_W, H4, H8, float64_to_uint32)
> -GEN_VEXT_V_ENV(vfncvt_xu_f_v_h, 2, 2)
> -GEN_VEXT_V_ENV(vfncvt_xu_f_v_w, 4, 4)
> +RVVCALL(OPFVV1, vfncvt_xu_f_w_b, NOP_UU_B, H1, H2, float16_to_uint8)
> +RVVCALL(OPFVV1, vfncvt_xu_f_w_h, NOP_UU_H, H2, H4, float32_to_uint16)
> +RVVCALL(OPFVV1, vfncvt_xu_f_w_w, NOP_UU_W, H4, H8, float64_to_uint32)
> +GEN_VEXT_V_ENV(vfncvt_xu_f_w_b, 1, 1)
> +GEN_VEXT_V_ENV(vfncvt_xu_f_w_h, 2, 2)
> +GEN_VEXT_V_ENV(vfncvt_xu_f_w_w, 4, 4)
>
> /* vfncvt.x.f.v vd, vs2, vm # Convert double-width float to signed integer.
> */
> -RVVCALL(OPFVV1, vfncvt_x_f_v_h, NOP_UU_H, H2, H4, float32_to_int16)
> -RVVCALL(OPFVV1, vfncvt_x_f_v_w, NOP_UU_W, H4, H8, float64_to_int32)
> -GEN_VEXT_V_ENV(vfncvt_x_f_v_h, 2, 2)
> -GEN_VEXT_V_ENV(vfncvt_x_f_v_w, 4, 4)
> +RVVCALL(OPFVV1, vfncvt_x_f_w_b, NOP_UU_B, H1, H2, float16_to_int8)
> +RVVCALL(OPFVV1, vfncvt_x_f_w_h, NOP_UU_H, H2, H4, float32_to_int16)
> +RVVCALL(OPFVV1, vfncvt_x_f_w_w, NOP_UU_W, H4, H8, float64_to_int32)
> +GEN_VEXT_V_ENV(vfncvt_x_f_w_b, 1, 1)
> +GEN_VEXT_V_ENV(vfncvt_x_f_w_h, 2, 2)
> +GEN_VEXT_V_ENV(vfncvt_x_f_w_w, 4, 4)
>
> /* vfncvt.f.xu.v vd, vs2, vm # Convert double-width unsigned integer to
> float */
> -RVVCALL(OPFVV1, vfncvt_f_xu_v_h, NOP_UU_H, H2, H4, uint32_to_float16)
> -RVVCALL(OPFVV1, vfncvt_f_xu_v_w, NOP_UU_W, H4, H8, uint64_to_float32)
> -GEN_VEXT_V_ENV(vfncvt_f_xu_v_h, 2, 2)
> -GEN_VEXT_V_ENV(vfncvt_f_xu_v_w, 4, 4)
> +RVVCALL(OPFVV1, vfncvt_f_xu_w_h, NOP_UU_H, H2, H4, uint32_to_float16)
> +RVVCALL(OPFVV1, vfncvt_f_xu_w_w, NOP_UU_W, H4, H8, uint64_to_float32)
> +GEN_VEXT_V_ENV(vfncvt_f_xu_w_h, 2, 2)
> +GEN_VEXT_V_ENV(vfncvt_f_xu_w_w, 4, 4)
>
> /* vfncvt.f.x.v vd, vs2, vm # Convert double-width integer to float. */
> -RVVCALL(OPFVV1, vfncvt_f_x_v_h, NOP_UU_H, H2, H4, int32_to_float16)
> -RVVCALL(OPFVV1, vfncvt_f_x_v_w, NOP_UU_W, H4, H8, int64_to_float32)
> -GEN_VEXT_V_ENV(vfncvt_f_x_v_h, 2, 2)
> -GEN_VEXT_V_ENV(vfncvt_f_x_v_w, 4, 4)
> +RVVCALL(OPFVV1, vfncvt_f_x_w_h, NOP_UU_H, H2, H4, int32_to_float16)
> +RVVCALL(OPFVV1, vfncvt_f_x_w_w, NOP_UU_W, H4, H8, int64_to_float32)
> +GEN_VEXT_V_ENV(vfncvt_f_x_w_h, 2, 2)
> +GEN_VEXT_V_ENV(vfncvt_f_x_w_w, 4, 4)
>
> /* vfncvt.f.f.v vd, vs2, vm # Convert double float to single-width float. */
> static uint16_t vfncvtffv16(uint32_t a, float_status *s)
> @@ -3852,10 +3857,10 @@ static uint16_t vfncvtffv16(uint32_t a, float_status
> *s)
> return float32_to_float16(a, true, s);
> }
>
> -RVVCALL(OPFVV1, vfncvt_f_f_v_h, NOP_UU_H, H2, H4, vfncvtffv16)
> -RVVCALL(OPFVV1, vfncvt_f_f_v_w, NOP_UU_W, H4, H8, float64_to_float32)
> -GEN_VEXT_V_ENV(vfncvt_f_f_v_h, 2, 2)
> -GEN_VEXT_V_ENV(vfncvt_f_f_v_w, 4, 4)
> +RVVCALL(OPFVV1, vfncvt_f_f_w_h, NOP_UU_H, H2, H4, vfncvtffv16)
> +RVVCALL(OPFVV1, vfncvt_f_f_w_w, NOP_UU_W, H4, H8, float64_to_float32)
> +GEN_VEXT_V_ENV(vfncvt_f_f_w_h, 2, 2)
> +GEN_VEXT_V_ENV(vfncvt_f_f_w_w, 4, 4)
>
> /*
> *** Vector Reduction Operations
> --
> 2.25.1
>
>
- [PATCH v8 58/78] target/riscv: rvv-1.0: remove integer extract instruction, (continued)
- [PATCH v8 58/78] target/riscv: rvv-1.0: remove integer extract instruction, frank . chang, 2021/10/15
- [PATCH v8 59/78] target/riscv: rvv-1.0: floating-point min/max instructions, frank . chang, 2021/10/15
- [PATCH v8 60/78] target/riscv: introduce floating-point rounding mode enum, frank . chang, 2021/10/15
- [PATCH v8 61/78] target/riscv: rvv-1.0: floating-point/integer type-convert instructions, frank . chang, 2021/10/15
- [PATCH v8 62/78] target/riscv: rvv-1.0: widening floating-point/integer type-convert, frank . chang, 2021/10/15
- [PATCH v8 63/78] target/riscv: add "set round to odd" rounding mode helper function, frank . chang, 2021/10/15
- [PATCH v8 64/78] target/riscv: rvv-1.0: narrowing floating-point/integer type-convert, frank . chang, 2021/10/15
- Re: [PATCH v8 64/78] target/riscv: rvv-1.0: narrowing floating-point/integer type-convert,
Alistair Francis <=
- [PATCH v8 65/78] target/riscv: rvv-1.0: relax RV_VLEN_MAX to 1024-bits, frank . chang, 2021/10/15
- [PATCH v8 66/78] target/riscv: rvv-1.0: implement vstart CSR, frank . chang, 2021/10/15
- [PATCH v8 67/78] target/riscv: rvv-1.0: trigger illegal instruction exception if frm is not valid, frank . chang, 2021/10/15
- [PATCH v8 68/78] target/riscv: rvv-1.0: set mstatus.SD bit when writing vector CSRs, frank . chang, 2021/10/15
- [PATCH v8 69/78] target/riscv: gdb: support vector registers for rv64 & rv32, frank . chang, 2021/10/15
- [PATCH v8 70/78] target/riscv: rvv-1.0: floating-point reciprocal square-root estimate instruction, frank . chang, 2021/10/15
- [PATCH v8 71/78] target/riscv: rvv-1.0: floating-point reciprocal estimate instruction, frank . chang, 2021/10/15
- [PATCH v8 72/78] target/riscv: set mstatus.SD bit when writing fp CSRs, frank . chang, 2021/10/15