Add OpenMP-target implementation

This code has been ported from https://github.com/UoB-HPC/bude-benchmark, the gpu branch.
The code works, but performance is low.

Contributes to #3

openmp-target/Makefile

@@ -0,0 +1,27 @@
+SHELL := bash
+.SHELLFLAGS := -eu -o pipefail -c
+.DELETE_ON_ERROR:
+
+MAKEFLAGS += --warn-undefined-variables --no-builtin-rules
+
+# -------
+
+CC = cc
+
+ARCH = sm_70
+
+CFLAGS = -O3 -fopenmp -fopenmp-targets=nvptx64 -Xopenmp-target -march=$(ARCH)
+LDFLAGS = -lm
+
+OBJ = $(patsubst %.c,%.o,$(wildcard *.c))
+
+bude: Makefile $(OBJ)
+	$(CC) $(CFLAGS) $(OBJ) -o bude $(LDFLAGS)
+
+%.o: %.c Makefile make.deps
+	$(CC) $(CFLAGS) -c $<
+
+.PHONY: clean
+
+clean:
+	rm -f bude *.ptx *.cub *.lst *.o *.optrpt

openmp-target/README.md

@@ -0,0 +1,32 @@
+# BUDE OpenMP Target
+
+This is a GPU implementation of BUDE using OpenMP target.
+
+## Building
+
+This implementations supports the following compilers:
+
+* CCE (Cray)
+* LLVM 10+ (built with offloading support)
+
+```
+make
+```
+
+The default compiler and flags can be overridden using `CC` and `CFLAGS`, respectively.
+
+### Target Architecture
+
+By default, the NVIDIA V100 is targetted, but this can be changed by setting the `ARCH` parameter to the name of the target GPU architecture.
+
+For example, to target a P100:
+
+```
+make ARCH=sm_60
+```
+
+## Running
+
+This implementation has no special run-time options.
+The `-n` and `-i` parameters are available.
+Run `bude -h` for a help message.