diff options
Diffstat (limited to 'target/arm/mve_helper.c')
| -rw-r--r-- | target/arm/mve_helper.c | 25 |
1 files changed, 25 insertions, 0 deletions
diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index 91139d2..20c7145 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -518,6 +518,24 @@ static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) #define DO_UQSUB_H(n, m, s) do_sat_bhw((int64_t)n - m, 0, UINT16_MAX, s) #define DO_UQSUB_W(n, m, s) do_sat_bhw((int64_t)n - m, 0, UINT32_MAX, s) +/* + * For QDMULH and QRDMULH we simplify "double and shift by esize" into + * "shift by esize-1", adjusting the QRDMULH rounding constant to match. + */ +#define DO_QDMULH_B(n, m, s) do_sat_bhw(((int64_t)n * m) >> 7, \ + INT8_MIN, INT8_MAX, s) +#define DO_QDMULH_H(n, m, s) do_sat_bhw(((int64_t)n * m) >> 15, \ + INT16_MIN, INT16_MAX, s) +#define DO_QDMULH_W(n, m, s) do_sat_bhw(((int64_t)n * m) >> 31, \ + INT32_MIN, INT32_MAX, s) + +#define DO_QRDMULH_B(n, m, s) do_sat_bhw(((int64_t)n * m + (1 << 6)) >> 7, \ + INT8_MIN, INT8_MAX, s) +#define DO_QRDMULH_H(n, m, s) do_sat_bhw(((int64_t)n * m + (1 << 14)) >> 15, \ + INT16_MIN, INT16_MAX, s) +#define DO_QRDMULH_W(n, m, s) do_sat_bhw(((int64_t)n * m + (1 << 30)) >> 31, \ + INT32_MIN, INT32_MAX, s) + #define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \ void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ uint32_t rm) \ @@ -585,6 +603,13 @@ DO_2OP_SAT_SCALAR(vqsubs_scalarb, 1, int8_t, DO_SQSUB_B) DO_2OP_SAT_SCALAR(vqsubs_scalarh, 2, int16_t, DO_SQSUB_H) DO_2OP_SAT_SCALAR(vqsubs_scalarw, 4, int32_t, DO_SQSUB_W) +DO_2OP_SAT_SCALAR(vqdmulh_scalarb, 1, int8_t, DO_QDMULH_B) +DO_2OP_SAT_SCALAR(vqdmulh_scalarh, 2, int16_t, DO_QDMULH_H) +DO_2OP_SAT_SCALAR(vqdmulh_scalarw, 4, int32_t, DO_QDMULH_W) +DO_2OP_SAT_SCALAR(vqrdmulh_scalarb, 1, int8_t, DO_QRDMULH_B) +DO_2OP_SAT_SCALAR(vqrdmulh_scalarh, 2, int16_t, DO_QRDMULH_H) +DO_2OP_SAT_SCALAR(vqrdmulh_scalarw, 4, int32_t, DO_QRDMULH_W) + static inline uint32_t do_vbrsrb(uint32_t n, uint32_t m) { m &= 0xff; |