Merge pull request #3304 from xianyi/develop

Merge develop int0 0.3.0 for release 0.3.16

.github/workflows/nightly-Homebrew-build.yml

@@ -43,11 +43,6 @@ jobs:
       - name: Update Homebrew
         if: github.event_name != 'pull_request'
         run: brew update || true
-
-      - name: unlink installed gcc to allow updating
-        run: |
-          brew unlink gcc@8
-          brew unlink gcc@9
           
       - name: Install prerequisites
         run: brew install --fetch-HEAD --HEAD --only-dependencies --keep-tmp openblas

.travis.yml

@@ -224,12 +224,21 @@ matrix:
       before_script:
         - COMMON_FLAGS="DYNAMIC_ARCH=1 NUM_THREADS=32"
         - brew update
-        - brew install gcc@10
       script:
         - travis_wait 45 make QUIET_MAKE=1 $COMMON_FLAGS $BTYPE
       env:
-        - BTYPE="TARGET=NEHALEM BINARY=64 INTERFACE64=1 FC=gfortran-10"
-        
+        - BTYPE="TARGET=HASWELL USE_OPENMP=1 BINARY=64 INTERFACE64=1 CC=gcc-10 FC=gfortran-10"
+
+    - <<: *test-macos
+      osx_image: xcode12
+      before_script:
+        - COMMON_FLAGS="DYNAMIC_ARCH=1 NUM_THREADS=32"
+        - brew update
+      script:
+        - travis_wait 45 make QUIET_MAKE=1 $COMMON_FLAGS $BTYPE
+      env:
+        - BTYPE="TARGET=NEHALEM BINARY=64 INTERFACE64=1 FC=gfortran-10"      
+
   #  - <<: *test-macos
   #    osx_image: xcode10
   #    env:

CMakeLists.txt

@@ -6,7 +6,7 @@ cmake_minimum_required(VERSION 2.8.5)
 project(OpenBLAS C ASM)
 set(OpenBLAS_MAJOR_VERSION 0)
 set(OpenBLAS_MINOR_VERSION 3)
-set(OpenBLAS_PATCH_VERSION 14)
+set(OpenBLAS_PATCH_VERSION 16)
 set(OpenBLAS_VERSION "${OpenBLAS_MAJOR_VERSION}.${OpenBLAS_MINOR_VERSION}.${OpenBLAS_PATCH_VERSION}")
 
 # Adhere to GNU filesystem layout conventions

CONTRIBUTORS.md

@@ -194,3 +194,6 @@ In chronological order:
 
 * PingTouGe Semiconductor Co., Ltd.
   * [2020-10] Add RISC-V Vector (0.7.1) support. Optimize BLAS kernels for Xuantie C910
+
+* River Dillon <oss@outerpassage.net>
+  * [2021-07-10] fix compilation with musl libc

Changelog.txt

@@ -1,4 +1,102 @@
 OpenBLAS ChangeLog
+====================================================================
+Version 0.3.16
+ 11-Jul-2021
+
+common:
+ - drastically reduced the stack size requirements for running the LAPACK
+   testsuite (Reference-LAPACK PR 553)
+ - fixed spurious test failures in the LAPACK testsuite (Reference-LAPACK
+   PR 564)
+ - expressly setting DYNAMIC_ARCH=0 no longer enables dynamic_arch mode 
+ - improved performance of xGER, xSPR, xSPR2, xSYR, xSYR2, xTRSV, SGEMV_N 
+   and DGEMV_N, for small input sizes and consecutive arguments
+ - improved performance of xGETRF, xPORTF and xPOTRI for small input sizes
+   by disabling multithreading
+ - fixed installing with BSD versions of the "install" utility
+
+RISCV:
+ - fixed the implementation of xIMIN
+ - improved the performance of DSDOT
+ - fixed linking of the tests on C910V with current vendor gcc
+
+POWER:
+- fixed SBGEMM computation for some odd value inputs
+- fixed compilation for PPCG4, PPC970, POWER3, POWER4 and POWER5
+
+x86_64:
+ - improved performance of SGEMV_N and SGEMV_T for small N on AVX512-capable cpus
+ - worked around a miscompilation of ZGEMM/ZTRMM on Sandybridge with old gcc
+   versions
+ - fixed compilation with MS Visual Studio versions older than 2017
+ - fixed macro name collision with winnt.h from the latest Win10 SDK
+ - added cpu type autodetection for Intel Ice Lake SP
+ - fixed cpu type autodetection for Intel Tiger Lake
+ - added cpu type autodetection for recent Centaur/Zhaoxin models
+ - fixed compilation with musl libc
+
+ARM64:
+- fixed compilation with gcc/gfortran on the Apple M1
+- fixed linking of the tests on FreeBSD
+- fixed missing restore of a register in the recently rewritten DNRM2 kernel
+  for ThunderX2 and Neoverse N1 that could cause spurious failures in e.g.
+  DGEEV
+- added compiler optimization flags for the EMAG8180
+- added initial support for Cortex A55
+
+ARM:
+- fixed linking of the tests on FreeBSD
+
+====================================================================
+Version 0.3.15
+  2-May-2021
+
+common:
+ - imported improvements and bugfixes from Reference-LAPACK 3.9.1
+ - imported LAPACKE interface fixes from Reference-LAPACK PRs 534 + 537
+ - fixed a problem in the cpu detection of 0.3.14 that prevented cross-compilation
+ - fixed a sequence problem in the generation of softlinks to the library in GMAKE
+
+RISC V:
+ - fixed compilation on RISCV (missing entry in getarch)
+ - fixed a potential division by zero in CROTG and ZROTG
+
+POWER:
+ - fixed LAPACK testsuite failures seen with the NVIDIA HPC compiler
+ - improved CGEMM, DGEMM and ZGEMM performance on POWER10
+ - added an optimized ZGEMV kernel for POWER10
+ - fixed a potential division by zero in CROTG and ZROTG
+
+x86_64:
+ - added support for Intel Control-flow Enforcement Technology (CET)
+ - reverted the DOMATCOPY_RT code to the generic C version
+ - fixed a bug in the AVX512 SGEMM kernel introduced in 0.3.14
+ - fixed misapplication of -msse flag to non-SSE cpus in DYNAMIC_ARCH
+ - added support for compilation of the benchmarks on older OSX versions
+ - fix propagation of the NO_AVX512 option in CMAKE builds
+ - fix compilation of the AVX512 SGEMM kernel with clang-cl on Windows
+ - fixed compilation of the CTESTs with INTERFACE64=1 (random faults on OSX)
+ - corrected the Haswell DROT kernel to require AVX2/FMA3 rather than AVX512
+
+ARM:
+ - fixed a potential division by zero in CROTG and ZROTG
+ - fixed a potential overflow in IMATCOPY/ZIMATCOPY and the CTESTs
+
+ARM64:
+ - fixed spurious reads outside the array in the SGEMM tcopy macro
+ - fixed a potential division by zero in CROTG and ZROTG
+ - fixed a segmentation fault in DYNAMIC_ARCH builds (reappeared in 0.3.14)
+
+MIPS
+ - fixed a potential division by zero in CROTG and ZROTG
+ - fixed a potential overflow in IMATCOPY/ZIMATCOPY and the CTESTs
+
+MIPS64:
+ - fixed a potential division by zero in CROTG and ZROTG
+
+SPARC:
+ - fixed a potential division by zero in CROTG and ZROTG
+
 ====================================================================
 Version 0.3.14
  17-Mar-2021

Makefile

@@ -167,7 +167,6 @@ ifeq ($(NO_SHARED), 1)
 	$(error OpenBLAS: neither static nor shared are enabled.)
 endif
 endif
-	@-ln -fs $(LIBNAME) $(LIBPREFIX).$(LIBSUFFIX)
 	@for d in $(SUBDIRS) ; \
 	do if test -d $$d; then \
 	  $(MAKE) -C $$d $(@F) || exit 1 ; \
@@ -196,6 +195,7 @@ endif
 ifdef USE_THREAD
 	@echo USE_THREAD=$(USE_THREAD) >>  Makefile.conf_last
 endif
+	@-ln -fs $(LIBNAME) $(LIBPREFIX).$(LIBSUFFIX)
 	@touch lib.grd
 
 prof : prof_blas prof_lapack

Makefile.arm64

@@ -57,6 +57,28 @@ endif
 endif
 endif
 
+# Use a53 tunings because a55 is only available in GCC>=8.1
+ifeq ($(CORE), CORTEXA55)
+ifeq ($(GCCVERSIONGTEQ7), 1)
+ifeq ($(GCCVERSIONGTEQ8), 1)
+CCOMMON_OPT += -march=armv8.2-a -mtune=cortex-a55
+ifneq ($(F_COMPILER), NAG)
+FCOMMON_OPT += -march=armv8.2-a -mtune=cortex-a55
+endif
+else
+CCOMMON_OPT += -march=armv8.2-a -mtune=cortex-a53
+ifneq ($(F_COMPILER), NAG)
+FCOMMON_OPT += -march=armv8.2-a -mtune=cortex-a53
+endif
+endif
+else
+CCOMMON_OPT += -march=armv8-a -mtune=cortex-a53
+ifneq ($(F_COMPILER), NAG)
+FCOMMON_OPT += -march=armv8-a -mtune=cortex-a53
+endif
+endif
+endif
+
 ifeq ($(CORE), THUNDERX)
 CCOMMON_OPT += -march=armv8-a -mtune=thunderx
 ifneq ($(F_COMPILER), NAG)
@@ -107,4 +129,13 @@ FCOMMON_OPT += -march=armv8.2-a -mtune=tsv110
 endif
 endif
 endif
+
+ifeq ($(GCCVERSIONGTEQ9), 1)
+ifeq ($(CORE), EMAG8180)
+CCOMMON_OPT += -march=armv8-a -mtune=emag
+ifneq ($(F_COMPILER), NAG)
+FCOMMON_OPT += -march=armv8-a -mtune=emag
+endif
+endif
+endif
 endif

Makefile.install

@@ -74,17 +74,17 @@ endif
 ifneq ($(OSNAME), AIX)
 ifndef NO_LAPACKE
 	@echo Copying LAPACKE header files to $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)
-	@-install -pm644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapack.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapack.h"
-	@-install -pm644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapacke.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke.h"
-	@-install -pm644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapacke_config.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke_config.h"
-	@-install -pm644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapacke_mangling_with_flags.h.in "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke_mangling.h"
-	@-install -pm644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapacke_utils.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke_utils.h"
+	@-install -m644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapack.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapack.h"
+	@-install -m644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapacke.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke.h"
+	@-install -m644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapacke_config.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke_config.h"
+	@-install -m644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapacke_mangling_with_flags.h.in "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke_mangling.h"
+	@-install -m644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapacke_utils.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke_utils.h"
 endif
 
 #for install static library
 ifneq ($(NO_STATIC),1)
 	@echo Copying the static library to $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
-	@install -pm644 $(LIBNAME) "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)"
+	@install -m644 $(LIBNAME) "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)"
 	@cd "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)" ; \
 	ln -fs $(LIBNAME) $(LIBPREFIX).$(LIBSUFFIX)
 endif
@@ -92,7 +92,7 @@ endif
 ifneq ($(NO_SHARED),1)
 	@echo Copying the shared library to $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
 ifeq ($(OSNAME), $(filter $(OSNAME),Linux SunOS Android Haiku FreeBSD DragonFly))
-	@install -pm755 $(LIBSONAME) "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)"
+	@install -m755 $(LIBSONAME) "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)"
 	@cd "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)" ; \
 	ln -fs $(LIBSONAME) $(LIBPREFIX).so ; \
 	ln -fs $(LIBSONAME) $(LIBPREFIX).so.$(MAJOR_VERSION)

Makefile.rule

@@ -3,7 +3,7 @@
 #
 
 # This library's version
-VERSION = 0.3.14
+VERSION = 0.3.16
 
 # If you set the suffix, the library name will be libopenblas_$(LIBNAMESUFFIX).a
 # and libopenblas_$(LIBNAMESUFFIX).so. Meanwhile, the soname in shared library

Makefile.system

@@ -333,6 +333,7 @@ GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
 GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
 GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
 GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
+GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
 GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
 GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
 GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
@@ -380,6 +381,12 @@ ifeq ($(OSNAME), AIX)
 EXTRALIB	+= -lm
 endif
 
+ifeq ($(OSNAME), FreeBSD)
+ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
+EXTRALIB	+= -lm
+endif
+endif
+
 ifeq ($(OSNAME), WINNT)
 NEED_PIC = 0
 NO_EXPRECISION = 1
@@ -619,6 +626,7 @@ DYNAMIC_CORE += CORTEXA57
 DYNAMIC_CORE += CORTEXA72
 DYNAMIC_CORE += CORTEXA73
 DYNAMIC_CORE += NEOVERSEN1
+DYNAMIC_CORE += CORTEXA55
 DYNAMIC_CORE += FALKOR
 DYNAMIC_CORE += THUNDERX
 DYNAMIC_CORE += THUNDERX2T99

Makefile.x86

@@ -1,10 +1,21 @@
 # COMPILER_PREFIX = mingw32-
 
-ifdef HAVE_SSE
-CCOMMON_OPT += -msse
-FCOMMON_OPT += -msse
+ifneq ($(DYNAMIC_ARCH),1)
+ADD_CPUFLAGS = 1
+else
+ifdef TARGET_CORE
+ADD_CPUFLAGS = 1
+endif
 endif
 
+ifdef ADD_CPUFLAGS
+ifdef HAVE_SSE
+CCOMMON_OPT += -msse
+ifneq ($(F_COMPILER), NAG)
+FCOMMON_OPT += -msse
+endif
+endif
+endif
 
 ifeq ($(OSNAME), Interix)
 ARFLAGS		= -m x86

Makefile.x86_64

@@ -8,6 +8,16 @@ endif
 endif
 endif
 
+
+ifneq ($(DYNAMIC_ARCH),1)
+ADD_CPUFLAGS = 1
+else
+ifdef TARGET_CORE
+ADD_CPUFLAGS = 1
+endif
+endif
+
+ifdef ADD_CPUFLAGS
 ifdef HAVE_SSE3
 CCOMMON_OPT += -msse3
 ifneq ($(F_COMPILER), NAG)
@@ -44,7 +54,6 @@ endif
 endif
 
 ifeq ($(CORE), SKYLAKEX)
-ifndef DYNAMIC_ARCH
 ifndef NO_AVX512
 CCOMMON_OPT += -march=skylake-avx512
 ifneq ($(F_COMPILER), NAG)
@@ -62,10 +71,8 @@ endif
 endif
 endif
 endif
-endif
 
 ifeq ($(CORE), COOPERLAKE)
-ifndef DYNAMIC_ARCH
 ifndef NO_AVX512
 ifeq ($(C_COMPILER), GCC)
 # cooperlake support was added in 10.1
@@ -88,7 +95,6 @@ endif
 endif
 endif
 endif
-endif
 
 ifdef HAVE_AVX2
 ifndef NO_AVX2
@@ -120,6 +126,7 @@ endif
 endif
 endif
 
+endif
 
 
 ifeq ($(OSNAME), Interix)

README.md

@@ -27,7 +27,7 @@ We provide official binary packages for the following platform:
 
   * Windows x86/x86_64
 
-You can download them from [file hosting on sourceforge.net](https://sourceforge.net/projects/openblas/files/).
+You can download them from [file hosting on sourceforge.net](https://sourceforge.net/projects/openblas/files/) or from the Releases section of the github project page, [https://github.com/xianyi/OpenBLAS/releases](https://github.com/xianyi/OpenBLAS/releases).
 
 ## Installation from Source
 

TargetList.txt

@@ -92,6 +92,7 @@ CORTEXA57
 CORTEXA72
 CORTEXA73
 NEOVERSEN1
+CORTEXA55
 EMAG8180
 FALKOR
 THUNDERX

appveyor.yml

@@ -47,6 +47,7 @@ environment:
 install:
   - if [%COMPILER%]==[clang-cl] call %CONDA_INSTALL_LOCN%\Scripts\activate.bat
   - if [%COMPILER%]==[clang-cl] conda config --add channels conda-forge --force
+  - if [%COMPILER%]==[clang-cl] conda config --set auto_update_conda false
   - if [%COMPILER%]==[clang-cl] conda install --yes --quiet clangdev cmake ninja flang=11.0.1
   - if [%COMPILER%]==[clang-cl] call "C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Auxiliary\Build\vcvarsall.bat" x64
   - if [%COMPILER%]==[clang-cl] set "LIB=%CONDA_INSTALL_LOCN%\Library\lib;%LIB%"

azure-pipelines.yml

@@ -4,7 +4,15 @@ trigger:
   branches:
     include:
       - develop
-
+resources:
+  containers:
+      - container: oneapi-hpckit
+        image: intel/oneapi-hpckit:latest
+        options: '-v /usr/bin/sudo:/usr/bin/sudo -v /usr/lib/sudo/libsudo_util.so.0:/usr/lib/sudo/libsudo_util.so.0 -v /usr/lib/sudo/sudoers.so:/usr/lib/sudo/sudoers.so'
+      - container: oneapi-basekit
+        image: intel/oneapi-basekit:latest
+        options: '-v /usr/bin/sudo:/usr/bin/sudo -v /usr/lib/sudo/libsudo_util.so.0:/usr/lib/sudo/libsudo_util.so.0 -v /usr/lib/sudo/sudoers.so:/usr/lib/sudo/sudoers.so'
+ 
 jobs:
 # manylinux1 is useful to test because the
 # standard Docker container uses an old version
@@ -68,4 +76,92 @@ jobs:
       dir
       openblas_utest.exe
   
-      
+- job: OSX_OpenMP
+  pool:
+     vmImage: 'macOS-10.15'
+  steps:   
+  - script: |
+      brew update
+      make TARGET=CORE2 DYNAMIC_ARCH=1 USE_OPENMP=1 INTERFACE64=1 CC=gcc-10 FC=gfortran-10
+     
+- job: OSX_GCC_Nothreads
+  pool:
+     vmImage: 'macOS-10.15'
+  steps:   
+  - script: |
+      brew update
+      make USE_THREADS=0 CC=gcc-10 FC=gfortran-10
+     
+- job: OSX_OpenMP_Clang
+  pool:
+     vmImage: 'macOS-10.15'
+  variables:
+     LD_LIBRARY_PATH: /usr/local/opt/llvm/lib
+     LIBRARY_PATH: /usr/local/opt/llvm/lib
+  steps:   
+  - script: |
+      brew update
+      brew install llvm libomp
+      make TARGET=CORE2 USE_OPENMP=1 INTERFACE64=1 DYNAMIC_ARCH=1 CC=/usr/local/opt/llvm/bin/clang FC=gfortran-10
+
+- job: OSX_Ifort_Clang
+  pool:
+     vmImage: 'macOS-10.15'
+  variables:
+     LD_LIBRARY_PATH: /usr/local/opt/llvm/lib
+     MACOS_HPCKIT_URL: https://registrationcenter-download.intel.com/akdlm/irc_nas/17643/m_HPCKit_p_2021.2.0.2903_offline.dmg
+     LIBRARY_PATH: /usr/local/opt/llvm/lib
+     MACOS_FORTRAN_COMPONENTS: intel.oneapi.mac.ifort-compiler
+  steps:   
+  - script: |
+      brew update
+      brew install llvm libomp
+      sudo mkdir -p /opt/intel
+      sudo chown $USER /opt/intel
+    displayName: prepare for cache restore
+  - task: Cache@2
+    inputs:
+      path: /opt/intel/oneapi
+      key: '"install" | "$(MACOS_HPCKIT_URL)" | "$(MACOS_FORTRAN_COMPONENTS)"'
+      cacheHitVar: CACHE_RESTORED
+  - script: | 
+      curl --output webimage.dmg --url $(MACOS_HPCKIT_URL) --retry 5 --retry-delay 5
+      hdiutil attach webimage.dmg
+      sudo /Volumes/"$(basename "$(MACOS_HPCKIT_URL)" .dmg)"/bootstrapper.app/Contents/MacOS/bootstrapper -s --action install --components="$(MACOS_FORTRAN_COMPONENTS)" --eula=accept --continue-with-optional-error=yes --log-dir=.
+      installer_exit_code=$?
+      hdiutil detach /Volumes/"$(basename "$URL" .dmg)" -quiet
+      exit $installer_exit_code
+    displayName: install
+    condition: ne(variables.CACHE_RESTORED, 'true')
+  - script: | 
+      source /opt/intel/oneapi/setvars.sh
+      make CC=/usr/local/opt/llvm/bin/clang FC=ifort
+ 
+- job: OSX_NDK_ARMV7
+  pool:
+     vmImage: 'macOS-10.15'
+  steps:   
+  - script: | 
+      brew update
+      brew install --cask android-ndk
+      export ANDROID_NDK_HOME=/usr/local/share/android-ndk
+      make TARGET=ARMV7 ONLY_CBLAS=1 CC=$ANDROID_NDK_HOME/toolchains/llvm/prebuilt/darwin-x86_64/bin/armv7a-linux-androideabi21-clang AR=$ANDROID_NDK_HOME/toolchains/llvm/prebuilt/darwin-x86_64/bin/arm-linux-androideabi-ar HOSTCC=gcc ARM_SOFTFP_ABI=1 -j4
+ 
+- job: ALPINE_MUSL
+  pool:
+     vmImage: 'ubuntu-latest'
+  steps:
+  - script: |
+        wget 'https://raw.githubusercontent.com/alpinelinux/alpine-chroot-install/v0.9.0/alpine-chroot-install' \
+          && echo 'e5dfbbdc0c4b3363b99334510976c86bfa6cb251  alpine-chroot-install' | sha1sum -c || exit 1
+        alpine() { /alpine/enter-chroot -u "$USER" "$@"; }
+        sudo sh alpine-chroot-install -p 'build-base gfortran perl linux-headers sudo'
+        alpine make DYNAMIC_ARCH=1 BINARY=64
+        alpine make DYNAMIC_ARCH=1 BINARY=64 PREFIX=mytestdir install
+        alpine ls -l mytestdir/include
+        alpine echo "// tests that inclusion of openblas_config.h works with musl" >test_install.c
+        alpine echo "#include <openblas_config.h>" >>test_install.c
+        alpine echo "int main(){" >> test_install.c
+        alpine echo "cpu_set_t* cpu_set = NULL;}" >>test_install.c
+        alpine gcc -Imytestdir/include test_install.c -Lmytestdir/lib -lopenblas -lpthread -lgfortran -o test_install
+        

benchmark/bench.h

@@ -3,6 +3,8 @@
 #include <time.h>
 #ifdef __CYGWIN32__
 #include <sys/time.h>
+#elif defined(__APPLE__)
+#include <mach/mach_time.h>
 #endif
 #include "common.h"
 

benchmark/getri.c

@@ -72,13 +72,17 @@ int main(int argc, char *argv[]){
   FLOAT *a,*work;
   FLOAT wkopt[4];
   blasint *ipiv;
-  blasint m, i, j, info,lwork;
+  blasint m, i, j, l, info,lwork;
 
   int from =   1;
   int to   = 200;
   int step =   1;
+  int loops =  1;
 
-  double time1;
+  double time1,timeg;
+  
+  char *p;
+  char btest = 'I';
 
   argc--;argv++;
 
@@ -86,6 +90,9 @@ int main(int argc, char *argv[]){
   if (argc > 0) { to       = MAX(atol(*argv), from);	argc--; argv++;}
   if (argc > 0) { step     = atol(*argv);		argc--; argv++;}
 
+  if ((p = getenv("OPENBLAS_TEST"))) btest=*p;
+  
+  if ((p = getenv("OPENBLAS_LOOPS"))) loops=*p;
 
   fprintf(stderr, "From : %3d  To : %3d Step = %3d\n", from, to, step);
 
@@ -124,32 +131,41 @@ int main(int argc, char *argv[]){
   fprintf(stderr, "   SIZE           FLops           Time          Lwork\n");
 
   for(m = from; m <= to; m += step){
-
+    timeg = 0.;
     fprintf(stderr, " %6d : ", (int)m);
 
-    GETRF (&m, &m, a, &m, ipiv, &info);
+    for (l = 0; l < loops; l++) {
 
+    if (btest == 'F') begin();
+    GETRF (&m, &m, a, &m, ipiv, &info);
+    if (btest == 'F') {
+      end();
+      timeg += getsec();
+    }
     if (info) {
       fprintf(stderr, "Matrix is not singular .. %d\n", info);
       exit(1);
     }
 
-    begin();
+    if (btest == 'I') begin();
 
     lwork = -1;
     GETRI(&m, a, &m, ipiv, wkopt, &lwork, &info);
 
     lwork = (blasint)wkopt[0];
     GETRI(&m, a, &m, ipiv, work, &lwork, &info);
-    end();
+    if (btest == 'I') end();
 
     if (info) {
       fprintf(stderr, "failed compute inverse matrix .. %d\n", info);
       exit(1);
     }
 
-    time1 = getsec();
-
+    if (btest == 'I') 
+      timeg += getsec();
+    
+    } // loops
+    time1 = timeg/(double)loops;
     fprintf(stderr,
 	    " %10.2f MFlops : %10.2f Sec : %d\n",
 	    COMPSIZE * COMPSIZE * (4.0/3.0 * (double)m * (double)m *(double)m - (double)m *(double)m + 5.0/3.0* (double)m) / time1 * 1.e-6,time1,lwork);

benchmark/linpack.c

@@ -72,17 +72,21 @@ int main(int argc, char *argv[]){
   FLOAT *a, *b;
   blasint *ipiv;
 
-  blasint m, i, j, info;
+  blasint m, i, j, l, info;
   blasint unit =   1;
 
   int from =   1;
   int to   = 200;
   int step =   1;
+  int loops =  1;
 
   FLOAT maxerr;
 
-  double time1, time2;
+  double time1, time2, timeg1,timeg2;
 
+  char *p;
+  if ((p = getenv("OPENBLAS_LOOPS"))) loops=*p;
+  
   argc--;argv++;
 
   if (argc > 0) { from     = atol(*argv);		argc--; argv++;}
@@ -110,9 +114,9 @@ int main(int argc, char *argv[]){
   fprintf(stderr, "   SIZE       Residual     Decompose            Solve           Total\n");
 
   for(m = from; m <= to; m += step){
-
+    timeg1 = timeg2 = 0.;
     fprintf(stderr, " %6d : ", (int)m);
-
+    for (l = 0; l < loops; l++) {
     for(j = 0; j < m; j++){
       for(i = 0; i < m * COMPSIZE; i++){
 	a[(long)i + (long)j * (long)m * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
@@ -138,7 +142,7 @@ int main(int argc, char *argv[]){
       exit(1);
     }
 
-    time1 = getsec();
+    timeg1 += getsec();
 
     begin();
 
@@ -151,8 +155,10 @@ int main(int argc, char *argv[]){
       exit(1);
     }
 
-    time2 = getsec();
-
+    timeg2 += getsec();
+    } //loops
+    time1=timeg1/(double)loops;
+    time2=timeg2/(double)loops;
     maxerr = 0.;
 
     for(i = 0; i < m; i++){

benchmark/potrf.c

@@ -99,14 +99,15 @@ int main(int argc, char *argv[]){
   char *p;
   char btest = 'F';
 
-  blasint m, i, j, info, uplos=0;
-  double flops;
+  blasint m, i, j, l, info, uplos=0;
+  double flops = 0.;
 
   int from =   1;
   int to   = 200;
   int step =   1;
+  int loops =  1;
 
-  double time1;
+  double time1, timeg;
 
   argc--;argv++;
 
@@ -119,6 +120,8 @@ int main(int argc, char *argv[]){
 
   if ((p = getenv("OPENBLAS_TEST"))) btest=*p;
 
+  if ((p = getenv("OPENBLAS_LOOPS"))) loops=*p;
+
   fprintf(stderr, "From : %3d  To : %3d Step = %3d Uplo = %c\n", from, to, step,*uplo[uplos]);
 
   if (( a    = (FLOAT *)malloc(sizeof(FLOAT) * to * to * COMPSIZE)) == NULL){
@@ -129,19 +132,21 @@ int main(int argc, char *argv[]){
     fprintf(stderr,"Out of Memory!!\n");exit(1);
   }
 
-  for(m = from; m <= to; m += step){
 
+  for(m = from; m <= to; m += step){
+    timeg=0.;
+	  for (l = 0; l < loops; l++) {
 #ifndef COMPLEX
       if (uplos & 1) {
 	for (j = 0; j < m; j++) {
 	  for(i = 0; i < j; i++)     a[(long)i + (long)j * (long)m] = 0.;
-	                             a[(long)j + (long)j * (long)m] = ((double) rand() / (double) RAND_MAX) + 8.;
+	  a[(long)j + (long)j * (long)m] = ((double) rand() / (double) RAND_MAX) + 8.;
 	  for(i = j + 1; i < m; i++) a[(long)i + (long)j * (long)m] = ((double) rand() / (double) RAND_MAX) - 0.5;
 	}
       } else {
 	for (j = 0; j < m; j++) {
 	  for(i = 0; i < j; i++)     a[(long)i + (long)j * (long)m] = ((double) rand() / (double) RAND_MAX) - 0.5;
-	                             a[(long)j + (long)j * (long)m] = ((double) rand() / (double) RAND_MAX) + 8.;
+	  a[(long)j + (long)j * (long)m] = ((double) rand() / (double) RAND_MAX) + 8.;
 	  for(i = j + 1; i < m; i++) a[(long)i + (long)j * (long)m] = 0.;
 	}
       }
@@ -192,8 +197,8 @@ int main(int argc, char *argv[]){
 	exit(1);
       }
 
-      time1 = getsec();
-      flops = COMPSIZE * COMPSIZE * (1.0/3.0 * (double)m * (double)m *(double)m +1.0/2.0* (double)m *(double)m + 1.0/6.0* (double)m) / time1 * 1.e-6;
+      if ( btest == 'F')
+	      timeg += getsec();
 
       if ( btest == 'S' )
       {
@@ -214,9 +219,7 @@ int main(int argc, char *argv[]){
 		fprintf(stderr, "Potrs info = %d\n", info);
 		exit(1);
         }
-        time1 = getsec();
-        flops = COMPSIZE * COMPSIZE * (2.0 * (double)m * (double)m *(double)m ) / time1 * 1.e-6;
-
+        timeg += getsec();
       }
 	
       if ( btest == 'I' )
@@ -232,11 +235,17 @@ int main(int argc, char *argv[]){
 		fprintf(stderr, "Potri info = %d\n", info);
 		exit(1);
         }
-
-        time1 = getsec();
-        flops = COMPSIZE * COMPSIZE * (2.0/3.0 * (double)m * (double)m *(double)m +1.0/2.0* (double)m *(double)m + 5.0/6.0* (double)m) / time1 * 1.e-6;
+        timeg += getsec();
       }
-	
+    } // loops
+
+      time1 = timeg/(double)loops;
+	if ( btest == 'F')
+		flops = COMPSIZE * COMPSIZE * (1.0/3.0 * (double)m * (double)m *(double)m +1.0/2.0* (double)m *(double)m + 1.0/6.0* (double)m) / time1 * 1.e-6;
+	if ( btest == 'S')
+		flops = COMPSIZE * COMPSIZE * (2.0 * (double)m * (double)m *(double)m ) / time1 * 1.e-6;
+	if ( btest == 'I')
+		flops = COMPSIZE * COMPSIZE * (2.0/3.0 * (double)m * (double)m *(double)m +1.0/2.0* (double)m *(double)m + 5.0/6.0* (double)m) / time1 * 1.e-6;
       fprintf(stderr, "%8d : %10.2f MFlops : %10.3f Sec : Test=%c\n",m,flops ,time1,btest);
 
 

benchmark/syr2.c

@@ -46,14 +46,17 @@ int main(int argc, char *argv[]){
 
   if ((p = getenv("OPENBLAS_UPLO"))) uplo=*p; 
 
-  blasint m, i, j;
+  blasint m, i, j, l;
   blasint inc_x= 1;
   blasint inc_y= 1;
   int from =   1;
   int to   = 200;
   int step =   1;
+  int loops =  1;
 
-  double time1;
+  if ((p = getenv("OPENBLAS_LOOPS"))) loops=*p;
+
+  double time1,timeg;
 
   argc--;argv++;
 
@@ -85,8 +88,9 @@ int main(int argc, char *argv[]){
 
   for(m = from; m <= to; m += step)
   {
-	
+    timeg = 0.;	
     fprintf(stderr, " %6d : ", (int)m);
+    for (l = 0; l < loops; l++) {
 	for(i = 0; i < m * COMPSIZE * abs(inc_x); i++){
 	x[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
    	}
@@ -107,8 +111,10 @@ int main(int argc, char *argv[]){
 
     end();
 
-    time1 = getsec();
+    timeg += getsec();
+    } // loops
 
+    time1 = timeg/(double)loops;
     fprintf(stderr,
 	    " %10.2f MFlops\n",
 	    COMPSIZE * COMPSIZE * 2. * (double)m * (double)m / time1 * 1.e-6);

benchmark/syrk.c

@@ -56,17 +56,20 @@ int main(int argc, char *argv[]){
 
   char uplo='U';
   char trans='N';
-
+  
   if ((p = getenv("OPENBLAS_UPLO"))) uplo=*p;
   if ((p = getenv("OPENBLAS_TRANS"))) trans=*p;
 
-  blasint m, i, j;
+  blasint m, i, j, l;
 
   int from =   1;
   int to   = 200;
   int step =   1;
+  int loops =  1;
 
-  double time1;
+  if ((p = getenv("OPENBLAS_LOOPS"))) loops=*p;
+
+  double time1,timeg;
 
   argc--;argv++;
 
@@ -95,9 +98,12 @@ int main(int argc, char *argv[]){
 
   for(m = from; m <= to; m += step)
   {
+    timeg = 0.;
 
     fprintf(stderr, " %6d : ", (int)m);
 
+    for(l = 0; l < loops; l++) {
+
     for(j = 0; j < m; j++){
       for(i = 0; i < m * COMPSIZE; i++){
 	a[(long)i + (long)j * (long)m * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
@@ -111,8 +117,10 @@ int main(int argc, char *argv[]){
 
     end();
 
-    time1 = getsec();
-
+    timeg += getsec();
+    
+    } //loops
+    time1 = timeg / (double)loops;
     fprintf(stderr,
 	    " %10.2f MFlops\n",
 	    COMPSIZE * COMPSIZE * 1. * (double)m * (double)m * (double)m / time1 * 1.e-6);

cmake/arch.cmake

@@ -44,7 +44,7 @@ endif ()
 
 if (DYNAMIC_ARCH)
   if (ARM64)
-    set(DYNAMIC_CORE ARMV8 CORTEXA53 CORTEXA57 CORTEXA72 CORTEXA73 FALKOR THUNDERX THUNDERX2T99 TSV110 EMAG8180 NEOVERSEN1 THUNDERX3T110)
+	  set(DYNAMIC_CORE ARMV8 CORTEXA53 CORTEXA55 CORTEXA57 CORTEXA72 CORTEXA73 FALKOR THUNDERX THUNDERX2T99 TSV110 EMAG8180 NEOVERSEN1 THUNDERX3T110)
     if (DYNAMIC_LIST)
 	    set(DYNAMIC_CORE ARMV8 ${DYNAMIC_LIST})
     endif ()

cmake/lapack.cmake

@@ -66,7 +66,7 @@ set(SLASRC
    slaqgb.f slaqge.f slaqp2.f slaqps.f slaqsb.f slaqsp.f slaqsy.f
    slaqr0.f slaqr1.f slaqr2.f slaqr3.f slaqr4.f slaqr5.f
    slaqtr.f slar1v.f slar2v.f ilaslr.f ilaslc.f
-   slarf.f  slarfb.f slarfg.f slarfgp.f slarft.f slarfx.f slarfy.f slargv.f
+   slarf.f  slarfb.f slarfb_gett.f slarfg.f slarfgp.f slarft.f slarfx.f slarfy.f slargv.f
    slarrv.f slartv.f
    slarz.f  slarzb.f slarzt.f slasy2.f
    slasyf.f slasyf_rook.f slasyf_rk.f slasyf_aa.f
@@ -112,14 +112,14 @@ set(SLASRC
    sgeqrt.f sgeqrt2.f sgeqrt3.f sgemqrt.f
    stpqrt.f stpqrt2.f stpmqrt.f stprfb.f
    sgelqt.f sgelqt3.f sgemlqt.f
-   sgetsls.f sgeqr.f slatsqr.f slamtsqr.f sgemqr.f
+   sgetsls.f sgetsqrhrt.f sgeqr.f slatsqr.f slamtsqr.f sgemqr.f
    sgelq.f slaswlq.f slamswlq.f sgemlq.f
    stplqt.f stplqt2.f stpmlqt.f
    ssytrd_2stage.f ssytrd_sy2sb.f ssytrd_sb2st.F ssb2st_kernels.f
    ssyevd_2stage.f ssyev_2stage.f ssyevx_2stage.f ssyevr_2stage.f
    ssbev_2stage.f ssbevx_2stage.f ssbevd_2stage.f ssygv_2stage.f
    sgesvdq.f slaorhr_col_getrfnp.f
-   slaorhr_col_getrfnp2.f sorgtsqr.f sorhr_col.f )
+   slaorhr_col_getrfnp2.f sorgtsqr.f sorgtsqr_row.f sorhr_col.f )
 
 set(SXLASRC sgesvxx.f sgerfsx.f sla_gerfsx_extended.f sla_geamv.f
    sla_gercond.f sla_gerpvgrw.f ssysvxx.f ssyrfsx.f
@@ -171,7 +171,7 @@ set(CLASRC
    claqhb.f claqhe.f claqhp.f claqp2.f claqps.f claqsb.f
    claqr0.f claqr1.f claqr2.f claqr3.f claqr4.f claqr5.f
    claqsp.f claqsy.f clar1v.f clar2v.f ilaclr.f ilaclc.f
-   clarf.f  clarfb.f clarfg.f clarfgp.f clarft.f
+   clarf.f  clarfb.f clarfb_gett.f clarfg.f clarfgp.f clarft.f
    clarfx.f clarfy.f clargv.f clarnv.f clarrv.f clartg.f clartv.f
    clarz.f  clarzb.f clarzt.f clascl.f claset.f clasr.f  classq.f
    clasyf.f clasyf_rook.f clasyf_rk.f clasyf_aa.f
@@ -209,14 +209,14 @@ set(CLASRC
    cgeqrt.f cgeqrt2.f cgeqrt3.f cgemqrt.f
    ctpqrt.f ctpqrt2.f ctpmqrt.f ctprfb.f
    cgelqt.f cgelqt3.f cgemlqt.f
-   cgetsls.f cgeqr.f clatsqr.f clamtsqr.f cgemqr.f
+   cgetsls.f cgetsqrhrt.f cgeqr.f clatsqr.f clamtsqr.f cgemqr.f
    cgelq.f claswlq.f clamswlq.f cgemlq.f
    ctplqt.f ctplqt2.f ctpmlqt.f
    chetrd_2stage.f chetrd_he2hb.f chetrd_hb2st.F chb2st_kernels.f
    cheevd_2stage.f cheev_2stage.f cheevx_2stage.f cheevr_2stage.f
    chbev_2stage.f chbevx_2stage.f chbevd_2stage.f chegv_2stage.f
    cgesvdq.f claunhr_col_getrfnp.f claunhr_col_getrfnp2.f 
-   cungtsqr.f cunhr_col.f )
+   cungtsqr.f cungtsqr_row.f cunhr_col.f )
 
 set(CXLASRC cgesvxx.f cgerfsx.f cla_gerfsx_extended.f cla_geamv.f
    cla_gercond_c.f cla_gercond_x.f cla_gerpvgrw.f
@@ -253,7 +253,7 @@ set(DLASRC
    dlaqgb.f dlaqge.f dlaqp2.f dlaqps.f dlaqsb.f dlaqsp.f dlaqsy.f
    dlaqr0.f dlaqr1.f dlaqr2.f dlaqr3.f dlaqr4.f dlaqr5.f
    dlaqtr.f dlar1v.f dlar2v.f iladlr.f iladlc.f
-   dlarf.f  dlarfb.f dlarfg.f dlarfgp.f dlarft.f dlarfx.f dlarfy.f
+   dlarf.f  dlarfb.f dlarfb_gett.f dlarfg.f dlarfgp.f dlarft.f dlarfx.f dlarfy.f
    dlargv.f dlarrv.f dlartv.f
    dlarz.f  dlarzb.f dlarzt.f dlasy2.f
    dlasyf.f dlasyf_rook.f dlasyf_rk.f dlasyf_aa.f
@@ -300,14 +300,14 @@ set(DLASRC
    dgeqrt.f dgeqrt2.f dgeqrt3.f dgemqrt.f
    dtpqrt.f dtpqrt2.f dtpmqrt.f dtprfb.f
    dgelqt.f dgelqt3.f dgemlqt.f
-   dgetsls.f dgeqr.f dlatsqr.f dlamtsqr.f dgemqr.f
+   dgetsls.f dgetsqrhrt.f dgeqr.f dlatsqr.f dlamtsqr.f dgemqr.f
    dgelq.f dlaswlq.f dlamswlq.f dgemlq.f
    dtplqt.f dtplqt2.f dtpmlqt.f
    dsytrd_2stage.f dsytrd_sy2sb.f dsytrd_sb2st.F dsb2st_kernels.f
    dsyevd_2stage.f dsyev_2stage.f dsyevx_2stage.f dsyevr_2stage.f
    dsbev_2stage.f dsbevx_2stage.f dsbevd_2stage.f dsygv_2stage.f
    dcombssq.f dgesvdq.f dlaorhr_col_getrfnp.f
-   dlaorhr_col_getrfnp2.f dorgtsqr.f dorhr_col.f )
+   dlaorhr_col_getrfnp2.f dorgtsqr.f dorgtsqr_row.f dorhr_col.f )
 
 set(DXLASRC dgesvxx.f dgerfsx.f dla_gerfsx_extended.f dla_geamv.f
    dla_gercond.f dla_gerpvgrw.f dsysvxx.f dsyrfsx.f
@@ -360,7 +360,7 @@ set(ZLASRC
    zlaqhb.f zlaqhe.f zlaqhp.f zlaqp2.f zlaqps.f zlaqsb.f
    zlaqr0.f zlaqr1.f zlaqr2.f zlaqr3.f zlaqr4.f zlaqr5.f
    zlaqsp.f zlaqsy.f zlar1v.f zlar2v.f ilazlr.f ilazlc.f
-   zlarcm.f zlarf.f  zlarfb.f
+   zlarcm.f zlarf.f  zlarfb.f zlarfb_gett.f
    zlarfg.f zlarfgp.f zlarft.f
    zlarfx.f zlarfy.f zlargv.f zlarnv.f zlarrv.f zlartg.f zlartv.f
    zlarz.f  zlarzb.f zlarzt.f zlascl.f zlaset.f zlasr.f
@@ -402,13 +402,13 @@ set(ZLASRC
    ztpqrt.f ztpqrt2.f ztpmqrt.f ztprfb.f
    ztplqt.f ztplqt2.f ztpmlqt.f
    zgelqt.f zgelqt3.f zgemlqt.f
-   zgetsls.f zgeqr.f zlatsqr.f zlamtsqr.f zgemqr.f
+   zgetsls.f zgetsqrhrt.f zgeqr.f zlatsqr.f zlamtsqr.f zgemqr.f
    zgelq.f zlaswlq.f zlamswlq.f zgemlq.f
    zhetrd_2stage.f zhetrd_he2hb.f zhetrd_hb2st.F zhb2st_kernels.f
    zheevd_2stage.f zheev_2stage.f zheevx_2stage.f zheevr_2stage.f
    zhbev_2stage.f zhbevx_2stage.f zhbevd_2stage.f zhegv_2stage.f
    zgesvdq.f zlaunhr_col_getrfnp.f zlaunhr_col_getrfnp2.f
-   zungtsqr.f zunhr_col.f)
+   zungtsqr.f zungtsqr_row.f zunhr_col.f)
 
 set(ZXLASRC zgesvxx.f zgerfsx.f zla_gerfsx_extended.f zla_geamv.f
    zla_gercond_c.f zla_gercond_x.f zla_gerpvgrw.f zsysvxx.f zsyrfsx.f

cmake/lapacke.cmake

@@ -114,6 +114,8 @@ set(CSRC
   lapacke_cgetrs_work.c
   lapacke_cgetsls.c
   lapacke_cgetsls_work.c
+  lapacke_cgetsqrhrt.c
+  lapacke_cgetsqrhrt_work.c 
   lapacke_cggbak.c
   lapacke_cggbak_work.c
   lapacke_cggbal.c
@@ -590,6 +592,8 @@ set(CSRC
   lapacke_cungrq_work.c
   lapacke_cungtr.c
   lapacke_cungtr_work.c
+  lapacke_cungtsqr_row.c
+  lapacke_cungtsqr_row_work.c
   lapacke_cunmbr.c
   lapacke_cunmbr_work.c
   lapacke_cunmhr.c
@@ -735,6 +739,8 @@ set(DSRC
   lapacke_dgetrs_work.c
   lapacke_dgetsls.c
   lapacke_dgetsls_work.c
+  lapacke_dgetsqrhrt.c
+  lapacke_dgetsqrhrt_work.c
   lapacke_dggbak.c
   lapacke_dggbak_work.c
   lapacke_dggbal.c
@@ -862,6 +868,8 @@ set(DSRC
   lapacke_dorgrq_work.c
   lapacke_dorgtr.c
   lapacke_dorgtr_work.c
+  lapacke_dorgtsqr_row.c
+  lapacke_dorgtsqr_row_work.c
   lapacke_dormbr.c
   lapacke_dormbr_work.c
   lapacke_dormhr.c
@@ -1309,6 +1317,8 @@ set(SSRC
   lapacke_sgetrs_work.c
   lapacke_sgetsls.c
   lapacke_sgetsls_work.c
+  lapacke_sgetsqrhrt.c
+  lapacke_sgetsqrhrt_work.c
   lapacke_sggbak.c
   lapacke_sggbak_work.c
   lapacke_sggbal.c
@@ -1435,6 +1445,8 @@ set(SSRC
   lapacke_sorgrq_work.c
   lapacke_sorgtr.c
   lapacke_sorgtr_work.c
+  lapacke_sorgtsqr_row.c
+  lapacke_sorgtsqr_row_work.c
   lapacke_sormbr.c
   lapacke_sormbr_work.c
   lapacke_sormhr.c
@@ -1877,6 +1889,8 @@ set(ZSRC
   lapacke_zgetrs_work.c
   lapacke_zgetsls.c
   lapacke_zgetsls_work.c
+  lapacke_zgetsqrhrt.c
+  lapacke_zgetsqrhrt_work.c
   lapacke_zggbak.c
   lapacke_zggbak_work.c
   lapacke_zggbal.c
@@ -2351,6 +2365,8 @@ set(ZSRC
   lapacke_zungrq_work.c
   lapacke_zungtr.c
   lapacke_zungtr_work.c
+  lapacke_zungtsqr_row.c
+  lapacke_zungtsqr_row_work.c
   lapacke_zunmbr.c
   lapacke_zunmbr_work.c
   lapacke_zunmhr.c

cmake/prebuild.cmake

@@ -177,7 +177,7 @@ if (DEFINED CORE AND CMAKE_CROSSCOMPILING AND NOT (${HOST_OS} STREQUAL "WINDOWSS
     set(ZGEMM_UNROLL_M 4)
     set(ZGEMM_UNROLL_N 4)
     set(SYMV_P 16)
-  elseif ("${TCORE}" STREQUAL "CORTEXA57" OR "${TCORE}" STREQUAL "CORTEXA53")
+  elseif ("${TCORE}" STREQUAL "CORTEXA57" OR "${TCORE}" STREQUAL "CORTEXA53" OR "${TCORE}" STREQUAL "CORTEXA55")
     file(APPEND ${TARGET_CONF_TEMP}
       "#define L1_CODE_SIZE\t32768\n"
       "#define L1_CODE_LINESIZE\t64\n"

cmake/system.cmake

@@ -39,7 +39,7 @@ if (DEFINED BINARY AND DEFINED TARGET AND BINARY EQUAL 32)
   if (${TARGET} STREQUAL "BULLDOZER" OR ${TARGET} STREQUAL "PILEDRIVER" OR ${TARGET} STREQUAL "ZEN")
     set(TARGET "BARCELONA")
   endif ()
-  if (${TARGET} STREQUAL "ARMV8" OR ${TARGET} STREQUAL "CORTEXA57" OR ${TARGET} STREQUAL "CORTEXA53")
+  if (${TARGET} STREQUAL "ARMV8" OR ${TARGET} STREQUAL "CORTEXA57" OR ${TARGET} STREQUAL "CORTEXA53" OR ${TARGET} STREQUAL "CORTEXA55")
     set(TARGET "ARMV7")
   endif ()
 endif ()
@@ -299,6 +299,10 @@ if (NO_AVX2)
   set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_AVX2")
 endif ()
 
+if (NO_AVX512)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_AVX512")
+endif ()
+
 if (USE_THREAD)
   # USE_SIMPLE_THREADED_LEVEL3 = 1
   # NO_AFFINITY = 1

cmake/utils.cmake

@@ -254,6 +254,19 @@ function(GenerateNamedObjects sources_in)
       # now add the object and set the defines
       set(obj_defines ${defines_in})
 
+      list(FIND obj_defines "RC" def_idx)
+      if (${def_idx} GREATER -1) 
+	      #	      list(REMOVE_AT ${obj_defines} ${def_idx})
+	      list (REMOVE_ITEM obj_defines "RC")
+	      list(APPEND obj_defines  "RC=RC")
+      endif ()
+      list(FIND obj_defines "CR" def_idx)
+      if (${def_idx} GREATER -1) 
+	      #	      list(REMOVE_AT ${obj_defines} ${def_idx})
+	      list (REMOVE_ITEM obj_defines "CR")
+	      list(APPEND obj_defines  "CR=CR")
+      endif ()
+
       if (use_cblas)
         set(obj_name "cblas_${obj_name}")
         list(APPEND obj_defines "CBLAS")

common.h

@@ -416,6 +416,15 @@ please https://github.com/xianyi/OpenBLAS/issues/246
 #include "common_alpha.h"
 #endif
 
+#if (defined(ARCH_X86) || defined(ARCH_X86_64)) && defined(__CET__) && defined(__has_include)
+#if __has_include(<cet.h>)
+#include <cet.h>
+#endif
+#endif
+#ifndef _CET_ENDBR
+#define _CET_ENDBR
+#endif
+
 #ifdef ARCH_X86
 #include "common_x86.h"
 #endif

common_interface.h

@@ -709,6 +709,13 @@ int BLASFUNC(cpotrf)(char *, blasint *, float  *, blasint *, blasint *);
 int BLASFUNC(zpotrf)(char *, blasint *, double *, blasint *, blasint *);
 int BLASFUNC(xpotrf)(char *, blasint *, xdouble *, blasint *, blasint *);
 
+int BLASFUNC(spotri)(char *, blasint *, float  *, blasint *, blasint *);
+int BLASFUNC(dpotri)(char *, blasint *, double *, blasint *, blasint *);
+int BLASFUNC(qpotri)(char *, blasint *, xdouble *, blasint *, blasint *);
+int BLASFUNC(cpotri)(char *, blasint *, float  *, blasint *, blasint *);
+int BLASFUNC(zpotri)(char *, blasint *, double *, blasint *, blasint *);
+int BLASFUNC(xpotri)(char *, blasint *, xdouble *, blasint *, blasint *);
+
 int BLASFUNC(spotrs)(char *, blasint *, blasint *, float   *, blasint *, float   *, blasint *, blasint *);
 int BLASFUNC(dpotrs)(char *, blasint *, blasint *, double  *, blasint *, double  *, blasint *, blasint *);
 int BLASFUNC(qpotrs)(char *, blasint *, blasint *, xdouble *, blasint *, xdouble *, blasint *, blasint *);

common_x86.h

@@ -340,7 +340,8 @@ REALNAME:
 	.align 16; \
 	.globl REALNAME ;\
        .type REALNAME, @function; \
-REALNAME:
+REALNAME: \
+	_CET_ENDBR
 
 #ifdef PROFILE
 #define PROFCODE call mcount

common_x86_64.h

@@ -451,7 +451,8 @@ REALNAME:
 	.align 512; \
 	.globl REALNAME ;\
        .type REALNAME, @function; \
-REALNAME:
+REALNAME: \
+	_CET_ENDBR
 
 #ifdef PROFILE
 #define PROFCODE call *mcount@GOTPCREL(%rip)

cpuid_arm64.c

@@ -36,6 +36,7 @@ size_t length=sizeof(value);
 #define CPU_ARMV8       	1
 // Arm
 #define CPU_CORTEXA53     2
+#define CPU_CORTEXA55     14
 #define CPU_CORTEXA57     3
 #define CPU_CORTEXA72     4
 #define CPU_CORTEXA73     5
@@ -67,7 +68,8 @@ static char *cpuname[] = {
   "EMAG8180",
   "NEOVERSEN1",
   "THUNDERX3T110",
-  "VORTEX"	
+  "VORTEX",
+  "CORTEXA55"
 };
 
 static char *cpuname_lower[] = {
@@ -84,7 +86,8 @@ static char *cpuname_lower[] = {
   "emag8180",
   "neoversen1",
   "thunderx3t110",
-  "vortex"	
+  "vortex",
+  "cortexa55"
 };
 
 int get_feature(char *search)
@@ -161,6 +164,8 @@ int detect(void)
         return CPU_CORTEXA73;
       else if (strstr(cpu_part, "0xd0c"))
         return CPU_NEOVERSEN1;
+      else if (strstr(cpu_part, "0xd05"))
+	return CPU_CORTEXA55;
     }
     // Qualcomm
     else if (strstr(cpu_implementer, "0x51") && strstr(cpu_part, "0xc00"))
@@ -281,6 +286,7 @@ void get_cpuconfig(void)
 	{
 
     case CPU_CORTEXA53:
+    case CPU_CORTEXA55:			
       printf("#define %s\n", cpuname[d]);
       // Fall-through
 		case CPU_ARMV8:

cpuid_x86.c

@@ -283,6 +283,7 @@ int get_vendor(void){
   if (!strcmp(vendor, "CyrixInstead")) return VENDOR_CYRIX;
   if (!strcmp(vendor, "NexGenDriven")) return VENDOR_NEXGEN;
   if (!strcmp(vendor, "CentaurHauls")) return VENDOR_CENTAUR;
+  if (!strcmp(vendor, "  Shanghai  ")) return VENDOR_CENTAUR;
   if (!strcmp(vendor, "RiseRiseRise")) return VENDOR_RISE;
   if (!strcmp(vendor, " SiS SiS SiS")) return VENDOR_SIS;
   if (!strcmp(vendor, "GenuineTMx86")) return VENDOR_TRANSMETA;
@@ -1398,6 +1399,17 @@ int get_cpuname(void){
 	    return CPUTYPE_SANDYBRIDGE;
 	  else
 	  return CPUTYPE_NEHALEM;
+	case 10: // Ice Lake SP
+	  if(support_avx512_bf16())
+            return CPUTYPE_COOPERLAKE;	
+          if(support_avx512())
+            return CPUTYPE_SKYLAKEX;
+          if(support_avx2())
+            return CPUTYPE_HASWELL;
+          if(support_avx())
+	    return CPUTYPE_SANDYBRIDGE;
+	  else
+	  return CPUTYPE_NEHALEM;	
         }
       break;
       case 7: // family 6 exmodel 7
@@ -1620,7 +1632,9 @@ int get_cpuname(void){
     case 0x6:
       return CPUTYPE_NANO;
       break;
-
+    case 0x7:
+      return CPUTYPE_NEHALEM;
+      break;
     }
     return CPUTYPE_VIAC3;
   }
@@ -2112,7 +2126,22 @@ int get_coretype(void){
 #endif
 	  else
 	    return CORE_NEHALEM;
-#endif			
+#endif
+	if (model == 10)
+#ifndef NO_AVX512
+	  if(support_avx512_bf16())
+            return CORE_COOPERLAKE;
+	  return CORE_SKYLAKEX;
+#else
+	  if(support_avx())
+#ifndef NO_AVX2
+	    return CORE_HASWELL;
+#else
+	    return CORE_SANDYBRIDGE;
+#endif
+	  else
+	    return CORE_NEHALEM;
+#endif	
         break;    	
       case 7:
         if (model == 10) 
@@ -2135,13 +2164,13 @@ int get_coretype(void){
       case 8:
        if (model == 12) { // Tiger Lake
           if(support_avx512())
-            return CPUTYPE_SKYLAKEX;
+            return CORE_SKYLAKEX;
           if(support_avx2())
-            return CPUTYPE_HASWELL;
+            return CORE_HASWELL;
           if(support_avx())
-            return CPUTYPE_SANDYBRIDGE;
+            return CORE_SANDYBRIDGE;
           else
-          return CPUTYPE_NEHALEM;
+          return CORE_NEHALEM;
         }
         if (model == 14) { // Kaby Lake 
 	  if(support_avx())
@@ -2259,6 +2288,9 @@ int get_coretype(void){
     case 0x6:
       return CORE_NANO;
       break;
+    case 0x7:
+      return CORE_NEHALEM;
+      break;
     }
     return CORE_VIAC3;
   }

ctest/c_cblas2.c

@@ -20,7 +20,7 @@ void F77_cgemv(int *order, char *transp, int *m, int *n,
   get_transpose_type(transp, &trans);
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A  = (CBLAS_TEST_COMPLEX *)malloc( (*m)*LDA*sizeof( CBLAS_TEST_COMPLEX) );
+     A  = (CBLAS_TEST_COMPLEX *)malloc( (*m)*(size_t)LDA*sizeof( CBLAS_TEST_COMPLEX) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ ){
            A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -50,7 +50,7 @@ void F77_cgbmv(int *order, char *transp, int *m, int *n, int *kl, int *ku,
   get_transpose_type(transp, &trans);
   if (*order == TEST_ROW_MJR) {
      LDA = *ku+*kl+2;
-     A=( CBLAS_TEST_COMPLEX* )malloc((*n+*kl)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+     A=( CBLAS_TEST_COMPLEX* )malloc((*n+*kl)*(size_t)LDA*sizeof(CBLAS_TEST_COMPLEX));
      for( i=0; i<*ku; i++ ){
         irow=*ku+*kl-i;
         jcol=(*ku)-i;
@@ -94,7 +94,7 @@ void F77_cgeru(int *order, int *m, int *n, CBLAS_TEST_COMPLEX *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A=(CBLAS_TEST_COMPLEX*)malloc((*m)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+     A=(CBLAS_TEST_COMPLEX*)malloc((*m)*(size_t)LDA*sizeof(CBLAS_TEST_COMPLEX));
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ ){
            A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -122,7 +122,7 @@ void F77_cgerc(int *order, int *m, int *n, CBLAS_TEST_COMPLEX *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+     A=(CBLAS_TEST_COMPLEX* )malloc((*m)*(size_t)LDA*sizeof(CBLAS_TEST_COMPLEX ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ ){
            A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -154,7 +154,7 @@ void F77_chemv(int *order, char *uplow, int *n, CBLAS_TEST_COMPLEX *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A = (CBLAS_TEST_COMPLEX *)malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+     A = (CBLAS_TEST_COMPLEX *)malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_COMPLEX));
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ ){
            A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -190,7 +190,7 @@ int i,irow,j,jcol,LDA;
 		 *incx, beta, y, *incy );
      else {
         LDA = *k+2;
-        A =(CBLAS_TEST_COMPLEX*)malloc((*n+*k)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        A =(CBLAS_TEST_COMPLEX*)malloc((*n+*k)*(size_t)LDA*sizeof(CBLAS_TEST_COMPLEX));
         if (uplo == CblasUpper) {
            for( i=0; i<*k; i++ ){
               irow=*k-i;
@@ -251,8 +251,8 @@ void F77_chpmv(int *order, char *uplow, int *n, CBLAS_TEST_COMPLEX *alpha,
 	         beta, y, *incy);
      else {
         LDA = *n;
-        A = (CBLAS_TEST_COMPLEX* )malloc(LDA*LDA*sizeof(CBLAS_TEST_COMPLEX ));
-        AP = (CBLAS_TEST_COMPLEX* )malloc( (((LDA+1)*LDA)/2)*
+        A = (CBLAS_TEST_COMPLEX* )malloc((size_t)LDA*LDA*sizeof(CBLAS_TEST_COMPLEX ));
+        AP = (CBLAS_TEST_COMPLEX* )malloc( ((((size_t)LDA+1)*LDA)/2)*
 	        sizeof( CBLAS_TEST_COMPLEX ));
         if (uplo == CblasUpper) {
            for( j=0, k=0; j<*n; j++ )
@@ -311,7 +311,7 @@ void F77_ctbmv(int *order, char *uplow, char *transp, char *diagn,
 	x, *incx);
      else {
         LDA = *k+2;
-        A=(CBLAS_TEST_COMPLEX *)malloc((*n+*k)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        A=(CBLAS_TEST_COMPLEX *)malloc((*n+*k)*(size_t)LDA*sizeof(CBLAS_TEST_COMPLEX));
         if (uplo == CblasUpper) {
            for( i=0; i<*k; i++ ){
               irow=*k-i;
@@ -375,7 +375,7 @@ void F77_ctbsv(int *order, char *uplow, char *transp, char *diagn,
 	         *incx);
      else {
         LDA = *k+2;
-        A=(CBLAS_TEST_COMPLEX*)malloc((*n+*k)*LDA*sizeof(CBLAS_TEST_COMPLEX ));
+        A=(CBLAS_TEST_COMPLEX*)malloc((*n+*k)*(size_t)LDA*sizeof(CBLAS_TEST_COMPLEX ));
         if (uplo == CblasUpper) {
            for( i=0; i<*k; i++ ){
               irow=*k-i;
@@ -436,8 +436,8 @@ void F77_ctpmv(int *order, char *uplow, char *transp, char *diagn,
         cblas_ctpmv( CblasRowMajor, UNDEFINED, trans, diag, *n, ap, x, *incx );
      else {
         LDA = *n;
-        A=(CBLAS_TEST_COMPLEX*)malloc(LDA*LDA*sizeof(CBLAS_TEST_COMPLEX));
-        AP=(CBLAS_TEST_COMPLEX*)malloc((((LDA+1)*LDA)/2)*
+        A=(CBLAS_TEST_COMPLEX*)malloc((size_t)LDA*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        AP=(CBLAS_TEST_COMPLEX*)malloc(((((size_t)LDA+1)*LDA)/2)*
 	 	sizeof(CBLAS_TEST_COMPLEX));
         if (uplo == CblasUpper) {
            for( j=0, k=0; j<*n; j++ )
@@ -491,8 +491,8 @@ void F77_ctpsv(int *order, char *uplow, char *transp, char *diagn,
         cblas_ctpsv( CblasRowMajor, UNDEFINED, trans, diag, *n, ap, x, *incx );
      else {
         LDA = *n;
-        A=(CBLAS_TEST_COMPLEX*)malloc(LDA*LDA*sizeof(CBLAS_TEST_COMPLEX));
-        AP=(CBLAS_TEST_COMPLEX*)malloc((((LDA+1)*LDA)/2)*
+        A=(CBLAS_TEST_COMPLEX*)malloc((size_t)LDA*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        AP=(CBLAS_TEST_COMPLEX*)malloc(((((size_t)LDA+1)*LDA)/2)*
 		sizeof(CBLAS_TEST_COMPLEX));
      	if (uplo == CblasUpper) {
            for( j=0, k=0; j<*n; j++ )
@@ -544,7 +544,7 @@ void F77_ctrmv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA=*n+1;
-     A=(CBLAS_TEST_COMPLEX*)malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+     A=(CBLAS_TEST_COMPLEX*)malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_COMPLEX));
      for( i=0; i<*n; i++ )
        for( j=0; j<*n; j++ ) {
 	  A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -573,7 +573,7 @@ void F77_ctrsv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A =(CBLAS_TEST_COMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+     A =(CBLAS_TEST_COMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_COMPLEX ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ ) {
            A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -601,8 +601,8 @@ void F77_chpr(int *order, char *uplow, int *n, float *alpha,
         cblas_chpr(CblasRowMajor, UNDEFINED, *n, *alpha, x, *incx, ap );
      else {
         LDA = *n;
-        A = (CBLAS_TEST_COMPLEX* )malloc(LDA*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
-        AP = ( CBLAS_TEST_COMPLEX* )malloc( (((LDA+1)*LDA)/2)*
+        A = (CBLAS_TEST_COMPLEX* )malloc((size_t)LDA*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+        AP = ( CBLAS_TEST_COMPLEX* )malloc( ((((size_t)LDA+1)*LDA)/2)*
 		sizeof( CBLAS_TEST_COMPLEX ));
         if (uplo == CblasUpper) {
            for( j=0, k=0; j<*n; j++ )
@@ -678,8 +678,8 @@ void F77_chpr2(int *order, char *uplow, int *n, CBLAS_TEST_COMPLEX *alpha,
 		     *incy, ap );
      else {
         LDA = *n;
-        A=(CBLAS_TEST_COMPLEX*)malloc( LDA*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
-        AP=(CBLAS_TEST_COMPLEX*)malloc( (((LDA+1)*LDA)/2)*
+        A=(CBLAS_TEST_COMPLEX*)malloc( (size_t)LDA*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+        AP=(CBLAS_TEST_COMPLEX*)malloc( ((((size_t)LDA+1)*LDA)/2)*
 	sizeof( CBLAS_TEST_COMPLEX ));
         if (uplo == CblasUpper) {
            for( j=0, k=0; j<*n; j++ )
@@ -750,7 +750,7 @@ void F77_cher(int *order, char *uplow, int *n, float *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A=(CBLAS_TEST_COMPLEX*)malloc((*n)*LDA*sizeof( CBLAS_TEST_COMPLEX ));
+     A=(CBLAS_TEST_COMPLEX*)malloc((*n)*(size_t)LDA*sizeof( CBLAS_TEST_COMPLEX ));
 
      for( i=0; i<*n; i++ )
        for( j=0; j<*n; j++ ) {
@@ -784,7 +784,7 @@ void F77_cher2(int *order, char *uplow, int *n, CBLAS_TEST_COMPLEX *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A= ( CBLAS_TEST_COMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+     A= ( CBLAS_TEST_COMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_COMPLEX ) );
 
      for( i=0; i<*n; i++ )
        for( j=0; j<*n; j++ ) {

ctest/c_dblas2.c

@@ -19,7 +19,7 @@ void F77_dgemv(int *order, char *transp, int *m, int *n, double *alpha,
   get_transpose_type(transp, &trans);
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( double* )malloc( (*m)*LDA*sizeof( double ) );
+     A   = ( double* )malloc( (*m)*(size_t)LDA*sizeof( double ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ )
            A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -43,7 +43,7 @@ void F77_dger(int *order, int *m, int *n, double *alpha, double *x, int *incx,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( double* )malloc( (*m)*LDA*sizeof( double ) );
+     A   = ( double* )malloc( (*m)*(size_t)LDA*sizeof( double ) );
 
      for( i=0; i<*m; i++ ) {
        for( j=0; j<*n; j++ )
@@ -74,7 +74,7 @@ void F77_dtrmv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( double* )malloc( (*n)*LDA*sizeof( double ) );
+     A   = ( double* )malloc( (*n)*(size_t)LDA*sizeof( double ) );
      for( i=0; i<*n; i++ )
        for( j=0; j<*n; j++ )
          A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -102,7 +102,7 @@ void F77_dtrsv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( double* )malloc( (*n)*LDA*sizeof( double ) );
+     A   = ( double* )malloc( (*n)*(size_t)LDA*sizeof( double ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -123,7 +123,7 @@ void F77_dsymv(int *order, char *uplow, int *n, double *alpha, double *a,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( double* )malloc( (*n)*LDA*sizeof( double ) );
+     A   = ( double* )malloc( (*n)*(size_t)LDA*sizeof( double ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -146,7 +146,7 @@ void F77_dsyr(int *order, char *uplow, int *n, double *alpha, double *x,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( double* )malloc( (*n)*LDA*sizeof( double ) );
+     A   = ( double* )malloc( (*n)*(size_t)LDA*sizeof( double ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -170,7 +170,7 @@ void F77_dsyr2(int *order, char *uplow, int *n, double *alpha, double *x,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( double* )malloc( (*n)*LDA*sizeof( double ) );
+     A   = ( double* )malloc( (*n)*(size_t)LDA*sizeof( double ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -196,7 +196,7 @@ void F77_dgbmv(int *order, char *transp, int *m, int *n, int *kl, int *ku,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *ku+*kl+2;
-     A   = ( double* )malloc( (*n+*kl)*LDA*sizeof( double ) );
+     A   = ( double* )malloc( (*n+*kl)*(size_t)LDA*sizeof( double ) );
      for( i=0; i<*ku; i++ ){
         irow=*ku+*kl-i;
         jcol=(*ku)-i;
@@ -236,7 +236,7 @@ void F77_dtbmv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *k+1;
-     A = ( double* )malloc( (*n+*k)*LDA*sizeof( double ) );
+     A = ( double* )malloc( (*n+*k)*(size_t)LDA*sizeof( double ) );
      if (uplo == CblasUpper) {
         for( i=0; i<*k; i++ ){
            irow=*k-i;
@@ -282,7 +282,7 @@ void F77_dtbsv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *k+1;
-     A = ( double* )malloc( (*n+*k)*LDA*sizeof( double ) );
+     A = ( double* )malloc( (*n+*k)*(size_t)LDA*sizeof( double ) );
      if (uplo == CblasUpper) {
         for( i=0; i<*k; i++ ){
         irow=*k-i;
@@ -325,7 +325,7 @@ void F77_dsbmv(int *order, char *uplow, int *n, int *k, double *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *k+1;
-     A   = ( double* )malloc( (*n+*k)*LDA*sizeof( double ) );
+     A   = ( double* )malloc( (*n+*k)*(size_t)LDA*sizeof( double ) );
      if (uplo == CblasUpper) {
         for( i=0; i<*k; i++ ){
            irow=*k-i;
@@ -369,8 +369,8 @@ void F77_dspmv(int *order, char *uplow, int *n, double *alpha, double *ap,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n;
-     A   = ( double* )malloc( LDA*LDA*sizeof( double ) );
-     AP  = ( double* )malloc( (((LDA+1)*LDA)/2)*sizeof( double ) );
+     A   = ( double* )malloc( (size_t)LDA*LDA*sizeof( double ) );
+     AP  = ( double* )malloc( ((((size_t)LDA+1)*LDA)/2)*sizeof( double ) );
      if (uplo == CblasUpper) {
         for( j=0, k=0; j<*n; j++ )
            for( i=0; i<j+1; i++, k++ )
@@ -411,8 +411,8 @@ void F77_dtpmv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n;
-     A   = ( double* )malloc( LDA*LDA*sizeof( double ) );
-     AP  = ( double* )malloc( (((LDA+1)*LDA)/2)*sizeof( double ) );
+     A   = ( double* )malloc( (size_t)LDA*LDA*sizeof( double ) );
+     AP  = ( double* )malloc( ((((size_t)LDA+1)*LDA)/2)*sizeof( double ) );
      if (uplo == CblasUpper) {
         for( j=0, k=0; j<*n; j++ )
            for( i=0; i<j+1; i++, k++ )
@@ -451,8 +451,8 @@ void F77_dtpsv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n;
-     A   = ( double* )malloc( LDA*LDA*sizeof( double ) );
-     AP  = ( double* )malloc( (((LDA+1)*LDA)/2)*sizeof( double ) );
+     A   = ( double* )malloc( (size_t)LDA*LDA*sizeof( double ) );
+     AP  = ( double* )malloc( ((((size_t)LDA+1)*LDA)/2)*sizeof( double ) );
      if (uplo == CblasUpper) {
         for( j=0, k=0; j<*n; j++ )
            for( i=0; i<j+1; i++, k++ )
@@ -488,8 +488,8 @@ void F77_dspr(int *order, char *uplow, int *n, double *alpha, double *x,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n;
-     A   = ( double* )malloc( LDA*LDA*sizeof( double ) );
-     AP  = ( double* )malloc( (((LDA+1)*LDA)/2)*sizeof( double ) );
+     A   = ( double* )malloc( (size_t)LDA*LDA*sizeof( double ) );
+     AP  = ( double* )malloc( ((((size_t)LDA+1)*LDA)/2)*sizeof( double ) );
      if (uplo == CblasUpper) {
         for( j=0, k=0; j<*n; j++ )
            for( i=0; i<j+1; i++, k++ )
@@ -540,8 +540,8 @@ void F77_dspr2(int *order, char *uplow, int *n, double *alpha, double *x,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n;
-     A   = ( double* )malloc( LDA*LDA*sizeof( double ) );
-     AP  = ( double* )malloc( (((LDA+1)*LDA)/2)*sizeof( double ) );
+     A   = ( double* )malloc( (size_t)LDA*LDA*sizeof( double ) );
+     AP  = ( double* )malloc( ((((size_t)LDA+1)*LDA)/2)*sizeof( double ) );
      if (uplo == CblasUpper) {
         for( j=0, k=0; j<*n; j++ )
            for( i=0; i<j+1; i++, k++ )

ctest/c_dblas3.c

@@ -26,34 +26,34 @@ void F77_dgemm(int *order, char *transpa, char *transpb, int *m, int *n,
   if (*order == TEST_ROW_MJR) {
      if (transa == CblasNoTrans) {
         LDA = *k+1;
-        A = (double *)malloc( (*m)*LDA*sizeof( double ) );
+        A = (double *)malloc( (*m)*(size_t)LDA*sizeof( double ) );
         for( i=0; i<*m; i++ )
            for( j=0; j<*k; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else {
         LDA = *m+1;
-        A   = ( double* )malloc( LDA*(*k)*sizeof( double ) );
+        A   = ( double* )malloc( (size_t)LDA*(*k)*sizeof( double ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*m; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      if (transb == CblasNoTrans) {
         LDB = *n+1;
-        B   = ( double* )malloc( (*k)*LDB*sizeof( double ) );
+        B   = ( double* )malloc( (*k)*(size_t)LDB*sizeof( double ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ )
               B[i*LDB+j]=b[j*(*ldb)+i];
      }
      else {
         LDB = *k+1;
-        B   = ( double* )malloc( LDB*(*n)*sizeof( double ) );
+        B   = ( double* )malloc( (size_t)LDB*(*n)*sizeof( double ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ )
               B[i*LDB+j]=b[j*(*ldb)+i];
      }
      LDC = *n+1;
-     C   = ( double* )malloc( (*m)*LDC*sizeof( double ) );
+     C   = ( double* )malloc( (*m)*(size_t)LDC*sizeof( double ) );
      for( j=0; j<*n; j++ )
         for( i=0; i<*m; i++ )
            C[i*LDC+j]=c[j*(*ldc)+i];
@@ -89,25 +89,25 @@ void F77_dsymm(int *order, char *rtlf, char *uplow, int *m, int *n,
   if (*order == TEST_ROW_MJR) {
      if (side == CblasLeft) {
         LDA = *m+1;
-        A   = ( double* )malloc( (*m)*LDA*sizeof( double ) );
+        A   = ( double* )malloc( (*m)*(size_t)LDA*sizeof( double ) );
         for( i=0; i<*m; i++ )
            for( j=0; j<*m; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else{
         LDA = *n+1;
-        A   = ( double* )malloc( (*n)*LDA*sizeof( double ) );
+        A   = ( double* )malloc( (*n)*(size_t)LDA*sizeof( double ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*n; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      LDB = *n+1;
-     B   = ( double* )malloc( (*m)*LDB*sizeof( double ) );
+     B   = ( double* )malloc( (*m)*(size_t)LDB*sizeof( double ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ )
            B[i*LDB+j]=b[j*(*ldb)+i];
      LDC = *n+1;
-     C   = ( double* )malloc( (*m)*LDC*sizeof( double ) );
+     C   = ( double* )malloc( (*m)*(size_t)LDC*sizeof( double ) );
      for( j=0; j<*n; j++ )
         for( i=0; i<*m; i++ )
            C[i*LDC+j]=c[j*(*ldc)+i];
@@ -143,20 +143,20 @@ void F77_dsyrk(int *order, char *uplow, char *transp, int *n, int *k,
   if (*order == TEST_ROW_MJR) {
      if (trans == CblasNoTrans) {
         LDA = *k+1;
-        A   = ( double* )malloc( (*n)*LDA*sizeof( double ) );
+        A   = ( double* )malloc( (*n)*(size_t)LDA*sizeof( double ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else{
         LDA = *n+1;
-        A   = ( double* )malloc( (*k)*LDA*sizeof( double ) );
+        A   = ( double* )malloc( (*k)*(size_t)LDA*sizeof( double ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      LDC = *n+1;
-     C   = ( double* )malloc( (*n)*LDC*sizeof( double ) );
+     C   = ( double* )malloc( (*n)*(size_t)LDC*sizeof( double ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            C[i*LDC+j]=c[j*(*ldc)+i];
@@ -191,8 +191,8 @@ void F77_dsyr2k(int *order, char *uplow, char *transp, int *n, int *k,
      if (trans == CblasNoTrans) {
         LDA = *k+1;
         LDB = *k+1;
-        A   = ( double* )malloc( (*n)*LDA*sizeof( double ) );
-        B   = ( double* )malloc( (*n)*LDB*sizeof( double ) );
+        A   = ( double* )malloc( (*n)*(size_t)LDA*sizeof( double ) );
+        B   = ( double* )malloc( (*n)*(size_t)LDB*sizeof( double ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ ) {
               A[i*LDA+j]=a[j*(*lda)+i];
@@ -202,8 +202,8 @@ void F77_dsyr2k(int *order, char *uplow, char *transp, int *n, int *k,
      else {
         LDA = *n+1;
         LDB = *n+1;
-        A   = ( double* )malloc( LDA*(*k)*sizeof( double ) );
-        B   = ( double* )malloc( LDB*(*k)*sizeof( double ) );
+        A   = ( double* )malloc( (size_t)LDA*(*k)*sizeof( double ) );
+        B   = ( double* )malloc( (size_t)LDB*(*k)*sizeof( double ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ ){
               A[i*LDA+j]=a[j*(*lda)+i];
@@ -211,7 +211,7 @@ void F77_dsyr2k(int *order, char *uplow, char *transp, int *n, int *k,
            }
      }
      LDC = *n+1;
-     C   = ( double* )malloc( (*n)*LDC*sizeof( double ) );
+     C   = ( double* )malloc( (*n)*(size_t)LDC*sizeof( double ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            C[i*LDC+j]=c[j*(*ldc)+i];
@@ -249,20 +249,20 @@ void F77_dtrmm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
   if (*order == TEST_ROW_MJR) {
      if (side == CblasLeft) {
         LDA = *m+1;
-        A   = ( double* )malloc( (*m)*LDA*sizeof( double ) );
+        A   = ( double* )malloc( (*m)*(size_t)LDA*sizeof( double ) );
         for( i=0; i<*m; i++ )
            for( j=0; j<*m; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else{
         LDA = *n+1;
-        A   = ( double* )malloc( (*n)*LDA*sizeof( double ) );
+        A   = ( double* )malloc( (*n)*(size_t)LDA*sizeof( double ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*n; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      LDB = *n+1;
-     B   = ( double* )malloc( (*m)*LDB*sizeof( double ) );
+     B   = ( double* )malloc( (*m)*(size_t)LDB*sizeof( double ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ )
            B[i*LDB+j]=b[j*(*ldb)+i];
@@ -300,20 +300,20 @@ void F77_dtrsm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
   if (*order == TEST_ROW_MJR) {
      if (side == CblasLeft) {
         LDA = *m+1;
-        A   = ( double* )malloc( (*m)*LDA*sizeof( double ) );
+        A   = ( double* )malloc( (*m)*(size_t)LDA*sizeof( double ) );
         for( i=0; i<*m; i++ )
            for( j=0; j<*m; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else{
         LDA = *n+1;
-        A   = ( double* )malloc( (*n)*LDA*sizeof( double ) );
+        A   = ( double* )malloc( (*n)*(size_t)LDA*sizeof( double ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*n; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      LDB = *n+1;
-     B   = ( double* )malloc( (*m)*LDB*sizeof( double ) );
+     B   = ( double* )malloc( (*m)*(size_t)LDB*sizeof( double ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ )
            B[i*LDB+j]=b[j*(*ldb)+i];

ctest/c_sblas2.c

@@ -19,7 +19,7 @@ void F77_sgemv(int *order, char *transp, int *m, int *n, float *alpha,
   get_transpose_type(transp, &trans);
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( float* )malloc( (*m)*LDA*sizeof( float ) );
+     A   = ( float* )malloc( (*m)*(size_t)LDA*sizeof( float ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ )
            A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -43,7 +43,7 @@ void F77_sger(int *order, int *m, int *n, float *alpha, float *x, int *incx,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( float* )malloc( (*m)*LDA*sizeof( float ) );
+     A   = ( float* )malloc( (*m)*(size_t)LDA*sizeof( float ) );
 
      for( i=0; i<*m; i++ ) {
        for( j=0; j<*n; j++ )
@@ -74,7 +74,7 @@ void F77_strmv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( float* )malloc( (*n)*LDA*sizeof( float ) );
+     A   = ( float* )malloc( (*n)*(size_t)LDA*sizeof( float ) );
      for( i=0; i<*n; i++ )
        for( j=0; j<*n; j++ )
          A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -102,7 +102,7 @@ void F77_strsv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( float* )malloc( (*n)*LDA*sizeof( float ) );
+     A   = ( float* )malloc( (*n)*(size_t)LDA*sizeof( float ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -123,7 +123,7 @@ void F77_ssymv(int *order, char *uplow, int *n, float *alpha, float *a,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( float* )malloc( (*n)*LDA*sizeof( float ) );
+     A   = ( float* )malloc( (*n)*(size_t)LDA*sizeof( float ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -146,7 +146,7 @@ void F77_ssyr(int *order, char *uplow, int *n, float *alpha, float *x,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( float* )malloc( (*n)*LDA*sizeof( float ) );
+     A   = ( float* )malloc( (*n)*(size_t)LDA*sizeof( float ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -170,7 +170,7 @@ void F77_ssyr2(int *order, char *uplow, int *n, float *alpha, float *x,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A   = ( float* )malloc( (*n)*LDA*sizeof( float ) );
+     A   = ( float* )malloc( (*n)*(size_t)LDA*sizeof( float ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            A[ LDA*i+j ]=a[ (*lda)*j+i ];
@@ -196,7 +196,7 @@ void F77_sgbmv(int *order, char *transp, int *m, int *n, int *kl, int *ku,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *ku+*kl+2;
-     A   = ( float* )malloc( (*n+*kl)*LDA*sizeof( float ) );
+     A   = ( float* )malloc( (*n+*kl)*(size_t)LDA*sizeof( float ) );
      for( i=0; i<*ku; i++ ){
         irow=*ku+*kl-i;
         jcol=(*ku)-i;
@@ -236,7 +236,7 @@ void F77_stbmv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *k+1;
-     A = ( float* )malloc( (*n+*k)*LDA*sizeof( float ) );
+     A = ( float* )malloc( (*n+*k)*(size_t)LDA*sizeof( float ) );
      if (uplo == CblasUpper) {
         for( i=0; i<*k; i++ ){
            irow=*k-i;
@@ -282,7 +282,7 @@ void F77_stbsv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *k+1;
-     A = ( float* )malloc( (*n+*k)*LDA*sizeof( float ) );
+     A = ( float* )malloc( (*n+*k)*(size_t)LDA*sizeof( float ) );
      if (uplo == CblasUpper) {
         for( i=0; i<*k; i++ ){
         irow=*k-i;
@@ -325,7 +325,7 @@ void F77_ssbmv(int *order, char *uplow, int *n, int *k, float *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *k+1;
-     A   = ( float* )malloc( (*n+*k)*LDA*sizeof( float ) );
+     A   = ( float* )malloc( (*n+*k)*(size_t)LDA*sizeof( float ) );
      if (uplo == CblasUpper) {
         for( i=0; i<*k; i++ ){
            irow=*k-i;
@@ -369,8 +369,8 @@ void F77_sspmv(int *order, char *uplow, int *n, float *alpha, float *ap,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n;
-     A   = ( float* )malloc( LDA*LDA*sizeof( float ) );
-     AP  = ( float* )malloc( (((LDA+1)*LDA)/2)*sizeof( float ) );
+     A   = ( float* )malloc( (size_t)LDA*LDA*sizeof( float ) );
+     AP  = ( float* )malloc( ((((size_t)LDA+1)*LDA)/2)*sizeof( float ) );
      if (uplo == CblasUpper) {
         for( j=0, k=0; j<*n; j++ )
            for( i=0; i<j+1; i++, k++ )
@@ -410,8 +410,8 @@ void F77_stpmv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n;
-     A   = ( float* )malloc( LDA*LDA*sizeof( float ) );
-     AP  = ( float* )malloc( (((LDA+1)*LDA)/2)*sizeof( float ) );
+     A   = ( float* )malloc( (size_t)LDA*LDA*sizeof( float ) );
+     AP  = ( float* )malloc( ((((size_t)LDA+1)*LDA)/2)*sizeof( float ) );
      if (uplo == CblasUpper) {
         for( j=0, k=0; j<*n; j++ )
            for( i=0; i<j+1; i++, k++ )
@@ -449,8 +449,8 @@ void F77_stpsv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n;
-     A   = ( float* )malloc( LDA*LDA*sizeof( float ) );
-     AP  = ( float* )malloc( (((LDA+1)*LDA)/2)*sizeof( float ) );
+     A   = ( float* )malloc( (size_t)LDA*LDA*sizeof( float ) );
+     AP  = ( float* )malloc( ((((size_t)LDA+1)*LDA)/2)*sizeof( float ) );
      if (uplo == CblasUpper) {
         for( j=0, k=0; j<*n; j++ )
            for( i=0; i<j+1; i++, k++ )
@@ -485,8 +485,8 @@ void F77_sspr(int *order, char *uplow, int *n, float *alpha, float *x,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n;
-     A   = ( float* )malloc( LDA*LDA*sizeof( float ) );
-     AP  = ( float* )malloc( (((LDA+1)*LDA)/2)*sizeof( float ) );
+     A   = ( float* )malloc( (size_t)LDA*LDA*sizeof( float ) );
+     AP  = ( float* )malloc( ((((size_t)LDA+1)*LDA)/2)*sizeof( float ) );
      if (uplo == CblasUpper) {
         for( j=0, k=0; j<*n; j++ )
            for( i=0; i<j+1; i++, k++ )
@@ -536,8 +536,8 @@ void F77_sspr2(int *order, char *uplow, int *n, float *alpha, float *x,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n;
-     A   = ( float* )malloc( LDA*LDA*sizeof( float ) );
-     AP  = ( float* )malloc( (((LDA+1)*LDA)/2)*sizeof( float ) );
+     A   = ( float* )malloc( (size_t)LDA*LDA*sizeof( float ) );
+     AP  = ( float* )malloc( ((((size_t)LDA+1)*LDA)/2)*sizeof( float ) );
      if (uplo == CblasUpper) {
         for( j=0, k=0; j<*n; j++ )
            for( i=0; i<j+1; i++, k++ )

ctest/c_sblas3.c

@@ -23,34 +23,34 @@ void F77_sgemm(int *order, char *transpa, char *transpb, int *m, int *n,
   if (*order == TEST_ROW_MJR) {
      if (transa == CblasNoTrans) {
         LDA = *k+1;
-        A = (float *)malloc( (*m)*LDA*sizeof( float ) );
+        A = (float *)malloc( (*m)*(size_t)LDA*sizeof( float ) );
         for( i=0; i<*m; i++ )
            for( j=0; j<*k; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else {
         LDA = *m+1;
-        A   = ( float* )malloc( LDA*(*k)*sizeof( float ) );
+        A   = ( float* )malloc( (size_t)LDA*(*k)*sizeof( float ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*m; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      if (transb == CblasNoTrans) {
         LDB = *n+1;
-        B   = ( float* )malloc( (*k)*LDB*sizeof( float ) );
+        B   = ( float* )malloc( (*k)*(size_t)LDB*sizeof( float ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ )
               B[i*LDB+j]=b[j*(*ldb)+i];
      }
      else {
         LDB = *k+1;
-        B   = ( float* )malloc( LDB*(*n)*sizeof( float ) );
+        B   = ( float* )malloc( (size_t)LDB*(*n)*sizeof( float ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ )
               B[i*LDB+j]=b[j*(*ldb)+i];
      }
      LDC = *n+1;
-     C   = ( float* )malloc( (*m)*LDC*sizeof( float ) );
+     C   = ( float* )malloc( (*m)*(size_t)LDC*sizeof( float ) );
      for( j=0; j<*n; j++ )
         for( i=0; i<*m; i++ )
            C[i*LDC+j]=c[j*(*ldc)+i];
@@ -85,25 +85,25 @@ void F77_ssymm(int *order, char *rtlf, char *uplow, int *m, int *n,
   if (*order == TEST_ROW_MJR) {
      if (side == CblasLeft) {
         LDA = *m+1;
-        A   = ( float* )malloc( (*m)*LDA*sizeof( float ) );
+        A   = ( float* )malloc( (*m)*(size_t)LDA*sizeof( float ) );
         for( i=0; i<*m; i++ )
            for( j=0; j<*m; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else{
         LDA = *n+1;
-        A   = ( float* )malloc( (*n)*LDA*sizeof( float ) );
+        A   = ( float* )malloc( (*n)*(size_t)LDA*sizeof( float ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*n; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      LDB = *n+1;
-     B   = ( float* )malloc( (*m)*LDB*sizeof( float ) );
+     B   = ( float* )malloc( (*m)*(size_t)LDB*sizeof( float ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ )
            B[i*LDB+j]=b[j*(*ldb)+i];
      LDC = *n+1;
-     C   = ( float* )malloc( (*m)*LDC*sizeof( float ) );
+     C   = ( float* )malloc( (*m)*(size_t)LDC*sizeof( float ) );
      for( j=0; j<*n; j++ )
         for( i=0; i<*m; i++ )
            C[i*LDC+j]=c[j*(*ldc)+i];
@@ -139,20 +139,20 @@ void F77_ssyrk(int *order, char *uplow, char *transp, int *n, int *k,
   if (*order == TEST_ROW_MJR) {
      if (trans == CblasNoTrans) {
         LDA = *k+1;
-        A   = ( float* )malloc( (*n)*LDA*sizeof( float ) );
+        A   = ( float* )malloc( (*n)*(size_t)LDA*sizeof( float ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else{
         LDA = *n+1;
-        A   = ( float* )malloc( (*k)*LDA*sizeof( float ) );
+        A   = ( float* )malloc( (*k)*(size_t)LDA*sizeof( float ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      LDC = *n+1;
-     C   = ( float* )malloc( (*n)*LDC*sizeof( float ) );
+     C   = ( float* )malloc( (*n)*(size_t)LDC*sizeof( float ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            C[i*LDC+j]=c[j*(*ldc)+i];
@@ -187,8 +187,8 @@ void F77_ssyr2k(int *order, char *uplow, char *transp, int *n, int *k,
      if (trans == CblasNoTrans) {
         LDA = *k+1;
         LDB = *k+1;
-        A   = ( float* )malloc( (*n)*LDA*sizeof( float ) );
-        B   = ( float* )malloc( (*n)*LDB*sizeof( float ) );
+        A   = ( float* )malloc( (*n)*(size_t)LDA*sizeof( float ) );
+        B   = ( float* )malloc( (*n)*(size_t)LDB*sizeof( float ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ ) {
               A[i*LDA+j]=a[j*(*lda)+i];
@@ -198,8 +198,8 @@ void F77_ssyr2k(int *order, char *uplow, char *transp, int *n, int *k,
      else {
         LDA = *n+1;
         LDB = *n+1;
-        A   = ( float* )malloc( LDA*(*k)*sizeof( float ) );
-        B   = ( float* )malloc( LDB*(*k)*sizeof( float ) );
+        A   = ( float* )malloc( (size_t)LDA*(*k)*sizeof( float ) );
+        B   = ( float* )malloc( (size_t)LDB*(*k)*sizeof( float ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ ){
               A[i*LDA+j]=a[j*(*lda)+i];
@@ -207,7 +207,7 @@ void F77_ssyr2k(int *order, char *uplow, char *transp, int *n, int *k,
            }
      }
      LDC = *n+1;
-     C   = ( float* )malloc( (*n)*LDC*sizeof( float ) );
+     C   = ( float* )malloc( (*n)*(size_t)LDC*sizeof( float ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ )
            C[i*LDC+j]=c[j*(*ldc)+i];
@@ -245,20 +245,20 @@ void F77_strmm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
   if (*order == TEST_ROW_MJR) {
      if (side == CblasLeft) {
         LDA = *m+1;
-        A   = ( float* )malloc( (*m)*LDA*sizeof( float ) );
+        A   = ( float* )malloc( (*m)*(size_t)LDA*sizeof( float ) );
         for( i=0; i<*m; i++ )
            for( j=0; j<*m; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else{
         LDA = *n+1;
-        A   = ( float* )malloc( (*n)*LDA*sizeof( float ) );
+        A   = ( float* )malloc( (*n)*(size_t)LDA*sizeof( float ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*n; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      LDB = *n+1;
-     B   = ( float* )malloc( (*m)*LDB*sizeof( float ) );
+     B   = ( float* )malloc( (*m)*(size_t)LDB*sizeof( float ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ )
            B[i*LDB+j]=b[j*(*ldb)+i];
@@ -296,20 +296,20 @@ void F77_strsm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
   if (*order == TEST_ROW_MJR) {
      if (side == CblasLeft) {
         LDA = *m+1;
-        A   = ( float* )malloc( (*m)*LDA*sizeof( float ) );
+        A   = ( float* )malloc( (*m)*(size_t)LDA*sizeof( float ) );
         for( i=0; i<*m; i++ )
            for( j=0; j<*m; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else{
         LDA = *n+1;
-        A   = ( float* )malloc( (*n)*LDA*sizeof( float ) );
+        A   = ( float* )malloc( (*n)*(size_t)LDA*sizeof( float ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*n; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      LDB = *n+1;
-     B   = ( float* )malloc( (*m)*LDB*sizeof( float ) );
+     B   = ( float* )malloc( (*m)*(size_t)LDB*sizeof( float ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ )
            B[i*LDB+j]=b[j*(*ldb)+i];

ctest/c_zblas2.c

@@ -20,7 +20,7 @@ void F77_zgemv(int *order, char *transp, int *m, int *n,
   get_transpose_type(transp, &trans);
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A  = (CBLAS_TEST_ZOMPLEX *)malloc( (*m)*LDA*sizeof( CBLAS_TEST_ZOMPLEX) );
+     A  = (CBLAS_TEST_ZOMPLEX *)malloc( (*m)*(size_t)LDA*sizeof( CBLAS_TEST_ZOMPLEX) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ ){
            A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -50,7 +50,7 @@ void F77_zgbmv(int *order, char *transp, int *m, int *n, int *kl, int *ku,
   get_transpose_type(transp, &trans);
   if (*order == TEST_ROW_MJR) {
      LDA = *ku+*kl+2;
-     A=( CBLAS_TEST_ZOMPLEX* )malloc((*n+*kl)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+     A=( CBLAS_TEST_ZOMPLEX* )malloc((*n+*kl)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
      for( i=0; i<*ku; i++ ){
         irow=*ku+*kl-i;
         jcol=(*ku)-i;
@@ -94,7 +94,7 @@ void F77_zgeru(int *order, int *m, int *n, CBLAS_TEST_ZOMPLEX *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A=(CBLAS_TEST_ZOMPLEX*)malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+     A=(CBLAS_TEST_ZOMPLEX*)malloc((*m)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ ){
            A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -122,7 +122,7 @@ void F77_zgerc(int *order, int *m, int *n, CBLAS_TEST_ZOMPLEX *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     A=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ ){
            A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -154,7 +154,7 @@ void F77_zhemv(int *order, char *uplow, int *n, CBLAS_TEST_ZOMPLEX *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A = (CBLAS_TEST_ZOMPLEX *)malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+     A = (CBLAS_TEST_ZOMPLEX *)malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ ){
            A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -190,7 +190,7 @@ int i,irow,j,jcol,LDA;
 		 *incx, beta, y, *incy );
      else {
         LDA = *k+2;
-        A =(CBLAS_TEST_ZOMPLEX*)malloc((*n+*k)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        A =(CBLAS_TEST_ZOMPLEX*)malloc((*n+*k)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
         if (uplo == CblasUpper) {
            for( i=0; i<*k; i++ ){
               irow=*k-i;
@@ -251,8 +251,8 @@ void F77_zhpmv(int *order, char *uplow, int *n, CBLAS_TEST_ZOMPLEX *alpha,
 	         beta, y, *incy);
      else {
         LDA = *n;
-        A = (CBLAS_TEST_ZOMPLEX* )malloc(LDA*LDA*sizeof(CBLAS_TEST_ZOMPLEX ));
-        AP = (CBLAS_TEST_ZOMPLEX* )malloc( (((LDA+1)*LDA)/2)*
+        A = (CBLAS_TEST_ZOMPLEX* )malloc((size_t)LDA*LDA*sizeof(CBLAS_TEST_ZOMPLEX ));
+        AP = (CBLAS_TEST_ZOMPLEX* )malloc( ((((size_t)LDA+1)*LDA)/2)*
 	        sizeof( CBLAS_TEST_ZOMPLEX ));
         if (uplo == CblasUpper) {
            for( j=0, k=0; j<*n; j++ )
@@ -311,7 +311,7 @@ void F77_ztbmv(int *order, char *uplow, char *transp, char *diagn,
 	x, *incx);
      else {
         LDA = *k+2;
-        A=(CBLAS_TEST_ZOMPLEX *)malloc((*n+*k)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        A=(CBLAS_TEST_ZOMPLEX *)malloc((*n+*k)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
         if (uplo == CblasUpper) {
            for( i=0; i<*k; i++ ){
               irow=*k-i;
@@ -375,7 +375,7 @@ void F77_ztbsv(int *order, char *uplow, char *transp, char *diagn,
 	         *incx);
      else {
         LDA = *k+2;
-        A=(CBLAS_TEST_ZOMPLEX*)malloc((*n+*k)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ));
+        A=(CBLAS_TEST_ZOMPLEX*)malloc((*n+*k)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ));
         if (uplo == CblasUpper) {
            for( i=0; i<*k; i++ ){
               irow=*k-i;
@@ -436,8 +436,8 @@ void F77_ztpmv(int *order, char *uplow, char *transp, char *diagn,
         cblas_ztpmv( CblasRowMajor, UNDEFINED, trans, diag, *n, ap, x, *incx );
      else {
         LDA = *n;
-        A=(CBLAS_TEST_ZOMPLEX*)malloc(LDA*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
-        AP=(CBLAS_TEST_ZOMPLEX*)malloc((((LDA+1)*LDA)/2)*
+        A=(CBLAS_TEST_ZOMPLEX*)malloc((size_t)LDA*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        AP=(CBLAS_TEST_ZOMPLEX*)malloc(((((size_t)LDA+1)*LDA)/2)*
 	 	sizeof(CBLAS_TEST_ZOMPLEX));
         if (uplo == CblasUpper) {
            for( j=0, k=0; j<*n; j++ )
@@ -491,8 +491,8 @@ void F77_ztpsv(int *order, char *uplow, char *transp, char *diagn,
         cblas_ztpsv( CblasRowMajor, UNDEFINED, trans, diag, *n, ap, x, *incx );
      else {
         LDA = *n;
-        A=(CBLAS_TEST_ZOMPLEX*)malloc(LDA*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
-        AP=(CBLAS_TEST_ZOMPLEX*)malloc((((LDA+1)*LDA)/2)*
+        A=(CBLAS_TEST_ZOMPLEX*)malloc((size_t)LDA*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        AP=(CBLAS_TEST_ZOMPLEX*)malloc(((((size_t)LDA+1)*LDA)/2)*
 		sizeof(CBLAS_TEST_ZOMPLEX));
      	if (uplo == CblasUpper) {
            for( j=0, k=0; j<*n; j++ )
@@ -544,7 +544,7 @@ void F77_ztrmv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA=*n+1;
-     A=(CBLAS_TEST_ZOMPLEX*)malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+     A=(CBLAS_TEST_ZOMPLEX*)malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
      for( i=0; i<*n; i++ )
        for( j=0; j<*n; j++ ) {
 	  A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -573,7 +573,7 @@ void F77_ztrsv(int *order, char *uplow, char *transp, char *diagn,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A =(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     A =(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ ) {
            A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
@@ -601,8 +601,8 @@ void F77_zhpr(int *order, char *uplow, int *n, double *alpha,
         cblas_zhpr(CblasRowMajor, UNDEFINED, *n, *alpha, x, *incx, ap );
      else {
         LDA = *n;
-        A = (CBLAS_TEST_ZOMPLEX* )malloc(LDA*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
-        AP = ( CBLAS_TEST_ZOMPLEX* )malloc( (((LDA+1)*LDA)/2)*
+        A = (CBLAS_TEST_ZOMPLEX* )malloc((size_t)LDA*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        AP = ( CBLAS_TEST_ZOMPLEX* )malloc( ((((size_t)LDA+1)*LDA)/2)*
 		sizeof( CBLAS_TEST_ZOMPLEX ));
         if (uplo == CblasUpper) {
            for( j=0, k=0; j<*n; j++ )
@@ -678,8 +678,8 @@ void F77_zhpr2(int *order, char *uplow, int *n, CBLAS_TEST_ZOMPLEX *alpha,
 		     *incy, ap );
      else {
         LDA = *n;
-        A=(CBLAS_TEST_ZOMPLEX*)malloc( LDA*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
-        AP=(CBLAS_TEST_ZOMPLEX*)malloc( (((LDA+1)*LDA)/2)*
+        A=(CBLAS_TEST_ZOMPLEX*)malloc( (size_t)LDA*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        AP=(CBLAS_TEST_ZOMPLEX*)malloc( ((((size_t)LDA+1)*LDA)/2)*
 	sizeof( CBLAS_TEST_ZOMPLEX ));
         if (uplo == CblasUpper) {
            for( j=0, k=0; j<*n; j++ )
@@ -750,7 +750,7 @@ void F77_zher(int *order, char *uplow, int *n, double *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A=(CBLAS_TEST_ZOMPLEX*)malloc((*n)*LDA*sizeof( CBLAS_TEST_ZOMPLEX ));
+     A=(CBLAS_TEST_ZOMPLEX*)malloc((*n)*(size_t)LDA*sizeof( CBLAS_TEST_ZOMPLEX ));
 
      for( i=0; i<*n; i++ )
        for( j=0; j<*n; j++ ) {
@@ -784,7 +784,7 @@ void F77_zher2(int *order, char *uplow, int *n, CBLAS_TEST_ZOMPLEX *alpha,
 
   if (*order == TEST_ROW_MJR) {
      LDA = *n+1;
-     A= ( CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     A= ( CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
 
      for( i=0; i<*n; i++ )
        for( j=0; j<*n; j++ ) {

ctest/c_zblas3.c

@@ -26,7 +26,7 @@ void F77_zgemm(int *order, char *transpa, char *transpb, int *m, int *n,
   if (*order == TEST_ROW_MJR) {
      if (transa == CblasNoTrans) {
         LDA = *k+1;
-        A=(CBLAS_TEST_ZOMPLEX*)malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        A=(CBLAS_TEST_ZOMPLEX*)malloc((*m)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*m; i++ )
            for( j=0; j<*k; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -35,7 +35,7 @@ void F77_zgemm(int *order, char *transpa, char *transpb, int *m, int *n,
      }
      else {
         LDA = *m+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc(LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((size_t)LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*k; i++ )
            for( j=0; j<*m; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -45,7 +45,7 @@ void F77_zgemm(int *order, char *transpa, char *transpb, int *m, int *n,
 
      if (transb == CblasNoTrans) {
         LDB = *n+1;
-        B=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*LDB*sizeof(CBLAS_TEST_ZOMPLEX) );
+        B=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*(size_t)LDB*sizeof(CBLAS_TEST_ZOMPLEX) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ ) {
               B[i*LDB+j].real=b[j*(*ldb)+i].real;
@@ -54,7 +54,7 @@ void F77_zgemm(int *order, char *transpa, char *transpb, int *m, int *n,
      }
      else {
         LDB = *k+1;
-        B=(CBLAS_TEST_ZOMPLEX* )malloc(LDB*(*n)*sizeof(CBLAS_TEST_ZOMPLEX));
+        B=(CBLAS_TEST_ZOMPLEX* )malloc((size_t)LDB*(*n)*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ ) {
               B[i*LDB+j].real=b[j*(*ldb)+i].real;
@@ -63,7 +63,7 @@ void F77_zgemm(int *order, char *transpa, char *transpb, int *m, int *n,
      }
 
      LDC = *n+1;
-     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_ZOMPLEX));
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*(size_t)LDC*sizeof(CBLAS_TEST_ZOMPLEX));
      for( j=0; j<*n; j++ )
         for( i=0; i<*m; i++ ) {
            C[i*LDC+j].real=c[j*(*ldc)+i].real;
@@ -103,7 +103,7 @@ void F77_zhemm(int *order, char *rtlf, char *uplow, int *m, int *n,
   if (*order == TEST_ROW_MJR) {
      if (side == CblasLeft) {
         LDA = *m+1;
-        A= (CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        A= (CBLAS_TEST_ZOMPLEX* )malloc((*m)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*m; i++ )
            for( j=0; j<*m; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -112,7 +112,7 @@ void F77_zhemm(int *order, char *rtlf, char *uplow, int *m, int *n,
      }
      else{
         LDA = *n+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*n; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -120,14 +120,14 @@ void F77_zhemm(int *order, char *rtlf, char *uplow, int *m, int *n,
            }
      }
      LDB = *n+1;
-     B=(CBLAS_TEST_ZOMPLEX* )malloc( (*m)*LDB*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     B=(CBLAS_TEST_ZOMPLEX* )malloc( (*m)*(size_t)LDB*sizeof(CBLAS_TEST_ZOMPLEX ) );
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ ) {
            B[i*LDB+j].real=b[j*(*ldb)+i].real;
            B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
         }
      LDC = *n+1;
-     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*(size_t)LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
      for( j=0; j<*n; j++ )
         for( i=0; i<*m; i++ ) {
            C[i*LDC+j].real=c[j*(*ldc)+i].real;
@@ -167,25 +167,25 @@ void F77_zsymm(int *order, char *rtlf, char *uplow, int *m, int *n,
   if (*order == TEST_ROW_MJR) {
      if (side == CblasLeft) {
         LDA = *m+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*m; i++ )
            for( j=0; j<*m; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      else{
         LDA = *n+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*n; j++ )
               A[i*LDA+j]=a[j*(*lda)+i];
      }
      LDB = *n+1;
-     B=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDB*sizeof(CBLAS_TEST_ZOMPLEX ));
+     B=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*(size_t)LDB*sizeof(CBLAS_TEST_ZOMPLEX ));
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ )
            B[i*LDB+j]=b[j*(*ldb)+i];
      LDC = *n+1;
-     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_ZOMPLEX));
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*(size_t)LDC*sizeof(CBLAS_TEST_ZOMPLEX));
      for( j=0; j<*n; j++ )
         for( i=0; i<*m; i++ )
            C[i*LDC+j]=c[j*(*ldc)+i];
@@ -221,7 +221,7 @@ void F77_zherk(int *order, char *uplow, char *transp, int *n, int *k,
   if (*order == TEST_ROW_MJR) {
      if (trans == CblasNoTrans) {
         LDA = *k+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -230,7 +230,7 @@ void F77_zherk(int *order, char *uplow, char *transp, int *n, int *k,
      }
      else{
         LDA = *n+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -238,7 +238,7 @@ void F77_zherk(int *order, char *uplow, char *transp, int *n, int *k,
            }
      }
      LDC = *n+1;
-     C=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ ) {
            C[i*LDC+j].real=c[j*(*ldc)+i].real;
@@ -277,7 +277,7 @@ void F77_zsyrk(int *order, char *uplow, char *transp, int *n, int *k,
   if (*order == TEST_ROW_MJR) {
      if (trans == CblasNoTrans) {
         LDA = *k+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -286,7 +286,7 @@ void F77_zsyrk(int *order, char *uplow, char *transp, int *n, int *k,
      }
      else{
         LDA = *n+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -294,7 +294,7 @@ void F77_zsyrk(int *order, char *uplow, char *transp, int *n, int *k,
            }
      }
      LDC = *n+1;
-     C=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ ) {
            C[i*LDC+j].real=c[j*(*ldc)+i].real;
@@ -333,8 +333,8 @@ void F77_zher2k(int *order, char *uplow, char *transp, int *n, int *k,
      if (trans == CblasNoTrans) {
         LDA = *k+1;
         LDB = *k+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ));
-        B=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDB*sizeof(CBLAS_TEST_ZOMPLEX ));
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ));
+        B=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDB*sizeof(CBLAS_TEST_ZOMPLEX ));
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -346,8 +346,8 @@ void F77_zher2k(int *order, char *uplow, char *transp, int *n, int *k,
      else {
         LDA = *n+1;
         LDB = *n+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc( LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX ) );
-        B=(CBLAS_TEST_ZOMPLEX* )malloc( LDB*(*k)*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        A=(CBLAS_TEST_ZOMPLEX* )malloc( (size_t)LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        B=(CBLAS_TEST_ZOMPLEX* )malloc( (size_t)LDB*(*k)*sizeof(CBLAS_TEST_ZOMPLEX ) );
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ ){
 	      A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -357,7 +357,7 @@ void F77_zher2k(int *order, char *uplow, char *transp, int *n, int *k,
            }
      }
      LDC = *n+1;
-     C=(CBLAS_TEST_ZOMPLEX* )malloc( (*n)*LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     C=(CBLAS_TEST_ZOMPLEX* )malloc( (*n)*(size_t)LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ ) {
            C[i*LDC+j].real=c[j*(*ldc)+i].real;
@@ -397,8 +397,8 @@ void F77_zsyr2k(int *order, char *uplow, char *transp, int *n, int *k,
      if (trans == CblasNoTrans) {
         LDA = *k+1;
         LDB = *k+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
-        B=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDB*sizeof(CBLAS_TEST_ZOMPLEX));
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        B=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDB*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*n; i++ )
            for( j=0; j<*k; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -410,8 +410,8 @@ void F77_zsyr2k(int *order, char *uplow, char *transp, int *n, int *k,
      else {
         LDA = *n+1;
         LDB = *n+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc(LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
-        B=(CBLAS_TEST_ZOMPLEX* )malloc(LDB*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((size_t)LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
+        B=(CBLAS_TEST_ZOMPLEX* )malloc((size_t)LDB*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*k; i++ )
            for( j=0; j<*n; j++ ){
 	      A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -421,7 +421,7 @@ void F77_zsyr2k(int *order, char *uplow, char *transp, int *n, int *k,
            }
      }
      LDC = *n+1;
-     C=(CBLAS_TEST_ZOMPLEX* )malloc( (*n)*LDC*sizeof(CBLAS_TEST_ZOMPLEX));
+     C=(CBLAS_TEST_ZOMPLEX* )malloc( (*n)*(size_t)LDC*sizeof(CBLAS_TEST_ZOMPLEX));
      for( i=0; i<*n; i++ )
         for( j=0; j<*n; j++ ) {
            C[i*LDC+j].real=c[j*(*ldc)+i].real;
@@ -463,7 +463,7 @@ void F77_ztrmm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
   if (*order == TEST_ROW_MJR) {
      if (side == CblasLeft) {
         LDA = *m+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*m; i++ )
            for( j=0; j<*m; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -472,7 +472,7 @@ void F77_ztrmm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
      }
      else{
         LDA = *n+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*n; i++ )
            for( j=0; j<*n; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -480,7 +480,7 @@ void F77_ztrmm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
            }
      }
      LDB = *n+1;
-     B=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDB*sizeof(CBLAS_TEST_ZOMPLEX));
+     B=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*(size_t)LDB*sizeof(CBLAS_TEST_ZOMPLEX));
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ ) {
            B[i*LDB+j].real=b[j*(*ldb)+i].real;
@@ -522,7 +522,7 @@ void F77_ztrsm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
   if (*order == TEST_ROW_MJR) {
      if (side == CblasLeft) {
         LDA = *m+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc( (*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        A=(CBLAS_TEST_ZOMPLEX* )malloc( (*m)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
         for( i=0; i<*m; i++ )
            for( j=0; j<*m; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -531,7 +531,7 @@ void F77_ztrsm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
      }
      else{
         LDA = *n+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*(size_t)LDA*sizeof(CBLAS_TEST_ZOMPLEX));
         for( i=0; i<*n; i++ )
            for( j=0; j<*n; j++ ) {
               A[i*LDA+j].real=a[j*(*lda)+i].real;
@@ -539,7 +539,7 @@ void F77_ztrsm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
            }
      }
      LDB = *n+1;
-     B=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDB*sizeof(CBLAS_TEST_ZOMPLEX));
+     B=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*(size_t)LDB*sizeof(CBLAS_TEST_ZOMPLEX));
      for( i=0; i<*m; i++ )
         for( j=0; j<*n; j++ ) {
            B[i*LDB+j].real=b[j*(*ldb)+i].real;

ctest/constant.c

@@ -1,3 +1,4 @@
+#include "cblas_test.h"
 int CBLAS_CallFromC;
 int RowMajorStrg;
 

driver/level3/Makefile

@@ -425,7 +425,7 @@ cgemm_rr.$(SUFFIX) : gemm.c level3.c  ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 cgemm_rc.$(SUFFIX) : gemm.c level3.c  ../../param.h
-	$(CC) $(CFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 cgemm_cn.$(SUFFIX) : gemm.c level3.c  ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -473,7 +473,7 @@ zgemm_rr.$(SUFFIX) : gemm.c level3.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 zgemm_rc.$(SUFFIX) : gemm.c level3.c ../../param.h
-	$(CC) $(CFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 zgemm_cn.$(SUFFIX) : gemm.c level3.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -521,7 +521,7 @@ xgemm_rr.$(SUFFIX) : gemm.c level3.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 xgemm_rc.$(SUFFIX) : gemm.c level3.c ../../param.h
-	$(CC) $(CFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 xgemm_cn.$(SUFFIX) : gemm.c level3.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -632,7 +632,7 @@ cgemm_thread_rr.$(SUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 cgemm_thread_rc.$(SUFFIX) : gemm.c level3_thread.c ../../param.h
-	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 cgemm_thread_cn.$(SUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -680,7 +680,7 @@ zgemm_thread_rr.$(SUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 zgemm_thread_rc.$(SUFFIX) : gemm.c level3_thread.c ../../param.h
-	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 zgemm_thread_cn.$(SUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -728,7 +728,7 @@ xgemm_thread_rr.$(SUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 xgemm_thread_rc.$(SUFFIX) : gemm.c level3_thread.c ../../param.h
-	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 xgemm_thread_cn.$(SUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -1895,7 +1895,7 @@ cgemm3m_rr.$(SUFFIX) : gemm3m.c gemm3m_level3.c
 	$(CC) $(CFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 cgemm3m_rc.$(SUFFIX) : gemm3m.c gemm3m_level3.c
-	$(CC) $(CFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 cgemm3m_cn.$(SUFFIX) : gemm3m.c gemm3m_level3.c
 	$(CC) $(CFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -1943,7 +1943,7 @@ zgemm3m_rr.$(SUFFIX) : gemm3m.c gemm3m_level3.c
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 zgemm3m_rc.$(SUFFIX) : gemm3m.c gemm3m_level3.c
-	$(CC) $(CFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 zgemm3m_cn.$(SUFFIX) : gemm3m.c gemm3m_level3.c
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -1991,7 +1991,7 @@ xgemm3m_rr.$(SUFFIX) : gemm3m.c gemm3m_level3.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 xgemm3m_rc.$(SUFFIX) : gemm3m.c gemm3m_level3.c ../../param.h
-	$(CC) $(CFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 xgemm3m_cn.$(SUFFIX) : gemm3m.c gemm3m_level3.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -2048,7 +2048,7 @@ cgemm3m_thread_rr.$(SUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 cgemm3m_thread_rc.$(SUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
-	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 cgemm3m_thread_cn.$(SUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -2096,7 +2096,7 @@ zgemm3m_thread_rr.$(SUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 zgemm3m_thread_rc.$(SUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
-	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 zgemm3m_thread_cn.$(SUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -2144,7 +2144,7 @@ xgemm3m_thread_rr.$(SUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 xgemm3m_thread_rc.$(SUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
-	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 xgemm3m_thread_cn.$(SUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(CFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -2817,7 +2817,7 @@ cgemm_rr.$(PSUFFIX) : gemm.c level3.c  ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 cgemm_rc.$(PSUFFIX) : gemm.c level3.c  ../../param.h
-	$(CC) $(PFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 cgemm_cn.$(PSUFFIX) : gemm.c level3.c  ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -2865,7 +2865,7 @@ zgemm_rr.$(PSUFFIX) : gemm.c level3.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 zgemm_rc.$(PSUFFIX) : gemm.c level3.c ../../param.h
-	$(CC) $(PFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 zgemm_cn.$(PSUFFIX) : gemm.c level3.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -2913,7 +2913,7 @@ xgemm_rr.$(PSUFFIX) : gemm.c level3.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 xgemm_rc.$(PSUFFIX) : gemm.c level3.c ../../param.h
-	$(CC) $(PFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 xgemm_cn.$(PSUFFIX) : gemm.c level3.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -3025,7 +3025,7 @@ cgemm_thread_rr.$(PSUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 cgemm_thread_rc.$(PSUFFIX) : gemm.c level3_thread.c ../../param.h
-	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 cgemm_thread_cn.$(PSUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -3073,7 +3073,7 @@ zgemm_thread_rr.$(PSUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 zgemm_thread_rc.$(PSUFFIX) : gemm.c level3_thread.c ../../param.h
-	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 zgemm_thread_cn.$(PSUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -3121,7 +3121,7 @@ xgemm_thread_rr.$(PSUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 xgemm_thread_rc.$(PSUFFIX) : gemm.c level3_thread.c ../../param.h
-	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 xgemm_thread_cn.$(PSUFFIX) : gemm.c level3_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -4288,7 +4288,7 @@ cgemm3m_rr.$(PSUFFIX) : gemm3m.c gemm3m_level3.c
 	$(CC) $(PFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 cgemm3m_rc.$(PSUFFIX) : gemm3m.c gemm3m_level3.c
-	$(CC) $(PFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 cgemm3m_cn.$(PSUFFIX) : gemm3m.c gemm3m_level3.c
 	$(CC) $(PFLAGS) $(BLOCKS) -c -UDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -4336,7 +4336,7 @@ zgemm3m_rr.$(PSUFFIX) : gemm3m.c gemm3m_level3.c
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 zgemm3m_rc.$(PSUFFIX) : gemm3m.c gemm3m_level3.c
-	$(CC) $(PFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 zgemm3m_cn.$(PSUFFIX) : gemm3m.c gemm3m_level3.c
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -4384,7 +4384,7 @@ xgemm3m_rr.$(PSUFFIX) : gemm3m.c gemm3m_level3.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 xgemm3m_rc.$(PSUFFIX) : gemm3m.c gemm3m_level3.c ../../param.h
-	$(CC) $(PFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 xgemm3m_cn.$(PSUFFIX) : gemm3m.c gemm3m_level3.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DXDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -4441,7 +4441,7 @@ cgemm3m_thread_rr.$(PSUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 cgemm3m_thread_rc.$(PSUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
-	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 cgemm3m_thread_cn.$(PSUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -UDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -4489,7 +4489,7 @@ zgemm3m_thread_rr.$(PSUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 zgemm3m_thread_rc.$(PSUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
-	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 zgemm3m_thread_cn.$(PSUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DDOUBLE -DCOMPLEX -DCN $< -o $(@F)
@@ -4537,7 +4537,7 @@ xgemm3m_thread_rr.$(PSUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRR $< -o $(@F)
 
 xgemm3m_thread_rc.$(PSUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
-	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRC $< -o $(@F)
+	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DRC=RC $< -o $(@F)
 
 xgemm3m_thread_cn.$(PSUFFIX) : gemm3m.c level3_gemm3m_thread.c ../../param.h
 	$(CC) $(PFLAGS) $(BLOCKS) -c -DTHREADED_LEVEL3 -DXDOUBLE -DCOMPLEX -DCN $< -o $(@F)

driver/others/dynamic.c

@@ -404,6 +404,7 @@ static int get_vendor(void){
   if (!strcmp(vendor.vchar, "GenuineIntel")) return VENDOR_INTEL;
   if (!strcmp(vendor.vchar, "AuthenticAMD")) return VENDOR_AMD;
   if (!strcmp(vendor.vchar, "CentaurHauls")) return VENDOR_CENTAUR;
+  if (!strcmp(vendor.vchar, "  Shanghai  ")) return VENDOR_CENTAUR;
   if (!strcmp(vendor.vchar, "HygonGenuine")) return VENDOR_HYGON;
 
   if ((eax == 0) || ((eax & 0x500) != 0)) return VENDOR_INTEL;
@@ -621,6 +622,22 @@ static gotoblas_t *get_coretype(void){
 	    return &gotoblas_NEHALEM;
 	  }
         }
+	if (model == 10) {
+          // Ice Lake SP
+	   if(support_avx512_bf16())
+             return &gotoblas_COOPERLAKE;
+          if (support_avx512()) 
+	    return &gotoblas_SKYLAKEX;
+	  if(support_avx2())
+	    return &gotoblas_HASWELL;
+	  if(support_avx()) {
+	    openblas_warning(FALLBACK_VERBOSE, SANDYBRIDGE_FALLBACK);
+	    return &gotoblas_SANDYBRIDGE;
+	  } else {
+	    openblas_warning(FALLBACK_VERBOSE, NEHALEM_FALLBACK);
+	    return &gotoblas_NEHALEM;
+	  }
+        }      
         return NULL;  
       case 7:
 	if (model == 10) // Goldmont Plus
@@ -808,6 +825,9 @@ static gotoblas_t *get_coretype(void){
     switch (family) {
     case 0x6:
       return &gotoblas_NANO;
+      break;
+    case 0x7:
+      return &gotoblas_NEHALEM;
     }
   }
 

driver/others/dynamic_arm64.c

@@ -99,6 +99,11 @@ extern gotoblas_t  gotoblas_NEOVERSEN1;
 #else
 #define gotoblas_NEOVERSEN1 gotoblas_ARMV8
 #endif
+#ifdef DYN_CORTEX_A55
+extern gotoblas_t  gotoblas_CORTEXA55;
+#else
+#define gotoblas_CORTEXA55 gotoblas_ARMV8
+#endif
 #else
 extern gotoblas_t  gotoblas_CORTEXA53;
 extern gotoblas_t  gotoblas_CORTEXA57;
@@ -111,11 +116,12 @@ extern gotoblas_t  gotoblas_TSV110;
 extern gotoblas_t  gotoblas_EMAG8180;
 extern gotoblas_t  gotoblas_NEOVERSEN1;
 extern gotoblas_t  gotoblas_THUNDERX3T110;
+extern gotoblas_t  gotoblas_CORTEXA55;
 #endif
 
 extern void openblas_warning(int verbose, const char * msg);
 
-#define NUM_CORETYPES   12
+#define NUM_CORETYPES   13
 
 /*
  * In case asm/hwcap.h is outdated on the build system, make sure
@@ -126,7 +132,7 @@ extern void openblas_warning(int verbose, const char * msg);
 #endif
 
 #define get_cpu_ftr(id, var) ({					\
-		__asm__ ("mrs %0, "#id : "=r" (var));		\
+		__asm__ __volatile__ ("mrs %0, "#id : "=r" (var));		\
 	})
 
 static char *corename[] = {
@@ -142,6 +148,7 @@ static char *corename[] = {
   "emag8180",
   "neoversen1",
   "thunderx3t110",
+  "cortexa55",
   "unknown"
 };
 
@@ -158,6 +165,7 @@ char *gotoblas_corename(void) {
   if (gotoblas == &gotoblas_EMAG8180)     return corename[ 9];
   if (gotoblas == &gotoblas_NEOVERSEN1)   return corename[10];
   if (gotoblas == &gotoblas_THUNDERX3T110) return corename[11];
+  if (gotoblas == &gotoblas_CORTEXA55)    return corename[12];
   return corename[NUM_CORETYPES];
 }
 
@@ -189,6 +197,7 @@ static gotoblas_t *force_coretype(char *coretype) {
     case  9: return (&gotoblas_EMAG8180);
     case 10: return (&gotoblas_NEOVERSEN1);
     case 11: return (&gotoblas_THUNDERX3T110);
+    case 12: return (&gotoblas_CORTEXA55);
   }
   snprintf(message, 128, "Core not found: %s\n", coretype);
   openblas_warning(1, message);
@@ -247,6 +256,8 @@ static gotoblas_t *get_coretype(void) {
           return &gotoblas_CORTEXA73;
         case 0xd0c: // Neoverse N1
           return &gotoblas_NEOVERSEN1;
+	case 0xd05: // Cortex A55
+	  return &gotoblas_CORTEXA55;
       }
       break;
     case 0x42: // Broadcom

driver/others/memory.c

@@ -1702,7 +1702,6 @@ inline int atoi(const char *str) { return 0; }
 #include <sys/sysinfo.h>
 #include <sched.h>
 #include <errno.h>
-#include <linux/unistd.h>
 #include <sys/syscall.h>
 #include <sys/time.h>
 #include <sys/resource.h>

exports/Makefile

@@ -139,9 +139,13 @@ endif
 ifneq (,$(filter 1 2,$(NOFORTRAN)))
 #only build without Fortran
 	$(CC) $(CFLAGS) $(LDFLAGS) -all_load -headerpad_max_install_names -install_name "$(CURDIR)/../$(INTERNALNAME)" -dynamiclib -o ../$(LIBDYNNAME) $< -Wl,-exported_symbols_list,osx.def  $(FEXTRALIB)
+else
+ifeq ($(F_COMPILER), INTEL)
+	$(FC) $(FFLAGS) $(LDFLAGS) -all-load -headerpad_max_install_names -install_name "$(CURDIR)/../$(INTERNALNAME)" -dynamiclib -o ../$(LIBDYNNAME) $< -Wl,-exported_symbols_list,osx.def
 else
 	$(FC) $(FFLAGS) $(LDFLAGS) -all_load -headerpad_max_install_names -install_name "$(CURDIR)/../$(INTERNALNAME)" -dynamiclib -o ../$(LIBDYNNAME) $< -Wl,-exported_symbols_list,osx.def  $(FEXTRALIB)
 endif
+endif
 
 dllinit.$(SUFFIX) : dllinit.c
 	$(CC) $(CFLAGS) -c -o $(@F) -s $<

f_check

@@ -314,11 +314,11 @@ if ($link ne "") {
 
     $link =~ s/\-Y\sP\,/\-Y/g;
     
-    $link =~ s/\-R\s*/\-rpath\@/g;
+    $link =~ s/\-R\s*/\-rpath\%/g;
 
-    $link =~ s/\-rpath\s+/\-rpath\@/g;
+    $link =~ s/\-rpath\s+/\-rpath\%/g;
 
-    $link =~ s/\-rpath-link\s+/\-rpath-link\@/g;
+    $link =~ s/\-rpath-link\s+/\-rpath-link\%/g;
 
     @flags = split(/[\s\,\n]/, $link);
     # remove leading and trailing quotes from each flag.
@@ -344,13 +344,13 @@ if ($link ne "") {
 	}
 
 
-	if ($flags =~ /^\-rpath\@/) {
-	    $flags =~ s/\@/\,/g;
+	if ($flags =~ /^\-rpath\%/) {
+	    $flags =~ s/\%/\,/g;
 	    $linker_L .= "-Wl,". $flags . " " ;
 	}
 
-	if ($flags =~ /^\-rpath-link\@/) {
-	    $flags =~ s/\@/\,/g;
+	if ($flags =~ /^\-rpath-link\%/) {
+	    $flags =~ s/\%/\,/g;
 	    $linker_L .= "-Wl,". $flags . " " ;
 	}
 	if ($flags =~ /-lgomp/ && $ENV{"CC"} =~ /clang/) {
@@ -391,10 +391,6 @@ if ($link ne "") {
 
 }
 
-if ($vendor eq "INTEL"){
-    $linker_a .= "-lgfortran"
-}
-
 if ($vendor eq "FLANG"){
     $linker_a .= "-lflang"
 }

getarch.c

@@ -1159,6 +1159,21 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #else
 #endif
 
+#ifdef FORCE_CORTEXA55
+#define FORCE
+#define ARCHITECTURE    "ARM64"
+#define SUBARCHITECTURE "CORTEXA55"
+#define SUBDIRNAME      "arm64"
+#define ARCHCONFIG   "-DCORTEXA55 " \
+       "-DL1_CODE_SIZE=16384 -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=3 " \
+       "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=64 -DL1_DATA_ASSOCIATIVE=2 " \
+       "-DL2_SIZE=65536 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=16 " \
+       "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
+       "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
+#define LIBNAME   "cortexa55"
+#define CORENAME  "CORTEXA55"
+#else
+#endif
 
 #ifdef FORCE_FALKOR
 #define FORCE

getarch_2nd.c

@@ -4,7 +4,15 @@
 #else
 #include "config_kernel.h"
 #endif
-#include "common.h"
+#if (defined(__WIN32__) || defined(__WIN64__) || defined(__CYGWIN32__) || defined(__CYGWIN64__) || defined(_WIN32) || defined(_WIN64)) && defined(__64BIT__)
+typedef long long BLASLONG;
+typedef unsigned long long BLASULONG;
+#else
+typedef long BLASLONG;
+typedef unsigned long BLASULONG;
+#endif
+
+#include "param.h"
 
 int main(int argc, char **argv) {
 

interface/gemm.c

@@ -49,6 +49,8 @@
 #define ERROR_NAME "QGEMM "
 #elif defined(DOUBLE)
 #define ERROR_NAME "DGEMM "
+#elif defined(BFLOAT16)
+#define ERROR_NAME "SBGEMM "
 #else
 #define ERROR_NAME "SGEMM "
 #endif
@@ -124,6 +126,7 @@ void NAME(char *TRANSA, char *TRANSB,
 
 #ifdef SMP
   double MNK;
+#if defined(USE_SIMPLE_THREADED_LEVEL3) || !defined(NO_AFFINITY)
 #ifndef COMPLEX
 #ifdef XDOUBLE
   int mode  =  BLAS_XDOUBLE | BLAS_REAL;
@@ -142,6 +145,7 @@ void NAME(char *TRANSA, char *TRANSB,
 #endif
 #endif
 #endif
+#endif
 
 #if defined(SMP) && !defined(NO_AFFINITY) && !defined(USE_SIMPLE_THREADED_LEVEL3)
   int nodes;

interface/gemv.c

@@ -202,6 +202,11 @@ void CNAME(enum CBLAS_ORDER order,
 
   if (alpha == ZERO) return;
 
+  if (trans == 0 && incx == 1 && incy == 1 && m*n < 2304 *GEMM_MULTITHREAD_THRESHOLD) {
+    GEMV_N(m, n, 0, alpha, a, lda, x, incx, y, incy, NULL);
+    return;
+  }    
+
   IDEBUG_START;
 
   FUNCTION_PROFILE_START();

interface/ger.c

@@ -164,6 +164,11 @@ void CNAME(enum CBLAS_ORDER order,
   if (m == 0 || n == 0) return;
   if (alpha == 0.) return;
 
+  if (incx == 1 && incy == 1 && 1L*m*n <= 2048 *GEMM_MULTITHREAD_THRESHOLD) {
+    GER(m, n, 0, alpha, x, incx, y, incy, a, lda, NULL);
+    return;
+  }  
+
   IDEBUG_START;
 
   FUNCTION_PROFILE_START();

interface/imatcopy.c

@@ -150,9 +150,9 @@ void CNAME( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows,
 #endif
 
 	if ( *lda >  *ldb )
-		msize = (*lda) * (*ldb)  * sizeof(FLOAT);
+		msize = (size_t)(*lda) * (*ldb)  * sizeof(FLOAT);
 	else
-		msize = (*ldb) * (*ldb)  * sizeof(FLOAT);
+		msize = (size_t)(*ldb) * (*ldb)  * sizeof(FLOAT);
 
 	b = malloc(msize);
 	if ( b == NULL )

interface/lapack/getrf.c

@@ -95,7 +95,14 @@ int NAME(blasint *M, blasint *N, FLOAT *a, blasint *ldA, blasint *ipiv, blasint
 
 #ifdef SMP
   args.common = NULL;
-  args.nthreads = num_cpu_avail(4);
+#ifndef DOUBLE
+  if (args.m*args.n < 40000)
+#else
+  if (args.m*args.n < 10000)
+#endif
+	args.nthreads=1;
+  else
+	args.nthreads = num_cpu_avail(4);
 
   if (args.nthreads == 1) {
 #endif

interface/lapack/potrf.c

@@ -112,6 +112,13 @@ int NAME(char *UPLO, blasint *N, FLOAT *a, blasint *ldA, blasint *Info){
 
 #ifdef SMP
   args.common = NULL;
+#ifndef DOUBLE
+  if (args.n <128)
+#else
+  if (args.n <64)
+#endif
+    args.nthreads = 1;
+  else
   args.nthreads = num_cpu_avail(4);
 
   if (args.nthreads == 1) {

interface/lapack/potri.c

@@ -121,6 +121,9 @@ int NAME(char *UPLO, blasint *N, FLOAT *a, blasint *ldA, blasint *Info){
 
 #ifdef SMP
   args.common = NULL;
+  if (args.n < 180)
+    args.nthreads = 1;
+  else
   args.nthreads = num_cpu_avail(4);
 
   if (args.nthreads == 1) {

interface/lapack/zgetrf.c

@@ -95,7 +95,10 @@ int NAME(blasint *M, blasint *N, FLOAT *a, blasint *ldA, blasint *ipiv, blasint
 
 #ifdef SMP
   args.common = NULL;
-  args.nthreads = num_cpu_avail(4);
+  if (args.m*args.n <10000)
+	args.nthreads = 1;
+  else
+	args.nthreads = num_cpu_avail(4);
 
   if (args.nthreads == 1) {
 #endif

interface/lapack/zpotrf.c

@@ -112,6 +112,13 @@ int NAME(char *UPLO, blasint *N, FLOAT *a, blasint *ldA, blasint *Info){
 
 #ifdef SMP
   args.common = NULL;
+#ifndef DOUBLE
+  if (args.n < 64)
+#else
+  if (args.n < 64)
+#endif
+    args.nthreads = 1;
+  else
   args.nthreads = num_cpu_avail(4);
 
   if (args.nthreads == 1) {

interface/lapack/zpotri.c

@@ -121,6 +121,15 @@ int NAME(char *UPLO, blasint *N, FLOAT *a, blasint *ldA, blasint *Info){
 
 #ifdef SMP
   args.nthreads = num_cpu_avail(4);
+#ifndef DOUBLE
+  if (args.n < 200)
+#else
+  if (args.n < 150)
+#endif
+	  args.nthreads=1;
+  else
+#endif
+  args.nthreads = num_cpu_avail(4);
 
   if (args.nthreads == 1) {
 #endif

interface/spr.c

@@ -167,6 +167,26 @@ void CNAME(enum CBLAS_ORDER order,
 
   FUNCTION_PROFILE_START();
 
+  if (incx == 1 && n <100) {
+    blasint i;
+    if (uplo==0) {
+      for (i = 0; i < n; i++){
+        if (x[i] != ZERO) {
+          AXPYU_K(i + 1, 0, 0, alpha * x[i], x,     1, a, 1, NULL, 0);
+        }
+        a += i + 1;
+      }
+    } else { 
+      for (i = 0; i < n; i++){
+        if (x[i] != ZERO) {
+          AXPYU_K(n - i, 0, 0, alpha * x[i], x + i, 1, a, 1, NULL, 0);
+        }
+        a += n - i;
+      }
+    }
+    return;
+  }
+
   if (incx < 0 ) x -= (n - 1) * incx;
 
   buffer = (FLOAT *)blas_memory_alloc(1);

interface/spr2.c

@@ -168,6 +168,24 @@ void CNAME(enum CBLAS_ORDER order,
 
   if (alpha == ZERO) return;
 
+  if (incx == 1 && incy == 1 && n < 50) {
+    blasint i;
+    if (!uplo) {
+      for (i = 0; i < n; i++){
+        AXPYU_K(i + 1, 0, 0, alpha * x[i], y,     1, a, 1, NULL, 0);
+        AXPYU_K(i + 1, 0, 0, alpha * y[i], x,     1, a, 1, NULL, 0);
+        a += i + 1;
+      }
+    } else {
+      for (i = 0; i < n; i++){
+	AXPYU_K(n - i, 0, 0, alpha * x[i], y + i, 1, a, 1, NULL, 0);
+        AXPYU_K(n - i, 0, 0, alpha * y[i], x + i, 1, a, 1, NULL, 0);
+        a += n - i;
+      }
+    }
+    return;
+  }
+
   IDEBUG_START;
 
   FUNCTION_PROFILE_START();

interface/syr.c

@@ -168,7 +168,28 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint n, FLOAT alpha,
   IDEBUG_START;
 
   FUNCTION_PROFILE_START();
+#if 1
+  if (incx == 1 && n < 100) {
+    BLASLONG i;
 
+    if (uplo == 0) {
+      for (i = 0; i < n; i++){
+        if (x[i] != ZERO) {
+          AXPYU_K(i + 1, 0, 0, alpha * x[i], x,     1, a, 1, NULL, 0);
+        }
+        a += lda;
+      }  
+    } else {
+      for (i = 0; i < n; i++){
+        if (x[i] != ZERO) {
+          AXPYU_K(n - i, 0, 0, alpha * x[i], x + i, 1, a, 1, NULL, 0);
+        }
+        a += 1 + lda;
+      }
+    }
+    return;
+  } 
+#endif
   if (incx < 0 ) x -= (n - 1) * incx;
 
   buffer = (FLOAT *)blas_memory_alloc(1);

interface/syr2.c

@@ -170,6 +170,25 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint n, FLOAT alpha,
 
   IDEBUG_START;
 
+  if (incx == 1 && incy == 1 && n < 100) {
+    blasint i;
+    if (!uplo) {
+      for (i = 0; i < n; i++){
+        AXPYU_K(i + 1, 0, 0, alpha * x[i], y,     1, a, 1, NULL, 0);
+        AXPYU_K(i + 1, 0, 0, alpha * y[i], x,     1, a, 1, NULL, 0);
+        a += lda;
+      }
+    } else {
+      for (i = 0; i < n; i++){
+        AXPYU_K(n - i, 0, 0, alpha * x[i], y + i, 1, a, 1, NULL, 0);
+        AXPYU_K(n - i, 0, 0, alpha * y[i], x + i, 1, a, 1, NULL, 0);
+        a += 1 + lda;
+      }
+    }
+    return;
+  }
+
+	  
   FUNCTION_PROFILE_START();
 
   if (incx < 0 ) x -= (n - 1) * incx;

interface/syrk.c

@@ -354,6 +354,17 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Tr
 #endif
 
   args.common = NULL;
+#ifndef COMPLEX
+#ifdef DOUBLE
+  if (args.n < 100)
+#else
+  if (args.n < 200)
+#endif
+#else
+  if (args.n < 65)
+#endif
+  args.nthreads = 1;
+  else
   args.nthreads = num_cpu_avail(3);
 
   if (args.nthreads == 1) {

interface/trsv.c

@@ -188,6 +188,12 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,
 
   if (n == 0) return;
 
+  if (incx == 1 && trans == 0 && n < 50) {
+    buffer = NULL;
+    (trsv[(trans<<2) | (uplo<<1) | unit])(n, a, lda, x, incx, buffer);
+    return;  
+  }
+
   IDEBUG_START;
 
   FUNCTION_PROFILE_START();

interface/zimatcopy.c

@@ -172,9 +172,9 @@ void CNAME( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows,
 #endif
 
 	if ( *lda >  *ldb )
-                msize = (*lda) * (*ldb)  * sizeof(FLOAT) * 2;
+                msize = (size_t)(*lda) * (*ldb)  * sizeof(FLOAT) * 2;
         else
-                msize = (*ldb) * (*ldb)  * sizeof(FLOAT) * 2;
+                msize = (size_t)(*ldb) * (*ldb)  * sizeof(FLOAT) * 2;
 
         b = malloc(msize);
         if ( b == NULL )

interface/zrotg.c

@@ -79,8 +79,12 @@ void NAME(FLOAT *DA, FLOAT *DB, FLOAT *C, FLOAT *S){
       aa_i = fabs(da_r);
     }
 
-    scale = (aa_i / aa_r);
-    ada = aa_r * sqrt(ONE + scale * scale);
+    if (aa_r == ZERO) {
+	ada = 0.;
+    } else {
+        scale = (aa_i / aa_r);
+        ada = aa_r * sqrt(ONE + scale * scale);
+    }
 
     bb_r = fabs(db_r);
     bb_i = fabs(db_i);
@@ -90,9 +94,12 @@ void NAME(FLOAT *DA, FLOAT *DB, FLOAT *C, FLOAT *S){
       bb_i = fabs(bb_r);
     }
 
-    scale = (bb_i / bb_r);
-    adb = bb_r * sqrt(ONE + scale * scale);
-
+    if (bb_r == ZERO) {
+	adb = 0.;
+    } else {
+    	scale = (bb_i / bb_r);
+    	adb = bb_r * sqrt(ONE + scale * scale);
+    }
     scale = ada + adb;
 
     aa_r    = da_r / scale;

interface/zsyr.c

@@ -172,6 +172,32 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, int n, FLOAT alpha, FLO
 
   if ((alpha_r == ZERO) && (alpha_i == ZERO)) return;
 
+  if (incx == 1 && n < 50) {
+    blasint i;
+    if (!uplo) {
+      for (i = 0; i < n; i++){
+        if ((x[i * 2 + 0] != ZERO) || (x[i * 2 + 1] != ZERO)) {
+          AXPYU_K(i + 1, 0, 0,
+              alpha_r * x[i * 2 + 0] - alpha_i * x[i * 2 + 1],
+              alpha_i * x[i * 2 + 0] + alpha_r * x[i * 2 + 1],
+              x,         1, a, 1, NULL, 0);
+        }
+        a += lda;
+      }
+    } else {
+      for (i = 0; i < n; i++){
+        if ((x[i * 2 + 0] != ZERO) || (x[i * 2 + 1] != ZERO)) {
+          AXPYU_K(n - i, 0, 0,
+              alpha_r * x[i * 2 + 0] - alpha_i * x[i * 2 + 1],
+              alpha_i * x[i * 2 + 0] + alpha_r * x[i * 2 + 1],
+              x + i * 2, 1, a, 1, NULL, 0);
+        }
+        a += 2 + lda;
+      }
+    }
+    return;
+  }
+
   IDEBUG_START;
 
   FUNCTION_PROFILE_START();

interface/ztrsv.c

@@ -199,6 +199,12 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,
 
   if (n == 0) return;
 
+  if (incx == 1 && trans == 0 && n < 50) {
+    buffer = NULL;
+    (trsv[(trans<<2) | (uplo<<1) | unit])(n, a, lda, x, incx, buffer);
+    return;
+  }
+
   IDEBUG_START;
 
   FUNCTION_PROFILE_START();

kernel/Makefile.L3

@@ -818,6 +818,8 @@ ifeq ($(OS), AIX)
 	m4 zgemm_kernel_n.s > zgemm_kernel_n_nomacros.s
 	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DNN zgemm_kernel_n_nomacros.s -o $@
 	rm zgemm_kernel_n.s zgemm_kernel_n_nomacros.s
+else ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DDOUBLE -DCOMPLEX -DNN $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DNN $< -o $@
 endif
@@ -828,6 +830,8 @@ ifeq ($(OS), AIX)
 	m4 zgemm_kernel_l.s > zgemm_kernel_l_nomacros.s
 	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DCN zgemm_kernel_l_nomacros.s -o $@
 	rm zgemm_kernel_l.s zgemm_kernel_l_nomacros.s
+else ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DDOUBLE -DCOMPLEX -DCN $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DCN $< -o $@
 endif
@@ -838,6 +842,8 @@ ifeq ($(OS), AIX)
 	m4 zgemm_kernel_r.s > zgemm_kernel_r_nomacros.s
 	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DNC zgemm_kernel_r_nomacros.s -o $@
 	rm zgemm_kernel_r.s zgemm_kernel_r_nomacros.s
+else ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DDOUBLE -DCOMPLEX -DNC $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DNC $< -o $@
 endif
@@ -848,6 +854,8 @@ ifeq ($(OS), AIX)
 	m4 zgemm_kernel_b.s > zgemm_kernel_b_nomacros.s
 	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DCC zgemm_kernel_b_nomacros.s -o $@
 	rm zgemm_kernel_b.s zgemm_kernel_b_nomacros.s
+else ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DDOUBLE -DCOMPLEX -DCC $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DCC $< -o $@
 endif
@@ -1044,6 +1052,8 @@ ifeq ($(OS), AIX)
 	m4 ztrmm_kernel_ln.s > ztrmm_kernel_ln_nomacros.s
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -UTRANSA -UCONJ -DNN ztrmm_kernel_ln_nomacros.s -o $@
 	rm ztrmm_kernel_ln.s ztrmm_kernel_ln_nomacros.s
+else ifeq ($(CORE), SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -UTRANSA -UCONJ -DNN $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -UTRANSA -UCONJ -DNN $< -o $@
 endif
@@ -1054,6 +1064,8 @@ ifeq ($(OS), AIX)
 	m4 ztrmm_kernel_lt.s > ztrmm_kernel_lt_nomacros.s
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -DTRANSA -UCONJ -DNN ztrmm_kernel_lt_nomacros.s -o $@
 	rm ztrmm_kernel_lt.s ztrmm_kernel_lt_nomacros.s
+else ifeq ($(CORE), SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -DTRANSA -UCONJ -DNN $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -DTRANSA -UCONJ -DNN $< -o $@
 endif
@@ -1064,6 +1076,8 @@ ifeq ($(OS), AIX)
 	m4 ztrmm_kernel_lr.s > ztrmm_kernel_lr_nomacros.s
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -UTRANSA -DCONJ -DCN ztrmm_kernel_lr_nomacros.s -o $@
 	rm ztrmm_kernel_lr.s ztrmm_kernel_lr_nomacros.s
+else ifeq ($(CORE), SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -UTRANSA -DCONJ -DCN $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -UTRANSA -DCONJ -DCN $< -o $@
 endif
@@ -1074,6 +1088,8 @@ ifeq ($(OS), AIX)
 	m4 ztrmm_kernel_lc.s >ztrmm_kernel_lc_nomacros.s
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -DTRANSA -DCONJ -DCN ztrmm_kernel_lc_nomacros.s -o $@
 	rm ztrmm_kernel_lc.s ztrmm_kernel_lc_nomacros.s 
+else ifeq ($(CORE), SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -DTRANSA -DCONJ -DCN $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -DTRANSA -DCONJ -DCN $< -o $@
 endif
@@ -1084,6 +1100,8 @@ ifeq ($(OS), AIX)
 	m4 ztrmm_kernel_rn.s > ztrmm_kernel_rn_nomacros.s
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -UTRANSA -UCONJ -DNN ztrmm_kernel_rn_nomacros.s -o $@
 	rm ztrmm_kernel_rn.s ztrmm_kernel_rn_nomacros.s
+else ifeq ($(CORE), SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -UTRANSA -UCONJ -DNN $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -UTRANSA -UCONJ -DNN $< -o $@
 endif
@@ -1094,6 +1112,8 @@ ifeq ($(OS), AIX)
 	m4 ztrmm_kernel_rt.s > ztrmm_kernel_rt_nomacros.s
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -DTRANSA -UCONJ -DNN ztrmm_kernel_rt_nomacros.s -o $@
 	rm ztrmm_kernel_rt.s ztrmm_kernel_rt_nomacros.s
+else ifeq ($(CORE), SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -DTRANSA -UCONJ -DNN $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -DTRANSA -UCONJ -DNN $< -o $@
 endif
@@ -1104,6 +1124,8 @@ ifeq ($(OS), AIX)
 	m4 ztrmm_kernel_rr.s > ztrmm_kernel_rr_nomacros.s
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -UTRANSA -DCONJ -DNC ztrmm_kernel_rr_nomacros.s -o $@
 	rm ztrmm_kernel_rr.s ztrmm_kernel_rr_nomacros.s
+else ifeq ($(CORE), SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -UTRANSA -DCONJ -DNC $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -UTRANSA -DCONJ -DNC $< -o $@
 endif
@@ -1114,6 +1136,8 @@ ifeq ($(OS), AIX)
 	m4 ztrmm_kernel_rc.s > ztrmm_kernel_rc_nomacros.s
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -DTRANSA -DCONJ -DNC ztrmm_kernel_rc_nomacros.s -o $@
 	rm ztrmm_kernel_rc.s ztrmm_kernel_rc_nomacros.s
+else ifeq ($(CORE), SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -DTRANSA -DCONJ -DNC $< -o $@
 else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -DTRANSA -DCONJ -DNC $< -o $@
 endif
@@ -1187,29 +1211,55 @@ $(KDIR)ctrmm_kernel_RC$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(CGEMMKERNEL)
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -UDOUBLE -DCOMPLEX -ULEFT -DTRANSA -DCONJ -DNC $< -o $@
 
 $(KDIR)ztrmm_kernel_LN$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZGEMMKERNEL)
+ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -UTRANSA -UCONJ -DNN $< -o $@
+else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -UTRANSA -UCONJ -DNN $< -o $@
+endif
 
 $(KDIR)ztrmm_kernel_LT$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZGEMMKERNEL)
+ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -DTRANSA -UCONJ -DNN $< -o $@
+else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -DTRANSA -UCONJ -DNN $< -o $@
-
+endif
 $(KDIR)ztrmm_kernel_LR$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZGEMMKERNEL)
+ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -UTRANSA -DCONJ -DCN $< -o $@
+else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -UTRANSA -DCONJ -DCN $< -o $@
-
+endif
 $(KDIR)ztrmm_kernel_LC$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZGEMMKERNEL)
+ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -DTRANSA -DCONJ -DCN $< -o $@
+else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -DLEFT -DTRANSA -DCONJ -DCN $< -o $@
-
+endif
 $(KDIR)ztrmm_kernel_RN$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZGEMMKERNEL)
+ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -UTRANSA -UCONJ -DNN $< -o $@
+else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -UTRANSA -UCONJ -DNN $< -o $@
-
+endif
 $(KDIR)ztrmm_kernel_RT$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZGEMMKERNEL)
+ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -DTRANSA -UCONJ -DNN $< -o $@
+else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -DTRANSA -UCONJ -DNN $< -o $@
-
+endif
 $(KDIR)ztrmm_kernel_RR$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZGEMMKERNEL)
+ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -UTRANSA -DCONJ -DNC $< -o $@
+else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -UTRANSA -DCONJ -DNC $< -o $@
-
+endif
 $(KDIR)ztrmm_kernel_RC$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZGEMMKERNEL)
+ifeq ($(CORE),SANDYBRIDGE)
+	$(CC) $(filter-out -mavx,$(CFLAGS)) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -DTRANSA -DCONJ -DNC $< -o $@
+else
 	$(CC) $(CFLAGS) -c -DTRMMKERNEL -DDOUBLE -DCOMPLEX -ULEFT -DTRANSA -DCONJ -DNC $< -o $@
 endif
+endif
 
 
 

kernel/arm64/KERNEL.CORTEXA55

@@ -0,0 +1,196 @@
+SAMINKERNEL  = ../arm/amin.c
+DAMINKERNEL  = ../arm/amin.c
+CAMINKERNEL  = ../arm/zamin.c
+ZAMINKERNEL  = ../arm/zamin.c
+
+SMAXKERNEL   = ../arm/max.c
+DMAXKERNEL   = ../arm/max.c
+
+SMINKERNEL   = ../arm/min.c
+DMINKERNEL   = ../arm/min.c
+
+ISAMINKERNEL = ../arm/iamin.c
+IDAMINKERNEL = ../arm/iamin.c
+ICAMINKERNEL = ../arm/izamin.c
+IZAMINKERNEL = ../arm/izamin.c
+
+ISMAXKERNEL  = ../arm/imax.c
+IDMAXKERNEL  = ../arm/imax.c
+
+ISMINKERNEL  = ../arm/imin.c
+IDMINKERNEL  = ../arm/imin.c
+
+STRSMKERNEL_LN	=  ../generic/trsm_kernel_LN.c
+STRSMKERNEL_LT	=  ../generic/trsm_kernel_LT.c
+STRSMKERNEL_RN	=  ../generic/trsm_kernel_RN.c
+STRSMKERNEL_RT	=  ../generic/trsm_kernel_RT.c
+
+DTRSMKERNEL_LN	= ../generic/trsm_kernel_LN.c
+DTRSMKERNEL_LT	= ../generic/trsm_kernel_LT.c
+DTRSMKERNEL_RN	= ../generic/trsm_kernel_RN.c
+DTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
+
+CTRSMKERNEL_LN	= ../generic/trsm_kernel_LN.c
+CTRSMKERNEL_LT	= ../generic/trsm_kernel_LT.c
+CTRSMKERNEL_RN	= ../generic/trsm_kernel_RN.c
+CTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
+
+ZTRSMKERNEL_LN	= ../generic/trsm_kernel_LN.c
+ZTRSMKERNEL_LT	= ../generic/trsm_kernel_LT.c
+ZTRSMKERNEL_RN	= ../generic/trsm_kernel_RN.c
+ZTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
+
+SAMAXKERNEL  = amax.S
+DAMAXKERNEL  = amax.S
+CAMAXKERNEL  = zamax.S
+ZAMAXKERNEL  = zamax.S
+
+SAXPYKERNEL  = axpy.S
+DAXPYKERNEL  = axpy.S
+CAXPYKERNEL  = zaxpy.S
+ZAXPYKERNEL  = zaxpy.S
+
+SROTKERNEL   = rot.S
+DROTKERNEL   = rot.S
+CROTKERNEL   = zrot.S
+ZROTKERNEL   = zrot.S
+
+SSCALKERNEL  = scal.S
+DSCALKERNEL  = scal.S
+CSCALKERNEL  = zscal.S
+ZSCALKERNEL  = zscal.S
+
+SGEMVNKERNEL = gemv_n.S
+DGEMVNKERNEL = gemv_n.S
+CGEMVNKERNEL = zgemv_n.S
+ZGEMVNKERNEL = zgemv_n.S
+
+SGEMVTKERNEL = gemv_t.S
+DGEMVTKERNEL = gemv_t.S
+CGEMVTKERNEL = zgemv_t.S
+ZGEMVTKERNEL = zgemv_t.S
+
+
+SASUMKERNEL    = asum.S
+DASUMKERNEL    = asum.S
+CASUMKERNEL    = casum.S
+ZASUMKERNEL    = zasum.S
+
+SCOPYKERNEL    = copy.S
+DCOPYKERNEL    = copy.S
+CCOPYKERNEL    = copy.S
+ZCOPYKERNEL    = copy.S
+
+SSWAPKERNEL    = swap.S
+DSWAPKERNEL    = swap.S
+CSWAPKERNEL    = swap.S
+ZSWAPKERNEL    = swap.S
+
+ISAMAXKERNEL   = iamax.S
+IDAMAXKERNEL   = iamax.S
+ICAMAXKERNEL   = izamax.S
+IZAMAXKERNEL   = izamax.S
+
+SNRM2KERNEL    = nrm2.S
+DNRM2KERNEL    = nrm2.S
+CNRM2KERNEL    = znrm2.S
+ZNRM2KERNEL    = znrm2.S
+
+ifneq ($(C_COMPILER), PGI)
+SDOTKERNEL   = ../generic/dot.c
+else
+SDOTKERNEL   = dot.S
+endif
+DDOTKERNEL   = dot.S
+ifneq ($(C_COMPILER), PGI)
+CDOTKERNEL   = zdot.S
+ZDOTKERNEL   = zdot.S
+else
+CDOTKERNEL = ../arm/zdot.c
+ZDOTKERNEL = ../arm/zdot.c
+endif
+DSDOTKERNEL  = dot.S
+
+DGEMM_BETA     = dgemm_beta.S
+SGEMM_BETA     = sgemm_beta.S
+
+ifeq ($(SGEMM_UNROLL_M)x$(SGEMM_UNROLL_N), 8x8)
+SGEMMKERNEL    =  sgemm_kernel_$(SGEMM_UNROLL_M)x$(SGEMM_UNROLL_N)_cortexa53.S
+STRMMKERNEL    =  strmm_kernel_$(SGEMM_UNROLL_M)x$(SGEMM_UNROLL_N)_cortexa53.S
+else
+SGEMMKERNEL    =  sgemm_kernel_$(SGEMM_UNROLL_M)x$(SGEMM_UNROLL_N).S
+STRMMKERNEL    =  strmm_kernel_$(SGEMM_UNROLL_M)x$(SGEMM_UNROLL_N).S
+endif
+ifneq ($(SGEMM_UNROLL_M), $(SGEMM_UNROLL_N))
+ifeq ($(SGEMM_UNROLL_M), 16)
+SGEMMITCOPY    =  sgemm_tcopy_$(SGEMM_UNROLL_M).S
+else
+SGEMMITCOPY    =  ../generic/gemm_tcopy_$(SGEMM_UNROLL_M).c
+endif
+ifeq ($(SGEMM_UNROLL_M), 4)
+SGEMMINCOPY    =  sgemm_ncopy_$(SGEMM_UNROLL_M).S
+else
+SGEMMINCOPY    =  ../generic/gemm_ncopy_$(SGEMM_UNROLL_M).c
+endif
+SGEMMINCOPYOBJ =  sgemm_incopy$(TSUFFIX).$(SUFFIX)
+SGEMMITCOPYOBJ =  sgemm_itcopy$(TSUFFIX).$(SUFFIX)
+endif
+
+SGEMMOTCOPY    =  sgemm_tcopy_$(SGEMM_UNROLL_N).S
+SGEMMONCOPY    =  sgemm_ncopy_$(SGEMM_UNROLL_N).S
+SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
+SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+DGEMMKERNEL    =  dgemm_kernel_$(DGEMM_UNROLL_M)x$(DGEMM_UNROLL_N).S
+DTRMMKERNEL    =  dtrmm_kernel_$(DGEMM_UNROLL_M)x$(DGEMM_UNROLL_N).S
+
+ifneq ($(DGEMM_UNROLL_M), $(DGEMM_UNROLL_N))
+
+ifeq ($(DGEMM_UNROLL_M), 8)
+DGEMMINCOPY    =  dgemm_ncopy_$(DGEMM_UNROLL_M).S
+DGEMMITCOPY    =  dgemm_tcopy_$(DGEMM_UNROLL_M).S
+else
+DGEMMINCOPY    =  ../generic/gemm_ncopy_$(DGEMM_UNROLL_M).c
+DGEMMITCOPY    =  ../generic/gemm_tcopy_$(DGEMM_UNROLL_M).c
+endif
+
+DGEMMINCOPYOBJ =  dgemm_incopy$(TSUFFIX).$(SUFFIX)
+DGEMMITCOPYOBJ =  dgemm_itcopy$(TSUFFIX).$(SUFFIX)
+endif
+
+ifeq ($(DGEMM_UNROLL_N), 4)
+DGEMMONCOPY    =  dgemm_ncopy_$(DGEMM_UNROLL_N).S
+DGEMMOTCOPY    =  dgemm_tcopy_$(DGEMM_UNROLL_N).S
+else
+DGEMMONCOPY    =  ../generic/gemm_ncopy_$(DGEMM_UNROLL_N).c
+DGEMMOTCOPY    =  ../generic/gemm_tcopy_$(DGEMM_UNROLL_N).c
+endif
+
+DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
+DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+CGEMMKERNEL    =  cgemm_kernel_$(CGEMM_UNROLL_M)x$(CGEMM_UNROLL_N).S
+CTRMMKERNEL    =  ctrmm_kernel_$(CGEMM_UNROLL_M)x$(CGEMM_UNROLL_N).S
+ifneq ($(CGEMM_UNROLL_M), $(CGEMM_UNROLL_N))
+CGEMMINCOPY    =  ../generic/zgemm_ncopy_$(CGEMM_UNROLL_M).c
+CGEMMITCOPY    =  ../generic/zgemm_tcopy_$(CGEMM_UNROLL_M).c
+CGEMMINCOPYOBJ =  cgemm_incopy$(TSUFFIX).$(SUFFIX)
+CGEMMITCOPYOBJ =  cgemm_itcopy$(TSUFFIX).$(SUFFIX)
+endif
+CGEMMONCOPY    =  ../generic/zgemm_ncopy_$(CGEMM_UNROLL_N).c
+CGEMMOTCOPY    =  ../generic/zgemm_tcopy_$(CGEMM_UNROLL_N).c
+CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
+CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+ZGEMMKERNEL    =  zgemm_kernel_$(ZGEMM_UNROLL_M)x$(ZGEMM_UNROLL_N).S
+ZTRMMKERNEL    =  ztrmm_kernel_$(ZGEMM_UNROLL_M)x$(ZGEMM_UNROLL_N).S
+ifneq ($(ZGEMM_UNROLL_M), $(ZGEMM_UNROLL_N))
+ZGEMMINCOPY    =  ../generic/zgemm_ncopy_$(ZGEMM_UNROLL_M).c
+ZGEMMITCOPY    =  ../generic/zgemm_tcopy_$(ZGEMM_UNROLL_M).c
+ZGEMMINCOPYOBJ =  zgemm_incopy$(TSUFFIX).$(SUFFIX)
+ZGEMMITCOPYOBJ =  zgemm_itcopy$(TSUFFIX).$(SUFFIX)
+endif
+ZGEMMONCOPY    =  ../generic/zgemm_ncopy_$(ZGEMM_UNROLL_N).c
+ZGEMMOTCOPY    =  ../generic/zgemm_tcopy_$(ZGEMM_UNROLL_N).c
+ZGEMMONCOPYOBJ =  zgemm_oncopy$(TSUFFIX).$(SUFFIX)
+ZGEMMOTCOPYOBJ =  zgemm_otcopy$(TSUFFIX).$(SUFFIX)

kernel/arm64/dznrm2_thunderx2t99.c

@@ -321,7 +321,7 @@ static void nrm2_compute(BLASLONG n, FLOAT *x, BLASLONG inc_x,
 	: "cc",
 	  "memory",
 	  "x0", "x1", "x2", "x3", "x4", "x5", "x6",
-	  "d0", "d1", "d2", "d3", "d4", "d5", "d6", "d7", "d8"
+	  "d0", "d1", "d2", "d3", "d4", "d5", "d6", "d7", "d8", REGINF
 	);
 
 }

kernel/arm64/sgemm_tcopy_16.S

@@ -270,11 +270,6 @@ All rights reserved.
 	ldr	s1, [A02]
 	ldr	s2, [A03]
 	ldr	s3, [A04]
-	
-	add	A01, A01, #4
-	add	A02, A02, #4
-	add	A03, A03, #4
-	add	A04, A04, #4
 
 	stp	s0, s1, [B04]
 	add	B04, B04, #8
@@ -285,11 +280,6 @@ All rights reserved.
 	ldr	s5, [A06]
 	ldr	s6, [A07]
 	ldr	s7, [A08]
-	
-	ldr	d4, [A05], #8
-	ldr	d5, [A06], #8
-	ldr	d6, [A07], #8
-	ldr	d7, [A08], #8
 
 	stp	s4, s5, [B04]
 	add	B04, B04, #8

kernel/power/KERNEL.POWER10

@@ -169,8 +169,13 @@ ZROTKERNEL   = zrot.c
 #
 SSCALKERNEL  = sscal.c
 DSCALKERNEL  = dscal.c
+ifeq ($(C_COMPILER), PGI)
+CSCALKERNEL  = ../arm/zscal.c
+ZSCALKERNEL  = ../arm/zscal.c
+else
 CSCALKERNEL  = zscal.c
 ZSCALKERNEL  = zscal.c
+endif
 #
 SSWAPKERNEL  = sswap.c
 DSWAPKERNEL  = dswap.c
@@ -181,7 +186,7 @@ ZSWAPKERNEL  = zswap.c
 SGEMVNKERNEL = sgemv_n.c
 DGEMVNKERNEL = dgemv_n_power10.c
 CGEMVNKERNEL = cgemv_n.c
-ZGEMVNKERNEL = zgemv_n_4.c
+ZGEMVNKERNEL =  zgemv_n_power10.c
 #
 SGEMVTKERNEL = sgemv_t.c
 DGEMVTKERNEL = dgemv_t_power10.c

kernel/power/KERNEL.POWER5

@@ -54,3 +54,8 @@ ZTRSMKERNEL_LN	=  ztrsm_kernel_LN.S
 ZTRSMKERNEL_LT	=  ztrsm_kernel_LT.S
 ZTRSMKERNEL_RN	=  ztrsm_kernel_LT.S
 ZTRSMKERNEL_RT	=  ztrsm_kernel_RT.S
+
+CROTKERNEL = ../arm/zrot.c
+ZROTKERNEL = ../arm/zrot.c
+SGEMVNKERNEL = ../arm/gemv_n.c
+SGEMVTKERNEL = ../arm/gemv_t.c

kernel/power/KERNEL.POWER8

@@ -242,8 +242,13 @@ ZROTKERNEL   = zrot.c
 #
 SSCALKERNEL  = sscal.c
 DSCALKERNEL  = dscal.c
+ifeq ($(C_COMPILER), PGI)
+CSCALKERNEL  = ../arm/zscal.c
+ZSCALKERNEL  = ../arm/zscal.c
+else
 CSCALKERNEL  = zscal.c
 ZSCALKERNEL  = zscal.c
+endif
 #
 SSWAPKERNEL  = sswap.c
 DSWAPKERNEL  = dswap.c

kernel/power/KERNEL.POWER9

@@ -166,8 +166,13 @@ ZROTKERNEL   = zrot.c
 #
 SSCALKERNEL  = sscal.c
 DSCALKERNEL  = dscal.c
+ifeq ($(C_COMPILER), PGI)
+CSCALKERNEL  = ../arm/zscal.c
+ZSCALKERNEL  = ../arm/zscal.c
+else
 CSCALKERNEL  = zscal.c
 ZSCALKERNEL  = zscal.c
+endif
 #
 SSWAPKERNEL  = sswap.c
 DSWAPKERNEL  = dswap.c

kernel/power/KERNEL.PPC440

@@ -16,11 +16,11 @@ ZASUMKERNEL	= zasum_ppc440.S
 SAXPYKERNEL	=  axpy_ppc440.S
 DAXPYKERNEL	=  axpy_ppc440.S
 ifneq ($(__BYTE_ORDER__),__ORDER_BIG_ENDIAN__)
-CAXPYKERNEL	= ../arm/zaxpy.c
-ZAXPYKERNEL	= ../arm/zaxpy.c
-else
 CAXPYKERNEL	= zaxpy_ppc440.S
 ZAXPYKERNEL	= zaxpy_ppc440.S
+else
+CAXPYKERNEL	= ../arm/zaxpy.c
+ZAXPYKERNEL	= ../arm/zaxpy.c
 endif
 
 SDOTKERNEL	=  dot_ppc440.S

kernel/power/KERNEL.PPCG4

@@ -15,8 +15,13 @@ ZASUMKERNEL	= zasum_ppc440.S
 
 SAXPYKERNEL	=  axpy_ppc440.S
 DAXPYKERNEL	=  axpy_ppc440.S
+ifneq ($(__BYTE_ORDER__),__ORDER_BIG_ENDIAN__)
 CAXPYKERNEL	= zaxpy_ppc440.S
 ZAXPYKERNEL	= zaxpy_ppc440.S
+else
+CAXPYKERNEL    = ../arm/zaxpy.c
+ZAXPYKERNEL    = ../arm/zaxpy.c
+endif
 
 SDOTKERNEL	=  dot_ppc440.S
 DDOTKERNEL	=  dot_ppc440.S

kernel/power/cdot.c

@@ -28,7 +28,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #else
 
 #include "common.h"
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "cdot_microk_power10.c"
 #else
 #ifndef HAVE_KERNEL_8
@@ -120,7 +120,7 @@ OPENBLAS_COMPLEX_FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLA
 
     if ((inc_x == 1) && (inc_y == 1)) {
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
         BLASLONG n1 = n & -16;
 #else
         BLASLONG n1 = n & -8;

kernel/power/cswap.c

@@ -39,8 +39,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8)  || defined(POWER9)
 #include "cswap_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "cswap_microk_power10.c"
+#elif defined(POWER10)
+#include "cswap_microk_power8.c"
 #endif
 #endif
 

kernel/power/dasum.c

@@ -49,8 +49,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "dasum_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "dasum_microk_power10.c"
+#elif defined(POWER10)
+#include "dasum_microk_power8.c"
 #endif
 #endif
 
@@ -112,7 +114,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
 	if ( inc_x == 1 )
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 16 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 3) & 0x3;

kernel/power/dgemm_kernel_power10.c

@@ -190,10 +190,9 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT * A, FLOAT * B,
 	  __vector_quad acc0, acc1, acc2, acc3, acc4,acc5,acc6,acc7;
 	  BLASLONG l = 0;
 	  vec_t *rowA = (vec_t *) & AO[0];
-	  vec_t *rb = (vec_t *) & BO[0];
 	  __vector_pair rowB, rowB1;
-	  __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
-	  __builtin_vsx_assemble_pair (&rowB1, rb[3], rb[2]);
+	  rowB = *((__vector_pair *)((void *)&BO[0]));
+	  rowB1 = *((__vector_pair *)((void *)&BO[4]));
 	  __builtin_mma_xvf64ger (&acc0, rowB, rowA[0]);
 	  __builtin_mma_xvf64ger (&acc1, rowB1, rowA[0]);
 	  __builtin_mma_xvf64ger (&acc2, rowB, rowA[1]);
@@ -205,9 +204,8 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT * A, FLOAT * B,
 	  for (l = 1; l < temp; l++)
 	    {
 	      rowA = (vec_t *) & AO[l << 3];
-	      rb = (vec_t *) & BO[l << 3];
-	      __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
-	      __builtin_vsx_assemble_pair (&rowB1, rb[3], rb[2]);
+	      rowB = *((__vector_pair *)((void *)&BO[l << 3]));
+	      rowB1 = *((__vector_pair *)((void *)&BO[(l << 3) + 4]));
 	      __builtin_mma_xvf64gerpp (&acc0, rowB, rowA[0]);
 	      __builtin_mma_xvf64gerpp (&acc1, rowB1, rowA[0]);
 	      __builtin_mma_xvf64gerpp (&acc2, rowB, rowA[1]);
@@ -247,9 +245,8 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT * A, FLOAT * B,
 	  BLASLONG l = 0;
 	  vec_t *rowA = (vec_t *) & AO[0];
 	  __vector_pair rowB, rowB1;
-	  vec_t *rb = (vec_t *) & BO[0];
-	  __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
-	  __builtin_vsx_assemble_pair (&rowB1, rb[3], rb[2]);
+	  rowB = *((__vector_pair *)((void *)&BO[0]));
+	  rowB1 = *((__vector_pair *)((void *)&BO[4]));
 	  __builtin_mma_xvf64ger (&acc0, rowB, rowA[0]);
 	  __builtin_mma_xvf64ger (&acc1, rowB1, rowA[0]);
 	  __builtin_mma_xvf64ger (&acc2, rowB, rowA[1]);
@@ -257,9 +254,8 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT * A, FLOAT * B,
 	  for (l = 1; l < temp; l++)
 	    {
 	      rowA = (vec_t *) & AO[l << 2];
-	      rb = (vec_t *) & BO[l << 3];
-	      __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
-	      __builtin_vsx_assemble_pair (&rowB1, rb[3], rb[2]);
+	      rowB = *((__vector_pair *)((void *)&BO[l << 3]));
+	      rowB1 = *((__vector_pair *)((void *)&BO[(l << 3) + 4]));
 	      __builtin_mma_xvf64gerpp (&acc0, rowB, rowA[0]);
 	      __builtin_mma_xvf64gerpp (&acc1, rowB1, rowA[0]);
 	      __builtin_mma_xvf64gerpp (&acc2, rowB, rowA[1]);
@@ -291,17 +287,15 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT * A, FLOAT * B,
 	  BLASLONG l = 0;
 	  vec_t *rowA = (vec_t *) & AO[0];
 	  __vector_pair rowB, rowB1;
-	  vec_t *rb = (vec_t *) & BO[0];
-	  __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
-	  __builtin_vsx_assemble_pair (&rowB1, rb[3], rb[2]);
+	  rowB = *((__vector_pair *)((void *)&BO[0]));
+	  rowB1 = *((__vector_pair *)((void *)&BO[4]));
 	  __builtin_mma_xvf64ger (&acc0, rowB, rowA[0]);
 	  __builtin_mma_xvf64ger (&acc1, rowB1, rowA[0]);
 	  for (l = 1; l < temp; l++)
 	    {
 	      rowA = (vec_t *) & AO[l << 1];
-	      rb = (vec_t *) & BO[l << 3];
-	      __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
-	      __builtin_vsx_assemble_pair (&rowB1, rb[3], rb[2]);
+	      rowB = *((__vector_pair *)((void *)&BO[l << 3]));
+	      rowB1 = *((__vector_pair *)((void *)&BO[(l << 3) + 4]));
 	      __builtin_mma_xvf64gerpp (&acc0, rowB, rowA[0]);
 	      __builtin_mma_xvf64gerpp (&acc1, rowB1, rowA[0]);
 	    }
@@ -403,8 +397,7 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT * A, FLOAT * B,
 	  BLASLONG l = 0;
 	  vec_t *rowA = (vec_t *) & AO[0];
 	  __vector_pair rowB;
-	  vec_t *rb = (vec_t *) & BO[0];
-	  __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
+	  rowB = *((__vector_pair *)((void *)&BO[0]));
 	  __builtin_mma_xvf64ger (&acc0, rowB, rowA[0]);
 	  __builtin_mma_xvf64ger (&acc1, rowB, rowA[1]);
 	  __builtin_mma_xvf64ger (&acc2, rowB, rowA[2]);
@@ -412,8 +405,7 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT * A, FLOAT * B,
 	  for (l = 1; l < temp; l++)
 	    {
 	      rowA = (vec_t *) & AO[l << 3];
-	      rb = (vec_t *) & BO[l << 2];
-	      __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
+	      rowB = *((__vector_pair *)((void *)&BO[l << 2]));
 	      __builtin_mma_xvf64gerpp (&acc0, rowB, rowA[0]);
 	      __builtin_mma_xvf64gerpp (&acc1, rowB, rowA[1]);
 	      __builtin_mma_xvf64gerpp (&acc2, rowB, rowA[2]);
@@ -445,15 +437,13 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT * A, FLOAT * B,
 	  BLASLONG l = 0;
 	  vec_t *rowA = (vec_t *) & AO[0];
 	  __vector_pair rowB;
-	  vec_t *rb = (vec_t *) & BO[0];
-	  __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
+	  rowB = *((__vector_pair *)((void *)&BO[0]));
 	  __builtin_mma_xvf64ger (&acc0, rowB, rowA[0]);
 	  __builtin_mma_xvf64ger (&acc1, rowB, rowA[1]);
 	  for (l = 1; l < temp; l++)
 	    {
 	      rowA = (vec_t *) & AO[l << 2];
-	      rb = (vec_t *) & BO[l << 2];
-	      __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
+	      rowB = *((__vector_pair *)((void *)&BO[l << 2]));
 	      __builtin_mma_xvf64gerpp (&acc0, rowB, rowA[0]);
 	      __builtin_mma_xvf64gerpp (&acc1, rowB, rowA[1]);
 	    }
@@ -481,14 +471,12 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT * A, FLOAT * B,
 	  BLASLONG l = 0;
 	  vec_t *rowA = (vec_t *) & AO[0];
 	  __vector_pair rowB;
-	  vec_t *rb = (vec_t *) & BO[0];
-	  __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
+	  rowB = *((__vector_pair *)((void *)&BO[0]));
 	  __builtin_mma_xvf64ger (&acc0, rowB, rowA[0]);
 	  for (l = 1; l < temp; l++)
 	    {
 	      rowA = (vec_t *) & AO[l << 1];
-	      rb = (vec_t *) & BO[l << 2];
-	      __builtin_vsx_assemble_pair (&rowB, rb[1], rb[0]);
+	      rowB = *((__vector_pair *)((void *)&BO[l << 2]));
 	      __builtin_mma_xvf64gerpp (&acc0, rowB, rowA[0]);
 	    }
 	  SAVE_ACC (&acc0, 0);

kernel/power/drot.c

@@ -42,8 +42,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "drot_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "drot_microk_power10.c"
+#elif defined(POWER10)
+#include "drot_microk_power8.c"
 #endif
 #endif
 
@@ -117,7 +119,7 @@ int CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT
 	if ( (inc_x == 1) && (inc_y == 1) )
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 16 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)y & (uintptr_t)0x1F)) >> 3) & 0x3;

kernel/power/dscal.c

@@ -38,8 +38,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "dscal_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "dscal_microk_power10.c"
+#elif defined(POWER10)
+#include "dscal_microk_power8.c"
 #endif
 #endif
 
@@ -102,7 +104,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 		if ( da == 0.0 )
 		{		
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 			if ( n >= 16 )
 			{
 				BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 3) & 0x3;
@@ -136,7 +138,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 		else
 		{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 			if ( n >= 16 )
 			{
 				BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 3) & 0x3;

kernel/power/dswap.c

@@ -38,8 +38,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "dswap_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "swap_microk_power10.c"
+#elif defined(POWER10)
+#include "dswap_microk_power8.c"
 #endif
 #endif
 
@@ -117,7 +119,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT dummy3, FLOAT *x,
 	if ( (inc_x == 1) && (inc_y == 1 ))
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 32 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)y & (uintptr_t)0x1F)) >> 3) & 0x3;

kernel/power/gemv_n.S

@@ -159,6 +159,11 @@
 #define PREFETCHSIZE_C  16
 #endif
 
+#ifdef POWER3
+#define PREFETCHSIZE_A  16
+#define PREFETCHSIZE_C  16
+#endif
+
 #ifdef POWER4
 #define PREFETCHSIZE_A  16
 #define PREFETCHSIZE_C  16

kernel/power/gemv_t.S

@@ -124,6 +124,11 @@
 #define PREFETCHSIZE_C  16
 #endif
 
+#ifdef POWER3
+#define PREFETCHSIZE_A  16
+#define PREFETCHSIZE_C  16
+#endif
+
 #ifdef POWER4
 #define PREFETCHSIZE_A  48
 #define PREFETCHSIZE_C  16

kernel/power/sasum.c

@@ -49,8 +49,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "sasum_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "sasum_microk_power10.c"
+#elif defined(POWER10)
+#include "sasum_microk_power8.c"
 #endif
 #endif
 
@@ -112,7 +114,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
 	if ( inc_x == 1 )
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 32 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 2) & 0x7;

kernel/power/sbgemm_kernel_power10.c

@@ -49,17 +49,11 @@ typedef __vector unsigned char  vec_t;
 typedef FLOAT v4sf_t __attribute__ ((vector_size (16)));
 typedef FLOAT v2sf_t __attribute__ ((vector_size (8)));
 
-vector char mask =
-  { 0x0, 0x1, 0x8, 0x9, 0x2, 0x3, 0xa, 0xb, 0x4, 0x5, 0xc, 0xd, 0x6, 0x7, 0xe,
-  0xf
-};
-
 /* 
  * BFLOAT16 xvbf16ger2pp instruction needs 4×2 matrix of
  * bfloat16 floating-point values as input. Hence this
  * merging is needed on A and B matrices. 
  */
-#define MERGE_ROW(x) vec_perm(x, x, mask)
 #define MERGE_HIGH(x, y) (vec_t) vec_mergeh ((vector short)x, (vector short)y)
 #define MERGE_LOW(x, y) (vec_t) vec_mergel ((vector short)x, (vector short)y)
 
@@ -104,6 +98,30 @@ vector char mask =
 	  rowC = (v2sf_t *) &CO[7* ldc+J]; \
           rowC[0] += result[6] * alpha;
 
+ #define  SAVE4x2_ACC_SCALAR(ACC) {                             \
+           __builtin_mma_disassemble_acc ((void *)result, ACC); \
+           res[0] = result[0] * alpha;                          \
+           res[1] = result[1] * alpha;                          \
+           res[2] = result[2] * alpha;                          \
+           res[3] = result[3] * alpha;                          \
+           CO[0 * ldc] += res[0][0];                            \
+           CO[1 * ldc] += res[1][0];                            \
+           CO[2 * ldc] += res[2][0];                            \
+           CO[3 * ldc] += res[3][0];                            \
+ }
+
+ #define  SAVE4x2_ACC1_SCALAR(ACC) {                            \
+           __builtin_mma_disassemble_acc ((void *)result, ACC); \
+           res[0] = result[0] * alpha;                          \
+           res[1] = result[1] * alpha;                          \
+           res[2] = result[2] * alpha;                          \
+           res[3] = result[3] * alpha;                          \
+           CO[4 * ldc] += res[0][0];                            \
+           CO[5 * ldc] += res[1][0];                            \
+           CO[6 * ldc] += res[2][0];                            \
+           CO[7 * ldc] += res[3][0];                            \
+}
+
 #define MMA __builtin_mma_xvbf16ger2pp
 
 #define  SAVE2x4_ACC(ACC, J)  \
@@ -179,8 +197,8 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 4;
 	      vec_t *rowA = (vec_t *) & (AO[l << 1]);
 	      vec_t *rowB = (vec_t *) & (BO[l]);
-	      vec_t rowB_h = MERGE_HIGH (rowB[0], rowB[1]);
-	      vec_t rowB_l = MERGE_LOW (rowB[0], rowB[1]);
+	      vec_t rowB_h = MERGE_HIGH (rowB[0], vzero);
+	      vec_t rowB_l = MERGE_LOW (rowB[0], vzero);
 	      vec_t rowA_h = MERGE_HIGH (rowA[0], vzero);
 	      vec_t rowA_l = MERGE_LOW (rowA[0], vzero);
 	      vec_t rowA2_h = MERGE_HIGH (rowA[1], vzero);
@@ -231,8 +249,8 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 4;
 	      vec_t *rowA = (vec_t *) & (AO[l]);
 	      vec_t *rowB = (vec_t *) & (BO[l]);
-	      vec_t rowB_h = MERGE_HIGH (rowB[0], rowB[1]);
-	      vec_t rowB_l = MERGE_LOW (rowB[0], rowB[1]);
+	      vec_t rowB_h = MERGE_HIGH (rowB[0], vzero);
+	      vec_t rowB_l = MERGE_LOW (rowB[0], vzero);
 	      vec_t rowA_h = MERGE_HIGH (rowA[0], vzero);
 	      vec_t rowA_l = MERGE_LOW (rowA[0], vzero);
 	      MMA (&acc0, rowB_h, rowA_h);
@@ -271,8 +289,8 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 	      vector short rowA =
 		{ AO[l + 0], 0, AO[l + 1], 0, AO[l + 2], 0, AO[l + 3], 0 };
 	      vec_t *rowB = (vec_t *) & (BO[l << 1]);
-	      MMA (&acc0, MERGE_HIGH (rowB[0], rowB[1]), (vec_t) rowA);
-	      MMA (&acc1, MERGE_LOW (rowB[0], rowB[1]), (vec_t) rowA);
+	      MMA (&acc0, MERGE_HIGH (rowB[0], vzero), (vec_t) rowA);
+	      MMA (&acc1, MERGE_LOW (rowB[0], vzero), (vec_t) rowA);
 	    }
 	  SAVE_ACC (&acc0, 0);
 	  SAVE_ACC1 (&acc1, 0);
@@ -306,8 +324,8 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 2;
 	      vector short rowA = { AO[l + 0], 0, AO[l + 1], 0, 0, 0, 0, 0 };
 	      vec_t *rowB = (vec_t *) & (BO[(l << 2)]);
-	      MMA (&acc0, MERGE_HIGH (rowB[0], rowB[1]), (vec_t) rowA);
-	      MMA (&acc1, MERGE_LOW (rowB[0], rowB[1]), (vec_t) rowA);
+	      MMA (&acc0, MERGE_HIGH (rowB[0], vzero), (vec_t) rowA);
+	      MMA (&acc1, MERGE_LOW (rowB[0], vzero), (vec_t) rowA);
 	    }
 	  SAVE4x2_ACC (&acc0, 0);
 	  SAVE4x2_ACC1 (&acc1, 0);
@@ -319,7 +337,7 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 	{
 	  IFLOAT *BO = B;
 	  v2sf_t *rowC;
-	  v2sf_t result[8];
+	  v4sf_t result[4], res[4];
 	  __vector_quad acc0, acc1;
 	  __builtin_mma_xxsetaccz (&acc0);
 	  __builtin_mma_xxsetaccz (&acc1);
@@ -338,11 +356,11 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 1;
 	      vector short rowA = { AO[l], 0, 0, 0, 0, 0, 0, 0 };
 	      vec_t *rowB = (vec_t *) & (BO[(l << 3)]);
-	      MMA (&acc0, MERGE_HIGH (rowB[0], rowB[1]), (vec_t) rowA);
-	      MMA (&acc1, MERGE_LOW (rowB[0], rowB[1]), (vec_t) rowA);
+	      MMA (&acc0, MERGE_HIGH (rowB[0], vzero), (vec_t) rowA);
+	      MMA (&acc1, MERGE_LOW (rowB[0], vzero), (vec_t) rowA);
 	    }
-	  SAVE4x2_ACC (&acc0, 0);
-	  SAVE4x2_ACC1 (&acc1, 0);
+	  SAVE4x2_ACC_SCALAR (&acc0);
+	  SAVE4x2_ACC1_SCALAR (&acc1);
 	  CO += 1;
 	  AO += k;
 	  BO += (k << 3);
@@ -387,16 +405,16 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 3;
 	      vec_t *rowA = (vec_t *) & (AO[(l << 2)]);
 	      vec_t *rowA1 = (vec_t *) & (A1[(l << 2)]);
-	      vec_t *rowB = (vec_t *) & (BO[l]);
-	      vec_t rowB_mrg = MERGE_ROW (rowB[0]);
-	      MMA (&acc0, rowB_mrg, MERGE_HIGH (rowA[0], vzero));
-	      MMA (&acc1, rowB_mrg, MERGE_LOW (rowA[0], vzero));
-	      MMA (&acc2, rowB_mrg, MERGE_HIGH (rowA[1], vzero));
-	      MMA (&acc3, rowB_mrg, MERGE_LOW (rowA[1], vzero));
-	      MMA (&acc4, rowB_mrg, MERGE_HIGH (rowA1[0], vzero));
-	      MMA (&acc5, rowB_mrg, MERGE_LOW (rowA1[0], vzero));
-	      MMA (&acc6, rowB_mrg, MERGE_HIGH (rowA1[1], vzero));
-	      MMA (&acc7, rowB_mrg, MERGE_LOW (rowA1[1], vzero));
+	      vector short rowB_mrg =
+		{ BO[l], 0, BO[l + 1], 0, BO[l + 2], 0, BO[l + 3], 0 };
+	      MMA (&acc0, (vec_t)rowB_mrg, MERGE_HIGH (rowA[0], vzero));
+	      MMA (&acc1, (vec_t)rowB_mrg, MERGE_LOW (rowA[0], vzero));
+	      MMA (&acc2, (vec_t)rowB_mrg, MERGE_HIGH (rowA[1], vzero));
+	      MMA (&acc3, (vec_t)rowB_mrg, MERGE_LOW (rowA[1], vzero));
+	      MMA (&acc4, (vec_t)rowB_mrg, MERGE_HIGH (rowA1[0], vzero));
+	      MMA (&acc5, (vec_t)rowB_mrg, MERGE_LOW (rowA1[0], vzero));
+	      MMA (&acc6, (vec_t)rowB_mrg, MERGE_HIGH (rowA1[1], vzero));
+	      MMA (&acc7, (vec_t)rowB_mrg, MERGE_LOW (rowA1[1], vzero));
 	    }
 
 	  SAVE_ACC (&acc0, 0);
@@ -436,12 +454,12 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 	      if (k > 1)
 		l = (k / 2) << 3;
 	      vec_t *rowA = (vec_t *) & (AO[(l << 2)]);
-	      vec_t *rowB = (vec_t *) & (BO[l]);
-	      vec_t rowB_mrg = MERGE_ROW (rowB[0]);
-	      MMA (&acc0, rowB_mrg, MERGE_HIGH (rowA[0], vzero));
-	      MMA (&acc1, rowB_mrg, MERGE_LOW (rowA[0], vzero));
-	      MMA (&acc2, rowB_mrg, MERGE_HIGH (rowA[1], vzero));
-	      MMA (&acc3, rowB_mrg, MERGE_LOW (rowA[1], vzero));
+	      vector short rowB_mrg =
+		{ BO[l], 0, BO[l + 1], 0, BO[l + 2], 0, BO[l + 3], 0 };
+	      MMA (&acc0, (vec_t)rowB_mrg, MERGE_HIGH (rowA[0], vzero));
+	      MMA (&acc1, (vec_t)rowB_mrg, MERGE_LOW (rowA[0], vzero));
+	      MMA (&acc2, (vec_t)rowB_mrg, MERGE_HIGH (rowA[1], vzero));
+	      MMA (&acc3, (vec_t)rowB_mrg, MERGE_LOW (rowA[1], vzero));
 	    }
 
 	  SAVE_ACC (&acc0, 0);
@@ -475,9 +493,10 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 3;
 	      vec_t *rowA = (vec_t *) & (AO[l << 1]);
 	      vec_t *rowB = (vec_t *) & (BO[l]);
-	      vec_t rowB_mrg = MERGE_ROW (rowB[0]);
-	      MMA (&acc0, rowB_mrg, MERGE_HIGH (rowA[0], vzero));
-	      MMA (&acc1, rowB_mrg, MERGE_LOW (rowA[0], vzero));
+	      vector short rowB_mrg =
+		{ BO[l], 0, BO[l + 1], 0, BO[l + 2], 0, BO[l + 3], 0 };
+	      MMA (&acc0, (vec_t)rowB_mrg, MERGE_HIGH (rowA[0], vzero));
+	      MMA (&acc1, (vec_t)rowB_mrg, MERGE_LOW (rowA[0], vzero));
 	    }
 	  SAVE_ACC (&acc0, 0);
 	  SAVE_ACC (&acc1, 4);
@@ -505,8 +524,9 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 3;
 	      vector short rowA =
 		{ AO[l], 0, AO[l + 1], 0, AO[l + 2], 0, AO[l + 3], 0 };
-	      vec_t *rowB = (vec_t *) & (BO[l]);
-	      MMA (&acc0, MERGE_ROW (rowB[0]), (vec_t) rowA);
+	      vector short rowB_mrg =
+		{ BO[l], 0, BO[l + 1], 0, BO[l + 2], 0, BO[l + 3], 0 };
+	      MMA (&acc0, (vec_t)(rowB_mrg), (vec_t) rowA);
 	    }
 	  SAVE_ACC (&acc0, 0);
 	  CO += 4;
@@ -536,8 +556,11 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 	      if (k > 1)
 		l = (k / 2) << 2;
 	      vector short rowA = { AO[l], 0, AO[l + 1], 0, 0, 0, 0, 0 };
-	      vec_t *rowB = (vec_t *) & (BO[l << 1]);
-	      MMA (&acc0, MERGE_ROW (rowB[0]), (vec_t) rowA);
+	      vector short rowB_mrg =
+		{ BO[(l<<1)], 0, BO[(l<<1) + 1], 0, BO[(l<<1) + 2], 0,
+		BO[(l<<1) + 3], 0
+	      };
+	      MMA (&acc0, (vec_t)(rowB_mrg), (vec_t) rowA);
 	    }
 	  SAVE4x2_ACC (&acc0, 0);
 	  CO += 2;
@@ -548,7 +571,7 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 	{
 	  IFLOAT *BO = B;
 	  v2sf_t *rowC;
-	  v2sf_t result[8];
+	  v4sf_t result[4], res[4];
 	  __vector_quad acc0;
 	  BLASLONG l = 0;
 	  __builtin_mma_xxsetaccz (&acc0);
@@ -566,10 +589,13 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 	      if (k > 1)
 		l = (k / 2) << 1;
 	      vector short rowA = { AO[l], 0, 0, 0, 0, 0, 0, 0 };
-	      vec_t *rowB = (vec_t *) & (BO[l << 2]);
-	      MMA (&acc0, MERGE_ROW (rowB[0]), (vec_t) rowA);
+	      vector short rowB_mrg =
+		{ BO[(l<<2) + 0], 0, BO[(l<<2) + 1], 0, BO[(l <<2) + 2], 0,
+		BO[(l<<2) + 3], 0
+	      };
+	      MMA (&acc0, (vec_t)(rowB_mrg), (vec_t) rowA);
 	    }
-	  SAVE4x2_ACC (&acc0, 0);
+	  SAVE4x2_ACC_SCALAR (&acc0);
 	  AO += k;
 	  BO += (k << 2);
 	  CO += 1;
@@ -620,14 +646,14 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 	      vector short rowB = { BO[l + 0], 0, BO[l + 1], 0, 0, 0, 0, 0 };
 	      vec_t *rowA = (vec_t *) & (AO[l << 3]);
 	      vec_t *rowA1 = (vec_t *) & (A1[l << 3]);
-	      MMA (&acc0, (vec_t) rowB, MERGE_HIGH (rowA[0], rowA[2]));
-	      MMA (&acc1, (vec_t) rowB, MERGE_LOW (rowA[0], rowA[2]));
-	      MMA (&acc2, (vec_t) rowB, MERGE_HIGH (rowA[1], rowA[3]));
-	      MMA (&acc3, (vec_t) rowB, MERGE_LOW (rowA[1], rowA[3]));
-	      MMA (&acc4, (vec_t) rowB, MERGE_HIGH (rowA1[0], rowA1[2]));
-	      MMA (&acc5, (vec_t) rowB, MERGE_LOW (rowA1[0], rowA1[2]));
-	      MMA (&acc6, (vec_t) rowB, MERGE_HIGH (rowA1[1], rowA1[3]));
-	      MMA (&acc7, (vec_t) rowB, MERGE_LOW (rowA1[1], rowA1[3]));
+	      MMA (&acc0, (vec_t) rowB, MERGE_HIGH (rowA[0], vzero));
+	      MMA (&acc1, (vec_t) rowB, MERGE_LOW (rowA[0], vzero));
+	      MMA (&acc2, (vec_t) rowB, MERGE_HIGH (rowA[1], vzero));
+	      MMA (&acc3, (vec_t) rowB, MERGE_LOW (rowA[1], vzero));
+	      MMA (&acc4, (vec_t) rowB, MERGE_HIGH (rowA1[0], vzero));
+	      MMA (&acc5, (vec_t) rowB, MERGE_LOW (rowA1[0], vzero));
+	      MMA (&acc6, (vec_t) rowB, MERGE_HIGH (rowA1[1], vzero));
+	      MMA (&acc7, (vec_t) rowB, MERGE_LOW (rowA1[1], vzero));
 	    }
 	  SAVE2x4_ACC (&acc0, 0);
 	  SAVE2x4_ACC (&acc1, 4);
@@ -669,10 +695,10 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 2;
 	      vector short rowB = { BO[l + 0], 0, BO[l + 1], 0, 0, 0, 0, 0 };
 	      vec_t *rowA = (vec_t *) & (AO[l << 3]);
-	      MMA (&acc0, (vec_t) rowB, MERGE_HIGH (rowA[0], rowA[2]));
-	      MMA (&acc1, (vec_t) rowB, MERGE_LOW (rowA[0], rowA[2]));
-	      MMA (&acc2, (vec_t) rowB, MERGE_HIGH (rowA[1], rowA[3]));
-	      MMA (&acc3, (vec_t) rowB, MERGE_LOW (rowA[1], rowA[3]));
+	      MMA (&acc0, (vec_t) rowB, MERGE_HIGH (rowA[0], vzero ));
+	      MMA (&acc1, (vec_t) rowB, MERGE_LOW (rowA[0], vzero));
+	      MMA (&acc2, (vec_t) rowB, MERGE_HIGH (rowA[1], vzero));
+	      MMA (&acc3, (vec_t) rowB, MERGE_LOW (rowA[1], vzero));
 	    }
 	  SAVE2x4_ACC (&acc0, 0);
 	  SAVE2x4_ACC (&acc1, 4);
@@ -708,8 +734,8 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 2;
 	      vector short rowB = { BO[l + 0], 0, BO[l + 1], 0, 0, 0, 0, 0 };
 	      vec_t *rowA = (vec_t *) & (AO[(l << 2)]);
-	      MMA (&acc0, (vec_t) rowB, MERGE_HIGH (rowA[0], rowA[1]));
-	      MMA (&acc1, (vec_t) rowB, MERGE_LOW (rowA[0], rowA[1]));
+	      MMA (&acc0, (vec_t) rowB, MERGE_HIGH (rowA[0], vzero));
+	      MMA (&acc1, (vec_t) rowB, MERGE_LOW (rowA[0], vzero));
 	    }
 	  SAVE2x4_ACC (&acc0, 0);
 	  SAVE2x4_ACC (&acc1, 4);
@@ -740,8 +766,10 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 	      if (k > 1)
 		l = (k / 2) << 2;
 	      vector short rowB = { BO[l + 0], 0, BO[l + 1], 0, 0, 0, 0, 0 };
-	      vec_t *rowA = (vec_t *) & (AO[l << 1]);
-	      MMA (&acc0, (vec_t) rowB, MERGE_ROW (rowA[0]));
+	      vector short rowA =
+	        { AO[(l << 1)], 0, AO[(l << 1) + 1] , 0 , AO[(l<<1) + 2],
+	        0, AO[(l << 1) + 3], 0 };
+	      MMA (&acc0, (vec_t) rowB, (vec_t)(rowA));
 	    }
 	  SAVE2x4_ACC (&acc0, 0);
 	  CO += 4;
@@ -829,10 +857,10 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 1;
 	      vector short rowB = { BO[l], 0, 0, 0, 0, 0, 0, 0 };
 	      vec_t *rowA = (vec_t *) & (AO[(l << 4)]);
-	      MMA (&acc0, (vec_t) rowB, MERGE_HIGH (rowA[0], rowA[2]));
-	      MMA (&acc1, (vec_t) rowB, MERGE_LOW (rowA[0], rowA[2]));
-	      MMA (&acc2, (vec_t) rowB, MERGE_HIGH (rowA[1], rowA[3]));
-	      MMA (&acc3, (vec_t) rowB, MERGE_LOW (rowA[1], rowA[3]));
+	      MMA (&acc0, (vec_t) rowB, MERGE_HIGH (rowA[0], vzero));
+	      MMA (&acc1, (vec_t) rowB, MERGE_LOW (rowA[0], vzero));
+	      MMA (&acc2, (vec_t) rowB, MERGE_HIGH (rowA[1], vzero));
+	      MMA (&acc3, (vec_t) rowB, MERGE_LOW (rowA[1], vzero));
 	    }
 	  rowC = (v4sf_t *) &CO[0];
 	  __builtin_mma_disassemble_acc ((void *)result, &acc0);
@@ -871,8 +899,8 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 		l = (k / 2) << 1;
 	      vector short rowB = { BO[l], 0, 0, 0, 0, 0, 0, 0 };
 	      vec_t *rowA = (vec_t *) & (AO[(l << 3)]);
-	      MMA (&acc0, (vec_t) rowB, MERGE_HIGH (rowA[0], rowA[1]));
-	      MMA (&acc1, (vec_t) rowB, MERGE_LOW (rowA[0], rowA[1]));
+	      MMA (&acc0, (vec_t) rowB, MERGE_HIGH (rowA[0], vzero));
+	      MMA (&acc1, (vec_t) rowB, MERGE_LOW (rowA[0], vzero));
 	    }
 	  rowC = (v4sf_t *) &CO[0];
 	  __builtin_mma_disassemble_acc ((void *)result, &acc0);
@@ -904,8 +932,10 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, IFLOAT * A,
 	      if (k > 1)
 		l = (k / 2) << 1;
 	      vector short rowB = { BO[l], 0, 0, 0, 0, 0, 0, 0 };
-	      vec_t *rowA = (vec_t *) & (AO[(l << 2)]);
-	      MMA (&acc0, (vec_t) rowB, MERGE_ROW (rowA[0]));
+	      vector short rowA =
+	        { AO[(l << 2)], 0, AO[(l << 2) + 1] , 0 ,
+		AO[(l << 2) + 2], 0, AO[(l << 2) + 3], 0 };
+	      MMA (&acc0, (vec_t) rowB, (vec_t)(rowA));
 	    }
 	  rowC = (v4sf_t *) &CO[0];
 	  __builtin_mma_disassemble_acc ((void *)result, &acc0);

kernel/power/srot.c

@@ -42,8 +42,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "srot_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "srot_microk_power10.c"
+#elif defined(POWER10)
+#include "srot_microk_power8.c"
 #endif
 #endif
 
@@ -117,7 +119,7 @@ int CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT
 	if ( (inc_x == 1) && (inc_y == 1) )
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 16 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)y & (uintptr_t)0x1F)) >> 2) & 0x7;

kernel/power/sscal.c

@@ -38,8 +38,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "sscal_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "sscal_microk_power10.c"
+#elif defined(POWER10)
+#include "sscal_microk_power8.c"
 #endif
 #endif
 
@@ -104,7 +106,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 		if ( da == 0.0 )
 		{		
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 			if ( n >= 32 )
 			{
 				BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 2) & 0x7;
@@ -138,7 +140,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 		else
 		{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 			if ( n >= 32 )
 			{
 				BLASLONG align = ((32 - ((uintptr_t)x & (uintptr_t)0x1F)) >> 2) & 0x7;

kernel/power/sswap.c

@@ -38,8 +38,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "sswap_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "swap_microk_power10.c"
+#elif defined(POWER10)
+#include "sswap_microk_power8.c"
 #endif
 #endif
 
@@ -117,7 +119,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT dummy3, FLOAT *x,
 	if ( (inc_x == 1) && (inc_y == 1 ))
 	{
 
-#if defined(POWER10)
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 		if ( n >= 64 )
 		{
 			BLASLONG align = ((32 - ((uintptr_t)y & (uintptr_t)0x1F)) >> 2) & 0x7;

kernel/power/zgemv_n.S

@@ -155,6 +155,11 @@
 #define PREFETCHSIZE_C  16
 #endif
 
+#ifdef POWER3
+#define PREFETCHSIZE_A  34
+#define PREFETCHSIZE_C  16
+#endif
+
 #ifdef POWER4
 #define PREFETCHSIZE_A  34
 #define PREFETCHSIZE_C  16

kernel/power/zgemv_t.S

@@ -129,6 +129,11 @@
 #define PREFETCHSIZE_C  16
 #endif
 
+#ifdef POWER3
+#define PREFETCHSIZE_A  34
+#define PREFETCHSIZE_C  16
+#endif
+
 #ifdef POWER4
 #define PREFETCHSIZE_A  34
 #define PREFETCHSIZE_C  16

kernel/power/zgemv_t_4.c

@@ -43,6 +43,134 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #elif HAVE_KERNEL_4x4_VEC
 
+#if defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
+typedef __vector unsigned char  vec_t;
+typedef FLOAT v4sf_t __attribute__ ((vector_size (16)));
+
+
+static void zgemv_kernel_4x4(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOAT *y, FLOAT alpha_r, FLOAT alpha_i) {
+    BLASLONG i;
+    FLOAT *a0, *a1, *a2, *a3;
+    a0 = ap;
+    a1 = ap + lda;
+    a2 = a1 + lda;
+    a3 = a2 + lda;
+    __vector_quad acc0, acc1, acc2, acc3;;
+    __vector_quad acc4, acc5, acc6, acc7;
+    v4sf_t result[4];
+    __vector_pair *Va0, *Va1, *Va2, *Va3;
+    i = 0;
+    n = n << 1;
+    __builtin_mma_xxsetaccz (&acc0);
+    __builtin_mma_xxsetaccz (&acc1);
+    __builtin_mma_xxsetaccz (&acc2);
+    __builtin_mma_xxsetaccz (&acc3);
+    __builtin_mma_xxsetaccz (&acc4);
+    __builtin_mma_xxsetaccz (&acc5);
+    __builtin_mma_xxsetaccz (&acc6);
+    __builtin_mma_xxsetaccz (&acc7);
+    while (i < n) {
+
+	vec_t *rx = (vec_t *) & x[i];
+        Va0  = ((__vector_pair*)((void*)&a0[i]));
+        Va1  = ((__vector_pair*)((void*)&a1[i]));
+        Va2  = ((__vector_pair*)((void*)&a2[i]));
+        Va3  = ((__vector_pair*)((void*)&a3[i]));
+
+        __builtin_mma_xvf64gerpp (&acc0, Va0[0], rx[0]);
+        __builtin_mma_xvf64gerpp (&acc1, Va1[0], rx[0]);
+        __builtin_mma_xvf64gerpp (&acc2, Va2[0], rx[0]);
+        __builtin_mma_xvf64gerpp (&acc3, Va3[0], rx[0]);
+        __builtin_mma_xvf64gerpp (&acc4, Va0[0], rx[1]);
+        __builtin_mma_xvf64gerpp (&acc5, Va1[0], rx[1]);
+        __builtin_mma_xvf64gerpp (&acc6, Va2[0], rx[1]);
+        __builtin_mma_xvf64gerpp (&acc7, Va3[0], rx[1]);
+        __builtin_mma_xvf64gerpp (&acc0, Va0[1], rx[2]);
+        __builtin_mma_xvf64gerpp (&acc1, Va1[1], rx[2]);
+        __builtin_mma_xvf64gerpp (&acc2, Va2[1], rx[2]);
+        __builtin_mma_xvf64gerpp (&acc3, Va3[1], rx[2]);
+        __builtin_mma_xvf64gerpp (&acc4, Va0[1], rx[3]);
+        __builtin_mma_xvf64gerpp (&acc5, Va1[1], rx[3]);
+        __builtin_mma_xvf64gerpp (&acc6, Va2[1], rx[3]);
+        __builtin_mma_xvf64gerpp (&acc7, Va3[1], rx[3]);
+        i += 8;
+
+    }
+#if ( !defined(CONJ) && !defined(XCONJ) ) || ( defined(CONJ) && defined(XCONJ) )
+    __builtin_mma_disassemble_acc ((void *)result, &acc0);
+    register FLOAT temp_r0 = result[0][0] - result[1][1];
+    register FLOAT temp_i0 = result[0][1] + result[1][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc4);
+    temp_r0 += result[2][0] - result[3][1];
+    temp_i0 += result[2][1] + result[3][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc1);
+    register FLOAT temp_r1 = result[0][0] - result[1][1];
+    register FLOAT temp_i1 = result[0][1] + result[1][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc5);
+    temp_r1 += result[2][0] - result[3][1];
+    temp_i1 += result[2][1] + result[3][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc2);
+    register FLOAT temp_r2 = result[0][0] - result[1][1];
+    register FLOAT temp_i2 = result[0][1] + result[1][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc6);
+    temp_r2 += result[2][0] - result[3][1];
+    temp_i2 += result[2][1] + result[3][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc3);
+    register FLOAT temp_r3 = result[0][0] - result[1][1];
+    register FLOAT temp_i3 = result[0][1] + result[1][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc7);
+    temp_r3 += result[2][0] - result[3][1];
+    temp_i3 += result[2][1] + result[3][0];
+#else
+    __builtin_mma_disassemble_acc ((void *)result, &acc0);
+    register FLOAT temp_r0 = result[0][0] + result[1][1];
+    register FLOAT temp_i0 = result[0][1] - result[1][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc4);
+    temp_r0 += result[2][0] + result[3][1];
+    temp_i0 += result[2][1] - result[3][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc1);
+    register FLOAT temp_r1 = result[0][0] + result[1][1];
+    register FLOAT temp_i1 = result[0][1] - result[1][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc5);
+    temp_r1 += result[2][0] + result[3][1];
+    temp_i1 += result[2][1] - result[3][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc2);
+    register FLOAT temp_r2 = result[0][0] + result[1][1];
+    register FLOAT temp_i2 = result[0][1] - result[1][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc6);
+    temp_r2 += result[2][0] + result[3][1];
+    temp_i2 += result[2][1] - result[3][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc3);
+    register FLOAT temp_r3 = result[0][0] + result[1][1];
+    register FLOAT temp_i3 = result[0][1] - result[1][0];
+    __builtin_mma_disassemble_acc ((void *)result, &acc7);
+    temp_r3 += result[2][0] + result[3][1];
+    temp_i3 += result[2][1] - result[3][0];
+#endif
+#if !defined(XCONJ)
+
+    y[0] += alpha_r * temp_r0 - alpha_i * temp_i0;
+    y[1] += alpha_r * temp_i0 + alpha_i * temp_r0;
+    y[2] += alpha_r * temp_r1 - alpha_i * temp_i1;
+    y[3] += alpha_r * temp_i1 + alpha_i * temp_r1;
+    y[4] += alpha_r * temp_r2 - alpha_i * temp_i2;
+    y[5] += alpha_r * temp_i2 + alpha_i * temp_r2;
+    y[6] += alpha_r * temp_r3 - alpha_i * temp_i3;
+    y[7] += alpha_r * temp_i3 + alpha_i * temp_r3;
+
+#else
+
+    y[0] += alpha_r * temp_r0 + alpha_i * temp_i0;
+    y[1] -= alpha_r * temp_i0 - alpha_i * temp_r0;
+    y[2] += alpha_r * temp_r1 + alpha_i * temp_i1;
+    y[3] -= alpha_r * temp_i1 - alpha_i * temp_r1;
+    y[4] += alpha_r * temp_r2 + alpha_i * temp_i2;
+    y[5] -= alpha_r * temp_i2 - alpha_i * temp_r2;
+    y[6] += alpha_r * temp_r3 + alpha_i * temp_i3;
+    y[7] -= alpha_r * temp_i3 - alpha_i * temp_r3;
+#endif
+}
+#else
 static void zgemv_kernel_4x4(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOAT *y, FLOAT alpha_r, FLOAT alpha_i) {
     BLASLONG i;
     FLOAT *a0, *a1, *a2, *a3;
@@ -198,6 +326,7 @@ static void zgemv_kernel_4x4(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOA
 #endif
 }
 
+#endif
 #else
 
 static void zgemv_kernel_4x4(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOAT *y, FLOAT alpha_r, FLOAT alpha_i) {

kernel/power/zscal.c

@@ -43,12 +43,16 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(DOUBLE)
 #include "zscal_microk_power8.c"
 #endif
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #if defined(DOUBLE)
 #include "zscal_microk_power10.c"
 #else
 #include "cscal_microk_power10.c"
 #endif
+#elif defined(POWER10)
+#if defined(DOUBLE)
+#include "zscal_microk_power8.c"
+#endif
 #endif
 #endif
 

kernel/power/zswap.c

@@ -39,8 +39,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #if defined(__VEC__) || defined(__ALTIVEC__)
 #if defined(POWER8) || defined(POWER9)
 #include "zswap_microk_power8.c"
-#elif defined(POWER10)
+#elif defined(POWER10) && (__BYTE_ORDER__ != __ORDER_BIG_ENDIAN__)
 #include "cswap_microk_power10.c"
+#elif defined(POWER10)
+#include "zswap_microk_power8.c"
 #endif
 #endif