Commit adac766c authored by Andrey Filippov's avatar Andrey Filippov
Browse files

fixed NaN by adding fat zero to channel weights after exp()

parent edce489f

src/TileProcessor.cuh

+58 −28
Original line number Diff line number Diff line
@@ -41,22 +41,22 @@ 
#pragma once

#include "dtt8x8.cuh"

#define THREADSX         (DTT_SIZE)

#define NUM_CAMS                  4

#define NUM_PAIRS                 6

#define NUM_COLORS                3

#define IMG_WIDTH              2592

#define IMG_HEIGHT             1936

#define KERNELS_HOR             164

#define KERNELS_VERT            123

#define NUM_CAMS                  4

#define NUM_PAIRS                 6

#define NUM_COLORS                3

#define KERNELS_LSTEP             4

#define THREADS_PER_TILE          8

#define TILES_PER_BLOCK           4

#define CORR_THREADS_PER_TILE     8

#define CORR_TILES_PER_BLOCK      4

#define IMCLT_THREADS_PER_TILE   16

#define IMCLT_TILES_PER_BLOCK     4

#define TEXTURE_THREADS_PER_TILE  8

#define TEXTURE_TILES_PER_BLOCK   1

#define IMCLT_THREADS_PER_TILE   16

#define IMCLT_TILES_PER_BLOCK     4

#define CORR_NTILE_SHIFT          8 // higher bits - number of a pair, other bits tile number

#define CORR_PAIRS_MASK        0x3f// lower bits used to address correlation pair for the selected tile

#define CORR_TEXTURE_BIT          7 // bit 7 used to request texture for the tile
@@ -64,6 +64,7 @@ 
#define TASK_TEXTURE_BIT          3 // bit to request texture calculation int task field of struct tp_task

#define LIST_TEXTURE_BIT          7 // bit to request texture calculation

#define CORR_OUT_RAD              4

#define FAT_ZERO_WEIGHT           0.0001 // add to port weights to avoid nan



//7

//#define DEBUG1 1
@@ -74,7 +75,7 @@ 
//#define DEBUG6 1

#define DEBUG7 1

#define DEBUG8 1

//#define DEBUG9 1

#define DEBUG9 1





#endif
@@ -146,10 +147,10 @@ 
#define BAYER_RED_COL 1

//#define BAYER_BLUE_ROW (1 - BAYER_RED_ROW)

//#define BAYER_BLUE_COL (1 - BAYER_RED_COL)





#define DBG_TILE_X     40

#define DBG_TILE_Y     80

//#define DBG_TILE_X     40

//#define DBG_TILE_Y     80

#define DBG_TILE_X     49

#define DBG_TILE_Y     66



#define DBG_TILE     (DBG_TILE_Y * 324 + DBG_TILE_X)

//56494
@@ -503,6 +504,7 @@ __device__ void debayer_shot( 
__device__ void tile_combine_rgba(

		int     colors,        // number of colors

		float * mclt_tile,     // debayer

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

		float * rgba,          // result

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

		float * max_diff,      // maximal (weighted) deviation of each channel from the average /null
@@ -803,7 +805,7 @@ __global__ void convert_correct_tiles( 
			float           ** gpu_clt,            // [NUM_CAMS][TILESY][TILESX][NUM_COLORS][DTT_SIZE*DTT_SIZE]

			size_t             dstride,            // in floats (pixels)

			int                num_tiles,          // number of tiles in task

			int                lpf_mask)           // apply lpf to colors : bit 0 - red, bit 1 - blue, bit2 - green

			int                lpf_mask)           // apply lpf to colors : bit 0 - red, bit 1 - blue, bit2 - green. Now - always 0 !

{

	dim3 t = threadIdx;

	int tile_in_block = threadIdx.y;
@@ -902,7 +904,11 @@ __global__ void textures_gen( 
		return; // nothing to do

	}

	// get number of tile

	int tile_num = (gpu_texture_indices[tile_indx]) >> CORR_NTILE_SHIFT;

	int tile_code = gpu_texture_indices[tile_indx];

	if ((tile_code & (1 << CORR_TEXTURE_BIT)) == 0){

		return; // nothing to do

	}

	int tile_num = tile_code >> CORR_NTILE_SHIFT;

	__shared__ float mclt_tiles [NUM_CAMS][NUM_COLORS][2*DTT_SIZE][DTT_SIZE21];

	__shared__ union {

		float clt_tiles  [NUM_CAMS][NUM_COLORS][4][DTT_SIZE][DTT_SIZE1]; // NUM_CAMS == 4
@@ -1000,6 +1006,20 @@ __global__ void textures_gen( 
			__syncthreads();// __syncwarp();

		} else {

			// copy? - no, just remember to use mclt_tile, not mclt_dst

			// will have to copy mclt_tiles -> mclt_dst as they have different gaps

			// untested copy for mono mode

#pragma unroll

			for (int n = 0; n <= DTT_SIZE; n += DTT_SIZE){

				float * msp = mclt_tile + threadIdx.x + n;

				float * dst = mclt_dst +  threadIdx.x + n;

#pragma unroll

				for (int row = 0; row < DTT_SIZE2; row++){

					*dst = *msp;

					msp += DTT_SIZE21;

					dst += DTT_SIZE21;

				}

			}

			__syncthreads();

		}

#ifdef DEBUG77

		//		float * mclt_dst =  (float *) shr.mclt_debayer[camera_num][color];
@@ -1055,10 +1075,11 @@ __global__ void textures_gen( 
	}

	__syncthreads();// __syncwarp();

#endif



//	__shared__ float mclt_tiles [NUM_CAMS][NUM_COLORS][2*DTT_SIZE][DTT_SIZE21];

	tile_combine_rgba(

			colors,                    // int     colors,        // number of colors

			(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 *) 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 * ) 0,              // float * max_diff,      // maximal (weighted) deviation of each channel from the average /null
@@ -1638,7 +1659,7 @@ __device__ void convertCorrectTile( 
		float               * gpu_images,

		float               * gpu_clt,

		const int             color,

		const int             lpf_mask,

		const int             lpf_mask, // now 0

		const float           centerX,

		const float           centerY,

		const int             txy,
@@ -2711,6 +2732,7 @@ __device__ void debayer( 
__device__ void tile_combine_rgba(

		int     colors,        // number of colors

		float * mclt_tile,     // debayer // has gaps to align with union !

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

		float * rgba,          // result

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

		float * max_diff,      // maximal (weighted) deviation of each channel from the average /null
@@ -2858,7 +2880,7 @@ __device__ void tile_combine_rgba( 
				dc *= wnd2_inv; // to compensate fading near the edges

				d+= *(chn_weights + ncol) * dc * dc;

			}

			d = expf(-pair_dist2r[ipair] * d); // 0.5 for exact match, lower for mismatch. Add this weight to both ports involved

			d = expf(-pair_dist2r[ipair] * d) + (FAT_ZERO_WEIGHT); // 0.5 for exact match, lower for mismatch. Add this weight to both ports involved

			// Add weight to both channels in a pair

			*(port_weights_i + (DTT_SIZE2*DTT_SIZE21) * pair_ports[ipair][0]) +=d;

			*(port_weights_i + (DTT_SIZE2*DTT_SIZE21) * pair_ports[ipair][1]) +=d;
@@ -2964,7 +2986,7 @@ __device__ void tile_combine_rgba( 
			}

			// TODO: Should it use pair_dist2r ? no as it is relative?

			//				port_weights[ip][i] = Math.exp(-ksigma * d2[ip]);

			*(port_weights_i + (DTT_SIZE2*DTT_SIZE21) * cam) = expf(-ksigma * d2_ip);

			*(port_weights_i + (DTT_SIZE2*DTT_SIZE21) * cam) = expf(-ksigma * d2_ip) + (FAT_ZERO_WEIGHT);

		}

		// and now make a new average with those weights

		// Inserting dust remove here
@@ -3033,22 +3055,30 @@ __device__ void tile_combine_rgba( 




		///

		if (rbg_tile) {

			float k = 0.0;

			int rbga_offset = colors * (DTT_SIZE2*DTT_SIZE21); // padded in union !

#pragma unroll

			for (int cam = 0; cam < NUM_CAMS; cam++){

				k += *(port_weights_i + (DTT_SIZE2*DTT_SIZE21) * cam); // port_weights[ip][i];

			}

			k = 1.0/k;

			float * rbg_tile_i = rbg_tile + i;

#pragma unroll  // non-constant

			for (int ncol = 0; ncol < colors; ncol++) { // if (iclt_tile[0][ncol] != null) {

				float * rgba_col_i = rgba_i + ncol * (DTT_SIZE2*DTT_SIZE21);

			float * mclt_col_i = mclt_tile_i + MCLT_UNION_LEN * ncol;

//				float * mclt_col_i = mclt_tile_i + MCLT_UNION_LEN * ncol;

				float * rbg_col_i = rbg_tile_i + ncol * (DTT_SIZE2*DTT_SIZE21); // different gap between tiles than MCLT_UNION_LEN

				*rgba_col_i = 0.0; // color_avg[ncol][i] = 0;

#pragma unroll

				for (int cam = 0; cam < NUM_CAMS; cam++) {

				*rgba_col_i += k * *(port_weights_i + (DTT_SIZE2*DTT_SIZE21) * cam) * *(mclt_col_i + cam * colors_offset);

//					*rgba_col_i += k * *(port_weights_i + (DTT_SIZE2*DTT_SIZE21) * cam) * *(mclt_col_i + cam * colors_offset);

					*rgba_col_i += k * *(port_weights_i + (DTT_SIZE2*DTT_SIZE21) * cam) * *(rbg_col_i +  cam * rbga_offset);

				}

			}

		}

//	int colors_offset = colors * MCLT_UNION_LEN; // padded in union !



		// calculate alpha from channel weights. Start with just a sum of weights?

//		int used_ports = NUM_CAMS;

//		if (dust_remove){