+FDO:
icl, ifort : PASS1=-Qprof_gen PASS2=-Qprof_use
pgf90 : PASS1=-Mpfi PASS2=-Mpfo
ifort is the Intel Fortran compiler, icl is the Intel C++ compiler and
pgf90 is the PGI Fortran 90 compiler.
pgcc is the PGI C compiler.
ONESTEP is set to 1 for every compile with the PGI compilers.
Portability:
178.galgel: -Mfixed
Baseline: C : pgcc -fastsse -Mipa=fast,inline
Baseline: Fortran: pgf90 -fastsse -Mipa=fast,inline +FDO
Peak tuning:
168.wupwise: pgf90 -fastsse -Mipa=fast,inline -Mnovect
171.swim: ifort -Qipo -O3 -QaxN -QxW -Qunroll0 +FDO
172.mgrid: pgf90 -fastsse -Mipa=fast,inline
173.applu: ifort -Qipo -O3 -QaxN -QxW -auto +FDO
177.mesa: icl -Qipo -arch:SSE2 -Qunroll1 -Qansi_alias +FDO
-Qoption,f,-ip_ninl_max_stats=1500,-ip_ninl_max_total_stats=4500
178.galgel: pgf90 -fastsse -Mipa=fast,safe RM_SOURCES=lapak.f90
-Munix -lacml
179.art: pgcc basepeak=yes
183.equake: icl -fast -arch:SSE2 -QaxW +FDO
187.facerec: pgf90 -fastsse -Mipa=fast,inline +FDO
188.ammp: icl -Oa -arch:SSE2 -Zp4 -Qansi_alias
189.lucas: ifort -Qipo -QxW -Qunroll1
191.fma3d: pgf90 -Mipa=fast,inline -fastsse -Mnovect +FDO
200.sixtrack: pgf90 -fastsse -Mipa=fast,inline
301.apsi: pgf90 -fastsse -Mipa=fast,inline
|