Merge pull request #6 from xianyi/develop

merge develop

Author: martin-frbg

.travis.yml

@@ -85,8 +85,8 @@ jobs:
       sudo: true
       language: minimal
       before_install:
-        - "wget 'https://raw.githubusercontent.com/alpinelinux/alpine-chroot-install/v0.6.0/alpine-chroot-install' \
-              && echo 'a827a4ba3d0817e7c88bae17fe34e50204983d1e  alpine-chroot-install' | sha1sum -c || exit 1"
+        - "wget 'https://raw.githubusercontent.com/alpinelinux/alpine-chroot-install/v0.9.0/alpine-chroot-install' \
+          && echo 'e5dfbbdc0c4b3363b99334510976c86bfa6cb251  alpine-chroot-install' | sha1sum -c || exit 1"
         - alpine() { /alpine/enter-chroot -u "$USER" "$@"; }
       install:
         - sudo sh alpine-chroot-install -p 'build-base gfortran perl linux-headers'

Makefile.rule

@@ -152,6 +152,9 @@ NO_AFFINITY = 1
 # FUNCTION_PROFILE = 1
 
 # Support for IEEE quad precision(it's *real* REAL*16)( under testing)
+# This option should not be used - it is a holdover from unfinished code present
+# in the original GotoBLAS2 library that may be usable as a starting point but
+# is not even expected to compile in its present form.
 # QUAD_PRECISION = 1
 
 # Theads are still working for a while after finishing BLAS operation

c_check

@@ -205,7 +205,7 @@ $binformat    = bin64  if ($data =~ /BINARY_64/);
 $no_avx512= 0;
 if (($architecture eq "x86") || ($architecture eq "x86_64")) {
     $code = '"vbroadcastss -4 * 4(%rsi), %zmm2"';
-    print $tmpf "int main(void){ __asm__ volatile($code); }\n";
+    print $tmpf "#include <immintrin.h>\n\nint main(void){ __asm__ volatile($code); }\n";
     $args = " -march=skylake-avx512 -o $tmpf.o -x c $tmpf";
     my @cmd = ("$compiler_name $args >/dev/null 2>/dev/null");
     system(@cmd) == 0;

cmake/fc.cmake

@@ -3,6 +3,11 @@
 ## Description: Ported from portion of OpenBLAS/Makefile.system
 ##              Sets Fortran related variables.
 
+if (INTERFACE64)
+  set(SUFFIX64 64)
+  set(SUFFIX64_UNDERSCORE _64)
+endif()
+
 if (${F_COMPILER} STREQUAL "FLANG")
   set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_FLANG")
   if (BINARY64 AND INTERFACE64)

cmake/openblas.pc.in

@@ -1,10 +1,11 @@
 libdir=@CMAKE_INSTALL_FULL_LIBDIR@
+libsuffix=@SUFFIX64_UNDERSCORE@
 includedir=@CMAKE_INSTALL_FULL_INCLUDEDIR@
 
 openblas_config=USE_64BITINT=@USE_64BITINT@ NO_CBLAS=@NO_CBLAS@ NO_LAPACK=@NO_LAPACK@ NO_LAPACKE=@NO_LAPACKE@ DYNAMIC_ARCH=@DYNAMIC_ARCH@ DYNAMIC_OLDER=@DYNAMIC_OLDER@ NO_AFFINITY=@NO_AFFINITY@ USE_OPENMP=@USE_OPENMP@ @CORE@ MAX_THREADS=@NUM_THREADS@ 
 Name: OpenBLAS
 Description: OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version
 Version: @OPENBLAS_VERSION@
 URL: https://github.com/xianyi/OpenBLAS
-Libs: -L${libdir} -lopenblas
+Libs: -L${libdir} -lopenblas${libsuffix}
 Cflags: -I${includedir}

cmake/system_check.cmake

@@ -67,7 +67,7 @@ else()
 endif()
 
 if (X86_64 OR X86)
-  file(WRITE ${PROJECT_BINARY_DIR}/avx512.tmp "int main(void){ __asm__ volatile(\"vbroadcastss -4 * 4(%rsi), %zmm2\"); }")
+  file(WRITE ${PROJECT_BINARY_DIR}/avx512.tmp "#include <immintrin.h>\n\nint main(void){ __asm__ volatile(\"vbroadcastss -4 * 4(%rsi), %zmm2\"); }")
 execute_process(COMMAND ${CMAKE_C_COMPILER} -march=skylake-avx512 -v -o ${PROJECT_BINARY_DIR}/avx512.o -x c ${PROJECT_BINARY_DIR}/avx512.tmp OUTPUT_QUIET ERROR_QUIET RESULT_VARIABLE NO_AVX512)
 if (NO_AVX512 EQUAL 1)
 set (CCOMMON_OPT "${CCOMMON_OPT} -DNO_AVX512")

driver/others/memory.c

@@ -2587,20 +2587,20 @@ void *blas_memory_alloc(int procpos){
 
   position = 0;
 
+  LOCK_COMMAND(&alloc_lock);
   do {
 /*    if (!memory[position].used) { */
-      LOCK_COMMAND(&alloc_lock);
 /*      blas_lock(&memory[position].lock);*/
 
       if (!memory[position].used) goto allocation;
 
-      UNLOCK_COMMAND(&alloc_lock);
 /*      blas_unlock(&memory[position].lock);*/
 /*    } */
 
     position ++;
 
   } while (position < NUM_BUFFERS);
+  UNLOCK_COMMAND(&alloc_lock);
 
   goto error;
 

exports/Makefile

@@ -114,9 +114,9 @@ $(LIBDYNNAME) : ../$(LIBNAME).osx.renamed osx.def
 endif
 ifneq (,$(filter 1 2,$(NOFORTRAN)))
 #only build without Fortran
-	$(CC) $(CFLAGS) -all_load -headerpad_max_install_names -install_name "$(CURDIR)/../$(LIBDYNNAME)" -dynamiclib -o ../$(LIBDYNNAME) $< -Wl,-exported_symbols_list,osx.def  $(FEXTRALIB)
+	$(CC) $(CFLAGS) $(LDFLAGS) -all_load -headerpad_max_install_names -install_name "$(CURDIR)/../$(LIBDYNNAME)" -dynamiclib -o ../$(LIBDYNNAME) $< -Wl,-exported_symbols_list,osx.def  $(FEXTRALIB)
 else
-	$(FC) $(FFLAGS) -all_load -headerpad_max_install_names -install_name "$(CURDIR)/../$(LIBDYNNAME)" -dynamiclib -o ../$(LIBDYNNAME) $< -Wl,-exported_symbols_list,osx.def  $(FEXTRALIB)
+	$(FC) $(FFLAGS) $(LDFLAGS) -all_load -headerpad_max_install_names -install_name "$(CURDIR)/../$(LIBDYNNAME)" -dynamiclib -o ../$(LIBDYNNAME) $< -Wl,-exported_symbols_list,osx.def  $(FEXTRALIB)
 endif
 
 dllinit.$(SUFFIX) : dllinit.c

interface/lapack/laswp.c

@@ -97,7 +97,7 @@ int NAME(blasint *N, FLOAT *a, blasint *LDA, blasint *K1, blasint *K2, blasint *
 
   blas_level1_thread(mode, n, k1, k2, dummyalpha,
 		     a, lda, NULL, 0, ipiv, incx,
-		     laswp[flag], nthreads);
+		     (int(*)())laswp[flag], nthreads);
   }
 #endif
 

interface/lapack/zlaswp.c

@@ -96,7 +96,7 @@ int NAME(blasint *N, FLOAT *a, blasint *LDA, blasint *K1, blasint *K2, blasint *
   mode  =  BLAS_SINGLE  | BLAS_COMPLEX;
 #endif
 
-  blas_level1_thread(mode, n, k1, k2, dummyalpha, a, lda, NULL, 0, ipiv, incx, laswp[flag], nthreads);
+  blas_level1_thread(mode, n, k1, k2, dummyalpha, a, lda, NULL, 0, ipiv, incx, (int(*)())laswp[flag], nthreads);
   }
 #endif