Merge branch 'develop' into release-0.2.16

Author: xianyi

.travis.yml

@@ -25,6 +25,7 @@ before_install:
  - if [[ "$TARGET_BOX" == "LINUX32" ]]; then sudo apt-get install -qq  gcc-multilib gfortran-multilib; fi
 
 script: 
+ - set -e
  - make QUIET_MAKE=1 DYNAMIC_ARCH=1 TARGET=NEHALEM NUM_THREADS=32 $BTYPE
  - if [ "$TARGET_BOX" == "LINUX32" ] || [ "$TARGET_BOX" == "LINUX64" ]; then make -C test DYNAMIC_ARCH=1 TARGET=NEHALEM NUM_THREADS=32 $BTYPE; fi
  - if [ "$TARGET_BOX" == "LINUX32" ] || [ "$TARGET_BOX" == "LINUX64" ]; then make -C ctest DYNAMIC_ARCH=1 TARGET=NEHALEM NUM_THREADS=32 $BTYPE; fi

CONTRIBUTORS.md

@@ -141,5 +141,11 @@ In chronological order:
 * Martin Koehler <https://github.com/grisuthedragon/>
   * [2015-09-07] Improved imatcopy
 
+* Ashwin Sekhar T K <https://github.com/ashwinyes/>
+  * [2015-11-09] Assembly kernels for Cortex-A57 (ARMv8)
+  * [2015-11-20] lapack-test fixes for Cortex-A57
+  * [2016-03-14] Additional functional Assembly Kernels for Cortex-A57
+  * [2016-03-14] Optimize Dgemm 4x4 for Cortex-A57
+
 * [Your name or handle] <[email or website]>
   * [Date] [Brief summary of your changes]

Makefile

@@ -83,20 +83,20 @@ shared :
 ifndef NO_SHARED
 ifeq ($(OSNAME), $(filter $(OSNAME),Linux SunOS))
 	@$(MAKE) -C exports so
-	@-ln -fs $(LIBSONAME) $(LIBPREFIX).so
-	@-ln -fs $(LIBSONAME) $(LIBPREFIX).so.$(MAJOR_VERSION)
+	@ln -fs $(LIBSONAME) $(LIBPREFIX).so
+	@ln -fs $(LIBSONAME) $(LIBPREFIX).so.$(MAJOR_VERSION)
 endif
 ifeq ($(OSNAME), FreeBSD)
 	@$(MAKE) -C exports so
-	@-ln -fs $(LIBSONAME) $(LIBPREFIX).so
+	@ln -fs $(LIBSONAME) $(LIBPREFIX).so
 endif
 ifeq ($(OSNAME), NetBSD)
 	@$(MAKE) -C exports so
-	@-ln -fs $(LIBSONAME) $(LIBPREFIX).so
+	@ln -fs $(LIBSONAME) $(LIBPREFIX).so
 endif
 ifeq ($(OSNAME), Darwin)
 	@$(MAKE) -C exports dyn
-	@-ln -fs $(LIBDYNNAME) $(LIBPREFIX).dylib
+	@ln -fs $(LIBDYNNAME) $(LIBPREFIX).dylib
 endif
 ifeq ($(OSNAME), WINNT)
 	@$(MAKE) -C exports dll

USAGE.md

@@ -0,0 +1,199 @@
+# Notes on OpenBLAS usage
+## Usage
+
+#### Program is Terminated. Because you tried to allocate too many memory regions
+
+In OpenBLAS, we mange a pool of memory buffers and allocate the number of
+buffers as the following.
+```
+#define NUM_BUFFERS (MAX_CPU_NUMBER * 2)
+```
+This error indicates that the program exceeded the number of buffers.
+
+Please build OpenBLAS with larger `NUM_THREADS`. For example, `make
+NUM_THREADS=32` or `make NUM_THREADS=64`.  In `Makefile.system`, we will set
+`MAX_CPU_NUMBER=NUM_THREADS`.
+
+#### How can I use OpenBLAS in multi-threaded applications?
+
+If your application is already multi-threaded, it will conflict with OpenBLAS
+multi-threading. Thus, you must set OpenBLAS to use single thread in any of the
+following ways:
+
+* `export OPENBLAS_NUM_THREADS=1` in the environment variables.
+* Call `openblas_set_num_threads(1)` in the application on runtime.
+* Build OpenBLAS single thread version, e.g. `make USE_THREAD=0`
+
+If the application is parallelized by OpenMP, please use OpenBLAS built with
+`USE_OPENMP=1`
+
+#### How to choose TARGET manually at runtime when compiled with DYNAMIC_ARCH
+
+The environment variable which control the kernel selection is
+`OPENBLAS_CORETYPE` (see `driver/others/dynamic.c`) e.g. `export
+OPENBLAS_CORETYPE=Haswell` and the function `char* openblas_get_corename()`
+returns the used target.
+
+#### How could I disable OpenBLAS threading affinity on runtime?
+
+You can define the `OPENBLAS_MAIN_FREE` or `GOTOBLAS_MAIN_FREE` environment
+variable to disable threading affinity on runtime. For example, before the
+running,
+```
+export OPENBLAS_MAIN_FREE=1
+```
+
+Alternatively, you can disable affinity feature with enabling `NO_AFFINITY=1`
+in `Makefile.rule`.
+
+## Linking with the library
+
+* Link with shared library
+
+`gcc -o test test.c -I /your_path/OpenBLAS/include/ -L/your_path/OpenBLAS/lib -lopenblas`
+
+If the library is multithreaded, please add `-lpthread`. If the library
+contains LAPACK functions, please add `-lgfortran` or other Fortran libs.
+
+* Link with static library
+
+`gcc -o test test.c /your/path/libopenblas.a`
+
+You can download `test.c` from https://gist.github.com/xianyi/5780018
+
+On Linux, if OpenBLAS was compiled with threading support (`USE_THREAD=1` by
+default), custom programs statically linked against `libopenblas.a` should also
+link with the pthread library e.g.:
+
+```
+gcc -static -I/opt/OpenBLAS/include -L/opt/OpenBLAS/lib -o my_program my_program.c -lopenblas -lpthread
+```
+
+Failing to add the `-lpthread` flag will cause errors such as:
+
+```
+/opt/OpenBLAS/libopenblas.a(memory.o): In function `_touch_memory':
+memory.c:(.text+0x15): undefined reference to `pthread_mutex_lock'
+memory.c:(.text+0x41): undefined reference to `pthread_mutex_unlock'
+...
+```
+
+## Code examples
+
+#### Call CBLAS interface
+This example shows calling cblas_dgemm in C. https://gist.github.com/xianyi/6930656
+```
+#include <cblas.h>
+#include <stdio.h>
+
+void main()
+{
+  int i=0;
+  double A[6] = {1.0,2.0,1.0,-3.0,4.0,-1.0};
+  double B[6] = {1.0,2.0,1.0,-3.0,4.0,-1.0};
+  double C[9] = {.5,.5,.5,.5,.5,.5,.5,.5,.5};
+  cblas_dgemm(CblasColMajor, CblasNoTrans, CblasTrans,3,3,2,1,A, 3, B, 3,2,C,3);
+
+  for(i=0; i<9; i++)
+    printf("%lf ", C[i]);
+  printf("\n");
+}
+```
+`gcc -o test_cblas_open test_cblas_dgemm.c -I /your_path/OpenBLAS/include/ -L/your_path/OpenBLAS/lib -lopenblas -lpthread -lgfortran`
+
+#### Call BLAS Fortran interface
+
+This example shows calling dgemm Fortran interface in C. https://gist.github.com/xianyi/5780018
+
+```
+#include "stdio.h"
+#include "stdlib.h"
+#include "sys/time.h"
+#include "time.h"
+
+extern void dgemm_(char*, char*, int*, int*,int*, double*, double*, int*, double*, int*, double*, double*, int*);
+
+int main(int argc, char* argv[])
+{
+  int i;
+  printf("test!\n");
+  if(argc<4){
+    printf("Input Error\n");
+    return 1;
+  }
+
+  int m = atoi(argv[1]);
+  int n = atoi(argv[2]);
+  int k = atoi(argv[3]);
+  int sizeofa = m * k;
+  int sizeofb = k * n;
+  int sizeofc = m * n;
+  char ta = 'N';
+  char tb = 'N';
+  double alpha = 1.2;
+  double beta = 0.001;
+
+  struct timeval start,finish;
+  double duration;
+
+  double* A = (double*)malloc(sizeof(double) * sizeofa);
+  double* B = (double*)malloc(sizeof(double) * sizeofb);
+  double* C = (double*)malloc(sizeof(double) * sizeofc);
+
+  srand((unsigned)time(NULL));
+
+  for (i=0; i<sizeofa; i++)
+    A[i] = i%3+1;//(rand()%100)/10.0;
+
+  for (i=0; i<sizeofb; i++)
+    B[i] = i%3+1;//(rand()%100)/10.0;
+
+  for (i=0; i<sizeofc; i++)
+    C[i] = i%3+1;//(rand()%100)/10.0;
+  //#if 0
+  printf("m=%d,n=%d,k=%d,alpha=%lf,beta=%lf,sizeofc=%d\n",m,n,k,alpha,beta,sizeofc);
+  gettimeofday(&start, NULL);
+  dgemm_(&ta, &tb, &m, &n, &k, &alpha, A, &m, B, &k, &beta, C, &m);
+  gettimeofday(&finish, NULL);
+
+  duration = ((double)(finish.tv_sec-start.tv_sec)*1000000 + (double)(finish.tv_usec-start.tv_usec)) / 1000000;
+  double gflops = 2.0 * m *n*k;
+  gflops = gflops/duration*1.0e-6;
+
+  FILE *fp;
+  fp = fopen("timeDGEMM.txt", "a");
+  fprintf(fp, "%dx%dx%d\t%lf s\t%lf MFLOPS\n", m, n, k, duration, gflops);
+  fclose(fp);
+
+  free(A);
+  free(B);
+  free(C);
+  return 0;
+}
+```
+
+` gcc -o time_dgemm time_dgemm.c /your/path/libopenblas.a`
+
+` ./time_dgemm <m> <n> <k> `
+
+## Troubleshooting
+* Please read [Faq](https://github.com/xianyi/OpenBLAS/wiki/Faq) at first.
+* Please use gcc version 4.6 and above to compile Sandy Bridge AVX kernels on Linux/MingW/BSD.
+* Please use Clang version 3.1 and above to compile the library on Sandy Bridge microarchitecture. The Clang 3.0 will generate the wrong AVX binary code.
+* The number of CPUs/Cores should less than or equal to 256. On Linux x86_64(amd64), there is experimental support for up to 1024 CPUs/Cores and 128 numa nodes if you build the library with BIGNUMA=1.
+* OpenBLAS does not set processor affinity by default. On Linux, you can enable processor affinity by commenting the line NO_AFFINITY=1 in Makefile.rule. But this may cause [the conflict with R parallel](https://stat.ethz.ch/pipermail/r-sig-hpc/2012-April/001348.html).
+* On Loongson 3A. make test would be failed because of pthread_create error. The error code is EAGAIN. However, it will be OK when you run the same testcase on shell.
+
+## BLAS reference manual
+If you want to understand every BLAS function and definition, please read
+[Intel MKL reference manual](https://software.intel.com/sites/products/documentation/doclib/iss/2013/mkl/mklman/GUID-F7ED9FB8-6663-4F44-A62B-61B63C4F0491.htm)
+or [netlib.org](http://netlib.org/blas/)
+
+Here are [OpenBLAS extension functions](https://github.com/xianyi/OpenBLAS/wiki/OpenBLAS-Extensions)
+
+## How to reference OpenBLAS.
+
+You can reference our [papers](https://github.com/xianyi/OpenBLAS/wiki/publications).
+
+Alternatively, you can cite the OpenBLAS homepage http://www.openblas.net directly.
+

benchmark/Makefile

@@ -2134,7 +2134,7 @@ zgemm3m.$(SUFFIX) : gemm3m.c
 	$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
 
 smallscaling: smallscaling.c ../$(LIBNAME)
-	$(CC) $(CFLAGS) -lpthread -fopenmp -lm -o $(@F) $^
+	$(CC) $(CFLAGS) -o $(@F) $^ $(EXTRALIB) -fopenmp -lm
 
 clean ::
 	@rm -f *.goto *.mkl *.acml *.atlas *.veclib

benchmark/smallscaling.c

@@ -23,28 +23,32 @@ typedef struct {
 
 void * s_create_matrix(int size) {
     float * r = malloc(size * sizeof(double));
-    for(int i = 0; i < size; i++)
+    int i;
+    for(i = 0; i < size; i++)
         r[i] = 1e3 * i / size;
     return r;
 }
 
 void * c_create_matrix(int size) {
     float * r = malloc(size * 2 * sizeof(double));
-    for(int i = 0; i < 2 * size; i++)
+    int i;
+    for(i = 0; i < 2 * size; i++)
         r[i] = 1e3 * i / size;
     return r;
 }
 
 void * z_create_matrix(int size) {
     double * r = malloc(size * 2 * sizeof(double));
-    for(int i = 0; i < 2 * size; i++)
+    int i;
+    for(i = 0; i < 2 * size; i++)
         r[i] = 1e3 * i / size;
     return r;
 }
 
 void * d_create_matrix(int size) {
     double * r = malloc(size * sizeof(double));
-    for(int i = 0; i < size; i++)
+    int i;
+    for(i = 0; i < size; i++)
         r[i] = 1e3 * i / size;
     return r;
 }
@@ -188,4 +192,5 @@ int main(int argc, char * argv[]) {
             size *= inc_factor;
         }
     }
+    return(0);
 }

common.h

@@ -332,12 +332,13 @@ typedef int blasint;
 #endif
 #endif
 
-
+/*
 #ifdef PILEDRIVER
 #ifndef YIELDING
 #define YIELDING        __asm__ __volatile__ ("nop;nop;nop;nop;nop;nop;nop;nop;\n");
 #endif
 #endif
+*/
 
 /*
 #ifdef STEAMROLLER

common_power.h

@@ -236,7 +236,7 @@ static inline int blas_quickdivide(blasint x, blasint y){
 #define HAVE_PREFETCH
 #endif
 
-#if defined(POWER3) || defined(POWER6) || defined(PPCG4) || defined(CELL)
+#if defined(POWER3) || defined(POWER6) || defined(PPCG4) || defined(CELL) || defined(POWER8)
 #define DCBT_ARG	0
 #else
 #define DCBT_ARG	8
@@ -258,6 +258,13 @@ static inline int blas_quickdivide(blasint x, blasint y){
 #define L1_PREFETCH	dcbtst
 #endif
 
+#if defined(POWER8)
+#define L1_DUALFETCH
+#define L1_PREFETCHSIZE (16 + 128 * 100)
+#define L1_PREFETCH	dcbtst
+#endif
+
+#
 #ifndef L1_PREFETCH
 #define L1_PREFETCH	dcbt
 #endif
@@ -790,6 +797,8 @@ Lmcount$lazy_ptr:
 #define BUFFER_SIZE     (  2 << 20)
 #elif defined(PPC440FP2)
 #define BUFFER_SIZE     ( 16 << 20)
+#elif defined(POWER8)
+#define BUFFER_SIZE     ( 64 << 20)
 #else
 #define BUFFER_SIZE     ( 16 << 20)
 #endif

common_x86_64.h

@@ -396,7 +396,7 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 
 #define PROFCODE
 
-#define EPILOGUE .end	 REALNAME
+#define EPILOGUE .end
 #endif
 
 #if defined(OS_LINUX) || defined(OS_FREEBSD) || defined(OS_NETBSD) || defined(__ELF__) || defined(C_PGI)

cpuid_power.c

@@ -55,6 +55,7 @@
 #define CPUTYPE_POWER6     5
 #define CPUTYPE_CELL       6
 #define CPUTYPE_PPCG4	   7
+#define CPUTYPE_POWER8     8
 
 char *cpuname[] = {
   "UNKNOWN",
@@ -65,6 +66,7 @@ char *cpuname[] = {
   "POWER6",
   "CELL",
   "PPCG4",
+  "POWER8"
 };
 
 char *lowercpuname[] = {
@@ -76,6 +78,7 @@ char *lowercpuname[] = {
   "power6",
   "cell",
   "ppcg4",
+  "power8"
 };
 
 char *corename[] = {
@@ -87,6 +90,7 @@ char *corename[] = {
   "POWER6",
   "CELL",
   "PPCG4",
+  "POWER8"
 };
 
 int detect(void){
@@ -115,7 +119,7 @@ int detect(void){
   if (!strncasecmp(p, "POWER5", 6)) return CPUTYPE_POWER5;
   if (!strncasecmp(p, "POWER6", 6)) return CPUTYPE_POWER6;
   if (!strncasecmp(p, "POWER7", 6)) return CPUTYPE_POWER6;
-  if (!strncasecmp(p, "POWER8", 6)) return CPUTYPE_POWER6;
+  if (!strncasecmp(p, "POWER8", 6)) return CPUTYPE_POWER8;
   if (!strncasecmp(p, "Cell",   4)) return CPUTYPE_CELL;
   if (!strncasecmp(p, "7447",   4)) return CPUTYPE_PPCG4;