updating for CUDA CDP2

pom.xml

@@ -113,6 +113,15 @@
 				<artifactId>loci_tools</artifactId>
 				<version>6.1.0</version>
 			</dependency>
+			<!-- https://mvnrepository.com/artifact/ome/pom-bio-formats -->
+			<!-- Was source in attic for development -->
+			<dependency>
+			    <groupId>ome</groupId>
+			    <artifactId>pom-bio-formats</artifactId>
+			    <version>6.13.0</version>
+			    <type>pom</type>
+			</dependency>
+			
 <!-- 			
 			<dependency>
 			  <groupId>com.drewnoakes</groupId>

src/main/java/com/elphel/imagej/gpu/GpuQuad.java

@@ -2886,11 +2886,19 @@ public class GpuQuad{ // quad camera description
 			int ntt = cpu_num_texture_tiles[((pass & 3) << 1) + border_tile];
 			if (ntt > 0) {
 				int [] grid_texture = {(ntt + GPUTileProcessor.TEXTURE_TILES_PER_BLOCK-1) / GPUTileProcessor.TEXTURE_TILES_PER_BLOCK,1,1}; // TEXTURE_TILES_PER_BLOCK = 1
-
+				/* Pre CDP2
 				int ti_offset = (pass & 3) * (width * (tilesya >> 2)); //  (TILES-X * (TILES-YA >> 2));  // 1/4
 				if (border_tile != 0){
 					ti_offset += width * (tilesya >> 2) - ntt; // TILES-X * (TILES-YA >> 2) - ntt;
 				}
+				*/
+				// for CDP2
+			    int ti_offset = (pass & 3) * (width * (tilesya >> 2)); //  (TILES-X * (TILES-YA >> 2));  // 1/4
+			    if (border_tile != 0){
+//			    	ti_offset += width * (tilesya >> 2) - ntt; // TILES-X * (TILES-YA >> 2) - ntt;
+			    	ti_offset += width * (tilesya >> 2); // TILES-X * (TILES-YA >> 2) - ntt;
+			    	ti_offset = - ti_offset; // does not depend on results of the previous kernel, but is negative
+			    }
 				int shared_size = host_get_textures_shared_size( // in bytes
 						num_cams,     // int                num_cams,     // actual number of cameras
 						num_colors,   // int                num_colors,   // actual number of colors: 3 for RGB, 1 for LWIR/mono
@@ -2923,8 +2931,11 @@ public class GpuQuad{ // quad camera description
 						Pointer.to(new int[] {num_cams}),                // int         num_cams,
 						Pointer.to(gpu_woi),                             // int       * woi,            // min_x, min_y, max_x, max_y
 						Pointer.to(gpu_clt),                             // float    ** gpu_clt,        // [num_cams] ->[TILESY][TILESX][num_colors][DTT_SIZE*DTT_SIZE]
-						Pointer.to(new int[] {ntt}),                     // size_t      num_texture_tiles,// number of texture tiles to process
-						Pointer.to(new int[] {ti_offset}),                     // size_t      num_texture_tiles,// number of texture tiles to process
+//						Pointer.to(new int[] {ntt}),                     // size_t      num_texture_tiles,// number of texture tiles to process
+//						Pointer.to(gpu_num_texture_tiles[((pass & 3) << 1) + border_tile)]),               // int       * num_texture_tiles,// number of texture tiles to process
+						Pointer.to(gpu_num_texture_tiles).withByteOffset(
+								(((pass & 3) << 1) + border_tile)*Sizeof.INT),               // int       * num_texture_tiles,// number of texture tiles to process
+						Pointer.to(new int[] {ti_offset}),               // int         gpu_texture_indices_offset, // add to gpu_texture_indices (now complicated: if negative - add *(pnum_texture_tiles) and negate
 						Pointer.to(gpu_texture_indices_ovlp),            //  gpu_texture_indices_offset,// add to gpu_texture_indices
 						Pointer.to(gpu_geometry_correction),             // struct gc * gpu_geometry_correction,
 						Pointer.to(new int[]   {num_colors}),            // int         colors,         // number of colors (3/1)
@@ -3174,7 +3185,9 @@ public class GpuQuad{ // quad camera description
 				Pointer.to(new int[] {num_cams}),                // int         num_cams,
 				Pointer.to(new int[]   {0}),  // Pointer.to(gpu_woi),  // int       * woi,            // min_x, min_y, max_x, max_y
 				Pointer.to(gpu_clt),                             // float    ** gpu_clt,        // [num_cams] ->[TILESY][TILESX][num_colors][DTT_SIZE*DTT_SIZE]
-				Pointer.to(new int[] {cpu_pnum_texture_tiles[0]}),                     // size_t      num_texture_tiles,// number of texture tiles to process
+//				Pointer.to(new int[] {cpu_pnum_texture_tiles[0]}),                     // size_t      num_texture_tiles,// number of texture tiles to process
+//				Pointer.to(new int[] {ntt}),                     // size_t      num_texture_tiles,// number of texture tiles to process
+				Pointer.to(gpu_num_texture_tiles),               // int       * num_texture_tiles,// number of texture tiles to process
 				Pointer.to(new int[] {0}),                     // size_t      num_texture_tiles,// number of texture tiles to process
 				Pointer.to(gpu_texture_indices),          // int   * nonoverlap_list,    // pointer to the calculated number of non-zero tiles
 				Pointer.to(gpu_geometry_correction),             // struct gc * gpu_geometry_correction,