diff --git a/step3 b/step3 new file mode 100644 index 0000000..7af8ca9 --- /dev/null +++ b/step3 @@ -0,0 +1,68 @@ +#include +#include +#include + +float inputImage[5][5] = { + {1, 2, 3, 4, 5}, + {6, 7, 8, 9, 10}, + {11, 12, 13, 14, 15}, + {16, 17, 18, 19, 20}, + {21, 22, 23, 24, 25} +}; + +float kx[3] = { 1.0f / 4, 2.0f / 4, 1.0f / 4 }; +float ky[3] = { 1.0f / 4, 2.0f / 4, 1.0f / 4 }; +float buf[3][3] = { 0 }; + +void applySeparableGaussianBlurNEON(float src[][5], float dst[][5], float kx[], float ky[]) { + float32x4_t kx_vec = vld1q_f32(kx); + float32x4_t ky_vec = vld1q_f32(ky); + + for (int i = 1; i < 3; ++i) { + for (int j = 0; j < 3; j += 4) { + float32x4_t left = vld1q_f32(&src[i - 1][j + 1]); + float32x4_t center = vld1q_f32(&src[i][j + 1]); + float32x4_t right = vld1q_f32(&src[i + 1][j + 1]); + + float32x4_t result_left = vmulq_f32(left, kx_vec); + float32x4_t result_center = vmulq_f32(center, kx_vec); + float32x4_t result_right = vmulq_f32(right, kx_vec); + + float32x4_t total_result = vaddq_f32(result_left, result_center); + total_result = vaddq_f32(total_result, result_right); + + vst1q_f32(&buf[i][j], total_result); + } + } + + for (int i = 0; i < 3; ++i) { + for (int j = 1; j < 3; j++) { + float32x4_t buf_data = vld1q_f32(&buf[i][j - 1]); + + float32x4_t result = vmulq_f32(buf_data, ky_vec); + + vst1q_f32(&dst[i + 1][j + 1], result); + } + } +} + +int main() { + float outputImage[5][5] = {0}; + clock_t start, end; + + start = clock(); + applySeparableGaussianBlurNEON(inputImage, outputImage, kx, ky); + end = clock(); + + double timeSpent = (double)(end - start) / CLOCKS_PER_SEC; + + for (int i = 0; i < 5; i++) { + for (int j = 0; j < 5; j++) { + printf("%.6f ", outputImage[i][j]); + } + printf("\n"); + } + + printf("Time spent: %.6f seconds\n", timeSpent); + return 0; +} \ No newline at end of file