updated to jcuda 12.6

pom.xml

@@ -58,7 +58,8 @@
                 <groupId>org.jcuda</groupId>
                 <artifactId>jcuda</artifactId>
                 <!-- <version>10.1.0</version> -->
-                <version>11.2.0</version>
+                <!--<version>11.2.0</version> -->
+                <version>12.6.0</version>
             </dependency>
             <!--
             	As of 2018/09/11 TF for GPU on Maven supports CUDA 9.0 (vs latest 9.2)

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

@@ -6,7 +6,7 @@ package com.elphel.imagej.gpu;
 ** GPU acceleration for the Tile Processor
 **
 **
-** Copyright (C) 2018 Elphel, Inc.
+** Copyright (C) 2018-2025 Elphel, Inc.
 **
 ** -----------------------------------------------------------------------------**
 **
@@ -72,16 +72,18 @@ import jcuda.nvrtc.JNvrtc;
 import jcuda.nvrtc.nvrtcProgram;
 
 public class GPUTileProcessor {
-	public static boolean USE_DS_DP = false; // Use Dynamic Shared memory with Dynamic Parallelism (not implemented)  
+	public static boolean USE_DS_DP = true; // false; // Use Dynamic Shared memory with Dynamic Parallelism (not implemented)  
 	String LIBRARY_PATH = "/usr/local/cuda/targets/x86_64-linux/lib/libcudadevrt.a"; // linux
 	// Can be downloaded and twice extracted from
 	// https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-cudart-dev-11-2_11.2.152-1_amd64.deb
 	// First deb itself, then data.tar.xz, and it will have usr/local/cuda/targets/x86_64-linux/lib/libcudadevrt.a inside
 	// Found "cuda-cudart-dev" on https://ubuntu.pkgs.org/
 	static String GPU_RESOURCE_DIR =              "kernels";
-	static String [] GPU_KERNEL_FILES = {"dtt8x8.cuh","TileProcessor.cuh"};
+//	static String [] GPU_KERNEL_FILES = {"dtt8x8.cuh","TileProcessor.cuh"}; // was never used and dtt8x8.cuh had incorrect name
+//	static String [] GPU_KERNEL_FILES = {"dtt8x8.cu","TileProcessor.cu"};
 	// "*" - generated defines, first index - separately compiled unit
-	static String [][] GPU_SRC_FILES = {{"*","dtt8x8.h","dtt8x8.cu","geometry_correction.h","geometry_correction.cu","TileProcessor.h","TileProcessor.cuh"}};
+//	static String [][] GPU_SRC_FILES = {{"*","dtt8x8.h","dtt8x8.cu","geometry_correction.h","geometry_correction.cu","TileProcessor.h","TileProcessor.cuh"}};
+	static String [][] GPU_SRC_FILES = {{"*","dtt8x8.h","dtt8x8.cu","geometry_correction.h","geometry_correction.cu","TileProcessor.h","TileProcessor.cu"}};
 	static String GPU_CONVERT_DIRECT_NAME =        "convert_direct";      // name in C code
 	static String GPU_IMCLT_ALL_NAME =             "imclt_rbg_all";
 	static String GPU_CORRELATE2D_NAME =           "correlate2D";         // name in C code
@@ -89,7 +91,7 @@ public class GPUTileProcessor {
 	static String GPU_CORR2D_COMBINE_NAME =        "corr2D_combine";      // name in C code
 	static String GPU_CORR2D_NORMALIZE_NAME =      "corr2D_normalize";    // name in C code
 	static String GPU_TEXTURES_NAME =              "textures_nonoverlap"; // name in C code
-	static String GPU_RBGA_NAME =                  "generate_RBGA";       // name in C code
+	static String GPU_RBGA_NAME =                  "generate_RBGA";       // name in C code //// *** Modified 2025 *** ////
 	static String GPU_ROT_DERIV =                  "calc_rot_deriv";      // calculate rotation matrices and derivatives
 	static String GPU_SET_TILES_OFFSETS =          "get_tiles_offsets";   // calculate pixel offsets and disparity distortions
 	static String GPU_CALCULATE_TILES_OFFSETS =    "calculate_tiles_offsets";   // calculate pixel offsets and disparity distortions
@@ -100,7 +102,7 @@ public class GPUTileProcessor {
 	static String GPU_MARK_TEXTURE_NEIGHBOR_NAME = "mark_texture_neighbor_tiles";
 	static String GPU_GEN_TEXTURE_LIST_NAME =      "gen_texture_list";
 	static String GPU_CLEAR_TEXTURE_RBGA_NAME =    "clear_texture_rbga";
-	static String GPU_TEXTURES_ACCUMULATE_NAME =   "textures_accumulate";
+	static String GPU_TEXTURES_ACCUMULATE_NAME =   "textures_accumulate";  //// *** Modified 2025 *** ////
 	static String GPU_CREATE_NONOVERLAP_LIST_NAME ="create_nonoverlap_list";
 	static String GPU_ERASE_CLT_TILES_NAME =       "erase_clt_tiles";
 
@@ -298,7 +300,7 @@ public class GPUTileProcessor {
         ClassLoader classLoader = getClass().getClassLoader();
 
         String [] kernelSources = new String[GPU_SRC_FILES.length];
-        boolean show_source = false; // true;
+        boolean show_source = true; // false; // true;
         for (int cunit = 0; cunit < kernelSources.length; cunit++) {
         	kernelSources[cunit] = ""; // use StringBuffer?
             for (String src_file:GPU_SRC_FILES[cunit]) {
@@ -370,7 +372,7 @@ public class GPUTileProcessor {
         GPU_CORR2D_COMBINE_kernel =          functions[4];
         GPU_CORR2D_NORMALIZE_kernel =        functions[5];
         GPU_TEXTURES_kernel=                 functions[6];
-        GPU_RBGA_kernel=                     functions[7];
+        GPU_RBGA_kernel=                     functions[7];  //// *** Modified 2025 *** ////
         GPU_ROT_DERIV_kernel =               functions[8];
         GPU_CALCULATE_TILES_OFFSETS_kernel = functions[9];
         GPU_CALC_REVERSE_DISTORTION_kernel = functions[10];
@@ -380,7 +382,7 @@ public class GPUTileProcessor {
         GPU_MARK_TEXTURE_NEIGHBOR_kernel =   functions[13];
         GPU_GEN_TEXTURE_LIST_kernel =        functions[14];
         GPU_CLEAR_TEXTURE_RBGA_kernel =      functions[15];
-        GPU_TEXTURES_ACCUMULATE_kernel =     functions[16];
+        GPU_TEXTURES_ACCUMULATE_kernel =     functions[16]; //// *** Modified 2025 *** ////
         GPU_CREATE_NONOVERLAP_LIST_kernel =  functions[17];
         GPU_ERASE_CLT_TILES_kernel =         functions[18];
         
@@ -504,7 +506,7 @@ public class GPUTileProcessor {
     		// Use the NVRTC to create a program by compiling the source code
     		nvrtcProgram program = new nvrtcProgram();
     		nvrtcCreateProgram(	program, sourceCode, null, 0, null, null);
-    		String options[] = {"--gpu-architecture=compute_"+capability};
+    		String options[] = {"--gpu-architecture=compute_"+capability,"--extensible-whole-program"};
 
     		try {
     			nvrtcCompileProgram(program, options.length, options);