fix vst1.f32 of coeff sum at eltwise_arm layer

In line 414: "vmla.f32 q1, q0, %q6 \n", destination register is q1 instead of q0, So, replace the {d0-d1} of line 416 with {d2-d3}.
XiaoguangHu01 · Mar 11, 2018 · 28154dc · 28154dc
1 parent 57f89a0
commit 28154dc
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/src/layer/arm/eltwise_arm.cpp b/src/layer/arm/eltwise_arm.cpp
@@ -413,7 +413,7 @@ int Eltwise_arm::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>&
                         "vld1.f32   {d2-d3}, [%2 :128]  \n"
                         "vmla.f32   q1, q0, %q6         \n"
                         "subs       %0, #1              \n"
-                        "vst1.f32   {d0-d1}, [%2 :128]! \n"
+                        "vst1.f32   {d2-d3}, [%2 :128]! \n"
                         "bne        0b                  \n"
                         : "=r"(nn),     // %0
                           "=r"(ptr),    // %1