diff --git a/Makefile b/Makefile
index 10fc9f3e..a46c691a 100644
--- a/Makefile
+++ b/Makefile
@@ -133,21 +133,24 @@ TESTSSRCS := $(addprefix $(TESTSDIR)/,bootstrap_test.cc allgather_test.cu)
 TESTSOBJS := $(patsubst %.cc,%.o,$(TESTSSRCS)) $(patsubst %.cu,%.o,$(TESTSSRCS))
 TESTSOBJTARGETS := $(TESTSOBJS:%=$(BUILDDIR)/$(OBJDIR)/%)
 TESTSBINS       := $(patsubst %.o,$(BUILDDIR)/$(BINDIR)/%,$(TESTSOBJS))
-TESTSPERFBIN := $(BUILDDIR)/$(BINDIR)/allgather_test3
+
+MSCLLPPTESTSOBJSDIR:= $(BUILDDIR)/$(OBJDIR)/$(TESTSDIR)
+MSCLLPPTESTBINFILESLIST := allgather_test3
+MSCLLPPTESTBINS         := $(MSCLLPPTESTBINFILESLIST:%=$(BUILDDIR)/$(BINDIR)/$(TESTSDIR)/%_perf)
 
 INCLUDE := -Isrc -Isrc/include
 
-.PHONY: all build lib tests clean
+.PHONY: all build lib tests mscclpp_test clean
 
 all: build
 
-build: lib tests
+build: lib tests mscclpp_test
 
 lib: $(LIBOBJTARGETS) $(INCTARGETS) $(LIBTARGET)
 
 tests: $(TESTSBINS)
 
-allgather_perf: $(TESTSPERFBIN)
+mscclpp_test: $(LIBTARGET) $(MSCLLPPTESTBINS)
 
 cpplint:
 	clang-format-12 -style=file --verbose --Werror --dry-run $(CPPSOURCES)
@@ -175,9 +178,6 @@ $(LIBTARGET): $(LIBOBJTARGETS)
 	ln -sf $(LIBTARGET) $(BUILDDIR)/$(LIBDIR)/$(LIBSONAME)
 
 # Compile .cc tests
-$(TESTSPERFBIN): $(BUILDDIR)/$(OBJDIR)/$(TESTSDIR)/allgather_test3.o $(BUILDDIR)/$(OBJDIR)/$(TESTSDIR)/common.o $(BUILDDIR)/$(OBJDIR)/$(TESTSDIR)/timer.o
-	$(NVCC) -o $@ $^ $(MPI_LDFLAGS) -L$(BUILDDIR)/$(LIBDIR) -lmscclpp
-
 $(BUILDDIR)/$(OBJDIR)/$(TESTSDIR)/%.o: $(TESTSDIR)/%.cc $(INCTARGETS)
 	@mkdir -p $(@D)
 	$(CXX) -o $@ -I$(BUILDDIR)/$(INCDIR) $(MPI_INC) $(CXXFLAGS) -c $< $(MPI_MACRO)
@@ -192,5 +192,10 @@ $(BUILDDIR)/$(BINDIR)/%: $(BUILDDIR)/$(OBJDIR)/%.o $(LIBTARGET)
 	@mkdir -p $(@D)
 	$(NVCC) -o $@ $< $(MPI_LDFLAGS) -L$(BUILDDIR)/$(LIBDIR) -lmscclpp
 
+# Comile mscclpp_test
+$(BUILDDIR)/$(BINDIR)/$(TESTSDIR)/%_perf: $(MSCLLPPTESTSOBJSDIR)/%.o $(MSCLLPPTESTSOBJSDIR)/common.o $(MSCLLPPTESTSOBJSDIR)/timer.o
+	@mkdir -p $(@D)
+	$(NVCC) -o $@ $^ $(MPI_LDFLAGS) -L$(BUILDDIR)/$(LIBDIR) -lmscclpp
+
 clean:
 	rm -rf $(BUILDDIR)
diff --git a/tests/common.cu b/tests/common.cu
index 4343c181..4c819c98 100644
--- a/tests/common.cu
+++ b/tests/common.cu
@@ -30,13 +30,13 @@ static size_t stepBytes = 1*1024*1024;
 static size_t stepFactor = 1;
 static int datacheck = 1;
 static int warmup_iters = 10;
-static int iters = 20;
+static int iters = 100;
 static int timeout = 0;
 static int report_cputime = 0;
 // Report average iteration time: (0=RANK0,1=AVG,2=MIN,3=MAX)
 static int average = 1;
 static std::string ip_port;
-static int cudaGraphLaunches = 10;
+static int cudaGraphLaunches = 15;
 
 #define NUM_BLOCKS 32