portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-10 07:55:35 +00:00
Code Releases Activity

Merge branch 'master' of https://github.com/paboyle/Grid

Browse Source
This commit is contained in:
paboyle 2016-03-02 07:04:03 -08:00
commit 325e745daa
12 changed files with 337 additions and 2513 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
.gitignore vendored
View File

@ -95,6 +95,7 @@ Thumbs.db
# build directory #
###################
build/*
build_debug/*
# IDE related files #
#####################

2714
configure vendored
View File

File diff suppressed because it is too large Load Diff

2
configure.ac
View File

@ -71,7 +71,7 @@ AC_CHECK_FUNCS([gettimeofday])
AC_ARG_ENABLE([simd],[AC_HELP_STRING([--enable-simd=SSE4|AVX|AVXFMA4|AVX2|AVX512|IMCI],\
[Select instructions to be SSE4.0, AVX 1.0, AVX 2.0+FMA, AVX 512, IMCI])],\
[ac_SIMD=${enable_simd}],[ac_SIMD=AVX2])
[ac_SIMD=${enable_simd}],[ac_SIMD=DEBUG])
supported=no

1
lib/AlignedAllocator.h
View File

@ -36,7 +36,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
#include <malloc.h>
#endif
#include <immintrin.h>
#ifdef HAVE_MM_MALLOC_H
#include <mm_malloc.h>
#endif

33
lib/Config.h.in
View File

@ -30,15 +30,6 @@
/* GRID_DEFAULT_PRECISION is SINGLE */
#undef GRID_DEFAULT_PRECISION_SINGLE
/* Support Altivec instructions */
#undef HAVE_ALTIVEC
/* Support AVX (Advanced Vector Extensions) instructions */
#undef HAVE_AVX
/* Support AVX2 (Advanced Vector Extensions 2) instructions */
#undef HAVE_AVX2
/* Define to 1 if you have the declaration of `be64toh', and to 0 if you
don't. */
#undef HAVE_DECL_BE64TOH
@ -53,9 +44,6 @@
/* Define to 1 if you have the <execinfo.h> header file. */
#undef HAVE_EXECINFO_H
/* Support FMA3 (Fused Multiply-Add) instructions */
#undef HAVE_FMA
/* Define to 1 if you have the `gettimeofday' function. */
#undef HAVE_GETTIMEOFDAY
@ -74,30 +62,9 @@
/* Define to 1 if you have the <memory.h> header file. */
#undef HAVE_MEMORY_H
/* Support mmx instructions */
#undef HAVE_MMX
/* Define to 1 if you have the <mm_malloc.h> header file. */
#undef HAVE_MM_MALLOC_H
/* Support SSE (Streaming SIMD Extensions) instructions */
#undef HAVE_SSE
/* Support SSE2 (Streaming SIMD Extensions 2) instructions */
#undef HAVE_SSE2
/* Support SSE3 (Streaming SIMD Extensions 3) instructions */
#undef HAVE_SSE3
/* Support SSSE4.1 (Streaming SIMD Extensions 4.1) instructions */
#undef HAVE_SSE4_1
/* Support SSSE4.2 (Streaming SIMD Extensions 4.2) instructions */
#undef HAVE_SSE4_2
/* Support SSSE3 (Supplemental Streaming SIMD Extensions 3) instructions */
#undef HAVE_SSSE3
/* Define to 1 if you have the <stdint.h> header file. */
#undef HAVE_STDINT_H

9
lib/Init.cc
View File

@ -45,9 +45,6 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#include <algorithm>
#include <iterator>
#define __X86_64
namespace Grid {
//////////////////////////////////////////////////////
@ -291,12 +288,11 @@ void Grid_sa_signal_handler(int sig,siginfo_t *si,void * ptr)
// Linux/Posix
#ifdef __linux__
// And x86 64bit
ucontext_t * uc= (ucontext_t *)ptr;
#ifdef __x86_64__
ucontext_t * uc= (ucontext_t *)ptr;
struct sigcontext *sc = (struct sigcontext *)&uc->uc_mcontext;
printf(" instruction %llx\n",(unsigned long long)sc->rip);
#define REG(A) printf(" %s %lx\n",#A,sc-> A);
REG(rdi);
REG(rsi);
REG(rbp);
@ -316,6 +312,7 @@ void Grid_sa_signal_handler(int sig,siginfo_t *si,void * ptr)
REG(r13);
REG(r14);
REG(r15);
#endif
#endif
BACKTRACE();
exit(0);

23
lib/PerfCount.h
View File

@ -43,8 +43,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#else
#include <sys/syscall.h>
#endif
namespace Grid {
namespace Grid {
#ifdef __linux__
static long perf_event_open(struct perf_event_attr *hw_event, pid_t pid,
@ -58,6 +58,23 @@ static long perf_event_open(struct perf_event_attr *hw_event, pid_t pid,
}
#endif
#ifdef __bgq__
inline uint64_t cyclecount(void){
uint64_t tmp;
asm volatile ("mfspr %0,0x10C" : "=&r" (tmp) );
return tmp;
}
#elif defined __x86_64__
#include <x86intrin.h>
inline uint64_t cyclecount(void){
return __rdtsc();
}
#else
#warning No cycle counter implemented for this architecture
inline uint64_t cyclecount(void){
return 0;
}
#endif
class PerformanceCounter {
private:
@ -149,7 +166,7 @@ public:
ioctl(fd, PERF_EVENT_IOC_RESET, 0);
ioctl(fd, PERF_EVENT_IOC_ENABLE, 0);
}
begin =__rdtsc();
begin =cyclecount();
#else
begin = 0;
#endif
@ -162,7 +179,7 @@ public:
ioctl(fd, PERF_EVENT_IOC_DISABLE, 0);
::read(fd, &count, sizeof(long long));
}
elapsed = __rdtsc() - begin;
elapsed = cyclecount() - begin;
#else
elapsed = 0;
#endif

6
lib/qcd/action/fermion/WilsonKernelsAsm.cc
View File

@ -118,12 +118,12 @@ void WilsonKernels<Impl >::DiracOptAsmDhopSite(StencilImpl &st,DoubledGaugeField
StencilEntry *SE;
//#define STAMP(i) timers[i] = __rdtsc() ;
#define STAMP(i) //timers[i] = __rdtsc() ;
//#define STAMP(i) timers[i] = cyclecount() ;
#define STAMP(i) //timers[i] = cyclecount() ;
MASK_REGS;
first = __rdtsc();
first = cyclecount();
SE=st.GetEntry(ptype,Xm,ss);

12
lib/serialisation/MacroMagic.h
View File

@ -120,7 +120,7 @@ THE SOFTWARE.
\
\
template <typename T>\
static void write(Writer<T> &WR,const std::string &s, const cname &obj){ \
static inline void write(Writer<T> &WR,const std::string &s, const cname &obj){ \
push(WR,s);\
GRID_MACRO_EVAL(GRID_MACRO_MAP(GRID_MACRO_WRITE_MEMBER,__VA_ARGS__)) \
pop(WR);\
@ -128,14 +128,14 @@ THE SOFTWARE.
\
\
template <typename T>\
static void read(Reader<T> &RD,const std::string &s, cname &obj){ \
static inline void read(Reader<T> &RD,const std::string &s, cname &obj){ \
push(RD,s);\
GRID_MACRO_EVAL(GRID_MACRO_MAP(GRID_MACRO_READ_MEMBER,__VA_ARGS__)) \
pop(RD);\
} \
\
\
friend std::ostream & operator << (std::ostream &os, const cname &obj ) { \
friend inline std::ostream & operator << (std::ostream &os, const cname &obj ) { \
os<<"class "<<#cname<<" {"<<std::endl;\
GRID_MACRO_EVAL(GRID_MACRO_MAP(GRID_MACRO_OS_WRITE_MEMBER,__VA_ARGS__)) \
os<<"}"; \
@ -165,7 +165,7 @@ namespace Grid {
class EnumIO<name> {\
public:\
template <typename T>\
static void write(Writer<T> &WR,const std::string &s, const name &obj){ \
static inline void write(Writer<T> &WR,const std::string &s, const name &obj){ \
switch (obj) {\
GRID_MACRO_EVAL(GRID_MACRO_MAP(GRID_MACRO_ENUMCASE,__VA_ARGS__))\
default: Grid::write(WR,s,#undefname); break;\
@ -173,7 +173,7 @@ namespace Grid {
}\
\
template <typename T>\
static void read(Reader<T> &RD,const std::string &s, name &obj){ \
static inline void read(Reader<T> &RD,const std::string &s, name &obj){ \
std::string buf;\
Grid::read(RD, s, buf);\
if (buf == #undefname) {obj = name::undefname;}\
@ -182,7 +182,7 @@ namespace Grid {
}\
};\
\
std::ostream & operator << (std::ostream &os, const name &obj ) { \
inline std::ostream & operator << (std::ostream &os, const name &obj ) { \
switch (obj) {\
GRID_MACRO_EVAL(GRID_MACRO_MAP(GRID_MACRO_ENUMCASEIO,__VA_ARGS__))\
default: os << #undefname; break;\

14
lib/serialisation/XmlIO.cc
View File

@ -80,6 +80,20 @@ void XmlReader::pop(void)
node_ = node_.parent();
}
bool XmlReader::nextElement(const std::string &s)
{
if (node_.next_sibling(s.c_str()))
{
node_ = node_.next_sibling(s.c_str());
return true;
}
else
{
return false;
}
}
template <>
void XmlReader::readDefault(const string &s, string &output)
{

1
lib/serialisation/XmlIO.h
View File

@ -68,6 +68,7 @@ namespace Grid
virtual ~XmlReader(void) = default;
void push(const std::string &s);
void pop(void);
bool nextElement(const std::string &s);
template <typename U>
void readDefault(const std::string &s, U &output);
template <typename U>

32
lib/simd/Grid_empty.h
View File

@ -35,6 +35,7 @@ Author: neo <cossu@post.kek.jp>
// Time-stamp: <2015-06-09 14:28:02 neo>
//----------------------------------------------------------------------
namespace Grid {
namespace Optimization {
template<class vtype>
@ -243,6 +244,36 @@ namespace Optimization {
//////////////////////////////////////////////
// Some Template specialization
struct Permute{
static inline float Permute0(float in){
return in;
};
static inline float Permute1(float in){
return in;
};
static inline float Permute2(float in){
return in;
};
static inline float Permute3(float in){
return in;
};
static inline double Permute0(double in){
return in;
};
static inline double Permute1(double in){
return in;
};
static inline double Permute2(double in){
return in;
};
static inline double Permute3(double in){
return in;
};
};
template < typename vtype >
void permute(vtype &a, vtype b, int perm) {
};
@ -282,7 +313,6 @@ namespace Optimization {
//////////////////////////////////////////////////////////////////////////////////////
// Here assign types
namespace Grid {
typedef float SIMD_Ftype; // Single precision type
typedef double SIMD_Dtype; // Double precision type