Remove trailing whitespace

PiperOrigin-RevId: 276185628
diff --git a/src/f32-conv-hwc/3x3s2p1c3x4-neonfma-2x2.c b/src/f32-conv-hwc/3x3s2p1c3x4-neonfma-2x2.c
index 298beec..9a62d7e 100644
--- a/src/f32-conv-hwc/3x3s2p1c3x4-neonfma-2x2.c
+++ b/src/f32-conv-hwc/3x3s2p1c3x4-neonfma-2x2.c
@@ -622,7 +622,7 @@
               vst1_f32((float*) ((uintptr_t) o0_tmp + output_width_stride), vo0x1_lo);
               vo0x1_lo = vget_high_f32(vo0x1);
             }
-      
+
             vst1_f32(o1_tmp, vo1x0_lo); o1_tmp += 2;
             vo1x0_lo = vget_high_f32(vo1x0);
             vst1_f32(o0_tmp, vo0x0_lo); o0_tmp += 2;
diff --git a/src/im2col.c b/src/im2col.c
index 22f5dd3..73b041f 100644
--- a/src/im2col.c
+++ b/src/im2col.c
@@ -35,7 +35,7 @@
           for (size_t kernel_x = 0; kernel_x < kernel_width; kernel_x++) {
             const size_t input_x = output_x * subsampling_width + kernel_x * dilation_width - input_padding_left;
             if (input_x < output_width) {
-              memcpy(output, 
+              memcpy(output,
                 (const void*) ((uintptr_t) input + (input_y * input_width + input_x) * input_pixel_stride_in_bytes),
                 group_input_channels_in_bytes);
             } else {
diff --git a/src/math/exp-avx2-p5.c b/src/math/exp-avx2-p5.c
index c3d21c9..683b10f 100644
--- a/src/math/exp-avx2-p5.c
+++ b/src/math/exp-avx2-p5.c
@@ -42,7 +42,7 @@
     const __m256 vx = _mm256_loadu_ps(input);
 
     // Compute reduced argument n := round(x / log(2)).
-    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the 
+    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the
     // large number back. The first addition is combined with multiplication by log2e into a single FMA instruction.
     // The trick with adding large number is valid only within certain bounds (|x| <= 2**22), but thats ok, because
     // inputs outside of [-103.97207, 88.72283] underflow or overflow expf(x) anyway. We fixup the result for such
diff --git a/src/math/exp-avx2-perm-p3.c b/src/math/exp-avx2-perm-p3.c
index 2ee2c60..7984765 100644
--- a/src/math/exp-avx2-perm-p3.c
+++ b/src/math/exp-avx2-perm-p3.c
@@ -43,7 +43,7 @@
     const __m256 vx = _mm256_loadu_ps(input);
 
     // Compute reduced argument n := round(x * 8 / log(2)).
-    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the 
+    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the
     // large number back. The first addition is combined with multiplication by log2e into a single FMA instruction.
     // The trick with adding large number is valid only within certain bounds (|x| <= 2**22), but thats ok, because
     // inputs outside of [-103.97207, 88.72283] underflow or overflow expf(x) anyway. We fixup the result for such
diff --git a/src/math/exp-avx2-perm-p4.c b/src/math/exp-avx2-perm-p4.c
index c53d607..2f69868 100644
--- a/src/math/exp-avx2-perm-p4.c
+++ b/src/math/exp-avx2-perm-p4.c
@@ -46,7 +46,7 @@
     const __m256 vx = _mm256_loadu_ps(input);
 
     // Compute reduced argument n := round(x * 8 / log(2)).
-    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the 
+    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the
     // large number back. The first addition is combined with multiplication by log2e into a single FMA instruction.
     // The trick with adding large number is valid only within certain bounds (|x| <= 2**22), but thats ok, because
     // inputs outside of [-103.97207, 88.72283] underflow or overflow expf(x) anyway. We fixup the result for such
diff --git a/src/math/exp-avx512f-p5.c b/src/math/exp-avx512f-p5.c
index 9b2faac..4b2c083 100644
--- a/src/math/exp-avx512f-p5.c
+++ b/src/math/exp-avx512f-p5.c
@@ -42,7 +42,7 @@
     const __m512 vx = _mm512_loadu_ps(input);
 
     // Compute reduced argument n := round(x / log(2)).
-    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the 
+    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the
     // large number back. The first addition is combined with multiplication by log2e into a single FMA instruction.
     // The trick with adding large number is valid only within certain bounds (|x| <= 2**22), but thats ok, because
     // inputs outside of [-103.97207, 88.72283] underflow or overflow expf(x) anyway. We fixup the result for such
diff --git a/src/math/exp-avx512f-perm-p3.c b/src/math/exp-avx512f-perm-p3.c
index c014392..ff82d1b 100644
--- a/src/math/exp-avx512f-perm-p3.c
+++ b/src/math/exp-avx512f-perm-p3.c
@@ -45,7 +45,7 @@
     const __m512 vx = _mm512_loadu_ps(input);
 
     // Compute reduced argument n := round(x * 16 / log(2)).
-    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the 
+    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the
     // large number back. The first addition is combined with multiplication by log2e into a single FMA instruction.
     // The trick with adding large number is valid only within certain bounds (|x| <= 2**22), but thats ok, because
     // inputs outside of [-103.97207, 88.72283] underflow or overflow expf(x) anyway. We fixup the result for such
diff --git a/src/math/expminus-avx2-p5.c b/src/math/expminus-avx2-p5.c
index 8c05304..5d4fe4e 100644
--- a/src/math/expminus-avx2-p5.c
+++ b/src/math/expminus-avx2-p5.c
@@ -35,7 +35,7 @@
     const __m256 vx = _mm256_loadu_ps(input);
 
     // Compute reduced argument n := round(x / log(2)).
-    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the 
+    // We do it by adding a large number (magic bias), which cause rounding of result to an integer, then subtracing the
     // large number back. The first addition is combined with multiplication by log2e into a single FMA instruction.
     // The trick with adding large number is valid only within certain bounds (|x| <= 2**22), but thats ok, because
     // inputs outside of [-103.97207, 0.0] underflow expf(x) anyway. We fixup the result for such inputs at the very end
diff --git a/src/q8-gemm/2x4c8-sse2.c b/src/q8-gemm/2x4c8-sse2.c
index 8551ad6..55e962b 100644
--- a/src/q8-gemm/2x4c8-sse2.c
+++ b/src/q8-gemm/2x4c8-sse2.c
@@ -149,7 +149,7 @@
     const __m128i vq31prod1x0123 = _mm_shuffle_epi32(vq31prod1x0213, _MM_SHUFFLE(3, 1, 2, 0));
 
     const __m128i vremainder_mask = _mm_load_si128((const __m128i*) params->sse2.remainder_mask);
-    
+
     const __m128i vrem0x0123 =
       _mm_add_epi32(_mm_and_si128(vq31prod0x0123, vremainder_mask), _mm_cmpgt_epi32(_mm_setzero_si128(), vq31prod0x0123));
     const __m128i vrem1x0123 =
@@ -158,9 +158,9 @@
     const __m128i vremainder_threshold = _mm_load_si128((const __m128i*) params->sse2.remainder_threshold);
     const __m128i vshift = _mm_load_si128((const __m128i*) params->sse2.shift);
 
-    vacc0x0123 = 
+    vacc0x0123 =
       _mm_sub_epi32(_mm_sra_epi32(vq31prod0x0123, vshift), _mm_cmpgt_epi32(vrem0x0123, vremainder_threshold));
-    vacc1x0123 = 
+    vacc1x0123 =
       _mm_sub_epi32(_mm_sra_epi32(vq31prod1x0123, vshift), _mm_cmpgt_epi32(vrem1x0123, vremainder_threshold));
 
     const __m128i voutput_zero_point = _mm_load_si128((const __m128i*) params->sse2.output_zero_point);
diff --git a/src/u8-maxpool/9p8q-neon.c b/src/u8-maxpool/9p8q-neon.c
index 28c19ea..9deaf24 100644
--- a/src/u8-maxpool/9p8q-neon.c
+++ b/src/u8-maxpool/9p8q-neon.c
@@ -129,7 +129,7 @@
         }
       }
     }
-    
+
     for (ptrdiff_t m = (ptrdiff_t) ks - 9; m > 0; m -= 8) {
       const uint8_t* i0 = *input++;
       const uint8_t* i1 = *input++;
diff --git a/src/u8-maxpool/9p8q-scalar.c b/src/u8-maxpool/9p8q-scalar.c
index 198dce7..e442f75 100644
--- a/src/u8-maxpool/9p8q-scalar.c
+++ b/src/u8-maxpool/9p8q-scalar.c
@@ -89,7 +89,7 @@
         *o++ = vout;
       } while (--k != 0);
     }
-    
+
     for (ptrdiff_t m = (ptrdiff_t) ks - 9; m > 0; m -= 8) {
       const uint8_t* i0 = *input++;
       const uint8_t* i1 = *input++;
diff --git a/src/u8-maxpool/9p8q-sse2.c b/src/u8-maxpool/9p8q-sse2.c
index 1805a69..8903457 100644
--- a/src/u8-maxpool/9p8q-sse2.c
+++ b/src/u8-maxpool/9p8q-sse2.c
@@ -133,7 +133,7 @@
         }
       }
     }
-    
+
     for (ptrdiff_t m = (ptrdiff_t) ks - 9; m > 0; m -= 8) {
       const uint8_t* i0 = *input++;
       const uint8_t* i1 = *input++;
diff --git a/test/average-pooling-operator-tester.h b/test/average-pooling-operator-tester.h
index f12e280..fb28701 100644
--- a/test/average-pooling-operator-tester.h
+++ b/test/average-pooling-operator-tester.h
@@ -412,7 +412,7 @@
                 const size_t iy = oy * stride_height() + py - padding_top();
                 for (size_t px = 0; px < pooling_width(); px++) {
                   const size_t ix = ox * stride_width() + px - padding_left();
-                  if (ix < input_width() && iy < input_height()) {                  
+                  if (ix < input_width() && iy < input_height()) {
                     acc += double(int32_t(input[((i * input_height() + iy) * input_width() + ix) * input_pixel_stride() + c]) - int32_t(input_zero_point()));
                   }
                 }
@@ -491,12 +491,12 @@
           for (size_t ox = 0; ox < output_width(); ox++) {
             for (size_t c = 0; c < channels(); c++) {
               float acc = 0.0f;
-              int32_t n = 0; 
+              int32_t n = 0;
               for (size_t py = 0; py < pooling_height(); py++) {
                 const size_t iy = oy * stride_height() + py - padding_top();
                 for (size_t px = 0; px < pooling_width(); px++) {
                   const size_t ix = ox * stride_width() + px - padding_left();
-                  if (ix < input_width() && iy < input_height()) {                  
+                  if (ix < input_width() && iy < input_height()) {
                     acc += input[((i * input_height() + iy) * input_width() + ix) * input_pixel_stride() + c];
                     n += 1;
                   }
@@ -597,7 +597,7 @@
                 const size_t iy = oy * stride_height() + py - padding_top();
                 for (size_t px = 0; px < pooling_width(); px++) {
                   const size_t ix = ox * stride_width() + px - padding_left();
-                  if (ix < input_width() && iy < input_height()) {                  
+                  if (ix < input_width() && iy < input_height()) {
                     acc += double(int32_t(input[((i * input_height() + iy) * input_width() + ix) * input_pixel_stride() + c]) - int32_t(input_zero_point()));
                   }
                 }
@@ -667,7 +667,7 @@
                 const size_t iy = oy * stride_height() + py - padding_top();
                 for (size_t px = 0; px < pooling_width(); px++) {
                   const size_t ix = ox * stride_width() + px - padding_left();
-                  if (ix < next_input_width() && iy < next_input_height()) {                  
+                  if (ix < next_input_width() && iy < next_input_height()) {
                     acc += double(int32_t(input[((i * next_input_height() + iy) * next_input_width() + ix) * input_pixel_stride() + c]) - int32_t(input_zero_point()));
                   }
                 }
@@ -742,7 +742,7 @@
                 const size_t iy = oy * stride_height() + py - padding_top();
                 for (size_t px = 0; px < pooling_width(); px++) {
                   const size_t ix = ox * stride_width() + px - padding_left();
-                  if (ix < input_width() && iy < input_height()) {                  
+                  if (ix < input_width() && iy < input_height()) {
                     acc += input[((i * input_height() + iy) * input_width() + ix) * input_pixel_stride() + c];
                     n += 1;
                   }
@@ -825,7 +825,7 @@
                 const size_t iy = oy * stride_height() + py - padding_top();
                 for (size_t px = 0; px < pooling_width(); px++) {
                   const size_t ix = ox * stride_width() + px - padding_left();
-                  if (ix < next_input_width() && iy < next_input_height()) {                  
+                  if (ix < next_input_width() && iy < next_input_height()) {
                     acc += input[((i * next_input_height() + iy) * next_input_width() + ix) * input_pixel_stride() + c];
                     n += 1;
                   }
diff --git a/test/clamp-microkernel-tester.h b/test/clamp-microkernel-tester.h
index 1bcc4c9..e43da37 100644
--- a/test/clamp-microkernel-tester.h
+++ b/test/clamp-microkernel-tester.h
@@ -103,7 +103,6 @@
           output_params = xnn_compute_scalar_u8_output_params(qmin(), qmax());
           break;
       }
-          
 
       // Compute reference results.
       for (size_t i = 0; i < n(); i++) {
diff --git a/test/f32-gavgpool.cc b/test/f32-gavgpool.cc
index 95c3f5c..8ebb7f2 100644
--- a/test/f32-gavgpool.cc
+++ b/test/f32-gavgpool.cc
@@ -64,7 +64,7 @@
       GAvgPoolMicrokernelTester()
         .m(7)
         .n(n)
-        .Test(xnn_f32_gavgpool_ukernel_up7__neon);    
+        .Test(xnn_f32_gavgpool_ukernel_up7__neon);
     }
   }
 
@@ -130,7 +130,7 @@
       GAvgPoolMicrokernelTester()
         .m(7)
         .n(n)
-        .Test(xnn_f32_gavgpool_ukernel_up7__neon);    
+        .Test(xnn_f32_gavgpool_ukernel_up7__neon);
     }
   }
 
@@ -231,7 +231,7 @@
         .m(m)
         .n(4)
         .Test(xnn_f32_gavgpool_ukernel_mp7p7q__neon);
-    } 
+    }
   }
 
   TEST(F32_GAVGPOOL_MP7p7q__NEON, n_eq_4_multipass_fulltile_with_x_stride) {
@@ -241,7 +241,7 @@
         .m(m)
         .n(4)
         .Test(xnn_f32_gavgpool_ukernel_mp7p7q__neon);
-    } 
+    }
   }
 
   TEST(F32_GAVGPOOL_MP7p7q__NEON, n_div_4_2pass_fulltile) {
@@ -274,7 +274,7 @@
           .m(m)
           .n(n)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 
@@ -287,7 +287,7 @@
           .n(n)
           .x_stride(131)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 
@@ -343,7 +343,7 @@
           .m(m)
           .n(n)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 
@@ -356,7 +356,7 @@
           .n(n)
           .x_stride(23)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 
@@ -412,7 +412,7 @@
           .m(m)
           .n(n)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 
@@ -425,7 +425,7 @@
           .n(n)
           .x_stride(23)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 #endif  // XNN_ARCH_ARM || XNN_ARCH_ARM64
@@ -483,7 +483,7 @@
       GAvgPoolMicrokernelTester()
         .m(7)
         .n(n)
-        .Test(xnn_f32_gavgpool_ukernel_up7__sse);    
+        .Test(xnn_f32_gavgpool_ukernel_up7__sse);
     }
   }
 
@@ -549,7 +549,7 @@
       GAvgPoolMicrokernelTester()
         .m(7)
         .n(n)
-        .Test(xnn_f32_gavgpool_ukernel_up7__sse);    
+        .Test(xnn_f32_gavgpool_ukernel_up7__sse);
     }
   }
 
@@ -650,7 +650,7 @@
         .m(m)
         .n(4)
         .Test(xnn_f32_gavgpool_ukernel_mp7p7q__sse);
-    } 
+    }
   }
 
   TEST(F32_GAVGPOOL_MP7p7q__SSE2, n_eq_4_multipass_fulltile_with_x_stride) {
@@ -660,7 +660,7 @@
         .m(m)
         .n(4)
         .Test(xnn_f32_gavgpool_ukernel_mp7p7q__sse);
-    } 
+    }
   }
 
   TEST(F32_GAVGPOOL_MP7p7q__SSE2, n_div_4_2pass_fulltile) {
@@ -693,7 +693,7 @@
           .m(m)
           .n(n)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__sse);
-      } 
+      }
     }
   }
 
@@ -706,7 +706,7 @@
           .n(n)
           .x_stride(131)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__sse);
-      } 
+      }
     }
   }
 
@@ -762,7 +762,7 @@
           .m(m)
           .n(n)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__sse);
-      } 
+      }
     }
   }
 
@@ -775,7 +775,7 @@
           .n(n)
           .x_stride(23)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__sse);
-      } 
+      }
     }
   }
 
@@ -831,7 +831,7 @@
           .m(m)
           .n(n)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__sse);
-      } 
+      }
     }
   }
 
@@ -844,7 +844,7 @@
           .n(n)
           .x_stride(23)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__sse);
-      } 
+      }
     }
   }
 #endif  // XNN_ARCH_X86 || XNN_ARCH_X86_64
@@ -902,7 +902,7 @@
       GAvgPoolMicrokernelTester()
         .m(7)
         .n(n)
-        .Test(xnn_f32_gavgpool_ukernel_up7__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);    
+        .Test(xnn_f32_gavgpool_ukernel_up7__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);
     }
   }
 
@@ -968,7 +968,7 @@
       GAvgPoolMicrokernelTester()
         .m(7)
         .n(n)
-        .Test(xnn_f32_gavgpool_ukernel_up7__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);    
+        .Test(xnn_f32_gavgpool_ukernel_up7__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);
     }
   }
 
@@ -1069,7 +1069,7 @@
         .m(m)
         .n(4)
         .Test(xnn_f32_gavgpool_ukernel_mp7p7q__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);
-    } 
+    }
   }
 
   TEST(F32_GAVGPOOL_MP7p7q__PSIMD, n_eq_4_multipass_fulltile_with_x_stride) {
@@ -1079,7 +1079,7 @@
         .m(m)
         .n(4)
         .Test(xnn_f32_gavgpool_ukernel_mp7p7q__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);
-    } 
+    }
   }
 
   TEST(F32_GAVGPOOL_MP7p7q__PSIMD, n_div_4_2pass_fulltile) {
@@ -1112,7 +1112,7 @@
           .m(m)
           .n(n)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);
-      } 
+      }
     }
   }
 
@@ -1125,7 +1125,7 @@
           .n(n)
           .x_stride(131)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);
-      } 
+      }
     }
   }
 
@@ -1181,7 +1181,7 @@
           .m(m)
           .n(n)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);
-      } 
+      }
     }
   }
 
@@ -1194,7 +1194,7 @@
           .n(n)
           .x_stride(23)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);
-      } 
+      }
     }
   }
 
@@ -1250,7 +1250,7 @@
           .m(m)
           .n(n)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);
-      } 
+      }
     }
   }
 
@@ -1263,7 +1263,7 @@
           .n(n)
           .x_stride(23)
           .Test(xnn_f32_gavgpool_ukernel_mp7p7q__psimd, GAvgPoolMicrokernelTester::Variant::Scalar);
-      } 
+      }
     }
   }
 #endif  // !XNN_ARCH_WASM && !XNN_ARCH_ASMJS
@@ -1314,7 +1314,7 @@
     GAvgPoolMicrokernelTester()
       .m(7)
       .n(n)
-      .Test(xnn_f32_gavgpool_ukernel_up7__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);    
+      .Test(xnn_f32_gavgpool_ukernel_up7__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);
   }
 }
 
@@ -1405,7 +1405,7 @@
       .m(m)
       .n(1)
         .Test(xnn_f32_gavgpool_ukernel_mp7p7q__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);
-  } 
+  }
 }
 
 TEST(F32_GAVGPOOL_MP7p7q__SCALAR, n_eq_1_multipass_fulltile_with_x_stride) {
@@ -1414,7 +1414,7 @@
       .m(m)
       .n(1)
         .Test(xnn_f32_gavgpool_ukernel_mp7p7q__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);
-  } 
+  }
 }
 
 TEST(F32_GAVGPOOL_MP7p7q__SCALAR, n_gt_1_2pass_fulltile) {
@@ -1464,7 +1464,7 @@
         .m(m)
         .n(n)
             .Test(xnn_f32_gavgpool_ukernel_mp7p7q__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);
-    } 
+    }
   }
 }
 
@@ -1476,6 +1476,6 @@
         .n(n)
             .x_stride(23)
         .Test(xnn_f32_gavgpool_ukernel_mp7p7q__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);
-    } 
+    }
   }
 }
diff --git a/test/max-pooling-operator-tester.h b/test/max-pooling-operator-tester.h
index b411e14..eb312fb 100644
--- a/test/max-pooling-operator-tester.h
+++ b/test/max-pooling-operator-tester.h
@@ -414,7 +414,7 @@
                 const size_t iy = oy * stride_height() + py * dilation_height() - padding_top();
                 for (size_t px = 0; px < pooling_width(); px++) {
                   const size_t ix = ox * stride_width() + px * dilation_width() - padding_left();
-                  if (ix < input_width() && iy < input_height()) {                  
+                  if (ix < input_width() && iy < input_height()) {
                     max_value = std::max(max_value,
                       input[((i * input_height() + iy) * input_width() + ix) * input_pixel_stride() + c]);
                   }
diff --git a/test/q8-gavgpool.cc b/test/q8-gavgpool.cc
index 79a27bf..b3056ad 100644
--- a/test/q8-gavgpool.cc
+++ b/test/q8-gavgpool.cc
@@ -119,7 +119,7 @@
       GAvgPoolMicrokernelTester()
         .m(7)
         .n(n)
-        .Test(xnn_q8_gavgpool_ukernel_up7__neon);    
+        .Test(xnn_q8_gavgpool_ukernel_up7__neon);
     }
   }
 
@@ -245,7 +245,7 @@
       GAvgPoolMicrokernelTester()
         .m(7)
         .n(n)
-        .Test(xnn_q8_gavgpool_ukernel_up7__neon);    
+        .Test(xnn_q8_gavgpool_ukernel_up7__neon);
     }
   }
 
@@ -458,7 +458,7 @@
         .m(m)
         .n(8)
         .Test(xnn_q8_gavgpool_ukernel_mp7p7q__neon);
-    } 
+    }
   }
 
   TEST(Q8_GAVGPOOL_MP7p7q__NEON, n_eq_8_multipass_fulltile_with_x_stride) {
@@ -468,7 +468,7 @@
         .m(m)
         .n(8)
         .Test(xnn_q8_gavgpool_ukernel_mp7p7q__neon);
-    } 
+    }
   }
 
   TEST(Q8_GAVGPOOL_MP7p7q__NEON, n_div_8_2pass_fulltile) {
@@ -501,7 +501,7 @@
           .m(m)
           .n(n)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 
@@ -514,7 +514,7 @@
           .n(n)
           .x_stride(131)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 
@@ -630,7 +630,7 @@
           .m(m)
           .n(n)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 
@@ -643,7 +643,7 @@
           .n(n)
           .x_stride(23)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 
@@ -759,7 +759,7 @@
           .m(m)
           .n(n)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 
@@ -772,7 +772,7 @@
           .n(n)
           .x_stride(23)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__neon);
-      } 
+      }
     }
   }
 #endif  // XNN_ARCH_ARM || XNN_ARCH_ARM64
@@ -881,7 +881,7 @@
       GAvgPoolMicrokernelTester()
         .m(7)
         .n(n)
-        .Test(xnn_q8_gavgpool_ukernel_up7__sse2);    
+        .Test(xnn_q8_gavgpool_ukernel_up7__sse2);
     }
   }
 
@@ -1007,7 +1007,7 @@
       GAvgPoolMicrokernelTester()
         .m(7)
         .n(n)
-        .Test(xnn_q8_gavgpool_ukernel_up7__sse2);    
+        .Test(xnn_q8_gavgpool_ukernel_up7__sse2);
     }
   }
 
@@ -1220,7 +1220,7 @@
         .m(m)
         .n(8)
         .Test(xnn_q8_gavgpool_ukernel_mp7p7q__sse2);
-    } 
+    }
   }
 
   TEST(Q8_GAVGPOOL_MP7p7q__SSE2, n_eq_8_multipass_fulltile_with_x_stride) {
@@ -1230,7 +1230,7 @@
         .m(m)
         .n(8)
         .Test(xnn_q8_gavgpool_ukernel_mp7p7q__sse2);
-    } 
+    }
   }
 
   TEST(Q8_GAVGPOOL_MP7p7q__SSE2, n_div_8_2pass_fulltile) {
@@ -1263,7 +1263,7 @@
           .m(m)
           .n(n)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__sse2);
-      } 
+      }
     }
   }
 
@@ -1276,7 +1276,7 @@
           .n(n)
           .x_stride(131)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__sse2);
-      } 
+      }
     }
   }
 
@@ -1392,7 +1392,7 @@
           .m(m)
           .n(n)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__sse2);
-      } 
+      }
     }
   }
 
@@ -1405,7 +1405,7 @@
           .n(n)
           .x_stride(23)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__sse2);
-      } 
+      }
     }
   }
 
@@ -1521,7 +1521,7 @@
           .m(m)
           .n(n)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__sse2);
-      } 
+      }
     }
   }
 
@@ -1534,7 +1534,7 @@
           .n(n)
           .x_stride(23)
           .Test(xnn_q8_gavgpool_ukernel_mp7p7q__sse2);
-      } 
+      }
     }
   }
 #endif  // XNN_ARCH_X86 || XNN_ARCH_X86_64
@@ -1632,7 +1632,7 @@
     GAvgPoolMicrokernelTester()
       .m(7)
       .n(n)
-      .Test(xnn_q8_gavgpool_ukernel_up7__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);    
+      .Test(xnn_q8_gavgpool_ukernel_up7__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);
   }
 }
 
@@ -1831,7 +1831,7 @@
       .m(m)
       .n(1)
       .Test(xnn_q8_gavgpool_ukernel_mp7p7q__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);
-  } 
+  }
 }
 
 TEST(Q8_GAVGPOOL_MP7p7q__SCALAR, n_eq_1_multipass_fulltile_with_x_stride) {
@@ -1840,7 +1840,7 @@
       .m(m)
       .n(1)
       .Test(xnn_q8_gavgpool_ukernel_mp7p7q__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);
-  } 
+  }
 }
 
 TEST(Q8_GAVGPOOL_MP7p7q__SCALAR, n_gt_1_2pass_fulltile) {
@@ -1946,7 +1946,7 @@
         .m(m)
         .n(n)
         .Test(xnn_q8_gavgpool_ukernel_mp7p7q__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);
-    } 
+    }
   }
 }
 
@@ -1958,6 +1958,6 @@
         .n(n)
         .x_stride(23)
         .Test(xnn_q8_gavgpool_ukernel_mp7p7q__scalar, GAvgPoolMicrokernelTester::Variant::Scalar);
-    } 
+    }
   }
 }