Removed OMP in sapporo

Makefile

@@ -2,16 +2,15 @@
 # This is the unified Makefile for the hybrid code                             #
 ################################################################################
 
-RUNDIR       ?= phigrape
+RUNDIR       ?= run
 ETICS_LMAX   ?= 2
 ETICS_NMAX   ?= 10
 GPUARCH      ?= sm_75
 CUDAHOME     ?= /usr/local/cuda
 MPIHOME      ?= /home/meiron/local
 OPTIMIZATION ?= 3
-CFLAGS       += -mcmodel=large
 CXXFLAGS     += $(CFLAGS)
-CUDAFLAGS     =
+CUDAFLAGS    ?=
 
 CXX          ?= g++
 MPICC        ?= mpicc
@@ -23,18 +22,24 @@ default: hybrid-sapporo
 
 hybrid-yebisu: etics yebisu
 	make -C grapite yebisu
-	make -C $(RUNDIR) grapite
+	make -C phigrape grapite
+	mkdir -p $(RUNDIR)/
+	cp -rp phigrape/phigrape $(RUNDIR)/
 
 hybrid-sapporo: etics sapporo
 	make -C grapite sapporo
-	make -C $(RUNDIR) grapite
+	make -C phigrape grapite
+	cp -rp phigrape/phigrape $(RUNDIR)/
 
 plain-yebisu: yebisu
-	make -C $(RUNDIR) yebisu
+	make -C phigrape yebisu
+	mkdir -p $(RUNDIR)/
+	cp -rp phigrape/phigrape $(RUNDIR)/
 
 plain-sapporo: sapporo
 	cp -rp sapporo2/testCodes/src/g6lib.h sapporo2/lib/grape6.h
-	make -C $(RUNDIR) sapporo
+	make -C phigrape sapporo
+	cp -rp phigrape/phigrape $(RUNDIR)/
 
 etics:
 	make -C etics/src library
@@ -44,7 +49,9 @@ yebisu:
 
 sapporo:
 	make -C sapporo2/lib
-	cd $(RUNDIR) && ln -s ../sapporo2/lib/CUDAKernels CUDA
+	mkdir -p $(RUNDIR)/
+	rm -rf $(RUNDIR)/CUDA
+	cp -rp ../sapporo2/lib/CUDAKernels $(RUNDIR)/CUDA
 
 
 ##################
@@ -58,6 +65,6 @@ clean:
 	make -C yebisu clean
 	make -C sapporo2/lib clean
 	make -C grapite clean
-	make -C $(RUNDIR) clean
+	make -C phigrape clean
 
 .PHONY: default hybrid-yebisu hybrid-sapporo plain-yebisu plain-yebisu etics yebisu sapporo

grapite

@@ -1 +1 @@
-Subproject commit 0d155035ce04cb12c1a1f09e6ae875f589c80281
+Subproject commit aa58cd9b21892ece6729fbdf9062746286ae2374

phigrape

@@ -1 +1 @@
-Subproject commit 11c0db23a3c3680fccffec7e53779a16e603f64c
+Subproject commit 0a2344563d4916a0947f4da7f661a5206ace098e

sapporo2/.gitignore

@@ -0,0 +1,5 @@
+*
+
+!.gitignore
+!get_sapporo.sh
+!patch.diff

sapporo2/patch.diff

@@ -20,19 +20,105 @@ diff -ruN
  __device__ __forceinline__ double atomicMin(double *address, double val)
  {
 diff -ruN
+--- lib/include/cudadev.h
++++
+@@ -80,7 +80,7 @@
+       assert(false);\
+     } }
+ 
+-#define cuSafeCall( call )       CU_SAFE_CALL_NO_SYNC(call);
++#define cuSafeCall( call )       call;
+ 
+   class context {
+   protected:
+diff -ruN
+--- lib/include/sapdevclass.h
++++
+@@ -33,7 +33,10 @@
+ 
+ #include <cassert>
+ #include <iostream>
+-#include <omp.h>
++#define omp_get_num_procs()   1
++#define omp_get_thread_num()  0
++#define omp_get_num_threads() 1
++#define omp_get_max_threads() 1
+ 
+ #include <sys/time.h>
+ 
+diff -ruN
+--- lib/interfaces/sapporoG6lib.cpp
++++
+@@ -45,34 +45,9 @@
+   
+   int g6_open_(int *id)
+   {
+-    //Check for a config file if its there use it  
+-    id = id;    //Make the compiler happy
+-    int *devList = NULL;
+-    int how_many = 0;
+-    FILE *fd;
+-    if ((fd = fopen("sapporo2.config", "r"))) {
+-      char line[256];
+-      fprintf(stderr, "sapporo2::open - config file is found\n");
+-      if(fgets(line, 256, fd) != NULL)
+-        sscanf(line, "%d", &how_many);
+-
+-      //Read the devices we want to use
+-      if(how_many > 0)
+-      {
+-        devList = new int[how_many];
+-        for (int i = 0; i < how_many; i++) {
+-            if(fgets(line, 256, fd) != NULL)
+-              sscanf(line, "%d", &devList[i]);
+-        }
+-      }
+-    } else {
+-      fprintf(stderr," sapporo2::open - no config file is found \n");
+-      how_many = 0;
+-    }
+-    int res = g6_open_special(how_many, devList);
+-    
+-    delete[] devList;
+-    
++    int devList = *id;
++    int how_many = 1;
++    int res = g6_open_special(how_many, &devList);
+     return res;
+   }
+ 
+diff -ruN
 --- lib/Makefile
 +++
-@@ -5,7 +5,8 @@
+@@ -1,11 +1,13 @@
+-CXX = g++
+-CC  = gcc
+-LD  = g++ 
+-F90  = ifort
++CXX ?= g++
++CC  ?= gcc
++LD  ?= g++ 
++F90 ?= ifort
  
  .SUFFIXES: .o .cpp .ptx .cu
  
 -CUDA_TK  ?= /usr/local/cuda
 +CUDAHOME ?= /usr/local/cuda
 +CUDA_TK   = $(CUDAHOME)
++OPTIMIZATION ?= 3
  
  
  testRunFlags1=
-@@ -36,17 +37,8 @@
+@@ -24,7 +26,7 @@
+ 	testRunFlags3="-D TIMING_STATS=1"
+ endif
+ 
+-OFLAGS = -g -O3 -Wall -Wextra -Wstrict-aliasing=2 -fopenmp
++OFLAGS = -g -O$(OPTIMIZATION) -Wall -Wextra -Wstrict-aliasing=2
+ 
+ #Use below if compiling with CPU_SUPPORT (SSE)
+ #CXXFLAGS += ${testRunFlags} -fPIC $(OFLAGS) -I$(CUDA_TK)/include -msse4
+@@ -36,23 +38,14 @@
  NVCC      = $(CUDA_TK)/bin/nvcc
  
  
@@ -48,7 +134,14 @@ diff -ruN
 -#NVCCFLAGS ?= -arch sm_30
 -#NVCCFLAGS = -arch sm_20
 +GPUARCH   ?= sm_75
-+NVCCFLAGS  = -arch $(GPUARCH)
++NVCCFLAGS  = -arch $(GPUARCH) -ccbin $(CXX)
  NVCCFLAGS +=  ${testRunFlags}
  
  # Use with Mac OS X
+ # NVCCFLAGS = -arch sm_12 -Xcompiler="-Duint=unsigned\ int"
+ 
+-LDFLAGS = -lcuda -fopenmp 
++LDFLAGS = -lcuda
+ 
+ 
+ INCLUDEPATH = ./include