POL5892 fix gpu 2

CMakeLists.txt

@@ -57,6 +57,8 @@ if(USEGPU)
 		endif()
 	elseif (CUDAToolkit_VERSION_MAJOR STREQUAL "12")
 		set(CUDA_ARCH_LIST "50;52;53;60;61;62;70;72;75;80;86;89;90")
+	elseif (CUDAToolkit_VERSION_MAJOR STREQUAL "13")
+		set(CUDA_ARCH_LIST "75;80;86;87;88;89;90;100;103;120;121")
 	else() # some old CUDA version (<10)
 		set(CUDA_ARCH_LIST "50") # 5.0 is the oldest non-deprecated CC as of 2024-09-06
 	endif()
@@ -132,6 +134,9 @@ option(RUN_GTEST "Downloads google unit test API and runs google test scripts to
 
 #==== Source files
 set(SOURCE
+	# headers
+	src/nyx/raw_tiff.h
+	# sources
 	src/nyx/3rdparty/dsyevj3.cpp
 	src/nyx/features/basic_morphology.cpp
 	src/nyx/features/caliper_feret.cpp

src/nyx/cache.cpp

@@ -9,6 +9,11 @@
 	return false; \
 } \
 
+#define OKDETL(x,detl) if (x == false) \
+{ \
+	return std::string("error due to ") + detl + " at " + __FILE__ + ":" + std::to_string(__LINE__); \
+} \
+
 #define OKV(x) if (x == false) \
 { \
 	std::cerr << "gpu cache related error in " << __FILE__ << ":" << __LINE__; \
@@ -76,7 +81,7 @@
 		return szb;
 	}
 
-	bool GpusideCache::allocate_gpu_cache(
+	std::optional<std::string> GpusideCache::allocate_gpu_cache(
 		// out
 		GpuCache<Pixel2>& clouds,	// geo moments
 		GpuCache<Pixel2>& konturs,
@@ -107,14 +112,13 @@
 		int gabor_ker_side)
 	{
 		using_contour =
-		using_erosion =
-		using_gabor =
-		using_moments = false;
+			using_erosion =
+			using_gabor =
+			using_moments = false;
 
 		//****** plan GPU memory
-
 		size_t amt = 0;
-		OK(gpu_get_free_mem(amt));
+		OKDETL(gpu_get_free_mem(amt), "gpu_get_free_mem()");
 
 		int n_gabFilters = n_gabor_filters + 1;		// '+1': an extra filter for the baseline signal
 
@@ -130,7 +134,7 @@
 			n_rois, roi_w, roi_h, n_gabFilters, gabor_ker_side);
 
 		batch_len = n_rois;
-		size_t critAmt = amt * 0.75; // 75% GPU RAM as critical RAM
+		size_t critAmt = float(amt) * 0.75; // 75% GPU RAM as critical RAM
 
 		if (critAmt < szb)
 		{
@@ -139,10 +143,7 @@
 			{
 				// failed to find a batch ?
 				if (try_nrois == 0)
-				{
-					std::cerr << "error: cannot make a ROI batch \n";
-					return false;
-				}
+					return "cannot rightsize a ROI batch: even 1 ROI requires over critical memory amount " + std::to_string(critAmt);
 
 				size_t ccl = roi_area * try_nrois;	// combined cloud length
 				size_t try_szb = ram_comsumption_szb(
@@ -162,7 +163,7 @@
 			// have we found a compromise ?
 			if (batch_len < n_rois)
 			{
-				return false;
+				return "cannot rightsize a ROI batch: batch_len " + std::to_string(batch_len) + " < n_rois " + std::to_string(n_rois);
 			}
 		}
 
@@ -172,17 +173,17 @@
 
 		// ROI clouds (always on)
 
-		OK(clouds.clear());
-		OK(clouds.alloc(batch_roi_cloud_len, batch_len));
+		OKDETL(clouds.clear(), "clouds.clear()");
+		OKDETL(clouds.alloc(batch_roi_cloud_len, batch_len), "clouds.alloc()");
 
 		// contours
 
 		if (needContour)
 		{
 			using_contour = true;
 
-			OK(konturs.clear());
-			OK(konturs.alloc(batch_roi_cloud_len, batch_len));
+			OKDETL(konturs.clear(), "konturs.clear()");
+			OKDETL(konturs.alloc(batch_roi_cloud_len, batch_len), "konturs.alloc()");
 		}
 
 		// moments
@@ -192,17 +193,17 @@
 			using_moments = true;
 
 			// moments / real intensities
-			OK(allocate_on_device((void**)realintens, sizeof(RealPixIntens) * batch_roi_cloud_len));
+			OKDETL(allocate_on_device((void**)realintens, sizeof(RealPixIntens) * batch_roi_cloud_len), "allocate_on_device(moments1)");
 
 			// moments / pre-reduce
-			OK(allocate_on_device((void**)prereduce, sizeof(double) * batch_roi_cloud_len * 16));	// 16 is the max number of simultaneous totals calculated by a kernel, e.g. RM00-33
+			OKDETL(allocate_on_device((void**)prereduce, sizeof(double) * batch_roi_cloud_len * 16), "allocate_on_device(moments2)");	// 16 is the max number of simultaneous totals calculated by a kernel, e.g. RM00-33
 
 			// moments / intermediate
-			OK(intermediate.alloc(GpusideState::__COUNT__, batch_len));
+			OKDETL(intermediate.alloc(GpusideState::__COUNT__, batch_len), "allocate_on_device(moments3)");
 
 			// moments / CUB DeviceReduce's temp buffer
-			OK(devicereduce_evaluate_buffer_szb(devicereduce_buf_szb, batch_roi_cloud_len));
-			OK(allocate_on_device((void**)devicereduce_buf, devicereduce_buf_szb));
+			OKDETL(devicereduce_evaluate_buffer_szb(devicereduce_buf_szb, batch_roi_cloud_len), "allocate_on_device(moments4)");
+			OKDETL(allocate_on_device((void**)devicereduce_buf, devicereduce_buf_szb), "allocate_on_device(moments5)");
 		}
 
 		// erosion / image matrices 1 and 2
@@ -213,9 +214,9 @@
 
 			// imat1 is shared by erosion and Gabor
 			if (!*imat1)
-				OK(allocate_on_device((void**)imat1, sizeof(imat1[0]) * roi_w * roi_h));
+				OKDETL(allocate_on_device((void**)imat1, sizeof(imat1[0]) * roi_w * roi_h), "allocate_on_device(eros1)");
 
-			OK(allocate_on_device((void**)imat2, sizeof(imat2[0]) * roi_w * roi_h));
+			OKDETL(allocate_on_device((void**)imat2, sizeof(imat2[0]) * roi_w * roi_h), "allocate_on_device(eros2)");
 		}
 
 		// Gabor
@@ -226,16 +227,16 @@
 
 			// imat1 is shared by erosion and Gabor
 			if (!*imat1)
-				OK(allocate_on_device((void**)imat1, sizeof(imat1[0]) * roi_w * roi_h));
+				OKDETL(allocate_on_device((void**)imat1, sizeof(imat1[0]) * roi_w * roi_h), "allocate_on_device(gabor)");
 
 			size_t gabTotlen = (roi_w + gabor_ker_side - 1) * (roi_h + gabor_ker_side - 1) * n_gabFilters;
-			OK(gabor_linear_image.alloc(gabTotlen, 1));
-			OK(gabor_result.alloc(gabTotlen, 1));
-			OK(gabor_linear_kernel.alloc(gabTotlen, 1));
-			OK(gabor_energy_image.alloc(gabTotlen, 1));
+			OKDETL(gabor_linear_image.alloc(gabTotlen, 1), "gabor_linear_image.alloc()");
+			OKDETL(gabor_result.alloc(gabTotlen, 1), "gabor_result.alloc()");
+			OKDETL(gabor_linear_kernel.alloc(gabTotlen, 1), "gabor_linear_kernel.alloc()");
+			OKDETL(gabor_energy_image.alloc(gabTotlen, 1), "gabor_energy_image.alloc()");
 		}
 
-		return true;
+		return std::nullopt;
 	}
 
 	void GpusideCache::send_roi_batch_data_2_gpu(

src/nyx/cache.h

@@ -6,6 +6,9 @@
 	#include "features/pixel.h"
 #endif
 
+#include <optional>
+#include <string>
+
 class CpusideCache
 {
 public:
@@ -148,7 +151,7 @@ class GpusideCache
 		int n_gabor_filters,
 		int gabor_ker_side);
 
-	bool allocate_gpu_cache(
+	std::optional<std::string> allocate_gpu_cache(
 		// out
 		GpuCache<Pixel2>& clouds,
 		GpuCache<Pixel2>& konturs,

src/nyx/features/contour.cpp

@@ -583,26 +583,29 @@ ContourFeature::ContourFeature() : FeatureMethod("ContourFeature")
 	// U - unordered subset of C subject to order
 	// S - ordered subset of C
 
-	std::list<Pixel2> U (C.begin(), C.end());
-
+
 	// fix X-crossing points
+
+	std::vector<Pixel2> C_fix = C;
 	int n_xxings = 0;
-	for (Pixel2 p : U)
+	for (const Pixel2& p : C)
 	{
-		Pixel2 pN (p.x, p.y-1, p.inten),
-			pS (p.x, p.y+1, p.inten),
-			pW (p.x-1, p.y, p.inten),
-			pE (p.x+1, p.y, p.inten);
-		auto itN = std::find(U.begin(), U.end(), pN);
-		auto itS = std::find(U.begin(), U.end(), pS);
-		auto itE = std::find(U.begin(), U.end(), pE);
-		auto itW = std::find(U.begin(), U.end(), pW);
-		if (itN != U.end() && itS != U.end() && itW != U.end() && itE != U.end())
+		Pixel2 pN (p.x, p.y - 1, p.inten),
+			pS (p.x, p.y + 1, p.inten),
+			pW (p.x - 1, p.y, p.inten),
+			pE (p.x + 1, p.y, p.inten);
+		auto itN = std::find (C_fix.begin(), C_fix.end(), pN);
+		auto itS = std::find (C_fix.begin(), C_fix.end(), pS);
+		auto itE = std::find (C_fix.begin(), C_fix.end(), pE);
+		auto itW = std::find (C_fix.begin(), C_fix.end(), pW);
+		if (itN != C_fix.end() && itS != C_fix.end() && itW != C_fix.end() && itE != C_fix.end())
 		{
-			U.remove(p);
+			auto itP = std::find (C_fix.begin(), C_fix.end(), p);
+			C_fix.erase (itP);
 			n_xxings++;
 		}
 	}
+	std::list<Pixel2> U (C_fix.begin(), C_fix.end());
 
 	// find contour by contour
 	while (!U.empty())

src/nyx/globals.h

@@ -53,7 +53,6 @@ namespace Nyxus
 	int processDataset_2D_wholeslide(
 		Environment & env,
 		const std::vector<std::string>& intensFiles,
-		const std::vector<std::string>& labelFiles,
 		int n_threads,
 		const SaveOption saveOption,
 		const std::string& outputPath);

src/nyx/gpu/erosion.cu

@@ -6,7 +6,7 @@
 #include <iostream>
 
 #include "gpu.h"
-#include "../cache.h"   //xxxxxxxxxx    #include "../gpucache.h"
+#include "../cache.h"
 
 #include "../helpers/timing.h"
 

src/nyx/gpu/gabor.cu

@@ -1,6 +1,6 @@
 #include <vector>
 #include "gabor.cuh"
-#include "../cache.h"     //xxxxxxxxxxxx      ../gpucache.h"
+#include "../cache.h"
 
 using namespace std;
 

src/nyx/gpu/geomoments_central.cu

@@ -4,7 +4,7 @@
 #include <cuda_runtime_api.h>
 #include <builtin_types.h>
 #include "gpu.h"
-#include "../cache.h"   //xxxxxxx       #include "../gpucache.h"
+#include "../cache.h"
 #include "../features/pixel.h"
 #include "geomoments.cuh"
 

src/nyx/gpu/geomoments_hu.cu

@@ -7,7 +7,7 @@
 
 #include "../features/image_matrix.h"
 #include "gpu.h"
-#include "../cache.h"   //xxxxxxxxx     #include "../gpucache.h"
+#include "../cache.h"
 #include "geomoments.cuh"
 
 namespace NyxusGpu

src/nyx/gpu/geomoments_main.cu

@@ -7,7 +7,7 @@
 
 #include "gpu.h"
 #include "geomoments.cuh"
-#include "../cache.h"   //xxxxxxxxxxxxx     #include "../gpucache.h"
+#include "../cache.h"
 
 #include "../helpers/timing.h"
 

src/nyx/gpu/geomoments_norm.cu

@@ -7,7 +7,7 @@
 
 #include "../features/image_matrix.h"
 #include "gpu.h"
-#include "../cache.h"   //xxxxxxxxxx    #include "../gpucache.h"
+#include "../cache.h"
 #include "geomoments.cuh"
 
 namespace NyxusGpu

src/nyx/gpu/geomoments_origin.cu

@@ -4,7 +4,7 @@
 #include <cuda_runtime_api.h>
 #include <builtin_types.h>
 #include "gpu.h"
-#include "../cache.h"   //xxxxx     #include "../gpucache.h"
+#include "../cache.h"
 #include "../features/pixel.h"
 #include "geomoments.cuh"
 

src/nyx/gpu/geomoments_raw.cu

@@ -4,7 +4,7 @@
 #include <cuda_runtime_api.h>
 #include <builtin_types.h>
 #include "gpu.h"
-#include "../cache.h"   //xxxxxxxxx     #include "../gpucache.h"
+#include "../cache.h"
 #include "geomoments.cuh"
 #include "../features/pixel.h"
 

src/nyx/gpu/geomoments_weighting.cu

@@ -113,7 +113,7 @@ namespace NyxusGpu
         if (pxIdx >= cloud_len)
             return;
 
-        StatsInt x = d_roicloud[pxIdx].x,
+        [[maybe_unused]] StatsInt x = d_roicloud[pxIdx].x,
             y = d_roicloud[pxIdx].y;
 
         const Pixel2 & p = d_roicloud[pxIdx], 
@@ -171,4 +171,4 @@ namespace NyxusGpu
         return true;
     }
 
-}
+}

src/nyx/gpu/helper_cuda.h

@@ -140,23 +140,23 @@ static const char* _cudaGetErrorEnum(cufftResult error) {
     case CUFFT_UNALIGNED_DATA:
         return "CUFFT_UNALIGNED_DATA";
 
-    case CUFFT_INCOMPLETE_PARAMETER_LIST:
-        return "CUFFT_INCOMPLETE_PARAMETER_LIST";
+    //case CUFFT_INCOMPLETE_PARAMETER_LIST:
+    //    return "CUFFT_INCOMPLETE_PARAMETER_LIST";
 
     case CUFFT_INVALID_DEVICE:
         return "CUFFT_INVALID_DEVICE";
 
-    case CUFFT_PARSE_ERROR:
-        return "CUFFT_PARSE_ERROR";
+    //case CUFFT_PARSE_ERROR:
+    //    return "CUFFT_PARSE_ERROR";
 
     case CUFFT_NO_WORKSPACE:
         return "CUFFT_NO_WORKSPACE";
 
     case CUFFT_NOT_IMPLEMENTED:
         return "CUFFT_NOT_IMPLEMENTED";
 
-    case CUFFT_LICENSE_ERROR:
-        return "CUFFT_LICENSE_ERROR";
+    //case CUFFT_LICENSE_ERROR:
+    //    return "CUFFT_LICENSE_ERROR";
 
     case CUFFT_NOT_SUPPORTED:
         return "CUFFT_NOT_SUPPORTED";
@@ -973,4 +973,4 @@ inline bool checkCudaCapabilities(int major_version, int minor_version) {
 
 // end of CUDA Helper Functions
 
-#endif  // COMMON_HELPER_CUDA_H_
+#endif  // COMMON_HELPER_CUDA_H_

src/nyx/gpu/reducers.cu

@@ -4,7 +4,7 @@
 #include <cuda_runtime_api.h>
 #include <builtin_types.h>
 #include "gpu.h"
-#include "../cache.h"   //xxxxxxxxxx    #include "../gpucache.h"
+#include "../cache.h"
 #include "../features/pixel.h"
 #include "geomoments.cuh"