ref #62. Added the user friendly message with USE_OPENMP=1. The users should use...

When OpenBLAS is compiled with USE_OPENMP=1, it ignores OPENBLAS_NUM_THREADS and GOTO_NUM_THREADS flags.Therefore, you should use OMP_NUM_THREADS.

Without setting OMP_NUM_THREADS, a process will use maximal number of threads on a computing node. Thus, if there are 2 processes on the computing node, the thread will contend against other threads on CPU cores. As a result, the application will hang.

diff --git a/Makefile b/Makefile
index 56d4910..af8b793 100644 (file)
--- a/Makefile
+++ b/Makefile
@@ -57,6 +57,13 @@ else
        @echo " (Multi threaded; Max num-threads is $(NUM_THREADS))"
 endif
 
+ifeq ($(USE_OPENMP), 1)
+       @echo
+       @echo " Use OpenMP in the multithreading. Becasue of ignoring OPENBLAS_NUM_THREADS and GOTO_NUM_THREADS flags, "
+       @echo " you should use OMP_NUM_THREADS environment variable to control the number of threads."
+       @echo
+endif
+
 ifeq ($(OSNAME), Darwin)
        @echo "WARNING: If you plan to use the dynamic library $(LIBDYNNAME), you must run:"
        @echo

diff --git a/README b/README
index 2487415..79ab48d 100644 (file)
--- a/README
+++ b/README
@@ -48,7 +48,7 @@ export OMP_NUM_THREADS=4
 
 The priorities are OPENBLAS_NUM_THREADS > GOTO_NUM_THREADS > OMP_NUM_THREADS.
 
-If you compile this lib with USE_OPENMP=1, you should only set OMP_NUM_THREADS environment variable.
+If you compile this lib with USE_OPENMP=1, you should set OMP_NUM_THREADS environment variable. OpenBLAS ignores OPENBLAS_NUM_THREADS and GOTO_NUM_THREADS with USE_OPENMP=1.
 
 4.2 Set the number of threads with calling functions. for example,
 void goto_set_num_threads(int num_threads);

diff --git a/ctest/Makefile b/ctest/Makefile
index c15a681..1e07bd1 100644 (file)
--- a/ctest/Makefile
+++ b/ctest/Makefile
@@ -36,22 +36,43 @@ ztestl3o = c_zblas3.o c_z3chke.o auxiliary.o c_xerbla.o constant.o
 all :: all1 all2 all3 
 
 all1: xscblat1 xdcblat1 xccblat1 xzcblat1
+ifeq ($(USE_OPENMP), 1)
+       OMP_NUM_THREADS=2 ./xscblat1
+       OMP_NUM_THREADS=2 ./xdcblat1
+       OMP_NUM_THREADS=2 ./xccblat1
+       OMP_NUM_THREADS=2 ./xzcblat1
+else
        OPENBLAS_NUM_THREADS=2 ./xscblat1
        OPENBLAS_NUM_THREADS=2 ./xdcblat1
        OPENBLAS_NUM_THREADS=2 ./xccblat1
        OPENBLAS_NUM_THREADS=2 ./xzcblat1
+endif
 
 all2: xscblat2 xdcblat2 xccblat2 xzcblat2
+ifeq ($(USE_OPENMP), 1)
+       OMP_NUM_THREADS=2 ./xscblat2 < sin2
+       OMP_NUM_THREADS=2 ./xdcblat2 < din2
+       OMP_NUM_THREADS=2 ./xccblat2 < cin2
+       OMP_NUM_THREADS=2 ./xzcblat2 < zin2
+else
        OPENBLAS_NUM_THREADS=2 ./xscblat2 < sin2
        OPENBLAS_NUM_THREADS=2 ./xdcblat2 < din2
        OPENBLAS_NUM_THREADS=2 ./xccblat2 < cin2
        OPENBLAS_NUM_THREADS=2 ./xzcblat2 < zin2
+endif
 
 all3: xscblat3 xdcblat3 xccblat3 xzcblat3
+ifeq ($(USE_OPENMP), 1)
+       OMP_NUM_THREADS=2 ./xscblat3 < sin3
+       OMP_NUM_THREADS=2 ./xdcblat3 < din3
+       OMP_NUM_THREADS=2 ./xccblat3 < cin3
+       OMP_NUM_THREADS=2 ./xzcblat3 < zin3
+else
        OPENBLAS_NUM_THREADS=2 ./xscblat3 < sin3
        OPENBLAS_NUM_THREADS=2 ./xdcblat3 < din3
        OPENBLAS_NUM_THREADS=2 ./xccblat3 < cin3
        OPENBLAS_NUM_THREADS=2 ./xzcblat3 < zin3
+endif
 
 clean ::
        rm -f x* 

diff --git a/test/Makefile b/test/Makefile
index a5308e4..2df499b 100644 (file)
--- a/test/Makefile
+++ b/test/Makefile
@@ -4,29 +4,46 @@ include ../Makefile.system
 all :: level1 level2 level3
 
 level1 : sblat1 dblat1 cblat1 zblat1
-       OPENBLAS_NUM_THREADS=1 ./sblat1
-       OPENBLAS_NUM_THREADS=1 ./dblat1
-       OPENBLAS_NUM_THREADS=1 ./cblat1
-       OPENBLAS_NUM_THREADS=1 ./zblat1
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./sblat1
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./dblat1
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./cblat1
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./zblat1
 ifdef SMP
+ifeq ($(USE_OPENMP), 1)
+       OMP_NUM_THREADS=2 ./sblat1
+       OMP_NUM_THREADS=2 ./dblat1
+       OMP_NUM_THREADS=2 ./cblat1
+       OMP_NUM_THREADS=2 ./zblat1
+else
        OPENBLAS_NUM_THREADS=2 ./sblat1
        OPENBLAS_NUM_THREADS=2 ./dblat1
        OPENBLAS_NUM_THREADS=2 ./cblat1
        OPENBLAS_NUM_THREADS=2 ./zblat1
 endif
+endif
 
 level2 : sblat2 dblat2 cblat2 zblat2
        rm -f ?BLAT2.SUMM
-       OPENBLAS_NUM_THREADS=1 ./sblat2 < ./sblat2.dat
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./sblat2 < ./sblat2.dat
        @$(GREP) -q FATAL SBLAT2.SUMM && cat SBLAT2.SUMM || exit 0
-       OPENBLAS_NUM_THREADS=1 ./dblat2 < ./dblat2.dat
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./dblat2 < ./dblat2.dat
        @$(GREP) -q FATAL DBLAT2.SUMM && cat DBLAT2.SUMM || exit 0
-       OPENBLAS_NUM_THREADS=1 ./cblat2 < ./cblat2.dat
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./cblat2 < ./cblat2.dat
        @$(GREP) -q FATAL CBLAT2.SUMM && cat CBLAT2.SUMM || exit 0
-       OPENBLAS_NUM_THREADS=1 ./zblat2 < ./zblat2.dat
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./zblat2 < ./zblat2.dat
        @$(GREP) -q FATAL ZBLAT2.SUMM && cat ZBLAT2.SUMM || exit 0
 ifdef SMP
        rm -f ?BLAT2.SUMM
+ifeq ($(USE_OPENMP), 1)
+       OMP_NUM_THREADS=2 ./sblat2 < ./sblat2.dat
+       @$(GREP) -q FATAL SBLAT2.SUMM && cat SBLAT2.SUMM || exit 0
+       OMP_NUM_THREADS=2 ./dblat2 < ./dblat2.dat
+       @$(GREP) -q FATAL DBLAT2.SUMM && cat DBLAT2.SUMM || exit 0
+       OMP_NUM_THREADS=2 ./cblat2 < ./cblat2.dat
+       @$(GREP) -q FATAL CBLAT2.SUMM && cat CBLAT2.SUMM || exit 0
+       OMP_NUM_THREADS=2 ./zblat2 < ./zblat2.dat
+       @$(GREP) -q FATAL ZBLAT2.SUMM && cat ZBLAT2.SUMM || exit 0
+else
        OPENBLAS_NUM_THREADS=2 ./sblat2 < ./sblat2.dat
        @$(GREP) -q FATAL SBLAT2.SUMM && cat SBLAT2.SUMM || exit 0
        OPENBLAS_NUM_THREADS=2 ./dblat2 < ./dblat2.dat
@@ -36,19 +53,30 @@ ifdef SMP
        OPENBLAS_NUM_THREADS=2 ./zblat2 < ./zblat2.dat
        @$(GREP) -q FATAL ZBLAT2.SUMM && cat ZBLAT2.SUMM || exit 0
 endif
+endif
 
 level3 : sblat3 dblat3 cblat3 zblat3
        rm -f ?BLAT3.SUMM
-       OPENBLAS_NUM_THREADS=1 ./sblat3 < ./sblat3.dat
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./sblat3 < ./sblat3.dat
        @$(GREP) -q FATAL SBLAT3.SUMM && cat SBLAT3.SUMM || exit 0
-       OPENBLAS_NUM_THREADS=1 ./dblat3 < ./dblat3.dat
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./dblat3 < ./dblat3.dat
        @$(GREP) -q FATAL DBLAT3.SUMM && cat DBLAT3.SUMM || exit 0
-       OPENBLAS_NUM_THREADS=1 ./cblat3 < ./cblat3.dat
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./cblat3 < ./cblat3.dat
        @$(GREP) -q FATAL CBLAT3.SUMM && cat CBLAT3.SUMM || exit 0
-       OPENBLAS_NUM_THREADS=1 ./zblat3 < ./zblat3.dat
+       OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./zblat3 < ./zblat3.dat
        @$(GREP) -q FATAL ZBLAT3.SUMM && cat ZBLAT3.SUMM || exit 0
 ifdef SMP
        rm -f ?BLAT3.SUMM
+ifeq ($(USE_OPENMP), 1)
+       OMP_NUM_THREADS=2 ./sblat3 < ./sblat3.dat
+       @$(GREP) -q FATAL SBLAT3.SUMM && cat SBLAT3.SUMM || exit 0
+       OMP_NUM_THREADS=2 ./dblat3 < ./dblat3.dat
+       @$(GREP) -q FATAL DBLAT3.SUMM && cat DBLAT3.SUMM || exit 0
+       OMP_NUM_THREADS=2 ./cblat3 < ./cblat3.dat
+       @$(GREP) -q FATAL CBLAT3.SUMM && cat CBLAT3.SUMM || exit 0
+       OMP_NUM_THREADS=2 ./zblat3 < ./zblat3.dat
+       @$(GREP) -q FATAL ZBLAT3.SUMM && cat ZBLAT3.SUMM || exit 0
+else
        OPENBLAS_NUM_THREADS=2 ./sblat3 < ./sblat3.dat
        @$(GREP) -q FATAL SBLAT3.SUMM && cat SBLAT3.SUMM || exit 0
        OPENBLAS_NUM_THREADS=2 ./dblat3 < ./dblat3.dat
@@ -58,6 +86,7 @@ ifdef SMP
        OPENBLAS_NUM_THREADS=2 ./zblat3 < ./zblat3.dat
        @$(GREP) -q FATAL ZBLAT3.SUMM && cat ZBLAT3.SUMM || exit 0
 endif
+endif
 
 FLDFLAGS = $(FFLAGS:-fPIC=)
 CEXTRALIB =