configure.ac

@@ -263,31 +263,6 @@ fi
 # do not warn about deprecated GPU architectures
 NVCCFLAGS+=" -Wno-deprecated-gpu-targets"
 
-# ------------------------------------------------------------------------------
-# Setup nvcc ccbin
-# ------------------------------------------------------------------------------
-
-AC_ARG_WITH([nvcc-ccbin],
-   [AS_HELP_STRING([--with-nvcc-ccbin=PATH],[path or executable name of host compiler for nvcc @<:@default=auto@:>@])],
-   [],
-   [with_nvcc_ccbin=auto])
-
-
-if test "x$with_nvcc_ccbin" != "xno"
-then
-   if test "x$with_nvcc_ccbin" != "xauto"
-   then
-      NVCCFLAGS+=" -ccbin $with_nvcc_ccbin"
-   else
-      case $host_os in
-      *darwin*)
-        NVCCFLAGS+=" -ccbin /usr/bin/clang"
-        ;;
-      esac
-   fi
-fi
-
-
 # ------------------------------------------------------------------------------
 # Setup SSE
 # ------------------------------------------------------------------------------
@@ -358,7 +333,7 @@ case $host_os in
   then
      NVCCFLAGS+=" -m64"
      AM_CXXFLAGS="$AM_CXXFLAGS -m64"
-     NVCCFLAGS+=" -D_POSIX_C_SOURCE"
+     NVCCFLAGS+=" -D_POSIX_C_SOURCE -ccbin /usr/bin/clang"
      CUDA_LIBS+=" -F/Library/Frameworks -framework CUDA"
   fi
 

debian/changelog

-libhmsbeagle (2.1.2+git20180619-1) UNRELEASED; urgency=medium
+libhmsbeagle (3.0.2+dfsg-1) UNRELEASED; urgency=medium
 
   * New upstream version
   * Force LC_ALL=C.UTF-8 encoding

debian/watch

 version=4
 
-opts="uversionmangle=s/.*date=(\d{4})-(\d\d)-(\d\d)T.*/2.1.2+git$1$2$3/, \
-    filenamemangle=s/.*date=(\d{4})-(\d\d)-(\d\d)T.*/2.1.2+git$1$2$3.tar.gz/, \
-    repack,compression=xz" \
-    https://qa.debian.org/cgi-bin/fakeupstream.cgi?upstream=github_commits_package_json/beagle-dev/beagle-lib \
-    .*/archive/(.*\.tar\.gz?.*)
-
-# Issue asking for release tags:
-# https://github.com/beagle-dev/beagle-lib/issues/115
+opts="uversionmangle=s/_/./g,repacksuffix=+dfsg,dversionmangle=s/\+dfsg//g,repack,compression=xz" \
+   https://github.com/beagle-dev/beagle-lib/releases .*/archive/v@ANY_VERSION@@ARCHIVE_EXT@

examples/synthetictest/synthetictest.cpp

@@ -1247,26 +1247,6 @@ void runBeagle(int resource,
             sumLogL += siteLogLs[i] * patternWeights[i];
         }
         fprintf(stdout, "\nsumLogL = %.5f\n", sumLogL);
-        if (calcderivs) {
-            double* siteSecondDerivs = (double*) malloc(sizeof(double) * nsites);
-            beagleGetSiteDerivatives(instance, siteLogLs, siteSecondDerivs);
-            sumLogL = 0.0;
-            fprintf(stdout, "site first derivs = ");
-            for (int i=0; i<nsites; i++) {
-                fprintf(stdout, "%.5f \t", siteLogLs[i]);
-                sumLogL += siteLogLs[i] * patternWeights[i];
-            }
-            fprintf(stdout, "\nsumFirstDerivs = %.5f\n", sumLogL);
-
-            sumLogL = 0.0;
-            fprintf(stdout, "site second derivs = ");
-            for (int i=0; i<nsites; i++) {
-                fprintf(stdout, "%.5f \t", siteSecondDerivs[i]);
-                sumLogL += siteSecondDerivs[i] * patternWeights[i];
-            }
-            fprintf(stdout, "\nsumSecondDerivs = %.5f\n", sumLogL);
-            free(siteSecondDerivs);
-        }
         free(siteLogLs);
     }
 

libhmsbeagle/CPU/BeagleCPUImpl.hpp

@@ -2551,10 +2551,9 @@ void BeagleCPUImpl<BEAGLE_CPU_GENERIC>::calcEdgeLogLikelihoodsSecondDerivByParti
 
                         w += kTransPaddedStateCount;
                     }
-                    v += kPartialsPaddedStateCount;
-                }
                     v += ((kPatternCount - endPattern) + startPattern) * kPartialsPaddedStateCount;
                 }
+            }
 
         } else { // Integrate against a partial at the child
 

libhmsbeagle/GPU/GPUInterfaceCUDA.cpp

@@ -203,7 +203,7 @@ int GPUInterface::Initialize() {
         SAFE_CUDA(cuDeviceGet(&tmpCudaDevice, i));
         SAFE_CUDA(cuDeviceGetAttribute(&capabilityMajor, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR, tmpCudaDevice));
         SAFE_CUDA(cuDeviceGetAttribute(&capabilityMinor, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR, tmpCudaDevice));
-        if (capabilityMajor >= 3 && capabilityMinor != 9999) {
+        if ((capabilityMajor > 1 && capabilityMinor != 9999) || (capabilityMajor == 1 && capabilityMinor > 0)) {
             resourceMap->insert(std::make_pair(currentDevice++, i));
         }
     }

libhmsbeagle/GPU/kernels/Makefile.am

@@ -19,17 +19,17 @@ BeagleCUDA_kernels.h: Makefile kernels4.cu kernelsX.cu kernelsAll.cu ../GPUImplD
 #
 # 	Compile 4-state model
 	$(NVCC) -o BeagleCUDA_kernels.ptx --default-stream per-thread -ptx -DCUDA -DSTATE_COUNT=4 \
-		$(srcdir)/kernels4.cu $(NVCCFLAGS) -DHAVE_CONFIG_H $(INCLUDE_DIRS) || { \rm BeagleCUDA_kernels.h; exit; }; \
-	echo "#define KERNELS_STRING_SP_4 \"" | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h; \
-	cat BeagleCUDA_kernels.ptx | sed 's/\"/\\"/g' | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h; \
+		$(srcdir)/kernels4.cu $(NVCCFLAGS) -DHAVE_CONFIG_H $(INCLUDE_DIRS)
+	echo "#define KERNELS_STRING_SP_4 \"" | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h
+	cat BeagleCUDA_kernels.ptx | sed 's/\"/\\"/g' | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h
 	echo "\"" >> BeagleCUDA_kernels.h
 #
 #	HERE IS THE LOOP FOR GENERIC KERNELS
 #
 	for s in $(STATE_COUNT_LIST); do \
 		echo "Making state count = $$s" ; \
-		$(NVCC) -o BeagleCUDA_kernels.ptx --default-stream per-thread -ptx -DCUDA -DSTATE_COUNT=$$s \
-			$(srcdir)/kernelsX.cu $(NVCCFLAGS) -DHAVE_CONFIG_H $(INCLUDE_DIRS) || { \rm BeagleCUDA_kernels.h; exit; }; \
+		($(NVCC) -o BeagleCUDA_kernels.ptx --default-stream per-thread -ptx -DCUDA -DSTATE_COUNT=$$s \
+			$(srcdir)/kernelsX.cu $(NVCCFLAGS) -DHAVE_CONFIG_H $(INCLUDE_DIRS)) || exit; \
 		echo "#define KERNELS_STRING_SP_$$s \"" | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h; \
 		cat BeagleCUDA_kernels.ptx | sed 's/\"/\\"/g' | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h; \
 		echo "\"" >> BeagleCUDA_kernels.h; \
@@ -40,17 +40,17 @@ BeagleCUDA_kernels.h: Makefile kernels4.cu kernelsX.cu kernelsAll.cu ../GPUImplD
 #
 # 	Compile 4-state model
 	$(NVCC) -o BeagleCUDA_kernels.ptx --default-stream per-thread -ptx -DCUDA -DSTATE_COUNT=4 -DDOUBLE_PRECISION \
-		$(srcdir)/kernels4.cu $(NVCCFLAGS) -DHAVE_CONFIG_H $(INCLUDE_DIRS) || { \rm BeagleCUDA_kernels.h; exit; }; \
-	echo "#define KERNELS_STRING_DP_4 \"" | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h; \
-	cat BeagleCUDA_kernels.ptx | sed 's/\"/\\"/g' | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h; \
+		$(srcdir)/kernels4.cu $(NVCCFLAGS) -DHAVE_CONFIG_H $(INCLUDE_DIRS)
+	echo "#define KERNELS_STRING_DP_4 \"" | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h
+	cat BeagleCUDA_kernels.ptx | sed 's/\"/\\"/g' | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h
 	echo "\"" >> BeagleCUDA_kernels.h
 #
 #	HERE IS THE LOOP FOR GENERIC KERNELS
 #
 	for s in $(STATE_COUNT_LIST); do \
 		echo "Making state count = $$s" ; \
-		$(NVCC) -o BeagleCUDA_kernels.ptx --default-stream per-thread -ptx -DCUDA -DSTATE_COUNT=$$s -DDOUBLE_PRECISION \
-			$(srcdir)/kernelsX.cu $(NVCCFLAGS) -DHAVE_CONFIG_H $(INCLUDE_DIRS) || { \rm BeagleCUDA_kernels.h; exit; }; \
+		($(NVCC) -o BeagleCUDA_kernels.ptx --default-stream per-thread -ptx -DCUDA -DSTATE_COUNT=$$s -DDOUBLE_PRECISION \
+			$(srcdir)/kernelsX.cu $(NVCCFLAGS) -DHAVE_CONFIG_H $(INCLUDE_DIRS)) || exit; \
 		echo "#define KERNELS_STRING_DP_$$s \"" | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h; \
 		cat BeagleCUDA_kernels.ptx | sed 's/\"/\\"/g' | sed 's/$$/\\n\\/' >> BeagleCUDA_kernels.h; \
 		echo "\"" >> BeagleCUDA_kernels.h; \