Fix tanh typo for rvv. (#5584)

* Fix tanh typo for rvv. * Fix tanh for rvv fp16.
Tencent · Jul 18, 2024 · 081a9c3 · 081a9c3
1 parent 569617f
commit 081a9c3
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 2 deletions.
diff --git a/src/layer/riscv/rvv_mathfun.h b/src/layer/riscv/rvv_mathfun.h
@@ -308,7 +308,7 @@ _RVV_FLOAT32_COS_OP(8, 4)
                                                                                                  \
         /* clamp the inputs to the range [-9, 9] since anything outside */                       \
         /* this range is -/+1.0f in single-precision.                   */                       \
-        x2 = vfmin_vf_f32m##LMUL(x, c_tanh_hi, vl);                                              \
+        x2 = vfmin_vf_f32m##LMUL(x2, c_tanh_hi, vl);                                             \
                                                                                                  \
         /* since the polynomials are odd/even, we need x**2. */                                  \
         vfloat32m##LMUL##_t z = vfmul_vv_f32m##LMUL(x2, x2, vl);                                 \

diff --git a/src/layer/riscv/rvv_mathfun_fp16s.h b/src/layer/riscv/rvv_mathfun_fp16s.h
@@ -308,7 +308,7 @@ _RVV_FLOAT16_COS_OP(8, 2)
                                                                                                  \
         /* clamp the inputs to the range [-9, 9] since anything outside */                       \
         /* this range is -/+1.0f in single-precision.                   */                       \
-        x2 = vfmin_vf_f16m##LMUL(x, c_tanh_hi, vl);                                              \
+        x2 = vfmin_vf_f16m##LMUL(x2, c_tanh_hi, vl);                                             \
                                                                                                  \
         /* since the polynomials are odd/even, we need x**2. */                                  \
         vfloat16m##LMUL##_t z = vfmul_vv_f16m##LMUL(x2, x2, vl);                                 \