floraachy
/
OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge branch 'develop' into loongarch64_rename_targets

This commit is contained in:
XiWeiGu 2022-08-29 10:17:38 +08:00 committed by GitHub
parent 26eee97b38 2847636ecd
commit acd48edf99
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
66 changed files with 369 additions and 197 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

50
.github/workflows/dynamic_arch.yml vendored
View File

@ -257,3 +257,53 @@ jobs:
- name: Run tests
timeout-minutes: 60
run: cd build && ctest
cross_build:
runs-on: ubuntu-22.04
strategy:
fail-fast: false
matrix:
include:
- target: mips64el
triple: mips64el-linux-gnuabi64
opts: DYNAMIC_ARCH=1
- target: riscv64
triple: riscv64-linux-gnu
opts: TARGET=RISCV64_GENERIC
- target: mipsel
triple: mipsel-linux-gnu
opts: TARGET=MIPS1004K
- target: alpha
triple: alpha-linux-gnu
opts: TARGET=EV4
steps:
- name: Checkout repository
uses: actions/checkout@v3
- name: Install Dependencies
run: |
sudo apt-get install -y ccache gcc-${{ matrix.triple }} gfortran-${{ matrix.triple }} libgomp1-${{ matrix.target }}-cross
- name: Compilation cache
uses: actions/cache@v3
with:
path: ~/.ccache
key: ccache-${{ runner.os }}-${{ matrix.target }}-${{ github.ref }}-${{ github.sha }}
restore-keys: |
ccache-${{ runner.os }}-${{ matrix.target }}-${{ github.ref }}
ccache-${{ runner.os }}-${{ matrix.target }}
- name: Configure ccache
run: |
# Limit the maximum size and switch on compression to avoid exceeding the total disk or cache quota (5 GB).
test -d ~/.ccache || mkdir -p ~/.ccache
echo "max_size = 300M" > ~/.ccache/ccache.conf
echo "compression = true" >> ~/.ccache/ccache.conf
ccache -s
- name: Build OpenBLAS
run: |
make -j$(nproc) HOSTCC="ccache gcc" CC="ccache ${{ matrix.triple }}-gcc" FC="ccache ${{ matrix.triple }}-gfortran" ARCH=${{ matrix.target }} ${{ matrix.opts }}

6
.travis.yml
View File

@ -30,7 +30,7 @@ matrix:
before_script: &common-before
- COMMON_FLAGS="DYNAMIC_ARCH=1 TARGET=POWER8 NUM_THREADS=32"
script:
- travis_wait 20 make QUIET_MAKE=1 $COMMON_FLAGS $BTYPE
- travis_wait 40 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
@ -104,7 +104,7 @@ matrix:
- sudo apt-get update
- sudo apt-get install gcc-9 gfortran-9 -y
script:
- travis_wait 20 make QUIET_MAKE=1 BINARY=64 USE_OPENMP=1 CC=gcc-9 FC=gfortran-9
- travis_wait 40 make QUIET_MAKE=1 BINARY=64 USE_OPENMP=1 CC=gcc-9 FC=gfortran-9
- make -C test $COMMON_FLAGS $BTYPE
- make -C ctest $COMMON_FLAGS $BTYPE
- make -C utest $COMMON_FLAGS $BTYPE
@ -121,7 +121,7 @@ matrix:
- sudo apt-get update
- sudo apt-get install gcc-9 gfortran-9 -y
script:
- travis_wait 20 make QUIET_MAKE=1 BUILD_BFLOAT16=1 BINARY=64 USE_OPENMP=1 CC=gcc-9 FC=gfortran-9
- travis_wait 40 make QUIET_MAKE=1 BUILD_BFLOAT16=1 BINARY=64 USE_OPENMP=1 CC=gcc-9 FC=gfortran-9
- make -C test $COMMON_FLAGS $BTYPE
- make -C ctest $COMMON_FLAGS $BTYPE
- make -C utest $COMMON_FLAGS $BTYPE

3
CONTRIBUTORS.md
View File

@ -211,4 +211,5 @@ In chronological order:
* PLCT Lab, Institute of Software Chinese Academy of Sciences
* [2022-03] Support RISC-V Vector Intrinisc 1.0 version.
* Pablo Romero <https://github.com/pablorcum>
* [2022-08] Fix building from sources for QNX

7
Makefile
View File

@ -110,6 +110,10 @@ ifeq ($(OSNAME), Darwin)
@echo "\"make PREFIX=/your_installation_path/ install\"."
@echo
@echo "(or set PREFIX in Makefile.rule and run make install."
@echo
@echo "Note that any flags passed to make during build should also be passed to make install"
@echo "to circumvent any install errors."
@echo
@echo "If you want to move the .dylib to a new location later, make sure you change"
@echo "the internal name of the dylib with:"
@echo
@ -118,6 +122,9 @@ endif
@echo
@echo "To install the library, you can run \"make PREFIX=/path/to/your/installation install\"."
@echo
@echo "Note that any flags passed to make during build should also be passed to make install"
@echo "to circumvent any install errors."
@echo
shared :
ifneq ($(NO_SHARED), 1)

42
Makefile.alpha
View File

@ -1,42 +1,24 @@
CPP = $(CC) -E
RANLIB = ranlib
ifeq ($(LIBSUBARCH), EV4)
LIBNAME = $(LIBPREFIX)_ev4.a
LIBNAME_P = $(LIBPREFIX)_ev4_p.a
endif
ifeq ($(LIBSUBARCH), EV5)
LIBNAME = $(LIBPREFIX)_ev5.a
LIBNAME_P = $(LIBPREFIX)_ev5_p.a
endif
ifeq ($(LIBSUBARCH), EV6)
LIBNAME = $(LIBPREFIX)_ev6.a
LIBNAME_P = $(LIBPREFIX)_ev6_p.a
endif
ifneq ($(COMPILER), NATIVE)
# GCC User
ifeq ($(LIBSUBARCH), EV4)
OPTION += -DEV4 -mcpu=ev4
ifeq ($(CORE), EV4)
CCOMMON_OPT += -mcpu=ev4
endif
ifeq ($(LIBSUBARCH), EV5)
OPTION += -DEV5 -mcpu=ev5
ifeq ($(CORE), EV5)
CCOMMON_OPT += -mcpu=ev5
endif
ifeq ($(LIBSUBARCH), EV6)
OPTION += -DEV6 -mcpu=ev6
ifeq ($(CORE), EV6)
CCOMMON_OPT += -mcpu=ev6
endif
else
# Compaq Compiler User
ifeq ($(LIBSUBARCH), EV4)
OPTION += -DEV4 -tune ev4 -arch ev4
ifeq ($(CORE), EV4)
CCOMMON_OPT += -tune ev4 -arch ev4
endif
ifeq ($(LIBSUBARCH), EV5)
OPTION += -DEV5 -tune ev5 -arch ev5
ifeq ($(CORE), EV5)
CCOMMON_OPT += -tune ev5 -arch ev5
endif
ifeq ($(LIBSUBARCH), EV6)
OPTION += -DEV6 -tune ev6 -arch ev6
ifeq ($(CORE), EV6)
CCOMMON_OPT += -tune ev6 -arch ev6
endif
endif

8
Makefile.prebuild
View File

@ -60,9 +60,9 @@ all: getarch_2nd
./getarch_2nd 1 >> $(TARGET_CONF)
$(TARGET_CONF): c_check$(SCRIPTSUFFIX) f_check$(SCRIPTSUFFIX) getarch
./c_check$(SCRIPTSUFFIX) $(TARGET_MAKE) $(TARGET_CONF) $(CC) $(TARGET_FLAGS) $(CFLAGS)
./c_check$(SCRIPTSUFFIX) $(TARGET_MAKE) $(TARGET_CONF) "$(CC)" "$(TARGET_FLAGS) $(CFLAGS)"
ifneq ($(ONLY_CBLAS), 1)
./f_check$(SCRIPTSUFFIX) $(TARGET_MAKE) $(TARGET_CONF) $(FC) $(TARGET_FLAGS)
./f_check$(SCRIPTSUFFIX) $(TARGET_MAKE) $(TARGET_CONF) "$(FC)" "$(TARGET_FLAGS)"
else
#When we only build CBLAS, we set NOFORTRAN=2
echo "NOFORTRAN=2" >> $(TARGET_MAKE)
@ -77,8 +77,8 @@ endif
getarch : getarch.c cpuid.S dummy $(CPUIDEMU)
avx512=$$(./c_check$(SCRIPTSUFFIX) - - $(CC) $(TARGET_FLAGS) $(CFLAGS) | grep NO_AVX512); \
rv64gv=$$(./c_check$(SCRIPTSUFFIX) - - $(CC) $(TARGET_FLAGS) $(CFLAGS) | grep NO_RV64GV); \
avx512=$$(./c_check$(SCRIPTSUFFIX) - - "$(CC)" "$(TARGET_FLAGS) $(CFLAGS)" | grep NO_AVX512); \
rv64gv=$$(./c_check$(SCRIPTSUFFIX) - - "$(CC)" "$(TARGET_FLAGS) $(CFLAGS)" | grep NO_RV64GV); \
$(HOSTCC) $(HOST_CFLAGS) $(EXFLAGS) $${avx512:+-D$${avx512}} $${rv64gv:+-D$${rv64gv}} -lm -o $(@F) getarch.c cpuid.S $(CPUIDEMU)
getarch_2nd : getarch_2nd.c $(TARGET_CONF) dummy

12
Makefile.system
View File

@ -677,7 +677,12 @@ endif
endif
ifeq ($(ARCH), mips64)
DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4
DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
ifdef DYNAMIC_LIST
override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
endif
endif
ifeq ($(ARCH), loongarch64)
@ -856,6 +861,11 @@ CCOMMON_OPT += -mabi=32
BINARY_DEFINED = 1
endif
ifneq (, $(filter $(CORE), MIPS64_GENERIC))
CCOMMON_OPT += -DNO_MSA
FCOMMON_OPT += -DNO_MSA
endif
ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
CCOMMON_OPT += -march=loongson3a
FCOMMON_OPT += -march=loongson3a

5
TargetList.txt
View File

@ -65,6 +65,7 @@ MIPS1004K
MIPS24K
4.MIPS64 CPU:
MIPS64_GENERIC
SICORTEX
LOONGSON3A
LOONGSON3B
@ -128,3 +129,7 @@ LA264
12. Elbrus E2000:
E2K
13. Alpha
EV4
EV5
EV6

4
c_check
View File

@ -31,8 +31,8 @@ flags="$*"
cross_suffix=""
if [ "`dirname $compiler_name`" != '.' ]; then
cross_suffix="$cross_suffix`dirname $compiler_name`/"
if [ "`dirname \"$compiler_name\"`" != '.' ]; then
cross_suffix="$cross_suffix`dirname \"$compiler_name\"`/"
fi
bn=`basename $compiler_name`

4
common.h
View File

@ -90,7 +90,7 @@ extern "C" {
#endif
#include <time.h>
#ifdef OS_LINUX
#if defined(OS_LINUX) || defined(OS_QNX)
#include <malloc.h>
#include <sched.h>
#endif
@ -107,7 +107,7 @@ extern "C" {
#endif
#endif
#ifdef OS_HAIKU
#if defined(OS_HAIKU) || defined(OS_QNX)
#define NO_SYSV_IPC
#endif

2
common_alpha.h
View File

@ -43,7 +43,7 @@
#define MB asm("mb")
#define WMB asm("wmb")
#define RMB asm("rmb")
#define RMB asm("mb")
static void __inline blas_lock(unsigned long *address){
#ifndef __DECC

2
common_macro.h
View File

@ -2612,7 +2612,7 @@
#ifndef ASSEMBLER
#if !defined(DYNAMIC_ARCH) \
&& (defined(ARCH_X86) || defined(ARCH_X86_64) || defined(ARCH_IA64) || defined(ARCH_MIPS64) || defined(ARCH_ARM64) \
|| defined(ARCH_LOONGARCH64) || defined(ARCH_E2K))
|| defined(ARCH_LOONGARCH64) || defined(ARCH_E2K) || defined(ARCH_ALPHA))
extern BLASLONG gemm_offset_a;
extern BLASLONG gemm_offset_b;
extern BLASLONG sbgemm_p;

12
common_mips64.h
View File

@ -86,7 +86,9 @@ static inline unsigned int rpcc(void){
//__asm__ __volatile__("dmfc0 %0, $25, 1": "=r"(tmp):: "memory");
//ret=tmp;
__asm__ __volatile__(".set push \n"
#if !defined(__mips_isa_rev) || __mips_isa_rev < 2
".set mips32r2\n"
#endif
"rdhwr %0, $2\n"
".set pop": "=r"(ret):: "memory");
@ -99,7 +101,9 @@ static inline unsigned int rpcc(void){
static inline int WhereAmI(void){
int ret=0;
__asm__ __volatile__(".set push \n"
#if !defined(__mips_isa_rev) || __mips_isa_rev < 2
".set mips32r2\n"
#endif
"rdhwr %0, $0\n"
".set pop": "=r"(ret):: "memory");
return ret;
@ -197,9 +201,15 @@ static inline int blas_quickdivide(blasint x, blasint y){
#if defined(ASSEMBLER) && !defined(NEEDPARAM)
#if defined(__mips_isa_rev) && __mips_isa_rev >= 6
#define ASSEMBLER_ARCH mips64r6
#else
#define ASSEMBLER_ARCH mips64
#endif
#define PROLOGUE \
.text ;\
.set mips64 ;\
.set ASSEMBLER_ARCH ;\
.align 5 ;\
.globl REALNAME ;\
.ent REALNAME ;\

5
cpuid_alpha.c
View File

@ -59,6 +59,11 @@ void get_subarchitecture(void){
printf("ev%d", implver() + 4);
}
void get_corename(void){
printf("EV%d", implver() + 4);
}
void get_subdirname(void){
printf("alpha");
}

29
cpuid_mips64.c
View File

@ -70,16 +70,18 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
/* or implied, of The University of Texas at Austin. */
/*********************************************************************/
#define CPU_UNKNOWN 0
#define CPU_SICORTEX 1
#define CPU_LOONGSON3R3 2
#define CPU_LOONGSON3R4 3
#define CPU_I6400 4
#define CPU_P6600 5
#define CPU_I6500 6
#define CPU_UNKNOWN 0
#define CPU_MIPS64_GENERIC 1
#define CPU_SICORTEX 2
#define CPU_LOONGSON3R3 3
#define CPU_LOONGSON3R4 4
#define CPU_I6400 5
#define CPU_P6600 6
#define CPU_I6500 7
static char *cpuname[] = {
"UNKNOWN",
"MIPS64_GENERIC"
"SICORTEX",
"LOONGSON3R3",
"LOONGSON3R4",
@ -113,8 +115,11 @@ int detect(void){
return CPU_SICORTEX;
}
}
return CPU_MIPS64_GENERIC;
#else
return CPU_UNKNOWN;
#endif
return CPU_UNKNOWN;
}
char *get_corename(void){
@ -136,8 +141,10 @@ void get_subarchitecture(void){
printf("P6600");
}else if(detect()==CPU_I6500){
printf("I6500");
}else{
}else if(detect()==CPU_SICORTEX){
printf("SICORTEX");
}else{
printf("MIPS64_GENERIC");
}
}
@ -215,8 +222,8 @@ void get_libname(void){
printf("p6600\n");
}else if(detect()==CPU_I6500) {
printf("i6500\n");
}else{
printf("mips64\n");
}else {
printf("mips64_generic\n");
}
}

6
ctest/CMakeLists.txt
View File

@ -40,7 +40,7 @@ else()
c_${float_char}blas1.c)
endif()
target_link_libraries(x${float_char}cblat1 ${OpenBLAS_LIBNAME})
if(${CMAKE_SYSTEM_NAME} MATCHES "Linux" OR ${CMAKE_SYSTEM_NAME} MATCHES "FreeBSD")
if(${CMAKE_SYSTEM_NAME} MATCHES "Linux" OR ${CMAKE_SYSTEM_NAME} MATCHES "FreeBSD" OR ${CMAKE_SYSTEM_NAME} MATCHES "QNX")
target_link_libraries(x${float_char}cblat1 m)
endif()
add_test(NAME "x${float_char}cblat1"
@ -65,7 +65,7 @@ else()
constant.c)
endif()
target_link_libraries(x${float_char}cblat2 ${OpenBLAS_LIBNAME})
if(${CMAKE_SYSTEM_NAME} MATCHES "Linux" OR ${CMAKE_SYSTEM_NAME} MATCHES "FreeBSD")
if(${CMAKE_SYSTEM_NAME} MATCHES "Linux" OR ${CMAKE_SYSTEM_NAME} MATCHES "FreeBSD" OR ${CMAKE_SYSTEM_NAME} MATCHES "QNX")
target_link_libraries(x${float_char}cblat2 m)
endif()
add_test(NAME "x${float_char}cblat2"
@ -90,7 +90,7 @@ else()
constant.c)
endif()
target_link_libraries(x${float_char}cblat3 ${OpenBLAS_LIBNAME})
if(${CMAKE_SYSTEM_NAME} MATCHES "Linux" OR ${CMAKE_SYSTEM_NAME} MATCHES "FreeBSD")
if(${CMAKE_SYSTEM_NAME} MATCHES "Linux" OR ${CMAKE_SYSTEM_NAME} MATCHES "FreeBSD" OR ${CMAKE_SYSTEM_NAME} MATCHES "QNX")
target_link_libraries(x${float_char}cblat3 m)
endif()
add_test(NAME "x${float_char}cblat3"

125
driver/others/dynamic_mips64.c
View File

@ -38,22 +38,48 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include <sys/resource.h>
#include "common.h"
#if (defined OS_LINUX || defined OS_ANDROID)
#include <asm/hwcap.h>
#include <sys/auxv.h>
#ifndef HWCAP_LOONGSON_CPUCFG
#define HWCAP_LOONGSON_CPUCFG (1 << 14)
#endif
#endif
#ifdef DYNAMIC_LIST
extern gotoblas_t gotoblas_MIPS64_GENERIC;
#ifdef DYN_LOONGSON3R3
extern gotoblas_t gotoblas_LOONGSON3R3;
#else
#define gotoblas_LOONGSON3R3 gotoblas_MIPS64_GENERIC
#endif
#ifdef DYN_LOONGSON3R4
extern gotoblas_t gotoblas_LOONGSON3R4;
#else
#define gotoblas_LOONGSON3R4 gotoblas_MIPS64_GENERIC
#endif
#else
extern gotoblas_t gotoblas_LOONGSON3R3;
extern gotoblas_t gotoblas_LOONGSON3R4;
extern gotoblas_t gotoblas_MIPS64_GENERIC;
#endif
extern void openblas_warning(int verbose, const char * msg);
#define NUM_CORETYPES 2
#define NUM_CORETYPES 3
static char *corename[] = {
"MIPS64_GENERIC"
"loongson3r3",
"loongson3r4",
"UNKNOWN"
};
char *gotoblas_corename(void) {
if (gotoblas == &gotoblas_LOONGSON3R3) return corename[0];
if (gotoblas == &gotoblas_LOONGSON3R4) return corename[1];
if (gotoblas == &gotoblas_MIPS64_GENERIC) return corename[0];
if (gotoblas == &gotoblas_LOONGSON3R3) return corename[1];
if (gotoblas == &gotoblas_LOONGSON3R4) return corename[2];
return corename[NUM_CORETYPES];
}
@ -73,77 +99,32 @@ static gotoblas_t *force_coretype(char *coretype) {
switch (found)
{
case 0: return (&gotoblas_LOONGSON3R3);
case 1: return (&gotoblas_LOONGSON3R4);
case 0: return (&gotoblas_MIPS64_GENERIC);
case 1: return (&gotoblas_LOONGSON3R3);
case 2: return (&gotoblas_LOONGSON3R4);
}
snprintf(message, 128, "Core not found: %s\n", coretype);
openblas_warning(1, message);
return NULL;
}
#if (defined OS_LINUX || defined OS_ANDROID)
#define MMI_MASK 0x00000010
#define MSA_MASK 0x00000020
int fd[2];
int support_cpucfg;
static void handler(int signum)
{
close(fd[1]);
exit(1);
}
/* Brief : Function to check if cpucfg supported on loongson
* Return: 1 supported
* 0 not supported
*/
static int cpucfg_test(void) {
pid_t pid;
int status = 0;
support_cpucfg = 0;
pipe(fd);
pid = fork();
if (pid == 0) { /* Subprocess */
struct sigaction act;
close(fd[0]);
/* Set signal action for SIGILL. */
act.sa_handler = handler;
sigaction(SIGILL,&act,NULL);
/* Execute cpucfg in subprocess. */
__asm__ volatile(
".insn \n\t"
".word (0xc8080118) \n\t"
:::
);
support_cpucfg = 1;
write(fd[1],&support_cpucfg,sizeof(support_cpucfg));
close(fd[1]);
exit(0);
} else if (pid > 0){ /* Parent process*/
close(fd[1]);
if ((waitpid(pid,&status,0) <= 0) ||
(read(fd[0],&support_cpucfg,sizeof(support_cpucfg)) <= 0))
support_cpucfg = 0;
close(fd[0]);
} else {
support_cpucfg = 0;
}
return support_cpucfg;
}
static gotoblas_t *get_coretype_from_cpucfg(void) {
int flag = 0;
__asm__ volatile(
".set push \n\t"
".set noat \n\t"
".insn \n\t"
"dli $8, 0x01 \n\t"
".word (0xc9084918) \n\t"
"usw $9, 0x00(%0) \n\t"
"dli $1, 0x01 \n\t"
".word (0xc8080118) \n\t"
"move %0, $1 \n\t"
".set pop \n\t"
: "=r"(flag)
:
:
: "r"(&flag)
: "memory"
);
if (flag & MSA_MASK)
return (&gotoblas_LOONGSON3R4);
@ -153,7 +134,7 @@ static gotoblas_t *get_coretype_from_cpucfg(void) {
}
static gotoblas_t *get_coretype_from_cpuinfo(void) {
#ifdef linux
#ifdef __linux
FILE *infile;
char buffer[512], *p;
@ -176,17 +157,19 @@ static gotoblas_t *get_coretype_from_cpuinfo(void) {
return NULL;
}
#endif
return NULL;
return NULL;
}
#endif
static gotoblas_t *get_coretype(void) {
int ret = 0;
ret = cpucfg_test();
if (ret == 1)
return get_coretype_from_cpucfg();
else
return get_coretype_from_cpuinfo();
#if (!defined OS_LINUX && !defined OS_ANDROID)
return NULL;
#else
if (!(getauxval(AT_HWCAP) & HWCAP_LOONGSON_CPUCFG))
return get_coretype_from_cpucfg();
else
return get_coretype_from_cpuinfo();
#endif
}
void gotoblas_dynamic_init(void) {
@ -208,9 +191,9 @@ void gotoblas_dynamic_init(void) {
if (gotoblas == NULL)
{
snprintf(coremsg, 128, "Falling back to loongson3r3 core\n");
snprintf(coremsg, 128, "Falling back to MIPS64_GENEIRC\n");
openblas_warning(1, coremsg);
gotoblas = &gotoblas_LOONGSON3R3;
gotoblas = &gotoblas_MIPS64_GENERIC;
}
if (gotoblas && gotoblas->init) {

16
exports/gensymbol
View File

@ -4000,6 +4000,22 @@ case "$p1" in
no_underscore_objs="$no_underscore_objs $misc_common_objs"
printf 'int main(void){\n'
for obj in $underscore_objs; do
[ "$obj" != "xerbla" ] && printf 'extern void %s%s%s%s();\n' \
"$symbolprefix" "$obj" "$bu" "$symbolsuffix"
done
for obj in $need_2underscore_objs; do
printf 'extern void %s%s%s%s%s();\n' \
"$symbolprefix" "$obj" "$bu" "$bu" "$symbolsuffix"
done
for obj in $no_underscore_objs; do
printf 'extern void %s%s%s();\n' \
"$symbolprefix" "$obj" "$symbolsuffix"
done
printf '\n'
for obj in $underscore_objs; do
[ "$obj" != "xerbla" ] && printf '%s%s%s%s();\n' \
"$symbolprefix" "$obj" "$bu" "$symbolsuffix"

12
exports/gensymbol.pl
View File

@ -3955,6 +3955,18 @@ if ($ARGV[0] eq "linktest") {
@no_underscore_objs = (@no_underscore_objs, @misc_common_objs);
print "int main(void){\n";
foreach $objs (@underscore_objs) {
print "extern void ", $symbolprefix, $objs, $bu, $symbolsuffix, "();\n" if $objs ne "xerbla";
}
foreach $objs (@need_2underscore_objs) {
print "extern void ", $symbolprefix, $objs, $bu, $bu, $symbolsuffix, "();\n";
}
foreach $objs (@no_underscore_objs) {
print "extern void ", $symbolprefix, $objs, $symbolsuffix, "();\n";
}
foreach $objs (@underscore_objs) {
print $symbolprefix, $objs, $bu, $symbolsuffix, "();\n" if $objs ne "xerbla";
}

56
getarch.c
View File

@ -131,6 +131,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
/* #define FORCE_PPC440 */
/* #define FORCE_PPC440FP2 */
/* #define FORCE_CELL */
/* #define FORCE_MIPS64_GENERIC */
/* #define FORCE_SICORTEX */
/* #define FORCE_LOONGSON3R3 */
/* #define FORCE_LOONGSON3R4 */
@ -146,6 +147,9 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
/* #define FORCE_SPARCV7 */
/* #define FORCE_ZARCH_GENERIC */
/* #define FORCE_Z13 */
/* #define FORCE_EV4 */
/* #define FORCE_EV5 */
/* #define FORCE_EV6 */
/* #define FORCE_GENERIC */
#ifdef FORCE_P2
@ -915,6 +919,20 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define CORENAME "CELL"
#endif
#ifdef FORCE_MIPS64_GENERIC
#define FORCE
#define ARCHITECTURE "MIPS"
#define SUBARCHITECTURE "MIPS64_GENERIC"
#define SUBDIRNAME "mips64"
#define ARCHCONFIG "-DMIPS64_GENERIC " \
"-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
"-DL2_SIZE=1048576 -DL2_LINESIZE=32 " \
"-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
#define LIBNAME "mips64_generic"
#define CORENAME "MIPS64_GENERIC"
#else
#endif
#ifdef FORCE_SICORTEX
#define FORCE
#define ARCHITECTURE "MIPS"
@ -1601,6 +1619,42 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define CORENAME "Z14"
#endif
#ifdef FORCE_EV4
#define FORCE
#define ARCHITECTURE "ALPHA"
#define SUBARCHITECTURE "ev4"
#define ARCHCONFIG "-DEV4 " \
"-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=32 " \
"-DL2_SIZE=2097152 -DL2_LINESIZE=32 " \
"-DDTB_DEFAULT_ENTRIES=32 -DDTB_SIZE=8192 "
#define LIBNAME "ev4"
#define CORENAME "EV4"
#endif
#ifdef FORCE_EV5
#define FORCE
#define ARCHITECTURE "ALPHA"
#define SUBARCHITECTURE "ev5"
#define ARCHCONFIG "-DEV5 " \
"-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=32 " \
"-DL2_SIZE=2097152 -DL2_LINESIZE=64 " \
"-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=8192 "
#define LIBNAME "ev5"
#define CORENAME "EV5"
#endif
#ifdef FORCE_EV6
#define FORCE
#define ARCHITECTURE "ALPHA"
#define SUBARCHITECTURE "ev6"
#define ARCHCONFIG "-DEV6 " \
"-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
"-DL2_SIZE=4194304 -DL2_LINESIZE=64 " \
"-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=8192 "
#define LIBNAME "ev6"
#define CORENAME "EV6"
#endif
#ifdef FORCE_C910V
#define FORCE
#define ARCHITECTURE "RISCV64"
@ -1777,7 +1831,7 @@ int main(int argc, char *argv[]){
#ifdef FORCE
printf("CORE=%s\n", CORENAME);
#else
#if defined(INTEL_AMD) || defined(POWER) || defined(__mips__) || defined(__arm__) || defined(__aarch64__) || defined(ZARCH) || defined(sparc) || defined(__loongarch__) || defined(__riscv)
#if defined(INTEL_AMD) || defined(POWER) || defined(__mips__) || defined(__arm__) || defined(__aarch64__) || defined(ZARCH) || defined(sparc) || defined(__loongarch__) || defined(__riscv) || defined(__alpha__)
printf("CORE=%s\n", get_corename());
#endif
#endif

2
kernel/alpha/amax.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define N $16
#define X $17

2
kernel/alpha/asum.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCHSIZE 88

2
kernel/alpha/axpy.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCHSIZE 40

2
kernel/alpha/cabs.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
.set noat
.set noreorder

2
kernel/alpha/cnrm2.S
View File

@ -39,7 +39,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCH_SIZE 80

2
kernel/alpha/copy.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define N $16
#define X $17

2
kernel/alpha/cscal.S
View File

@ -42,7 +42,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
.globl NAME
.ent NAME

2
kernel/alpha/dnrm2.S
View File

@ -39,7 +39,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCH_SIZE 80

2
kernel/alpha/dot.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCHSIZE 88

2
kernel/alpha/gemm_beta.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
.set noat
.set noreorder

2
kernel/alpha/gemm_kernel_4x4.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#if !defined(EV4) && !defined(EV5) && !defined(EV6)
#error "Architecture is not specified."

2
kernel/alpha/gemv_n.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define STACKSIZE 64
#define PREFETCHSIZE 32

2
kernel/alpha/gemv_t.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define STACKSIZE 64
#define PREFETCHSIZE 32

2
kernel/alpha/iamax.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define N $16
#define X $17

2
kernel/alpha/imax.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define N $16
#define X $17

2
kernel/alpha/izamax.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define N $16
#define X $17

2
kernel/alpha/lsame.S
View File

@ -36,7 +36,7 @@
/* or implied, of The University of Texas at Austin. */
/*********************************************************************/
#include "version.h"
.set noat
.set noreorder

2
kernel/alpha/max.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define N $16
#define X $17

2
kernel/alpha/rot.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define N $16
#define X $17

2
kernel/alpha/scal.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCHSIZE 88

2
kernel/alpha/snrm2.S
View File

@ -39,7 +39,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCH_SIZE 80

2
kernel/alpha/sum.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCHSIZE 88

2
kernel/alpha/swap.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
PROLOGUE
PROFCODE

2
kernel/alpha/trsm_kernel_4x4_LN.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#if !defined(EV4) && !defined(EV5) && !defined(EV6)
#error "Architecture is not specified."

2
kernel/alpha/trsm_kernel_4x4_LT.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#if !defined(EV4) && !defined(EV5) && !defined(EV6)
#error "Architecture is not specified."

2
kernel/alpha/trsm_kernel_4x4_RT.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#if !defined(EV4) && !defined(EV5) && !defined(EV6)
#error "Architecture is not specified."

2
kernel/alpha/zamax.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define N $16
#define X $17

2
kernel/alpha/zasum.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCHSIZE 88

2
kernel/alpha/zaxpy.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCHSIZE 40

2
kernel/alpha/zdot.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCHSIZE 88

2
kernel/alpha/zgemm_beta.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
.set noat
.set noreorder

2
kernel/alpha/zgemm_kernel_2x2.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#if !defined(EV4) && !defined(EV5) && !defined(EV6)
#error "Architecture is not specified."

2
kernel/alpha/zgemv_n.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define STACKSIZE 64
#define PREFETCHSIZE 32

2
kernel/alpha/zgemv_t.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define STACKSIZE 64
#define PREFETCHSIZE 32

2
kernel/alpha/znrm2.S
View File

@ -39,7 +39,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCH_SIZE 80

2
kernel/alpha/zrot.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define N $16
#define X $17

2
kernel/alpha/zscal.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCHSIZE 88

2
kernel/alpha/zsum.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#define PREFETCHSIZE 88

2
kernel/alpha/zswap.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
PROLOGUE
PROFCODE

2
kernel/alpha/ztrsm_kernel_2x2_LN.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#if !defined(EV4) && !defined(EV5) && !defined(EV6)
#error "Architecture is not specified."

2
kernel/alpha/ztrsm_kernel_2x2_LT.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#if !defined(EV4) && !defined(EV5) && !defined(EV6)
#error "Architecture is not specified."

2
kernel/alpha/ztrsm_kernel_2x2_RT.S
View File

@ -38,7 +38,7 @@
#define ASSEMBLER
#include "common.h"
#include "version.h"
#if !defined(EV4) && !defined(EV5) && !defined(EV6)
#error "Architecture is not specified."

40
kernel/mips64/KERNEL
View File

@ -42,50 +42,58 @@ endif
ifndef SGEMMKERNEL
SGEMMKERNEL = gemm_kernel.S
ifneq ($(SGEMM_UNROLL_M), $(SGEMM_UNROLL_N))
SGEMMINCOPY = ../generic/gemm_ncopy_2.c
SGEMMITCOPY = ../generic/gemm_tcopy_2.c
SGEMMINCOPYOBJ = sgemm_incopy$(TSUFFIX).$(SUFFIX)
SGEMMITCOPYOBJ = sgemm_itcopy$(TSUFFIX).$(SUFFIX)
endif
SGEMMONCOPY = ../generic/gemm_ncopy_8.c
SGEMMOTCOPY = ../generic/gemm_tcopy_8.c
SGEMMINCOPYOBJ = sgemm_incopy.o
SGEMMITCOPYOBJ = sgemm_itcopy.o
SGEMMONCOPYOBJ = sgemm_oncopy.o
SGEMMOTCOPYOBJ = sgemm_otcopy.o
SGEMMONCOPYOBJ = sgemm_oncopy$(TSUFFIX).$(SUFFIX)
SGEMMOTCOPYOBJ = sgemm_otcopy$(TSUFFIX).$(SUFFIX)
endif
ifndef DGEMMKERNEL
DGEMMKERNEL = gemm_kernel.S
ifneq ($(DGEMM_UNROLL_M), $(DGEMM_UNROLL_N))
DGEMMINCOPY = ../generic/gemm_ncopy_2.c
DGEMMITCOPY = ../generic/gemm_tcopy_2.c
DGEMMINCOPYOBJ = dgemm_incopy$(TSUFFIX).$(SUFFIX)
DGEMMITCOPYOBJ = dgemm_itcopy$(TSUFFIX).$(SUFFIX)
endif
DGEMMONCOPY = ../generic/gemm_ncopy_8.c
DGEMMOTCOPY = ../generic/gemm_tcopy_8.c
DGEMMINCOPYOBJ = dgemm_incopy.o
DGEMMITCOPYOBJ = dgemm_itcopy.o
DGEMMONCOPYOBJ = dgemm_oncopy.o
DGEMMOTCOPYOBJ = dgemm_otcopy.o
DGEMMONCOPYOBJ = dgemm_oncopy$(TSUFFIX).$(SUFFIX)
DGEMMOTCOPYOBJ = dgemm_otcopy$(TSUFFIX).$(SUFFIX)
endif
ifndef CGEMMKERNEL
CGEMMKERNEL = zgemm_kernel.S
ifneq ($(CGEMM_UNROLL_M), $(CGEMM_UNROLL_N))
CGEMMINCOPY = ../generic/zgemm_ncopy_1.c
CGEMMITCOPY = ../generic/zgemm_tcopy_1.c
CGEMMINCOPYOBJ = cgemm_incopy$(TSUFFIX).$(SUFFIX)
CGEMMITCOPYOBJ = cgemm_itcopy$(TSUFFIX).$(SUFFIX)
endif
CGEMMONCOPY = ../generic/zgemm_ncopy_4.c
CGEMMOTCOPY = ../generic/zgemm_tcopy_4.c
CGEMMINCOPYOBJ = cgemm_incopy.o
CGEMMITCOPYOBJ = cgemm_itcopy.o
CGEMMONCOPYOBJ = cgemm_oncopy.o
CGEMMOTCOPYOBJ = cgemm_otcopy.o
CGEMMONCOPYOBJ = cgemm_oncopy$(TSUFFIX).$(SUFFIX)
CGEMMOTCOPYOBJ = cgemm_otcopy$(TSUFFIX).$(SUFFIX)
endif
ifndef ZGEMMKERNEL
ZGEMMKERNEL = zgemm_kernel.S
ifneq ($(ZGEMM_UNROLL_M), $(ZGEMM_UNROLL_N))
ZGEMMINCOPY = ../generic/zgemm_ncopy_1.c
ZGEMMITCOPY = ../generic/zgemm_tcopy_1.c
ZGEMMINCOPYOBJ = zgemm_incopy$(TSUFFIX).$(SUFFIX)
ZGEMMITCOPYOBJ = zgemm_itcopy$(TSUFFIX).$(SUFFIX)
endif
ZGEMMONCOPY = ../generic/zgemm_ncopy_4.c
ZGEMMOTCOPY = ../generic/zgemm_tcopy_4.c
ZGEMMINCOPYOBJ = zgemm_incopy.o
ZGEMMITCOPYOBJ = zgemm_itcopy.o
ZGEMMONCOPYOBJ = zgemm_oncopy.o
ZGEMMOTCOPYOBJ = zgemm_otcopy.o
ZGEMMONCOPYOBJ = zgemm_oncopy$(TSUFFIX).$(SUFFIX)
ZGEMMOTCOPYOBJ = zgemm_otcopy$(TSUFFIX).$(SUFFIX)
endif
ifndef SGEMM_BETA

11
lapack-netlib/SRC/iparam2stage.c
View File

@ -717,11 +717,12 @@ integer iparam2stage_(integer *ispec, char *name__, char *opts, integer *ni,
ret_val = -1;
// s_copy(subnam, name__, (ftnlen)12, name_len);
strncpy(subnam,name__,13);
subnam[13]='\0';
for (int i=0;i<13;i++) subnam[i]=toupper(subnam[i]);
//fprintf(stderr,"iparam2stage, name__ gelesen #%s#\n",name__);
//fprintf(stderr,"iparam2stage, subnam gelesen #%s#\n",subnam);
strncpy(subnam,name__,13);
subnam[13]='\0';
{
int i;
for (i=0;i<13;i++) subnam[i]=toupper(subnam[i]);
}
#if 0

21
lapack-netlib/TESTING/MATGEN/Makefile
View File

@ -40,27 +40,40 @@ ifneq ($(C_LAPACK), 1)
$(FC) $(FFLAGS) -c -o $@ $<
endif
ifneq "$(or $(BUILD_SINGLE),$(BUILD_COMPLEX))" ""
SCATGEN = slatm1.o slatm7.o slaran.o slarnd.o
endif
ifeq ($(BUILD_SINGLE),1)
SMATGEN = slatms.o slatme.o slatmr.o slatmt.o \
slagge.o slagsy.o slakf2.o slarge.o slaror.o slarot.o slatm2.o \
slatm3.o slatm5.o slatm6.o slahilb.o
endif
ifeq ($(BUILD_COMPLEX),1)
CMATGEN = clatms.o clatme.o clatmr.o clatmt.o \
clagge.o claghe.o clagsy.o clakf2.o clarge.o claror.o clarot.o \
clatm1.o clarnd.o clatm2.o clatm3.o clatm5.o clatm6.o clahilb.o
endif
ifneq "$(or $(BUILD_DOUBLE),$(BUILD_COMPLEX16))" ""
DZATGEN = dlatm1.o dlatm7.o dlaran.o dlarnd.o
endif
ifeq ($(BUILD_DOUBLE),1)
DMATGEN = dlatms.o dlatme.o dlatmr.o dlatmt.o \
dlagge.o dlagsy.o dlakf2.o dlarge.o dlaror.o dlarot.o dlatm2.o \
dlatm3.o dlatm5.o dlatm6.o dlahilb.o
endif
ifeq ($(BUILD_COMPLEX16),1)
ZMATGEN = zlatms.o zlatme.o zlatmr.o zlatmt.o \
zlagge.o zlaghe.o zlagsy.o zlakf2.o zlarge.o zlaror.o zlarot.o \
zlatm1.o zlarnd.o zlatm2.o zlatm3.o zlatm5.o zlatm6.o zlahilb.o
endif
.PHONY: all
.NOTPARALLEL:
all: $(TMGLIB)
ALLOBJ = $(SMATGEN) $(CMATGEN) $(SCATGEN) $(DMATGEN) $(ZMATGEN) \
@ -107,9 +120,17 @@ cleanlib:
rm -f $(TMGLIB)
ifneq ($(C_LAPACK), 1)
ifeq ($(filter $(BUILD_SINGLE) $(BUILD_COMPLEX),1),)
slaran.o: slaran.f ; $(FC) $(FFLAGS_NOOPT) -c -o $@ $<
endif
ifeq ($(filter $(BUILD_DOUBLE) $(BUILD_COMPLEX16),1),)
dlaran.o: dlaran.f ; $(FC) $(FFLAGS_NOOPT) -c -o $@ $<
endif
else
ifeq ($(filter $(BUILD_SINGLE) $(BUILD_COMPLEX),1),)
slaran.o: slaran.c ; $(CC) $(CFLAGS) -O0 -c -o $@ $<
endif
ifeq ($(filter $(BUILD_DOUBLE) $(BUILD_COMPLEX16),1),)
dlaran.o: dlaran.c ; $(CC) $(CFLAGS) -O0 -c -o $@ $<
endif
endif

2
param.h
View File

@ -2951,7 +2951,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define SYMV_P 16
#endif
#if defined(P5600) || defined(MIPS1004K) || defined(MIPS24K) || defined(I6400) || defined(P6600) || defined(I6500)
#if defined(MIPS64_GENERIC) || defined(P5600) || defined(MIPS1004K) || defined(MIPS24K) || defined(I6400) || defined(P6600) || defined(I6500)
#define SNUMOPT 2
#define DNUMOPT 2