High Performance Computing Linpack Benchmark

 - needs an MPI implementation (mpich and lam are supported by this port)
 - needs an Basic Linear Algebra Subprograms implementation like atlas (default)
   or blas (both are supported by this port)

tested on alpha (5), amd64 (5) and i386 (4 and 5)

5 files changed, 153 insertions, 0 deletions

diff --git a/benchmarks/Makefile b/benchmarks/Makefile
index 5ff72aeb98f9..b136ee2cf8db 100644
--- a/benchmarks/Makefile
+++ b/benchmarks/Makefile
@@ -8,6 +8,7 @@
     SUBDIR += dbench
     SUBDIR += dbs
     SUBDIR += forkbomb
+    SUBDIR += hpl
     SUBDIR += iozone
     SUBDIR += iozone21
     SUBDIR += lmbench
diff --git a/benchmarks/hpl/Makefile b/benchmarks/hpl/Makefile
new file mode 100644
index 000000000000..3e0837cfc20d
--- /dev/null
+++ b/benchmarks/hpl/Makefile
@@ -0,0 +1,119 @@
+# New ports collection makefile for:	physfs
+# Date Created:		22 February 2003
+# Whom:			Oliver Lehmann <oliver@FreeBSD.org>
+#
+# $FreeBSD$
+#
+
+PORTNAME=	hpl
+PORTVERSION=	1.0a
+CATEGORIES=	benchmarks
+MASTER_SITES=	http://www.netlib.org/benchmark/hpl/
+DISTNAME=	${PORTNAME}
+EXTRACT_SUFX=	.tgz
+
+MAINTAINER=	oliver@FreeBSD.org
+COMMENT=	High Performance Computing Linpack Benchmark
+
+.if !defined(WITH_LAM)
+BUILD_DEPENDS=	${LOCALBASE}/mpich/bin/mpicc:${PORTSDIR}/net/mpich
+.else
+BUILD_DEPENDS=	${LOCALBASE}/bin/lamd:${PORTSDIR}/net/lam
+CC=		${LOCALBASE}/bin/mpicc
+FC=		${LOCALBASE}/bin/mpicc
+.endif
+
+RUN_DEPENDS=	${BUILD_DEPENDS}
+
+.if ${MACHINE_ARCH} == "amd64"
+# math/atlas is currently broken on amd64
+WITH_BLAS=	yes
+.endif
+
+.if defined(WITH_BLAS)
+LIB_DEPENDS+=	blas.1:${PORTSDIR}/math/blas
+.else
+LIB_DEPENDS+=	cblas.1:${PORTSDIR}/math/atlas
+.endif
+
+INSTALLS_SHLIB=	yes
+USE_REINPLACE=	yes
+HPL_ARCH?=	FreeBSD_PIV_CBLAS
+ALL_TARGET=	arch=${HPL_ARCH}
+MAN3=		HPL_abort.3 HPL_all_reduce.3 HPL_barrier.3 HPL_bcast.3 \
+		HPL_binit.3 HPL_broadcast.3 HPL_bwait.3 HPL_copyL.3 \
+		HPL_daxpy.3 HPL_dcopy.3 HPL_dgemm.3 HPL_dgemv.3 HPL_dger.3 \
+		HPL_dlacpy.3 HPL_dlamch.3 HPL_dlange.3 HPL_dlaprnt.3 \
+		HPL_dlaswp00N.3 HPL_dlaswp01N.3 HPL_dlaswp01T.3 HPL_dlaswp02N.3 \
+		HPL_dlaswp03N.3 HPL_dlaswp03T.3 HPL_dlaswp04N.3 HPL_dlaswp04T.3 \
+		HPL_dlaswp05N.3 HPL_dlaswp05T.3 HPL_dlaswp06N.3 HPL_dlaswp06T.3 \
+		HPL_dlaswp10N.3 HPL_dlatcpy.3 HPL_dlocmax.3 HPL_dlocswpN.3 \
+		HPL_dlocswpT.3 HPL_dmatgen.3 HPL_dscal.3 HPL_dswap.3 HPL_dtrsm.3 \
+		HPL_dtrsv.3 HPL_equil.3 HPL_fprintf.3 HPL_grid_exit.3 \
+		HPL_grid_info.3 HPL_grid_init.3 HPL_idamax.3 HPL_indxg2l.3 \
+		HPL_indxg2lp.3 HPL_indxg2p.3 HPL_indxl2g.3 HPL_infog2l.3 \
+		HPL_jumpit.3 HPL_ladd.3 HPL_lmul.3 HPL_logsort.3 HPL_max.3 \
+		HPL_min.3 HPL_numroc.3 HPL_numrocI.3 HPL_pabort.3 HPL_packL.3 \
+		HPL_pddriver.3 HPL_pdfact.3 HPL_pdgesv.3 HPL_pdgesv0.3 \
+		HPL_pdgesvK1.3 HPL_pdgesvK2.3 HPL_pdinfo.3 HPL_pdlamch.3 \
+		HPL_pdlange.3 HPL_pdlaprnt.3 HPL_pdlaswp00N.3 HPL_pdlaswp00T.3 \
+		HPL_pdlaswp01N.3 HPL_pdlaswp01T.3 HPL_pdmatgen.3 HPL_pdmxswp.3 \
+		HPL_pdpancrN.3 HPL_pdpancrT.3 HPL_pdpanel_disp.3 \
+		HPL_pdpanel_free.3 HPL_pdpanel_init.3 HPL_pdpanel_new.3 \
+		HPL_pdpanllN.3 HPL_pdpanllT.3 HPL_pdpanrlN.3 HPL_pdpanrlT.3 \
+		HPL_pdrpancrN.3 HPL_pdrpancrT.3 HPL_pdrpanllN.3 HPL_pdrpanllT.3 \
+		HPL_pdrpanrlN.3 HPL_pdrpanrlT.3 HPL_pdtest.3 HPL_pdtrsv.3 \
+		HPL_pdupdateNN.3 HPL_pdupdateNT.3 HPL_pdupdateTN.3 \
+		HPL_pdupdateTT.3 HPL_perm.3 HPL_pipid.3 HPL_plindx0.3 \
+		HPL_plindx1.3 HPL_plindx10.3 HPL_pnum.3 HPL_ptimer.3 \
+		HPL_ptimer_cputime.3 HPL_ptimer_walltime.3 HPL_pwarn.3 \
+		HPL_rand.3 HPL_recv.3 HPL_reduce.3 HPL_rollN.3 HPL_rollT.3 \
+		HPL_sdrv.3 HPL_send.3 HPL_setran.3 HPL_spreadN.3 HPL_spreadT.3 \
+		HPL_sum.3 HPL_timer.3 HPL_timer_cputime.3 HPL_timer_walltime.3 \
+		HPL_warn.3 HPL_xjumpm.3
+
+post-patch:
+	@${REINPLACE_CMD} -e 's|^TOPdir[ ]*=.*$$|TOPdir = ${WRKSRC}| ; \
+		s|^MPdir[ ]*=.*$$|MPdir = ${LOCALBASE}/mpich| ; \
+		s|^LAdir[ ]*=.*$$|LAdir = ${LOCALBASE}/lib| ; \
+		s|^CC[ ]*=.*$$|CC = ${CC}| ; \
+		s|^CCFLAGS[ ]*=.*$$|CCFLAGS = $$\(HPL_DEFS\) ${CFLAGS} -fomit-frame-pointer -funroll-loops| ; \
+		s|^LINKER[ ]*=.*$$|LINKER = ${FC}| ; \
+		s|^ARCHIVER[ ]*=.*$$|ARCHIVER = ${AR}|'\
+		${WRKSRC}/setup/Make.${HPL_ARCH}
+
+.if defined(WITH_LAM)
+	@${REINPLACE_CMD} -e 's|^MPdir[ ]*=.*$$|#&| ; s|^MPinc[ ]*=.*$$|#&| ; \
+		s|^MPlib[ ]*=.*$$|MPlib = /usr/lib/libg2c.a|' \
+		${WRKSRC}/setup/Make.${HPL_ARCH}
+.else
+	@${REINPLACE_CMD} -e 's|^MPdir[ ]*=.*$$|MPdir = ${LOCALBASE}/mpich|' \
+		${WRKSRC}/setup/Make.${HPL_ARCH}
+.endif
+
+.if defined(WITH_BLAS)
+	@${REINPLACE_CMD} -e 's|^HPL_OPTS[ ]*=.*$$|HPL_OPTS =| ; \
+		s|^LAlib[ ]*=.*$$|LAlib = $$\(LAdir\)/libblas.a|' \
+		${WRKSRC}/setup/Make.${HPL_ARCH}
+.endif
+
+	@${REINPLACE_CMD} -e 's|HPL\.dat|${PREFIX}/etc/${PORTNAME}/&|' \
+		${WRKSRC}/testing/ptest/HPL_pdinfo.c
+
+	@${CP} ${WRKSRC}/setup/Make.${HPL_ARCH} ${WRKSRC}/
+
+do-install:
+	${MKDIR} ${PREFIX}/etc/${PORTNAME}
+	${INSTALL_PROGRAM} ${WRKSRC}/bin/${HPL_ARCH}/xhpl ${PREFIX}/bin/
+	${INSTALL_DATA} ${WRKSRC}/bin/${HPL_ARCH}/HPL.dat ${PREFIX}/etc/${PORTNAME}/
+.for man in ${MAN3}
+	${INSTALL_MAN} ${WRKSRC}/man/man3/${man} ${MANPREFIX}/man/man3/
+.endfor
+.if !defined(NOPORTDOCS)
+	${MKDIR} ${DOCSDIR}
+.for doc in BUGS COPYRIGHT HISTORY README TODO TUNING
+	${INSTALL_DATA} ${WRKSRC}/${doc} ${DOCSDIR}
+.endfor
+.endif
+
+.include <bsd.port.mk>
diff --git a/benchmarks/hpl/distinfo b/benchmarks/hpl/distinfo
new file mode 100644
index 000000000000..83cb8a462d83
--- /dev/null
+++ b/benchmarks/hpl/distinfo
@@ -0,0 +1,2 @@
+MD5 (hpl.tgz) = 32b432af96e674dcb657ef1d63e6d591
+SIZE (hpl.tgz) = 523139
diff --git a/benchmarks/hpl/pkg-descr b/benchmarks/hpl/pkg-descr
new file mode 100644
index 000000000000..d1c43a126ff0
--- /dev/null
+++ b/benchmarks/hpl/pkg-descr
@@ -0,0 +1,21 @@
+HPL is a software package that solves a (random) dense linear system in double
+precision (64 bits) arithmetic on distributed-memory computers. It can thus be
+regarded as aportable as well as freely available implementation of the High
+Performance Computing Linpack Benchmark.
+
+The algorithm used by HPL can be summarized by the following keywords:
+Two-dimensional block-cyclic data distribution - Right-looking variant of the
+LU factorization with row partial pivoting featuring multiple look-ahead
+depths - Recursive panel factorization with pivot search and column broadcast
+combined - Various virtual panel broadcast topologies - bandwidth reducing
+swap-broadcast algorithm - backward substitution with look-ahead of depth 1.
+
+The HPL package provides a testing and timing program to quantify the accuracy
+of the obtained solution as well as the time it took to compute it. The best
+performance achievable by this software on your system depends on a large
+variety of factors. Nonetheless, with some restrictive assumptions on the
+interconnection network, the algorithm described here and its attached
+implementation are scalable in the sense that their parallel efficiency is
+maintained constant with respect to the per processor memory usage.
+
+WWW: http://www.netlib.org/benchmark/hpl/
diff --git a/benchmarks/hpl/pkg-plist b/benchmarks/hpl/pkg-plist
new file mode 100644
index 000000000000..4d6e9d52bdb7
--- /dev/null
+++ b/benchmarks/hpl/pkg-plist
@@ -0,0 +1,10 @@
+bin/xhpl
+etc/hpl/HPL.dat
+%%PORTDOCS%%%%DOCSDIR%%/BUGS
+%%PORTDOCS%%%%DOCSDIR%%/COPYRIGHT
+%%PORTDOCS%%%%DOCSDIR%%/HISTORY
+%%PORTDOCS%%%%DOCSDIR%%/README
+%%PORTDOCS%%%%DOCSDIR%%/TODO
+%%PORTDOCS%%%%DOCSDIR%%/TUNING
+%%PORTDOCS%%@dirrm %%DOCSDIR%%
+@dirrm etc/hpl