rebase? (#1)

* With the Intel compiler on Linux, prefer ifort for the final link step 

icc has known problems with mixed-language builds that ifort can handle just fine. Fixes #1956

* Rename operands to put lda on the input/output constraint list

* Fix wrong constraints in inline assembly

for #2009

* Fix inline assembly constraints

rework indices to allow marking argument lda4 as input and output. For #2009

* Fix inline assembly constraints

rework indices to allow marking argument lda as input and output.

* Fix inline assembly constraints

* Fix inline assembly constraints

* Fix inline assembly constraints in Bulldozer TRSM kernels

rework indices to allow marking i,as and bs as both input and output (marked operand n1 as well for simplicity). For #2009

* Correct range_n limiting

same bug as seen in #1388, somehow missed in corresponding PR #1389

* Allow multithreading TRMV again

revert workaround introduced for issue #1332 as the actual cause appears to be my incorrect fix from #1262 (see #1388)

* Fix error introduced during cleanup

* Reduce list of kernels in the dynamic arch build

to make compilation complete reliably within the 1h limit again

* init

* move fix to right place

* Fix missing -c option in AVX512 test

* Fix AVX512 test always returning false due to missing compiler option

* Make x86_32 imply NO_AVX2, NO_AVX512 in addition to NO_AVX

fixes #2033

* Keep xcode8.3 for osx BINARY=32 build

as xcode10 deprecated i386

* Make sure that AVX512 is disabled in 32bit builds

for #2033

* Improve handling of NO_STATIC and NO_SHARED

to avoid surprises from defining either as zero. Fixes #2035 by addressing some concerns from #1422

* init

* address warning introed with #1814 et al

* Restore locking optimizations for OpenMP case

restore another accidentally dropped part of #1468 that was missed in #2004 to address performance regression reported in #1461

* HiSilicon tsv110 CPUs optimization branch

add HiSilicon tsv110 CPUs  optimization branch

* add TARGET support for  HiSilicon tsv110 CPUs

* add TARGET support for HiSilicon tsv110 CPUs

* add TARGET support for HiSilicon tsv110 CPUs

* Fix module definition conflicts between LAPACK and ReLAPACK

for #2043

* Do not compile in AVX512 check if AVX support is disabled

xgetbv is function depends on NO_AVX being undefined - we could change that too, but that combo is unlikely to work anyway

* ctest.c : add __POWERPC__ for PowerMac

* Fix crash in sgemm SSE/nano kernel on x86_64

Fix bug #2047.

Signed-off-by: Celelibi <celelibi@gmail.com>

* param.h : enable defines for PPC970 on DarwinOS

fixes:
gemm.c: In function 'sgemm_':
../common_param.h:981:18: error: 'SGEMM_DEFAULT_P' undeclared (first use in this function)
 #define SGEMM_P  SGEMM_DEFAULT_P
                  ^

* common_power.h: force DCBT_ARG 0 on PPC970 Darwin

without this, we see
../kernel/power/gemv_n.S:427:Parameter syntax error
and many more similar entries

that relates to this assembly command
dcbt 8, r24, r18

this change makes the DCBT_ARG = 0
and openblas builds through to completion on PowerMac 970
Tests pass

* Make TARGET=GENERIC compatible with DYNAMIC_ARCH=1

for issue #2048

* make DYNAMIC_ARCH=1 package work on TSV110.

* make DYNAMIC_ARCH=1 package work on TSV110

* Add Intel Denverton

for #2048

* Add Intel Denverton

* Change 64-bit detection as explained in #2056

* Trivial typo fix

as suggested in #2022

* Disable the AVX512 DGEMM kernel (again)

Due to as yet unresolved errors seen in #1955 and #2029

* Use POSIX getenv on Cygwin

The Windows-native GetEnvironmentVariable cannot be relied on, as
Cygwin does not always copy environment variables set through Cygwin
to the Windows environment block, particularly after fork().

* Fix for #2063: The DllMain used in Cygwin did not run the thread memory
pool cleanup upon THREAD_DETACH which is needed when compiled with
USE_TLS=1.

* Also call CloseHandle on each thread, as well as on the event so as to not leak thread handles.

* AIX asm syntax changes needed for shared object creation

* power9 makefile. dgemm based on power8 kernel with following changes : 32x unrolled 16x4 kernel and 8x4 kernel using (lxv stxv butterfly rank1 update). improvement from 17 to 22-23gflops. dtrmm cases were added into dgemm itself

* Expose CBLAS interfaces for I?MIN and I?MAX

* Build CBLAS interfaces for I?MIN and I?MAX

* Add declarations for ?sum and cblas_?sum

* Add interface for ?sum (derived from ?asum)

* Add ?sum

* Add implementations of ssum/dsum and csum/zsum

as trivial copies of asum/zsasum with the fabs calls replaced by fmov to preserve code structure

* Add ARM implementations of ?sum

(trivial copies of the respective ?asum with the fabs calls removed)

* Add ARM64 implementations of ?sum

as trivial copies of the respective ?asum kernels with the fabs calls removed

* Add ia64 implementation of ?sum

as trivial copy of asum with the fabs calls removed

* Add MIPS implementation of ?sum

as trivial copy of ?asum with the fabs calls removed

* Add MIPS64 implementation of ?sum

as trivial copy of ?asum with the fabs replaced by mov to preserve code structure

* Add POWER implementation of ?sum

as trivial copy of ?asum with the fabs replaced by fmr to preserve code structure

* Add SPARC implementation of ?sum

as trivial copy of ?asum with the fabs replaced by fmov to preserve code structure

* Add x86 implementation of ?sum

as trivial copy of ?asum with the fabs calls removed

* Add x86_64 implementation of ?sum

as trivial copy of ?asum with the fabs calls removed

* Add ZARCH implementation of ?sum

as trivial copies of the respective ?asum kernels with the ABS and vflpsb calls removed

* Detect 32bit environment on 64bit ARM hardware

for #2056, using same approach as #2058

* Add cmake defaults for ?sum kernels

* Add ?sum

* Add ?sum definitions for generic kernel

* Add declarations for ?sum

* Add -lm and disable EXPRECISION support on *BSD

fixes #2075

* Add in runtime CPU detection for POWER.

* snprintf define consolidated to common.h

* Support INTERFACE64=1

* Add support for INTERFACE64 and fix XERBLA calls

1. Replaced all instances of "int" with "blasint"
2. Added string length as "hidden" third parameter in calls to fortran XERBLA

* Correct length of name string in xerbla call

* Avoid out-of-bounds accesses in LAPACK EIG tests

see https://github.com/Reference-LAPACK/lapack/issues/333

* Correct INFO=4 condition

* Disable reallocation of work array in xSYTRF

as it appears to cause memory management problems (seen in the LAPACK tests)

* Disable repeated recursion on Ab_BR in ReLAPACK xGBTRF

due to crashes in LAPACK tests

* sgemm/strmm

* Update Changelog with changes from 0.3.6

* Increment version to 0.3.7.dev

* Increment version to 0.3.7.dev

* Misc. typo fixes

Found via `codespell -q 3 -w -L ith,als,dum,nd,amin,nto,wis,ba -S ./relapack,./kernel,./lapack-netlib`

* Correct argument of CPU_ISSET for glibc <2.5

fixes #2104

* conflict resolve

* Revert reference/ fixes

* Revert Changelog.txt typos

* Disable the SkyLakeX DGEMMITCOPY kernel as well

as a stopgap measure for https://github.com/numpy/numpy/issues/13401 as mentioned in #1955

* Disable DGEMMINCOPY as well for now

#1955

* init

* Fix errors in cpu enumeration with glibc 2.6

for #2114

* Change two http links to https

Closes #2109

* remove redundant code #2113

* Set up CI with Azure Pipelines

[skip ci]

* TST: add native POWER8 to CI

* add native POWER8 testing to
Travis CI matrix with ppc64le
os entry

* Update link to IBM MASS library, update cpu support status

* first try migrating one of the arm builds from travis

* fix tabbing in azure commands

* Update azure-pipelines.yml

take out offending lines (although stolen from https://github.com/conda-forge/opencv-feedstock azure-pipelines fiie)

* Update azure-pipelines.yml

* Update azure-pipelines.yml

* Update azure-pipelines.yml

* Update azure-pipelines.yml

* DOC: Add Azure CI status badge

* Add ARMV6 build to azure CI setup (#2122)

using aytekinar's Alpine image and docker script from the Travis setup

[skip ci]

* TST: Azure manylinux1 & clean-up

* remove some of the steps & comments
from the original Azure yml template

* modify the trigger section to use
develop since OpenBLAS primarily uses
this branch; use the same batching
behavior as downstream projects NumPy/
SciPy

* remove Travis emulated ARMv6 gcc build
because this now happens in Azure

* use documented Ubuntu vmImage name for Azure
and add in a manylinux1 test run to the matrix

[skip appveyor]

* Add NO_AFFINITY to available options on Linux, and set it to ON

to match the gmake default. Fixes second part of #2114

* Replace ISMIN and ISAMIN kernels on all x86_64 platforms (#2125)

* Mark iamax_sse.S as unsuitable for MIN due to issue #2116
* Use iamax.S rather than iamax_sse.S for ISMIN/ISAMIN on all x86_64 as workaround for #2116

* Move ARMv8 gcc build from Travis to Azure

* Move ARMv8 gcc build from Travis to Azure

* Update .travis.yml

* Test drone CI

* install make

* remove sudo

* Install gcc

* Install perl

* Install gfortran and add a clang job

* gfortran->gcc-gfortran

* Switch to ubuntu and parallel jobs

* apt update

* Fix typo

* update yes

* no need of gcc in clang build

* Add a cmake build as well

* Add cmake builds and print options

* build without lapack on cmake

* parallel build

* See if ubuntu 19.04 fixes the ICE

* Remove qemu armv8 builds

* arm32 build

* Fix typo

* TST: add SkylakeX AVX512 CI test

* adapt the C-level reproducer code for some
recent SkylakeX AVX512 kernel issues, provided
by Isuru Fernando and modified by Martin Kroeker,
for usage in the utest suite

* add an Intel SDE SkylakeX emulation utest run to
the Azure CI matrix; a custom Docker build was required
because Ubuntu image provided by Azure does not support
AVX512VL instructions

* Add option USE_LOCKING for single-threaded build with locking support

for calling from concurrent threads

* Add option USE_LOCKING for single-threaded build with locking support

* Add option USE_LOCKING for SMP-like locking in USE_THREAD=0 builds

* Add option USE_LOCKING but keep default settings intact

* Remove unrelated change

* Do not try ancient PGI hacks with recent versions of that compiler

should fix #2139

* Build and run utests in any case, they do their own checks for fortran availability

* Add softfp support in min/max kernels

fix for #1912

* Revert "Add softfp support in min/max kernels"

* Separate implementations of AMAX and IAMAX on arm

As noted in #1912 and comment on #1942, the combined implementation happens to "do the right thing" on hardfp, but cannot return both value and index on softfp where they would have to share the return register

* Ensure correct output for DAMAX with softfp

* Use generic kernels for complex (I)AMAX to support softfp

* improved zgemm power9 based on power8

* upload thread safety test folder

* hook up c++ thread safety test (main Makefile)

*  add c++ thread test option to Makefile.rule

* Document NO_AVX512 

for #2151

* sgemm pipeline improved, zgemm rewritten without inner packs, ABI lxvx v20 fixed with vs52

* Fix detection of AVX512 capable compilers in getarch

21eda8b5 introduced a check in getarch.c to test if the compiler is capable of
AVX512. This check currently fails, since the used __AVX2__ macro is only
defined if getarch itself was compiled with AVX2/AVX512 support. Make sure this
is the case by building getarch with -march=native on x86_64. It is only
supposed to run on the build host anyway.

* c_check: Unlink correct file

* power9 zgemm ztrmm optimized

* conflict resolve

* Add gfortran workaround for ABI violations in LAPACKE

for #2154 (see gcc bug 90329)

* Add gfortran workaround for ABI violations

for #2154 (see gcc bug 90329)

* Add gfortran workaround for potential ABI violation 

for #2154

* Update fc.cmake

* Remove any inadvertent use of -march=native from DYNAMIC_ARCH builds

from #2143, -march=native precludes use of more specific options like -march=skylake-avx512 in individual kernels, and defeats the purpose of dynamic arch anyway.

* Avoid unintentional activation of TLS code via USE_TLS=0

fixes #2149

* Do not force gcc options on non-gcc compilers

fixes compile failure with pgi 18.10 as reported on OpenBLAS-users

* Update Makefile.x86_64

* Zero ecx with a mov instruction

PGI assembler does not like the initialization in the constraints.

* Fix mov syntax

* new sgemm 8x16

* Update dtrmm_kernel_16x4_power8.S

* PGI compiler does not like -march=native

* Fix build on FreeBSD/powerpc64.

Signed-off-by: Piotr Kubaj <pkubaj@anongoth.pl>

* Fix build for PPC970 on FreeBSD pt. 1

FreeBSD needs DCBT_ARG=0 as well.

* Fix build for PPC970 on FreeBSD pt.2

FreeBSD needs those macros too.

* cgemm/ctrmm power9

* Utest needs CBLAS but not necessarily FORTRAN

* Add mingw builds to Appveyor config

* Add getarch flags to disable AVX on x86

(and other small fixes to match Makefile behaviour)

* Make disabling DYNAMIC_ARCH on unsupported systems work

needs to be unset in the cache for the change to have any effect

* Mingw32 needs leading underscore on object names

(also copy BUNDERSCORE settings for FORTRAN from the corresponding Makefile)

.drone.yml

@@ -0,0 +1,143 @@
+---
+kind: pipeline
+name: arm64_gcc_make
+
+platform:
+  os: linux
+  arch: arm64
+
+steps:
+- name: Build and Test
+  image: ubuntu:19.04
+  environment:
+    CC: gcc
+    COMMON_FLAGS: 'DYNAMIC_ARCH=1 TARGET=ARMV8 NUM_THREADS=32'
+  commands:
+    - echo "MAKE_FLAGS:= $COMMON_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC gfortran perl
+    - $CC --version
+    - make QUIET_MAKE=1 $COMMON_FLAGS
+    - make -C test $COMMON_FLAGS
+    - make -C ctest $COMMON_FLAGS
+    - make -C utest $COMMON_FLAGS
+
+---
+kind: pipeline
+name: arm32_gcc_make
+
+platform:
+  os: linux
+  arch: arm
+
+steps:
+- name: Build and Test
+  image: ubuntu:19.04
+  environment:
+    CC: gcc
+    COMMON_FLAGS: 'DYNAMIC_ARCH=1 TARGET=ARMV6 NUM_THREADS=32'
+  commands:
+    - echo "MAKE_FLAGS:= $COMMON_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC gfortran perl
+    - $CC --version
+    - make QUIET_MAKE=1 $COMMON_FLAGS
+    - make -C test $COMMON_FLAGS
+    - make -C ctest $COMMON_FLAGS
+    - make -C utest $COMMON_FLAGS
+
+---
+kind: pipeline
+name: arm64_clang_make
+
+platform:
+  os: linux
+  arch: arm64
+
+steps:
+- name: Build and Test
+  image: ubuntu:18.04
+  environment:
+    CC: clang
+    COMMON_FLAGS: 'DYNAMIC_ARCH=1 TARGET=ARMV8 NUM_THREADS=32'
+  commands:
+    - echo "MAKE_FLAGS:= $COMMON_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC gfortran perl
+    - $CC --version
+    - make QUIET_MAKE=1 $COMMON_FLAGS
+    - make -C test $COMMON_FLAGS
+    - make -C ctest $COMMON_FLAGS
+    - make -C utest $COMMON_FLAGS
+
+---
+kind: pipeline
+name: arm32_clang_cmake
+
+platform:
+  os: linux
+  arch: arm
+
+steps:
+- name: Build and Test
+  image: ubuntu:18.04
+  environment:
+    CC: clang
+    CMAKE_FLAGS: '-DDYNAMIC_ARCH=1 -DTARGET=ARMV6 -DNUM_THREADS=32 -DNOFORTRAN=ON -DBUILD_WITHOUT_LAPACK=ON'
+  commands:
+    - echo "CMAKE_FLAGS:= $CMAKE_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC g++ perl cmake
+    - $CC --version
+    - mkdir build && cd build
+    - cmake $CMAKE_FLAGS ..
+    - make -j
+    - ctest
+
+---
+kind: pipeline
+name: arm64_gcc_cmake
+
+platform:
+  os: linux
+  arch: arm64
+
+steps:
+- name: Build and Test
+  image: ubuntu:18.04
+  environment:
+    CC: gcc
+    CMAKE_FLAGS: '-DDYNAMIC_ARCH=1 -DTARGET=ARMV8 -DNUM_THREADS=32 -DNOFORTRAN=ON -DBUILD_WITHOUT_LAPACK=ON'
+  commands:
+    - echo "CMAKE_FLAGS:= $CMAKE_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC g++ perl cmake
+    - $CC --version
+    - mkdir build && cd build
+    - cmake $CMAKE_FLAGS ..
+    - make -j
+    - ctest
+
+---
+kind: pipeline
+name: arm64_clang_cmake
+
+platform:
+  os: linux
+  arch: arm64
+
+steps:
+- name: Build and Test
+  image: ubuntu:18.04
+  environment:
+    CC: clang
+    CMAKE_FLAGS: '-DDYNAMIC_ARCH=1 -DTARGET=ARMV8 -DNUM_THREADS=32 -DNOFORTRAN=ON -DBUILD_WITHOUT_LAPACK=ON'
+  commands:
+    - echo "CMAKE_FLAGS:= $CMAKE_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC g++ perl cmake
+    - $CC --version
+    - mkdir build && cd build
+    - cmake $CMAKE_FLAGS ..
+    - make -j
+    - ctest

.travis.yml

@@ -25,6 +25,15 @@ matrix:
         - TARGET_BOX=LINUX64
         - BTYPE="BINARY=64"
 
+    - <<: *test-ubuntu
+      os: linux-ppc64le
+      before_script:
+        - COMMON_FLAGS="DYNAMIC_ARCH=1 TARGET=POWER8 NUM_THREADS=32"
+      env:
+        # for matrix annotation only
+        - TARGET_BOX=PPC64LE_LINUX
+        - BTYPE="BINARY=64 USE_OPENMP=1"
+
     - <<: *test-ubuntu
       env:
         - TARGET_BOX=LINUX64
@@ -160,45 +169,10 @@ matrix:
         - BTYPE="BINARY=64 INTERFACE64=1"
 
     - <<: *test-macos
+      osx_image: xcode8.3
       env:
         - BTYPE="BINARY=32"
 
-    - &emulated-arm
-      dist: trusty
-      sudo: required
-      services: docker
-      env: IMAGE_ARCH=arm32 TARGET_ARCH=ARMV6 COMPILER=gcc
-      name: "Emulated Build for ARMV6 with gcc"
-      before_install: sudo docker run --rm --privileged multiarch/qemu-user-static:register --reset
-      script: |
-        echo "FROM openblas/alpine:${IMAGE_ARCH}
-        COPY . /tmp/openblas
-        RUN mkdir /tmp/openblas/build                             &&  \
-            cd /tmp/openblas/build                                &&  \
-            CC=${COMPILER} cmake -D DYNAMIC_ARCH=OFF                  \
-                                 -D TARGET=${TARGET_ARCH}             \
-                                 -D BUILD_SHARED_LIBS=ON              \
-                                 -D BUILD_WITHOUT_LAPACK=ON           \
-                                 -D BUILD_WITHOUT_CBLAS=ON            \
-                                 -D CMAKE_BUILD_TYPE=Release ../  &&  \
-            cmake --build ." > Dockerfile
-        docker build .
-    - <<: *emulated-arm
-      env: IMAGE_ARCH=arm32 TARGET_ARCH=ARMV6 COMPILER=clang
-      name: "Emulated Build for ARMV6 with clang"
-    - <<: *emulated-arm
-      env: IMAGE_ARCH=arm64 TARGET_ARCH=ARMV8 COMPILER=gcc
-      name: "Emulated Build for ARMV8 with gcc"
-    - <<: *emulated-arm
-      env: IMAGE_ARCH=arm64 TARGET_ARCH=ARMV8 COMPILER=clang
-      name: "Emulated Build for ARMV8 with clang"
-
-  allow_failures:
-    - env: IMAGE_ARCH=arm32 TARGET_ARCH=ARMV6 COMPILER=gcc
-    - env: IMAGE_ARCH=arm32 TARGET_ARCH=ARMV6 COMPILER=clang
-    - env: IMAGE_ARCH=arm64 TARGET_ARCH=ARMV8 COMPILER=gcc
-    - env: IMAGE_ARCH=arm64 TARGET_ARCH=ARMV8 COMPILER=clang
-
 # whitelist
 branches:
   only:

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 6.dev)
+set(OpenBLAS_PATCH_VERSION 7.dev)
 set(OpenBLAS_VERSION "${OpenBLAS_MAJOR_VERSION}.${OpenBLAS_MINOR_VERSION}.${OpenBLAS_PATCH_VERSION}")
 
 # Adhere to GNU filesystem layout conventions
@@ -20,9 +20,14 @@ if(MSVC)
 option(BUILD_WITHOUT_LAPACK "Do not build LAPACK and LAPACKE (Only BLAS or CBLAS)" ON)
 endif()
 option(BUILD_WITHOUT_CBLAS "Do not build the C interface (CBLAS) to the BLAS functions" OFF)
-option(DYNAMIC_ARCH "Include support for multiple CPU targets, with automatic selection at runtime (x86/x86_64 only)" OFF)
-option(DYNAMIC_OLDER "Include specific support for older cpu models (Penryn,Dunnington,Atom,Nano,Opteron) with DYNAMIC_ARCH" OFF)
+option(DYNAMIC_ARCH "Include support for multiple CPU targets, with automatic selection at runtime (x86/x86_64, aarch64 or ppc only)" OFF)
+option(DYNAMIC_OLDER "Include specific support for older x86 cpu models (Penryn,Dunnington,Atom,Nano,Opteron) with DYNAMIC_ARCH" OFF)
 option(BUILD_RELAPACK "Build with ReLAPACK (recursive implementation of several LAPACK functions on top of standard LAPACK)" OFF)
+if(${CMAKE_SYSTEM_NAME} MATCHES "Linux")
+option(NO_AFFINITY "Disable support for CPU affinity masks to avoid binding processes from e.g. R or numpy/scipy to a single core" ON)
+else()
+set(NO_AFFINITY 1)
+endif()
 
 # Add a prefix or suffix to all exported symbol names in the shared library.
 # Avoids conflicts with other BLAS libraries, especially when using
@@ -42,6 +47,19 @@ endif()
 
 #######
 
+if(MSVC AND MSVC_STATIC_CRT)
+    set(CompilerFlags
+            CMAKE_CXX_FLAGS
+            CMAKE_CXX_FLAGS_DEBUG
+            CMAKE_CXX_FLAGS_RELEASE
+            CMAKE_C_FLAGS
+            CMAKE_C_FLAGS_DEBUG
+            CMAKE_C_FLAGS_RELEASE
+            )
+    foreach(CompilerFlag ${CompilerFlags})
+      string(REPLACE "/MD" "/MT" ${CompilerFlag} "${${CompilerFlag}}")
+    endforeach()
+endif()
 
 message(WARNING "CMake support is experimental. It does not yet support all build options and may not produce the same Makefiles that OpenBLAS ships with.")
 
@@ -62,10 +80,10 @@ endif ()
 
 set(SUBDIRS	${BLASDIRS})
 if (NOT NO_LAPACK)
-  list(APPEND SUBDIRS lapack)
   if(BUILD_RELAPACK)
     list(APPEND SUBDIRS relapack/src)
   endif()
+  list(APPEND SUBDIRS lapack)
 endif ()
 
 # set which float types we want to build for
@@ -134,7 +152,7 @@ endif ()
 
 # Only generate .def for dll on MSVC and always produce pdb files for debug and release
 if(MSVC)
-  if (${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION} LESS 3.4)
+  if (${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION} VERSION_LESS 3.4)
     set(OpenBLAS_DEF_FILE "${PROJECT_BINARY_DIR}/openblas.def")
   endif()
   set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} /Zi")
@@ -149,15 +167,9 @@ if (${DYNAMIC_ARCH})
   endforeach()
 endif ()
 
-# Only build shared libs for MSVC
-if (MSVC)
-  set(BUILD_SHARED_LIBS ON)
-endif()
-
-
 # add objects to the openblas lib
 add_library(${OpenBLAS_LIBNAME} ${LA_SOURCES} ${LAPACKE_SOURCES} ${RELA_SOURCES} ${TARGET_OBJS} ${OpenBLAS_DEF_FILE})
-target_include_directories(${OpenBLAS_LIBNAME} INTERFACE $<INSTALL_INTERFACE:include>)
+target_include_directories(${OpenBLAS_LIBNAME} INTERFACE $<INSTALL_INTERFACE:include/openblas${SUFFIX64}>)
 
 # Android needs to explicitly link against libm
 if(ANDROID)
@@ -166,7 +178,7 @@ endif()
 
 # Handle MSVC exports
 if(MSVC AND BUILD_SHARED_LIBS)
-  if (${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION} LESS 3.4)
+  if (${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION} VERSION_LESS 3.4)
     include("${PROJECT_SOURCE_DIR}/cmake/export.cmake")
   else()
     # Creates verbose .def file (51KB vs 18KB)
@@ -199,7 +211,8 @@ if (USE_THREAD)
   target_link_libraries(${OpenBLAS_LIBNAME} ${CMAKE_THREAD_LIBS_INIT})
 endif()
 
-if (MSVC OR NOT NOFORTRAN)
+#if (MSVC OR NOT NOFORTRAN)
+if (NOT NO_CBLAS)
   # Broken without fortran on unix
   add_subdirectory(utest)
 endif()
@@ -217,6 +230,14 @@ set_target_properties(${OpenBLAS_LIBNAME} PROPERTIES
   SOVERSION ${OpenBLAS_MAJOR_VERSION}
 )
 
+if (BUILD_SHARED_LIBS AND BUILD_RELAPACK)
+  if (NOT MSVC)
+    target_link_libraries(${OpenBLAS_LIBNAME} "-Wl,-allow-multiple-definition")
+  else()
+    target_link_libraries(${OpenBLAS_LIBNAME} "/FORCE:MULTIPLE")
+  endif()
+endif()
+
 if (BUILD_SHARED_LIBS AND NOT ${SYMBOLPREFIX}${SYMBOLSUFIX} STREQUAL "")
 if (NOT DEFINED ARCH)
   set(ARCH_IN "x86_64")
@@ -314,7 +335,7 @@ install (FILES ${OPENBLAS_CONFIG_H} DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
 if(NOT NOFORTRAN)
   message(STATUS "Generating f77blas.h in ${CMAKE_INSTALL_INCLUDEDIR}")
 
-  set(F77BLAS_H ${CMAKE_BINARY_DIR}/f77blas.h)
+  set(F77BLAS_H ${CMAKE_BINARY_DIR}/generated/f77blas.h)
   file(WRITE  ${F77BLAS_H} "#ifndef OPENBLAS_F77BLAS_H\n")
   file(APPEND ${F77BLAS_H} "#define OPENBLAS_F77BLAS_H\n")
   file(APPEND ${F77BLAS_H} "#include \"openblas_config.h\"\n")
@@ -327,10 +348,11 @@ endif()
 if(NOT NO_CBLAS)
 	message (STATUS "Generating cblas.h in ${CMAKE_INSTALL_INCLUDEDIR}")
 
+	set(CBLAS_H ${CMAKE_BINARY_DIR}/generated/cblas.h)
 	file(READ ${CMAKE_CURRENT_SOURCE_DIR}/cblas.h CBLAS_H_CONTENTS)
 	string(REPLACE "common" "openblas_config" CBLAS_H_CONTENTS_NEW "${CBLAS_H_CONTENTS}")
-	file(WRITE ${CMAKE_BINARY_DIR}/cblas.tmp "${CBLAS_H_CONTENTS_NEW}")
-	install (FILES ${CMAKE_BINARY_DIR}/cblas.tmp DESTINATION ${CMAKE_INSTALL_INCLUDEDIR} RENAME cblas.h)
+	file(WRITE ${CBLAS_H} "${CBLAS_H_CONTENTS_NEW}")
+	install (FILES ${CBLAS_H} DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
 endif()
 
 if(NOT NO_LAPACKE)

CONTRIBUTORS.md

@@ -167,4 +167,7 @@ In chronological order:
   * [2017-02-26] ztrmm kernel for IBM z13
   * [2017-03-13] strmm and ctrmm kernel for IBM z13
   * [2017-09-01] initial Blas Level-1,2 (double precision) for IBM z13
-
+  * [2018-03-07] added missing Blas Level 1-2  (double precision) simd codes
+  * [2019-02-01] added missing Blas Level-1,2 (single precision)  simd codes
+  * [2019-03-14] power9 dgemm/dtrmm kernel
+  * [2019-04-29] power9 sgemm/strmm kernel 

Changelog.txt

@@ -1,4 +1,82 @@
 OpenBLAS ChangeLog
+====================================================================
+Version 0.3.6
+29-Apr-2019
+
+common:
+	* the build tools now check that a given cpu TARGET is actually valid
+	* the build-time check of system features (c_check) has been made
+  	  less dependent on particular perl features (this should mainly
+  	  benefit building on Windows)
+	* several problem with the ReLAPACK integration were fixed,
+	  including INTERFACE64 support and building a shared library
+	* building with CMAKE on BSD systems was improved
+	* a non-absolute SUM function was added based on the
+  	  existing optimized code for ASUM
+	* CBLAS interfaces to the IxMIN and IxMAX functions were added
+	* a name clash between LAPACKE and BOOST headers was resolved
+	* CMAKE builds with OpenMP failed to include the appropriate getrf_parallel
+	  kernels
+	* a crash on thread (key) deletion with the USE_TLS=1 memory management
+	  option was fixed
+	* restored several earlier fixes, in particular for OpenMP performance,
+  	  building on BSD, and calling fork on CYGWIN, which had inadvertently
+  	  been dropped in the 0.3.3 rewrite of the memory management code.
+
+x86_64:
+	* the AVX512 DGEMM kernel has been disabled again due to unsolved problems
+	* building with old versions of MSVC was fixed
+	* it is now possible to build a static library on Windows with CMAKE
+	* accessing environment variables on CYGWIN at run time was fixed
+	* the CMAKE build system now recognizes 32bit userspace on 64bit hardware
+	* Intel "Denverton" atom and Hygon "Dhyana" zen CPUs are now autodetected
+	* building for DYNAMIC_ARCH with a DYNAMIC_LIST of targets is now supported
+  	  with CMAKE as well
+	* building for DYNAMIC_ARCH with GENERIC as the default target is now supported
+	* a buffer overflow in the SSE GEMM kernel for Intel Nano targets was fixed
+	* assembly bugs involving undeclared modification of input operands were fixed
+  	  in the AXPY, DOT, GEMV, GER, SCAL, SYMV and TRSM microkernels for Nehalem,
+	  Sandybridge, Haswell, Bulldozer and Piledriver. These would typically cause
+	  test failures or segfaults when compiled with recent versions of gcc from 8 onward.
+	* a similar bug was fixed in the blas_quickdivide code used to split workloads
+	  in most functions
+	* a bug in the IxMIN implementation for the GENERIC target made it return the result of IxMAX
+	* fixed building on SkylakeX systems when either the compiler or the (emulated) operating
+	  environment does not support AVX512
+	* improved GEMM performance on ZEN targets
+
+x86:
+	* build failures caused by the recently added checks for AVX512 were fixed
+	* an inline assembly bug involving undeclared modification of an input argument was
+  	  fixed in the blas_quickdivide code used to split workloads in most functions
+	* a bug in the IMIN implementation for the GENERIC target made it return the result of IMAX
+
+MIPS32:
+	* a bug in the IMIN implementation made it return the result of IMAX
+
+POWER:
+	* single precision BLAS1/2 functions have received optimized POWER8 kernels
+	* POWER9 is now a separate target, with an optimized DGEMM/DTRMM kernel
+	* building on PPC970 systems under OSX Leopard or Tiger is now supported
+	* out-of-bounds memory accesses in the gemm_beta microkernels were fixed
+	* building a shared library on AIX is now supported for POWER6
+	* DYNAMIC_ARCH support has been added for POWER6 and newer
+
+ARMv7:
+	* corrected xDOT behaviour with zero INC_X or INC_Y
+	* a bug in the IMIN implementation made it return the result of IMAX
+
+ARMv8:
+	* added support for HiSilicon TSV110 cpus
+	* the CMAKE build system now recognizes 32bit userspace on 64bit hardware
+	* cross-compilation with CMAKE now works again
+	* a bug in the IMIN implementation made it return the result of IMAX
+	* ARMV8 builds with the BINARY=32 option are now automatically handled as ARMV7
+
+IBM Z:
+	* optimized microkernels for single precicion BLAS1/2 functions have been added
+	  for both Z13 and Z14
+
 ====================================================================
 Version 0.3.5
 31-Dec-2018

Makefile

@@ -34,7 +34,7 @@ endif
 
 LAPACK_NOOPT := $(filter-out -O0 -O1 -O2 -O3 -Ofast,$(LAPACK_FFLAGS))
 
-SUBDIRS_ALL = $(SUBDIRS) test ctest utest exports benchmark ../laswp ../bench
+SUBDIRS_ALL = $(SUBDIRS) test ctest utest exports benchmark ../laswp ../bench cpp_thread_test
 
 .PHONY : all libs netlib $(RELA) test ctest shared install
 .NOTPARALLEL : all libs $(RELA) prof lapack-test install blas-test
@@ -96,7 +96,7 @@ endif
 	@echo
 
 shared :
-ifndef NO_SHARED
+ifneq ($(NO_SHARED), 1)
 ifeq ($(OSNAME), $(filter $(OSNAME),Linux SunOS Android Haiku))
 	@$(MAKE) -C exports so
 	@ln -fs $(LIBSONAME) $(LIBPREFIX).so
@@ -123,10 +123,13 @@ ifeq ($(NOFORTRAN), $(filter 0,$(NOFORTRAN)))
 	touch $(LIBNAME)
 ifndef NO_FBLAS
 	$(MAKE) -C test all
-	$(MAKE) -C utest all
 endif
+	$(MAKE) -C utest all
 ifndef NO_CBLAS
 	$(MAKE) -C ctest all
+ifeq ($(CPP_THREAD_SAFETY_TEST), 1)
+	$(MAKE) -C cpp_thread_test all
+endif
 endif
 endif
 

Makefile.arm64

@@ -38,3 +38,8 @@ ifeq ($(CORE), THUNDERX2T99)
 CCOMMON_OPT += -march=armv8.1-a -mtune=thunderx2t99
 FCOMMON_OPT += -march=armv8.1-a -mtune=thunderx2t99
 endif
+
+ifeq ($(CORE), TSV110)
+CCOMMON_OPT += -march=armv8.2-a -mtune=tsv110
+FCOMMON_OPT += -march=armv8.2-a -mtune=tsv110
+endif

Makefile.install

@@ -58,14 +58,14 @@ ifndef NO_LAPACKE
 endif
 
 #for install static library
-ifndef NO_STATIC
+ifneq ($(NO_STATIC),1)
 	@echo Copying the static library to $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
 	@install -pm644 $(LIBNAME) "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)"
 	@cd "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)" ; \
 	ln -fs $(LIBNAME) $(LIBPREFIX).$(LIBSUFFIX)
 endif
 #for install shared library
-ifndef NO_SHARED
+ifneq ($(NO_SHARED),1)
 	@echo Copying the shared library to $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
 ifeq ($(OSNAME), $(filter $(OSNAME),Linux SunOS Android Haiku))
 	@install -pm755 $(LIBSONAME) "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)"
@@ -106,14 +106,14 @@ ifndef NO_LAPACKE
 endif
 
 #for install static library
-ifndef NO_STATIC
+ifneq ($(NO_STATIC),1)
 	@echo Copying the static library to $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
 	@installbsd -c -m 644 $(LIBNAME) "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)"
 	@cd "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)" ; \
 	ln -fs $(LIBNAME) $(LIBPREFIX).$(LIBSUFFIX)
 endif
 #for install shared library
-ifndef NO_SHARED
+ifneq ($(NO_SHARED),1)
 	@echo Copying the shared library to $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
 	@installbsd -c -m 755 $(LIBSONAME) "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)"
 	@cd "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)" ; \
@@ -138,7 +138,7 @@ endif
 	@echo "SET(OpenBLAS_VERSION \"${VERSION}\")" > "$(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG)"
 	@echo "SET(OpenBLAS_INCLUDE_DIRS ${OPENBLAS_INCLUDE_DIR})" >> "$(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG)"
 
-ifndef NO_SHARED
+ifneq ($(NO_SHARED),1)
 #ifeq logical or
 ifeq ($(OSNAME), $(filter $(OSNAME),Linux FreeBSD NetBSD OpenBSD DragonFly))
 	@echo "SET(OpenBLAS_LIBRARIES ${OPENBLAS_LIBRARY_DIR}/$(LIBPREFIX).so)" >> "$(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG)"

Makefile.power

@@ -9,7 +9,15 @@ else
 USE_OPENMP = 1
 endif
 
-
+ifeq ($(CORE), POWER9)
+ifeq ($(USE_OPENMP), 1)
+COMMON_OPT += -Ofast -mcpu=power9 -mtune=power9 -mvsx -malign-power -DUSE_OPENMP -fno-fast-math -fopenmp
+FCOMMON_OPT += -O2 -frecursive -mcpu=power9 -mtune=power9 -malign-power -DUSE_OPENMP -fno-fast-math -fopenmp
+else
+COMMON_OPT += -Ofast -mcpu=power9 -mtune=power9 -mvsx -malign-power -fno-fast-math
+FCOMMON_OPT += -O2 -frecursive -mcpu=power9 -mtune=power9 -malign-power -fno-fast-math
+endif
+endif
 
 ifeq ($(CORE), POWER8)
 ifeq ($(USE_OPENMP), 1)
@@ -21,6 +29,10 @@ FCOMMON_OPT += -O2 -frecursive -mcpu=power8 -mtune=power8 -malign-power -fno-fas
 endif
 endif
 
+# workaround for C->FORTRAN ABI violation in LAPACKE
+ifeq ($(F_COMPILER), GFORTRAN)
+FCOMMON_OPT += -fno-optimize-sibling-calls
+endif
 
 FLAMEPATH	= $(HOME)/flame/lib
 

Makefile.rule

@@ -3,7 +3,7 @@
 #
 
 # This library's version
-VERSION = 0.3.6.dev
+VERSION = 0.3.7.dev
 
 # If you set the suffix, the library name will be libopenblas_$(LIBNAMESUFFIX).a
 # and libopenblas_$(LIBNAMESUFFIX).so. Meanwhile, the soname in shared library
@@ -58,6 +58,12 @@ VERSION = 0.3.6.dev
 # For force setting for multi  threaded, specify USE_THREAD = 1
 # USE_THREAD = 0
 
+# If you want to build a single-threaded OpenBLAS, but expect to call this
+# from several concurrent threads in some other program, comment this in for
+# thread safety. (This is done automatically for USE_THREAD=1 , and should not
+# be necessary when USE_OPENMP=1)
+# USE_LOCKING = 1
+
 # If you're going to use this library with OpenMP, please comment it in.
 # This flag is always set for POWER8. Don't set USE_OPENMP = 0 if you're targeting POWER8.
 # USE_OPENMP = 1
@@ -157,6 +163,10 @@ NO_AFFINITY = 1
 # Don't use Haswell optimizations if binutils is too old (e.g. RHEL6)
 # NO_AVX2 = 1
 
+# Don't use SkylakeX optimizations if binutils or compiler are too old (the build
+# system will try to determine this automatically)
+# NO_AVX512 = 1
+
 # Don't use parallel make.
 # NO_PARALLEL_MAKE = 1
 
@@ -181,17 +191,17 @@ NO_AFFINITY = 1
 # time out to improve performance. This number should be from 4 to 30
 # which corresponds to (1 << n) cycles. For example, if you set to 26,
 # thread will be running for (1 << 26) cycles(about 25ms on 3.0GHz
-# system). Also you can control this mumber by THREAD_TIMEOUT
+# system). Also you can control this number by THREAD_TIMEOUT
 # CCOMMON_OPT	+= -DTHREAD_TIMEOUT=26
 
-# Using special device driver for mapping physically contigous memory
+# Using special device driver for mapping physically contiguous memory
 # to the user space. If bigphysarea is enabled, it will use it.
 # DEVICEDRIVER_ALLOCATION = 1
 
 # If you need to synchronize FP CSR between threads (for x86/x86_64 only).
 # CONSISTENT_FPCSR = 1
 
-# If any gemm arguement m, n or k is less or equal this threshold, gemm will be execute
+# If any gemm argument m, n or k is less or equal this threshold, gemm will be execute
 # with single thread. (Actually in recent versions this is a factor proportional to the
 # number of floating point operations necessary for the given problem size, no longer
 # an individual dimension). You can use this setting to avoid the overhead of multi-
@@ -199,7 +209,7 @@ NO_AFFINITY = 1
 # been reported to be optimal for certain workloads (50 is the recommended value for Julia).
 # GEMM_MULTITHREAD_THRESHOLD = 4
 
-# If you need santy check by comparing reference BLAS. It'll be very
+# If you need sanity check by comparing results to reference BLAS. It'll be very
 # slow (Not implemented yet).
 # SANITY_CHECK = 1
 
@@ -239,6 +249,21 @@ COMMON_PROF = -pg
 # SYMBOLPREFIX=
 # SYMBOLSUFFIX=
 
+# Run a C++ based thread safety tester after the build is done.
+# This is mostly intended as a developer feature to spot regressions, but users and
+# package maintainers can enable this if they have doubts about the thread safety of
+# the library, given the configuration in this file.
+# By default, the thread safety tester launches 52 concurrent calculations at the same
+# time.
+#
+# Please note that the test uses ~1300 MiB of RAM for the DGEMM test.
+#
+# The test requires CBLAS to be built, a C++11 capable compiler and the presence of
+# an OpenMP implementation. If you are cross-compiling this test will probably not
+# work at all.
+#
+# CPP_THREAD_SAFETY_TEST = 1
+
 #
 #  End of user configuration
 #

Makefile.system

@@ -9,6 +9,11 @@ ifndef TOPDIR
 TOPDIR = .
 endif
 
+# If ARCH is not set, we use the host system's architecture.
+ifndef ARCH
+ARCH := $(shell uname -m)
+endif
+
 # Catch conflicting usage of ARCH in some BSD environments
 ifeq ($(ARCH), amd64)
 override ARCH=x86_64
@@ -137,7 +142,12 @@ endif
 endif
 
 
-
+# On x86_64 build getarch with march=native. This is required to detect AVX512 support in getarch.
+ifeq ($(ARCH), x86_64)
+ifneq ($(C_COMPILER), PGI)
+GETARCH_FLAGS += -march=native
+endif
+endif
 
 ifdef INTERFACE64
 ifneq ($(INTERFACE64), 0)
@@ -155,7 +165,8 @@ GETARCH_FLAGS	+= -DNO_AVX
 endif
 
 ifeq ($(BINARY), 32)
-GETARCH_FLAGS	+= -DNO_AVX
+GETARCH_FLAGS	+= -DNO_AVX -DNO_AVX2 -DNO_AVX512
+NO_AVX512 = 1
 endif
 
 ifeq ($(NO_AVX2), 1)
@@ -236,6 +247,10 @@ SMP = 1
 endif
 endif
 
+ifeq ($(SMP), 1)
+USE_LOCKING = 
+endif
+
 ifndef NEED_PIC
 NEED_PIC = 1
 endif
@@ -387,6 +402,12 @@ ifneq ($(MAX_STACK_ALLOC), 0)
 CCOMMON_OPT	+= -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
 endif
 
+ifdef USE_LOCKING
+ifneq ($(USE_LOCKING), 0)
+CCOMMON_OPT	+= -DUSE_LOCKING
+endif
+endif
+
 #
 #  Architecture dependent settings
 #
@@ -527,6 +548,12 @@ DYNAMIC_CORE += THUNDERX
 DYNAMIC_CORE += THUNDERX2T99
 endif
 
+ifeq ($(ARCH), power)
+DYNAMIC_CORE = POWER6
+DYNAMIC_CORE += POWER8
+DYNAMIC_CORE += POWER9
+endif
+
 # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
 ifndef DYNAMIC_CORE
 override DYNAMIC_ARCH=
@@ -737,6 +764,8 @@ CCOMMON_OPT += -DF_INTERFACE_GFORT
 FCOMMON_OPT += -Wall
 # make single-threaded LAPACK calls thread-safe #1847
 FCOMMON_OPT += -frecursive
+# work around ABI problem with passing single-character arguments
+FCOMMON_OPT += -fno-optimize-sibling-calls
 #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
 ifneq ($(NO_LAPACK), 1)
 EXTRALIB += -lgfortran
@@ -1042,7 +1071,7 @@ ifdef USE_SIMPLE_THREADED_LEVEL3
 CCOMMON_OPT	+= -DUSE_SIMPLE_THREADED_LEVEL3
 endif
 
-ifdef USE_TLS
+ifeq ($(USE_TLS), 1)
 CCOMMON_OPT += -DUSE_TLS
 endif
 

Makefile.x86_64

@@ -28,11 +28,15 @@ endif
 ifeq ($(CORE), HASWELL)
 ifndef DYNAMIC_ARCH
 ifndef NO_AVX2
+ifeq ($(C_COMPILER), GCC)
 CCOMMON_OPT += -mavx2
+endif
+ifeq ($(F_COMPILER), GFORTRAN)
 FCOMMON_OPT += -mavx2
 endif
 endif
 endif
+endif
 
 
 

Makefile.zarch

@@ -4,3 +4,7 @@ CCOMMON_OPT += -march=z13 -mzvector
 FCOMMON_OPT += -march=z13 -mzvector
 endif
 
+ifeq ($(CORE), Z14)
+CCOMMON_OPT += -march=z14 -mzvector
+FCOMMON_OPT += -march=z14 -mzvector
+endif

README.md

@@ -6,11 +6,13 @@ Travis CI: [![Build Status](https://travis-ci.org/xianyi/OpenBLAS.svg?branch=dev
 
 AppVeyor: [![Build status](https://ci.appveyor.com/api/projects/status/09sohd35n8nkkx64/branch/develop?svg=true)](https://ci.appveyor.com/project/xianyi/openblas/branch/develop)
 
+[![Build Status](https://dev.azure.com/xianyi/OpenBLAS/_apis/build/status/xianyi.OpenBLAS?branchName=develop)](https://dev.azure.com/xianyi/OpenBLAS/_build/latest?definitionId=1&branchName=develop)
+
 ## Introduction
 
 OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.
 
-Please read the documentation on the OpenBLAS wiki pages: <http://github.com/xianyi/OpenBLAS/wiki>.
+Please read the documentation on the OpenBLAS wiki pages: <https://github.com/xianyi/OpenBLAS/wiki>.
 
 ## Binary Packages
 
@@ -22,7 +24,7 @@ You can download them from [file hosting on sourceforge.net](https://sourceforge
 
 ## Installation from Source
 
-Download from project homepage, http://xianyi.github.com/OpenBLAS/, or check out the code
+Download from project homepage, https://xianyi.github.com/OpenBLAS/, or check out the code
 using Git from https://github.com/xianyi/OpenBLAS.git.
 
 ### Dependencies
@@ -63,9 +65,7 @@ A debug version can be built using `make DEBUG=1`.
 
 ### Compile with MASS support on Power CPU (optional)
 
-The [IBM MASS](http://www-01.ibm.com/software/awdtools/mass/linux/mass-linux.html) library
-consists of a set of mathematical functions for C, C++, and Fortran applications that are
-are tuned for optimum performance on POWER architectures.
+The [IBM MASS](https://www.ibm.com/support/home/product/W511326D80541V01/other_software/mathematical_acceleration_subsystem) library consists of a set of mathematical functions for C, C++, and Fortran applications that are tuned for optimum performance on POWER architectures.
 OpenBLAS with MASS requires a 64-bit, little-endian OS on POWER.
 The library can be installed as shown:
 
@@ -115,6 +115,7 @@ Please read `GotoBLAS_01Readme.txt`.
 - **AMD Bulldozer**: x86-64 ?GEMM FMA4 kernels. (Thanks to Werner Saar)
 - **AMD PILEDRIVER**: Uses Bulldozer codes with some optimizations.
 - **AMD STEAMROLLER**: Uses Bulldozer codes with some optimizations.
+- **AMD ZEN**: Uses Haswell codes with some optimizations.
 
 #### MIPS64
 
@@ -133,11 +134,13 @@ Please read `GotoBLAS_01Readme.txt`.
 
 #### PPC/PPC64
 
-- **POWER8**: Optmized Level-3 BLAS and some Level-1, only with `USE_OPENMP=1`
+- **POWER8**: Optimized BLAS, only for PPC64LE (Little Endian), only with `USE_OPENMP=1`
+- **POWER9**: Optimized Level-3 BLAS (real) and some Level-1,2. PPC64LE with OpenMP only. 
 
 #### IBM zEnterprise System
 
 - **Z13**: Optimized Level-3 BLAS and Level-1,2 (double precision)
+- **Z14**: Optimized Level-3 BLAS and Level-1,2 (single precision)
 
 ### Supported OS
 

TargetList.txt

@@ -48,6 +48,7 @@ POWER5
 POWER6
 POWER7
 POWER8
+POWER9
 PPCG4
 PPC970
 PPC970MP
@@ -90,7 +91,9 @@ CORTEXA73
 FALKOR
 THUNDERX
 THUNDERX2T99
+TSV110
 
 9.System Z:
 ZARCH_GENERIC
 Z13
+Z14

appveyor.yml

@@ -35,7 +35,14 @@ environment:
       DYNAMIC_ARCH: ON
       WITH_FORTRAN: no
     - COMPILER: cl
-
+    - COMPILER: MinGW64-gcc-7.2.0-mingw
+      DYNAMIC_ARCH: OFF
+      WITH_FORTRAN: ignore
+    - COMPILER: MinGW64-gcc-7.2.0
+    - APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2015
+      COMPILER: MinGW-gcc-5.3.0
+      WITH_FORTRAN: ignore
+    
 install:
   - if [%COMPILER%]==[clang-cl] call %CONDA_INSTALL_LOCN%\Scripts\activate.bat
   - if [%COMPILER%]==[clang-cl] conda config --add channels conda-forge --force
@@ -52,10 +59,17 @@ install:
 before_build:
   - ps: if (-Not (Test-Path .\build)) { mkdir build }
   - cd build
+  - set PATH=%PATH:C:\Program Files\Git\usr\bin;=%
+  - if [%COMPILER%]==[MinGW-gcc-5.3.0] set PATH=C:\MinGW\bin;C:\msys64\usr\bin;C:\mingw-w64\x86_64-7.2.0-posix-seh-rt_v5-rev1\mingw64\bin;%PATH%
+  - if [%COMPILER%]==[MinGW64-gcc-7.2.0-mingw] set PATH=C:\MinGW\bin;C:\mingw-w64\x86_64-7.2.0-posix-seh-rt_v5-rev1\mingw64\bin;%PATH%
+  - if [%COMPILER%]==[MinGW64-gcc-7.2.0] set PATH=C:\msys64\usr\bin;C:\mingw-w64\x86_64-7.2.0-posix-seh-rt_v5-rev1\mingw64\bin;%PATH%
   - if [%COMPILER%]==[cl] cmake -G "Visual Studio 15 2017 Win64" ..
-  - if [%WITH_FORTRAN%]==[no] cmake -G "Ninja" -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_C_COMPILER=clang-cl ..
+  - if [%COMPILER%]==[MinGW64-gcc-7.2.0-mingw] cmake -G "MinGW Makefiles" -DNOFORTRAN=1 ..
+  - if [%COMPILER%]==[MinGW64-gcc-7.2.0] cmake -G "MSYS Makefiles"  -DBINARY=32 -DNOFORTRAN=1 ..
+  - if [%COMPILER%]==[MinGW-gcc-5.3.0] cmake -G "MSYS Makefiles" -DNOFORTRAN=1 ..
+  - if [%WITH_FORTRAN%]==[no] cmake -G "Ninja" -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_C_COMPILER=clang-cl -DMSVC_STATIC_CRT=ON ..
   - if [%WITH_FORTRAN%]==[yes] cmake -G "Ninja" -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_C_COMPILER=clang-cl -DCMAKE_Fortran_COMPILER=flang -DBUILD_WITHOUT_LAPACK=no -DNOFORTRAN=0 ..
-  - if [%DYNAMIC_ARCH%]==[ON] cmake -DDYNAMIC_ARCH=ON ..
+  - if [%DYNAMIC_ARCH%]==[ON] cmake -DDYNAMIC_ARCH=ON -DDYNAMIC_LIST='CORE2;NEHALEM;SANDYBRIDGE;BULLDOZER;HASWELL' ..
 
 build_script:
   - cmake --build .
@@ -64,3 +78,4 @@ test_script:
   - echo Running Test
   - cd utest
   - openblas_utest
+  

azure-pipelines.yml

@@ -0,0 +1,51 @@
+trigger:
+  # start a new build for every push
+  batch: False
+  branches:
+    include:
+      - develop
+
+jobs:
+# manylinux1 is useful to test because the
+# standard Docker container uses an old version
+# of gcc / glibc
+- job: manylinux1_gcc
+  pool:
+    vmImage: 'ubuntu-16.04'
+  steps:
+  - script: |
+      echo "FROM quay.io/pypa/manylinux1_x86_64
+        COPY . /tmp/openblas
+        RUN cd /tmp/openblas                                      &&  \
+            COMMON_FLAGS='DYNAMIC_ARCH=1 TARGET=NEHALEM NUM_THREADS=32' && \
+            BTYPE='BINARY=64' CC=gcc && \
+            make QUIET_MAKE=1 $COMMON_FLAGS $BTYPE && \
+            make -C test $COMMON_FLAGS $BTYPE && \
+            make -C ctest $COMMON_FLAGS $BTYPE && \
+            make -C utest $COMMON_FLAGS $BTYPE" > Dockerfile
+      docker build .
+    displayName: Run manylinux1 docker build
+- job: Intel_SDE_skx
+  pool:
+    vmImage: 'ubuntu-16.04'
+  steps:
+  - script: |
+      # at the time of writing the available Azure Ubuntu vm image
+      # does not support AVX512VL, so use more recent LTS version
+      echo "FROM ubuntu:bionic
+      COPY . /tmp/openblas
+      RUN apt-get -y update && apt-get -y install \\
+          cmake \\
+          gfortran \\
+          make \\
+          wget
+      RUN mkdir /tmp/SDE && cd /tmp/SDE && \\
+          mkdir sde-external-8.35.0-2019-03-11-lin && \\
+          wget --quiet -O sde-external-8.35.0-2019-03-11-lin.tar.bz2 https://www.dropbox.com/s/fopsnzj67572sj5/sde-external-8.35.0-2019-03-11-lin.tar.bz2?dl=0 && \\
+          tar -xjvf sde-external-8.35.0-2019-03-11-lin.tar.bz2 -C /tmp/SDE/sde-external-8.35.0-2019-03-11-lin --strip-components=1
+      RUN cd /tmp/openblas && CC=gcc make QUIET_MAKE=1 DYNAMIC_ARCH=1 NUM_THREADS=32 BINARY=64
+      CMD cd /tmp/openblas && echo 0 > /proc/sys/kernel/yama/ptrace_scope && CC=gcc OPENBLAS_VERBOSE=2 /tmp/SDE/sde-external-8.35.0-2019-03-11-lin/sde64 -cpuid_in /tmp/SDE/sde-external-8.35.0-2019-03-11-lin/misc/cpuid/skx/cpuid.def -- make -C utest DYNAMIC_ARCH=1 NUM_THREADS=32 BINARY=64" > Dockerfile
+      docker build -t intel_sde .
+      # we need a privileged docker run for sde process attachment
+      docker run --privileged intel_sde
+    displayName: 'Run AVX512 SkylakeX docker build / test'

benchmark/gemm.c

@@ -207,7 +207,7 @@ int main(int argc, char *argv[]){
   for (i = 0; i < m * n * COMPSIZE; i++) {
     c[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
   }
- 
+
   fprintf(stderr, "          SIZE                   Flops             Time\n");
 
   for (i = from; i <= to; i += step) {

benchmark/scripts/R/deig.R

@@ -2,6 +2,8 @@
 
 argv <- commandArgs(trailingOnly = TRUE)
 
+if (!is.null(options("matprod")[[1]])) options(matprod = "blas")
+
 nfrom <- 128
 nto <- 2048
 nstep <- 128
@@ -19,7 +21,6 @@ if (length(argv) > 0) {
       loops <- as.numeric(argv[z])
     }
   }
-
 }
 
 p <- Sys.getenv("OPENBLAS_LOOPS")
@@ -27,29 +28,21 @@ if (p != "") {
   loops <- as.numeric(p)
 }
 
-
-cat(sprintf(
-  "From %.0f To %.0f Step=%.0f Loops=%.0f\n",
-  nfrom,
-  nto,
-  nstep,
-  loops
-))
+cat(sprintf("From %.0f To %.0f Step=%.0f Loops=%.0f\n", nfrom, nto, nstep, loops))
 cat(sprintf("      SIZE             Flops                   Time\n"))
 
 n <- nfrom
 while (n <= nto) {
-  A <- matrix(rnorm(n * n), ncol = n, nrow = n)
+  A <- matrix(rnorm(n * n), nrow = n)
   ev <- 0
   z <- system.time(for (l in 1:loops) {
     ev <- eigen(A)
   })
 
-  mflops <- (26.66 * n * n * n) * loops / (z[3] * 1.0e6)
+  mflops <- (26.66 * n * n * n) * loops / (z[3] * 1e+06)
 
   st <- sprintf("%.0fx%.0f :", n, n)
   cat(sprintf("%20s %10.2f MFlops %10.6f sec\n", st, mflops, z[3]))
 
   n <- n + nstep
-
 }

benchmark/scripts/R/dgemm.R

@@ -2,6 +2,8 @@
 
 argv <- commandArgs(trailingOnly = TRUE)
 
+if (!is.null(options("matprod")[[1]])) options(matprod = "blas")
+
 nfrom <- 128
 nto <- 2048
 nstep <- 128
@@ -19,7 +21,6 @@ if (length(argv) > 0) {
       loops <- as.numeric(argv[z])
     }
   }
-
 }
 
 p <- Sys.getenv("OPENBLAS_LOOPS")
@@ -27,26 +28,13 @@ if (p != "") {
   loops <- as.numeric(p)
 }
 
-
-cat(sprintf(
-  "From %.0f To %.0f Step=%.0f Loops=%.0f\n",
-  nfrom,
-  nto,
-  nstep,
-  loops
-))
+cat(sprintf("From %.0f To %.0f Step=%.0f Loops=%.0f\n", nfrom, nto, nstep, loops))
 cat(sprintf("      SIZE             Flops                   Time\n"))
 
 n <- nfrom
 while (n <= nto) {
-  A <- matrix(runif(n * n),
-              ncol = n,
-              nrow = n,
-              byrow = TRUE)
-  B <- matrix(runif(n * n),
-              ncol = n,
-              nrow = n,
-              byrow = TRUE)
+  A <- matrix(runif(n * n), nrow = n)
+  B <- matrix(runif(n * n), nrow = n)
   C <- 1
 
   z <- system.time(for (l in 1:loops) {
@@ -54,11 +42,10 @@ while (n <= nto) {
     l <- l + 1
   })
 
-  mflops <- (2.0 * n * n * n) * loops / (z[3] * 1.0e6)
+  mflops <- (2.0 * n * n * n) * loops / (z[3] * 1e+06)
 
   st <- sprintf("%.0fx%.0f :", n, n)
   cat(sprintf("%20s %10.2f MFlops %10.6f sec\n", st, mflops, z[3]))
 
   n <- n + nstep
-
 }

benchmark/scripts/R/dsolve.R

@@ -2,6 +2,8 @@
 
 argv <- commandArgs(trailingOnly = TRUE)
 
+if (!is.null(options("matprod")[[1]])) options(matprod = "blas")
+
 nfrom <- 128
 nto <- 2048
 nstep <- 128
@@ -19,7 +21,6 @@ if (length(argv) > 0) {
       loops <- as.numeric(argv[z])
     }
   }
-
 }
 
 p <- Sys.getenv("OPENBLAS_LOOPS")
@@ -27,31 +28,22 @@ if (p != "") {
   loops <- as.numeric(p)
 }
 
-
-cat(sprintf(
-  "From %.0f To %.0f Step=%.0f Loops=%.0f\n",
-  nfrom,
-  nto,
-  nstep,
-  loops
-))
+cat(sprintf("From %.0f To %.0f Step=%.0f Loops=%.0f\n", nfrom, nto, nstep, loops))
 cat(sprintf("      SIZE             Flops                   Time\n"))
 
 n <- nfrom
 while (n <= nto) {
-  A <- matrix(rnorm(n * n), ncol = n, nrow = n)
-  B <- matrix(rnorm(n * n), ncol = n, nrow = n)
+  A <- matrix(rnorm(n * n), nrow = n)
+  B <- matrix(rnorm(n * n), nrow = n)
 
   z <- system.time(for (l in 1:loops) {
     solve(A, B)
   })
 
-  mflops <-
-    (2.0 / 3.0 * n * n * n + 2.0 * n * n * n) * loops / (z[3] * 1.0e6)
+  mflops <- (8.0 / 3 * n * n * n) * loops / (z[3] * 1e+06)
 
   st <- sprintf("%.0fx%.0f :", n, n)
   cat(sprintf("%20s %10.2f MFlops %10.6f sec\n", st, mflops, z[3]))
 
   n <- n + nstep
-
 }

c_check

@@ -1,7 +1,7 @@
 #!/usr/bin/perl
 
-use File::Basename;
-use File::Temp qw(tempfile);
+#use File::Basename;
+# use File::Temp qw(tempfile);
 
 # Checking cross compile
 $hostos   = `uname -s | sed -e s/\-.*//`;    chop($hostos);
@@ -12,7 +12,7 @@ $hostarch = "arm64" if ($hostarch eq "aarch64");
 $hostarch = "power" if ($hostarch =~ /^(powerpc|ppc).*/);
 $hostarch = "zarch" if ($hostarch eq "s390x");
 
-$tmpf = new File::Temp( UNLINK => 1 );
+#$tmpf = new File::Temp( UNLINK => 1 );
 $binary = $ENV{"BINARY"};
 
 $makefile = shift(@ARGV);
@@ -31,12 +31,25 @@ if ($?) {
 
 $cross_suffix = "";
 
-if (dirname($compiler_name) ne ".") {
-    $cross_suffix .= dirname($compiler_name) . "/";
-}
+eval "use File::Basename";
+if ($@){ 
+    warn "could not load PERL module File::Basename, emulating its functionality";
+    my $dirnam = substr($compiler_name, 0, rindex($compiler_name, "/")-1 );
+    if ($dirnam ne ".") {
+	$cross_suffix .= $dirnam . "/";
+    }
+    my $basnam = substr($compiler_name, rindex($compiler_name,"/")+1, length($compiler_name)-rindex($compiler_name,"/")-1);
+	if ($basnam =~ /([^\s]*-)(.*)/) {
+	$cross_suffix .= $1;
+    }
+} else {
+    if (dirname($compiler_name) ne ".") {
+	$cross_suffix .= dirname($compiler_name) . "/";
+    }
 
-if (basename($compiler_name) =~ /([^\s]*-)(.*)/) {
-    $cross_suffix .= $1;
+    if (basename($compiler_name) =~ /([^\s]*-)(.*)/) {
+	$cross_suffix .= $1;
+    }
 }
 
 $compiler = "";
@@ -171,20 +184,26 @@ if ($?) {
 
 $have_msa = 0;
 if (($architecture eq "mips") || ($architecture eq "mips64")) {
-    $code = '"addvi.b $w0, $w1, 1"';
-    $msa_flags = "-mmsa -mfp64 -msched-weight -mload-store-pairs";
-    print $tmpf "#include <msa.h>\n\n";
-    print $tmpf "void main(void){ __asm__ volatile($code); }\n";
-
-    $args = "$msa_flags -o $tmpf.o -x c $tmpf";
-    my @cmd = ("$compiler_name $args");
-    system(@cmd) == 0;
-    if ($? != 0) {
-        $have_msa = 0;
+    eval "use File::Temp qw(tempfile)";
+    if ($@){ 
+	warn "could not load PERL module File::Temp, so could not check MSA capatibility";
     } else {
-        $have_msa = 1;
+	$tmpf = new File::Temp( UNLINK => 1 );
+	$code = '"addvi.b $w0, $w1, 1"';
+	$msa_flags = "-mmsa -mfp64 -msched-weight -mload-store-pairs";
+	print $tmpf "#include <msa.h>\n\n";
+	print $tmpf "void main(void){ __asm__ volatile($code); }\n";
+
+	$args = "$msa_flags -o $tmpf.o -x c $tmpf";
+	my @cmd = ("$compiler_name $args");
+	system(@cmd) == 0;
+	if ($? != 0) {
+	    $have_msa = 0;
+	} else {
+	    $have_msa = 1;
+	}
+	unlink("$tmpf.o");
     }
-    unlink("$tmpf.o");
 }
 
 $architecture = x86    if ($data =~ /ARCH_X86/);
@@ -204,17 +223,25 @@ $binformat    = bin64  if ($data =~ /BINARY_64/);
 
 $no_avx512= 0;
 if (($architecture eq "x86") || ($architecture eq "x86_64")) {
-    $code = '"vbroadcastss -4 * 4(%rsi), %zmm2"';
-    print $tmpf "#include <immintrin.h>\n\nint main(void){ __asm__ volatile($code); }\n";
-    $args = " -march=skylake-avx512 -o $tmpf.o -x c $tmpf";
-    my @cmd = ("$compiler_name $args >/dev/null 2>/dev/null");
-    system(@cmd) == 0;
-    if ($? != 0) {
-	$no_avx512 = 1;
-    } else {
+    eval "use File::Temp qw(tempfile)";
+    if ($@){ 
+	warn "could not load PERL module File::Temp, so could not check compiler compatibility with AVX512";
 	$no_avx512 = 0;
+    } else {
+#	$tmpf = new File::Temp( UNLINK => 1 );
+	($fh,$tmpf) = tempfile( UNLINK => 1 );
+	$code = '"vbroadcastss -4 * 4(%rsi), %zmm2"';
+	print $tmpf "#include <immintrin.h>\n\nint main(void){ __asm__ volatile($code); }\n";
+	$args = " -march=skylake-avx512 -c -o $tmpf.o -x c $tmpf";
+	my @cmd = ("$compiler_name $args >/dev/null 2>/dev/null");
+	system(@cmd) == 0;
+	if ($? != 0) {
+	    $no_avx512 = 1;
+	} else {
+	    $no_avx512 = 0;
+	}
+	unlink("$tmpf.o");
     }
-    unlink("tmpf.o");
 }
 
 $data = `$compiler_name -S ctest1.c && grep globl ctest1.s | head -n 1 && rm -f ctest1.s`;

cblas.h

@@ -73,6 +73,11 @@ double cblas_dasum (OPENBLAS_CONST blasint n, OPENBLAS_CONST double *x, OPENBLAS
 float  cblas_scasum(OPENBLAS_CONST blasint n, OPENBLAS_CONST void  *x, OPENBLAS_CONST blasint incx);
 double cblas_dzasum(OPENBLAS_CONST blasint n, OPENBLAS_CONST void *x, OPENBLAS_CONST blasint incx);
 
+float  cblas_ssum (OPENBLAS_CONST blasint n, OPENBLAS_CONST float  *x, OPENBLAS_CONST blasint incx);
+double cblas_dsum (OPENBLAS_CONST blasint n, OPENBLAS_CONST double *x, OPENBLAS_CONST blasint incx);
+float  cblas_scsum(OPENBLAS_CONST blasint n, OPENBLAS_CONST void  *x, OPENBLAS_CONST blasint incx);
+double cblas_dzsum(OPENBLAS_CONST blasint n, OPENBLAS_CONST void *x, OPENBLAS_CONST blasint incx);
+
 float  cblas_snrm2 (OPENBLAS_CONST blasint N, OPENBLAS_CONST float  *X, OPENBLAS_CONST blasint incX);
 double cblas_dnrm2 (OPENBLAS_CONST blasint N, OPENBLAS_CONST double *X, OPENBLAS_CONST blasint incX);
 float  cblas_scnrm2(OPENBLAS_CONST blasint N, OPENBLAS_CONST void  *X, OPENBLAS_CONST blasint incX);
@@ -88,6 +93,16 @@ CBLAS_INDEX cblas_idamin(OPENBLAS_CONST blasint n, OPENBLAS_CONST double *x, OPE
 CBLAS_INDEX cblas_icamin(OPENBLAS_CONST blasint n, OPENBLAS_CONST void  *x, OPENBLAS_CONST blasint incx);
 CBLAS_INDEX cblas_izamin(OPENBLAS_CONST blasint n, OPENBLAS_CONST void *x, OPENBLAS_CONST blasint incx);
 
+CBLAS_INDEX cblas_ismax(OPENBLAS_CONST blasint n, OPENBLAS_CONST float  *x, OPENBLAS_CONST blasint incx);
+CBLAS_INDEX cblas_idmax(OPENBLAS_CONST blasint n, OPENBLAS_CONST double *x, OPENBLAS_CONST blasint incx);
+CBLAS_INDEX cblas_icmax(OPENBLAS_CONST blasint n, OPENBLAS_CONST void  *x, OPENBLAS_CONST blasint incx);
+CBLAS_INDEX cblas_izmax(OPENBLAS_CONST blasint n, OPENBLAS_CONST void *x, OPENBLAS_CONST blasint incx);
+
+CBLAS_INDEX cblas_ismin(OPENBLAS_CONST blasint n, OPENBLAS_CONST float  *x, OPENBLAS_CONST blasint incx);
+CBLAS_INDEX cblas_idmin(OPENBLAS_CONST blasint n, OPENBLAS_CONST double *x, OPENBLAS_CONST blasint incx);
+CBLAS_INDEX cblas_icmin(OPENBLAS_CONST blasint n, OPENBLAS_CONST void  *x, OPENBLAS_CONST blasint incx);
+CBLAS_INDEX cblas_izmin(OPENBLAS_CONST blasint n, OPENBLAS_CONST void *x, OPENBLAS_CONST blasint incx);
+
 void cblas_saxpy(OPENBLAS_CONST blasint n, OPENBLAS_CONST float alpha, OPENBLAS_CONST float *x, OPENBLAS_CONST blasint incx, float *y, OPENBLAS_CONST blasint incy);
 void cblas_daxpy(OPENBLAS_CONST blasint n, OPENBLAS_CONST double alpha, OPENBLAS_CONST double *x, OPENBLAS_CONST blasint incx, double *y, OPENBLAS_CONST blasint incy);
 void cblas_caxpy(OPENBLAS_CONST blasint n, OPENBLAS_CONST void *alpha, OPENBLAS_CONST void *x, OPENBLAS_CONST blasint incx, void *y, OPENBLAS_CONST blasint incy);

cmake/arch.cmake

@@ -73,11 +73,16 @@ if (DYNAMIC_ARCH)
     endif ()
     if (NOT NO_AVX512)
       set(DYNAMIC_CORE ${DYNAMIC_CORE} SKYLAKEX)
+      string(REGEX REPLACE "-march=native" "" CMAKE_C_FLAGS ${CMAKE_C_FLAGS})
+   endif ()
+    if (DYNAMIC_LIST)
+	set(DYNAMIC_CORE PRESCOTT ${DYNAMIC_LIST})
     endif ()
   endif ()
 
   if (NOT DYNAMIC_CORE)
-    unset(DYNAMIC_ARCH)
+    message (STATUS "DYNAMIC_ARCH is not supported on this architecture, removing from options")
+    unset(DYNAMIC_ARCH CACHE)
   endif ()
 endif ()
 

cmake/fc.cmake

@@ -44,7 +44,10 @@ endif ()
 
 if (${F_COMPILER} STREQUAL "GFORTRAN")
   set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_GFORT")
+  # ensure reentrancy of lapack codes
   set(FCOMMON_OPT "${FCOMMON_OPT} -Wall -frecursive")
+  # work around ABI violation in passing string arguments from C
+  set(FCOMMON_OPT "${FCOMMON_OPT} -fno-optimize-sibling-calls")
   #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
   if (NOT NO_LAPACK)
     set(EXTRALIB "{EXTRALIB} -lgfortran")

cmake/kernel.cmake

@@ -1,7 +1,7 @@
 # helper functions for the kernel CMakeLists.txt
 
 
-# Set the default filenames for L1 objects. Most of these will be overriden by the appropriate KERNEL file.
+# Set the default filenames for L1 objects. Most of these will be overridden by the appropriate KERNEL file.
 macro(SetDefaultL1)
   set(SAMAXKERNEL amax.S)
   set(DAMAXKERNEL amax.S)
@@ -107,6 +107,12 @@ macro(SetDefaultL1)
   set(DAXPBYKERNEL ../arm/axpby.c)
   set(CAXPBYKERNEL ../arm/zaxpby.c)
   set(ZAXPBYKERNEL ../arm/zaxpby.c)
+  set(SSUMKERNEL sum.S)
+  set(DSUMKERNEL sum.S)
+  set(CSUMKERNEL zsum.S)
+  set(ZSUMKERNEL zsum.S)
+  set(QSUMKERNEL sum.S)
+  set(XSUMKERNEL zsum.S)
 endmacro ()
 
 macro(SetDefaultL2)
@@ -162,4 +168,4 @@ macro(SetDefaultL3)
   set(DGEADD_KERNEL ../generic/geadd.c)
   set(CGEADD_KERNEL ../generic/zgeadd.c)
   set(ZGEADD_KERNEL ../generic/zgeadd.c)
-endmacro ()
+endmacro ()

cmake/os.cmake

@@ -8,6 +8,11 @@ if (${CMAKE_SYSTEM_NAME} STREQUAL "Linux")
   set(NO_EXPRECISION 1)
 endif ()
 
+if (${CMAKE_SYSTEM_NAME} MATCHES "FreeBSD|OpenBSD|NetBSD|DragonFly")
+  set(EXTRALIB "${EXTRALIB} -lm")
+  set(NO_EXPRECISION 1)
+endif ()
+
 if (${CMAKE_SYSTEM_NAME} STREQUAL "AIX")
   set(EXTRALIB "${EXTRALIB} -lm")
 endif ()

cmake/prebuild.cmake

@@ -59,6 +59,9 @@ set(FU "")
 if (APPLE OR (MSVC AND NOT ${CMAKE_C_COMPILER_ID} MATCHES "Clang"))
   set(FU "_")
 endif()
+if(MINGW AND NOT MINGW64)
+  set(FU "_")
+endif()
 
 set(COMPILER_ID ${CMAKE_C_COMPILER_ID})
 if (${COMPILER_ID} STREQUAL "GNU")
@@ -82,6 +85,11 @@ endif ()
 # f_check
 if (NOT NOFORTRAN)
   include("${PROJECT_SOURCE_DIR}/cmake/f_check.cmake")
+else ()
+ file(APPEND ${TARGET_CONF_TEMP}
+   "#define BUNDERSCORE _\n"
+   "#define NEEDBUNDERSCORE 1\n")
+ set(BU "_")
 endif ()
 
 # Cannot run getarch on target if we are cross-compiling

cmake/system.cmake

@@ -65,6 +65,18 @@ if (DEFINED TARGET)
   set(GETARCH_FLAGS "-DFORCE_${TARGET}")
 endif ()
 
+# On x86_64 build getarch with march=native. This is required to detect AVX512 support in getarch.
+if (X86_64)
+  set(GETARCH_FLAGS "${GETARCH_FLAGS} -march=native")
+endif ()
+
+# On x86 no AVX support is available
+if (X86 OR X86_64)
+if ((DEFINED BINARY AND BINARY EQUAL 32) OR ("$CMAKE_SIZEOF_VOID_P}" EQUAL "4"))
+  set(GETARCH_FLAGS "${GETARCH_FLAGS} -DNO_AVX -DNO_AVX2 -DNO_AVX512")
+endif ()
+endif ()
+
 if (INTERFACE64)
   message(STATUS "Using 64-bit integers.")
   set(GETARCH_FLAGS	"${GETARCH_FLAGS} -DUSE64BITINT")
@@ -136,10 +148,16 @@ endif ()
 
 if (USE_THREAD)
   message(STATUS "Multi-threading enabled with ${NUM_THREADS} threads.")
+else()
+  if (${USE_LOCKING})
+    set(CCOMMON_OPT "${CCOMMON_OPT} -DUSE_LOCKING")
+  endif ()
 endif ()
 
 include("${PROJECT_SOURCE_DIR}/cmake/prebuild.cmake")
-
+if (DEFINED BINARY)
+  message(STATUS "Compiling a ${BINARY}-bit binary.")
+endif ()
 if (NOT DEFINED NEED_PIC)
   set(NEED_PIC 1)
 endif ()
@@ -156,6 +174,9 @@ include("${PROJECT_SOURCE_DIR}/cmake/cc.cmake")
 if (NOT NOFORTRAN)
   # Fortran Compiler dependent settings
   include("${PROJECT_SOURCE_DIR}/cmake/fc.cmake")
+else ()
+set(NO_LAPACK 1)
+set(NO_LAPACKE 1)
 endif ()
 
 if (BINARY64)
@@ -181,12 +202,24 @@ if (NEED_PIC)
 endif ()
 
 if (DYNAMIC_ARCH)
-  set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_ARCH")
-  if (DYNAMIC_OLDER)
-    set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_OLDER")
+  if (X86 OR X86_64 OR ARM64 OR PPC)
+    set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_ARCH")
+    if (DYNAMIC_OLDER)
+      set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_OLDER")
+    endif ()
+  else ()
+    unset (DYNAMIC_ARCH)
+    message (STATUS "DYNAMIC_ARCH is not supported on the target architecture, removing")
   endif ()
 endif ()
 
+if (DYNAMIC_LIST)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_LIST")
+  foreach(DCORE ${DYNAMIC_LIST})
+    set(CCOMMON_OPT "${CCOMMON_OPT} -DDYN_${DCORE}")
+  endforeach ()
+endif ()
+
 if (NO_LAPACK)
   set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_LAPACK")
   #Disable LAPACK C interface
@@ -276,7 +309,7 @@ endif ()
 
 set(KERNELDIR	"${PROJECT_SOURCE_DIR}/kernel/${ARCH}")
 
-# TODO: nead to convert these Makefiles
+# TODO: need to convert these Makefiles
 # include ${PROJECT_SOURCE_DIR}/cmake/${ARCH}.cmake
 
 if (${CORE} STREQUAL "PPC440")

cmake/system_check.cmake

@@ -39,13 +39,21 @@ elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc.*|power.*|Power.*")
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "mips64.*")
   set(MIPS64 1)
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "amd64.*|x86_64.*|AMD64.*")
-  set(X86_64 1)
+  if("${CMAKE_SIZEOF_VOID_P}" EQUAL "8")
+    set(X86_64 1)
+  else()
+    set(X86 1)
+  endif()
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "i686.*|i386.*|x86.*|amd64.*|AMD64.*")
   set(X86 1)
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(arm.*|ARM.*)")
   set(ARM 1)
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64.*|AARCH64.*)")
-  set(ARM64 1)
+  if("${CMAKE_SIZEOF_VOID_P}" EQUAL "8")
+    set(ARM64 1)
+  else()
+    set(ARM 1)
+  endif()
 endif()
 
 if (X86_64)
@@ -78,7 +86,7 @@ endif()
 
 if (X86_64 OR X86)
   file(WRITE ${PROJECT_BINARY_DIR}/avx512.tmp "#include <immintrin.h>\n\nint main(void){ __asm__ volatile(\"vbroadcastss -4 * 4(%rsi), %zmm2\"); }")
-execute_process(COMMAND ${CMAKE_C_COMPILER} -march=skylake-avx512 -v -o ${PROJECT_BINARY_DIR}/avx512.o -x c ${PROJECT_BINARY_DIR}/avx512.tmp OUTPUT_QUIET ERROR_QUIET RESULT_VARIABLE NO_AVX512)
+execute_process(COMMAND ${CMAKE_C_COMPILER} -march=skylake-avx512 -c -v -o ${PROJECT_BINARY_DIR}/avx512.o -x c ${PROJECT_BINARY_DIR}/avx512.tmp OUTPUT_QUIET ERROR_QUIET RESULT_VARIABLE NO_AVX512)
 if (NO_AVX512 EQUAL 1)
 set (CCOMMON_OPT "${CCOMMON_OPT} -DNO_AVX512")
 endif()

cmake/utils.cmake

@@ -89,7 +89,7 @@ function(AllCombinations list_in absent_codes_in)
   set(CODES_OUT ${CODES_OUT} PARENT_SCOPE)
 endfunction ()
 
-# generates object files for each of the sources, using the BLAS naming scheme to pass the funciton name as a preprocessor definition
+# generates object files for each of the sources, using the BLAS naming scheme to pass the function name as a preprocessor definition
 # @param sources_in the source files to build from
 # @param defines_in (optional) preprocessor definitions that will be applied to all objects
 # @param name_in (optional) if this is set this name will be used instead of the filename. Use a * to indicate where the float character should go, if no star the character will be prepended.

common.h

@@ -85,6 +85,8 @@ extern "C" {
 
 #if !defined(_MSC_VER)
 #include <unistd.h>
+#elif _MSC_VER < 1900
+#define snprintf _snprintf
 #endif
 #include <time.h>
 
@@ -129,7 +131,7 @@ extern "C" {
 #include <time.h>
 #include <unistd.h>
 #include <math.h>
-#ifdef SMP
+#if defined(SMP) || defined(USE_LOCKING)
 #include <pthread.h>
 #endif
 #endif
@@ -198,7 +200,7 @@ extern "C" {
 #error "You can't specify both LOCK operation!"
 #endif
 
-#ifdef SMP
+#if defined(SMP) || defined(USE_LOCKING)
 #define USE_PTHREAD_LOCK
 #undef	USE_PTHREAD_SPINLOCK
 #endif
@@ -348,6 +350,11 @@ typedef int blasint;
 #endif
 #endif
 
+#ifdef POWER9
+#ifndef YIELDING
+#define YIELDING        __asm__ __volatile__ ("nop;nop;nop;nop;nop;nop;nop;nop;\n");
+#endif
+#endif
 
 /*
 #ifdef PILEDRIVER
@@ -439,7 +446,7 @@ please https://github.com/xianyi/OpenBLAS/issues/246
 typedef char env_var_t[MAX_PATH];
 #define readenv(p, n) 0
 #else
-#ifdef OS_WINDOWS
+#if defined(OS_WINDOWS) && !defined(OS_CYGWIN_NT)
 typedef char env_var_t[MAX_PATH];
 #define readenv(p, n) GetEnvironmentVariable((LPCTSTR)(n), (LPTSTR)(p), sizeof(p))
 #else

common_c.h

@@ -19,6 +19,7 @@
 #define	CDOTC_K			cdotc_k
 #define	CNRM2_K			cnrm2_k
 #define	CSCAL_K			cscal_k
+#define	CSUM_K			csum_k
 #define	CSWAP_K			cswap_k
 #define	CROT_K			csrot_k
 
@@ -249,6 +250,7 @@
 #define	CDOTC_K			gotoblas -> cdotc_k
 #define	CNRM2_K			gotoblas -> cnrm2_k
 #define	CSCAL_K			gotoblas -> cscal_k
+#define	CSUM_K			gotoblas -> csum_k
 #define	CSWAP_K			gotoblas -> cswap_k
 #define	CROT_K			gotoblas -> csrot_k
 

common_d.h

@@ -19,6 +19,7 @@
 #define	DDOTC_K			ddot_k
 #define	DNRM2_K			dnrm2_k
 #define	DSCAL_K			dscal_k
+#define	DSUM_K			dsum_k
 #define	DSWAP_K			dswap_k
 #define	DROT_K			drot_k
 
@@ -174,6 +175,7 @@
 #define	DDOTC_K			gotoblas -> ddot_k
 #define	DNRM2_K			gotoblas -> dnrm2_k
 #define	DSCAL_K			gotoblas -> dscal_k
+#define	DSUM_K			gotoblas -> dsum_k
 #define	DSWAP_K			gotoblas -> dswap_k
 #define	DROT_K			gotoblas -> drot_k
 

common_interface.h

@@ -122,6 +122,13 @@ xdouble BLASFUNC(qasum) (blasint *, xdouble *, blasint *);
 double BLASFUNC(dzasum)(blasint *, double *, blasint *);
 xdouble BLASFUNC(qxasum)(blasint *, xdouble *, blasint *);
 
+FLOATRET  BLASFUNC(ssum) (blasint *, float  *, blasint *);
+FLOATRET  BLASFUNC(scsum)(blasint *, float  *, blasint *);
+double BLASFUNC(dsum) (blasint *, double *, blasint *);
+xdouble BLASFUNC(qsum) (blasint *, xdouble *, blasint *);
+double BLASFUNC(dzsum)(blasint *, double *, blasint *);
+xdouble BLASFUNC(qxsum)(blasint *, xdouble *, blasint *);
+
 blasint    BLASFUNC(isamax)(blasint *, float  *, blasint *);
 blasint    BLASFUNC(idamax)(blasint *, double *, blasint *);
 blasint    BLASFUNC(iqamax)(blasint *, xdouble *, blasint *);

common_level1.h

@@ -100,6 +100,13 @@ float   casum_k (BLASLONG, float  *, BLASLONG);
 double  zasum_k (BLASLONG, double *, BLASLONG);
 xdouble xasum_k (BLASLONG, xdouble *, BLASLONG);
 
+float   ssum_k (BLASLONG, float  *, BLASLONG);
+double  dsum_k (BLASLONG, double *, BLASLONG);
+xdouble qsum_k (BLASLONG, xdouble *, BLASLONG);
+float   csum_k (BLASLONG, float  *, BLASLONG);
+double  zsum_k (BLASLONG, double *, BLASLONG);
+xdouble xsum_k (BLASLONG, xdouble *, BLASLONG);
+
 float   samax_k (BLASLONG, float  *, BLASLONG);
 double  damax_k (BLASLONG, double *, BLASLONG);
 xdouble qamax_k (BLASLONG, xdouble *, BLASLONG);

common_macro.h

@@ -66,6 +66,7 @@
 #define	DOTC_K			QDOTC_K
 #define	NRM2_K			QNRM2_K
 #define	SCAL_K			QSCAL_K
+#define	SUM_K			QSUM_K
 #define	SWAP_K			QSWAP_K
 #define	ROT_K			QROT_K
 
@@ -356,6 +357,7 @@
 #define	DOTC_K			DDOTC_K
 #define	NRM2_K			DNRM2_K
 #define	SCAL_K			DSCAL_K
+#define	SUM_K			DSUM_K
 #define	SWAP_K			DSWAP_K
 #define	ROT_K			DROT_K
 
@@ -658,6 +660,7 @@
 #define	DOTC_K			SDOTC_K
 #define	NRM2_K			SNRM2_K
 #define	SCAL_K			SSCAL_K
+#define	SUM_K			SSUM_K
 #define	SWAP_K			SSWAP_K
 #define	ROT_K			SROT_K
 
@@ -962,6 +965,7 @@
 #define	DOTC_K			XDOTC_K
 #define	NRM2_K			XNRM2_K
 #define	SCAL_K			XSCAL_K
+#define	SUM_K			XSUM_K
 #define	SWAP_K			XSWAP_K
 #define	ROT_K			XROT_K
 
@@ -1363,6 +1367,7 @@
 #define	DOTC_K			ZDOTC_K
 #define	NRM2_K			ZNRM2_K
 #define	SCAL_K			ZSCAL_K
+#define	SUM_K			ZSUM_K
 #define	SWAP_K			ZSWAP_K
 #define	ROT_K			ZROT_K
 
@@ -1785,6 +1790,7 @@
 #define	DOTC_K			CDOTC_K
 #define	NRM2_K			CNRM2_K
 #define	SCAL_K			CSCAL_K
+#define	SUM_K			CSUM_K
 #define	SWAP_K			CSWAP_K
 #define	ROT_K			CROT_K
 

common_param.h

@@ -63,6 +63,7 @@ BLASLONG (*ismin_k) (BLASLONG, float *, BLASLONG);
 
   float  (*snrm2_k) (BLASLONG, float *, BLASLONG);
   float  (*sasum_k) (BLASLONG, float *, BLASLONG);
+  float  (*ssum_k)  (BLASLONG, float *, BLASLONG);
   int    (*scopy_k) (BLASLONG, float *, BLASLONG, float *, BLASLONG);
   float  (*sdot_k)  (BLASLONG, float *, BLASLONG, float *, BLASLONG);
   double (*dsdot_k) (BLASLONG, float *, BLASLONG, float *, BLASLONG);
@@ -154,6 +155,7 @@ BLASLONG (*idmin_k) (BLASLONG, double *, BLASLONG);
 
   double (*dnrm2_k) (BLASLONG, double *, BLASLONG);
   double (*dasum_k) (BLASLONG, double *, BLASLONG);
+  double (*dsum_k)  (BLASLONG, double *, BLASLONG);
   int    (*dcopy_k) (BLASLONG, double *, BLASLONG, double *, BLASLONG);
   double (*ddot_k)  (BLASLONG, double *, BLASLONG, double *, BLASLONG);
   int    (*drot_k)  (BLASLONG, double *, BLASLONG, double *, BLASLONG, double, double);
@@ -245,6 +247,7 @@ BLASLONG (*iqmin_k) (BLASLONG, xdouble *, BLASLONG);
 
  xdouble (*qnrm2_k) (BLASLONG, xdouble *, BLASLONG);
  xdouble (*qasum_k) (BLASLONG, xdouble *, BLASLONG);
+ xdouble (*qsum_k)  (BLASLONG, xdouble *, BLASLONG);
   int    (*qcopy_k) (BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);
  xdouble (*qdot_k)  (BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);
   int    (*qrot_k)  (BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG, xdouble, xdouble);
@@ -332,6 +335,7 @@ BLASLONG (*icamin_k)(BLASLONG, float *, BLASLONG);
 
   float (*cnrm2_k) (BLASLONG, float *, BLASLONG);
   float (*casum_k) (BLASLONG, float *, BLASLONG);
+  float (*csum_k)  (BLASLONG, float *, BLASLONG);
   int    (*ccopy_k) (BLASLONG, float *, BLASLONG, float *, BLASLONG);
   openblas_complex_float (*cdotu_k) (BLASLONG, float *, BLASLONG, float *, BLASLONG);
   openblas_complex_float (*cdotc_k) (BLASLONG, float *, BLASLONG, float *, BLASLONG);
@@ -495,6 +499,7 @@ BLASLONG (*izamin_k)(BLASLONG, double *, BLASLONG);
 
   double (*znrm2_k) (BLASLONG, double *, BLASLONG);
   double (*zasum_k) (BLASLONG, double *, BLASLONG);
+  double (*zsum_k)  (BLASLONG, double *, BLASLONG);
   int    (*zcopy_k) (BLASLONG, double *, BLASLONG, double *, BLASLONG);
   openblas_complex_double (*zdotu_k) (BLASLONG, double *, BLASLONG, double *, BLASLONG);
   openblas_complex_double (*zdotc_k) (BLASLONG, double *, BLASLONG, double *, BLASLONG);
@@ -660,6 +665,7 @@ BLASLONG (*ixamin_k)(BLASLONG, xdouble *, BLASLONG);
 
   xdouble (*xnrm2_k) (BLASLONG, xdouble *, BLASLONG);
   xdouble (*xasum_k) (BLASLONG, xdouble *, BLASLONG);
+  xdouble (*xsum_k) (BLASLONG, xdouble *, BLASLONG);
   int    (*xcopy_k) (BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);
   openblas_complex_xdouble (*xdotu_k) (BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);
   openblas_complex_xdouble (*xdotc_k) (BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);

common_power.h

@@ -39,7 +39,7 @@
 #ifndef COMMON_POWER
 #define COMMON_POWER
 
-#if defined(POWER8)
+#if defined(POWER8) || defined(POWER9)
 #define MB		__asm__ __volatile__ ("eieio":::"memory")
 #define WMB		__asm__ __volatile__ ("eieio":::"memory")
 #else
@@ -241,7 +241,7 @@ static inline int blas_quickdivide(blasint x, blasint y){
 #define HAVE_PREFETCH
 #endif
 
-#if defined(POWER3) || defined(POWER6) || defined(PPCG4) || defined(CELL) || defined(POWER8)
+#if defined(POWER3) || defined(POWER6) || defined(PPCG4) || defined(CELL) || defined(POWER8) || defined(POWER9) || ( defined(PPC970) && ( defined(OS_DARWIN) || defined(OS_FREEBSD) ) )
 #define DCBT_ARG	0
 #else
 #define DCBT_ARG	8
@@ -263,7 +263,7 @@ static inline int blas_quickdivide(blasint x, blasint y){
 #define L1_PREFETCH	dcbtst
 #endif
 
-#if defined(POWER8)
+#if defined(POWER8) || defined(POWER9)
 #define L1_DUALFETCH
 #define L1_PREFETCHSIZE (16 + 128 * 100)
 #define L1_PREFETCH	dcbtst
@@ -499,7 +499,7 @@ static inline int blas_quickdivide(blasint x, blasint y){
 
 #if defined(ASSEMBLER) && !defined(NEEDPARAM)
 
-#ifdef OS_LINUX
+#if defined(OS_LINUX) || defined(OS_FREEBSD)
 #ifndef __64BIT__
 #define PROLOGUE \
 	.section .text;\
@@ -598,9 +598,14 @@ REALNAME:;\
 #ifndef __64BIT__
 #define PROLOGUE \
 	.machine "any";\
+	.toc;\
 	.globl .REALNAME;\
+	.globl REALNAME;\
+	.csect REALNAME[DS],3;\
+REALNAME:;\
+	.long .REALNAME, TOC[tc0], 0;\
 	.csect .text[PR],5;\
-.REALNAME:;
+.REALNAME:
 
 #define EPILOGUE \
 _section_.text:;\
@@ -611,9 +616,14 @@ _section_.text:;\
 
 #define PROLOGUE \
 	.machine "any";\
+	.toc;\
 	.globl .REALNAME;\
+	.globl REALNAME;\
+	.csect REALNAME[DS],3;\
+REALNAME:;\
+	.llong .REALNAME, TOC[tc0], 0;\
 	.csect .text[PR], 5;\
-.REALNAME:;
+.REALNAME:
 
 #define EPILOGUE \
 _section_.text:;\
@@ -774,7 +784,7 @@ Lmcount$lazy_ptr:
 
 #define HALT		mfspr	r0, 1023
 
-#ifdef OS_LINUX
+#if defined(OS_LINUX) || defined(OS_FREEBSD)
 #if defined(PPC440) || defined(PPC440FP2)
 #undef  MAX_CPU_NUMBER
 #define MAX_CPU_NUMBER 1
@@ -802,7 +812,7 @@ Lmcount$lazy_ptr:
 #define BUFFER_SIZE     (  2 << 20)
 #elif defined(PPC440FP2)
 #define BUFFER_SIZE     ( 16 << 20)
-#elif defined(POWER8)
+#elif defined(POWER8) || defined(POWER9)
 #define BUFFER_SIZE     ( 64 << 20)
 #else
 #define BUFFER_SIZE     ( 16 << 20)
@@ -819,7 +829,7 @@ Lmcount$lazy_ptr:
 #define MAP_ANONYMOUS MAP_ANON
 #endif
 
-#ifdef OS_LINUX
+#if defined(OS_LINUX) || defined(OS_FREEBSD)
 #ifndef __64BIT__
 #define FRAMESLOT(X) (((X) * 4) + 8)
 #else

common_q.h

@@ -19,6 +19,7 @@
 #define	QDOTC_K			qdot_k
 #define	QNRM2_K			qnrm2_k
 #define	QSCAL_K			qscal_k
+#define	QSUM_K			qsum_k
 #define	QSWAP_K			qswap_k
 #define	QROT_K			qrot_k
 
@@ -161,6 +162,7 @@
 #define	QDOTC_K			gotoblas -> qdot_k
 #define	QNRM2_K			gotoblas -> qnrm2_k
 #define	QSCAL_K			gotoblas -> qscal_k
+#define	QSUM_K			gotoblas -> qsum_k
 #define	QSWAP_K			gotoblas -> qswap_k
 #define	QROT_K			gotoblas -> qrot_k
 

common_s.h

@@ -12,6 +12,7 @@
 #define	ISMAX_K			ismax_k
 #define	ISMIN_K			ismin_k
 #define	SASUM_K			sasum_k
+#define	SSUM_K			ssum_k
 #define	SAXPYU_K		saxpy_k
 #define	SAXPYC_K		saxpy_k
 #define	SCOPY_K			scopy_k
@@ -170,6 +171,7 @@
 #define	ISMAX_K			gotoblas -> ismax_k
 #define	ISMIN_K			gotoblas -> ismin_k
 #define	SASUM_K			gotoblas -> sasum_k
+#define	SSUM_K			gotoblas -> ssum_k
 #define	SAXPYU_K		gotoblas -> saxpy_k
 #define	SAXPYC_K		gotoblas -> saxpy_k
 #define	SCOPY_K			gotoblas -> scopy_k

common_stackalloc.h

@@ -45,7 +45,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  * SIZE must be carefully chosen to be:
  * - as small as possible to maximize the number of stack allocation
  * - large enough to support all architectures and kernel
- * Chosing a too small SIZE will lead to a stack smashing.
+ * Choosing a SIZE too small will lead to a stack smashing.
  */
 #define STACK_ALLOC(SIZE, TYPE, BUFFER)                                        \
   /* make it volatile because some function (ex: dgemv_n.S) */                 \

common_x.h

@@ -19,6 +19,7 @@
 #define	XDOTC_K			xdotc_k
 #define	XNRM2_K			xnrm2_k
 #define	XSCAL_K			xscal_k
+#define	XSUM_K			xsum_k
 #define	XSWAP_K			xswap_k
 #define	XROT_K			xqrot_k
 
@@ -227,6 +228,7 @@
 #define	XDOTC_K			gotoblas -> xdotc_k
 #define	XNRM2_K			gotoblas -> xnrm2_k
 #define	XSCAL_K			gotoblas -> xscal_k
+#define	XSUM_K			gotoblas -> xsum_k
 #define	XSWAP_K			gotoblas -> xswap_k
 #define	XROT_K			gotoblas -> xqrot_k
 

common_x86.h

@@ -187,7 +187,7 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 	
   y = blas_quick_divide_table[y];
 
-  __asm__ __volatile__  ("mull %0" :"=d" (result) :"a"(x), "0" (y));
+  __asm__ __volatile__  ("mull %0" :"=d" (result), "+a"(x): "0" (y));
 
   return result;
 #endif
@@ -214,7 +214,7 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 #endif
 
 #if defined(PILEDRIVER) || defined(BULLDOZER) || defined(STEAMROLLER) || defined(EXCAVATOR)
-//Enable some optimazation for barcelona.
+//Enable some optimization for barcelona.
 #define BARCELONA_OPTIMIZATION
 #endif
 

common_x86_64.h

@@ -129,12 +129,13 @@ static __inline void cpuid(int op, int *eax, int *ebx, int *ecx, int *edx){
   *ecx=cpuinfo[2];
   *edx=cpuinfo[3];
 #else
-        __asm__ __volatile__("cpuid"
+        __asm__ __volatile__("mov $0, %%ecx;"
+			     "cpuid"
 			     : "=a" (*eax),
 			     "=b" (*ebx),
 			     "=c" (*ecx),
 			     "=d" (*edx)
-			     : "0" (op), "c"(0));
+			     : "0" (op));
 #endif
 }
 
@@ -210,7 +211,7 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 	
   y = blas_quick_divide_table[y];
 
-  __asm__ __volatile__  ("mull %0" :"=d" (result) :"a"(x), "0" (y));
+  __asm__ __volatile__  ("mull %0" :"=d" (result), "+a"(x) : "0" (y));
 
   return result;
 }
@@ -276,7 +277,7 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 #ifdef ASSEMBLER
 
 #if defined(PILEDRIVER) || defined(BULLDOZER) || defined(STEAMROLLER) || defined(EXCAVATOR)
-//Enable some optimazation for barcelona.
+//Enable some optimization for barcelona.
 #define BARCELONA_OPTIMIZATION
 #endif
 

common_z.h

@@ -19,6 +19,7 @@
 #define	ZDOTC_K			zdotc_k
 #define	ZNRM2_K			znrm2_k
 #define	ZSCAL_K			zscal_k
+#define	ZSUM_K			zsum_k
 #define	ZSWAP_K			zswap_k
 #define	ZROT_K			zdrot_k
 
@@ -249,6 +250,7 @@
 #define	ZDOTC_K			gotoblas -> zdotc_k
 #define	ZNRM2_K			gotoblas -> znrm2_k
 #define	ZSCAL_K			gotoblas -> zscal_k
+#define	ZSUM_K			gotoblas -> zsum_k
 #define	ZSWAP_K			gotoblas -> zswap_k
 #define	ZROT_K			gotoblas -> zdrot_k
 

cpp_thread_test/Makefile

@@ -0,0 +1,14 @@
+include ../Makefile.rule
+
+all :: dgemv_tester dgemm_tester
+
+dgemv_tester :
+	$(CXX) $(COMMON_OPT) -Wall -Wextra -Wshadow -fopenmp -std=c++11 dgemv_thread_safety.cpp ../libopenblas.a -lpthread -o dgemv_tester
+	./dgemv_tester
+
+dgemm_tester : dgemv_tester
+	$(CXX) $(COMMON_OPT) -Wall -Wextra -Wshadow -fopenmp -std=c++11 dgemm_thread_safety.cpp ../libopenblas.a -lpthread -o dgemm_tester
+	./dgemm_tester
+
+clean ::
+	rm -f dgemv_tester dgemm_tester

cpp_thread_test/cpp_thread_safety_common.h

@@ -0,0 +1,55 @@
+inline void pauser(){
+    /// a portable way to pause a program
+    std::string dummy;
+    std::cout << "Press enter to continue...";
+    std::getline(std::cin, dummy);
+}
+
+void FillMatrices(std::vector<std::vector<double>>& matBlock, std::mt19937_64& PRNG, std::uniform_real_distribution<double>& rngdist, const blasint randomMatSize, const uint32_t numConcurrentThreads, const uint32_t numMat){
+	for(uint32_t i=0; i<numMat; i++){
+		for(uint32_t j = 0; j < static_cast<uint32_t>(randomMatSize*randomMatSize); j++){
+			matBlock[i][j] = rngdist(PRNG);
+		}
+	}
+	for(uint32_t i=numMat; i<(numConcurrentThreads*numMat); i+=numMat){
+		for(uint32_t j=0; j<numMat; j++){
+			matBlock[i+j] = matBlock[j];
+		}
+	}
+}
+
+void FillVectors(std::vector<std::vector<double>>& vecBlock, std::mt19937_64& PRNG, std::uniform_real_distribution<double>& rngdist, const blasint randomMatSize, const uint32_t numConcurrentThreads, const uint32_t numVec){
+	for(uint32_t i=0; i<numVec; i++){
+		for(uint32_t j = 0; j < static_cast<uint32_t>(randomMatSize); j++){
+			vecBlock[i][j] = rngdist(PRNG);
+		}
+	}
+	for(uint32_t i=numVec; i<(numConcurrentThreads*numVec); i+=numVec){
+		for(uint32_t j=0; j<numVec; j++){
+			vecBlock[i+j] = vecBlock[j];
+		}
+	}
+}
+
+std::mt19937_64 InitPRNG(){
+	std::random_device rd;
+	std::mt19937_64 PRNG(rd()); //seed PRNG using /dev/urandom or similar OS provided RNG
+	std::uniform_real_distribution<double> rngdist{-1.0, 1.0};
+	//make sure the internal state of the PRNG is properly mixed by generating 10M random numbers
+	//PRNGs often have unreliable distribution uniformity and other statistical properties before their internal state is sufficiently mixed
+	for (uint32_t i=0;i<10000000;i++) rngdist(PRNG);
+	return PRNG;
+}
+
+void PrintMatrices(const std::vector<std::vector<double>>& matBlock, const blasint randomMatSize, const uint32_t numConcurrentThreads, const uint32_t numMat){
+	for (uint32_t i=0;i<numConcurrentThreads*numMat;i++){
+		std::cout<<i<<std::endl;
+		for (uint32_t j = 0; j < static_cast<uint32_t>(randomMatSize); j++){
+			for (uint32_t k = 0; k < static_cast<uint32_t>(randomMatSize); k++){
+				std::cout<<matBlock[i][j*randomMatSize + k]<<"  ";
+			}
+			std::cout<<std::endl;
+		}
+		std::cout<<std::endl;
+	}
+}

cpp_thread_test/dgemm_thread_safety.cpp

@@ -0,0 +1,92 @@
+#include <iostream>
+#include <vector>
+#include <random>
+#include <future>
+#include <omp.h>
+#include "../cblas.h"
+#include "cpp_thread_safety_common.h"
+
+void launch_cblas_dgemm(double* A, double* B, double* C, const blasint randomMatSize){
+	cblas_dgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, randomMatSize, randomMatSize, randomMatSize, 1.0, A, randomMatSize, B, randomMatSize, 0.1, C, randomMatSize);
+}
+
+int main(int argc, char* argv[]){
+	blasint randomMatSize = 1024; //dimension of the random square matrices used
+	uint32_t numConcurrentThreads = 52; //number of concurrent calls of the functions being tested
+	uint32_t numTestRounds = 16; //number of testing rounds before success exit
+	
+	if (argc > 4){
+		std::cout<<"ERROR: too many arguments for thread safety tester"<<std::endl;
+		abort();
+	}
+	
+	if(argc == 4){
+		std::vector<std::string> cliArgs;
+		for (int i = 1; i < argc; i++){
+			cliArgs.push_back(argv[i]);
+			std::cout<<argv[i]<<std::endl;
+		}
+		randomMatSize = std::stoul(cliArgs[0]);
+		numConcurrentThreads = std::stoul(cliArgs[1]);
+		numTestRounds = std::stoul(cliArgs[2]);
+	}
+	
+	std::uniform_real_distribution<double> rngdist{-1.0, 1.0};
+	std::vector<std::vector<double>> matBlock(numConcurrentThreads*3);
+	std::vector<std::future<void>> futureBlock(numConcurrentThreads);
+	
+	std::cout<<"*----------------------------*\n";
+	std::cout<<"| DGEMM thread safety tester |\n";
+	std::cout<<"*----------------------------*\n";
+	std::cout<<"Size of random matrices(N=M=K): "<<randomMatSize<<'\n';
+	std::cout<<"Number of concurrent calls into OpenBLAS : "<<numConcurrentThreads<<'\n';
+	std::cout<<"Number of testing rounds : "<<numTestRounds<<'\n';
+	std::cout<<"This test will need "<<(static_cast<uint64_t>(randomMatSize*randomMatSize)*numConcurrentThreads*3*8)/static_cast<double>(1024*1024)<<" MiB of RAM\n"<<std::endl;
+	
+	std::cout<<"Initializing random number generator..."<<std::flush;
+	std::mt19937_64 PRNG = InitPRNG();
+	std::cout<<"done\n";
+	
+	std::cout<<"Preparing to test CBLAS DGEMM thread safety\n";
+	std::cout<<"Allocating matrices..."<<std::flush;
+	for(uint32_t i=0; i<(numConcurrentThreads*3); i++){
+		matBlock[i].resize(randomMatSize*randomMatSize);
+	}
+	std::cout<<"done\n";
+	//pauser();
+	std::cout<<"Filling matrices with random numbers..."<<std::flush;
+	FillMatrices(matBlock, PRNG, rngdist, randomMatSize, numConcurrentThreads, 3);
+	//PrintMatrices(matBlock, randomMatSize, numConcurrentThreads, 3);
+	std::cout<<"done\n";
+	std::cout<<"Testing CBLAS DGEMM thread safety\n";
+	omp_set_num_threads(numConcurrentThreads);
+	for(uint32_t R=0; R<numTestRounds; R++){
+		std::cout<<"DGEMM round #"<<R<<std::endl;
+		std::cout<<"Launching "<<numConcurrentThreads<<" threads simultaneously using OpenMP..."<<std::flush;
+		#pragma omp parallel for default(none) shared(futureBlock, matBlock, randomMatSize, numConcurrentThreads)
+		for(uint32_t i=0; i<numConcurrentThreads; i++){
+			futureBlock[i] = std::async(std::launch::async, launch_cblas_dgemm, &matBlock[i*3][0], &matBlock[i*3+1][0], &matBlock[i*3+2][0], randomMatSize);
+			//launch_cblas_dgemm( &matBlock[i][0], &matBlock[i+1][0], &matBlock[i+2][0]);
+		}
+		std::cout<<"done\n";
+		std::cout<<"Waiting for threads to finish..."<<std::flush;
+		for(uint32_t i=0; i<numConcurrentThreads; i++){
+			futureBlock[i].get();
+		}
+		std::cout<<"done\n";
+		//PrintMatrices(matBlock, randomMatSize, numConcurrentThreads, 3);
+		std::cout<<"Comparing results from different threads..."<<std::flush;
+		for(uint32_t i=3; i<(numConcurrentThreads*3); i+=3){ //i is the index of matrix A, for a given thread
+			for(uint32_t j = 0; j < static_cast<uint32_t>(randomMatSize*randomMatSize); j++){
+				if (std::abs(matBlock[i+2][j] - matBlock[2][j]) > 1.0E-13){ //i+2 is the index of matrix C, for a given thread
+					std::cout<<"ERROR: one of the threads returned a different result! Index : "<<i+2<<std::endl;
+					std::cout<<"CBLAS DGEMM thread safety test FAILED!"<<std::endl;
+					return -1;
+				}
+			}
+		}
+		std::cout<<"OK!\n"<<std::endl;
+	}
+	std::cout<<"CBLAS DGEMM thread safety test PASSED!\n"<<std::endl;
+	return 0;
+}

cpp_thread_test/dgemv_thread_safety.cpp

@@ -0,0 +1,101 @@
+#include <iostream>
+#include <vector>
+#include <random>
+#include <future>
+#include <omp.h>
+#include "../cblas.h"
+#include "cpp_thread_safety_common.h"
+
+void launch_cblas_dgemv(double* A, double* x, double* y, const blasint randomMatSize){
+	const blasint inc = 1;
+	cblas_dgemv(CblasColMajor, CblasNoTrans, randomMatSize, randomMatSize, 1.0, A, randomMatSize, x, inc, 0.1, y, inc);
+}
+
+int main(int argc, char* argv[]){
+	blasint randomMatSize = 1024; //dimension of the random square matrices and vectors being used
+	uint32_t numConcurrentThreads = 52; //number of concurrent calls of the functions being tested
+	uint32_t numTestRounds = 16; //number of testing rounds before success exit
+	
+	if (argc > 4){
+		std::cout<<"ERROR: too many arguments for thread safety tester"<<std::endl;
+		abort();
+	}
+	if(argc == 4){
+		std::vector<std::string> cliArgs;
+		for (int i = 1; i < argc; i++){
+			cliArgs.push_back(argv[i]);
+			std::cout<<argv[i]<<std::endl;
+		}
+		randomMatSize = std::stoul(cliArgs.at(0));
+		numConcurrentThreads = std::stoul(cliArgs.at(1));
+		numTestRounds = std::stoul(cliArgs.at(2));
+	}
+	
+	std::uniform_real_distribution<double> rngdist{-1.0, 1.0};
+	std::vector<std::vector<double>> matBlock(numConcurrentThreads);
+	std::vector<std::vector<double>> vecBlock(numConcurrentThreads*2);
+	std::vector<std::future<void>> futureBlock(numConcurrentThreads);
+	
+	std::cout<<"*----------------------------*\n";
+	std::cout<<"| DGEMV thread safety tester |\n";
+	std::cout<<"*----------------------------*\n";
+	std::cout<<"Size of random matrices and vectors(N=M): "<<randomMatSize<<'\n';
+	std::cout<<"Number of concurrent calls into OpenBLAS : "<<numConcurrentThreads<<'\n';
+	std::cout<<"Number of testing rounds : "<<numTestRounds<<'\n';
+	std::cout<<"This test will need "<<((static_cast<uint64_t>(randomMatSize*randomMatSize)*numConcurrentThreads*8)+(static_cast<uint64_t>(randomMatSize)*numConcurrentThreads*8*2))/static_cast<double>(1024*1024)<<" MiB of RAM\n"<<std::endl;
+	
+	std::cout<<"Initializing random number generator..."<<std::flush;
+	std::mt19937_64 PRNG = InitPRNG();
+	std::cout<<"done\n";
+	
+	std::cout<<"Preparing to test CBLAS DGEMV thread safety\n";
+	std::cout<<"Allocating matrices..."<<std::flush;
+	for(uint32_t i=0; i<numConcurrentThreads; i++){
+		matBlock.at(i).resize(randomMatSize*randomMatSize);
+	}
+	std::cout<<"done\n";
+	std::cout<<"Allocating vectors..."<<std::flush;
+	for(uint32_t i=0; i<(numConcurrentThreads*2); i++){
+		vecBlock.at(i).resize(randomMatSize);
+	}
+	std::cout<<"done\n";
+	//pauser();
+	
+	std::cout<<"Filling matrices with random numbers..."<<std::flush;
+	FillMatrices(matBlock, PRNG, rngdist, randomMatSize, numConcurrentThreads, 1);
+	//PrintMatrices(matBlock, randomMatSize, numConcurrentThreads);
+	std::cout<<"done\n";
+	std::cout<<"Filling vectors with random numbers..."<<std::flush;
+	FillVectors(vecBlock, PRNG, rngdist, randomMatSize, numConcurrentThreads, 2);
+	std::cout<<"done\n";
+	
+	std::cout<<"Testing CBLAS DGEMV thread safety"<<std::endl;
+	omp_set_num_threads(numConcurrentThreads);
+	for(uint32_t R=0; R<numTestRounds; R++){
+		std::cout<<"DGEMV round #"<<R<<std::endl;
+		std::cout<<"Launching "<<numConcurrentThreads<<" threads simultaneously using OpenMP..."<<std::flush;
+		#pragma omp parallel for default(none) shared(futureBlock, matBlock, vecBlock, randomMatSize, numConcurrentThreads)
+		for(uint32_t i=0; i<numConcurrentThreads; i++){
+			futureBlock[i] = std::async(std::launch::async, launch_cblas_dgemv, &matBlock[i][0], &vecBlock[i*2][0], &vecBlock[i*2+1][0], randomMatSize);
+		}
+		std::cout<<"done\n";
+		std::cout<<"Waiting for threads to finish..."<<std::flush;
+		for(uint32_t i=0; i<numConcurrentThreads; i++){
+			futureBlock[i].get();
+		}
+		std::cout<<"done\n";
+		std::cout<<"Comparing results from different threads..."<<std::flush;
+		for(uint32_t i=2; i<(numConcurrentThreads*2); i+=2){ //i is the index of vector x, for a given thread
+			for(uint32_t j = 0; j < static_cast<uint32_t>(randomMatSize); j++){
+				if (std::abs(vecBlock[i+1][j] - vecBlock[1][j]) > 1.0E-13){ //i+1 is the index of vector y, for a given thread
+					std::cout<<"ERROR: one of the threads returned a different result! Index : "<<i+1<<std::endl;
+					std::cout<<"CBLAS DGEMV thread safety test FAILED!"<<std::endl;
+					return -1;
+				}
+			}
+		}
+		std::cout<<"OK!\n"<<std::endl;
+	}
+	std::cout<<"CBLAS DGEMV thread safety test PASSED!\n"<<std::endl;
+	return 0;
+}

cpuid_arm64.c

@@ -39,6 +39,8 @@
 // Cavium
 #define CPU_THUNDERX      7
 #define CPU_THUNDERX2T99  8
+//Hisilicon
+#define CPU_TSV110        9
 
 static char *cpuname[] = {
   "UNKNOWN",
@@ -49,7 +51,8 @@ static char *cpuname[] = {
   "CORTEXA73",
   "FALKOR",
   "THUNDERX",
-  "THUNDERX2T99"
+  "THUNDERX2T99",
+  "TSV110"
 };
 
 static char *cpuname_lower[] = {
@@ -61,7 +64,8 @@ static char *cpuname_lower[] = {
   "cortexa73",
   "falkor",
   "thunderx",
-  "thunderx2t99"
+  "thunderx2t99",
+  "tsv110"
 };
 
 int get_feature(char *search)
@@ -145,6 +149,9 @@ int detect(void)
 			return CPU_THUNDERX;
     else if (strstr(cpu_implementer, "0x43") && strstr(cpu_part, "0x0af"))
 			return CPU_THUNDERX2T99;
+    // HiSilicon
+    else if (strstr(cpu_implementer, "0x48") && strstr(cpu_part, "0xd01"))
+                        return CPU_TSV110;
 	}
 
 	p = (char *) NULL ;
@@ -286,6 +293,21 @@ void get_cpuconfig(void)
 			printf("#define DTB_DEFAULT_ENTRIES  64       \n");
 			printf("#define DTB_SIZE             4096     \n");
 			break;
+			
+		case CPU_TSV110:
+			printf("#define TSV110                        \n");
+			printf("#define L1_CODE_SIZE         65536    \n");
+			printf("#define L1_CODE_LINESIZE     64       \n");
+			printf("#define L1_CODE_ASSOCIATIVE  4        \n");
+			printf("#define L1_DATA_SIZE         65536    \n");
+			printf("#define L1_DATA_LINESIZE     64       \n");
+			printf("#define L1_DATA_ASSOCIATIVE  4        \n");
+			printf("#define L2_SIZE              524228   \n");
+			printf("#define L2_LINESIZE          64       \n");
+			printf("#define L2_ASSOCIATIVE       8        \n");
+			printf("#define DTB_DEFAULT_ENTRIES  64       \n");
+			printf("#define DTB_SIZE             4096     \n");
+			break;	
 	}
 }
 

cpuid_power.c

@@ -94,7 +94,7 @@ char *corename[] = {
   "CELL",
   "PPCG4",
   "POWER8",
-  "POWER8"   	
+  "POWER9"   	
 };
 
 int detect(void){
@@ -124,7 +124,7 @@ int detect(void){
   if (!strncasecmp(p, "POWER6", 6)) return CPUTYPE_POWER6;
   if (!strncasecmp(p, "POWER7", 6)) return CPUTYPE_POWER6;
   if (!strncasecmp(p, "POWER8", 6)) return CPUTYPE_POWER8;
-  if (!strncasecmp(p, "POWER9", 6)) return CPUTYPE_POWER8;
+  if (!strncasecmp(p, "POWER9", 6)) return CPUTYPE_POWER9;
   if (!strncasecmp(p, "Cell",   4)) return CPUTYPE_CELL;
   if (!strncasecmp(p, "7447",   4)) return CPUTYPE_PPCG4;
 
@@ -156,7 +156,7 @@ int detect(void){
   if (!strncasecmp(p, "POWER6", 6)) return CPUTYPE_POWER6;
   if (!strncasecmp(p, "POWER7", 6)) return CPUTYPE_POWER6;
   if (!strncasecmp(p, "POWER8", 6)) return CPUTYPE_POWER8;
-  if (!strncasecmp(p, "POWER9", 6)) return CPUTYPE_POWER8;
+  if (!strncasecmp(p, "POWER9", 6)) return CPUTYPE_POWER9;
   if (!strncasecmp(p, "Cell",   4)) return CPUTYPE_CELL;
   if (!strncasecmp(p, "7447",   4)) return CPUTYPE_PPCG4;
   return CPUTYPE_POWER5;
@@ -180,7 +180,7 @@ int id;
 __asm __volatile("mfpvr %0" : "=r"(id));
 switch ( id >> 16 ) {
   case 0x4e: // POWER9
-    return CPUTYPE_POWER8;
+    return CPUTYPE_POWER9;
     break;
   case 0x4d:
   case 0x4b: // POWER8/8E 

cpuid_x86.c

@@ -1359,6 +1359,8 @@ int get_cpuname(void){
 	    return CPUTYPE_NEHALEM;
 	case 12:
 	    // Apollo Lake
+	case 15:
+	    // Denverton		
 	    return CPUTYPE_NEHALEM;
 	}
 	break;
@@ -1376,9 +1378,9 @@ int get_cpuname(void){
         }
       break;  
       case 9:
-      case 8: 
+      case 8:      
         switch (model) {
-	case 14: // Kaby Lake
+	case 14: // Kaby Lake and refreshes
           if(support_avx2())
             return CPUTYPE_HASWELL;
           if(support_avx())

cpuid_zarch.c

@@ -27,9 +27,9 @@
 
 #include <string.h>
 
-#define CPU_GENERIC    	0
-#define CPU_Z13       	1
-#define CPU_Z14       	2
+#define CPU_GENERIC     0
+#define CPU_Z13         1
+#define CPU_Z14         2
 
 static char *cpuname[] = {
   "ZARCH_GENERIC",
@@ -64,10 +64,8 @@ int detect(void)
 
   if (strstr(p, "2964")) return CPU_Z13;
   if (strstr(p, "2965")) return CPU_Z13;
-
-  /* detect z14, but fall back to z13 */
-  if (strstr(p, "3906")) return CPU_Z13;
-  if (strstr(p, "3907")) return CPU_Z13;
+  if (strstr(p, "3906")) return CPU_Z14;
+  if (strstr(p, "3907")) return CPU_Z14;
 
   return CPU_GENERIC;
 }
@@ -116,7 +114,14 @@ void get_cpuconfig(void)
 	  break;
 	case CPU_Z14:
 	  printf("#define Z14\n");
+	  printf("#define L1_DATA_SIZE 131072\n");
+	  printf("#define L1_DATA_LINESIZE 256\n");
+	  printf("#define L1_DATA_ASSOCIATIVE 8\n");
+	  printf("#define L2_SIZE 4194304\n");
+	  printf("#define L2_LINESIZE 256\n");
+	  printf("#define L2_ASSOCIATIVE 8\n");
 	  printf("#define DTB_DEFAULT_ENTRIES 64\n");
+	  printf("#define DTB_SIZE 4096\n");
 	  break;
 	}
 }

ctest.c

@@ -113,7 +113,7 @@ ARCH_X86
 ARCH_X86_64
 #endif
 
-#if defined(__powerpc___) || defined(__PPC__) || defined(_POWER)
+#if defined(__powerpc___) || defined(__PPC__) || defined(_POWER) || defined(__POWERPC__)
 ARCH_POWER
 #endif
 

ctest/c_cblat1.f

@@ -577,7 +577,7 @@
       SUBROUTINE STEST1(SCOMP1,STRUE1,SSIZE,SFAC)
 *     ************************* STEST1 *****************************
 *
-*     THIS IS AN INTERFACE SUBROUTINE TO ACCOMODATE THE FORTRAN
+*     THIS IS AN INTERFACE SUBROUTINE TO ACCOMMODATE THE FORTRAN
 *     REQUIREMENT THAT WHEN A DUMMY ARGUMENT IS AN ARRAY, THE
 *     ACTUAL ARGUMENT MUST ALSO BE AN ARRAY OR AN ARRAY ELEMENT.
 *

ctest/c_dblat1.f

@@ -653,7 +653,7 @@
       SUBROUTINE STEST1(SCOMP1,STRUE1,SSIZE,SFAC)
 *     ************************* STEST1 *****************************
 *
-*     THIS IS AN INTERFACE SUBROUTINE TO ACCOMODATE THE FORTRAN
+*     THIS IS AN INTERFACE SUBROUTINE TO ACCOMMODATE THE FORTRAN
 *     REQUIREMENT THAT WHEN A DUMMY ARGUMENT IS AN ARRAY, THE
 *     ACTUAL ARGUMENT MUST ALSO BE AN ARRAY OR AN ARRAY ELEMENT.
 *

ctest/c_sblat1.f

@@ -653,7 +653,7 @@
       SUBROUTINE STEST1(SCOMP1,STRUE1,SSIZE,SFAC)
 *     ************************* STEST1 *****************************
 *
-*     THIS IS AN INTERFACE SUBROUTINE TO ACCOMODATE THE FORTRAN
+*     THIS IS AN INTERFACE SUBROUTINE TO ACCOMMODATE THE FORTRAN
 *     REQUIREMENT THAT WHEN A DUMMY ARGUMENT IS AN ARRAY, THE
 *     ACTUAL ARGUMENT MUST ALSO BE AN ARRAY OR AN ARRAY ELEMENT.
 *

ctest/c_zblat1.f

@@ -577,7 +577,7 @@
       SUBROUTINE STEST1(SCOMP1,STRUE1,SSIZE,SFAC)
 *     ************************* STEST1 *****************************
 *
-*     THIS IS AN INTERFACE SUBROUTINE TO ACCOMODATE THE FORTRAN
+*     THIS IS AN INTERFACE SUBROUTINE TO ACCOMMODATE THE FORTRAN
 *     REQUIREMENT THAT WHEN A DUMMY ARGUMENT IS AN ARRAY, THE
 *     ACTUAL ARGUMENT MUST ALSO BE AN ARRAY OR AN ARRAY ELEMENT.
 *

driver/level2/trmv_thread.c

@@ -346,7 +346,7 @@ int CNAME(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *x, BLASLONG incx, FLOAT *bu
 
     range_m[MAX_CPU_NUMBER - num_cpu - 1] = range_m[MAX_CPU_NUMBER - num_cpu] - width;
     range_n[num_cpu] = num_cpu * (((m + 15) & ~15) + 16);
-    if (range_n[num_cpu] > m) range_n[num_cpu] = m;
+    if (range_n[num_cpu] > m * num_cpu) range_n[num_cpu] = m * num_cpu;
 
     queue[num_cpu].mode    = mode;
     queue[num_cpu].routine = trmv_kernel;
@@ -386,7 +386,7 @@ int CNAME(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *x, BLASLONG incx, FLOAT *bu
 
     range_m[num_cpu + 1] = range_m[num_cpu] + width;
     range_n[num_cpu] = num_cpu * (((m + 15) & ~15) + 16);
-    if (range_n[num_cpu] > m) range_n[num_cpu] = m;
+    if (range_n[num_cpu] > m * num_cpu) range_n[num_cpu] = m * num_cpu;
 
     queue[num_cpu].mode    = mode;
     queue[num_cpu].routine = trmv_kernel;

driver/others/Makefile

@@ -18,8 +18,12 @@ ifeq ($(DYNAMIC_ARCH), 1)
 ifeq ($(ARCH),arm64)
 COMMONOBJS	+=  dynamic_arm64.$(SUFFIX)
 else
+ifeq ($(ARCH),power)
+COMMONOBJS	+=  dynamic_power.$(SUFFIX)
+else
 COMMONOBJS	+=  dynamic.$(SUFFIX)
 endif
+endif
 else
 COMMONOBJS	+=  parameter.$(SUFFIX)
 endif
@@ -78,8 +82,12 @@ ifeq ($(DYNAMIC_ARCH), 1)
 ifeq ($(ARCH),arm64)
 HPLOBJS = memory.$(SUFFIX) xerbla.$(SUFFIX) dynamic_arm64.$(SUFFIX)
 else
+ifeq ($(ARCH),power)
+HPLOBJS = memory.$(SUFFIX) xerbla.$(SUFFIX) dynamic_power.$(SUFFIX)
+else
 HPLOBJS = memory.$(SUFFIX) xerbla.$(SUFFIX) dynamic.$(SUFFIX)
 endif
+endif
 else
 HPLOBJS = memory.$(SUFFIX) xerbla.$(SUFFIX) parameter.$(SUFFIX)
 endif

driver/others/blas_server.c

@@ -109,7 +109,7 @@ extern unsigned int openblas_thread_timeout();
 /* equal to "OMP_NUM_THREADS - 1" and thread only wakes up when     */
 /* jobs is queued.                                                  */
 
-/* We need this grobal for cheking if initialization is finished.   */
+/* We need this global for checking if initialization is finished.  */
 int blas_server_avail   __attribute__((aligned(ATTRIBUTE_SIZE))) = 0;
 
 /* Local Variables */
@@ -150,8 +150,8 @@ static unsigned int thread_timeout = (1U << (THREAD_TIMEOUT));
 
 #ifdef MONITOR
 
-/* Monitor is a function to see thread's status for every seconds. */
-/* Usually it turns off and it's for debugging.                    */
+/* Monitor is a function to see thread's status for every second. */
+/* Usually it turns off and it's for debugging.                   */
 
 static pthread_t      monitor_thread;
 static int main_status[MAX_CPU_NUMBER];

driver/others/blas_server_win32.c

@@ -50,7 +50,7 @@
 
 /* This is a thread implementation for Win32 lazy implementation */
 
-/* Thread server common infomation */
+/* Thread server common information */
 typedef struct{
   CRITICAL_SECTION lock;
   HANDLE filled;
@@ -61,7 +61,7 @@ typedef struct{
 
 } blas_pool_t;
 
-/* We need this global for cheking if initialization is finished.   */
+/* We need this global for checking if initialization is finished.   */
 int blas_server_avail = 0;
 
 /* Local Variables */
@@ -461,13 +461,18 @@ int BLASFUNC(blas_thread_shutdown)(void){
     SetEvent(pool.killed);
 
     for(i = 0; i < blas_num_threads - 1; i++){
+      // Could also just use WaitForMultipleObjects
       WaitForSingleObject(blas_threads[i], 5);  //INFINITE);
 #ifndef OS_WINDOWSSTORE
 // TerminateThread is only available with WINAPI_DESKTOP and WINAPI_SYSTEM not WINAPI_APP in UWP
       TerminateThread(blas_threads[i],0);
 #endif
+      CloseHandle(blas_threads[i]);
     }
 
+    CloseHandle(pool.filled);
+    CloseHandle(pool.killed);
+
     blas_server_avail = 0;
   }
 

driver/others/dynamic.c

@@ -322,7 +322,7 @@ int support_avx2(){
 }
 
 int support_avx512(){
-#ifndef NO_AVX512
+#if !defined(NO_AVX) && !defined(NO_AVX512)
   int eax, ebx, ecx, edx;
   int ret=0;
 
@@ -566,8 +566,8 @@ static gotoblas_t *get_coretype(void){
 	    return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
 	  }
 	}
-	//Apollo Lake
-	if (model == 12) { 
+	//Apollo Lake or Denverton
+	if (model == 12 || model == 15) { 
 	  return &gotoblas_NEHALEM;
 	}	
 	return NULL;

driver/others/dynamic_power.c

@@ -0,0 +1,102 @@
+
+#include "common.h"
+
+extern gotoblas_t gotoblas_POWER6;
+extern gotoblas_t gotoblas_POWER8;
+extern gotoblas_t gotoblas_POWER9;
+
+extern void openblas_warning(int verbose, const char *msg);
+
+static char *corename[] = {
+	"unknown",
+	"POWER6",
+	"POWER8",
+	"POWER9"
+};
+
+#define NUM_CORETYPES 4
+
+char *gotoblas_corename(void) {
+	if (gotoblas == &gotoblas_POWER6)	return corename[1];
+	if (gotoblas == &gotoblas_POWER8)	return corename[2];
+	if (gotoblas == &gotoblas_POWER9)	return corename[3];
+	return corename[0];
+}
+
+static gotoblas_t *get_coretype(void) {
+
+	if (__builtin_cpu_is("power6") || __builtin_cpu_is("power6x"))
+		return &gotoblas_POWER6;
+	if (__builtin_cpu_is("power8"))
+		return &gotoblas_POWER8;
+	if (__builtin_cpu_is("power9"))
+		return &gotoblas_POWER9;
+	return NULL;
+}
+
+static gotoblas_t *force_coretype(char * coretype) {
+
+	int i ;
+	int found = -1;
+	char message[128];
+
+	for ( i = 0 ; i < NUM_CORETYPES; i++)
+	{
+		if (!strncasecmp(coretype, corename[i], 20))
+		{
+			found = i;
+			break;
+		}
+	}
+
+	switch (found)
+	{
+	case  1: return (&gotoblas_POWER6);
+	case  2: return (&gotoblas_POWER8);
+	case  3: return (&gotoblas_POWER9);
+	default: return NULL;
+	}
+	snprintf(message, 128, "Core not found: %s\n", coretype);
+	openblas_warning(1, message);
+}
+
+void gotoblas_dynamic_init(void) {
+
+	char coremsg[128];
+	char coren[22];
+	char *p;
+
+
+	if (gotoblas) return;
+
+	p = getenv("OPENBLAS_CORETYPE");
+	if ( p )
+	{
+		gotoblas = force_coretype(p);
+	}
+	else
+	{
+		gotoblas = get_coretype();
+	}
+
+	if (gotoblas == NULL)
+	{
+		snprintf(coremsg, 128, "Falling back to POWER8 core\n");
+		openblas_warning(1, coremsg);
+		gotoblas = &gotoblas_POWER8;
+	}
+
+	if (gotoblas && gotoblas -> init) {
+		strncpy(coren,gotoblas_corename(),20);
+		sprintf(coremsg, "Core: %s\n",coren);
+		openblas_warning(2, coremsg);
+		gotoblas -> init();
+	} else {
+		openblas_warning(0, "OpenBLAS : Architecture Initialization failed. No initialization function found.\n");
+		exit(1);
+	}
+}
+
+void gotoblas_dynamic_quit(void) {
+	gotoblas = NULL;
+}

driver/others/init.c

@@ -765,7 +765,7 @@ int gotoblas_set_affinity(int pos) {
 
   int mynode = 1;
 
-  /* if number of threads is larger than inital condition */
+  /* if number of threads is larger than initial condition */
   if (pos < 0) {
       sched_setaffinity(0, sizeof(cpu_orig_mask), &cpu_orig_mask[0]);
       return 0;
@@ -857,7 +857,14 @@ void gotoblas_affinity_init(void) {
   common -> shmid = pshmid;
 
   if (common -> magic != SH_MAGIC) {
+
+#if defined(__GLIBC_PREREQ)
+#if __GLIBC_PREREQ(2, 7)
     cpu_set_t *cpusetp;
+#else
+    cpu_set_t cpuset;
+#endif
+#endif    
     int nums;
     int ret;
 
@@ -890,7 +897,7 @@ void gotoblas_affinity_init(void) {
     }
     CPU_FREE(cpusetp);
 #else
-    ret = sched_getaffinity(0,sizeof(cpu_set_t), cpusetp);
+    ret = sched_getaffinity(0,sizeof(cpu_set_t), &cpuset);
     if (ret!=0) {
         common->num_procs = nums;
     } else {
@@ -898,11 +905,11 @@ void gotoblas_affinity_init(void) {
     int i;
     int n = 0;
     for (i=0;i<nums;i++)
-        if (CPU_ISSET(i,cpusetp)) n++;
+        if (CPU_ISSET(i,&cpuset)) n++;
     common->num_procs = n;
     }
 #else
-    common->num_procs = CPU_COUNT(sizeof(cpu_set_t),cpusetp);
+    common->num_procs = CPU_COUNT(&cpuset);
     }
 #endif
 

driver/others/memory.c

@@ -198,45 +198,68 @@ int get_num_procs(void);
 #else
 int get_num_procs(void) {
   static int nums = 0;
-cpu_set_t *cpusetp;
-size_t size;
-int ret;
-int i,n;
+  cpu_set_t cpuset,*cpusetp;
+  size_t size;
+  int ret;
+
+#if defined(__GLIBC_PREREQ)
+#if !__GLIBC_PREREQ(2, 7)
+  int i;
+#if !__GLIBC_PREREQ(2, 6)
+  int n;
+#endif
+#endif
+#endif
 
   if (!nums) nums = sysconf(_SC_NPROCESSORS_CONF);
 #if !defined(OS_LINUX)
-     return nums;
+  return nums;
 #endif
 
 #if !defined(__GLIBC_PREREQ)
-   return nums;
+  return nums;
 #else
  #if !__GLIBC_PREREQ(2, 3)
-   return nums;
+  return nums;
  #endif
 
  #if !__GLIBC_PREREQ(2, 7)
-  ret = sched_getaffinity(0,sizeof(cpu_set_t), cpusetp);
+  ret = sched_getaffinity(0,sizeof(cpuset), &cpuset);
   if (ret!=0) return nums;
   n=0;
   #if !__GLIBC_PREREQ(2, 6)
   for (i=0;i<nums;i++)
-     if (CPU_ISSET(i,cpusetp)) n++;
+     if (CPU_ISSET(i,&cpuset)) n++;
   nums=n;
   #else
-  nums = CPU_COUNT(sizeof(cpu_set_t),cpusetp);
+  nums = CPU_COUNT(sizeof(cpuset),&cpuset);
   #endif
   return nums;
  #else
-  cpusetp = CPU_ALLOC(nums);
-  if (cpusetp == NULL) return nums;
-  size = CPU_ALLOC_SIZE(nums);
-  ret = sched_getaffinity(0,size,cpusetp);
-  if (ret!=0) return nums;
-  ret = CPU_COUNT_S(size,cpusetp);
-  if (ret > 0 && ret < nums) nums = ret;
-  CPU_FREE(cpusetp);
-  return nums;
+  if (nums >= CPU_SETSIZE) {
+    cpusetp = CPU_ALLOC(nums);
+      if (cpusetp == NULL) {
+        return nums;
+      }
+    size = CPU_ALLOC_SIZE(nums);
+    ret = sched_getaffinity(0,size,cpusetp);
+    if (ret!=0) {
+      CPU_FREE(cpusetp);
+      return nums;
+    }
+    ret = CPU_COUNT_S(size,cpusetp);
+    if (ret > 0 && ret < nums) nums = ret;	
+    CPU_FREE(cpusetp);
+    return nums;
+  } else {
+    ret = sched_getaffinity(0,sizeof(cpuset),&cpuset);
+    if (ret!=0) {
+      return nums;
+    }
+    ret = CPU_COUNT(&cpuset);
+    if (ret > 0 && ret < nums) nums = ret;	
+    return nums;
+  }
  #endif
 #endif
 }
@@ -1290,6 +1313,13 @@ void blas_memory_free_nolock(void * map_address) {
   free(map_address);
 }
 
+#ifdef SMP
+void blas_thread_memory_cleanup(void) {
+    blas_memory_cleanup((void*)get_memory_table());
+}
+#endif
+
+
 void blas_shutdown(void){
 #ifdef SMP
   BLASFUNC(blas_thread_shutdown)();
@@ -1299,7 +1329,7 @@ void blas_shutdown(void){
   /* Only cleanupIf we were built for threading and TLS was initialized */
   if (local_storage_key)
 #endif
-    blas_memory_cleanup((void*)get_memory_table());
+    blas_thread_memory_cleanup();
 
 #ifdef SEEK_ADDRESS
   base_address      = 0UL;
@@ -1529,7 +1559,7 @@ BOOL APIENTRY DllMain(HMODULE hModule, DWORD  ul_reason_for_call, LPVOID lpReser
       break;
     case DLL_THREAD_DETACH:
 #if defined(SMP)
-      blas_memory_cleanup((void*)get_memory_table());
+      blas_thread_memory_cleanup();
 #endif
       break;
     case DLL_PROCESS_DETACH:
@@ -1592,6 +1622,7 @@ void gotoblas_dummy_for_PGI(void) {
   gotoblas_init();
   gotoblas_quit();
 
+#if __PGIC__ < 19
 #if 0
   asm ("\t.section\t.ctors,\"aw\",@progbits; .align 8; .quad gotoblas_init; .section .text");
   asm ("\t.section\t.dtors,\"aw\",@progbits; .align 8; .quad gotoblas_quit; .section .text");
@@ -1599,13 +1630,16 @@ void gotoblas_dummy_for_PGI(void) {
   asm (".section .init,\"ax\"; call gotoblas_init@PLT; .section .text");
   asm (".section .fini,\"ax\"; call gotoblas_quit@PLT; .section .text");
 #endif
+#endif
 }
 #endif
 
 #else
+/* USE_TLS / COMPILE_TLS not set */
+
 #include <errno.h>
 
-#ifdef OS_WINDOWS
+#if defined(OS_WINDOWS) && !defined(OS_CYGWIN_NT)
 #define ALLOC_WINDOWS
 #ifndef MEM_LARGE_PAGES
 #define MEM_LARGE_PAGES  0x20000000
@@ -1619,7 +1653,7 @@ void gotoblas_dummy_for_PGI(void) {
 #include <stdio.h>
 #include <fcntl.h>
 
-#ifndef OS_WINDOWS
+#if !defined(OS_WINDOWS) || defined(OS_CYGWIN_NT)
 #include <sys/mman.h>
 #ifndef NO_SYSV_IPC
 #include <sys/shm.h>
@@ -1639,7 +1673,7 @@ void gotoblas_dummy_for_PGI(void) {
 #include <sys/resource.h>
 #endif
 
-#if defined(OS_FREEBSD) || defined(OS_DARWIN)
+#if defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN)
 #include <sys/sysctl.h>
 #include <sys/resource.h>
 #endif
@@ -1678,9 +1712,12 @@ void gotoblas_dummy_for_PGI(void) {
 #elif (defined(OS_DARWIN) || defined(OS_SUNOS)) && defined(C_GCC)
 #define CONSTRUCTOR	__attribute__ ((constructor))
 #define DESTRUCTOR	__attribute__ ((destructor))
-#else
+#elif __GNUC__ && INIT_PRIORITY && ((GCC_VERSION >= 40300) || (CLANG_VERSION >= 20900))
 #define CONSTRUCTOR	__attribute__ ((constructor(101)))
 #define DESTRUCTOR	__attribute__ ((destructor(101)))
+#else
+#define CONSTRUCTOR	__attribute__ ((constructor))
+#define DESTRUCTOR	__attribute__ ((destructor))
 #endif
 
 #ifdef DYNAMIC_ARCH
@@ -1704,45 +1741,70 @@ void goto_set_num_threads(int num_threads) {};
 int get_num_procs(void);
 #else
 int get_num_procs(void) {
+
   static int nums = 0;
-cpu_set_t *cpusetp;
-size_t size;
-int ret;
-int i,n;
+  cpu_set_t cpuset,*cpusetp;
+  size_t size;
+  int ret;
+
+#if defined(__GLIBC_PREREQ)
+#if !__GLIBC_PREREQ(2, 7)
+  int i;
+#if !__GLIBC_PREREQ(2, 6)
+  int n;
+#endif
+#endif
+#endif
 
   if (!nums) nums = sysconf(_SC_NPROCESSORS_CONF);
 #if !defined(OS_LINUX)
-     return nums;
+  return nums;
 #endif
 
 #if !defined(__GLIBC_PREREQ)
-   return nums;
+  return nums;
 #else
  #if !__GLIBC_PREREQ(2, 3)
-   return nums;
+  return nums;
  #endif
 
  #if !__GLIBC_PREREQ(2, 7)
-  ret = sched_getaffinity(0,sizeof(cpu_set_t), cpusetp);
+  ret = sched_getaffinity(0,sizeof(cpuset), &cpuset);
   if (ret!=0) return nums;
   n=0;
   #if !__GLIBC_PREREQ(2, 6)
   for (i=0;i<nums;i++)
-     if (CPU_ISSET(i,cpusetp)) n++;
+     if (CPU_ISSET(i,&cpuset)) n++;
   nums=n;
   #else
-  nums = CPU_COUNT(sizeof(cpu_set_t),cpusetp);
+  nums = CPU_COUNT(sizeof(cpuset),&cpuset);
   #endif
   return nums;
  #else
-  cpusetp = CPU_ALLOC(nums);
-  if (cpusetp == NULL) return nums;
-  size = CPU_ALLOC_SIZE(nums);
-  ret = sched_getaffinity(0,size,cpusetp);
-  if (ret!=0) return nums;
-  nums = CPU_COUNT_S(size,cpusetp);
-  CPU_FREE(cpusetp);
-  return nums;
+  if (nums >= CPU_SETSIZE) {
+    cpusetp = CPU_ALLOC(nums);
+      if (cpusetp == NULL) {
+        return nums;
+      }
+    size = CPU_ALLOC_SIZE(nums);
+    ret = sched_getaffinity(0,size,cpusetp);
+    if (ret!=0) {
+      CPU_FREE(cpusetp);
+      return nums;
+    }
+    ret = CPU_COUNT_S(size,cpusetp);
+    if (ret > 0 && ret < nums) nums = ret;	
+    CPU_FREE(cpusetp);
+    return nums;
+  } else {
+    ret = sched_getaffinity(0,sizeof(cpuset),&cpuset);
+    if (ret!=0) {
+      return nums;
+    }
+    ret = CPU_COUNT(&cpuset);
+    if (ret > 0 && ret < nums) nums = ret;	
+    return nums;
+  }
  #endif
 #endif
 }
@@ -1756,7 +1818,7 @@ int get_num_procs(void) {
   return nums;
 }
 #endif
-	
+
 #ifdef OS_HAIKU
 int get_num_procs(void) {
   static int nums = 0;
@@ -1793,7 +1855,7 @@ int get_num_procs(void) {
 
 #endif
 
-#if defined(OS_FREEBSD)
+#if defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY)
 
 int get_num_procs(void) {
 
@@ -1870,7 +1932,7 @@ void openblas_fork_handler()
   //   http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60035
   // In the mean time build with USE_OPENMP=0 or link against another
   // implementation of OpenMP.
-#if !(defined(OS_WINDOWS) || defined(OS_ANDROID)) && defined(SMP_SERVER)
+#if !((defined(OS_WINDOWS) && !defined(OS_CYGWIN_NT)) || defined(OS_ANDROID)) && defined(SMP_SERVER)
   int err;
   err = pthread_atfork ((void (*)(void)) BLASFUNC(blas_thread_shutdown), NULL, NULL);
   if(err != 0)
@@ -1883,7 +1945,7 @@ extern int openblas_goto_num_threads_env();
 extern int openblas_omp_num_threads_env();
 
 int blas_get_cpu_number(void){
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN) || defined(OS_ANDROID)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
   int max_num;
 #endif
   int blas_goto_num   = 0;
@@ -1891,11 +1953,11 @@ int blas_get_cpu_number(void){
 
   if (blas_num_threads) return blas_num_threads;
 
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN) || defined(OS_ANDROID)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
   max_num = get_num_procs();
 #endif
 
-  blas_goto_num = 0;
+  // blas_goto_num = 0;
 #ifndef USE_OPENMP
   blas_goto_num=openblas_num_threads_env();
   if (blas_goto_num < 0) blas_goto_num = 0;
@@ -1907,7 +1969,7 @@ int blas_get_cpu_number(void){
 
 #endif
 
-  blas_omp_num = 0;
+  // blas_omp_num = 0;
   blas_omp_num=openblas_omp_num_threads_env();
   if (blas_omp_num < 0) blas_omp_num = 0;
 
@@ -1915,7 +1977,7 @@ int blas_get_cpu_number(void){
   else if (blas_omp_num > 0) blas_num_threads = blas_omp_num;
   else blas_num_threads = MAX_CPU_NUMBER;
 
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN) || defined(OS_ANDROID)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
   if (blas_num_threads > max_num) blas_num_threads = max_num;
 #endif
 
@@ -2002,11 +2064,15 @@ static void *alloc_mmap(void *address){
   }
 
   if (map_address != (void *)-1) {
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
     LOCK_COMMAND(&alloc_lock);
+#endif    
     release_info[release_pos].address = map_address;
     release_info[release_pos].func    = alloc_mmap_free;
     release_pos ++;
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
     UNLOCK_COMMAND(&alloc_lock);
+#endif    
   }
 
 #ifdef OS_LINUX
@@ -2148,14 +2214,18 @@ static void *alloc_mmap(void *address){
 #if defined(OS_LINUX) && !defined(NO_WARMUP)
   }
 #endif
-  LOCK_COMMAND(&alloc_lock);
 
   if (map_address != (void *)-1) {
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
+    LOCK_COMMAND(&alloc_lock);
+#endif
     release_info[release_pos].address = map_address;
     release_info[release_pos].func    = alloc_mmap_free;
     release_pos ++;
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
+    UNLOCK_COMMAND(&alloc_lock);
+#endif
   }
-  UNLOCK_COMMAND(&alloc_lock);
 
   return map_address;
 }
@@ -2523,7 +2593,7 @@ void *blas_memory_alloc(int procpos){
 
   int position;
 #if defined(WHEREAMI) && !defined(USE_OPENMP)
-  int mypos;
+  int mypos = 0;
 #endif
 
   void *map_address;
@@ -2554,6 +2624,11 @@ void *blas_memory_alloc(int procpos){
     NULL,
   };
   void *(**func)(void *address);
+
+#if defined(USE_OPENMP)
+  if (!memory_initialized) {
+#endif
+
   LOCK_COMMAND(&alloc_lock);
 
   if (!memory_initialized) {
@@ -2589,6 +2664,9 @@ void *blas_memory_alloc(int procpos){
 
   }
   UNLOCK_COMMAND(&alloc_lock);
+#if defined(USE_OPENMP)
+  }
+#endif
 
 #ifdef DEBUG
   printf("Alloc Start ...\n");
@@ -2603,13 +2681,17 @@ void *blas_memory_alloc(int procpos){
 
   do {
     if (!memory[position].used && (memory[position].pos == mypos)) {
+#if defined(SMP) && !defined(USE_OPENMP)
       LOCK_COMMAND(&alloc_lock);
-//      blas_lock(&memory[position].lock);
-
+#else      
+      blas_lock(&memory[position].lock);
+#endif
       if (!memory[position].used) goto allocation;
-
+#if defined(SMP) && !defined(USE_OPENMP)
       UNLOCK_COMMAND(&alloc_lock);
-//      blas_unlock(&memory[position].lock);
+#else
+      blas_unlock(&memory[position].lock);
+#endif      
     }
 
     position ++;
@@ -2621,21 +2703,26 @@ void *blas_memory_alloc(int procpos){
 
   position = 0;
 
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
   LOCK_COMMAND(&alloc_lock);
+#endif
   do {
-/*    if (!memory[position].used) { */
-/*      blas_lock(&memory[position].lock);*/
-
+#if defined(USE_OPENMP)	  
+    if (!memory[position].used) { 
+      blas_lock(&memory[position].lock);
+#endif
       if (!memory[position].used) goto allocation;
       
-/*      blas_unlock(&memory[position].lock);*/
-/*    } */
-
+#if defined(USE_OPENMP)
+      blas_unlock(&memory[position].lock);      
+    }
+#endif
     position ++;
 
   } while (position < NUM_BUFFERS);
-  UNLOCK_COMMAND(&alloc_lock);
-
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
+  UNLOCK_COMMAND(&alloc_lock);	
+#endif
   goto error;
 
   allocation :
@@ -2645,10 +2732,11 @@ void *blas_memory_alloc(int procpos){
 #endif
 
   memory[position].used = 1;
-
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
   UNLOCK_COMMAND(&alloc_lock);
-/*  blas_unlock(&memory[position].lock);*/
-
+#else
+  blas_unlock(&memory[position].lock);	
+#endif
   if (!memory[position].addr) {
     do {
 #ifdef DEBUG
@@ -2665,7 +2753,7 @@ void *blas_memory_alloc(int procpos){
 
 #ifdef ALLOC_DEVICEDRIVER
 	if ((*func ==  alloc_devicedirver) && (map_address == (void *)-1)) {
-	    fprintf(stderr, "OpenBLAS Warning ... Physically contigous allocation was failed.\n");
+	    fprintf(stderr, "OpenBLAS Warning ... Physically contiguous allocation was failed.\n");
 	}
 #endif
 
@@ -2693,9 +2781,13 @@ void *blas_memory_alloc(int procpos){
 
     } while ((BLASLONG)map_address == -1);
 
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
     LOCK_COMMAND(&alloc_lock);
+#endif    
     memory[position].addr = map_address;
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
     UNLOCK_COMMAND(&alloc_lock);
+#endif
 
 #ifdef DEBUG
     printf("  Mapping Succeeded. %p(%d)\n", (void *)memory[position].addr, position);
@@ -2749,8 +2841,9 @@ void blas_memory_free(void *free_area){
 #endif
 
   position = 0;
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
   LOCK_COMMAND(&alloc_lock);
-
+#endif
   while ((position < NUM_BUFFERS) && (memory[position].addr != free_area))
     position++;
 
@@ -2764,7 +2857,9 @@ void blas_memory_free(void *free_area){
   WMB;
 
   memory[position].used = 0;
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
   UNLOCK_COMMAND(&alloc_lock);
+#endif
 
 #ifdef DEBUG
   printf("Unmap Succeeded.\n\n");
@@ -2779,8 +2874,9 @@ void blas_memory_free(void *free_area){
   for (position = 0; position < NUM_BUFFERS; position++)
     printf("%4ld  %p : %d\n", position, memory[position].addr, memory[position].used);
 #endif
+#if (defined(SMP) || defined(USE_LOCKING)) && !defined(USE_OPENMP)
   UNLOCK_COMMAND(&alloc_lock);
-
+#endif
   return;
 }
 
@@ -2830,7 +2926,7 @@ void blas_shutdown(void){
 
 #if defined(OS_LINUX) && !defined(NO_WARMUP)
 
-#ifdef SMP
+#if defined(SMP) || defined(USE_LOCKING)
 #if   defined(USE_PTHREAD_LOCK)
 static pthread_mutex_t    init_lock = PTHREAD_MUTEX_INITIALIZER;
 #elif defined(USE_PTHREAD_SPINLOCK)
@@ -2855,7 +2951,7 @@ static void _touch_memory(blas_arg_t *arg, BLASLONG *range_m, BLASLONG *range_n,
     if (hot_alloc != 2) {
 #endif
 
-#ifdef SMP
+#if defined(SMP) || defined(USE_LOCKING)
   LOCK_COMMAND(&init_lock);
 #endif
 
@@ -2865,7 +2961,7 @@ static void _touch_memory(blas_arg_t *arg, BLASLONG *range_m, BLASLONG *range_n,
     size    -= PAGESIZE;
   }
 
-#ifdef SMP
+#if defined(SMP) || defined(USE_LOCKING)
   UNLOCK_COMMAND(&init_lock);
 #endif
 
@@ -3098,7 +3194,7 @@ void gotoblas_dummy_for_PGI(void) {
 
   gotoblas_init();
   gotoblas_quit();
-
+#if __PGIC__ < 19
 #if 0
   asm ("\t.section\t.ctors,\"aw\",@progbits; .align 8; .quad gotoblas_init; .section .text");
   asm ("\t.section\t.dtors,\"aw\",@progbits; .align 8; .quad gotoblas_quit; .section .text");
@@ -3106,6 +3202,7 @@ void gotoblas_dummy_for_PGI(void) {
   asm (".section .init,\"ax\"; call gotoblas_init@PLT; .section .text");
   asm (".section .fini,\"ax\"; call gotoblas_quit@PLT; .section .text");
 #endif
+#endif	
 }
 #endif
 

driver/others/openblas_get_config.c

@@ -35,12 +35,6 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #include <string.h>
 
-#if defined(_WIN32) && defined(_MSC_VER)
-#if _MSC_VER < 1900
-#define snprintf _snprintf
-#endif
-#endif
-
 static char* openblas_config_str=""
 "OpenBLAS "
  VERSION

exports/Makefile

@@ -141,6 +141,14 @@ else
 	$(OBJCOPY) --redefine-syms objcopy.def ../$(LIBNAME) ../$(LIBNAME).renamed
 ../$(LIBSONAME) : ../$(LIBNAME).renamed linktest.c
 endif
+
+ifeq ($(F_COMPILER), INTEL)
+	$(FC) $(FFLAGS) $(LDFLAGS) -shared -o ../$(LIBSONAME) \
+	-Wl,--whole-archive $< -Wl,--no-whole-archive \
+	-Wl,-soname,$(INTERNALNAME) $(EXTRALIB)
+	$(CC) $(CFLAGS) $(LDFLAGS) -w -o linktest linktest.c ../$(LIBSONAME) $(FEXTRALIB) && echo OK.
+else
+
 ifneq ($(C_COMPILER), LSB)
 	$(CC) $(CFLAGS) $(LDFLAGS) -shared -o ../$(LIBSONAME) \
 	-Wl,--whole-archive $< -Wl,--no-whole-archive \
@@ -152,6 +160,7 @@ else
 	-Wl,--whole-archive $< -Wl,--no-whole-archive \
 	-Wl,-soname,$(INTERNALNAME) $(EXTRALIB)
 	$(FC) $(CFLAGS) $(LDFLAGS) -w -o linktest linktest.c ../$(LIBSONAME) $(FEXTRALIB) && echo OK.
+endif
 endif
 	rm -f linktest
 

exports/dllinit.c

@@ -40,15 +40,25 @@
 
 void gotoblas_init(void);
 void gotoblas_quit(void);
+#if defined(SMP) && defined(USE_TLS)
+void blas_thread_memory_cleanup(void);
+#endif
 
 BOOL APIENTRY DllMain(HINSTANCE hInst, DWORD reason, LPVOID reserved) {
-
-  if (reason == DLL_PROCESS_ATTACH) {
-    gotoblas_init();
-  }
-
-  if (reason == DLL_PROCESS_DETACH) {
-    gotoblas_quit();
+  switch(reason) {
+      case DLL_PROCESS_ATTACH:
+        gotoblas_init();
+        break;
+      case DLL_PROCESS_DETACH:
+        gotoblas_quit();
+        break;
+      case DLL_THREAD_ATTACH:
+        break;
+      case DLL_THREAD_DETACH:
+#if defined(SMP) && defined(USE_TLS)
+        blas_thread_memory_cleanup();
+#endif
+        break;
   }
 
   return TRUE;

f_check

@@ -125,7 +125,7 @@ if ($compiler eq "") {
 	    $openmp = "-openmp";
 	}
 
-	# for embeded underscore name, e.g. zho_ge, it may append 2 underscores.
+	# for embedded underscore name, e.g. zho_ge, it may append 2 underscores.
 	$data = `$compiler -O2 -S ftest3.f > /dev/null 2>&1 && cat ftest3.s && rm -f ftest3.s`;
 	if ($data =~ / zho_ge__/) {
 	    $need2bu       = 1;

getarch.c

@@ -637,6 +637,18 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CORENAME  "POWER8"
 #endif
 
+#if defined(FORCE_POWER9) 
+#define FORCE
+#define ARCHITECTURE    "POWER"
+#define SUBARCHITECTURE "POWER9"
+#define SUBDIRNAME      "power"
+#define ARCHCONFIG   "-DPOWER9 " \
+		     "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=128 " \
+		     "-DL2_SIZE=4194304 -DL2_LINESIZE=128 " \
+		     "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
+#define LIBNAME   "power9"
+#define CORENAME  "POWER9"
+#endif
 
 #ifdef FORCE_PPCG4
 #define FORCE
@@ -1065,6 +1077,23 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #else
 #endif
 
+#ifdef FORCE_TSV110
+#define FORCE
+#define ARCHITECTURE    "ARM64"
+#define SUBARCHITECTURE "TSV110"
+#define SUBDIRNAME      "arm64"
+#define ARCHCONFIG   "-DTSV110 " \
+       "-DL1_CODE_SIZE=65536  -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=4 " \
+       "-DL1_DATA_SIZE=65536  -DL1_DATA_LINESIZE=64 -DL1_DATA_ASSOCIATIVE=4 " \
+       "-DL2_SIZE=524288 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=8 " \
+       "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
+       "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
+#define LIBNAME   "tsv110"
+#define CORENAME  "TSV110"
+#else
+#endif
+
+
 #ifdef FORCE_ZARCH_GENERIC
 #define FORCE
 #define ARCHITECTURE    "ZARCH"
@@ -1085,6 +1114,16 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CORENAME  "Z13"
 #endif
 
+#ifdef FORCE_Z14
+#define FORCE
+#define ARCHITECTURE    "ZARCH"
+#define SUBARCHITECTURE "Z14"
+#define ARCHCONFIG   "-DZ14 " \
+       "-DDTB_DEFAULT_ENTRIES=64"
+#define LIBNAME   "z14"
+#define CORENAME  "Z14"
+#endif
+
 #ifndef FORCE
 
 #ifdef USER_TARGET

interface/CMakeLists.txt

@@ -12,6 +12,7 @@ set(BLAS1_REAL_ONLY_SOURCES
   rotm.c rotmg.c # N.B. these do not have complex counterparts
   rot.c
   asum.c
+  sum.c
 )
 
 # these will have 'z' prepended for the complex version
@@ -23,7 +24,7 @@ set(BLAS1_MANGLED_SOURCES
   axpby.c
 )
 
-# TODO: USE_NETLIB_GEMV shoudl switch gemv.c to netlib/*gemv.f
+# TODO: USE_NETLIB_GEMV should switch gemv.c to netlib/*gemv.f
 # these all have 'z' sources for complex versions
 set(BLAS2_SOURCES
   gemv.c ger.c
@@ -124,6 +125,7 @@ foreach (float_type ${FLOAT_TYPES})
     GenerateNamedObjects("max.c" "USE_ABS;USE_MIN" "scamin" ${CBLAS_FLAG} "" "" true "COMPLEX")
     GenerateNamedObjects("max.c" "USE_ABS" "scamax" ${CBLAS_FLAG} "" "" true "COMPLEX")
     GenerateNamedObjects("asum.c" "" "scasum" ${CBLAS_FLAG} "" "" true "COMPLEX")
+    GenerateNamedObjects("sum.c" "" "scsum" ${CBLAS_FLAG} "" "" true "COMPLEX")
   endif ()
   if (${float_type} STREQUAL "ZCOMPLEX")
     GenerateNamedObjects("zscal.c" "SSCAL" "dscal" ${CBLAS_FLAG} "" "" false "ZCOMPLEX")
@@ -132,6 +134,7 @@ foreach (float_type ${FLOAT_TYPES})
     GenerateNamedObjects("max.c" "USE_ABS;USE_MIN" "dzamin" ${CBLAS_FLAG} "" "" true "ZCOMPLEX")
     GenerateNamedObjects("max.c" "USE_ABS" "dzamax" ${CBLAS_FLAG} "" "" true "ZCOMPLEX")
     GenerateNamedObjects("asum.c" "" "dzasum" ${CBLAS_FLAG} "" "" true "ZCOMPLEX")
+    GenerateNamedObjects("sum.c" "" "dzsum" ${CBLAS_FLAG} "" "" true "ZCOMPLEX")
   endif ()
 endforeach ()
 

interface/Makefile

@@ -25,7 +25,7 @@ SBLAS1OBJS    = \
 		saxpy.$(SUFFIX) sswap.$(SUFFIX) \
 		scopy.$(SUFFIX) sscal.$(SUFFIX) \
 		sdot.$(SUFFIX) sdsdot.$(SUFFIX) dsdot.$(SUFFIX) \
-		sasum.$(SUFFIX) snrm2.$(SUFFIX) \
+		sasum.$(SUFFIX) ssum.$(SUFFIX) snrm2.$(SUFFIX) \
 		smax.$(SUFFIX) samax.$(SUFFIX) ismax.$(SUFFIX) isamax.$(SUFFIX) \
 		smin.$(SUFFIX) samin.$(SUFFIX) ismin.$(SUFFIX) isamin.$(SUFFIX) \
 		srot.$(SUFFIX) srotg.$(SUFFIX) srotm.$(SUFFIX) srotmg.$(SUFFIX) \
@@ -51,7 +51,7 @@ DBLAS1OBJS    = \
 		daxpy.$(SUFFIX) dswap.$(SUFFIX) \
 		dcopy.$(SUFFIX) dscal.$(SUFFIX) \
 		ddot.$(SUFFIX) \
-		dasum.$(SUFFIX) dnrm2.$(SUFFIX) \
+		dasum.$(SUFFIX) dsum.$(SUFFIX) dnrm2.$(SUFFIX) \
 		dmax.$(SUFFIX) damax.$(SUFFIX) idmax.$(SUFFIX) idamax.$(SUFFIX) \
 		dmin.$(SUFFIX) damin.$(SUFFIX) idmin.$(SUFFIX) idamin.$(SUFFIX) \
 		drot.$(SUFFIX) drotg.$(SUFFIX) drotm.$(SUFFIX) drotmg.$(SUFFIX) \
@@ -76,7 +76,7 @@ CBLAS1OBJS    = \
 		caxpy.$(SUFFIX) caxpyc.$(SUFFIX) cswap.$(SUFFIX) \
 		ccopy.$(SUFFIX) cscal.$(SUFFIX) csscal.$(SUFFIX) \
 		cdotc.$(SUFFIX)  cdotu.$(SUFFIX) \
-		scasum.$(SUFFIX) scnrm2.$(SUFFIX) \
+		scasum.$(SUFFIX) scsum.$(SUFFIX) scnrm2.$(SUFFIX) \
 		scamax.$(SUFFIX) icamax.$(SUFFIX) \
 		scamin.$(SUFFIX) icamin.$(SUFFIX) \
 		csrot.$(SUFFIX) crotg.$(SUFFIX) \
@@ -105,7 +105,7 @@ ZBLAS1OBJS    = \
 		zaxpy.$(SUFFIX) zaxpyc.$(SUFFIX) zswap.$(SUFFIX) \
 		zcopy.$(SUFFIX) zscal.$(SUFFIX) zdscal.$(SUFFIX) \
 		zdotc.$(SUFFIX)  zdotu.$(SUFFIX) \
-		dzasum.$(SUFFIX)  dznrm2.$(SUFFIX) \
+		dzasum.$(SUFFIX) dzsum.$(SUFFIX) dznrm2.$(SUFFIX) \
 		dzamax.$(SUFFIX) izamax.$(SUFFIX) \
 		dzamin.$(SUFFIX) izamin.$(SUFFIX) \
 		zdrot.$(SUFFIX) zrotg.$(SUFFIX) \
@@ -146,7 +146,7 @@ QBLAS1OBJS    = \
 		qaxpy.$(SUFFIX) qswap.$(SUFFIX) \
 		qcopy.$(SUFFIX) qscal.$(SUFFIX) \
 		qdot.$(SUFFIX) \
-		qasum.$(SUFFIX) qnrm2.$(SUFFIX) \
+		qasum.$(SUFFIX) qsum.$(SUFFIX) qnrm2.$(SUFFIX) \
 		qmax.$(SUFFIX) qamax.$(SUFFIX) iqmax.$(SUFFIX) iqamax.$(SUFFIX) \
 		qmin.$(SUFFIX) qamin.$(SUFFIX) iqmin.$(SUFFIX) iqamin.$(SUFFIX) \
 		qrot.$(SUFFIX) qrotg.$(SUFFIX) qrotm.$(SUFFIX) qrotmg.$(SUFFIX) \
@@ -168,7 +168,7 @@ XBLAS1OBJS    = \
 		xaxpy.$(SUFFIX) xaxpyc.$(SUFFIX) xswap.$(SUFFIX) \
 		xcopy.$(SUFFIX) xscal.$(SUFFIX) xqscal.$(SUFFIX) \
 		xdotc.$(SUFFIX)  xdotu.$(SUFFIX) \
-		qxasum.$(SUFFIX) qxnrm2.$(SUFFIX) \
+		qxasum.$(SUFFIX) qxsum.$(SUFFIX) qxnrm2.$(SUFFIX) \
 		qxamax.$(SUFFIX) ixamax.$(SUFFIX) \
 		qxamin.$(SUFFIX) ixamin.$(SUFFIX) \
 		xqrot.$(SUFFIX) xrotg.$(SUFFIX) \
@@ -203,7 +203,7 @@ ifdef QUAD_PRECISION
 QBLAS1OBJS    = \
 		qaxpy.$(SUFFIX) qswap.$(SUFFIX) \
 		qcopy.$(SUFFIX) qscal.$(SUFFIX) \
-		qasum.$(SUFFIX) qnrm2.$(SUFFIX) \
+		qasum.$(SUFFIX) qsum.$(SUFFIX) qnrm2.$(SUFFIX) \
 		qmax.$(SUFFIX) qamax.$(SUFFIX) iqmax.$(SUFFIX) iqamax.$(SUFFIX) \
 		qmin.$(SUFFIX) qamin.$(SUFFIX) iqmin.$(SUFFIX) iqamin.$(SUFFIX) \
 		qrot.$(SUFFIX) qrotg.$(SUFFIX) qrotm.$(SUFFIX) qrotmg.$(SUFFIX) \
@@ -224,7 +224,7 @@ QBLAS3OBJS    = \
 XBLAS1OBJS    = \
 		xaxpy.$(SUFFIX) xaxpyc.$(SUFFIX) xswap.$(SUFFIX) \
 		xcopy.$(SUFFIX) xscal.$(SUFFIX) xqscal.$(SUFFIX) \
-		qxasum.$(SUFFIX) qxnrm2.$(SUFFIX) \
+		qxasum.$(SUFFIX) qxsum.$(SUFFIX) qxnrm2.$(SUFFIX) \
 		qxamax.$(SUFFIX) ixamax.$(SUFFIX) \
 		qxamin.$(SUFFIX) ixamin.$(SUFFIX) \
 		xqrot.$(SUFFIX) xrotg.$(SUFFIX) \
@@ -263,7 +263,8 @@ CSBLAS1OBJS   = \
 	cblas_isamax.$(SUFFIX) cblas_isamin.$(SUFFIX) cblas_sasum.$(SUFFIX) cblas_saxpy.$(SUFFIX) \
 	cblas_scopy.$(SUFFIX) cblas_sdot.$(SUFFIX) cblas_sdsdot.$(SUFFIX) cblas_dsdot.$(SUFFIX) \
 	cblas_srot.$(SUFFIX) cblas_srotg.$(SUFFIX) cblas_srotm.$(SUFFIX) cblas_srotmg.$(SUFFIX) \
-	cblas_sscal.$(SUFFIX) cblas_sswap.$(SUFFIX) cblas_snrm2.$(SUFFIX) cblas_saxpby.$(SUFFIX)
+	cblas_sscal.$(SUFFIX) cblas_sswap.$(SUFFIX) cblas_snrm2.$(SUFFIX) cblas_saxpby.$(SUFFIX) \
+	cblas_ismin.$(SUFFIX) cblas_ismax.$(SUFFIX) cblas_ssum.$(SUFFIX)
 
 CSBLAS2OBJS   = \
 	cblas_sgemv.$(SUFFIX) cblas_sger.$(SUFFIX) cblas_ssymv.$(SUFFIX) cblas_strmv.$(SUFFIX) \
@@ -280,7 +281,8 @@ CDBLAS1OBJS   = \
 	cblas_idamax.$(SUFFIX) cblas_idamin.$(SUFFIX) cblas_dasum.$(SUFFIX) cblas_daxpy.$(SUFFIX) \
 	cblas_dcopy.$(SUFFIX) cblas_ddot.$(SUFFIX) \
 	cblas_drot.$(SUFFIX) cblas_drotg.$(SUFFIX) cblas_drotm.$(SUFFIX) cblas_drotmg.$(SUFFIX) \
-	cblas_dscal.$(SUFFIX) cblas_dswap.$(SUFFIX) cblas_dnrm2.$(SUFFIX) cblas_daxpby.$(SUFFIX)
+	cblas_dscal.$(SUFFIX) cblas_dswap.$(SUFFIX) cblas_dnrm2.$(SUFFIX) cblas_daxpby.$(SUFFIX) \
+	cblas_idmin.$(SUFFIX) cblas_idmax.$(SUFFIX) cblas_dsum.$(SUFFIX)
 
 CDBLAS2OBJS   = \
 	cblas_dgemv.$(SUFFIX) cblas_dger.$(SUFFIX) cblas_dsymv.$(SUFFIX) cblas_dtrmv.$(SUFFIX) \
@@ -300,7 +302,8 @@ CCBLAS1OBJS   = \
 	cblas_cdotc_sub.$(SUFFIX) cblas_cdotu_sub.$(SUFFIX) \
 	cblas_cscal.$(SUFFIX) cblas_csscal.$(SUFFIX) \
 	cblas_cswap.$(SUFFIX) cblas_scnrm2.$(SUFFIX) \
-	cblas_caxpby.$(SUFFIX)
+	cblas_caxpby.$(SUFFIX) \
+	cblas_icmin.$(SUFFIX) cblas_icmax.$(SUFFIX) cblas_scsum.$(SUFFIX)
 
 CCBLAS2OBJS   = \
 	cblas_cgemv.$(SUFFIX) cblas_cgerc.$(SUFFIX) cblas_cgeru.$(SUFFIX) \
@@ -326,7 +329,9 @@ CZBLAS1OBJS   = \
 	cblas_zdotc_sub.$(SUFFIX) cblas_zdotu_sub.$(SUFFIX) \
 	cblas_zscal.$(SUFFIX) cblas_zdscal.$(SUFFIX) \
 	cblas_zswap.$(SUFFIX) cblas_dznrm2.$(SUFFIX) \
-	cblas_zaxpby.$(SUFFIX)
+	cblas_zaxpby.$(SUFFIX) \
+	cblas_izmin.$(SUFFIX) cblas_izmax.$(SUFFIX) cblas_dzsum.$(SUFFIX)
+
 
 CZBLAS2OBJS   = \
 	cblas_zgemv.$(SUFFIX) cblas_zgerc.$(SUFFIX) cblas_zgeru.$(SUFFIX) \
@@ -560,6 +565,24 @@ dzasum.$(SUFFIX) dzasum.$(PSUFFIX) : asum.c
 qxasum.$(SUFFIX) qxasum.$(PSUFFIX) : asum.c
 	$(CC) $(CFLAGS) -c $< -o $(@F)
 
+ssum.$(SUFFIX) ssum.$(PSUFFIX) : sum.c
+	$(CC) $(CFLAGS) -c $< -o $(@F)
+
+dsum.$(SUFFIX) dsum.$(PSUFFIX) : sum.c
+	$(CC) $(CFLAGS) -c $< -o $(@F)
+
+qsum.$(SUFFIX) qsum.$(PSUFFIX) : sum.c
+	$(CC) $(CFLAGS) -c $< -o $(@F)
+
+scsum.$(SUFFIX) scsum.$(PSUFFIX) : sum.c
+	$(CC) $(CFLAGS) -c $< -o $(@F)
+
+dzsum.$(SUFFIX) dzsum.$(PSUFFIX) : sum.c
+	$(CC) $(CFLAGS) -c $< -o $(@F)
+
+qxsum.$(SUFFIX) qxsum.$(PSUFFIX) : sum.c
+	$(CC) $(CFLAGS) -c $< -o $(@F)
+
 snrm2.$(SUFFIX) snrm2.$(PSUFFIX) : nrm2.c
 	$(CC) $(CFLAGS) -c $< -o $(@F)
 
@@ -1383,6 +1406,18 @@ cblas_ismin.$(SUFFIX) cblas_ismin.$(PSUFFIX) : imax.c
 cblas_idmin.$(SUFFIX) cblas_idmin.$(PSUFFIX) : imax.c
 	$(CC) $(CFLAGS) -DCBLAS -c -UUSE_ABS -DUSE_MIN $< -o $(@F)
 
+cblas_icmax.$(SUFFIX) cblas_icmax.$(PSUFFIX) : imax.c
+	$(CC) $(CFLAGS) -DCBLAS -c -UUSE_ABS -UUSE_MIN $< -o $(@F)
+
+cblas_izmax.$(SUFFIX) cblas_izmax.$(PSUFFIX) : imax.c
+	$(CC) $(CFLAGS) -DCBLAS -c -UUSE_ABS -UUSE_MIN $< -o $(@F)
+
+cblas_icmin.$(SUFFIX) cblas_icmin.$(PSUFFIX) : imax.c
+	$(CC) $(CFLAGS) -DCBLAS -c -UUSE_ABS -DUSE_MIN $< -o $(@F)
+
+cblas_izmin.$(SUFFIX) cblas_izmin.$(PSUFFIX) : imax.c
+	$(CC) $(CFLAGS) -DCBLAS -c -UUSE_ABS -DUSE_MIN $< -o $(@F)
+
 cblas_sasum.$(SUFFIX) cblas_sasum.$(PSUFFIX) : asum.c
 	$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
 
@@ -1395,6 +1430,18 @@ cblas_scasum.$(SUFFIX) cblas_scasum.$(PSUFFIX) : asum.c
 cblas_dzasum.$(SUFFIX) cblas_dzasum.$(PSUFFIX) : asum.c
 	$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
 
+cblas_ssum.$(SUFFIX) cblas_ssum.$(PSUFFIX) : sum.c
+	$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
+
+cblas_dsum.$(SUFFIX) cblas_dsum.$(PSUFFIX) : sum.c
+	$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
+
+cblas_scsum.$(SUFFIX) cblas_scsum.$(PSUFFIX) : sum.c
+	$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
+
+cblas_dzsum.$(SUFFIX) cblas_dzsum.$(PSUFFIX) : sum.c
+	$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
+
 cblas_sdsdot.$(SUFFIX) cblas_sdsdot.$(PSUFFIX) : sdsdot.c
 	$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
 
@@ -1402,7 +1449,7 @@ cblas_dsdot.$(SUFFIX) cblas_dsdot.$(PSUFFIX) : dsdot.c
 	$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
 
 cblas_sdot.$(SUFFIX) cblas_sdot.$(PSUFFIX) : dot.c
-	$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
+		$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
 
 cblas_ddot.$(SUFFIX) cblas_ddot.$(PSUFFIX) : dot.c
 	$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)

interface/axpy.c

@@ -91,7 +91,7 @@ void CNAME(blasint n, FLOAT alpha, FLOAT *x, blasint incx, FLOAT *y, blasint inc
   //disable multi-thread when incx==0 or incy==0
   //In that case, the threads would be dependent.
   //
-  //Temporarily work-around the low performance issue with small imput size &
+  //Temporarily work-around the low performance issue with small input size &
   //multithreads.
   if (incx == 0 || incy == 0 || n <= MULTI_THREAD_MINIMAL)
 	  nthreads = 1;

interface/sum.c

@@ -0,0 +1,97 @@
+/*********************************************************************/
+/* Copyright 2009, 2010 The University of Texas at Austin.           */
+/* All rights reserved.                                              */
+/*                                                                   */
+/* Redistribution and use in source and binary forms, with or        */
+/* without modification, are permitted provided that the following   */
+/* conditions are met:                                               */
+/*                                                                   */
+/*   1. Redistributions of source code must retain the above         */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer.                                                  */
+/*                                                                   */
+/*   2. Redistributions in binary form must reproduce the above      */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer in the documentation and/or other materials       */
+/*      provided with the distribution.                              */
+/*                                                                   */
+/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */
+/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */
+/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */
+/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */
+/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */
+/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */
+/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */
+/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */
+/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */
+/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */
+/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */
+/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */
+/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */
+/*    POSSIBILITY OF SUCH DAMAGE.                                    */
+/*                                                                   */
+/* The views and conclusions contained in the software and           */
+/* documentation are those of the authors and should not be          */
+/* interpreted as representing official policies, either expressed   */
+/* or implied, of The University of Texas at Austin.                 */
+/*********************************************************************/
+
+#include <stdio.h>
+#include "common.h"
+#ifdef FUNCTION_PROFILE
+#include "functable.h"
+#endif
+
+#ifndef CBLAS
+
+FLOATRET NAME(blasint *N, FLOAT *x, blasint *INCX){
+
+  BLASLONG n    = *N;
+  BLASLONG incx = *INCX;
+  FLOATRET ret;
+
+  PRINT_DEBUG_NAME;
+
+  if (n <= 0) return 0;
+
+  IDEBUG_START;
+
+  FUNCTION_PROFILE_START();
+
+  ret = (FLOATRET)SUM_K(n, x, incx);
+
+  FUNCTION_PROFILE_END(COMPSIZE, n, n);
+
+  IDEBUG_END;
+
+  return ret;
+}
+
+#else
+#ifdef COMPLEX
+FLOAT CNAME(blasint n, void *vx, blasint incx){
+  FLOAT *x = (FLOAT*) vx;
+#else
+FLOAT CNAME(blasint n, FLOAT *x, blasint incx){
+#endif
+
+  FLOAT ret;
+
+  PRINT_DEBUG_CNAME;
+
+  if (n <= 0) return 0;
+
+  IDEBUG_START;
+
+  FUNCTION_PROFILE_START();
+
+  ret = SUM_K(n, x, incx);
+
+  FUNCTION_PROFILE_END(COMPSIZE, n, n);
+
+  IDEBUG_END;
+
+  return ret;
+}
+
+#endif

interface/trmv.c

@@ -218,11 +218,8 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,
   buffer = (FLOAT *)blas_memory_alloc(1);
 
 #ifdef SMP
-/*  nthreads = num_cpu_avail(2);
+  nthreads = num_cpu_avail(2);
 
-FIXME trmv_thread was found to be broken, see issue 1332 */
-  nthreads = 1;
-  
   if (nthreads == 1) {
 #endif
 

interface/trsm.c

@@ -204,7 +204,7 @@ void NAME(char *SIDE, char *UPLO, char *TRANS, char *DIAG,
   if (side  < 0)                info =  1;
 
   if (info != 0) {
-    BLASFUNC(xerbla)(ERROR_NAME, &info, sizeof(ERROR_NAME));
+    BLASFUNC(xerbla)(ERROR_NAME, &info, sizeof(ERROR_NAME)-1);
     return;
   }
 

interface/zaxpy.c

@@ -99,7 +99,7 @@ void CNAME(blasint n, FLOAT *ALPHA, FLOAT *x, blasint incx, FLOAT *y, blasint in
   //disable multi-thread when incx==0 or incy==0
   //In that case, the threads would be dependent.
   //
-  //Temporarily work-around the low performance issue with small imput size &
+  //Temporarily work-around the low performance issue with small input size &
   //multithreads.
   if (incx == 0 || incy == 0 || n <= MULTI_THREAD_MINIMAL)
 	  nthreads = 1;

interface/ztrmv.c

@@ -239,9 +239,6 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,
   } else
       nthreads = 1;
 
-/* FIXME TRMV multithreading appears to be broken, see issue 1332*/
-  nthreads = 1;
-
   if(nthreads > 1) {
     buffer_size = n > 16 ? 0 : n * 4 + 40;
   }

kernel/CMakeLists.txt

@@ -65,6 +65,7 @@ function (build_core TARGET_CORE KDIR TSUFFIX KERNEL_DEFINITIONS)
       GenerateNamedObjects("${KERNELDIR}/${${float_char}SCALKERNEL}" "" "scal_k" false "" "" false ${float_type})
       GenerateNamedObjects("${KERNELDIR}/${${float_char}SWAPKERNEL}" "" "swap_k" false "" "" false ${float_type})
       GenerateNamedObjects("${KERNELDIR}/${${float_char}AXPBYKERNEL}" "" "axpby_k" false "" "" false ${float_type})
+      GenerateNamedObjects("${KERNELDIR}/${${float_char}SUMKERNEL}" "" "sum_k" false "" "" false ${float_type})
 
       if (${float_type} STREQUAL "COMPLEX" OR ${float_type} STREQUAL "ZCOMPLEX")
         GenerateNamedObjects("${KERNELDIR}/${${float_char}AXPYKERNEL}" "CONJ" "axpyc_k" false "" "" false ${float_type})

kernel/Makefile.L1

@@ -340,6 +340,32 @@ ifndef XSCALKERNEL
 XSCALKERNEL = zscal.S
 endif
 
+### SUM ###
+
+ifndef SSUMKERNEL
+SSUMKERNEL =  sum.S
+endif
+
+ifndef DSUMKERNEL
+DSUMKERNEL =  sum.S
+endif
+
+ifndef CSUMKERNEL
+CSUMKERNEL = zsum.S
+endif
+
+ifndef ZSUMKERNEL
+ZSUMKERNEL = zsum.S
+endif
+
+ifndef QSUMKERNEL
+QSUMKERNEL =  sum.S
+endif
+
+ifndef XSUMKERNEL
+XSUMKERNEL = zsum.S
+endif
+
 ### SWAP ###
 
 ifndef SSWAPKERNEL
@@ -453,7 +479,7 @@ endif
 SBLASOBJS	+= \
 	 samax_k$(TSUFFIX).$(SUFFIX)  samin_k$(TSUFFIX).$(SUFFIX)  smax_k$(TSUFFIX).$(SUFFIX)  smin_k$(TSUFFIX).$(SUFFIX) \
 	isamax_k$(TSUFFIX).$(SUFFIX) isamin_k$(TSUFFIX).$(SUFFIX) ismax_k$(TSUFFIX).$(SUFFIX) ismin_k$(TSUFFIX).$(SUFFIX) \
-	sasum_k$(TSUFFIX).$(SUFFIX) saxpy_k$(TSUFFIX).$(SUFFIX) scopy_k$(TSUFFIX).$(SUFFIX) \
+	sasum_k$(TSUFFIX).$(SUFFIX) ssum_k$(TSUFFIX).$(SUFFIX) saxpy_k$(TSUFFIX).$(SUFFIX) scopy_k$(TSUFFIX).$(SUFFIX) \
 	sdot_k$(TSUFFIX).$(SUFFIX) sdsdot_k$(TSUFFIX).$(SUFFIX) dsdot_k$(TSUFFIX).$(SUFFIX) \
 	snrm2_k$(TSUFFIX).$(SUFFIX) srot_k$(TSUFFIX).$(SUFFIX) sscal_k$(TSUFFIX).$(SUFFIX) sswap_k$(TSUFFIX).$(SUFFIX) \
 	saxpby_k$(TSUFFIX).$(SUFFIX)
@@ -463,31 +489,32 @@ DBLASOBJS	+= \
 	idamax_k$(TSUFFIX).$(SUFFIX) idamin_k$(TSUFFIX).$(SUFFIX) idmax_k$(TSUFFIX).$(SUFFIX) idmin_k$(TSUFFIX).$(SUFFIX) \
 	dasum_k$(TSUFFIX).$(SUFFIX) daxpy_k$(TSUFFIX).$(SUFFIX) dcopy_k$(TSUFFIX).$(SUFFIX) ddot_k$(TSUFFIX).$(SUFFIX) \
 	dnrm2_k$(TSUFFIX).$(SUFFIX) drot_k$(TSUFFIX).$(SUFFIX) dscal_k$(TSUFFIX).$(SUFFIX) dswap_k$(TSUFFIX).$(SUFFIX) \
-	daxpby_k$(TSUFFIX).$(SUFFIX)
+	daxpby_k$(TSUFFIX).$(SUFFIX) dsum_k$(TSUFFIX).$(SUFFIX)
 
 QBLASOBJS	+= \
 	 qamax_k$(TSUFFIX).$(SUFFIX)  qamin_k$(TSUFFIX).$(SUFFIX)  qmax_k$(TSUFFIX).$(SUFFIX)  qmin_k$(TSUFFIX).$(SUFFIX) \
 	iqamax_k$(TSUFFIX).$(SUFFIX) iqamin_k$(TSUFFIX).$(SUFFIX) iqmax_k$(TSUFFIX).$(SUFFIX) iqmin_k$(TSUFFIX).$(SUFFIX) \
 	qasum_k$(TSUFFIX).$(SUFFIX) qaxpy_k$(TSUFFIX).$(SUFFIX) qcopy_k$(TSUFFIX).$(SUFFIX) qdot_k$(TSUFFIX).$(SUFFIX) \
-	qnrm2_k$(TSUFFIX).$(SUFFIX) qrot_k$(TSUFFIX).$(SUFFIX) qscal_k$(TSUFFIX).$(SUFFIX) qswap_k$(TSUFFIX).$(SUFFIX)
+	qnrm2_k$(TSUFFIX).$(SUFFIX) qrot_k$(TSUFFIX).$(SUFFIX) qscal_k$(TSUFFIX).$(SUFFIX) qswap_k$(TSUFFIX).$(SUFFIX) \
+	qsum_k$(TSUFFIX).$(SUFFIX)
 
 CBLASOBJS	+= \
 	camax_k$(TSUFFIX).$(SUFFIX)  camin_k$(TSUFFIX).$(SUFFIX) icamax_k$(TSUFFIX).$(SUFFIX) icamin_k$(TSUFFIX).$(SUFFIX) \
 	casum_k$(TSUFFIX).$(SUFFIX) caxpy_k$(TSUFFIX).$(SUFFIX) caxpyc_k$(TSUFFIX).$(SUFFIX) ccopy_k$(TSUFFIX).$(SUFFIX) \
 	cdotc_k$(TSUFFIX).$(SUFFIX) cdotu_k$(TSUFFIX).$(SUFFIX) cnrm2_k$(TSUFFIX).$(SUFFIX) csrot_k$(TSUFFIX).$(SUFFIX) \
-	cscal_k$(TSUFFIX).$(SUFFIX) cswap_k$(TSUFFIX).$(SUFFIX) caxpby_k$(TSUFFIX).$(SUFFIX)
+	cscal_k$(TSUFFIX).$(SUFFIX) cswap_k$(TSUFFIX).$(SUFFIX) caxpby_k$(TSUFFIX).$(SUFFIX) csum_k$(TSUFFIX).$(SUFFIX)
 
 ZBLASOBJS	+= \
 	 zamax_k$(TSUFFIX).$(SUFFIX)  zamin_k$(TSUFFIX).$(SUFFIX) izamax_k$(TSUFFIX).$(SUFFIX) izamin_k$(TSUFFIX).$(SUFFIX) \
 	zasum_k$(TSUFFIX).$(SUFFIX) zaxpy_k$(TSUFFIX).$(SUFFIX) zaxpyc_k$(TSUFFIX).$(SUFFIX) zcopy_k$(TSUFFIX).$(SUFFIX) \
 	zdotc_k$(TSUFFIX).$(SUFFIX) zdotu_k$(TSUFFIX).$(SUFFIX) znrm2_k$(TSUFFIX).$(SUFFIX) zdrot_k$(TSUFFIX).$(SUFFIX) \
-	zscal_k$(TSUFFIX).$(SUFFIX) zswap_k$(TSUFFIX).$(SUFFIX) zaxpby_k$(TSUFFIX).$(SUFFIX)
+	zscal_k$(TSUFFIX).$(SUFFIX) zswap_k$(TSUFFIX).$(SUFFIX) zaxpby_k$(TSUFFIX).$(SUFFIX) zsum_k$(TSUFFIX).$(SUFFIX)
 
 XBLASOBJS	+= \
 	 xamax_k$(TSUFFIX).$(SUFFIX)  xamin_k$(TSUFFIX).$(SUFFIX) ixamax_k$(TSUFFIX).$(SUFFIX) ixamin_k$(TSUFFIX).$(SUFFIX) \
 	xasum_k$(TSUFFIX).$(SUFFIX) xaxpy_k$(TSUFFIX).$(SUFFIX) xaxpyc_k$(TSUFFIX).$(SUFFIX) xcopy_k$(TSUFFIX).$(SUFFIX) \
 	xdotc_k$(TSUFFIX).$(SUFFIX) xdotu_k$(TSUFFIX).$(SUFFIX) xnrm2_k$(TSUFFIX).$(SUFFIX) xqrot_k$(TSUFFIX).$(SUFFIX) \
-	xscal_k$(TSUFFIX).$(SUFFIX) xswap_k$(TSUFFIX).$(SUFFIX)
+	xscal_k$(TSUFFIX).$(SUFFIX) xswap_k$(TSUFFIX).$(SUFFIX) xsum_k$(TSUFFIX).$(SUFFIX)
 
 ### AMAX ###
 
@@ -617,7 +644,7 @@ $(KDIR)idmin_k$(TSUFFIX).$(SUFFIX)  $(KDIR)idmin_k$(TPSUFFIX).$(PSUFFIX)  : $(KE
 $(KDIR)iqmin_k$(TSUFFIX).$(SUFFIX)  $(KDIR)iqmin_k$(TPSUFFIX).$(PSUFFIX)  : $(KERNELDIR)/$(IQMINKERNEL)
 	$(CC) -c $(CFLAGS) -UCOMPLEX -DXDOUBLE -UUSE_ABS -DUSE_MIN $< -o $@
 
-
+### ASUM ###
 $(KDIR)sasum_k$(TSUFFIX).$(SUFFIX)   $(KDIR)sasum_k$(TPSUFFIX).$(PSUFFIX)   : $(KERNELDIR)/$(SASUMKERNEL)
 	$(CC) -c $(CFLAGS) -UCOMPLEX -UDOUBLE $< -o $@
 
@@ -636,6 +663,26 @@ $(KDIR)zasum_k$(TSUFFIX).$(SUFFIX)  $(KDIR)zasum_k$(TPSUFFIX).$(PSUFFIX)  : $(KE
 $(KDIR)xasum_k$(TSUFFIX).$(SUFFIX)  $(KDIR)xasum_k$(TPSUFFIX).$(PSUFFIX)  : $(KERNELDIR)/$(XASUMKERNEL)
 	$(CC) -c $(CFLAGS) -DCOMPLEX -DXDOUBLE $< -o $@
 
+### SUM ###
+$(KDIR)ssum_k$(TSUFFIX).$(SUFFIX)   $(KDIR)ssum_k$(TPSUFFIX).$(PSUFFIX)   : $(KERNELDIR)/$(SSUMKERNEL)
+	$(CC) -c $(CFLAGS) -UCOMPLEX -UDOUBLE $< -o $@
+
+$(KDIR)dsum_k$(TSUFFIX).$(SUFFIX)   $(KDIR)dsum_k$(TPSUFFIX).$(PSUFFIX)   : $(KERNELDIR)/$(DSUMKERNEL)
+	$(CC) -c $(CFLAGS) -UCOMPLEX -DDOUBLE $< -o $@
+
+$(KDIR)qsum_k$(TSUFFIX).$(SUFFIX)   $(KDIR)qsum_k$(TPSUFFIX).$(PSUFFIX)   : $(KERNELDIR)/$(QSUMKERNEL)
+	$(CC) -c $(CFLAGS) -UCOMPLEX -DXDOUBLE $< -o $@
+
+$(KDIR)csum_k$(TSUFFIX).$(SUFFIX)  $(KDIR)csum_k$(TPSUFFIX).$(PSUFFIX)  : $(KERNELDIR)/$(CSUMKERNEL)
+	$(CC) -c $(CFLAGS) -DCOMPLEX -UDOUBLE $< -o $@
+
+$(KDIR)zsum_k$(TSUFFIX).$(SUFFIX)  $(KDIR)zsum_k$(TPSUFFIX).$(PSUFFIX)  : $(KERNELDIR)/$(ZSUMKERNEL)
+	$(CC) -c $(CFLAGS) -DCOMPLEX -DDOUBLE $< -o $@
+
+$(KDIR)xsum_k$(TSUFFIX).$(SUFFIX)  $(KDIR)xsum_k$(TPSUFFIX).$(PSUFFIX)  : $(KERNELDIR)/$(XSUMKERNEL)
+	$(CC) -c $(CFLAGS) -DCOMPLEX -DXDOUBLE $< -o $@
+
+### AXPY ###
 $(KDIR)saxpy_k$(TSUFFIX).$(SUFFIX)  $(KDIR)saxpy_k$(TPSUFFIX).$(PSUFFIX)  : $(KERNELDIR)/$(SAXPYKERNEL)
 	$(CC) -c $(CFLAGS) -UCOMPLEX -UDOUBLE $< -o $@
 

kernel/Makefile.L3

@@ -24,7 +24,7 @@ ifeq ($(TARGET), LOONGSON3B)
 USE_TRMM = 1
 endif
 
-ifeq ($(TARGET), GENERIC)
+ifeq ($(CORE), GENERIC)
 USE_TRMM = 1
 endif
 
@@ -44,10 +44,18 @@ ifeq ($(CORE), POWER8)
 USE_TRMM = 1
 endif
 
+ifeq ($(CORE), POWER9)
+USE_TRMM = 1
+endif
+
 ifeq ($(ARCH), zarch)
 USE_TRMM = 1
 endif
 
+ifeq ($(CORE), Z14)
+USE_TRMM = 1
+endif
+
 
 
 

kernel/alpha/sum.S

@@ -0,0 +1,206 @@
+/*********************************************************************/
+/* Copyright 2009, 2010 The University of Texas at Austin.           */
+/* All rights reserved.                                              */
+/*                                                                   */
+/* Redistribution and use in source and binary forms, with or        */
+/* without modification, are permitted provided that the following   */
+/* conditions are met:                                               */
+/*                                                                   */
+/*   1. Redistributions of source code must retain the above         */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer.                                                  */
+/*                                                                   */
+/*   2. Redistributions in binary form must reproduce the above      */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer in the documentation and/or other materials       */
+/*      provided with the distribution.                              */
+/*                                                                   */
+/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */
+/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */
+/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */
+/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */
+/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */
+/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */
+/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */
+/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */
+/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */
+/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */
+/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */
+/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */
+/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */
+/*    POSSIBILITY OF SUCH DAMAGE.                                    */
+/*                                                                   */
+/* The views and conclusions contained in the software and           */
+/* documentation are those of the authors and should not be          */
+/* interpreted as representing official policies, either expressed   */
+/* or implied, of The University of Texas at Austin.                 */
+/*********************************************************************/
+
+#define ASSEMBLER
+#include "common.h"
+#include "version.h"
+
+#define PREFETCHSIZE	88
+
+#define N	$16
+#define X	$17
+#define INCX	$18
+#define I	$19
+
+#define s0	$f0
+#define s1	$f1
+#define s2	$f10
+#define s3	$f11
+
+#define a0	$f12
+#define a1	$f13
+#define a2	$f14
+#define a3	$f15
+#define a4	$f16
+#define a5	$f17
+#define a6	$f18
+#define a7	$f19
+
+#define t0	$f20
+#define t1	$f21
+#define t2	$f22
+#define t3	$f23
+
+	PROLOGUE
+	PROFCODE
+
+	fclr	s0
+	unop
+	fclr	t0
+	ble	N,  $L999
+
+	sra	N, 3, I
+	fclr	s1
+	fclr	s2
+	ble	I, $L15
+
+	LD	a0,  0 * SIZE(X)
+	fclr	t1
+	SXADDQ	INCX, X, X
+	fclr	t2
+
+	LD	a1,  0 * SIZE(X)
+	fclr	t3
+	SXADDQ	INCX, X, X
+	fclr	s3
+
+	LD	a2,  0 * SIZE(X)
+	SXADDQ	INCX, X, X
+	LD	a3,  0 * SIZE(X)
+	SXADDQ	INCX, X, X
+
+	LD	a4,  0 * SIZE(X)
+	SXADDQ	INCX, X, X
+	LD	a5,  0 * SIZE(X)
+	SXADDQ	INCX, X, X
+
+	lda	I,  -1(I)
+	ble	I, $L13
+	.align 4
+
+$L12:
+	ADD	s0, t0, s0
+	ldl	$31, PREFETCHSIZE * 2 * SIZE(X)
+	fmov	a0, t0
+	lda	I,  -1(I)
+
+	ADD	s1, t1, s1
+	LD	a6,  0 * SIZE(X)
+	fmov	a1, t1
+	SXADDQ	INCX, X, X
+
+	ADD	s2, t2, s2
+	LD	a7,  0 * SIZE(X)
+	fmov	a2, t2
+	SXADDQ	INCX, X, X
+
+	ADD	s3, t3, s3
+	LD	a0,  0 * SIZE(X)
+	fmov	a3, t3
+	SXADDQ	INCX, X, X
+
+	ADD	s0, t0, s0
+	LD	a1,  0 * SIZE(X)
+	fmov	a4, t0
+	SXADDQ	INCX, X, X
+
+	ADD	s1, t1, s1
+	LD	a2,  0 * SIZE(X)
+	fmov	a5, t1
+	SXADDQ	INCX, X, X
+
+	ADD	s2, t2, s2
+	LD	a3,  0 * SIZE(X)
+	fmov	a6, t2
+	SXADDQ	INCX, X, X
+
+	ADD	s3, t3, s3
+	LD	a4,  0 * SIZE(X)
+	fmov	a7, t3
+	SXADDQ	INCX, X, X
+
+	LD	a5,  0 * SIZE(X)
+	unop
+	SXADDQ	INCX, X, X
+	bne	I, $L12
+	.align 4
+
+$L13:
+	ADD	s0, t0, s0
+	LD	a6,  0 * SIZE(X)
+	fmov	a0, t0
+	SXADDQ	INCX, X, X
+
+	ADD	s1, t1, s1
+	LD	a7,  0 * SIZE(X)
+	fmov	a1, t1
+	SXADDQ	INCX, X, X
+
+	ADD	s2, t2, s2
+	fmov	a2, t2
+	ADD	s3, t3, s3
+	fmov	a3, t3
+
+	ADD	s0, t0, s0
+	fmov	a4, t0
+	ADD	s1, t1, s1
+	fmov	a5, t1
+	ADD	s2, t2, s2
+	fmov	a6, t2
+	ADD	s3, t3, s3
+	fmov	a7, t3
+
+	ADD	s1, t1, s1
+	ADD	s2, t2, s2
+	ADD	s3, t3, s3
+
+	ADD	s0, s1, s0
+	ADD	s2, s3, s2
+	.align 4
+
+$L15:
+	and	N, 7, I
+	ADD	s0, s2, s0
+	unop
+	ble	I, $L999
+	.align 4
+
+$L17:
+	ADD	s0, t0, s0
+	LD	a0,  0 * SIZE(X)
+	SXADDQ	INCX, X, X
+	fmov	a0, t0
+
+	lda	I,  -1(I)
+	bne	I, $L17
+	.align 4
+
+$L999:
+	ADD	s0, t0, s0
+	ret
+	EPILOGUE

kernel/alpha/zsum.S

@@ -0,0 +1,208 @@
+/*********************************************************************/
+/* Copyright 2009, 2010 The University of Texas at Austin.           */
+/* All rights reserved.                                              */
+/*                                                                   */
+/* Redistribution and use in source and binary forms, with or        */
+/* without modification, are permitted provided that the following   */
+/* conditions are met:                                               */
+/*                                                                   */
+/*   1. Redistributions of source code must retain the above         */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer.                                                  */
+/*                                                                   */
+/*   2. Redistributions in binary form must reproduce the above      */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer in the documentation and/or other materials       */
+/*      provided with the distribution.                              */
+/*                                                                   */
+/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */
+/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */
+/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */
+/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */
+/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */
+/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */
+/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */
+/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */
+/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */
+/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */
+/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */
+/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */
+/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */
+/*    POSSIBILITY OF SUCH DAMAGE.                                    */
+/*                                                                   */
+/* The views and conclusions contained in the software and           */
+/* documentation are those of the authors and should not be          */
+/* interpreted as representing official policies, either expressed   */
+/* or implied, of The University of Texas at Austin.                 */
+/*********************************************************************/
+
+#define ASSEMBLER
+#include "common.h"
+#include "version.h"
+
+#define PREFETCHSIZE	88
+
+#define N	$16
+#define X	$17
+#define INCX	$18
+#define I	$19
+
+#define s0	$f0
+#define s1	$f1
+#define s2	$f10
+#define s3	$f11
+
+#define a0	$f12
+#define a1	$f13
+#define a2	$f14
+#define a3	$f15
+#define a4	$f16
+#define a5	$f17
+#define a6	$f18
+#define a7	$f19
+
+#define t0	$f20
+#define t1	$f21
+#define t2	$f22
+#define t3	$f23
+
+	PROLOGUE
+	PROFCODE
+
+	fclr	s0
+	unop
+	fclr	t0
+	addq	INCX, INCX, INCX
+
+	fclr	s1
+	unop
+	fclr	t1
+	ble	N,  $L999
+
+	fclr	s2
+	sra	N, 2, I
+	fclr	s3
+	ble	I, $L15
+
+	LD	a0,  0 * SIZE(X)
+	fclr	t2
+	LD	a1,  1 * SIZE(X)
+	SXADDQ	INCX, X, X
+
+	LD	a2,  0 * SIZE(X)
+	fclr	t3
+	LD	a3,  1 * SIZE(X)
+	SXADDQ	INCX, X, X
+
+	LD	a4,  0 * SIZE(X)
+	LD	a5,  1 * SIZE(X)
+	SXADDQ	INCX, X, X
+	lda	I,  -1(I)
+
+	ble	I, $L13
+	.align 4
+
+$L12:
+	ADD	s0, t0, s0
+	ldl	$31, PREFETCHSIZE * SIZE(X)
+	fmov	a0, t0
+	lda	I,  -1(I)
+
+	ADD	s1, t1, s1
+	LD	a6,  0 * SIZE(X)
+	fmov	a1, t1
+	unop
+
+	ADD	s2, t2, s2
+	LD	a7,  1 * SIZE(X)
+	fmov	a2, t2
+	SXADDQ	INCX, X, X
+
+	ADD	s3, t3, s3
+	LD	a0,  0 * SIZE(X)
+	fmov	a3, t3
+	unop
+
+	ADD	s0, t0, s0
+	LD	a1,  1 * SIZE(X)
+	fmov	a4, t0
+	SXADDQ	INCX, X, X
+
+	ADD	s1, t1, s1
+	LD	a2,  0 * SIZE(X)
+	fmov	a5, t1
+	unop
+
+	ADD	s2, t2, s2
+	LD	a3,  1 * SIZE(X)
+	fmov	a6, t2
+	SXADDQ	INCX, X, X
+
+	ADD	s3, t3, s3
+	LD	a4,  0 * SIZE(X)
+	fmov	a7, t3
+	unop
+
+	LD	a5,  1 * SIZE(X)
+	unop
+	SXADDQ	INCX, X, X
+	bne	I, $L12
+	.align 4
+
+$L13:
+	ADD	s0, t0, s0
+	LD	a6,  0 * SIZE(X)
+	fmov	a0, t0
+
+	ADD	s1, t1, s1
+	LD	a7,  1 * SIZE(X)
+	fmov	a1, t1
+	SXADDQ	INCX, X, X
+
+	ADD	s2, t2, s2
+	fmov	a2, t2
+	ADD	s3, t3, s3
+	fmov	a3, t3
+
+	ADD	s0, t0, s0
+	fmov	a4, t0
+	ADD	s1, t1, s1
+	fmov	a5, t1
+	ADD	s2, t2, s2
+	fmov	a6, t2
+	ADD	s3, t3, s3
+	fmov	a7, t3
+
+	ADD	s2, t2, s2
+	ADD	s3, t3, s3
+
+	.align 4
+
+$L15:
+	ADD	s0, s2, s0
+	and	N, 3, I
+	ADD	s1, s3, s1
+	ble	I, $L999
+	.align 4
+
+$L17:
+	ADD	s0, t0, s0
+	LD	a0,  0 * SIZE(X)
+	fmov	a0, t0
+	lda	I,  -1(I)
+
+	ADD	s1, t1, s1
+	LD	a1,  1 * SIZE(X)
+	fmov	a1, t1
+	SXADDQ	INCX, X, X
+
+	bne	I, $L17
+	.align 4
+
+$L999:
+	ADD	s0, t0, s0
+	ADD	s1, t1, s1
+
+	ADD	s0, s1, s0
+	ret
+	EPILOGUE

kernel/arm/KERNEL.ARMV5

@@ -35,6 +35,11 @@ DASUMKERNEL  = ../arm/asum.c
 CASUMKERNEL  = ../arm/zasum.c
 ZASUMKERNEL  = ../arm/zasum.c
 
+SSUMKERNEL  = ../arm/sum.c
+DSUMKERNEL  = ../arm/sum.c
+CSUMKERNEL  = ../arm/zsum.c
+ZSUMKERNEL  = ../arm/zsum.c
+
 SAXPYKERNEL  = ../arm/axpy.c
 DAXPYKERNEL  = ../arm/axpy.c
 CAXPYKERNEL  = ../arm/zaxpy.c

kernel/arm/KERNEL.ARMV6

@@ -1,30 +1,30 @@
 include $(KERNELDIR)/KERNEL.ARMV5
 
-SAMAXKERNEL  = iamax_vfp.S
-DAMAXKERNEL  = iamax_vfp.S
-CAMAXKERNEL  = iamax_vfp.S
-ZAMAXKERNEL  = iamax_vfp.S
+SAMAXKERNEL  = amax_vfp.S
+DAMAXKERNEL  = amax_vfp.S
+#CAMAXKERNEL  = amax_vfp.S
+#ZAMAXKERNEL  = amax_vfp.S
 
-SAMINKERNEL  = iamax_vfp.S
-DAMINKERNEL  = iamax_vfp.S
-CAMINKERNEL  = iamax_vfp.S
-ZAMINKERNEL  = iamax_vfp.S
+SAMINKERNEL  = amax_vfp.S
+DAMINKERNEL  = amax_vfp.S
+#CAMINKERNEL  = amax_vfp.S
+#ZAMINKERNEL  = amax_vfp.S
 
-SMAXKERNEL   = iamax_vfp.S
-DMAXKERNEL   = iamax_vfp.S
+SMAXKERNEL   = amax_vfp.S
+DMAXKERNEL   = amax_vfp.S
 
-SMINKERNEL   = iamax_vfp.S
-DMINKERNEL   = iamax_vfp.S
+SMINKERNEL   = amax_vfp.S
+DMINKERNEL   = amax_vfp.S
 
 ISAMAXKERNEL = iamax_vfp.S
 IDAMAXKERNEL = iamax_vfp.S
-ICAMAXKERNEL = iamax_vfp.S
-IZAMAXKERNEL = iamax_vfp.S
+#ICAMAXKERNEL = iamax_vfp.S
+#IZAMAXKERNEL = iamax_vfp.S
 
 ISAMINKERNEL = iamax_vfp.S
 IDAMINKERNEL = iamax_vfp.S
-ICAMINKERNEL = iamax_vfp.S
-IZAMINKERNEL = iamax_vfp.S
+#ICAMINKERNEL = iamax_vfp.S
+#IZAMINKERNEL = iamax_vfp.S
 
 ISMAXKERNEL  = iamax_vfp.S
 IDMAXKERNEL  = iamax_vfp.S
@@ -37,6 +37,9 @@ DASUMKERNEL  = asum_vfp.S
 CASUMKERNEL  = asum_vfp.S
 ZASUMKERNEL  = asum_vfp.S
 
+SSUMKERNEL  = sum_vfp.S
+DSUMKERNEL  = sum_vfp.S
+
 SAXPYKERNEL  = axpy_vfp.S
 DAXPYKERNEL  = axpy_vfp.S
 CAXPYKERNEL  = axpy_vfp.S

kernel/arm/amax_vfp.S

@@ -0,0 +1,445 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+/**************************************************************************************
+* 2013/11/14 Saar
+* 	 BLASTEST 		: OK
+* 	 CTEST			: OK
+* 	 TEST			: OK
+*
+**************************************************************************************/
+
+#define ASSEMBLER
+#include "common.h"
+
+#define STACKSIZE 256
+
+#define	N	r0
+#define	X	r1
+#define	INC_X	r2
+
+#define I	r12
+
+#define X_PRE	512
+
+/**************************************************************************************
+* Macro definitions
+**************************************************************************************/
+
+#if	defined(USE_ABS)
+
+#if	defined(DOUBLE)
+
+#define	VABS(x0,x1)	vabs.f64	x0, x1
+
+#else
+
+#define	VABS(x0,x1)	vabs.f32	x0, x1
+
+#endif
+
+#else
+
+#define VABS(x0,x1)	nop
+
+#endif
+
+/*****************************************************************************************/
+
+#if	defined(USE_MIN)
+
+#define	MOVCOND		movlt
+
+#if	defined(DOUBLE)
+
+#define	VMOVCOND	vmovlt.f64
+
+#else
+
+#define	VMOVCOND	vmovlt.f32
+
+#endif
+
+#else
+
+#define	MOVCOND		movgt
+
+#if	defined(DOUBLE)
+
+#define	VMOVCOND	vmovgt.f64
+
+#else
+
+#define	VMOVCOND	vmovgt.f32
+
+#endif
+
+
+#endif
+
+
+/*****************************************************************************************/
+
+
+
+#if	!defined(COMPLEX)
+
+#if	defined(DOUBLE)
+
+.macro INIT_F
+
+	vldmia.f64	X!, { d0 }
+	VABS(   d0,  d0 )
+
+.endm
+
+.macro KERNEL_F1
+
+	vldmia.f64	X!, { d4 }
+	VABS(   d4,  d4 )
+	vcmpe.f64  	d4,  d0
+	vmrs		APSR_nzcv, fpscr
+	VMOVCOND	d0,  d4
+
+.endm
+
+.macro INIT_S
+
+	vldmia.f64	X, { d0 }
+	VABS(   d0,  d0 )
+	add	X, X, INC_X
+
+.endm
+
+
+.macro KERNEL_S1
+
+	vldmia.f64	X, { d4 }
+	VABS(   d4,  d4 )
+	vcmpe.f64  	d4,  d0
+	vmrs		APSR_nzcv, fpscr
+	VMOVCOND	d0,  d4
+	add	X, X, INC_X
+
+.endm
+
+#else
+
+.macro INIT_F
+
+	vldmia.f32	X!, { s0 }
+	VABS(   s0,  s0 )
+
+.endm
+
+.macro KERNEL_F1
+
+	vldmia.f32	X!, { s4 }
+	VABS(   s4,  s4 )
+	vcmpe.f32  	s4,  s0
+	vmrs		APSR_nzcv, fpscr
+	VMOVCOND	s0,  s4
+
+.endm
+
+.macro INIT_S
+
+	vldmia.f32	X, { s0 }
+	VABS(   s0,  s0 )
+	add	X, X, INC_X
+
+.endm
+
+
+.macro KERNEL_S1
+
+	vldmia.f32	X, { s4 }
+	VABS(   s4,  s4 )
+	vcmpe.f32  	s4,  s0
+	vmrs		APSR_nzcv, fpscr
+	VMOVCOND	s0,  s4
+	add	X, X, INC_X
+
+.endm
+
+
+
+
+#endif
+
+#else
+
+#if	defined(DOUBLE)
+
+.macro INIT_F
+
+	vldmia.f64	X!, { d0 -d1 }
+	vabs.f64   d0,  d0
+	vabs.f64   d1,  d1
+	vadd.f64   d0  , d0,  d1
+.endm
+
+
+.macro KERNEL_F1
+
+	vldmia.f64	X!, { d4 - d5 }
+	vabs.f64   d4,  d4
+	vabs.f64   d5,  d5
+	vadd.f64   d4  , d4,  d5
+	vcmpe.f64  	d4,  d0
+	vmrs		APSR_nzcv, fpscr
+	VMOVCOND	d0,  d4
+
+.endm
+
+.macro INIT_S
+
+	vldmia.f64	X, { d0 -d1 }
+	vabs.f64   d0,  d0
+	vabs.f64   d1,  d1
+	vadd.f64   d0  , d0,  d1
+	add	X, X, INC_X
+
+.endm
+
+
+
+.macro KERNEL_S1
+
+	vldmia.f64	X, { d4 - d5 }
+	vabs.f64   d4,  d4
+	vabs.f64   d5,  d5
+	vadd.f64   d4  , d4,  d5
+	vcmpe.f64  	d4,  d0
+	vmrs		APSR_nzcv, fpscr
+	VMOVCOND	d0,  d4
+	add	X, X, INC_X
+
+.endm
+
+#else
+
+.macro INIT_F
+
+	vldmia.f32	X!, { s0 -s1 }
+	vabs.f32   s0,  s0
+	vabs.f32   s1,  s1
+	vadd.f32   s0  , s0,  s1
+
+.endm
+
+
+.macro KERNEL_F1
+
+	vldmia.f32	X!, { s4 - s5 }
+	vabs.f32   s4,  s4
+	vabs.f32   s5,  s5
+	vadd.f32   s4  , s4,  s5
+	vcmpe.f32  	s4,  s0
+	vmrs		APSR_nzcv, fpscr
+	VMOVCOND	s0,  s4
+
+.endm
+
+.macro INIT_S
+
+	vldmia.f32	X, { s0 -s1 }
+	vabs.f32   s0,  s0
+	vabs.f32   s1,  s1
+	vadd.f32   s0  , s0,  s1
+	add	X, X, INC_X
+
+.endm
+
+
+
+.macro KERNEL_S1
+
+	vldmia.f32	X, { s4 - s5 }
+	vabs.f32   s4,  s4
+	vabs.f32   s5,  s5
+	vadd.f32   s4  , s4,  s5
+	vcmpe.f32  	s4,  s0
+	vmrs		APSR_nzcv, fpscr
+	VMOVCOND	s0,  s4
+	add	X, X, INC_X
+
+.endm
+
+
+
+
+#endif
+
+#endif
+
+/**************************************************************************************
+* End of macro definitions
+**************************************************************************************/
+
+	PROLOGUE
+
+	.align 5
+
+        movs    r12, #0                                          // clear floating point register
+        vmov    s0, r12
+#if     defined(DOUBLE)
+        vcvt.f64.f32    d0, s0
+#endif
+
+
+	cmp	N, #0
+	ble	amax_kernel_L999
+
+	cmp	INC_X, #0
+	beq	amax_kernel_L999
+
+
+	cmp	INC_X, #1
+	bne	amax_kernel_S_BEGIN
+
+
+amax_kernel_F_BEGIN:
+
+	INIT_F
+
+	subs	N, N , #1
+	ble	amax_kernel_L999
+
+	asrs	I, N, #2					// I = N / 4
+	ble	amax_kernel_F1
+
+	.align 5
+
+amax_kernel_F4:
+
+	pld	[ X, #X_PRE ]
+	KERNEL_F1
+	KERNEL_F1
+#if defined(COMPLEX) && defined(DOUBLE)
+	pld	[ X, #X_PRE ]
+#endif
+	KERNEL_F1
+	KERNEL_F1
+
+	subs	I, I, #1
+	ble	amax_kernel_F1
+
+
+#if defined(COMPLEX) || defined(DOUBLE)
+	pld	[ X, #X_PRE ]
+#endif
+	KERNEL_F1
+	KERNEL_F1
+#if defined(COMPLEX) && defined(DOUBLE)
+	pld	[ X, #X_PRE ]
+#endif
+	KERNEL_F1
+	KERNEL_F1
+
+	subs	I, I, #1
+	bne	amax_kernel_F4
+
+amax_kernel_F1:
+
+	ands	I, N, #3
+	ble	amax_kernel_L999
+
+amax_kernel_F10:
+
+	KERNEL_F1
+
+	subs    I, I, #1
+        bne     amax_kernel_F10
+
+	b	amax_kernel_L999
+
+amax_kernel_S_BEGIN:
+
+#if defined(COMPLEX)
+
+#if defined(DOUBLE)
+	lsl	INC_X, INC_X, #4				// INC_X * SIZE * 2
+#else
+	lsl	INC_X, INC_X, #3				// INC_X * SIZE * 2
+#endif
+
+#else
+
+#if defined(DOUBLE)
+	lsl	INC_X, INC_X, #3				// INC_X * SIZE
+#else
+	lsl	INC_X, INC_X, #2				// INC_X * SIZE
+#endif
+
+#endif
+
+	INIT_S
+
+	subs	N, N , #1
+	ble	amax_kernel_L999
+
+	asrs	I, N, #2					// I = N / 4
+	ble	amax_kernel_S1
+
+	.align 5
+
+amax_kernel_S4:
+
+	KERNEL_S1
+	KERNEL_S1
+	KERNEL_S1
+	KERNEL_S1
+
+	subs	I, I, #1
+	bne	amax_kernel_S4
+
+amax_kernel_S1:
+
+	ands	I, N, #3
+	ble	amax_kernel_L999
+
+amax_kernel_S10:
+
+	KERNEL_S1
+
+	subs    I, I, #1
+        bne     amax_kernel_S10
+
+
+amax_kernel_L999:
+#if !defined(__ARM_PCS_VFP)
+#if defined(DOUBLE)
+	vmov	r0, r1, d0
+#else	
+	vmov    r0, s0
+#endif
+#endif
+	bx	lr
+
+	EPILOGUE
+

kernel/arm/imin.c

@@ -53,7 +53,7 @@ BLASLONG CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
 
 	while(i < n)
 	{
-		if( x[ix] > minf )
+		if( x[ix] < minf )
 		{
 			min = i;
 			minf = x[ix];

kernel/arm/sum.c

@@ -0,0 +1,51 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+/**************************************************************************************
+* trivial copy of asum.c with the ABS() removed                                       *
+**************************************************************************************/
+
+
+#include "common.h"
+#include <math.h>
+
+FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
+{
+	BLASLONG i=0;
+	FLOAT sumf = 0.0;
+	if (n <= 0 || inc_x <= 0) return(sumf);
+
+	n *= inc_x;
+	while(i < n)
+	{
+		sumf += x[i];
+		i += inc_x;
+	}
+	return(sumf);
+}
+
+

kernel/arm/sum_vfp.S

@@ -0,0 +1,425 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+/**************************************************************************************
+* trivial copy of asum_vfp.S with the in-place vabs.f64 calls removed                                    *
+**************************************************************************************/
+
+#define ASSEMBLER
+#include "common.h"
+
+#define STACKSIZE 256
+
+#define	N	r0
+#define	X	r1
+#define	INC_X	r2
+
+
+#define I	r12
+
+#define X_PRE	512
+
+/**************************************************************************************
+* Macro definitions
+**************************************************************************************/
+
+#if	!defined(COMPLEX)
+
+#if	defined(DOUBLE)
+
+.macro KERNEL_F4
+
+	pld	[ X, #X_PRE  ]
+	vldmia.f64	X!, { d4 - d5 }
+	vadd.f64   d0  , d0,  d4
+	vldmia.f64	X!, { d6 - d7 }
+	vadd.f64   d1  , d1,  d5
+	vadd.f64   d0  , d0,  d6
+	vadd.f64   d1  , d1,  d7
+
+.endm
+
+.macro KERNEL_F1
+
+	vldmia.f64	X!, { d4 }
+	vadd.f64   d0  , d0,  d4
+
+.endm
+
+
+.macro KERNEL_S4
+
+	vldmia.f64	X, { d4 }
+	vadd.f64   d0  , d0,  d4
+	add	X, X, INC_X
+
+	vldmia.f64	X, { d4 }
+	vadd.f64   d0  , d0,  d4
+	add	X, X, INC_X
+
+	vldmia.f64	X, { d4 }
+	vadd.f64   d0  , d0,  d4
+	add	X, X, INC_X
+
+	vldmia.f64	X, { d4 }
+	vadd.f64   d0  , d0,  d4
+	add	X, X, INC_X
+
+.endm
+
+
+.macro KERNEL_S1
+
+	vldmia.f64	X, { d4 }
+	vadd.f64   d0  , d0,  d4
+	add	X, X, INC_X
+
+.endm
+
+#else
+
+.macro KERNEL_F4
+
+	vldmia.f32	X!, { s4 - s5 }
+	vadd.f32   s0  , s0,  s4
+	vldmia.f32	X!, { s6 - s7 }
+	vadd.f32   s1  , s1,  s5
+	vadd.f32   s0  , s0,  s6
+	vadd.f32   s1  , s1,  s7
+
+.endm
+
+.macro KERNEL_F1
+
+	vldmia.f32	X!, { s4 }
+	vadd.f32   s0  , s0,  s4
+
+.endm
+
+
+.macro KERNEL_S4
+
+	vldmia.f32	X, { s4 }
+	vadd.f32   s0  , s0,  s4
+	add	X, X, INC_X
+
+	vldmia.f32	X, { s4 }
+	vadd.f32   s0  , s0,  s4
+	add	X, X, INC_X
+
+	vldmia.f32	X, { s4 }
+	vadd.f32   s0  , s0,  s4
+	add	X, X, INC_X
+
+	vldmia.f32	X, { s4 }
+	vadd.f32   s0  , s0,  s4
+	add	X, X, INC_X
+
+.endm
+
+
+.macro KERNEL_S1
+
+	vldmia.f32	X, { s4 }
+	vadd.f32   s0  , s0,  s4
+	add	X, X, INC_X
+
+.endm
+
+
+#endif
+
+#else
+
+#if	defined(DOUBLE)
+
+.macro KERNEL_F4
+
+	pld	[ X, #X_PRE  ]
+	vldmia.f64	X!, { d4 - d5 }
+	vadd.f64   d0  , d0,  d4
+	vldmia.f64	X!, { d6 - d7 }
+	vadd.f64   d1  , d1,  d5
+	vadd.f64   d0  , d0,  d6
+	vadd.f64   d1  , d1,  d7
+
+	pld	[ X, #X_PRE  ]
+	vldmia.f64	X!, { d4 - d5 }
+	vadd.f64   d0  , d0,  d4
+	vldmia.f64	X!, { d6 - d7 }
+	vadd.f64   d1  , d1,  d5
+	vadd.f64   d0  , d0,  d6
+	vadd.f64   d1  , d1,  d7
+
+
+.endm
+
+.macro KERNEL_F1
+
+	vldmia.f64	X!, { d4 }
+	vadd.f64   d0  , d0,  d4
+
+	vldmia.f64	X!, { d4 }
+	vadd.f64   d0  , d0,  d4
+
+
+.endm
+
+
+.macro KERNEL_S4
+
+	vldmia.f64	X, { d4 -d5 }
+	vadd.f64   d0  , d0,  d4
+	vadd.f64   d0  , d0,  d5
+	add	X, X, INC_X
+
+	vldmia.f64	X, { d4 -d5 }
+	vadd.f64   d0  , d0,  d4
+	vadd.f64   d0  , d0,  d5
+	add	X, X, INC_X
+
+	vldmia.f64	X, { d4 -d5 }
+	vadd.f64   d0  , d0,  d4
+	vadd.f64   d0  , d0,  d5
+	add	X, X, INC_X
+
+	vldmia.f64	X, { d4 -d5 }
+	vadd.f64   d0  , d0,  d4
+	vadd.f64   d0  , d0,  d5
+	add	X, X, INC_X
+
+.endm
+
+
+.macro KERNEL_S1
+
+	vldmia.f64	X, { d4 -d5 }
+	vadd.f64   d0  , d0,  d4
+	vadd.f64   d0  , d0,  d5
+	add	X, X, INC_X
+
+.endm
+
+#else
+
+.macro KERNEL_F4
+
+	pld	[ X, #X_PRE  ]
+	vldmia.f32	X!, { s4 - s5 }
+	vadd.f32   s0  , s0,  s4
+	vldmia.f32	X!, { s6 - s7 }
+	vadd.f32   s1  , s1,  s5
+	vadd.f32   s0  , s0,  s6
+	vadd.f32   s1  , s1,  s7
+
+	vldmia.f32	X!, { s4 - s5 }
+	vadd.f32   s0  , s0,  s4
+	vldmia.f32	X!, { s6 - s7 }
+	vadd.f32   s1  , s1,  s5
+	vadd.f32   s0  , s0,  s6
+	vadd.f32   s1  , s1,  s7
+
+
+.endm
+
+.macro KERNEL_F1
+
+	vldmia.f32	X!, { s4 }
+	vadd.f32   s0  , s0,  s4
+
+	vldmia.f32	X!, { s4 }
+	vadd.f32   s0  , s0,  s4
+
+.endm
+
+
+.macro KERNEL_S4
+
+	vldmia.f32	X, { s4 -s5 }
+	vadd.f32   s0  , s0,  s4
+	vadd.f32   s0  , s0,  s5
+	add	X, X, INC_X
+
+	vldmia.f32	X, { s4 -s5 }
+	vadd.f32   s0  , s0,  s4
+	vadd.f32   s0  , s0,  s5
+	add	X, X, INC_X
+
+	vldmia.f32	X, { s4 -s5 }
+	vadd.f32   s0  , s0,  s4
+	vadd.f32   s0  , s0,  s5
+	add	X, X, INC_X
+
+	vldmia.f32	X, { s4 -s5 }
+	vadd.f32   s0  , s0,  s4
+	vadd.f32   s0  , s0,  s5
+	add	X, X, INC_X
+
+.endm
+
+
+.macro KERNEL_S1
+
+	vldmia.f32	X, { s4 -s5 }
+	vadd.f32   s0  , s0,  s4
+	vadd.f32   s0  , s0,  s5
+	add	X, X, INC_X
+
+.endm
+
+#endif
+
+#endif
+
+/**************************************************************************************
+* End of macro definitions
+**************************************************************************************/
+
+	PROLOGUE
+
+	.align 5
+
+        movs    r12, #0                                          // clear floating point register
+        vmov    s0, r12
+        vmov    s1, r12
+#if     defined(DOUBLE)
+        vcvt.f64.f32    d0, s0
+        vcvt.f64.f32    d1, s1
+#endif
+
+	cmp	N, #0
+	ble	asum_kernel_L999
+
+	cmp	INC_X, #0
+	beq	asum_kernel_L999
+
+	cmp	INC_X, #1
+	bne	asum_kernel_S_BEGIN
+
+
+asum_kernel_F_BEGIN:
+
+	asrs	I, N, #2					// I = N / 4
+	ble	asum_kernel_F1
+
+	.align 5
+
+asum_kernel_F4:
+
+#if !defined(DOUBLE) && !defined(COMPLEX)
+	pld	[ X, #X_PRE  ]
+#endif
+	KERNEL_F4
+
+	subs	I, I, #1
+	ble	asum_kernel_F1
+
+	KERNEL_F4
+
+	subs	I, I, #1
+	bne	asum_kernel_F4
+
+asum_kernel_F1:
+
+	ands	I, N, #3
+	ble	asum_kernel_L999
+
+asum_kernel_F10:
+
+	KERNEL_F1
+
+	subs    I, I, #1
+        bne     asum_kernel_F10
+
+	b	asum_kernel_L999
+
+asum_kernel_S_BEGIN:
+
+#if defined(COMPLEX)
+
+#if defined(DOUBLE)
+	lsl	INC_X, INC_X, #4				// INC_X * SIZE * 2
+#else
+	lsl	INC_X, INC_X, #3				// INC_X * SIZE * 2
+#endif
+
+#else
+
+#if defined(DOUBLE)
+	lsl	INC_X, INC_X, #3				// INC_X * SIZE
+#else
+	lsl	INC_X, INC_X, #2				// INC_X * SIZE
+#endif
+
+#endif
+
+	asrs	I, N, #2					// I = N / 4
+	ble	asum_kernel_S1
+
+	.align 5
+
+asum_kernel_S4:
+
+	KERNEL_S4
+
+	subs	I, I, #1
+	bne	asum_kernel_S4
+
+asum_kernel_S1:
+
+	ands	I, N, #3
+	ble	asum_kernel_L999
+
+asum_kernel_S10:
+
+	KERNEL_S1
+
+	subs    I, I, #1
+        bne     asum_kernel_S10
+
+
+asum_kernel_L999:
+
+
+#if defined(DOUBLE)
+	vadd.f64	d0 , d0, d1				// set return value
+#else
+	vadd.f32	s0 , s0, s1				// set return value
+#endif
+
+#if !defined(__ARM_PCS_VFP)
+#if !defined(DOUBLE)
+	vmov	r0, s0
+#else
+	vmov	r0, r1, d0
+#endif
+#endif
+
+	bx	lr
+
+	EPILOGUE
+

kernel/arm/zsum.c

@@ -0,0 +1,57 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+/**************************************************************************************
+* trivial copy of zasum.c with the ABS() removed                                      *
+**************************************************************************************/
+
+
+#include "common.h"
+#include <math.h>
+
+#define CSUM1(x,i)	x[i]+x[i+1]
+
+FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
+{
+	BLASLONG i=0;
+	FLOAT sumf = 0.0;
+	BLASLONG inc_x2;
+
+	if (n <= 0 || inc_x <= 0) return(sumf);
+
+	inc_x2 = 2 * inc_x;
+
+	n *= inc_x2;
+	while(i < n)
+	{
+		sumf += CSUM1(x,i);
+		i += inc_x2;
+	}
+	return(sumf);
+}
+
+

kernel/arm64/KERNEL.TSV110

@@ -0,0 +1,175 @@
+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
+
+STRMMKERNEL	= ../generic/trmmkernel_4x4.c
+DTRMMKERNEL	= ../generic/trmmkernel_2x2.c
+CTRMMKERNEL	= ../generic/ztrmmkernel_2x2.c
+ZTRMMKERNEL	= ../generic/ztrmmkernel_2x2.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
+
+ISAMAXKERNEL = iamax.S
+IDAMAXKERNEL = iamax.S
+ICAMAXKERNEL = izamax.S
+IZAMAXKERNEL = izamax.S
+
+SASUMKERNEL  = asum.S
+DASUMKERNEL  = asum.S
+CASUMKERNEL  = casum.S
+ZASUMKERNEL  = zasum.S
+
+SAXPYKERNEL  = axpy.S
+DAXPYKERNEL  = axpy.S
+CAXPYKERNEL  = zaxpy.S
+ZAXPYKERNEL  = zaxpy.S
+
+SCOPYKERNEL  = copy.S
+DCOPYKERNEL  = copy.S
+CCOPYKERNEL  = copy.S
+ZCOPYKERNEL  = copy.S
+
+SDOTKERNEL   = dot.S
+DDOTKERNEL   = dot.S
+CDOTKERNEL   = zdot.S
+ZDOTKERNEL   = zdot.S
+DSDOTKERNEL  = dot.S
+
+SNRM2KERNEL  = nrm2.S
+DNRM2KERNEL  = nrm2.S
+CNRM2KERNEL  = znrm2.S
+ZNRM2KERNEL  = znrm2.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
+
+SSWAPKERNEL  = swap.S
+DSWAPKERNEL  = swap.S
+CSWAPKERNEL  = swap.S
+ZSWAPKERNEL  = swap.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
+
+SGEMMKERNEL    =  sgemm_kernel_$(SGEMM_UNROLL_M)x$(SGEMM_UNROLL_N).S
+STRMMKERNEL    =  strmm_kernel_$(SGEMM_UNROLL_M)x$(SGEMM_UNROLL_N).S
+ifneq ($(SGEMM_UNROLL_M), $(SGEMM_UNROLL_N))
+SGEMMINCOPY    =  ../generic/gemm_ncopy_$(SGEMM_UNROLL_M).c
+SGEMMITCOPY    =  ../generic/gemm_tcopy_$(SGEMM_UNROLL_M).c
+SGEMMINCOPYOBJ =  sgemm_incopy$(TSUFFIX).$(SUFFIX)
+SGEMMITCOPYOBJ =  sgemm_itcopy$(TSUFFIX).$(SUFFIX)
+endif
+SGEMMONCOPY    =  ../generic/gemm_ncopy_$(SGEMM_UNROLL_N).c
+SGEMMOTCOPY    =  ../generic/gemm_tcopy_$(SGEMM_UNROLL_N).c
+SGEMMONCOPYOBJ =  sgemm_oncopy$(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/csum.S

@@ -0,0 +1,164 @@
+/*******************************************************************************
+Copyright (c) 2019, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*******************************************************************************/
+
+#define ASSEMBLER
+#include "common.h"
+
+#define	N	x0	/* vector length */
+#define	X	x1	/* X vector address */
+#define	INC_X	x2	/* X stride */
+#define I	x5	/* loop variable */
+
+/*******************************************************************************
+* Macro definitions
+*******************************************************************************/
+
+#define REG0	wzr
+#define SUMF	s0
+#define TMPF	s1
+#define TMPVF	{v1.s}[0]
+#define SZ	4
+
+/******************************************************************************/
+
+.macro KERNEL_F1
+	ld1	{v1.2s}, [X], #8
+	ext	v2.8b, v1.8b, v1.8b, #4
+	fadd	TMPF, TMPF, s2
+	fadd	SUMF, SUMF, TMPF
+.endm
+
+.macro KERNEL_F8
+	ld1	{v1.4s, v2.4s, v3.4s, v4.4s}, [X]
+	add	X, X, #64
+
+	PRFM	PLDL1KEEP, [X, #1024]
+
+	fadd	v1.4s, v1.4s, v2.4s
+	fadd	v3.4s, v3.4s, v4.4s
+	fadd	v0.4s, v0.4s, v1.4s
+	fadd	v0.4s, v0.4s, v3.4s
+.endm
+
+.macro KERNEL_F8_FINALIZE
+	ext	v1.16b, v0.16b, v0.16b, #8
+	fadd	v0.2s, v0.2s, v1.2s
+	faddp	SUMF, v0.2s
+.endm
+
+.macro INIT_S
+	lsl	INC_X, INC_X, #3
+.endm
+
+.macro KERNEL_S1
+	ld1	{v1.2s}, [X], INC_X
+	ext	v2.8b, v1.8b, v1.8b, #4
+	fadd	TMPF, TMPF, s2
+	fadd	SUMF, SUMF, TMPF
+
+.endm
+
+/*******************************************************************************
+* End of macro definitions
+*******************************************************************************/
+
+	PROLOGUE
+
+	fmov	SUMF, REG0
+	fmov	s1, SUMF
+
+	cmp	N, xzr
+	ble	.Lcsum_kernel_L999
+	cmp	INC_X, xzr
+	ble	.Lcsum_kernel_L999
+
+	cmp	INC_X, #1
+	bne	.Lcsum_kernel_S_BEGIN
+
+.Lcsum_kernel_F_BEGIN:
+
+	asr	I, N, #3
+	cmp	I, xzr
+	beq	.Lcsum_kernel_F1
+
+.Lcsum_kernel_F8:
+
+	KERNEL_F8
+
+	subs	I, I, #1
+	bne	.Lcsum_kernel_F8
+
+	KERNEL_F8_FINALIZE
+
+.Lcsum_kernel_F1:
+
+	ands	I, N, #7
+	ble	.Lcsum_kernel_L999
+
+.Lcsum_kernel_F10:
+
+	KERNEL_F1
+
+	subs    I, I, #1
+        bne     .Lcsum_kernel_F10
+
+.Lcsum_kernel_L999:
+	ret
+
+.Lcsum_kernel_S_BEGIN:
+
+	INIT_S
+
+	asr	I, N, #2
+	cmp	I, xzr
+	ble	.Lcsum_kernel_S1
+
+.Lcsum_kernel_S4:
+
+	KERNEL_S1
+	KERNEL_S1
+	KERNEL_S1
+	KERNEL_S1
+
+	subs	I, I, #1
+	bne	.Lcsum_kernel_S4
+
+.Lcsum_kernel_S1:
+
+	ands	I, N, #3
+	ble	.Lcsum_kernel_L999
+
+.Lcsum_kernel_S10:
+
+	KERNEL_S1
+
+	subs    I, I, #1
+        bne     .Lcsum_kernel_S10
+
+	ret
+
+	EPILOGUE

kernel/arm64/sum.S

@@ -0,0 +1,186 @@
+/*******************************************************************************
+Copyright (c) 2019, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*******************************************************************************/
+
+#define ASSEMBLER
+#include "common.h"
+
+#define	N	x0	/* vector length */
+#define	X	x1	/* X vector address */
+#define	INC_X	x2	/* X stride */
+#define I	x5	/* loop variable */
+
+/*******************************************************************************
+* Macro definitions
+*******************************************************************************/
+
+#if !defined(DOUBLE)
+#define REG0	wzr
+#define SUMF	s0
+#define TMPF	s1
+#define TMPVF	{v1.s}[0]
+#define SZ	4
+#else
+#define REG0	xzr
+#define SUMF	d0
+#define TMPF	d1
+#define TMPVF	{v1.d}[0]
+#define SZ	8
+#endif
+
+/******************************************************************************/
+
+.macro KERNEL_F1
+	ldr	TMPF, [X], #SZ
+	fadd	SUMF, SUMF, TMPF
+.endm
+
+.macro KERNEL_F8
+#if !defined(DOUBLE)
+	ld1	{v1.4s, v2.4s}, [X], #32	// Load [X3, X2, X1, X0]
+	fadd	v1.4s, v1.4s, v2.4s		// [X3+X1, X2+X0]
+	fadd	v0.4s, v0.4s, v1.4s		// [X3+X1, X2+X0]
+	PRFM	PLDL1KEEP, [X, #1024]
+#else // DOUBLE
+	ld1	{v2.2d, v3.2d, v4.2d, v5.2d}, [X]
+	add	X, X, #64
+
+	PRFM	PLDL1KEEP, [X, #1024]
+
+	fadd	v2.2d, v2.2d, v3.2d
+	fadd	v4.2d, v4.2d, v5.2d
+	fadd	v0.2d, v0.2d, v2.2d
+	fadd	v0.2d, v0.2d, v4.2d
+#endif
+.endm
+
+.macro KERNEL_F8_FINALIZE
+#if !defined(DOUBLE)
+	ext	v1.16b, v0.16b, v0.16b, #8
+	fadd	v0.2s, v0.2s, v1.2s
+	faddp	SUMF, v0.2s
+#else
+	faddp	SUMF, v0.2d
+#endif
+.endm
+
+.macro INIT_S
+#if !defined(DOUBLE)
+	lsl	INC_X, INC_X, #2
+#else
+	lsl	INC_X, INC_X, #3
+#endif
+.endm
+
+.macro KERNEL_S1
+	ld1	TMPVF, [X], INC_X
+	fadd	SUMF, SUMF, TMPF
+.endm
+
+/*******************************************************************************
+* End of macro definitions
+*******************************************************************************/
+
+	PROLOGUE
+
+	fmov	SUMF, REG0
+#if !defined(DOUBLE)
+	fmov	s1, SUMF
+#else
+	fmov	d1, SUMF
+#endif
+
+	cmp	N, xzr
+	ble	.Lsum_kernel_L999
+	cmp	INC_X, xzr
+	ble	.Lsum_kernel_L999
+
+	cmp	INC_X, #1
+	bne	.Lsum_kernel_S_BEGIN
+
+.Lsum_kernel_F_BEGIN:
+
+	asr	I, N, #3
+	cmp	I, xzr
+	beq	.Lsum_kernel_F1
+
+.Lsum_kernel_F8:
+
+	KERNEL_F8
+
+	subs	I, I, #1
+	bne	.Lsum_kernel_F8
+
+	KERNEL_F8_FINALIZE
+
+.Lsum_kernel_F1:
+
+	ands	I, N, #7
+	ble	.Lsum_kernel_L999
+
+.Lsum_kernel_F10:
+
+	KERNEL_F1
+
+	subs    I, I, #1
+        bne     .Lsum_kernel_F10
+
+.Lsum_kernel_L999:
+	ret
+
+.Lsum_kernel_S_BEGIN:
+
+	INIT_S
+
+	asr	I, N, #2
+	cmp	I, xzr
+	ble	.Lsum_kernel_S1
+
+.Lsum_kernel_S4:
+
+	KERNEL_S1
+	KERNEL_S1
+	KERNEL_S1
+	KERNEL_S1
+
+	subs	I, I, #1
+	bne	.Lsum_kernel_S4
+
+.Lsum_kernel_S1:
+
+	ands	I, N, #3
+	ble	.Lsum_kernel_L999
+
+.Lsum_kernel_S10:
+
+	KERNEL_S1
+
+	subs    I, I, #1
+        bne     .Lsum_kernel_S10
+
+	ret
+
+	EPILOGUE

kernel/arm64/zsum.S

@@ -0,0 +1,158 @@
+/*******************************************************************************
+Copyright (c) 2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*******************************************************************************/
+
+#define ASSEMBLER
+#include "common.h"
+
+#define	N	x0	/* vector length */
+#define	X	x1	/* X vector address */
+#define	INC_X	x2	/* X stride */
+#define I	x5	/* loop variable */
+
+/*******************************************************************************
+* Macro definitions
+*******************************************************************************/
+
+#define REG0	xzr
+#define SUMF	d0
+#define TMPF	d1
+#define TMPVF	{v1.d}[0]
+#define SZ	8
+
+/******************************************************************************/
+
+.macro KERNEL_F1
+	ld1	{v1.2d}, [X], #16
+	faddp	TMPF, v1.2d
+	fadd	SUMF, SUMF, TMPF
+.endm
+
+.macro KERNEL_F4
+	ld1	{v1.2d, v2.2d, v3.2d, v4.2d}, [X], #64
+
+	fadd	v1.2d, v1.2d, v2.2d
+	fadd	v3.2d, v3.2d, v4.2d
+
+	fadd	v0.2d, v0.2d, v1.2d
+	fadd	v0.2d, v0.2d, v3.2d
+
+	PRFM	PLDL1KEEP, [X, #1024]
+.endm
+
+.macro KERNEL_F4_FINALIZE
+	faddp	SUMF, v0.2d
+.endm
+
+.macro INIT_S
+	lsl	INC_X, INC_X, #4
+.endm
+
+.macro KERNEL_S1
+	ld1	{v1.2d}, [X], INC_X
+	faddp	TMPF, v1.2d
+	fadd	SUMF, SUMF, TMPF
+.endm
+
+/*******************************************************************************
+* End of macro definitions
+*******************************************************************************/
+
+	PROLOGUE
+
+	fmov	SUMF, REG0
+
+	cmp	N, xzr
+	ble	.Lzsum_kernel_L999
+	cmp	INC_X, xzr
+	ble	.Lzsum_kernel_L999
+
+	cmp	INC_X, #1
+	bne	.Lzsum_kernel_S_BEGIN
+
+.Lzsum_kernel_F_BEGIN:
+
+	asr	I, N, #2
+	cmp	I, xzr
+	beq	.Lzsum_kernel_F1
+
+.Lzsum_kernel_F4:
+
+	KERNEL_F4
+
+	subs	I, I, #1
+	bne	.Lzsum_kernel_F4
+
+	KERNEL_F4_FINALIZE
+
+.Lzsum_kernel_F1:
+
+	ands	I, N, #3
+	ble	.Lzsum_kernel_L999
+
+.Lzsum_kernel_F10:
+
+	KERNEL_F1
+
+	subs    I, I, #1
+        bne     .Lzsum_kernel_F10
+
+.Lzsum_kernel_L999:
+	ret
+
+.Lzsum_kernel_S_BEGIN:
+
+	INIT_S
+
+	asr	I, N, #2
+	cmp	I, xzr
+	ble	.Lzsum_kernel_S1
+
+.Lzsum_kernel_S4:
+
+	KERNEL_S1
+	KERNEL_S1
+	KERNEL_S1
+	KERNEL_S1
+
+	subs	I, I, #1
+	bne	.Lzsum_kernel_S4
+
+.Lzsum_kernel_S1:
+
+	ands	I, N, #3
+	ble	.Lzsum_kernel_L999
+
+.Lzsum_kernel_S10:
+
+	KERNEL_S1
+
+	subs    I, I, #1
+        bne     .Lzsum_kernel_S10
+
+	ret
+
+	EPILOGUE

kernel/ia64/KERNEL

@@ -60,6 +60,10 @@ CASUMKERNEL  = asum.S
 ZASUMKERNEL  = asum.S
 XASUMKERNEL  = asum.S
 
+CSUMKERNEL  = sum.S
+ZSUMKERNEL  = sum.S
+XSUMKERNEL  = sum.S
+
 CNRM2KERNEL  = nrm2.S
 ZNRM2KERNEL  = nrm2.S
 XNRM2KERNEL  = nrm2.S

kernel/ia64/sum.S

@@ -0,0 +1,358 @@
+/*********************************************************************/
+/* Copyright 2009, 2010 The University of Texas at Austin.           */
+/* Copyright 2019, The OpenBLAS project                              */
+/* All rights reserved.                                              */
+/*                                                                   */
+/* Redistribution and use in source and binary forms, with or        */
+/* without modification, are permitted provided that the following   */
+/* conditions are met:                                               */
+/*                                                                   */
+/*   1. Redistributions of source code must retain the above         */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer.                                                  */
+/*                                                                   */
+/*   2. Redistributions in binary form must reproduce the above      */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer in the documentation and/or other materials       */
+/*      provided with the distribution.                              */
+/*                                                                   */
+/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */
+/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */
+/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */
+/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */
+/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */
+/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */
+/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */
+/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */
+/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */
+/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */
+/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */
+/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */
+/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */
+/*    POSSIBILITY OF SUCH DAMAGE.                                    */
+/*                                                                   */
+/* The views and conclusions contained in the software and           */
+/* documentation are those of the authors and should not be          */
+/* interpreted as representing official policies, either expressed   */
+/* or implied, of The University of Texas at Austin.                 */
+/*********************************************************************/
+
+#define ASSEMBLER
+#include "common.h"
+
+#ifdef XDOUBLE
+#define PREFETCH_SIZE ( 8 * 16 +  4)
+#elif defined(DOUBLE)
+#define PREFETCH_SIZE (16 * 16 +  8)
+#else
+#define PREFETCH_SIZE (32 * 16 + 16)
+#endif
+
+#ifndef COMPLEX
+#define COMPADD	0
+#define STRIDE INCX
+#else
+#define COMPADD	1
+#define STRIDE SIZE
+#endif
+
+#define PRE1	r2
+
+#define I	r17
+#define J	r18
+#define INCX16	r21
+
+#define PR	r30
+#define ARLC	r31
+
+#define N	r32
+#define X	r33
+#define INCX	r34
+
+
+	PROLOGUE
+	.prologue
+	PROFCODE
+	{ .mfi
+	adds	PRE1 = PREFETCH_SIZE * SIZE, X
+	mov	f8   = f0
+	.save ar.lc, ARLC
+	mov	ARLC = ar.lc
+	}
+	;;
+	.body
+#ifdef F_INTERFACE
+	{ .mmi
+	LDINT	N    = [N]
+	LDINT	INCX = [INCX]
+	nop.i	0
+	}
+	;;
+#ifndef USE64BITINT
+	{ .mii
+	nop.m	0
+	sxt4	N = N
+	sxt4	INCX = INCX
+	}
+	;;
+#endif
+#endif
+	{ .mmi
+	cmp.lt	p0, p6 = r0, INCX
+	cmp.lt	p0, p7 = r0, N
+	shr	I =  N, (4 - COMPADD)
+	}
+	{ .mbb
+	and	J = ((1 << (4 - COMPADD)) - 1), N
+	(p6) 	br.ret.sptk.many b0
+	(p7) 	br.ret.sptk.many b0
+	}
+	;;
+	{ .mfi
+	adds	I = -1, I
+	mov	f10 = f0
+	mov	PR = pr
+	}
+	{ .mfi
+	cmp.eq	p9, p0  =   r0, J
+	mov	f9  = f0
+	tbit.z	p0, p12 = N, 3 - COMPADD
+	}
+	;;
+	{ .mmi
+	cmp.eq	p16, p0 = r0, r0
+	cmp.ne	p17, p0 = r0, r0
+	mov	ar.ec= 3
+	}
+	{ .mfi
+	cmp.ne	p18, p0 = r0, r0
+	mov	f11 = f0
+	shl	INCX = INCX, BASE_SHIFT + COMPADD
+	}
+	;;
+	{ .mmi
+#ifdef XDOUBLE
+	shladd	INCX16  = INCX, (3 - COMPADD), r0
+#else
+	shladd	INCX16  = INCX, (4 - COMPADD), r0
+#endif
+	cmp.ne	p19, p0 = r0, r0
+	mov	ar.lc = I
+	}
+	{ .mmb
+	cmp.gt	p8 ,p0  =   r0, I
+#ifdef COMPLEX
+       adds	INCX = - SIZE, INCX
+#else
+	nop.m	0
+#endif
+	(p8) br.cond.dpnt  .L55
+	}
+	;;
+	.align 32
+
+.L52:
+	{ .mmf
+	(p16) lfetch.nt1 [PRE1], INCX16
+	(p16) LDFD	f32  = [X], STRIDE
+	}
+	{ .mfb
+	(p19) FADD	f8  = f8,  f71
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f35  = [X], INCX
+	}
+	{ .mfb
+	(p19) FADD	f9  = f9,  f74
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f38  = [X], STRIDE
+	}
+	{ .mfb
+	(p19) FADD	f10 = f10, f77
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f41  = [X], INCX
+	}
+	{ .mfb
+	(p19) FADD	f11 = f11, f80
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f44  = [X], STRIDE
+	}
+	{ .mfb
+	(p18) FADD	f8  = f8,  f34
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f47  = [X], INCX
+	}
+	{ .mfb
+	(p18) FADD	f9  = f9,  f37
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f50  = [X], STRIDE
+	}
+	{ .mfb
+	(p18) FADD	f10 = f10, f40
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f53  = [X], INCX
+	}
+	{ .mfb
+	(p18) FADD	f11 = f11, f43
+	}
+	;;
+	{ .mmf
+#ifdef XDOUBLE
+	(p16) lfetch.nt1 [PRE1], INCX16
+#endif
+	(p16) LDFD	f56  = [X], STRIDE
+	}
+	{ .mfb
+	(p18) FADD	f8  = f8,  f46
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f59  = [X], INCX
+	}
+	{ .mfb
+	(p18) FADD	f9  = f9,  f49
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f62  = [X], STRIDE
+	}
+	{ .mfb
+	(p18) FADD	f10 = f10, f52
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f65  = [X], INCX
+	}
+	{ .mfb
+	(p18) FADD	f11 = f11, f55
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f68  = [X], STRIDE
+	}
+	{ .mfb
+	(p18) FADD	f8  = f8,  f58
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f71  = [X], INCX
+	}
+	{ .mfb
+	(p18) FADD	f9  = f9,  f61
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f74  = [X], STRIDE
+	}
+	{ .mfb
+	(p18) FADD	f10 = f10, f64
+	}
+	;;
+	{ .mmf
+	(p16) LDFD	f77  = [X], INCX
+	}
+	{ .mfb
+	(p18) FADD	f11 = f11, f67
+	br.ctop.sptk.few .L52
+	}
+	;;
+	FADD	f8  = f8,  f71
+	FADD	f9  = f9,  f74
+	FADD	f10 = f10, f77
+	FADD	f11 = f11, f80
+	.align 32
+	;;
+.L55:
+	(p12) LDFD	f32  = [X], STRIDE
+	(p9) br.cond.dptk .L998
+	;;
+	(p12) LDFD	f33  = [X], INCX
+	;;
+	(p12) LDFD	f34  = [X], STRIDE
+	;;
+	(p12) LDFD	f35  = [X], INCX
+	tbit.z	p0, p13 = N, (2 - COMPADD)
+	;;
+	(p12) LDFD	f36  = [X], STRIDE
+	tbit.z	p0, p14 = N, (1 - COMPADD)
+	;;
+	(p12) LDFD	f37  = [X], INCX
+#ifndef COMPLEX
+	tbit.z	p0, p15 = N, 0
+#endif
+	;;
+	(p12) LDFD	f38  = [X], STRIDE
+	;;
+	(p12) LDFD	f39  = [X], INCX
+	;;
+	(p13) LDFD	f40  = [X], STRIDE
+	;;
+	(p13) LDFD	f41  = [X], INCX
+	;;
+	(p13) LDFD	f42  = [X], STRIDE
+	(p12) FADD	f8  = f8,  f32
+	;;
+	(p13) LDFD	f43  = [X], INCX
+	(p12) FADD	f9  = f9,  f33
+	;;
+	(p14) LDFD	f44  = [X], STRIDE
+	(p12) FADD	f10 = f10, f34
+	;;
+	(p14) LDFD	f45  = [X], INCX
+	(p12) FADD	f11 = f11, f35
+	;;
+#ifndef COMPLEX
+	(p15) LDFD	f46  = [X]
+#endif
+	(p12) FADD	f8  = f8,  f36
+	;;
+	(p12) FADD	f9  = f9,  f37
+	(p12) FADD	f10 = f10, f38
+	(p12) FADD	f11 = f11, f39
+	;;
+	(p13) FADD	f8  = f8,  f40
+	(p13) FADD	f9  = f9,  f41
+#ifndef COMPLEX
+#endif
+	(p13) FADD	f10 = f10, f42
+	;;
+	(p13) FADD	f11 = f11, f43
+	(p14) FADD	f8  = f8,  f44
+	(p14) FADD	f9  = f9,  f45
+#ifndef COMPLEX
+	(p15) FADD	f10 = f10, f46
+#endif
+	;;
+	.align 32
+
+.L998:
+	{ .mfi
+	FADD	f8  = f8,  f9
+	mov	ar.lc  = ARLC
+	}
+	{ .mmf
+	FADD	f10 = f10, f11
+	}
+	;;
+	{ .mii
+	mov	pr = PR, -65474
+	}
+	;;
+	{ .mfb
+	FADD	f8  = f8,  f10
+	br.ret.sptk.many b0
+	}
+	EPILOGUE

kernel/mips/KERNEL.P5600

@@ -30,6 +30,11 @@ IDMAXKERNEL  = ../mips/imax.c
 ISMINKERNEL  = ../mips/imin.c
 IDMINKERNEL  = ../mips/imin.c
 
+SSUMKERNEL  = ../mips/sum.c
+DSUMKERNEL  = ../mips/sum.c
+CSUMKERNEL  = ../mips/zsum.c
+ZSUMKERNEL  = ../mips/zsum.c
+
 ifdef HAVE_MSA
 SASUMKERNEL  = ../mips/sasum_msa.c
 DASUMKERNEL  = ../mips/dasum_msa.c

kernel/mips/imin.c

@@ -45,7 +45,7 @@ BLASLONG CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
 
 	while(i < n)
 	{
-		if( x[ix] > minf )
+		if( x[ix] < minf )
 		{
 			min = i;
 			minf = x[ix];