one or two 16x16 multiply as operands for an add/subtract operation with another 64-bit operand.
Signed-off-by: LIU Zhiwei <zhiwei_...@c-sky.com> --- target/riscv/helper.h | 11 ++ target/riscv/insn32.decode | 11 ++ target/riscv/insn_trans/trans_rvp.c.inc | 12 ++ target/riscv/packed_helper.c | 151 ++++++++++++++++++++++++ 4 files changed, 185 insertions(+) diff --git a/target/riscv/helper.h b/target/riscv/helper.h index c3c086bed0..87a0779842 100644 --- a/target/riscv/helper.h +++ b/target/riscv/helper.h @@ -1350,3 +1350,14 @@ DEF_HELPER_4(kmar64, i64, env, tl, tl, i64) DEF_HELPER_4(kmsr64, i64, env, tl, tl, i64) DEF_HELPER_4(ukmar64, i64, env, tl, tl, i64) DEF_HELPER_4(ukmsr64, i64, env, tl, tl, i64) + +DEF_HELPER_4(smalbb, i64, env, tl, tl, i64) +DEF_HELPER_4(smalbt, i64, env, tl, tl, i64) +DEF_HELPER_4(smaltt, i64, env, tl, tl, i64) +DEF_HELPER_4(smalda, i64, env, tl, tl, i64) +DEF_HELPER_4(smalxda, i64, env, tl, tl, i64) +DEF_HELPER_4(smalds, i64, env, tl, tl, i64) +DEF_HELPER_4(smalxds, i64, env, tl, tl, i64) +DEF_HELPER_4(smaldrs, i64, env, tl, tl, i64) +DEF_HELPER_4(smslda, i64, env, tl, tl, i64) +DEF_HELPER_4(smslxda, i64, env, tl, tl, i64) diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode index 5d123bbb97..d1668b34cb 100644 --- a/target/riscv/insn32.decode +++ b/target/riscv/insn32.decode @@ -953,3 +953,14 @@ kmar64 1001010 ..... ..... 001 ..... 1110111 @r kmsr64 1001011 ..... ..... 001 ..... 1110111 @r ukmar64 1011010 ..... ..... 001 ..... 1110111 @r ukmsr64 1011011 ..... ..... 001 ..... 1110111 @r + +smalbb 1000100 ..... ..... 001 ..... 1110111 @r +smalbt 1001100 ..... ..... 001 ..... 1110111 @r +smaltt 1010100 ..... ..... 001 ..... 1110111 @r +smalda 1000110 ..... ..... 001 ..... 1110111 @r +smalxda 1001110 ..... ..... 001 ..... 1110111 @r +smalds 1000101 ..... ..... 001 ..... 1110111 @r +smaldrs 1001101 ..... ..... 001 ..... 1110111 @r +smalxds 1010101 ..... ..... 001 ..... 1110111 @r +smslda 1010110 ..... ..... 001 ..... 1110111 @r +smslxda 1011110 ..... ..... 001 ..... 1110111 @r diff --git a/target/riscv/insn_trans/trans_rvp.c.inc b/target/riscv/insn_trans/trans_rvp.c.inc index 63b6810227..7c91bdc888 100644 --- a/target/riscv/insn_trans/trans_rvp.c.inc +++ b/target/riscv/insn_trans/trans_rvp.c.inc @@ -657,3 +657,15 @@ GEN_RVP_R_D64_ACC_OOL(kmar64); GEN_RVP_R_D64_ACC_OOL(kmsr64); GEN_RVP_R_D64_ACC_OOL(ukmar64); GEN_RVP_R_D64_ACC_OOL(ukmsr64); + +/* Signed 16-bit Multiply with 64-bit Add/Subtract Instructions */ +GEN_RVP_R_D64_ACC_OOL(smalbb); +GEN_RVP_R_D64_ACC_OOL(smalbt); +GEN_RVP_R_D64_ACC_OOL(smaltt); +GEN_RVP_R_D64_ACC_OOL(smalda); +GEN_RVP_R_D64_ACC_OOL(smalxda); +GEN_RVP_R_D64_ACC_OOL(smalds); +GEN_RVP_R_D64_ACC_OOL(smaldrs); +GEN_RVP_R_D64_ACC_OOL(smalxds); +GEN_RVP_R_D64_ACC_OOL(smslda); +GEN_RVP_R_D64_ACC_OOL(smslxda); diff --git a/target/riscv/packed_helper.c b/target/riscv/packed_helper.c index 59a06c604d..3330a2ecec 100644 --- a/target/riscv/packed_helper.c +++ b/target/riscv/packed_helper.c @@ -2375,3 +2375,154 @@ static inline void do_ukmsr64(CPURISCVState *env, void *vd, void *va, } RVPR64_ACC(ukmsr64, 1, 4); + +/* Signed 16-bit Multiply with 64-bit Add/Subtract Instructions */ +static inline void do_smalbb(CPURISCVState *env, void *vd, void *va, + void *vb, void *vc, uint8_t i) +{ + int64_t *d = vd, *c = vc; + int16_t *a = va, *b = vb; + + if (i == 0) { + *d = *c; + } + + *d += (int64_t)a[H2(i)] * b[H2(i)]; +} + +RVPR64_ACC(smalbb, 2, 2); + +static inline void do_smalbt(CPURISCVState *env, void *vd, void *va, + void *vb, void *vc, uint8_t i) +{ + int64_t *d = vd, *c = vc; + int16_t *a = va, *b = vb; + + if (i == 0) { + *d = *c; + } + + *d += (int64_t)a[H2(i)] * b[H2(i + 1)]; +} + +RVPR64_ACC(smalbt, 2, 2); + +static inline void do_smaltt(CPURISCVState *env, void *vd, void *va, + void *vb, void *vc, uint8_t i) +{ + int64_t *d = vd, *c = vc; + int16_t *a = va, *b = vb; + + if (i == 0) { + *d = *c; + } + + *d += (int64_t)a[H2(i + 1)] * b[H2(i + 1)]; +} + +RVPR64_ACC(smaltt, 2, 2); + +static inline void do_smalda(CPURISCVState *env, void *vd, void *va, + void *vb, void *vc, uint8_t i) +{ + int64_t *d = vd, *c = vc; + int16_t *a = va, *b = vb; + + if (i == 0) { + *d = *c; + } + + *d += (int64_t)a[H2(i)] * b[H2(i)] + (int64_t)a[H2(i + 1)] * b[H2(i + 1)]; +} + +RVPR64_ACC(smalda, 2, 2); + +static inline void do_smalxda(CPURISCVState *env, void *vd, void *va, + void *vb, void *vc, uint8_t i) +{ + int64_t *d = vd, *c = vc; + int16_t *a = va, *b = vb; + + if (i == 0) { + *d = *c; + } + + *d += (int64_t)a[H2(i)] * b[H2(i + 1)] + (int64_t)a[H2(i + 1)] * b[H2(i)]; +} + +RVPR64_ACC(smalxda, 2, 2); + +static inline void do_smalds(CPURISCVState *env, void *vd, void *va, + void *vb, void *vc, uint8_t i) +{ + int64_t *d = vd, *c = vc; + int16_t *a = va, *b = vb; + + if (i == 0) { + *d = *c; + } + + *d += (int64_t)a[H2(i + 1)] * b[H2(i + 1)] - (int64_t)a[H2(i)] * b[H2(i)]; +} + +RVPR64_ACC(smalds, 2, 2); + +static inline void do_smaldrs(CPURISCVState *env, void *vd, void *va, + void *vb, void *vc, uint8_t i) +{ + int64_t *d = vd, *c = vc; + int16_t *a = va, *b = vb; + + if (i == 0) { + *d = *c; + } + + *d += (int64_t)a[H2(i)] * b[H2(i)] - (int64_t)a[H2(i + 1)] * b[H2(i + 1)]; +} + +RVPR64_ACC(smaldrs, 2, 2); + +static inline void do_smalxds(CPURISCVState *env, void *vd, void *va, + void *vb, void *vc, uint8_t i) +{ + int64_t *d = vd, *c = vc; + int16_t *a = va, *b = vb; + + if (i == 0) { + *d = *c; + } + + *d += (int64_t)a[H2(i + 1)] * b[H2(i)] - (int64_t)a[H2(i)] * b[H2(i + 1)]; +} + +RVPR64_ACC(smalxds, 2, 2); + +static inline void do_smslda(CPURISCVState *env, void *vd, void *va, + void *vb, void *vc, uint8_t i) +{ + int64_t *d = vd, *c = vc; + int16_t *a = va, *b = vb; + + if (i == 0) { + *d = *c; + } + + *d -= (int64_t)a[H2(i)] * b[H2(i)] + (int64_t)a[H2(i + 1)] * b[H2(i + 1)]; +} + +RVPR64_ACC(smslda, 2, 2); + +static inline void do_smslxda(CPURISCVState *env, void *vd, void *va, + void *vb, void *vc, uint8_t i) +{ + int64_t *d = vd, *c = vc; + int16_t *a = va, *b = vb; + + if (i == 0) { + *d = *c; + } + + *d -= (int64_t)a[H2(i + 1)] * b[H2(i)] + (int64_t)a[H2(i)] * b[H2(i + 1)]; +} + +RVPR64_ACC(smslxda, 2, 2); -- 2.17.1