Releasev17

Notes for the v1.7

Timeline

• June 22 commit new features to trunk for 1.7.0
• June 29 trunk branches to 1.7.0
• Aug 15th 1.7.0 feature complete
• Oct 1 release of 1.7.0 (tentative)

Branch point gating

• ORTE State Machine (Complete)
• uGNI / Cray XE support
• Non-blocking Collectives
  • Fortran bindings
  • Implementation of good ones
  • bring in libnbc for basic
• grdma mpool
• June 22nd goal for all gating to be in
• June 29th goal for branching

Topics of Focus

• MPI-2.2 Compliance
• MPI-3 Compliance
• State-machine based ORTE
• Thread safety (progress and MPI)

• Turn ORTE progress thread on by default for now, see what breaks over next few weeks. If all is okay, then remove option

• BTL Resiliance
  • Failover (PML issue?)
  • Recovery from resource exhaustion
  • Flow control
• Process failure notification

Testing

• What do we want to test?
  • Need to come up with options list
  • Who can write tests?

Workshop

• Open MPI 1.7 release item
• PML code-forking issue

Stale code check

	Framework	Component	Responsible Organization
opal	backtrace	darwin	SNL
opal	backtrace	execinfo	SNL
opal	backtrace	none	SNL
opal	backtrace	printstack	SNL
opal	compress	bzip	IU
opal	compress	gzip	IU
opal	crs	blcr	UTK
opal	crs	dmtcp	UTK
opal	crs	none	UTK
opal	crs	self	UTK
opal	event	libevent2019	rhc
opal	hwloc	external	Cisco
opal	hwloc	hwloc132	Cisco
opal	if	bsdx_ipv4	rhc
opal	if	bsdx_ipv6	rhc
opal	if	linux_ipv6	rhc
opal	if	posix_ipv4	rhc
opal	if	solaris_ipv6	Oracle
opal	if	windows	HLRS
opal	installdirs	config	SNL
opal	installdirs	env	SNL
opal	installdirs	windows	SNL
opal	memchecker	valgrind	HLRS
opal	memory	linux	Mellanox
opal	memory	malloc_solaris	Oracle
opal	pstat	linux	rhc
opal	pstat	test	rhc
opal	shmem	mmap	Oracle/LANL
opal	shmem	posix	LANL
opal	shmem	sysv	LANL
opal	shmem	windows	HLRS
opal	timer	aix	IBM
opal	timer	altix	SNL
opal	timer	catamount	SNL
opal	timer	darwin	SNL
opal	timer	linux	SNL
opal	timer	solaris	Oracle
opal	timer	windows	HLRS
orte	common	pmi	LANL
orte	errmgr	default_app	rhc
orte	errmgr	default_hnp	rhc
orte	errmgr	default_orted	rhc
orte	ess	alps	LANL
orte	ess	cnos	Defunct (deleted from trunk)
orte	ess	env	rhc
orte	ess	generic	rhc
orte	ess	hnp	rhc
orte	ess	lsf	rhc
orte	ess	pmi	rhc
orte	ess	portals4_shmem	Defunct (deleted from trunk)
orte	ess	singleton	rhc
orte	ess	slurm	rhc
orte	ess	tm	rhc
orte	ess	tool	rhc
orte	filem	rsh	IU
orte	grpcomm	bad	rhc
orte	grpcomm	cnos	removed from trunk
orte	grpcomm	hier	rhc
orte	grpcomm	pmi	rhc
orte	grpcomm	portals4_shmem	removed from trunk
orte	iof	hnp	rhc
orte	iof	mr_hnp	rhc
orte	iof	mr_orted	rhc
orte	iof	orted	rhc
orte	iof	tool	rhc
orte	notifier	command	removed from trunk
orte	notifier	file	removed from trunk
orte	notifier	ftb	removed from trunk
orte	notifier	hnp	removed from trunk
orte	notifier	smtp	removed from trunk
orte	notifier	syslog	removed from trunk
orte	odls	default	rhc
orte	odls	process	HLRS
orte	oob	tcp	rhc
orte	oob	ud	LANL
orte	plm	alps	rhc
orte	plm	ccp	HLRS
orte	plm	lsf	IBM
orte	plm	poe	removed from trunk
orte	plm	process	HLRS
orte	plm	rsh	rhc
orte	plm	slurm	rhc
orte	plm	tm	rhc
orte	ras	alps	rhc
orte	ras	ccp	HLRS
orte	ras	gridengine	Oracle
orte	ras	loadleveler	IBM
orte	ras	lsf	IBM
orte	ras	simulator	rhc
orte	ras	slurm	rhc
orte	ras	tm	rhc
orte	rmaps	ppr	rhc
orte	rmaps	rank_file	rhc
orte	rmaps	resilient	rhc
orte	rmaps	round_robin	rhc
orte	rmaps	seq	rhc
orte	rml	ftrm	IU
orte	rml	oob	rhc
orte	routed	binomial	rhc
orte	routed	debruijn	LANL
orte	routed	radix	rhc
orte	sensor	file	rhc
orte	sensor	ft_tester	rhc
orte	sensor	heartbeat	rhc
orte	sensor	resusage	rhc
orte	snapc	full	IU
orte	sstore	central	IU
orte	sstore	stage	IU
orte	state	app	rhc
orte	state	hnp	rhc
orte	state	orted	rhc
ompi	allocator	basic	NVIDIA
ompi	allocator	bucket	NVIDIA
ompi	bml	r2	SNL
ompi	btl	elan	removed from trunk
ompi	btl	gm	removed from trunk
ompi	btl	mx	removed from 1.7
ompi	btl	ofud	removed from 1.7
ompi	btl	openib	Mellanox
ompi	btl	portals	removed from 1.7
ompi	btl	sctp	removed from 1.7
ompi	btl	self	UTK
ompi	btl	sm	LANL
ompi	btl	smcuda	NVIDIA
ompi	btl	tcp	UTK
ompi	btl	template	removed from 1.7
ompi	btl	udapl	Oracle (removed from 1.7)
ompi	btl	ugni	LANL
ompi	btl	vader	LANL
ompi	btl	wv	HLRS
ompi	coll	basic	UH
ompi	coll	demo	removed from 1.7
ompi	coll	fca	Mellanox
ompi	coll	hierarch	UH
ompi	coll	inter	UH
ompi	coll	self	Cisco
ompi	coll	sm	LANL
ompi	coll	sync	removed from trunk
ompi	coll	tuned	UTK
ompi	coll	libnbc	SNL
ompi	common	cuda	NVIDIA
ompi	common	mx	SNL
ompi	common	portals	removed from 1.7
ompi	common	sm	LANL
ompi	common	ugni	LANL
ompi	crcp	bkmrk	IU
ompi	dpm	orte	rhc
ompi	fbtl	posix	UH
ompi	fbtl	pvfs2	UH
ompi	fcache	ux	UH (probably remove before 1.7)
ompi	fcoll	dynamic	UH
ompi	fcoll	individual	UH
ompi	fcoll	static	UH
ompi	fcoll	two_phase	UH
ompi	fcoll	ylib	UH (no 1.7)
ompi	fs	lustre	UH
ompi	fs	pvfs2	UH
ompi	fs	ufs	UH
ompi	io	ompio	UH
ompi	io	romio	Cisco
ompi	mpool	fake	removed from 1.7
ompi	mpool	gpusm	NVIDIA
ompi	mpool	grdma	LANL
ompi	mpool	rdma	Mellanox (likely removed for grdma)
ompi	mpool	rgpusm	NVIDIA
ompi	mpool	sm	LANL
ompi	mtl	mx	SNL
ompi	mtl	mxm	Mellanox
ompi	mtl	portals	removed from 1.7
ompi	mtl	portals4	SNL
ompi	mtl	psm	SNL/Intel
ompi	op	example	Cisco (removed from 1.7)
ompi	op	x86	Cisco (removed from 1.7)
ompi	osc	pt2pt	remove from 1.7
ompi	osc	rdma	SNL
ompi	pml	bfo	Oracle
ompi	pml	cm	SNL
ompi	pml	crcpw	IU
ompi	pml	csum	removed from 1.7 (remove from trunk?)
ompi	pml	dr	removed from 1.7 (remove from trunk?)
ompi	pml	example	removed from 1.7
ompi	pml	ob1	LANL
ompi	pml	v	UTK
ompi	pubsub	orte	rhc
ompi	pubsub	pmi	rhc
ompi	rcache	rb	NVIDIA
ompi	rcache	vma	LANL
ompi	sharedfp	dummy	UH
ompi	topo	unity	Cisco
ompi	vprotocol	example	UTK
ompi	vprotocol	pessimist	UTK