Merge branch 'gemm-quantize-r' of github.com:nihui/ncnn into gemm-qua…

…ntize-r
Tencent · Sep 30, 2024 · 87d5cd4 · 87d5cd4
2 parents 55e0e57 + 93fad8d
commit 87d5cd4
Showing 1 changed file with 6 additions and 6 deletions.
diff --git a/src/layer/arm/gemm_int8_fp16s.h b/src/layer/arm/gemm_int8_fp16s.h
@@ -102,9 +102,9 @@ static void compute_A_tile_fp16_int8_scales(const Mat& A, Mat& scales, float B_s
                 _amax0 = vmaxq_f16(_amax0, vabsq_f16(_p));
                 p0 += 8;
             }
-            float32x4_t _absmax0 = vcvt_f32_f16(vget_low_f16(_amax0));
-            float32x4_t _absmax1 = vcvt_f32_f16(vget_high_f16(_amax0));
-#else // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+            float32x4_t _absmax0 = vcvt_f32_f16(vget_low_f16(_absmax));
+            float32x4_t _absmax1 = vcvt_f32_f16(vget_high_f16(_absmax));
+#else  // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
             const unsigned short* p0 = (const unsigned short*)A + (i + ii) * A_hstep;
 
             float32x4_t _absmax0 = vdupq_n_f32(0.f);
@@ -237,8 +237,8 @@ static void compute_A_tile_fp16_int8_scales(const Mat& A, Mat& scales, float B_s
                 _amax = vmax_f16(_amax, vabs_f16(_p));
                 p0 += 4;
             }
-            float32x4_t _absmax0 = vcvt_f32_f16(_amax);
-#else // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+            float32x4_t _absmax0 = vcvt_f32_f16(_absmax);
+#else  // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
             const unsigned short* p0 = (const unsigned short*)A + (i + ii) * A_hstep;
 
             float32x4_t _absmax0 = vdupq_n_f32(0.f);
@@ -1559,7 +1559,7 @@ static void transpose_compute_A_tile_fp16_int8_scales(const Mat& A, Mat& scales,
             }
             float16x4_t _aa = vmax_f16(vget_low_f16(_absmax), vget_high_f16(_absmax));
             float absmax = vmaxvq_f32(vcvt_f32_f16(_aa));
-#else // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+#else  // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
             const unsigned short* p0 = (const unsigned short*)A + (i + ii) * 8;
 
             float32x4_t _absmax0 = vdupq_n_f32(0.f);