diff --git a/src/TpHostGpu.cu b/src/TpHostGpu.cu
index 6cd44604d12e3c5d8ec86970661ed2b5e28fca76..4a471096f2fe0ca6324b57f4aae486c682ef8c0c 100644
--- a/src/TpHostGpu.cu
+++ b/src/TpHostGpu.cu
@@ -893,7 +893,8 @@ void TpHostGpu::testCorrelate2DIntraTD (int num_runs, int quad_combine){ // 886
(corr_size_combo * corr_size_combo) * sizeof(float),
num_corr_combo,
cudaMemcpyDeviceToHost));
- printf("Writing phase correlation data to %s\n", m_tpPaths.result_corr_quad_file);
+ printf("Writing phase correlation data to %s (%d x %d - not an image) \n", m_tpPaths.result_corr_quad_file,
+ (m_tpParams.tilesx * corr_size_combo),(m_tpParams.tilesya * corr_size_combo));
writeFloatsToFile(
cpu_corr_combo, // float * data, // allocated array
rslt_corr_size_combo, // int size, // length in elements
@@ -906,17 +907,9 @@ void TpHostGpu::testCorrelate2DIntraTD (int num_runs, int quad_combine){ // 886
m_gpu_num_corr_tiles,
sizeof(int),
cudaMemcpyDeviceToHost));
- // printf("Average CORR run time =%f ms, num cor tiles (new) = %d\n", avgTimeCORR, num_corrs);
-
- // int corr_size = 2 * CORR_OUT_RAD + 1;
- // int rslt_corr_size = num_corrs * corr_size * corr_size;
- // float * cpu_corr = (float *)malloc(rslt_corr_size * sizeof(float));
-
-// int num_corr_indices = num_pairs_inter * m_tpParams.num_tiles;
int num_corr_indices = m_tpParams.num_pairs * m_tpParams.num_tiles;
int rslt_corr_size = num_corrs * m_tpParams.corr_length; // corr_size * corr_size;
int corr_img_size = num_corr_indices * 16*16; // NAN
-// float * corr_img = (float *)malloc(corr_img_size * sizeof(float));
float * cpu_corr = (float *)malloc(rslt_corr_size * sizeof(float));
int * cpu_corr_indices = (int *) malloc(num_corr_indices * sizeof(int));
checkCudaErrors(cudaMemcpy2D(
@@ -964,7 +957,8 @@ void TpHostGpu::testCorrelate2DIntraTD (int num_runs, int quad_combine){ // 886
cpu_corr_td, // float * cpu_corr,
16); // num_sel_sensors); // int num_sel_sensors) // // Will not be used
- printf("Writing intrascene phase correlation TD data tp %s\n", m_tpPaths.result_intrascene_td);
+ printf("Writing intrascene phase correlation TD data tp %s (%d x %d x %d)\n", m_tpPaths.result_intrascene_td,
+ (m_tpParams.tilesx * 16),(m_tpParams.tilesya*16), m_tpParams.num_pairs);
writeFloatsToFile(
corr_img_td, // float * data, // allocated array
corr_img_size, // int size, // length in elements
@@ -1268,9 +1262,9 @@ void TpHostGpu::testTextures (
gpu_pnum_texture_tiles,
sizeof(int),
cudaMemcpyDeviceToHost));
- printf("cpu_pnum_texture_tiles = %d\n", cpu_pnum_texture_tiles);
- printf("tile_texture_layers = %d\n", m_tile_texture_layers);
} // if (use_dp) {
+ printf("cpu_pnum_texture_tiles = %d\n", cpu_pnum_texture_tiles);
+ printf("tile_texture_layers = %d\n", m_tile_texture_layers);
int rslt_texture_size = m_num_textures * m_tile_texture_size;
@@ -1328,29 +1322,20 @@ void TpHostGpu::testTextures (
cpu_diff_rgb_combo_out[nl * ntiles + ntile] = cpu_diff_rgb_combo[ntile * nlayers + nl];
}
}
- if (!use_dp) {
- printf("Writing phase texture data to %s\n", m_tpPaths.result_textures_file);
- writeFloatsToFile(
- non_overlap_layers, // float * data, // allocated array
- rslt_texture_size, // int size, // length in elements
- m_tpPaths.result_textures_file); // const char * path) // file path
- printf("Writing low-res data to %s\n", m_tpPaths.result_diff_rgb_combo_file);
- writeFloatsToFile(
- cpu_diff_rgb_combo_out, // cpu_diff_rgb_combo, // float * data, // allocated array
- diff_rgb_combo_size, // int size, // length in elements
- m_tpPaths.result_diff_rgb_combo_file); // const char * path) // file path
- } else { // if (!use_dp) {
- printf("Writing phase texture data to %s\n", m_tpPaths.result_textures_file_dp);
- writeFloatsToFile(
- non_overlap_layers, // float * data, // allocated array
- rslt_texture_size, // int size, // length in elements
- m_tpPaths.result_textures_file_dp); // const char * path) // file path
- printf("Writing low-res data to %s\n", m_tpPaths.result_diff_rgb_combo_file_dp);
- writeFloatsToFile(
- cpu_diff_rgb_combo_out, // cpu_diff_rgb_combo, // float * data, // allocated array
- diff_rgb_combo_size, // int size, // length in elements
- m_tpPaths.result_diff_rgb_combo_file_dp); // const char * path) // file path
- } // if (!use_dp) { else
+ const char * result_textures_file = use_dp ? m_tpPaths.result_textures_file_dp : m_tpPaths.result_textures_file;
+ const char * result_diff_rgb_combo_file = use_dp ? m_tpPaths.result_diff_rgb_combo_file_dp : m_tpPaths.result_diff_rgb_combo_file;
+ printf("Writing phase texture data to %s (%d x % d x %d)\n", result_textures_file,
+ m_tpParams.tilesx * m_tpParams.dtt_size2, m_tpParams.tilesy * m_tpParams.dtt_size2, m_tile_texture_layers);
+ writeFloatsToFile(
+ non_overlap_layers, // float * data, // allocated array
+ rslt_texture_size, // int size, // length in elements
+ result_textures_file); // const char * path) // file path
+ printf("Writing low-res data to %s (%d x %d x %d)\n", result_diff_rgb_combo_file,
+ m_tpParams.tilesx, m_tpParams.tilesy, nlayers);
+ writeFloatsToFile(
+ cpu_diff_rgb_combo_out, // cpu_diff_rgb_combo, // float * data, // allocated array
+ diff_rgb_combo_size, // int size, // length in elements
+ result_diff_rgb_combo_file); // const char * path) // file path
if ( m_tpParams.debug_tile) {
if (debug) {
int texture_offset = m_tpParams.dbg_tile * m_tile_texture_size;
@@ -1518,19 +1503,14 @@ void TpHostGpu::testTexturesRGBA (
rgba_height * rbga_slices,
cudaMemcpyDeviceToHost));
printf("Import raw, real, little-endian, 18 images 648x520\n");
- if (!use_dp) {
- printf("Writing RBGA texture slices to %s\n", m_tpPaths.result_textures_rgba_file);
- writeFloatsToFile(
- cpu_textures_rgba, // float * data, // allocated array
- rslt_rgba_size, // int size, // length in elements
- m_tpPaths.result_textures_rgba_file); // const char * path) // file path
- } else { // if (use_dp) {
- printf("Writing RBGA texture slices to %s\n", m_tpPaths.result_textures_rgba_file_dp);
+ const char * result_textures_rgba_file = use_dp ? m_tpPaths.result_textures_rgba_file_dp : m_tpPaths.result_textures_rgba_file;
+
+ printf("Writing RBGA texture slices to %s (%d x %d x %d)\n", result_textures_rgba_file,
+ rgba_woi_width, rgba_woi_height, rbga_slices);
writeFloatsToFile(
cpu_textures_rgba, // float * data, // allocated array
rslt_rgba_size, // int size, // length in elements
- m_tpPaths.result_textures_rgba_file_dp); // const char * path) // file path
- } // if (use_dp) { // else
+ result_textures_rgba_file); // const char * path) // file path
if ( m_tpParams.debug_tile) {
if (debug) {
@@ -1565,7 +1545,7 @@ void TpHostGpu::saveClt(
m_gpu_clt_h[ncam],
rslt_size * sizeof(float),
cudaMemcpyDeviceToHost));
- printf("Writing %s to %s\n", prompt, paths[ncam]);
+ printf("Writing %s to %s (%d * %d * 4 * %d * %d)\n", prompt, paths[ncam], m_tpParams.tilesx, m_tpParams.tilesy, m_tpParams.dtt_size, m_tpParams.dtt_size);
writeFloatsToFile(cpu_clt, // float * data, // allocated array
rslt_size, // int size, // length in elements
paths[ncam]); // const char * path) // file path
@@ -1589,7 +1569,8 @@ void TpHostGpu::saveRgb(
(m_tpParams.img_width + m_tpParams.dtt_size) * sizeof(float),
m_tpParams.num_colors* (m_tpParams.img_height + m_tpParams.dtt_size),
cudaMemcpyDeviceToHost));
- printf("Writing %s to %s\n", prompt, paths[ncam]);
+ printf("Writing %s to %s (%d * %d * %d)\n", prompt, paths[ncam],
+ (m_tpParams.img_width + m_tpParams.dtt_size), (m_tpParams.img_height + m_tpParams.dtt_size), m_tpParams.num_colors);
writeFloatsToFile( // will have margins
cpu_corr_image, // float * data, // allocated array
rslt_img_size, // int size, // length in elements
@@ -1689,7 +1670,9 @@ void TpHostGpu::saveInterCorrFile(
cpu_corr_indices, // int * cpu_corr_indices,
cpu_corr_td, // float * cpu_corr,
num_sel_sensors); // int num_sel_sensors)
- printf("Writing %s TD data to %s\n", prompt, path);
+ int num_slices = num_corr_indices / (m_tpParams.tilesx * m_tpParams.tilesy);
+ printf("Writing %s TD data to %s (%d x %d x %d)\n", prompt, path,
+ m_tpParams.tilesx * m_tpParams.dtt_size2 , m_tpParams.tilesy * m_tpParams.dtt_size2 , num_slices);
writeFloatsToFile(
corr_img_td, // float * data, // allocated array
corr_img_size, // int size, // length in elements
@@ -1727,11 +1710,13 @@ void TpHostGpu::saveInterCorrIndicesFile(
int tx = ctt % m_tpParams.tilesx;
corr_index_img[cpair * m_tpParams.tilesx * m_tpParams.tilesy + m_tpParams.tilesx * ty + tx] = ict; // cpu_corr_indices[ict];
}
- printf("Writing %s to %s\n",prompt,path);
+ int num_slices = num_corr_indices / (m_tpParams.tilesx * m_tpParams.tilesy);
+ printf("Writing %s to %s (%d x %d x %d - not an image)\n",prompt,path,
+ m_tpParams.tilesx, m_tpParams.tilesy, num_slices);
writeFloatsToFile(
corr_index_img, // float * data, // allocated array
corr_index_img_length, // int size, // length in elements
- "clt/aux_inter-indices.raw"); // const char * path) // file path
+ path); // const char * path) // file path
free (corr_index_img);
free (cpu_corr_indices);
}
diff --git a/src/test_tp.cu b/src/test_tp.cu
index c88b12353559c479be2e47a26df3cd83578c7c88..194013283d4f3cf5c289b94fae8f01bce96cdd03 100644
--- a/src/test_tp.cu
+++ b/src/test_tp.cu
@@ -126,7 +126,7 @@ int main(int argc, char **argv)
0.0, // const float target_disparity, // DBG_DISPARITY == 0.0
0.0, // const float scale, // 0.0
0, // int quad_combine,
- 0, // int use_dp,
+ 1, // int use_dp,
0); // int debug);
return 0;
/* */