diff --git a/Makefile b/Makefile
index f7796604..3e90be86 100644
--- a/Makefile
+++ b/Makefile
@@ -51,6 +51,8 @@ src/scs_version.o: src/scs_version.c $(INC_FILES)
 $(DIRSRC)/private.o: $(DIRSRC)/private.c  $(DIRSRC)/private.h
 $(INDIRSRC)/indirect/private.o: $(INDIRSRC)/private.c $(INDIRSRC)/private.h
 $(MKLSRC)/private.o: $(MKLSRC)/private.c  $(MKLSRC)/private.h
+$(HIPSRC)/private.o: $(HIPSRC)/private.c  $(HIPSRC)/private.h
+	$(HIPCC) $(CFLAGS) $(HIPCFLAGS) -I$(HIPSRC) -c $(HIPSRC)/private.c -o $@
 $(LINSYS)/scs_matrix.o: $(LINSYS)/scs_matrix.c $(LINSYS)/scs_matrix.h
 $(LINSYS)/csparse.o: $(LINSYS)/csparse.c $(LINSYS)/csparse.h
 
@@ -69,6 +71,11 @@ $(OUT)/libscsmkl.a: $(SCS_O) $(SCS_OBJECTS) $(MKLSRC)/private.o $(LINSYS)/scs_ma
 	$(ARCHIVE) $@ $^
 	- $(RANLIB) $@
 
+$(OUT)/libscship.a: $(SCS_O) $(SCS_OBJECTS) $(HIPSRC)/private.o $(LINSYS)/scs_matrix.o $(LINSYS)/csparse.o
+	mkdir -p $(OUT)
+	$(ARCHIVE) $@ $^
+	- $(RANLIB) $@
+
 $(OUT)/libscsdir.$(SHARED): $(SCS_O) $(SCS_OBJECTS) $(DIRSRC)/private.o $(AMD_OBJS) $(LDL_OBJS) $(LINSYS)/scs_matrix.o $(LINSYS)/csparse.o
 	mkdir -p $(OUT)
 	$(CC) $(CFLAGS) -shared -Wl,$(SONAME),$(@:$(OUT)/%=%) -o $@ $^ $(LDFLAGS) $(BLASLDFLAGS)
@@ -81,6 +88,10 @@ $(OUT)/libscsmkl.$(SHARED): $(SCS_O) $(SCS_OBJECTS) $(MKLSRC)/private.o $(LINSYS
 	mkdir -p $(OUT)
 	$(CC) $(CFLAGS) -shared -Wl,$(SONAME),$(@:$(OUT)/%=%) -o $@ $^ $(LDFLAGS) $(MKLFLAGS)
 
+$(OUT)/libscship.$(SHARED): $(SCS_O) $(SCS_OBJECTS) $(HIPSRC)/private.o $(LINSYS)/scs_matrix.o $(LINSYS)/csparse.o
+	mkdir -p $(OUT)
+	$(CC) $(CFLAGS) -shared -Wl,$(SONAME),$(@:$(OUT)/%=%) -o $@ $^ $(LDFLAGS) $(HIPLDFLAGS)
+
 $(OUT)/demo_socp_direct: test/random_socp_prob.c $(OUT)/libscsdir.a
 	$(CC) $(CFLAGS) -o $@ $^ $(LDFLAGS) $(BLASLDFLAGS)
 
@@ -90,6 +101,9 @@ $(OUT)/demo_socp_indirect: test/random_socp_prob.c $(OUT)/libscsindir.a
 $(OUT)/demo_socp_mkl: test/random_socp_prob.c $(OUT)/libscsmkl.a
 	$(CC) $(CFLAGS) -o $@ $^ $(LDFLAGS) $(MKLFLAGS)
 
+$(OUT)/demo_socp_hip: test/random_socp_prob.c $(OUT)/libscship.a
+	$(CC) $(CFLAGS) -o $@ $^ $(LDFLAGS) $(BLASLDFLAGS) $(HIPLDFLAGS)
+
 $(OUT)/run_from_file_direct: test/run_from_file.c $(OUT)/libscsdir.a
 	$(CC) $(CFLAGS) -o $@ $^ $(LDFLAGS) $(BLASLDFLAGS)
 
@@ -108,7 +122,8 @@ $(OUT)/run_tests_direct: test/run_tests.c $(OUT)/libscsdir.a
 	$(CC) $(CFLAGS) -o $@ $^ $(LDFLAGS) $(BLASLDFLAGS) -Itest
 $(OUT)/run_tests_mkl: test/run_tests.c $(OUT)/libscsmkl.a
 	$(CC) $(CFLAGS) -o $@ $^ $(LDFLAGS) $(MKLFLAGS) -Itest
-
+$(OUT)/run_tests_hip: test/run_tests.c $(OUT)/libscship.a
+	$(CC) $(CFLAGS) -o $@ $^ $(LDFLAGS) $(BLASLDFLAGS) $(HIPLDFLAGS) -Itest
 
 .PHONY: test_gpu
 test_gpu: $(OUT)/run_tests_gpu_indirect # $(OUT)/run_tests_gpu_direct
@@ -120,6 +135,8 @@ ifndef MKLROOT
 	$(error MKLROOT is undefined, set MKLROOT to the MKL install location)
 endif
 
+.PHONY:
+hip: $(OUT)/libscship.a $(OUT)/libscship.$(SHARED) $(OUT)/run_tests_hip $(OUT)/demo_socp_hip
 
 $(OUT)/run_tests_gpu_indirect: test/run_tests.c $(OUT)/libscsgpuindir.a
 	$(CC) $(CFLAGS) -o $@ $^ $(LDFLAGS) $(BLASLDFLAGS) $(CULDFLAGS) -Itest
diff --git a/scs.mk b/scs.mk
index 11ebd599..3c6885e2 100644
--- a/scs.mk
+++ b/scs.mk
@@ -9,6 +9,7 @@ endif
 #CC = i686-w64-mingw32-gcc -m32
 #CC = x86_64-w64-mingw32-gcc-4.8
 CUCC = $(CC) #Don't need to use nvcc, since using cuda blas APIs
+HIPCC = $(HIP_PATH)/bin/hipcc
 
 # For GPU must add cuda libs to path, e.g.
 # export DYLD_LIBRARY_PATH=/usr/local/cuda/lib:$DYLD_LIBRARY_PATH
@@ -70,6 +71,7 @@ INDIRSRC = $(LINSYS)/cpu/indirect
 GPUDIR = $(LINSYS)/gpu/direct
 GPUINDIR = $(LINSYS)/gpu/indirect
 MKLSRC = $(LINSYS)/mkl/direct
+HIPSRC = $(LINSYS)/hip/direct
 
 EXTSRC = $(LINSYS)/external
 
@@ -135,6 +137,11 @@ endif
 # to work for all combinations of platform / compiler / threading options.
 MKLFLAGS = -L$(MKLROOT) -L$(MKLROOT)/lib -Wl,--no-as-needed -lmkl_rt -lmkl_gnu_thread -lmkl_core -lgomp -lpthread -ldl
 
+HIP_PLATFORM =HIP_PLATFORM_AMD
+HIP_PATH = /opt/rocm
+HIPLDFLAGS = -L$(HIP_PATH)/lib -lamdhip64 -lhipsparse
+HIPCFLAGS = -D__$(HIP_PLATFORM)__ -I$(HIP_PATH)/include -Wno-extra-semi -Wno-strict-prototypes
+
 ############ OPENMP: ############
 # set USE_OPENMP = 1 to allow openmp (multi-threaded matrix multiplies):
 # set the number of threads to, for example, 4 by entering the command: