???:Intel? ?????(MKL)

1
???Intel? ?????(MKL)
??? http//www.njyangqs.com/
????????????
2
??

• ??
• ????
• ???
• ???

3
???

• ??? Intel MKL
• Intel ?????????????/?????
• ?Intel processors??
• ????????SMP?????????

4
??

• Intel???????
• ??,??,????!
• Intel??????????????
• ??
• ???? (BLAS, LAPACK)
• ????/?????(BLAS, LAPACK)
• ?????????(dgemm)
• ?????, ????, ??, ????(FFTs)
• ????,?????????? (VML)?????????(VSL)
• ?Intel????????????

5
??

• Intel ??????? ????
• ?????Intel ?????(Intel MKL) ??
• ????????Intel MKL
• ????n?????????.

X Y Z W
X Y Z W
4x4 ????

????
6
??

• Intel ???????
• BLAS (?????????)
• Level 1 BLAS ??-????
• 15?????
• 48???
• Level 2 BLAS ??-????
• 26?????
• 66???
• Level 3 BLAS ??-????
• 9?????
• 30???
• Extended BLAS ??????level 1 BLAS
• 8?????
• 24???

7
??

• Intel ???????
• LAPACK (?????)
• ???????????. ????????
• ?????1000???????????
• DFTs (???????)
• ????, ????
• ???
• VML (?????)
• ???????
• ????libm??,????
• VSL (??????)
• ?????????

8
??

• Intel ???????
• BLAS?LAPACK ??Fortran.
• ????????
• VSL?VML?Fortran?C??
• DFTs?Fortran 95?C??
• cblas?????C/C?????BLAS

9
??

• Intel ?????(Intel MKL) ??
• ??32??64?Intel???
• ???????????
• ?????

Windows Linux
??? Intel, CVF, Microsoft Intel, Gnu
? .dll, .lib .a, .so
10
??

• ??
• ????
• ???
• ???

11
????

• ???????
• ?????????????
• ??????? ?????????????
• CPU ?????, ????
• Cache ???????cache???? ??cache??
• TLBs ?????????????
• ???? ??????????.
• ??? ?????????????
• ?? ?????????(????).

12
????

• ???
• ???Intel ?????(Intel MKL) ?????,??
• ??????????
• ???level 1?level 2 BLAS ??????( O(n) )
• ??????????
• Level 3 BLAS ( O(n3) )
• LAPACK ( O(n3) )
• FFTs ( O(n log(n) )
• VML, VSL ? ????????
• ??????????OpenMP.
• ??Intel MKL????????????????

13
??

• ??
• ????
• ???
• ???

14
???

• ??Intel ?????(Intel MKL)
• ??1 ifort, BLAS, IA-32???
• ifort myprog.f mkl_c.lib
• ??2 CVF, LAPACK, IA-32???
• f77 myprog.f mkl_s.lib
• ??3 ???????????DLL???????C??
• link myprog.obj mkl_c_dll.lib
• ????????????????????????

15
???

• ????
• ?????????

Roll Your Own
for( i 0 i lt n i ) for( j 0 j lt m
j ) for( k 0 k lt kk k ) cij
aik bkj
ddot
for( i 0 i lt n i ) for( j 0 j lt m
j ) cij cblas_ddot( n, ai,
incx,b0j, incy)
16
???

• ????
• DGEMV/DGEMM

dgemv
for( i 0 i lt n i ) cblas_dgemv(
CBLAS_RowMajor, CBLAS_NoTrans, m, n,
alpha, a, lda, b0i, ldb, beta,
c0i, ldc )
dgemm
Cblas_dgemm( CblasColMajor, CblasNoTrans,
CblasNoTrans, m, n, kk, alpha, b, ldb, a, lda,
beta, c, ldc )
17
???

• ?? 1 DGEMM
• ?????C??,DDOT,DGEMG?DGEMM?????????
• ???MKL/BLAS????????

18
??

• ??
• ????
• ???
• ???

19
???

• LAPACK???Intel ?????
• ?????LAPACK??
• ??? ????????????
• ????
• ??????(Amdahls law t t?? t??/p)
• ????????????????,??????
• ?????????
• NETLIB LAPACK????????????????????,??????Intel
  MKL?,??????????,??????????????

20
???

• ???????(DFTs)
• 1?,2?,3?(???????)
• ?????
• ?????????????????,????????????????????????,???????
  ??????????
• ??????,?????
• ???????,????????????
• C?F90??

21
???

• ?Intel ?????????????
• ????3???
• ???????.
• Status DftiCreateDescriptor(MDH, )
• ?????(????).
• Status DftiCommitDescriptor(MDH)
• ????.
• Status DftiComputeForward(MDH, X)
• ?????(???)

22
???

• ?????(VML)???/??
• ????? ?????? ?libm,????(??)
• ?? ?Fortran?C????
• ?????
• ???( lt 1 ulp )
• ????, ??( lt 4 ulps )
• ??????v(-a), sin(0),?
• ???? ??????libm

23
???

• VML ?????
• ?????????(???????).
• ??, ??
• ??????????????
• ????????????????????.

24
???

• ??????(VSL)
• ???????(RNGs)
• ????????
• VSL???????????
• ??????? ????
• ?????????BRNG???
• 5????RNGs (BRNGs) ?,??,???
• MCG31, R250, MRG32, MCG59, WH

25
???

• ???RNGs
• Gaussian (two methods) ??
• Exponential ??
• Laplace ??????
• Weibull ????
• Cauchy ????
• Rayleigh ??
• Lognormal ????
• Gumbel Gumbel??

26
???

• ??VSL
• ???????
• ???????
• VSLStreamStatePtr stream
• ?????.
• vslNewStream(stream,VSL_BRNG_MC_G31, seed )
• ????RNGs.
• vsRngUniform( 0, stream, size, out, start, end
  )
• ?????(??).
• vslDeleteStream(stream)

27
???

• ?? ????????p
• ???????,Spr2
• ??????,??,1/4??????????????????,???????1,??????
  ??????????,???1/4????????????1/4???????
• ??????1/4???????????????????? x2y21

1
-1
1
0
-1
28
???
int main() unsigned int iter200000000
int i,j double x, y double
dUnderCurve0.0 double pi0.0 srand( 0)
for (i0iltiteri) x(double)rand()/(d
ouble)RAND_MAX y(double)rand()/(double)R
AND_MAX if (xx yy lt 1.0)
dUnderCurve pi dUnderCurve /
(double) iter 4 return 0

• ?? ????????p

1
-1
1
0
-1
29
???
unsigned int iter200000000 int i,j double
x, y double dUnderCurve0.0 double pi0.0
double rBLOCK_SIZE2
VSLStreamStatePtr stream vslNewStream(
stream, BRNG, (int) clock() )
for(j0jltiter/BLOCK_SIZE j)
vdRngUniform( METHOD, stream, BLOCK_SIZE2, r,
0.0, 1.0 ) for (i0 iltBLOCK_SIZE i)
xri
yriBLOCK_SIZE if (xx yy
lt 1.0) dUnderCurve
vslDeleteStream( stream
) pi dUnderCurve / (double) iter 4

• ?? ????????p

1
-1
1
0
-1