fix softmax arm fp16s sum error, fix #5340 (#5393)

Tencent · Mar 30, 2024 · 167501f · 167501f
1 parent 6595743
commit 167501f
Showing 1 changed file with 4 additions and 4 deletions.
diff --git a/src/layer/arm/softmax_arm_asimdhp.cpp b/src/layer/arm/softmax_arm_asimdhp.cpp
@@ -255,7 +255,7 @@ int Softmax_arm::forward_inplace_fp16s(Mat& bottom_top_blob, const Option& opt)
                     float16x8_t _ss01 = vpaddq_f16(_p0, _p1);
                     float16x8_t _ss23 = vpaddq_f16(_p2, _p3);
                     float16x8_t _ss2 = vpaddq_f16(_ss01, _ss23);
-                    _sum = vadd_f16(_sum, vpmax_f16(vget_low_f16(_ss2), vget_high_f16(_ss2)));
+                    _sum = vadd_f16(_sum, vpadd_f16(vget_low_f16(_ss2), vget_high_f16(_ss2)));
                     vst1_f16(sumptr, _sum);
                     ptr += 32;
                     maxptr += 4;
@@ -292,7 +292,7 @@ int Softmax_arm::forward_inplace_fp16s(Mat& bottom_top_blob, const Option& opt)
                     vst1q_f16(ptr, _p0);
                     vst1q_f16(ptr + 8, _p1);
                     float16x8_t _ss2 = vpaddq_f16(_p0, _p1);
-                    _sum = vadd_f16(_sum, vpmax_f16(vget_low_f16(_ss2), vget_high_f16(_ss2)));
+                    _sum = vadd_f16(_sum, vpadd_f16(vget_low_f16(_ss2), vget_high_f16(_ss2)));
                     vst1_f16(sumptr, _sum);
                     ptr += 16;
                     maxptr += 4;
@@ -743,7 +743,7 @@ int Softmax_arm::forward_inplace_fp16s(Mat& bottom_top_blob, const Option& opt)
                     float16x8_t _ss01 = vpaddq_f16(_p0, _p1);
                     float16x8_t _ss23 = vpaddq_f16(_p2, _p3);
                     float16x8_t _ss2 = vpaddq_f16(_ss01, _ss23);
-                    _sum = vadd_f16(_sum, vpmax_f16(vget_low_f16(_ss2), vget_high_f16(_ss2)));
+                    _sum = vadd_f16(_sum, vpadd_f16(vget_low_f16(_ss2), vget_high_f16(_ss2)));
                     vst1_f16(sumptr, _sum);
                     ptr += 32;
                     sumptr += 4;
@@ -768,7 +768,7 @@ int Softmax_arm::forward_inplace_fp16s(Mat& bottom_top_blob, const Option& opt)
                     float16x8_t _p1 = vld1q_f16(ptr + 8);
                     float16x4_t _sum = vld1_f16(sumptr);
                     float16x8_t _ss2 = vpaddq_f16(_p0, _p1);
-                    _sum = vadd_f16(_sum, vpmax_f16(vget_low_f16(_ss2), vget_high_f16(_ss2)));
+                    _sum = vadd_f16(_sum, vpadd_f16(vget_low_f16(_ss2), vget_high_f16(_ss2)));
                     vst1_f16(sumptr, _sum);
                     ptr += 16;
                     sumptr += 4;