The Multicore Programming Challenge课件.ppt

The Multicore Programming Challenge;Agenda;The Power Wall;The Memory Wall;So Now We have Multicore ;Take-Up in Enterprise Server Market;What Is Hard About MC Programming? ;Manycore is Coming, Ready or Not;Agenda;Application Developer Needs;Does It Matter? Of Course!;Programming Model: Some Requirements;Some Programming Approaches;First Thoughts: Ends of Spectrum;PGAS Languages;HPCS Languages; ; ;Shared Memory Models 1: PThreads;Agenda;Shared Memory Models 2: OpenMP* ;Shared Memory Models 2: OpenMP ;Cart3D OpenMP Scaling;The OpenMP ARB;Oct 1997 – 1.0 Fortran Oct 1998 – 1.0 C/C++ Nov 1999 – 1.1 Fortran (interpretations added) Nov 2000 – 2.0 Fortran Mar 2002 – 2.0 C/C++ May 2005 – 2.5 Fortran/C/C++ (mostly a merge) ?? 2008 – 3.0 Fortran/C/C++ (extensions) Original goals: Ease of use, incremental approach to parallelization “Adequate” speedup on small SMPs, ability to write scalable code on large SMPs with corresponding effort As far as possible, parallel code “compatible” with serial code ; OpenMP 3.0;Pointer -chasing Loops in OpenMP?;OpenMP 3.0 Introduces Tasks;More and More Threads;What is Missing? 1: Locality;Locality Support: An Easy Way;OpenMP Locality: Thread Subteams;BT-MZ Performance with Subteams;What is Missing? 2: Synchronization ;Agenda;Compiling for Multicore;Implementation Not Easy Everywhere;OpenUH Compiler Infrastructure;OpenMP Compiler Optimizations;;OpenMP Macro-Pipelined Execution;!$OMP PARALLEL !$OMP DO do i=1,imt RHOKX(imt,i) = 0.0 enddo !$OMP ENDDO !$OMP DO do i=1, imt do j=1, jmt if (k .le. KMU(j,i)) then RHOKX(j,i) = DXUR(j,i)*p5*RHOKX(j,i) endif enddo enddo !$OMP ENDDO !$OMP DO do i=1, imt do j=1, jmt if (k KMU(j,i)) then RHOKX(j,i) = 0.0 endif enddo enddo !$OMP ENDDO if (k == 1) then !$OMP DO do i=1, imt do j=1, jmt