Message ID | 20200317150653.9008-19-zhiwei_liu@c-sky.com |
---|---|
State | New |
Headers | show |
Series | target/riscv: support vector extension v0.7.1 | expand |
On Tue, Mar 17, 2020 at 8:43 AM LIU Zhiwei <zhiwei_liu@c-sky.com> wrote: > > Signed-off-by: LIU Zhiwei <zhiwei_liu@c-sky.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Alistair > --- > target/riscv/helper.h | 33 +++++ > target/riscv/insn32.decode | 8 ++ > target/riscv/insn_trans/trans_rvv.inc.c | 10 ++ > target/riscv/vector_helper.c | 156 ++++++++++++++++++++++++ > 4 files changed, 207 insertions(+) > > diff --git a/target/riscv/helper.h b/target/riscv/helper.h > index c7d4ff185a..f42a12eef3 100644 > --- a/target/riscv/helper.h > +++ b/target/riscv/helper.h > @@ -525,3 +525,36 @@ DEF_HELPER_6(vmax_vx_b, void, ptr, ptr, tl, ptr, env, i32) > DEF_HELPER_6(vmax_vx_h, void, ptr, ptr, tl, ptr, env, i32) > DEF_HELPER_6(vmax_vx_w, void, ptr, ptr, tl, ptr, env, i32) > DEF_HELPER_6(vmax_vx_d, void, ptr, ptr, tl, ptr, env, i32) > + > +DEF_HELPER_6(vmul_vv_b, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmul_vv_h, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmul_vv_w, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmul_vv_d, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulh_vv_b, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulh_vv_h, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulh_vv_w, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulh_vv_d, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulhu_vv_b, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulhu_vv_h, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulhu_vv_w, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulhu_vv_d, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulhsu_vv_b, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulhsu_vv_h, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulhsu_vv_w, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmulhsu_vv_d, void, ptr, ptr, ptr, ptr, env, i32) > +DEF_HELPER_6(vmul_vx_b, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmul_vx_h, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmul_vx_w, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmul_vx_d, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulh_vx_b, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulh_vx_h, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulh_vx_w, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulh_vx_d, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulhu_vx_b, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulhu_vx_h, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulhu_vx_w, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulhu_vx_d, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulhsu_vx_b, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulhsu_vx_h, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulhsu_vx_w, void, ptr, ptr, tl, ptr, env, i32) > +DEF_HELPER_6(vmulhsu_vx_d, void, ptr, ptr, tl, ptr, env, i32) > diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode > index aafbdc6be7..abfed469bc 100644 > --- a/target/riscv/insn32.decode > +++ b/target/riscv/insn32.decode > @@ -363,6 +363,14 @@ vmaxu_vv 000110 . ..... ..... 000 ..... 1010111 @r_vm > vmaxu_vx 000110 . ..... ..... 100 ..... 1010111 @r_vm > vmax_vv 000111 . ..... ..... 000 ..... 1010111 @r_vm > vmax_vx 000111 . ..... ..... 100 ..... 1010111 @r_vm > +vmul_vv 100101 . ..... ..... 010 ..... 1010111 @r_vm > +vmul_vx 100101 . ..... ..... 110 ..... 1010111 @r_vm > +vmulh_vv 100111 . ..... ..... 010 ..... 1010111 @r_vm > +vmulh_vx 100111 . ..... ..... 110 ..... 1010111 @r_vm > +vmulhu_vv 100100 . ..... ..... 010 ..... 1010111 @r_vm > +vmulhu_vx 100100 . ..... ..... 110 ..... 1010111 @r_vm > +vmulhsu_vv 100110 . ..... ..... 010 ..... 1010111 @r_vm > +vmulhsu_vx 100110 . ..... ..... 110 ..... 1010111 @r_vm > > vsetvli 0 ........... ..... 111 ..... 1010111 @r2_zimm > vsetvl 1000000 ..... ..... 111 ..... 1010111 @r > diff --git a/target/riscv/insn_trans/trans_rvv.inc.c b/target/riscv/insn_trans/trans_rvv.inc.c > index 53c49ee15c..c276beabd6 100644 > --- a/target/riscv/insn_trans/trans_rvv.inc.c > +++ b/target/riscv/insn_trans/trans_rvv.inc.c > @@ -1452,3 +1452,13 @@ GEN_OPIVX_TRANS(vminu_vx, opivx_check) > GEN_OPIVX_TRANS(vmin_vx, opivx_check) > GEN_OPIVX_TRANS(vmaxu_vx, opivx_check) > GEN_OPIVX_TRANS(vmax_vx, opivx_check) > + > +/* Vector Single-Width Integer Multiply Instructions */ > +GEN_OPIVV_GVEC_TRANS(vmul_vv, mul) > +GEN_OPIVV_TRANS(vmulh_vv, opivv_check) > +GEN_OPIVV_TRANS(vmulhu_vv, opivv_check) > +GEN_OPIVV_TRANS(vmulhsu_vv, opivv_check) > +GEN_OPIVX_GVEC_TRANS(vmul_vx, muls) > +GEN_OPIVX_TRANS(vmulh_vx, opivx_check) > +GEN_OPIVX_TRANS(vmulhu_vx, opivx_check) > +GEN_OPIVX_TRANS(vmulhsu_vx, opivx_check) > diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c > index 32c2760a8a..56ba9a7422 100644 > --- a/target/riscv/vector_helper.c > +++ b/target/riscv/vector_helper.c > @@ -852,6 +852,10 @@ GEN_VEXT_AMO(vamomaxuw_v_w, uint32_t, uint32_t, idx_w, clearl) > #define OP_UUU_H uint16_t, uint16_t, uint16_t, uint16_t, uint16_t > #define OP_UUU_W uint32_t, uint32_t, uint32_t, uint32_t, uint32_t > #define OP_UUU_D uint64_t, uint64_t, uint64_t, uint64_t, uint64_t > +#define OP_SUS_B int8_t, uint8_t, int8_t, uint8_t, int8_t > +#define OP_SUS_H int16_t, uint16_t, int16_t, uint16_t, int16_t > +#define OP_SUS_W int32_t, uint32_t, int32_t, uint32_t, int32_t > +#define OP_SUS_D int64_t, uint64_t, int64_t, uint64_t, int64_t > > /* operation of two vector elements */ > typedef void opivv2_fn(void *vd, void *vs1, void *vs2, int i); > @@ -1585,3 +1589,155 @@ GEN_VEXT_VX(vmax_vx_b, 1, 1, clearb) > GEN_VEXT_VX(vmax_vx_h, 2, 2, clearh) > GEN_VEXT_VX(vmax_vx_w, 4, 4, clearl) > GEN_VEXT_VX(vmax_vx_d, 8, 8, clearq) > + > +/* Vector Single-Width Integer Multiply Instructions */ > +#define DO_MUL(N, M) (N * M) > +RVVCALL(OPIVV2, vmul_vv_b, OP_SSS_B, H1, H1, H1, DO_MUL) > +RVVCALL(OPIVV2, vmul_vv_h, OP_SSS_H, H2, H2, H2, DO_MUL) > +RVVCALL(OPIVV2, vmul_vv_w, OP_SSS_W, H4, H4, H4, DO_MUL) > +RVVCALL(OPIVV2, vmul_vv_d, OP_SSS_D, H8, H8, H8, DO_MUL) > +GEN_VEXT_VV(vmul_vv_b, 1, 1, clearb) > +GEN_VEXT_VV(vmul_vv_h, 2, 2, clearh) > +GEN_VEXT_VV(vmul_vv_w, 4, 4, clearl) > +GEN_VEXT_VV(vmul_vv_d, 8, 8, clearq) > + > +static int8_t do_mulh_b(int8_t s2, int8_t s1) > +{ > + return (int16_t)s2 * (int16_t)s1 >> 8; > +} > + > +static int16_t do_mulh_h(int16_t s2, int16_t s1) > +{ > + return (int32_t)s2 * (int32_t)s1 >> 16; > +} > + > +static int32_t do_mulh_w(int32_t s2, int32_t s1) > +{ > + return (int64_t)s2 * (int64_t)s1 >> 32; > +} > + > +static int64_t do_mulh_d(int64_t s2, int64_t s1) > +{ > + uint64_t hi_64, lo_64; > + > + muls64(&lo_64, &hi_64, s1, s2); > + return hi_64; > +} > + > +static uint8_t do_mulhu_b(uint8_t s2, uint8_t s1) > +{ > + return (uint16_t)s2 * (uint16_t)s1 >> 8; > +} > + > +static uint16_t do_mulhu_h(uint16_t s2, uint16_t s1) > +{ > + return (uint32_t)s2 * (uint32_t)s1 >> 16; > +} > + > +static uint32_t do_mulhu_w(uint32_t s2, uint32_t s1) > +{ > + return (uint64_t)s2 * (uint64_t)s1 >> 32; > +} > + > +static uint64_t do_mulhu_d(uint64_t s2, uint64_t s1) > +{ > + uint64_t hi_64, lo_64; > + > + mulu64(&lo_64, &hi_64, s2, s1); > + return hi_64; > +} > + > +static int8_t do_mulhsu_b(int8_t s2, uint8_t s1) > +{ > + return (int16_t)s2 * (uint16_t)s1 >> 8; > +} > + > +static int16_t do_mulhsu_h(int16_t s2, uint16_t s1) > +{ > + return (int32_t)s2 * (uint32_t)s1 >> 16; > +} > + > +static int32_t do_mulhsu_w(int32_t s2, uint32_t s1) > +{ > + return (int64_t)s2 * (uint64_t)s1 >> 32; > +} > + > +static int64_t do_mulhsu_d(int64_t s2, uint64_t s1) > +{ > + uint64_t hi_64, lo_64, abs_s2 = s2; > + > + if (s2 < 0) { > + abs_s2 = -s2; > + } > + mulu64(&lo_64, &hi_64, abs_s2, s1); > + if (s2 < 0) { > + lo_64 = ~lo_64; > + hi_64 = ~hi_64; > + if (lo_64 == UINT64_MAX) { > + lo_64 = 0; > + hi_64 += 1; > + } else { > + lo_64 += 1; > + } > + } > + > + return hi_64; > +} > + > +RVVCALL(OPIVV2, vmulh_vv_b, OP_SSS_B, H1, H1, H1, do_mulh_b) > +RVVCALL(OPIVV2, vmulh_vv_h, OP_SSS_H, H2, H2, H2, do_mulh_h) > +RVVCALL(OPIVV2, vmulh_vv_w, OP_SSS_W, H4, H4, H4, do_mulh_w) > +RVVCALL(OPIVV2, vmulh_vv_d, OP_SSS_D, H8, H8, H8, do_mulh_d) > +RVVCALL(OPIVV2, vmulhu_vv_b, OP_UUU_B, H1, H1, H1, do_mulhu_b) > +RVVCALL(OPIVV2, vmulhu_vv_h, OP_UUU_H, H2, H2, H2, do_mulhu_h) > +RVVCALL(OPIVV2, vmulhu_vv_w, OP_UUU_W, H4, H4, H4, do_mulhu_w) > +RVVCALL(OPIVV2, vmulhu_vv_d, OP_UUU_D, H8, H8, H8, do_mulhu_d) > +RVVCALL(OPIVV2, vmulhsu_vv_b, OP_SUS_B, H1, H1, H1, do_mulhsu_b) > +RVVCALL(OPIVV2, vmulhsu_vv_h, OP_SUS_H, H2, H2, H2, do_mulhsu_h) > +RVVCALL(OPIVV2, vmulhsu_vv_w, OP_SUS_W, H4, H4, H4, do_mulhsu_w) > +RVVCALL(OPIVV2, vmulhsu_vv_d, OP_SUS_D, H8, H8, H8, do_mulhsu_d) > +GEN_VEXT_VV(vmulh_vv_b, 1, 1, clearb) > +GEN_VEXT_VV(vmulh_vv_h, 2, 2, clearh) > +GEN_VEXT_VV(vmulh_vv_w, 4, 4, clearl) > +GEN_VEXT_VV(vmulh_vv_d, 8, 8, clearq) > +GEN_VEXT_VV(vmulhu_vv_b, 1, 1, clearb) > +GEN_VEXT_VV(vmulhu_vv_h, 2, 2, clearh) > +GEN_VEXT_VV(vmulhu_vv_w, 4, 4, clearl) > +GEN_VEXT_VV(vmulhu_vv_d, 8, 8, clearq) > +GEN_VEXT_VV(vmulhsu_vv_b, 1, 1, clearb) > +GEN_VEXT_VV(vmulhsu_vv_h, 2, 2, clearh) > +GEN_VEXT_VV(vmulhsu_vv_w, 4, 4, clearl) > +GEN_VEXT_VV(vmulhsu_vv_d, 8, 8, clearq) > + > +RVVCALL(OPIVX2, vmul_vx_b, OP_SSS_B, H1, H1, DO_MUL) > +RVVCALL(OPIVX2, vmul_vx_h, OP_SSS_H, H2, H2, DO_MUL) > +RVVCALL(OPIVX2, vmul_vx_w, OP_SSS_W, H4, H4, DO_MUL) > +RVVCALL(OPIVX2, vmul_vx_d, OP_SSS_D, H8, H8, DO_MUL) > +RVVCALL(OPIVX2, vmulh_vx_b, OP_SSS_B, H1, H1, do_mulh_b) > +RVVCALL(OPIVX2, vmulh_vx_h, OP_SSS_H, H2, H2, do_mulh_h) > +RVVCALL(OPIVX2, vmulh_vx_w, OP_SSS_W, H4, H4, do_mulh_w) > +RVVCALL(OPIVX2, vmulh_vx_d, OP_SSS_D, H8, H8, do_mulh_d) > +RVVCALL(OPIVX2, vmulhu_vx_b, OP_UUU_B, H1, H1, do_mulhu_b) > +RVVCALL(OPIVX2, vmulhu_vx_h, OP_UUU_H, H2, H2, do_mulhu_h) > +RVVCALL(OPIVX2, vmulhu_vx_w, OP_UUU_W, H4, H4, do_mulhu_w) > +RVVCALL(OPIVX2, vmulhu_vx_d, OP_UUU_D, H8, H8, do_mulhu_d) > +RVVCALL(OPIVX2, vmulhsu_vx_b, OP_SUS_B, H1, H1, do_mulhsu_b) > +RVVCALL(OPIVX2, vmulhsu_vx_h, OP_SUS_H, H2, H2, do_mulhsu_h) > +RVVCALL(OPIVX2, vmulhsu_vx_w, OP_SUS_W, H4, H4, do_mulhsu_w) > +RVVCALL(OPIVX2, vmulhsu_vx_d, OP_SUS_D, H8, H8, do_mulhsu_d) > +GEN_VEXT_VX(vmul_vx_b, 1, 1, clearb) > +GEN_VEXT_VX(vmul_vx_h, 2, 2, clearh) > +GEN_VEXT_VX(vmul_vx_w, 4, 4, clearl) > +GEN_VEXT_VX(vmul_vx_d, 8, 8, clearq) > +GEN_VEXT_VX(vmulh_vx_b, 1, 1, clearb) > +GEN_VEXT_VX(vmulh_vx_h, 2, 2, clearh) > +GEN_VEXT_VX(vmulh_vx_w, 4, 4, clearl) > +GEN_VEXT_VX(vmulh_vx_d, 8, 8, clearq) > +GEN_VEXT_VX(vmulhu_vx_b, 1, 1, clearb) > +GEN_VEXT_VX(vmulhu_vx_h, 2, 2, clearh) > +GEN_VEXT_VX(vmulhu_vx_w, 4, 4, clearl) > +GEN_VEXT_VX(vmulhu_vx_d, 8, 8, clearq) > +GEN_VEXT_VX(vmulhsu_vx_b, 1, 1, clearb) > +GEN_VEXT_VX(vmulhsu_vx_h, 2, 2, clearh) > +GEN_VEXT_VX(vmulhsu_vx_w, 4, 4, clearl) > +GEN_VEXT_VX(vmulhsu_vx_d, 8, 8, clearq) > -- > 2.23.0 >
On 3/17/20 8:06 AM, LIU Zhiwei wrote: > +static int64_t do_mulhsu_d(int64_t s2, uint64_t s1) > +{ > + uint64_t hi_64, lo_64, abs_s2 = s2; > + > + if (s2 < 0) { > + abs_s2 = -s2; > + } > + mulu64(&lo_64, &hi_64, abs_s2, s1); > + if (s2 < 0) { > + lo_64 = ~lo_64; > + hi_64 = ~hi_64; > + if (lo_64 == UINT64_MAX) { > + lo_64 = 0; > + hi_64 += 1; > + } else { > + lo_64 += 1; > + } > + } > + > + return hi_64; > +} Missed the improvement here. See tcg_gen_mulsu2_i64. Otherwise, Reviewed-by: Richard Henderson <richard.henderson@linaro.org> r~
On 2020/3/28 8:06, Richard Henderson wrote: > On 3/17/20 8:06 AM, LIU Zhiwei wrote: >> +static int64_t do_mulhsu_d(int64_t s2, uint64_t s1) >> +{ >> + uint64_t hi_64, lo_64, abs_s2 = s2; >> + >> + if (s2 < 0) { >> + abs_s2 = -s2; >> + } >> + mulu64(&lo_64, &hi_64, abs_s2, s1); >> + if (s2 < 0) { >> + lo_64 = ~lo_64; >> + hi_64 = ~hi_64; >> + if (lo_64 == UINT64_MAX) { >> + lo_64 = 0; >> + hi_64 += 1; >> + } else { >> + lo_64 += 1; >> + } >> + } >> + >> + return hi_64; >> +} > Missed the improvement here. See tcg_gen_mulsu2_i64. Though I have not gotten the principle, the code in tcg_gen_mulsu2_i64 is much tidier. Thanks for pointing that. Zhiwei > Otherwise, > Reviewed-by: Richard Henderson <richard.henderson@linaro.org> > > > r~
On 3/28/20 8:17 AM, LIU Zhiwei wrote: >> Missed the improvement here. See tcg_gen_mulsu2_i64. > Though I have not gotten the principle, the code in tcg_gen_mulsu2_i64 is much > tidier. Let A = signed operand, B = unsigned operand P = unsigned product If the sign bit A is set, then P is too large. In that case we subtract 2**64 * B to fix that: HI_P -= (A < 0 ? B : 0) where the conditional is computed as (A >> 63) & B. r~
On 2020/3/28 23:47, Richard Henderson wrote: > On 3/28/20 8:17 AM, LIU Zhiwei wrote: >>> Missed the improvement here. See tcg_gen_mulsu2_i64. >> Though I have not gotten the principle, the code in tcg_gen_mulsu2_i64 is much >> tidier. > Let A = signed operand, > B = unsigned operand > P = unsigned product > > If the sign bit A is set, then P is too large. > In that case we subtract 2**64 * B to fix that: > > HI_P -= (A < 0 ? B : 0) > > where the conditional is computed as (A >> 63) & B. I think I get it. LET A = 2 ** 64 - X THEN X = 2 ** 64 - A SIGNED_P = -X * B if (A * B == P) then (2 ** 64 - X) * B == P 2 **64 * B - X * B == P -X *B == P - 2**64*B HI_P -= (A < 0 ? B :0) Zhiwei > > r~
On 2020/3/29 0:13, LIU Zhiwei wrote: > > > On 2020/3/28 23:47, Richard Henderson wrote: >> On 3/28/20 8:17 AM, LIU Zhiwei wrote: >>>> Missed the improvement here. See tcg_gen_mulsu2_i64. >>> Though I have not gotten the principle, the code in >>> tcg_gen_mulsu2_i64 is much >>> tidier. >> Let A = signed operand, >> B = unsigned operand >> P = unsigned product >> >> If the sign bit A is set, then P is too large. >> In that case we subtract 2**64 * B to fix that: >> >> HI_P -= (A < 0 ? B : 0) >> >> where the conditional is computed as (A >> 63) & B. > > I think I get it. > > LET A = 2 ** 64 - X > > THEN > > X = 2 ** 64 - A > SIGNED_P = -X * B > > if (A * B == P) then > > (2 ** 64 - X) * B == P > 2 **64 * B - X * B == P > > -X *B == P - 2**64*B > > HI_P -= (A < 0 ? B :0) > It's confusing here. I paste the clearer code. /* * Let A = signed operand, * B = unsigned operand * P = mulu64(A, B), unsigned product * * LET X = 2 ** 64 - A, 2's complement of A * SP = signed product * THEN * IF A < 0 * SP = -X * B * = -(2 ** 64 - A) * B * = A * B - 2 ** 64 * B * = P - 2 ** 64 * B * ELSE * SP = P * THEN * HI_P -= (A < 0 ? B : 0) */ static int64_t do_mulhsu_d(int64_t s2, uint64_t s1) { uint64_t hi_64, lo_64; mulu64(&lo_64, &hi_64, s2, s1); hi_64 -= s2 < 0 ? s1 : 0; return hi_64; } Zhiwei > Zhiwei >> >> r~ >
diff --git a/target/riscv/helper.h b/target/riscv/helper.h index c7d4ff185a..f42a12eef3 100644 --- a/target/riscv/helper.h +++ b/target/riscv/helper.h @@ -525,3 +525,36 @@ DEF_HELPER_6(vmax_vx_b, void, ptr, ptr, tl, ptr, env, i32) DEF_HELPER_6(vmax_vx_h, void, ptr, ptr, tl, ptr, env, i32) DEF_HELPER_6(vmax_vx_w, void, ptr, ptr, tl, ptr, env, i32) DEF_HELPER_6(vmax_vx_d, void, ptr, ptr, tl, ptr, env, i32) + +DEF_HELPER_6(vmul_vv_b, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmul_vv_h, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmul_vv_w, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmul_vv_d, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulh_vv_b, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulh_vv_h, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulh_vv_w, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulh_vv_d, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulhu_vv_b, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulhu_vv_h, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulhu_vv_w, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulhu_vv_d, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulhsu_vv_b, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulhsu_vv_h, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulhsu_vv_w, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmulhsu_vv_d, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(vmul_vx_b, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmul_vx_h, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmul_vx_w, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmul_vx_d, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulh_vx_b, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulh_vx_h, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulh_vx_w, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulh_vx_d, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulhu_vx_b, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulhu_vx_h, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulhu_vx_w, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulhu_vx_d, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulhsu_vx_b, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulhsu_vx_h, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulhsu_vx_w, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(vmulhsu_vx_d, void, ptr, ptr, tl, ptr, env, i32) diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode index aafbdc6be7..abfed469bc 100644 --- a/target/riscv/insn32.decode +++ b/target/riscv/insn32.decode @@ -363,6 +363,14 @@ vmaxu_vv 000110 . ..... ..... 000 ..... 1010111 @r_vm vmaxu_vx 000110 . ..... ..... 100 ..... 1010111 @r_vm vmax_vv 000111 . ..... ..... 000 ..... 1010111 @r_vm vmax_vx 000111 . ..... ..... 100 ..... 1010111 @r_vm +vmul_vv 100101 . ..... ..... 010 ..... 1010111 @r_vm +vmul_vx 100101 . ..... ..... 110 ..... 1010111 @r_vm +vmulh_vv 100111 . ..... ..... 010 ..... 1010111 @r_vm +vmulh_vx 100111 . ..... ..... 110 ..... 1010111 @r_vm +vmulhu_vv 100100 . ..... ..... 010 ..... 1010111 @r_vm +vmulhu_vx 100100 . ..... ..... 110 ..... 1010111 @r_vm +vmulhsu_vv 100110 . ..... ..... 010 ..... 1010111 @r_vm +vmulhsu_vx 100110 . ..... ..... 110 ..... 1010111 @r_vm vsetvli 0 ........... ..... 111 ..... 1010111 @r2_zimm vsetvl 1000000 ..... ..... 111 ..... 1010111 @r diff --git a/target/riscv/insn_trans/trans_rvv.inc.c b/target/riscv/insn_trans/trans_rvv.inc.c index 53c49ee15c..c276beabd6 100644 --- a/target/riscv/insn_trans/trans_rvv.inc.c +++ b/target/riscv/insn_trans/trans_rvv.inc.c @@ -1452,3 +1452,13 @@ GEN_OPIVX_TRANS(vminu_vx, opivx_check) GEN_OPIVX_TRANS(vmin_vx, opivx_check) GEN_OPIVX_TRANS(vmaxu_vx, opivx_check) GEN_OPIVX_TRANS(vmax_vx, opivx_check) + +/* Vector Single-Width Integer Multiply Instructions */ +GEN_OPIVV_GVEC_TRANS(vmul_vv, mul) +GEN_OPIVV_TRANS(vmulh_vv, opivv_check) +GEN_OPIVV_TRANS(vmulhu_vv, opivv_check) +GEN_OPIVV_TRANS(vmulhsu_vv, opivv_check) +GEN_OPIVX_GVEC_TRANS(vmul_vx, muls) +GEN_OPIVX_TRANS(vmulh_vx, opivx_check) +GEN_OPIVX_TRANS(vmulhu_vx, opivx_check) +GEN_OPIVX_TRANS(vmulhsu_vx, opivx_check) diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c index 32c2760a8a..56ba9a7422 100644 --- a/target/riscv/vector_helper.c +++ b/target/riscv/vector_helper.c @@ -852,6 +852,10 @@ GEN_VEXT_AMO(vamomaxuw_v_w, uint32_t, uint32_t, idx_w, clearl) #define OP_UUU_H uint16_t, uint16_t, uint16_t, uint16_t, uint16_t #define OP_UUU_W uint32_t, uint32_t, uint32_t, uint32_t, uint32_t #define OP_UUU_D uint64_t, uint64_t, uint64_t, uint64_t, uint64_t +#define OP_SUS_B int8_t, uint8_t, int8_t, uint8_t, int8_t +#define OP_SUS_H int16_t, uint16_t, int16_t, uint16_t, int16_t +#define OP_SUS_W int32_t, uint32_t, int32_t, uint32_t, int32_t +#define OP_SUS_D int64_t, uint64_t, int64_t, uint64_t, int64_t /* operation of two vector elements */ typedef void opivv2_fn(void *vd, void *vs1, void *vs2, int i); @@ -1585,3 +1589,155 @@ GEN_VEXT_VX(vmax_vx_b, 1, 1, clearb) GEN_VEXT_VX(vmax_vx_h, 2, 2, clearh) GEN_VEXT_VX(vmax_vx_w, 4, 4, clearl) GEN_VEXT_VX(vmax_vx_d, 8, 8, clearq) + +/* Vector Single-Width Integer Multiply Instructions */ +#define DO_MUL(N, M) (N * M) +RVVCALL(OPIVV2, vmul_vv_b, OP_SSS_B, H1, H1, H1, DO_MUL) +RVVCALL(OPIVV2, vmul_vv_h, OP_SSS_H, H2, H2, H2, DO_MUL) +RVVCALL(OPIVV2, vmul_vv_w, OP_SSS_W, H4, H4, H4, DO_MUL) +RVVCALL(OPIVV2, vmul_vv_d, OP_SSS_D, H8, H8, H8, DO_MUL) +GEN_VEXT_VV(vmul_vv_b, 1, 1, clearb) +GEN_VEXT_VV(vmul_vv_h, 2, 2, clearh) +GEN_VEXT_VV(vmul_vv_w, 4, 4, clearl) +GEN_VEXT_VV(vmul_vv_d, 8, 8, clearq) + +static int8_t do_mulh_b(int8_t s2, int8_t s1) +{ + return (int16_t)s2 * (int16_t)s1 >> 8; +} + +static int16_t do_mulh_h(int16_t s2, int16_t s1) +{ + return (int32_t)s2 * (int32_t)s1 >> 16; +} + +static int32_t do_mulh_w(int32_t s2, int32_t s1) +{ + return (int64_t)s2 * (int64_t)s1 >> 32; +} + +static int64_t do_mulh_d(int64_t s2, int64_t s1) +{ + uint64_t hi_64, lo_64; + + muls64(&lo_64, &hi_64, s1, s2); + return hi_64; +} + +static uint8_t do_mulhu_b(uint8_t s2, uint8_t s1) +{ + return (uint16_t)s2 * (uint16_t)s1 >> 8; +} + +static uint16_t do_mulhu_h(uint16_t s2, uint16_t s1) +{ + return (uint32_t)s2 * (uint32_t)s1 >> 16; +} + +static uint32_t do_mulhu_w(uint32_t s2, uint32_t s1) +{ + return (uint64_t)s2 * (uint64_t)s1 >> 32; +} + +static uint64_t do_mulhu_d(uint64_t s2, uint64_t s1) +{ + uint64_t hi_64, lo_64; + + mulu64(&lo_64, &hi_64, s2, s1); + return hi_64; +} + +static int8_t do_mulhsu_b(int8_t s2, uint8_t s1) +{ + return (int16_t)s2 * (uint16_t)s1 >> 8; +} + +static int16_t do_mulhsu_h(int16_t s2, uint16_t s1) +{ + return (int32_t)s2 * (uint32_t)s1 >> 16; +} + +static int32_t do_mulhsu_w(int32_t s2, uint32_t s1) +{ + return (int64_t)s2 * (uint64_t)s1 >> 32; +} + +static int64_t do_mulhsu_d(int64_t s2, uint64_t s1) +{ + uint64_t hi_64, lo_64, abs_s2 = s2; + + if (s2 < 0) { + abs_s2 = -s2; + } + mulu64(&lo_64, &hi_64, abs_s2, s1); + if (s2 < 0) { + lo_64 = ~lo_64; + hi_64 = ~hi_64; + if (lo_64 == UINT64_MAX) { + lo_64 = 0; + hi_64 += 1; + } else { + lo_64 += 1; + } + } + + return hi_64; +} + +RVVCALL(OPIVV2, vmulh_vv_b, OP_SSS_B, H1, H1, H1, do_mulh_b) +RVVCALL(OPIVV2, vmulh_vv_h, OP_SSS_H, H2, H2, H2, do_mulh_h) +RVVCALL(OPIVV2, vmulh_vv_w, OP_SSS_W, H4, H4, H4, do_mulh_w) +RVVCALL(OPIVV2, vmulh_vv_d, OP_SSS_D, H8, H8, H8, do_mulh_d) +RVVCALL(OPIVV2, vmulhu_vv_b, OP_UUU_B, H1, H1, H1, do_mulhu_b) +RVVCALL(OPIVV2, vmulhu_vv_h, OP_UUU_H, H2, H2, H2, do_mulhu_h) +RVVCALL(OPIVV2, vmulhu_vv_w, OP_UUU_W, H4, H4, H4, do_mulhu_w) +RVVCALL(OPIVV2, vmulhu_vv_d, OP_UUU_D, H8, H8, H8, do_mulhu_d) +RVVCALL(OPIVV2, vmulhsu_vv_b, OP_SUS_B, H1, H1, H1, do_mulhsu_b) +RVVCALL(OPIVV2, vmulhsu_vv_h, OP_SUS_H, H2, H2, H2, do_mulhsu_h) +RVVCALL(OPIVV2, vmulhsu_vv_w, OP_SUS_W, H4, H4, H4, do_mulhsu_w) +RVVCALL(OPIVV2, vmulhsu_vv_d, OP_SUS_D, H8, H8, H8, do_mulhsu_d) +GEN_VEXT_VV(vmulh_vv_b, 1, 1, clearb) +GEN_VEXT_VV(vmulh_vv_h, 2, 2, clearh) +GEN_VEXT_VV(vmulh_vv_w, 4, 4, clearl) +GEN_VEXT_VV(vmulh_vv_d, 8, 8, clearq) +GEN_VEXT_VV(vmulhu_vv_b, 1, 1, clearb) +GEN_VEXT_VV(vmulhu_vv_h, 2, 2, clearh) +GEN_VEXT_VV(vmulhu_vv_w, 4, 4, clearl) +GEN_VEXT_VV(vmulhu_vv_d, 8, 8, clearq) +GEN_VEXT_VV(vmulhsu_vv_b, 1, 1, clearb) +GEN_VEXT_VV(vmulhsu_vv_h, 2, 2, clearh) +GEN_VEXT_VV(vmulhsu_vv_w, 4, 4, clearl) +GEN_VEXT_VV(vmulhsu_vv_d, 8, 8, clearq) + +RVVCALL(OPIVX2, vmul_vx_b, OP_SSS_B, H1, H1, DO_MUL) +RVVCALL(OPIVX2, vmul_vx_h, OP_SSS_H, H2, H2, DO_MUL) +RVVCALL(OPIVX2, vmul_vx_w, OP_SSS_W, H4, H4, DO_MUL) +RVVCALL(OPIVX2, vmul_vx_d, OP_SSS_D, H8, H8, DO_MUL) +RVVCALL(OPIVX2, vmulh_vx_b, OP_SSS_B, H1, H1, do_mulh_b) +RVVCALL(OPIVX2, vmulh_vx_h, OP_SSS_H, H2, H2, do_mulh_h) +RVVCALL(OPIVX2, vmulh_vx_w, OP_SSS_W, H4, H4, do_mulh_w) +RVVCALL(OPIVX2, vmulh_vx_d, OP_SSS_D, H8, H8, do_mulh_d) +RVVCALL(OPIVX2, vmulhu_vx_b, OP_UUU_B, H1, H1, do_mulhu_b) +RVVCALL(OPIVX2, vmulhu_vx_h, OP_UUU_H, H2, H2, do_mulhu_h) +RVVCALL(OPIVX2, vmulhu_vx_w, OP_UUU_W, H4, H4, do_mulhu_w) +RVVCALL(OPIVX2, vmulhu_vx_d, OP_UUU_D, H8, H8, do_mulhu_d) +RVVCALL(OPIVX2, vmulhsu_vx_b, OP_SUS_B, H1, H1, do_mulhsu_b) +RVVCALL(OPIVX2, vmulhsu_vx_h, OP_SUS_H, H2, H2, do_mulhsu_h) +RVVCALL(OPIVX2, vmulhsu_vx_w, OP_SUS_W, H4, H4, do_mulhsu_w) +RVVCALL(OPIVX2, vmulhsu_vx_d, OP_SUS_D, H8, H8, do_mulhsu_d) +GEN_VEXT_VX(vmul_vx_b, 1, 1, clearb) +GEN_VEXT_VX(vmul_vx_h, 2, 2, clearh) +GEN_VEXT_VX(vmul_vx_w, 4, 4, clearl) +GEN_VEXT_VX(vmul_vx_d, 8, 8, clearq) +GEN_VEXT_VX(vmulh_vx_b, 1, 1, clearb) +GEN_VEXT_VX(vmulh_vx_h, 2, 2, clearh) +GEN_VEXT_VX(vmulh_vx_w, 4, 4, clearl) +GEN_VEXT_VX(vmulh_vx_d, 8, 8, clearq) +GEN_VEXT_VX(vmulhu_vx_b, 1, 1, clearb) +GEN_VEXT_VX(vmulhu_vx_h, 2, 2, clearh) +GEN_VEXT_VX(vmulhu_vx_w, 4, 4, clearl) +GEN_VEXT_VX(vmulhu_vx_d, 8, 8, clearq) +GEN_VEXT_VX(vmulhsu_vx_b, 1, 1, clearb) +GEN_VEXT_VX(vmulhsu_vx_h, 2, 2, clearh) +GEN_VEXT_VX(vmulhsu_vx_w, 4, 4, clearl) +GEN_VEXT_VX(vmulhsu_vx_d, 8, 8, clearq)
Signed-off-by: LIU Zhiwei <zhiwei_liu@c-sky.com> --- target/riscv/helper.h | 33 +++++ target/riscv/insn32.decode | 8 ++ target/riscv/insn_trans/trans_rvv.inc.c | 10 ++ target/riscv/vector_helper.c | 156 ++++++++++++++++++++++++ 4 files changed, 207 insertions(+)