Commit fdc9840a authored by Andrey Filippov's avatar Andrey Filippov
Browse files

adding generation of data for macroblocks

parent ee71f035

src/TileProcessor.cuh

+11 −8
Original line number Original line Diff line number Diff line
@@ -1196,7 +1196,8 @@ __global__ void generate_RBGA( 
			int               dust_remove,        // Do not reduce average weight when only one image differs much from the average

			int               dust_remove,        // Do not reduce average weight when only one image differs much from the average

			int               keep_weights,       // return channel weights after A in RGBA (was removed)

			int               keep_weights,       // return channel weights after A in RGBA (was removed)

			const size_t      texture_rbga_stride,     // in floats

			const size_t      texture_rbga_stride,     // in floats

			float           * gpu_texture_tiles)  // (number of colors +1 + ?)*16*16 rgba texture tiles

			float           * gpu_texture_tiles,  // (number of colors +1 + ?)*16*16 rgba texture tiles

			float           * gpu_diff_rgb_combo) // diff[NUM_CAMS], R[NUM_CAMS], B[NUM_CAMS],G[NUM_CAMS]

{

{

// TODO use atomic_add to increment	num_texture_tiles

// TODO use atomic_add to increment	num_texture_tiles

// TODO calculate woi

// TODO calculate woi
@@ -1329,7 +1330,9 @@ __global__ void generate_RBGA( 
						texture_rbga_stride,             // size_t      texture_rbg_stride, // in floats

						texture_rbga_stride,             // size_t      texture_rbg_stride, // in floats

						gpu_texture_tiles,               // float           * gpu_texture_rbg,     // (number of colors +1 + ?)*16*16 rgba texture tiles

						gpu_texture_tiles,               // float           * gpu_texture_rbg,     // (number of colors +1 + ?)*16*16 rgba texture tiles

			    		0,                               // size_t      texture_stride,     // in floats (now 256*4 = 1024)

			    		0,                               // size_t      texture_stride,     // in floats (now 256*4 = 1024)

						gpu_texture_tiles); // (float *) 0 );                   // float           * gpu_texture_tiles);  // (number of colors +1 + ?)*16*16 rgba texture tiles

						gpu_texture_tiles, //(float *)0);// float           * gpu_texture_tiles);  // (number of colors +1 + ?)*16*16 rgba texture tiles

						gpu_diff_rgb_combo);             // float           * gpu_diff_rgb_combo) // diff[NUM_CAMS], R[NUM_CAMS], B[NUM_CAMS],G[NUM_CAMS]



				cudaDeviceSynchronize(); // not needed yet, just for testing

				cudaDeviceSynchronize(); // not needed yet, just for testing

				/* */

				/* */

			}

			}
@@ -1788,8 +1791,8 @@ __global__ void textures_accumulate( 
		size_t            texture_rbg_stride, // in floats

		size_t            texture_rbg_stride, // in floats

		float           * gpu_texture_rbg,    // (number of colors +1 + ?)*16*16 rgba texture tiles

		float           * gpu_texture_rbg,    // (number of colors +1 + ?)*16*16 rgba texture tiles

		size_t            texture_stride,     // in floats (now 256*4 = 1024)

		size_t            texture_stride,     // in floats (now 256*4 = 1024)

		float           * gpu_texture_tiles)  // (number of colors +1 + ?)*16*16 rgba texture tiles

		float           * gpu_texture_tiles,  // (number of colors +1 + ?)*16*16 rgba texture tiles



		float           * gpu_diff_rgb_combo) // diff[NUM_CAMS], R[NUM_CAMS], B[NUM_CAMS],G[NUM_CAMS]

{

{

	//						(float *) gpu_geometry_correction ->pXY0,

	//						(float *) gpu_geometry_correction ->pXY0,

//	float weights[3] = {weight0, weight1, weight2};

//	float weights[3] = {weight0, weight1, weight2};
@@ -1997,8 +2000,8 @@ __global__ void textures_accumulate( 
			(float*) shr.mclt_debayer, // float * mclt_tile,     // debayer // has gaps to align with union !

			(float*) shr.mclt_debayer, // float * mclt_tile,     // debayer // has gaps to align with union !

			(float*) mclt_tiles,       // float * rbg_tile,      // if not null - original (not-debayered) rbg tile to use for the output

			(float*) mclt_tiles,       // float * rbg_tile,      // if not null - original (not-debayered) rbg tile to use for the output

			(float *) shr1.rgbaw,      // float * rgba,          // result

			(float *) shr1.rgbaw,      // float * rgba,          // result

			(float * ) 0,              // float * ports_rgb,     // average values of R,G,B for each camera (R0,R1,...,B2,B3) // null

			(float * ) ports_rgb,      // float * ports_rgb,     // average values of R,G,B for each camera (R0,R1,...,B2,B3) // null

			(float * ) 0,              // float * max_diff,      // maximal (weighted) deviation of each channel from the average /null

			(float * ) max_diff,       // float * max_diff,      // maximal (weighted) deviation of each channel from the average /null

			(float *) port_offsets,    // float * port_offsets,  // [port]{x_off, y_off} - just to scale pixel value differences

			(float *) port_offsets,    // float * port_offsets,  // [port]{x_off, y_off} - just to scale pixel value differences

			diff_sigma,                // float   diff_sigma,     // pixel value/pixel change

			diff_sigma,                // float   diff_sigma,     // pixel value/pixel change

			diff_threshold,            // float   diff_threshold, // pixel value/pixel change

			diff_threshold,            // float   diff_threshold, // pixel value/pixel change
@@ -2013,8 +2016,8 @@ __global__ void textures_accumulate( 
			(float*) shr.mclt_debayer, // float * mclt_tile,     // debayer // has gaps to align with union !

			(float*) shr.mclt_debayer, // float * mclt_tile,     // debayer // has gaps to align with union !

			(float*) mclt_tiles,       // float * rbg_tile,      // if not null - original (not-debayered) rbg tile to use for the output

			(float*) mclt_tiles,       // float * rbg_tile,      // if not null - original (not-debayered) rbg tile to use for the output

			(float *) shr1.rgbaw,      // float * rgba,          // result

			(float *) shr1.rgbaw,      // float * rgba,          // result

			(float * ) 0,              // float * ports_rgb,     // average values of R,G,B for each camera (R0,R1,...,B2,B3) // null

			(float * ) ports_rgb,      // float * ports_rgb,     // average values of R,G,B for each camera (R0,R1,...,B2,B3) // null

			(float * ) 0,              // float * max_diff,      // maximal (weighted) deviation of each channel from the average /null

			(float * ) max_diff,       // float * max_diff,      // maximal (weighted) deviation of each channel from the average /null

			(float *) port_offsets,    // float * port_offsets,  // [port]{x_off, y_off} - just to scale pixel value differences

			(float *) port_offsets,    // float * port_offsets,  // [port]{x_off, y_off} - just to scale pixel value differences

			diff_sigma,                // float   diff_sigma,     // pixel value/pixel change

			diff_sigma,                // float   diff_sigma,     // pixel value/pixel change

			diff_threshold,            // float   diff_threshold, // pixel value/pixel change

			diff_threshold,            // float   diff_threshold, // pixel value/pixel change

src/TileProcessor.h

+3 −1
Original line number Original line Diff line number Diff line
@@ -96,7 +96,9 @@ extern "C" __global__ void textures_accumulate( 
		size_t            texture_rbg_stride, // in floats

		size_t            texture_rbg_stride, // in floats

		float           * gpu_texture_rbg,    // (number of colors +1 + ?)*16*16 rgba texture tiles

		float           * gpu_texture_rbg,    // (number of colors +1 + ?)*16*16 rgba texture tiles

		size_t            texture_stride,     // in floats (now 256*4 = 1024)

		size_t            texture_stride,     // in floats (now 256*4 = 1024)

		float           * gpu_texture_tiles);  // (number of colors +1 + ?)*16*16 rgba texture tiles

		float           * gpu_texture_tiles,  // (number of colors +1 + ?)*16*16 rgba texture tiles

		float           * gpu_diff_rgb_combo); // diff[NUM_CAMS], R[NUM_CAMS], B[NUM_CAMS],G[NUM_CAMS]







extern "C"

extern "C"

__global__ void imclt_rbg_all(

__global__ void imclt_rbg_all(

src/test_tp.cu

+9 −4
Original line number Original line Diff line number Diff line
@@ -341,6 +341,7 @@ int main(int argc, char **argv) 
    int              * gpu_corr_indices;

    int              * gpu_corr_indices;





    float            * gpu_textures;

    float            * gpu_textures;

    float            * gpu_diff_rgb_combo;

    float            * gpu_textures_rbga;

    float            * gpu_textures_rbga;

    int              * gpu_texture_indices;

    int              * gpu_texture_indices;

    int              * gpu_woi;

    int              * gpu_woi;
@@ -587,7 +588,8 @@ int main(int argc, char **argv) 
    		&dstride_textures_rbga,              // in bytes ! for one rgba/ya 16x16 tile

    		&dstride_textures_rbga,              // in bytes ! for one rgba/ya 16x16 tile

			rgba_width,              // int width (floats),

			rgba_width,              // int width (floats),

			rgba_height * rbga_slices);               // int height);

			rgba_height * rbga_slices);               // int height);



//    checkCudaErrors(cudaMalloc((void **)&gpu_diff_rgb_combo,  TILESX * TILESY * NUM_CAMS * (NUM_COLS+1)* sizeof(float)));

    checkCudaErrors(cudaMalloc((void **)&gpu_diff_rgb_combo,  TILESX * TILESY * NUM_CAMS * (NUM_COLORS + 1) * sizeof(float)));





    // Now copy arrays of per-camera pointers to GPU memory to GPU itself

    // Now copy arrays of per-camera pointers to GPU memory to GPU itself
@@ -1094,7 +1096,9 @@ int main(int argc, char **argv) 
    			0, // const size_t      texture_rbg_stride, // in floats

    			0, // const size_t      texture_rbg_stride, // in floats

    			(float *) 0, // float           * gpu_texture_rbg,     // (number of colors +1 + ?)*16*16 rgba texture tiles

    			(float *) 0, // float           * gpu_texture_rbg,     // (number of colors +1 + ?)*16*16 rgba texture tiles

				dstride_textures/sizeof(float), // const size_t      texture_stride,     // in floats (now 256*4 = 1024)

				dstride_textures/sizeof(float), // const size_t      texture_stride,     // in floats (now 256*4 = 1024)

				gpu_textures);    // float           * gpu_texture_tiles);  // 4*16*16 rgba texture tiles

				gpu_textures,    // float           * gpu_texture_tiles);  // 4*16*16 rgba texture tiles

				gpu_diff_rgb_combo);             // float           * gpu_diff_rgb_combo) // diff[NUM_CAMS], R[NUM_CAMS], B[NUM_CAMS],G[NUM_CAMS]



    	getLastCudaError("Kernel failure");

    	getLastCudaError("Kernel failure");

    	checkCudaErrors(cudaDeviceSynchronize());

    	checkCudaErrors(cudaDeviceSynchronize());

    	printf("test pass: %d\n",i);

    	printf("test pass: %d\n",i);
@@ -1271,7 +1275,8 @@ int main(int argc, char **argv) 
	            1,                     // int               dust_remove,        // Do not reduce average weight when only one image differes much from the average

	            1,                     // int               dust_remove,        // Do not reduce average weight when only one image differes much from the average

	            0,                     // int               keep_weights,       // return channel weights after A in RGBA

	            0,                     // int               keep_weights,       // return channel weights after A in RGBA

				dstride_textures_rbga/sizeof(float), // 	const size_t      texture_rbga_stride,     // in floats

				dstride_textures_rbga/sizeof(float), // 	const size_t      texture_rbga_stride,     // in floats

				gpu_textures_rbga);    // 	float           * gpu_texture_tiles)    // (number of colors +1 + ?)*16*16 rgba texture tiles

				gpu_textures_rbga,     // 	float           * gpu_texture_tiles)    // (number of colors +1 + ?)*16*16 rgba texture tiles

				gpu_diff_rgb_combo);   // float           * gpu_diff_rgb_combo) // diff[NUM_CAMS], R[NUM_CAMS], B[NUM_CAMS],G[NUM_CAMS]





    	getLastCudaError("Kernel failure");

    	getLastCudaError("Kernel failure");

    	checkCudaErrors(cudaDeviceSynchronize());

    	checkCudaErrors(cudaDeviceSynchronize());
@@ -1362,9 +1367,9 @@ int main(int argc, char **argv) 
	checkCudaErrors(cudaFree(gpu_color_weights));

	checkCudaErrors(cudaFree(gpu_color_weights));

	checkCudaErrors(cudaFree(gpu_textures));

	checkCudaErrors(cudaFree(gpu_textures));

	checkCudaErrors(cudaFree(gpu_textures_rbga));

	checkCudaErrors(cudaFree(gpu_textures_rbga));

	checkCudaErrors(cudaFree(gpu_diff_rgb_combo));

	checkCudaErrors(cudaFree(gpu_woi));

	checkCudaErrors(cudaFree(gpu_woi));

	checkCudaErrors(cudaFree(gpu_num_texture_tiles));

	checkCudaErrors(cudaFree(gpu_num_texture_tiles));



	checkCudaErrors(cudaFree(gpu_geometry_correction));

	checkCudaErrors(cudaFree(gpu_geometry_correction));

    checkCudaErrors(cudaFree(gpu_correction_vector));

    checkCudaErrors(cudaFree(gpu_correction_vector));

    checkCudaErrors(cudaFree(gpu_rByRDist));

    checkCudaErrors(cudaFree(gpu_rByRDist));