add support for am processor, and option to use arm performance library

HPC_Sparse_Matrix.hpp

@@ -43,6 +43,11 @@
 #ifndef HPC_SPARSE_MATRIX_H
 #define HPC_SPARSE_MATRIX_H
 
+#if defined(USE_ARMPL)
+#include <armpl.h>
+#define LIBRARY_NAME "ArmPL"
+#endif
+
 // These constants are upper bounds that might need to be changes for 
 // pathological matrices, e.g., those with nearly dense rows/columns.
 
@@ -65,6 +70,10 @@ struct HPC_Sparse_Matrix_STRUCT {
   int ** ptr_to_inds_in_row;
   double ** ptr_to_diags;
 
+#if defined(USE_ARMPL) 
+  armpl_spmat_t mat_armpl;
+#endif
+
 #ifdef USING_MPI
   int num_external;
   int num_send_neighbors;

HPC_sparsemv.cpp

@@ -63,12 +63,27 @@ using std::endl;
 #include <cmath>
 #include "HPC_sparsemv.hpp"
 
+#if defined(USE_ARMPL)
+#include <armpl.h>
+#define LIBRARY_NAME "ArmPL"
+#endif
+
 int HPC_sparsemv( HPC_Sparse_Matrix *A, 
 		 const double * const x, double * const y)
 {
 
   const int nrow = (const int) A->local_nrow;
 
+#if defined(USE_ARMPL)
+  double alpha = 1.0, beta = 0.0;
+  armpl_status_t info = armpl_spmv_exec_d(ARMPL_SPARSE_OPERATION_NOTRANS, alpha, A->mat_armpl, x, beta, y);
+	if (info!=ARMPL_STATUS_SUCCESS) 
+  {
+    cerr << "ERROR: armpl_spmv_exec_d returned %d\n";
+    exit(1);
+  }
+
+#else
 #ifdef USING_OMP
 #pragma omp parallel for
 #endif
@@ -87,5 +102,7 @@ int HPC_sparsemv( HPC_Sparse_Matrix *A,
           sum += cur_vals[j]*x[cur_inds[j]];
       y[i] = sum;
     }
+#endif
   return(0);
 }
+

Makefile

@@ -47,8 +47,10 @@
 #
 # 0) Specify compiler and linker:
 
-CXX=/usr/local/bin/g++
-LINKER=/usr/local/bin/g++
+include config.mk
+
+#CXX=
+LINKER=mpicxx
 #CXX=mpicxx
 #LINKER=mpicxx
 
@@ -57,7 +59,7 @@ LINKER=/usr/local/bin/g++
 #    If you want to run the program with MPI, make sure USE_MPI is set 
 #    to -DUSING_MPI
 
-USE_MPI =
+USE_MPI = -DUSING_MPI
 #USE_MPI = -DUSING_MPI
 
 
@@ -68,17 +70,19 @@ USE_MPI =
 #    - Are not using MPI compiler wrappers
 #    Then specify the path to your MPI header file (include a -I)
 
-#MPI_INC = -I/usr/MPICH/SDK.gcc/include
+MPI_INC = -I/opt/amazon/openmpi/include
 
 
 # 3) Specify C++ compiler optimization flags (if any)
 #    Typically some reasonably high level of optimization should be used to 
 #    enhance performance.
 
 #IA32 with GCC: 
-#CPP_OPT_FLAGS = -O3 -funroll-all-loops -malign-double
+##CPP_OPT_FLAGS = -O3 -funroll-all-loops -malign-double
 CPP_OPT_FLAGS = -O3 -ftree-vectorize -ftree-vectorizer-verbose=2
-
+ifeq (arm,$(COMPILER))
+CPP_OPT_FLAGS = $(CFLAGS_OPT) -fPIC    # add -g for profiler 
+endif
 #
 # 4) MPI library:
 #    If you:
@@ -87,7 +91,7 @@ CPP_OPT_FLAGS = -O3 -ftree-vectorize -ftree-vectorizer-verbose=2
 #    - Are not using MPI compiler wrappers for linking
 #    Then specify the path to your MPI library (include -L and -l directives)
 
-#MPI_LIB = -L/usr/MPICH/SDK.gcc/lib -lmpich
+MPI_LIB = -L/opt/amazon/openmpi/lib
 
 #
 # 5) Build with OpenMP or not?
@@ -110,13 +114,18 @@ USE_OMP =
 SYS_LIB =-lm
 
 #
-# 6) Specify name if executable (optional):
+# 8) Arm PL library 
+ifeq (arm,$(COMPILER))
+ARMPL_LIB = -larmpl -L/shared/arm/armpl/armpl_22.0.2_gcc-11.2/lib
+endif
+#
+# 9) Specify name if executable (optional):
 
 TARGET = test_HPCCG
 
 ################### Derived Quantities (no modification required) ##############
 
-CXXFLAGS= $(CPP_OPT_FLAGS) $(OMP_FLAGS) $(USE_OMP) $(USE_MPI) $(MPI_INC)
+CXXFLAGS= $(CPP_OPT_FLAGS) $(OMP_FLAGS) $(CXXFLAGS_LIBRARY) $(USE_OMP) $(USE_MPI) $(MPI_INC) 
 
 LIB_PATHS= $(SYS_LIB)
 
@@ -129,7 +138,7 @@ TEST_CPP = main.cpp generate_matrix.cpp read_HPC_row.cpp \
 TEST_OBJ          = $(TEST_CPP:.cpp=.o)
 
 $(TARGET): $(TEST_OBJ)
-	$(LINKER) $(CPP_OPT_FLAGS) $(OMP_FLAGS) $(TEST_OBJ) $(LIB_PATHS) -o $(TARGET)
+	$(LINKER) $(CPP_OPT_FLAGS) $(OMP_FLAGS) $(TEST_OBJ) $(LIB_PATHS) $(ARMPL_LIB) -o $(TARGET)
 
 test:
 	@echo "Not implemented yet..."

config.mk

@@ -0,0 +1,207 @@
+#------------------------------------------------------------------------------------
+# Copyright (C) Arm Limited, 2019-2020 All rights reserved.
+#
+# The example code is provided to you as an aid to learning when working
+# with Arm-based technology, including but not limited to programming tutorials.
+#
+# Arm hereby grants to you, subject to the terms and conditions of this Licence,
+# a non-exclusive, non-transferable, non-sub-licensable, free-of-charge copyright
+# licence, to use, copy and modify the Software solely for the purpose of internal
+# demonstration and evaluation.
+#
+# You accept that the Software has not been tested by Arm therefore the Software
+# is provided "as is", without warranty of any kind, express or implied. In no
+# event shall the authors or copyright holders be liable for any claim, damages
+# or other liability, whether in action or contract, tort or otherwise, arising
+# from, out of or in connection with the Software or the use of Software.
+#------------------------------------------------------------------------------------
+
+COMPILER ?= arm
+
+#############################################################################
+# GNU compiler flags
+#############################################################################
+ifeq (gnu,$(COMPILER))
+
+LIBRARY ?= none
+
+CC = gcc
+CXX = g++
+FC = gfortran
+
+FLAGS_VERSION = --version
+
+CFLAGS_OPT    = -Ofast -mcpu=native
+
+CXXFLAGS_OPT    = $(CFLAGS_OPT)
+
+FFLAGS_OPT    = $(CFLAGS_OPT)
+
+else
+#############################################################################
+# Arm compiler flags
+#############################################################################
+ifeq (arm,$(COMPILER))
+
+LIBRARY ?= armpl
+
+CC = armclang
+CXX = armclang++
+FC = armflang
+
+FLAGS_VERSION = --version
+
+CFLAGS_OPT    = -Ofast -mcpu=native
+
+CXXFLAGS_OPT    = $(CFLAGS_OPT)
+
+FFLAGS_OPT    = $(CFLAGS_OPT)
+
+else
+#############################################################################
+# Fujitsu compiler flags
+#############################################################################
+ifeq (fujitsu,$(COMPILER))
+
+LIBRARY ?= ssl2
+
+CC = fcc
+CXX = FCC
+FC = frt
+
+FLAGS_VERSION = --version
+
+CFLAGS_OPT    = -Kfast
+
+CXXFLAGS_OPT    = $(CFLAGS_OPT)
+
+FFLAGS_OPT    = $(CFLAGS_OPT)
+
+else
+#############################################################################
+# Cray compiler flags
+#############################################################################
+ifeq (cray,$(COMPILER))
+
+LIBRARY ?= cblas
+
+CC = cc
+CXX = CC
+fc = ftn
+
+FLAGS_VERSION = --version
+
+CFLAGS_OPT    = -O3 
+
+CXXFLAGS_OPT    = $(CFLAGS_OPT)
+
+FFLAGS_OPT    = $(CFLAGS_OPT)
+
+else
+#############################################################################
+# Help message
+#############################################################################
+ifeq (help,$(COMPILER))
+
+$(info COMPILER can be one of: gnu, arm, fujitsu, cray)
+$(info  * gnu: GNU Compiler)
+$(info  * arm: Arm Compiler for Linux)
+$(info  * fujitsu: Fujitsu Compiler)
+$(info  * cray: Cray Compiler)
+$(error )
+
+else
+#############################################################################
+# Error handling
+#############################################################################
+
+$(error Invalid parameter: COMPILER=$(COMPILER))
+
+#############################################################################
+endif
+endif
+endif
+endif
+endif
+#############################################################################
+
+
+#############################################################################
+# Generic CBLAS library flags
+#############################################################################
+ifeq (none,$(LIBRARY))
+
+CFLAGS_LIBRARY   = 
+CXXFLAGS_LIBRARY = 
+FFLAGS_LIBRARY   = 
+LDFLAGS_LIBRARY  = -lm
+
+else
+#############################################################################
+# Arm Performance Library (ArmPL) library flags
+#############################################################################
+ifeq (armpl,$(LIBRARY))
+
+ifeq (arm,$(COMPILER))
+# The Arm compiler automatically knows where to find ArmPL and will 
+# automatically enable OpenMP if the -fopenmp flag is present
+CFLAGS_LIBRARY   = -armpl -DUSE_ARMPL
+CXXFLAGS_LIBRARY = -armpl -DUSE_ARMPL
+FFLAGS_LIBRARY   = -armpl -DUSE_ARMPL
+LDFLAGS_LIBRARY  = -armpl
+else
+# If not using the Arm compiler, make sure to load the ArmPL module.
+# ARMPL_INCLUDES and ARMPL_LIBRARIES are set by the ArmPL modulefile
+CFLAGS_LIBRARY   = -I$(ARMPL_INCLUDES) -DUSE_ARMPL
+CXXFLAGS_LIBRARY = -I$(ARMPL_INCLUDES) -DUSE_ARMPL
+FFLAGS_LIBRARY   = -I$(ARMPL_INCLUDES) -DUSE_ARMPL
+LDFLAGS_LIBRARY  = -L$(ARMPL_LIBRARIES) -larmpl -lamath -lastring
+endif
+
+else
+#############################################################################
+# Fujitsu SSL II library flags
+#############################################################################
+ifeq (ssl2,$(LIBRARY))
+
+ifeq (fujitsu,$(COMPILER))
+CFLAGS_LIBRARY   = -SSL2 -DUSE_SSL2
+CXXFLAGS_LIBRARY = -SSL2 -DUSE_SSL2
+FFLAGS_LIBRARY   = -SSL2 -DUSE_SSL2
+LDFLAGS_LIBRARY  = -SSL2
+else
+$(error SSL2 can only be used with the Fujitsu compiler)
+endif
+
+else
+#############################################################################
+# Error handling
+#############################################################################
+
+$(error Invalid parameter: LIBRARY=$(LIBRARY))
+
+#############################################################################
+endif
+endif
+endif
+#############################################################################
+
+#############################################################################
+# ArmIE Configuration
+#############################################################################
+
+ARMIE_EXE = armie
+ARMIE_SVE_VECTOR_BITS ?= 512
+ARMIE = $(ARMIE_EXE) -msve-vector-bits=$(ARMIE_SVE_VECTOR_BITS)
+ARMIE_MEMTRACE = $(ARMIE_EXE) -e libmemtrace_sve_$(ARMIE_SVE_VECTOR_BITS).so
+
+#############################################################################
+# Functions
+#############################################################################
+
+
+print_hline = @echo "------------------------------------------------"
+print_version = $(CC) $(FLAGS_VERSION)
+print_banner = $(call print_hline) @echo $1 $(call print_hline)
+
+