GPU: Improve debug / info messages

Author: davidrohr

GPU/GPUTracking/Base/cuda/GPUReconstructionCUDAGenRTC.cxx

@@ -85,6 +85,9 @@ int32_t GPUReconstructionCUDA::genRTC(std::string& filename, uint32_t& nCompile)
                                    "#define GPUCA_WARP_SIZE " + std::to_string(mWarpSize) + "\n";
   if (GetProcessingSettings().rtctech.printLaunchBounds || GetProcessingSettings().debugLevel >= 3) {
     GPUInfo("RTC Launch Bounds:\n%s", launchBounds.c_str());
+    if (GetProcessingSettings().rtctech.printLaunchBounds >= 2) {
+      return 1;
+    }
   }
 
   const std::string compilerVersions = getBackendVersions();

GPU/GPUTracking/DataCompression/GPUTPCClusterStatistics.cxx

@@ -228,7 +228,7 @@ void GPUTPCClusterStatistics::Finish()
   GPUInfo("Combined Sigma: %6.4f --> %6.4f (%6.4f%%)", eSigma, eSigmaCombined, eSigma > 1e-3 ? (100. * (eSigma - eSigmaCombined) / eSigma) : 0.f);
   GPUInfo("Combined Q: %6.4f --> %6.4f (%6.4f%%)", eQ, eQCombined, eQ > 1e-3 ? (100. * (eQ - eQCombined) / eQ) : 0.f);
 
-  printf("\nCombined Entropy: %7.4f   (Size %'13.0f, %'zu clusters)\nCombined Huffman: %7.4f   (Size %'13.0f, %f%%)\n\n", mEntropy / mNTotalClusters, mEntropy, mNTotalClusters, mHuffman / mNTotalClusters, mHuffman, 100. * (mHuffman - mEntropy) / mHuffman);
+  printf("\nCombined Entropy: %7.4f   (Size %'13.0f, %'zu clusters)\nCombined Huffman: %7.4f   (Size %'13.0f, %f%%)\n\n", mEntropy / mNTotalClusters, mEntropy / 8., mNTotalClusters, mHuffman / mNTotalClusters, mHuffman / 8., 100. * (mHuffman - mEntropy) / mHuffman);
 }
 
 float GPUTPCClusterStatistics::Analyze(std::vector<int32_t>& p, const char* name, bool count)

GPU/GPUTracking/Definitions/GPUSettingsList.h

@@ -233,7 +233,7 @@ BeginSubConfig(GPUSettingsProcessingRTCtechnical, rtctech, configStandalone.proc
 AddOption(runTest, int32_t, 0, "", 0, "Do not run the actual benchmark, but just test RTC compilation (1 full test, 2 test only compilation)")
 AddOption(cacheMutex, bool, true, "", 0, "Use a file lock to serialize access to the cache folder")
 AddOption(ignoreCacheValid, bool, false, "", 0, "If set, allows to use RTC cached code files even if they are not valid for the current source code / parameters")
-AddOption(printLaunchBounds, bool, false, "", 0, "Print launch bounds used for RTC code as debugging option")
+AddOption(printLaunchBounds, int32_t, false, "", 0, "Print launch bounds used for RTC code as debugging option, 2 for exit after printing", def(1))
 AddOption(allowOptimizedSlaveReconstruction, bool, false, "", 0, "Allow RTC with slave GPUReconstruction instances with optConstexpr and optSpecialcode")
 AddOption(cacheFolder, std::string, "./rtccache/", "", 0, "Folder in which the cache file is stored")
 AddOption(prependCommand, std::string, "", "", 0, "Prepend RTC compilation commands by this string")