Merge branch 'issue3_missing-phi-diffusion-code'

README.rst

@@ -84,6 +84,9 @@ performance and optimization of a specific application. CUDA hardware can be
 emulated on the CPU using the MCUDA_ framework. Proof-of-concept
 trials on GPU and KNL hardware can be run on Amazon's EC2_,
 Rescale's ScaleX_, and equivalent HPC cloud computing platforms.
+Most of the current generation of research supercomputers contain GPU or KNL
+accelerator hardware, including Argonne National Labs'
+Bebop_, NERSC Cori_, TACC Stampede2_, and XSEDE_.
 
 ===============
 Basic Algorithm
@@ -311,7 +314,7 @@ Work in Progress
         - [ ] ripening
 - [ ] KNL
     - [ ] OpenMP
-        - [ ] diffusion
+        - [x] diffusion
         - [ ] spinodal
         - [ ] dendrite
         - [ ] ripening
@@ -337,43 +340,47 @@ intended to imply that the entities, materials, or equipment are necessarily
 the best available for the purpose.
 
 .. _`Trevor Keller`: mailto::[email protected]
-.. _`National Institute of Standards and Technology`: http://www.nist.gov
-.. _`Source on GitHub`: https://github.com/usnistgov/hiperc
-.. _`hiperc_guide.pdf`: doc/hiperc_guide.pdf
-.. |`initial conc`| image:: common-diffusion/diffusion.00000.png
-.. |`final conc`| image:: common-diffusion/diffusion.10000.png
-.. |Source on GitHub| image:: https://assets-cdn.github.com/favicon.ico
-.. _`Source on GitHub`: https://github.com/usnistgov/hiperc
-.. |Documentation on readthedocs| image:: http://readthedocs.org/projects/hiperc/badge/?version=latest
-.. _`Documentation on readthedocs`: http://hiperc.readthedocs.io/en/latest/?badge=latest
-.. |Conversation on Gitter| image:: https://badges.gitter.im/usnistgov/hiperc.svg
+.. _Bebop: http://www.lcrc.anl.gov/systems/resources/bebop/
+.. _ccNUMA: https://en.wikipedia.org/wiki/Non-uniform_memory_access#Cache_coherent_NUMA
 .. _`Conversation on Gitter`: https://gitter.im/usnistgov/hiperc?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=body_badge
+.. |Conversation on Gitter| image:: https://badges.gitter.im/usnistgov/hiperc.svg
 .. _`Contributing Guidelines`: CONTRIBUTING.md
-.. _`file an issue`: https://github.com/usnistgov/hiperc/issues
-.. _OpenACC: https://www.openacc.org/
-.. _radeontop: https://github.com/clbr/radeontop
-.. _ccNUMA: https://en.wikipedia.org/wiki/Non-uniform_memory_access#Cache_coherent_NUMA
 .. _convolution: https://en.wikipedia.org/wiki/Discrete_Laplace_operator#Image_Processing
+.. _Cori: http://www.nersc.gov/users/computational-systems/cori/
 .. _Cray: http://www.cray.com/
 .. _CUDA: https://developer.nvidia.com/cuda-zone
+.. |Documentation on readthedocs| image:: http://readthedocs.org/projects/hiperc/badge/?version=latest
+.. _`Documentation on readthedocs`: http://hiperc.readthedocs.io/en/latest/?badge=latest
 .. _Doxygen: http://www.stack.nl/~dimitri/doxygen/
 .. _EC2: https://aws.amazon.com/ec2/Elastic-GPUs/
+.. _`file an issue`: https://github.com/usnistgov/hiperc/issues
+.. |`final conc`| image:: common-diffusion/diffusion.10000.png
 .. _Gitter: https://gitter.im/usnistgov/hiperc
 .. _GNU: https://gcc.gnu.org/
+.. _`hiperc_guide.pdf`: doc/hiperc_guide.pdf
 .. _htop: http://hisham.hm/htop/
+.. |`initial conc`| image:: common-diffusion/diffusion.00000.png
 .. _`intel_gpu_top`: https://github.com/ChrisCummins/intel-gpu-tools
 .. _KNC: https://ark.intel.com/products/codename/57721/Knights-Corner
 .. _KNL: https://ark.intel.com/products/codename/48999/Knights-Landing
 .. _LaTeX: https://www.latex-project.org/
 .. _Make: https://www.gnu.org/software/make/
 .. _MCUDA: http://impact.crhc.illinois.edu/mcuda.aspx
 .. _MIMD: https://en.wikipedia.org/wiki/MIMD
+.. _`National Institute of Standards and Technology`: http://www.nist.gov
 .. _nvidia-smi: https://developer.nvidia.com/nvidia-system-management-interface
+.. _OpenACC: https://www.openacc.org/
 .. _OpenCL: https://www.khronos.org/opencl/
 .. _OpenMP: http://www.openmp.org/
 .. _PGI: http://www.pgroup.com/
 .. _POSIX: http://www.opengroup.org/austin/papers/posix_faq.html
+.. _radeontop: https://github.com/clbr/radeontop
 .. _ScaleX: http://www.rescale.com/products/
 .. _SIMD: https://en.wikipedia.org/wiki/SIMD
+.. |Source on GitHub| image:: https://assets-cdn.github.com/favicon.ico
+.. _`Source on GitHub`: https://github.com/usnistgov/hiperc
+.. _Stampede2: https://www.tacc.utexas.edu/systems/stampede2
 .. _`Threading Building Blocks`: https://www.threadingbuildingblocks.org/
 .. _`Xeon Phi`: https://www.intel.com/content/www/us/en/products/processors/xeon-phi/xeon-phi-processors.html
+.. _XSEDE: https://www.xsede.org/ecosystem/resources
+

common-diffusion/timer.c

@@ -29,10 +29,13 @@
 #include "timer.h"
 
 #ifdef WIN32
-#define WIN32_LEAN_AND_MEAN
-#include <windows.h>
+	#define WIN32_LEAN_AND_MEAN
+	#include <windows.h>
 #else
-#include <sys/time.h>
+	#ifndef __USE_BSD
+		#define __USE_BSD
+	#endif
+	#include <sys/time.h>
 #endif
 
 /**

phi-openmp-diffusion/Makefile

@@ -0,0 +1,53 @@
+# Makefile for HiPerC diffusion code
+# Xeon Phi implementation
+
+CC = icc
+CFLAGS = -O3 -openmp -xmic-avx512 -I../common-diffusion
+QOPT = -qopt-prefetch=3 -qopt-report=4 -qopt-report-file="qopt.log"
+LINKS = -lpng
+# QOFF = -qoffload-arch=mic-avx512 # Only activate if offloading to KNL coprocessor
+
+OBJS = boundaries.o discretization.o mesh.o numerics.o output.o timer.o
+
+# Executable
+diffusion: phi_main.c $(OBJS)
+	$(CC) $(CFLAGS) $(OBJS) -include omp.h $< -o $@ $(QOPT) $(LINKS)
+
+# Xeon Phi objects
+boundaries.o: phi_boundaries.c
+	$(CC) $(CFLAGS) -c $< -o $@
+
+discretization.o: phi_discretization.c
+	$(CC) $(CFLAGS) -c $< -o $@
+
+# Common objects
+mesh.o: ../common-diffusion/mesh.c
+	$(CC) $(CFLAGS) -c $< -o $@
+
+numerics.o: ../common-diffusion/numerics.c
+	$(CC) $(CFLAGS) -c $< -o $@
+
+output.o: ../common-diffusion/output.c
+	$(CC) $(CFLAGS) -c $< -o $@
+
+timer.o: ../common-diffusion/timer.c
+	$(CC) $(CFLAGS) -c $< -o $@
+
+# Helper scripts
+.PHONY: run
+run: diffusion
+	/usr/bin/time -f' Time (%E wall, %U user, %S sys)' ./diffusion ../common-diffusion/params.txt
+
+.PHONY: cleanobjects
+cleanobjects:
+	rm -f diffusion *.o
+
+.PHONY: cleanoutputs
+cleanoutputs:
+	rm -f diffusion.*.csv diffusion.*.png runlog.csv
+
+.PHONY: clean
+clean: cleanobjects
+
+.PHONY: cleanall
+cleanall: cleanobjects cleanoutputs

phi-openmp-diffusion/README.md

@@ -26,9 +26,12 @@ To build this code, you must have installed
 These are usually available through the package manager. For example,
 ```apt-get install make libpng12-dev```. The Intel compiler is proprietary,
 but available to open-source developers at no cost. Hardware capable of AVX-512
-instructions, whether a primary Xeon CPU or a Xeon Phi, is also required.
-Proof-of-concept KNL code can be run on [Rescale's ScaleX][_scalex]
-and equivalent HPC cloud computing platforms.
+instructions, whether a primary Xeon CPU or a [Xeon Phi][_knl], is also required.
+Proof-of-concept KNL code can be run on [Rescale's ScaleX][_scalex] and
+equivalent HPC cloud computing platforms. [KNL][_knl] nodes are available on
+advanced research computing platforms, including Argonne National Labs'
+[Bebop][_bebop], NERSC [Cori][_cori], TACC [Stampede2][_tacc], and
+[XSEDE][_xsede].
 
 ## Customization
 
@@ -49,8 +52,13 @@ the [National Institute of Standards and Technology (NIST)](http://www.nist.gov)
 nor is it intended to imply that the entities, materials, or equipment are
 necessarily the best available for the purpose.
 
-[_make]:   https://www.gnu.org/software/make/
+[_bebop]:  http://www.lcrc.anl.gov/systems/resources/bebop/
+[_cori]:   http://www.nersc.gov/users/computational-systems/cori/
 [_gcc]:    https://gcc.gnu.org
 [_intel]:  https://software.intel.com/en-us/intel-compilers
+[_knl]:    https://www.intel.com/content/www/us/en/products/processors/xeon-phi/xeon-phi-processors.html
+[_make]:   https://www.gnu.org/software/make/
 [_png]:    http://www.libpng.org/pub/png/libpng.html
 [_scalex]: http://www.rescale.com/products/
+[_tacc]:   https://www.tacc.utexas.edu/systems/stampede2
+[_xsede]:  https://www.xsede.org/ecosystem/resources

phi-openmp-diffusion/phi_boundaries.c

@@ -0,0 +1,106 @@
+/**********************************************************************************
+ HiPerC: High Performance Computing Strategies for Boundary Value Problems
+ written by Trevor Keller and available from https://github.com/usnistgov/hiperc
+
+ This software was developed at the National Institute of Standards and Technology
+ by employees of the Federal Government in the course of their official duties.
+ Pursuant to title 17 section 105 of the United States Code this software is not
+ subject to copyright protection and is in the public domain. NIST assumes no
+ responsibility whatsoever for the use of this software by other parties, and makes
+ no guarantees, expressed or implied, about its quality, reliability, or any other
+ characteristic. We would appreciate acknowledgement if the software is used.
+
+ This software can be redistributed and/or modified freely provided that any
+ derivative works bear some notice that they are derived from it, and any modified
+ versions bear some notice that they have been modified.
+
+ Questions/comments to Trevor Keller ([email protected])
+ **********************************************************************************/
+
+
+/**
+ \file  phi_boundaries.c
+ \brief Implementation of boundary condition functions with OpenMP threading
+*/
+
+#include <math.h>
+#include <omp.h>
+#include "boundaries.h"
+
+void set_boundaries(fp_t bc[2][2])
+{
+	fp_t clo = 0.0, chi = 1.0;
+	bc[0][0] = clo; /* bottom boundary */
+	bc[0][1] = clo; /* top boundary */
+	bc[1][0] = chi; /* left boundary */
+	bc[1][1] = chi; /* right boundary */
+}
+
+void apply_initial_conditions(fp_t** conc, int nx, int ny, int nm, fp_t bc[2][2])
+{
+	#pragma omp parallel
+	{
+		int i, j;
+
+		#pragma omp for collapse(2)
+		for (j = 0; j < ny; j++)
+			for (i = 0; i < nx; i++)
+				conc[j][i] = bc[0][0];
+
+		#pragma omp for collapse(2)
+		for (j = 0; j < ny/2; j++)
+			for (i = 0; i < 1+nm/2; i++)
+				conc[j][i] = bc[1][0]; /* left half-wall */
+
+		#pragma omp for collapse(2)
+		for (j = ny/2; j < ny; j++)
+			for (i = nx-1-nm/2; i < nx; i++)
+				conc[j][i] = bc[1][1]; /* right half-wall */
+	}
+}
+
+void apply_boundary_conditions(fp_t** conc, int nx, int ny, int nm, fp_t bc[2][2])
+{
+	#pragma omp parallel
+	{
+		int i, ihi, ilo, j, jhi, jlo, offset;
+
+		/* apply fixed boundary values: sequence does not matter */
+
+		#pragma omp for collapse(2) private(i,j)
+		for (j = 0; j < ny/2; j++) {
+			for (i = 0; i < 1+nm/2; i++) {
+				conc[j][i] = bc[1][0]; /* left value */
+			}
+		}
+
+		#pragma omp for collapse(2) private(i,j)
+		for (j = ny/2; j < ny; j++) {
+			for (i = nx-1-nm/2; i < nx; i++) {
+				conc[j][i] = bc[1][1]; /* right value */
+			}
+		}
+
+		/* apply no-flux boundary conditions: inside to out, sequence matters */
+
+		for (offset = 0; offset < nm/2; offset++) {
+			ilo = nm/2 - offset;
+			ihi = nx - 1 - nm/2 + offset;
+			#pragma omp for private(j)
+			for (j = 0; j < ny; j++) {
+				conc[j][ilo-1] = conc[j][ilo]; /* left condition */
+				conc[j][ihi+1] = conc[j][ihi]; /* right condition */
+			}
+		}
+
+		for (offset = 0; offset < nm/2; offset++) {
+			jlo = nm/2 - offset;
+			jhi = ny - 1 - nm/2 + offset;
+			#pragma omp for private(i)
+			for (i = 0; i < nx; i++) {
+				conc[jlo-1][i] = conc[jlo][i]; /* bottom condition */
+				conc[jhi+1][i] = conc[jhi][i]; /* top condition */
+			}
+		}
+	}
+}