1 This file describes the most significant changes. For more detail, use
2 'git log' on a clone of the charm repository.
4 ================================================================================
5 What's new in Charm++ 6.8.0
6 ================================================================================
8 Over 600 bug fixes, improvements, and cleanups have been applied
9 across the entire system. Major changes are described below:
13 - Calls to entry methods taking a single fixed-size parameter can now
14 automatically be aggregated and routed through the TRAM library by
15 marking them with the [aggregate] attribute.
17 - Calls to parameter-marshalled entry methods with large array
18 arguments can ask for asynchronous zero-copy send behavior with an
19 `rdma' tag in the parameter's declaration.
21 - The runtime system now integrates an OpenMP runtime library so that
22 code using OpenMP parallelism will dispatch work to idle worker
23 threads within the Charm++ process.
25 - Applications can ask the runtime system to perform automatic
26 high-level end-of-run performance analysis by linking with the
27 `-tracemode perfReport' option.
29 - Added a new dynamic remapping/load-balancing strategy,
30 GreedyRefineLB, that offers high result quality and well bounded
33 - Improved and expanded topology-aware spanning tree generation
34 strategies, including support for runs on a torus with holes, such
35 as Blue Waters and other Cray XE/XK systems.
37 - Charm++ programs can now define their own main() function, rather
38 than using a generated implementation from a mainmodule/mainchare
39 combination. This extends the existing Charm++/MPI interoperation
42 - GPU manager now creates one instance per OS process and scales the
43 pre-allocated memory pool size according to the GPU memory size and
44 number of GPU manager instances on a physical node.
46 - Several GPU Manager API changes including:
48 * Replaced references to global variables in the GPU manager API with calls to
51 * The user is no longer required to specify a bufferID in dataInfo struct.
53 * Replaced calls to kernelSelect with direct invocation of functions passed
54 via the work request object (allows CUDA to be built with all programs).
56 - Added support for malleable jobs that can dynamically shrink and
57 expand the set of compute nodes hosting Charm++ processes.
59 - Greatly expanded and improved reduction operations:
61 * Added built-in reductions for all logical and bitwise operations
62 on integer and boolean input.
64 * Reductions over groups and chare arrays that apply commutative,
65 associative operations (e.g. MIN, MAX, SUM, AND, OR, XOR) are now
66 processed in a streaming fashion. This reduces the memory footprint of
67 reductions. User-defined reductions can opt into this mode as well.
69 * Added a new `Tuple' reducer that allows combining multiple reductions
70 of different input data and operations from a common set of source
71 objects to a single target callback.
73 * Added a new `Summary Statistics' reducer that provides count, mean,
74 and standard deviation using a numerically-stable streaming algorithm.
76 - Added a `++quiet' option to suppress charmrun and charm++ non-error
79 - Calls to chare array element entry methods with the [inline] tag now
80 avoid copying their arguments when the called method takes its
81 parameters by const&, offering a substantial reduction in overhead in
84 - Synchronous entry methods that block until completion (marked with
85 the [sync] attribute) can now return any type that defines a PUP
86 method, rather than only message types.
88 - Static (non-generated) header files are now warning-free for
89 gcc -Wall -Wextra -pedantic.
91 - Deprecated setReductionClient and CkSetReductionClient in favor of
92 explicitly passing callbacks to contribute calls.
94 - On C++ standard library implementations with support for
95 std::is_constructible (e.g. GCC libstdc++ >4.5), chare array
96 elements only need to define a constructor taking CkMigrateMessage*
97 if it will actually be migrated.
99 - The PUP serialization framework gained support for some C++11
100 library classes, including unique_ptr and unordered_map, when the
101 underlying types have PUP operators.
105 - More efficient implementations of message matching infrastructure, multiple
106 completion routines, and all varieties of reductions and gathers.
108 - Support for user-defined non-commutative reductions, MPI_IN_PLACE, cancelling
109 receive requests, MPI_THREAD_FUNNELED, PSCW synchronization for RMA, and more.
111 - Fixes to AMPI's extensions for load balancing and to Isomalloc on SMP builds.
113 - More robust derived datatype support, optimizations for truly contiguous types.
115 - ROMIO is now built on AMPI and linked in by ampicc by default.
117 Platforms and Portability
119 - The runtime system code now requires compiler support for C++11
120 R-value references and move constructors. This is not expected to be
121 incompatible with any currently supported compilers.
123 - The next feature release (anticipated to be 6.9.0 or 7.0) will require
124 full C++11 support from the compiler and standard library.
126 - Added support for IBM POWER8 systems with the PAMI communication API,
127 such as development/test platforms for the upcoming Sierra and Summit
128 supercomputers at LLNL and ORNL. Contributed by Sameer Kumar of IBM.
130 - Mac OS (darwin) builds now default to the modern libc++ standard
131 library instead of the older libstdc++.
133 - Blue Gene/Q build targets have been added for the `bgclang' compiler.
135 - Charm++ can now be built on Cray's CCE 8.5.4+.
137 - Charm++ will now build without custom configuration on Arch Linux
139 - Charmrun can automatically detect rank and node count from
140 Slurm/srun environment variables.
142 - Many obsolete architecture, network, and compiler support files have
143 been removed. These include:
145 * Sony/Toshiba/IBM Cell (including PlayStation 3)
147 * Intel IA-64 (Itanium)
148 * Intel x86-32 for Windows, Mac OS X (darwin), and Solaris
150 * Older IBM AIX/POWER configurations
151 * GCC 3 and KAI compilers
154 ================================================================================
155 What's new in Charm++ 6.7.1
156 ================================================================================
158 Changes in this release are primarily bug fixes for 6.7.0. The major exception
159 is AMPI, which has seen changes to its extension APIs and now complies with more
160 of the MPI standard. A brief list of changes follows:
164 - Startup and exit sequences are more robust
166 - Error and warning messages are generally more informative
168 - CkMulticast's set and concat reducers work correctly
172 - AMPI's extensions have been renamed to use the prefix AMPI_ instead of MPI_
173 and to generally follow MPI's naming conventions
175 - AMPI_Migrate(MPI_Info) is now used for dynamic load balancing and all fault
176 tolerance schemes (see the AMPI manual)
178 - AMPI officially supports MPI-2.2, and also implements the non-blocking
179 collectives and neighborhood collectives from MPI-3.1
181 Platforms and Portability
183 - Cray regularpages build target has been fixed
185 - Clang compiler target for BlueGene/Q systems added
187 - Comm. thread tracing for SMP mode added
189 - AMPI's compiler wrappers are easier to use with autoconf and cmake
191 ================================================================================
192 What's new in Charm++ 6.7.0
193 ================================================================================
195 Over 120 bugs fixed, spanning areas across the entire system
199 - New API for efficient formula-based distributed spare array creation
201 - CkLoop is now built by default
203 - CBase_Foo::pup need not be called from Foo::pup in user code anymore - runtime
204 code handles this automatically
206 - Error reporting and recovery in .ci files is greatly improved, providing more
207 precise line numbers and often column information
209 - Many data races occurring under shared-memory builds (smp, multicore) were
210 fixed, facilitating use of tools like ThreadSanitizer and Helgrind
214 - Further MPI standard compliance in AMPI allows users to build and run
215 Hypre-2.10.1 on AMPI with virtualization, migration, etc.
217 - Improved AMPI Fortran2003 PUP interface 'apup', similiar to C++'s STL PUP
219 Platforms and Portability
221 - Compiling Charm++ now requires support for C++11 variadic templates. In GCC,
222 this became available with version 4.3, released in 2008
224 - New machine target for multicore Linux ARM7: multicore-linux-arm7
226 - Preliminary support for POWER8 processors, in preparation for the upcoming
227 Summit and Sierra supercomputers
229 - The charmrun process launcher is now much more robust in the face of slow
230 or rate-limited connections to compute nodes
232 - PXSHM now auto-detects the node size, so the '+nodesize' is no longer needed
234 - Out-of-tree builds are now supported
238 - CommLib has been removed.
240 - CmiBool has been dropped in favor of C++'s bool
243 ================================================================================
244 What's new in Charm++ 6.6.1
245 ================================================================================
247 Changes in this release are primarily bug fixes for 6.6.0. A concise list of
248 affected components follows:
252 - Reductions with syncFT
254 - mpicxx based MPI builds
256 - Increased support for macros in CI file
258 - GNI + RDMA related communication
260 - MPI_STATUSES_IGNORE support for AMPIF
262 - Restart on different node count with chkpt
264 - Immediate msgs on multicore builds
266 ================================================================================
267 What's new in Charm++ 6.6.0
268 ================================================================================
270 - Machine target files for Cray XC systems ('gni-crayxc') have been added
272 - Interoperability with MPI code using native communication interfaces on Blue
273 Gene Q (PAMI) and Cray XE/XK/XC (uGNI) systems, in addition to the universal
274 MPI communication interface
276 - Support for partitioned jobs on all machine types, including TCP/IP and IB
277 Verbs networks using 'netlrts' and 'verbs' machine layers
279 - A substantially improved version of our asynchronous library, CkIO, for
280 parallel output of large files
282 - Narrowing the circumstances in which the runtime system will send
283 overhead-inducing ReductionStarting messages
285 - A new fully distributed load balancing strategy, DistributedLB, that produces
286 high quality results with very low latency
288 - An API for applications to feed custom per-object data to specialized load
289 balancing strategies (e.g. physical simulation coordinates)
291 - SMP builds on LRTS-based machine layers (pamilrts, gni, mpi, netlrts, verbs)
292 support tracing messages through communication threads
294 - Thread affinity mapping with +pemap now supports Intel's Hyperthreading more
297 - After restarting from a checkpoint, thread affinity will use new
298 +pemap/+commap arguments
300 - Queue order randomization options were added to assist in debugging race
301 conditions in application and runtime code
303 - The full runtime code and associated libraries can now compile under the C11
304 and C++11/14 standards.
306 - Numerous bug fixes, performance enhancements, and smaller improvements in the
307 provided runtime facilities
310 * The long-unsupported FEM library has been deprecated in favor of ParFUM
311 * The CmiBool typedefs have been deleted, as C++ bool has long been universal
312 * Future versions of the runtime system and libraries will require some degree
313 of support for C++11 features from compilers
315 ================================================================================
316 What's new in Charm++ 6.5.0
317 ================================================================================
319 - The Charm++ manual has been thoroughly revised to improve its organization,
320 comprehensiveness, and clarity, with many additional example code snippets
323 - The runtime system now includes the 'Metabalancer', which can provide
324 substantial performance improvements for applications that exhibit dynamic
325 load imbalance. It provides two primary benefits. First, it automatically
326 optimizes the frequency of load balancer invocation, to avoid work stoppage
327 when it will provide too little benefit. Second, calls to AtSync() are made
328 less synchronous, to further reduce overhead when the load balancer doesn't
329 need to run. To activate the Metabalancer, pass the option +MetaLB at
330 runtime. To get the full benefits, calls to AtSync() should be made at every
331 iteration, rather than at some arbitrary longer interval as was previously
334 - Many feature additions and usability improvements have been made in the
335 interface translator that generates code from .ci files:
336 * Charmxi now provides much better error reports, including more accurate
337 line numbers and clearer reasons for failure, including some semantic
338 problems that would otherwise appear when compiling the C++ code or even at
340 * A new SDAG construct 'case' has been added that defines a disjunction over a
341 set of 'when' clauses: only one 'when' out of a set will ever be triggered.
342 * Entry method templates are now supported. An example program can be found
343 in tests/charm++/method_templates/.
344 * SDAG keyword "atomic" has been deprecated in favor of the newly supported
345 keyword "serial". The two are synonymous, but "atomic" is now provided only
346 for backward compatibility.
347 * It is no longer necessary to call __sdag_init() in chares that contain SDAG
348 code - the generated code does this automatically. The function is left as
349 a no-op for compatibility, but may be removed in a future version.
350 * Code generated from .ci files is now primarily in .def.h files, with only
351 declarations in .decl.h. This improves debugging, speeds compilation,
352 provides clearer compiler output, and enables more complete encapsulation,
353 especially in SDAG code.
354 * Mainchare constructors are expected to take CkArgMsg*, and always have
355 been. However, charmxi would allow declarations with no argument, and
356 assume the message. This is now deprecated, and generates a warning.
358 - Projections tracing has been extended and improved in various ways
359 * The trace module can generate a record of network topology of the nodes in
360 a run for certain platforms (including Cray), which Projections can
362 * If the gzip library (libz) is available when Charm++ is compiled, traces
363 are compressed by default.
364 * If traces were flushed as a results of filled buffers during the run, a
365 warning will be printed at exit to indicate that the user should be wary of
366 interference that may have resulted.
367 * In SMP builds, it is now possible to trace message progression through the
368 communication threads. This is disabled by default to avoid overhead and
369 potential misleading interpretation.
371 - Array elements can be block-mapped at the SMP node level instead of at the
372 per-PE level (option "+useNodeBlkMapping").
374 - AMPI can now privatize global and static variables using TLS. This is
375 supported in C and C++ with __thread markings on the variable declarations
376 and definitions, and in Fortran with a patched version of the gfortran
377 compiler. To activate this feature, append '-tls' to the '-thread' option's
378 argument when you link your AMPI program.
380 - Charm can now be built to only support message priorities of a specific data
381 type. This enables an optimized message queue within the the runtime
382 system. Typical applications with medium sized compute grains may not benefit
383 noticeably when switching to the new scheduler. However, this may permit
384 further optimizations in later releases.
386 The new queue is enabled by specifying the data type of the message
387 priorities while building charm using --with-prio-type=dtype. Here, dtype can
388 be one of char, short, int, long, float, double and bitvec. Specifying bitvec
389 will permit arbitrary-length bitvector priorities, and is the current default
390 mode of operation. However, we may change this in a future release.
392 - Converse now provides a complete set of wrappers for
393 fopen/fread/fwrite/fclose to handle EINTR, which is not uncommon on the
394 increasingly-popular Lustre. They are named CmiF{open,read,write,close}, and
395 are available from C and C++ code.
397 - The utility class 'CkEntryOptions' now permits method chaining for cleaner
398 usage. This applies to all its set methods (setPriority, setQueueing,
399 setGroupDepID). Example usage can be found in examples/charm++/prio/pgm.C.
401 - When creating groups or chare arrays that depend on the previous construction
402 of another such entity on the local PE, it is now possible to declare that
403 dependence to the runtime. Creation messages whose dependence is not yet
404 satisfied will be buffered until it is.
406 - For any given chare class Foo and entry method Bar, the supporting class's
407 member CkIndex_Foo::Bar() is used to lookup/specify the entry method
408 index. This release adds a newer API for such members where the argument is a
409 function pointer of the same signature as the entry method. Those new
410 functions are used like CkIndex_Foo::idx_Bar(&Foo::Bar). This permits entry
411 point index lookup without instantiating temporary variables just to feed the
412 CkIndex_Foo::Bar() methods. In cases where Foo::Bar is overloaded, &Foo::Bar
413 must be cast to the desired type to disambiguate it.
415 - CkReduction::reducerType now have PUP methods defined; and can hence be
416 passed as parameter-marshalled arguments to entry methods.
418 - The runtime option +stacksize for controlling the allocation of user-level
419 threads' stacks now accepts shorthanded annotation such as 1M.
421 - The -optimize flag to the charmc compiler wrapper now passes more aggressive
422 options to the various underlying compilers than the previous '-O'.
424 - The charmc compiler wrapper now provides a flag -use-new-std to enable
425 support for C11 and C++11 where available. To use this in application code,
426 the runtime system must have been built with that flag as well.
428 - When using, CmiMemoryUsage(), the runtime can be instructed not to use the
429 underlying mallinfo() library call, which can be inaccurate in settings where
430 usage exceeds INT_MAX. This is accomplished by setting the environment
431 variable "MEMORYUSAGE_NO_MALLINFO".
433 - Experimental Features
434 * Initial implementation of a fast message-logging protocol. Use option
435 'mlogft' to build it.
436 * Message compression support for persistent message on Gemini machine layer.
437 * Node-level inter-PE loop/task parallelization is now supported through
439 * New temperature/CPU frequency aware load balancer
440 * Support interoperation of Charm++ and native MPI code through dynamically
441 switching control between the two
442 * API in centralized load balancers to get and set PE speed
443 * A new scheme for optimization of double in-memory checkpoint/restart.
444 * Message combining library for improved fine-grained communication
446 * Support for partitioning of allocated nodes into subsets that run
447 independent Charm++ instances but can interact with each other.
449 Platform-Specific Changes
450 -------------------------
453 * The gemini_gni network layer has been heavily tuned and optimized,
454 providing substantial improvements in performance, scalability, and
456 * The gemini_gni-crayxe machine layer supports a 'hugepages' option at build
457 time, rather than requiring manual configuration file editing.
458 * Persistent message optimizations can be used to reduce latency and
460 * Experimental support for 'urgent' sends, which are sent ahead of any other
461 outgoing messages queued for transmission.
463 - IBM Blue Gene Q: Experimental machine-layer support for the native PAMI
464 interface and MPI, with and without SMP support. This supports many new
465 systems, including LLNL's Sequoia, ALCF's Mira, and FZ Juelich's Juqueen.
467 There are three network-layer implementations for these systems: 'mpi',
468 'pami', and 'pamilrts'. The 'mpi' layer is stable, but its performance and
469 scalability suffers from the additional overhead of using MPI rather than
470 driving the interconnect directly. The 'pami' layer is well tested for NAMD,
471 but has shown instability for other applications. It is likely to be replaced
472 by the 'pamilrts' layer, which is more generally stable and seems to provide
473 the same performance, in the next release.
475 In addition to the common 'smp' option to build the runtime system with
476 shared memory support, there is an 'async' option which sometimes provides
477 better performance on SMP builds. This option passes tests on 'pamilrts', but
478 is still experimental.
480 Note: Applications that have large number of messages may crash in default
481 setup due to overflow in the low-level FIFOs. Environment variables
482 MUSPI_INJFIFOSIZE and PAMI_RGETINJFIFOSIZE can be set to avoid application
483 failures due to large number of small and large messages respectively. The
484 default value of these variable is 65536 which is sufficient for 1000
487 - Infiniband Verbs: Better support for more flavors of ibverbs libraries
490 * Experimental rendezvous protocol for better performance above some MPI
492 * Some tuning parameters ("+dynCapSend" and "+dynCapRecv") are now
493 configurable at job launch, rather than Charm++ compilation.
495 - PGI C++: Disable automatic 'using namespace std;'
497 - Charm++ now supports ARM, both non-smp and smp.
499 - Mac OS X: Compilation options to build and link correctly on newer versions
502 ================================================================================
503 What's new in Charm++ 6.4.0
504 ================================================================================
506 --------------------------------------------------------------------------------
508 --------------------------------------------------------------------------------
510 - Cray XE and XK systems using the Gemini network via either MPI
511 (mpi-crayxe) or the native uGNI (gemini_gni-crayxe)
513 - IBM Blue Gene Q, using MPI (mpi-bluegeneq) or PAMI (pami-bluegeneq)
515 - Clang, Cray, and Fujitsu compilers
517 - MPI-based machine layers can now run on >64k PEs
519 --------------------------------------------------------------------------------
521 --------------------------------------------------------------------------------
523 - Added a new [reductiontarget] attribute to enable
524 parameter-marshaled recipients of reduction messages
526 - Enabled pipelining of large messages in CkMulticast by default
528 - New load balancers added:
531 * Scotch graph partitioning based: ScotchLB and Refine and Topo variants
534 - Load balancing improvements:
536 * Allow reduced load database size using floats instead of doubles
537 * Improved hierarchical balancer
538 * Periodic balancing adapts its interval dynamically
539 * User code can request a callback when migration is complete
540 * More balancers properly consider object migratability and PE
541 availability and speed
542 * Instrumentation records multicasts
544 - Chare arrays support options that can enable some optimizations
546 - New 'completion detection' library for parallel process termination
547 detection, when the need for modularity excludes full quiescence
550 - New 'mesh streamer' library for fine-grain many-to-many collectives,
551 handling message bundling and network topology
553 - Memory pooling allocator performance and resource usage improved
556 - AMPI: More routines support MPI_IN_PLACE, and those that don't check
559 ================================================================================
560 What's new in Charm++ 6.2.1 (since 6.2.0)
561 ================================================================================
563 --------------------------------------------------------------------------------
564 New Supported Platforms:
565 --------------------------------------------------------------------------------
567 POWER7 with LAPI on Linux
569 Infiniband on PowerPC
571 --------------------------------------------------------------------------------
573 --------------------------------------------------------------------------------
575 - Better support for multicasts on groups
576 - Topology information gathering has been optimized
577 - Converse (seed) load balancers have many new optimizations applied
578 - CPU affinity can be set more easily using +pemap and +commap options
579 instead of the older +coremap
580 - HybridLB (hierarchical balancing for very large core-count systems)
581 has been substantially improved
582 - Load balancing infrastructure has further optimizations and bug fixes
583 - Object mappings can be read from a file, to allow offline
584 topology-aware placement
585 - Projections logs can be spread across multiple directories, speeding
586 up output when dealing with thousands of cores (+trace-subdirs N
587 will divide log files evenly among N subdirectories of the trace
588 root, named PROGNAME.projdir.K)
589 - AMPI now implements MPI_Issend
590 - AMPI's MPI_Alltoall uses a flooding algorithm more agressively,
591 versus pairwise exchange
592 - Virtualized ARMCI support has been extended to cover the functions
595 --------------------------------------------------------------------------------
596 Architecture-specific changes
597 --------------------------------------------------------------------------------
599 - LAPI SMP has many new optimizations applied
601 - Net builds support the use of clusters' mpiexec systems for job
602 launch, via the ++mpiexec option to charmrun
604 ================================================================================
605 What's new in Charm++ 6.2.0 (since 6.1)
606 ================================================================================
608 --------------------------------------------------------------------------------
609 New Supported Platforms:
610 --------------------------------------------------------------------------------
612 64-bit MIPS, such as SiCortex, using mpi-linux-mips64
614 Windows HPC cluster, using mpi-win32/mpi-win64
616 Mac OSX 10.6, Snow Leopard (32-bit and 64-bit).
618 --------------------------------------------------------------------------------
620 --------------------------------------------------------------------------------
623 - Smarter build/configure scripts
624 - A new interface for model-based load balancing
625 - new CPU topology API
626 - a general implementation of CmiMemoryUsage()
627 - Bug fix: Quiescence detection (QD) works with immediate messages
628 - New reduction functions implemented in Converse
629 - CCS (Converse Client-Server) can deliver message to more than one processor
630 - Added a memory-aware adaptive scheduler, which can be optionally
632 - Added preliminary support for automatic message prioritization
633 (disabled by default)
636 - Cross-array and cross-group sections
637 - Structured Dagger (SDAG): Support templated arguments properly
638 - Plain chares support checkpoint/restart (both in-memory and disk-based)
639 - Conditional packing of messages and parameters in SMP scenario
640 - Changes to the CkArrayIndex class hierarchy
641 -- sizeof() all CkArrayIndex* classes is now the same
642 -- Codes using custom array indices have to use placement-new to construct
643 their custom index. Refer example code: examples/charm++/hello/fancyarray/
644 -- *** Backward Incompatibility ***
645 CkArrayIndex[4D/5D/6D]::index are now of type int (instead of short)
646 However the data is stored as shorts. Access by casting
647 CkArrayIndexND::data() appropriately
648 -- *** Deprecated ***
649 The direct use of public data member
650 CkArrayIndexND::index (N=1..6) is deprecated. We reserve the right to
651 change/remove this variable in future releases of Charm++.
652 Instead, please access the indices via member function:
653 int CkArrayIndexND::data()
656 - Compilers renamed to avoid collision with host MPI (ampicc, ampiCC,
658 - Improved MPI standard conformance, and documentation of non-conformance
659 * Bug fixes in: MPI_Ssend, MPI_Cart_shift, MPI_Get_count
660 * Support MPI_IN_PLACE in MPI_(All)Reduce
661 * Define various missing constants
662 - Return the received message's tag in response to a non-blocking
663 wildcard receive, to support SuperLU
664 - Improved tracing for BigSim
666 Multiphase Shared Arrays (MSA)
667 - Typed handles to enforce phases
668 - Split-phase synchronization to enable message-driven execution
672 - Automatic tracing of API calls for simulation and analysis
675 - Wider support for architectures other than net- (in particular MPI layers)
676 - Improved support for large scale debugging (better scalability)
677 - Enhanced record/replay stability to handle various events, and to
678 signal unexpected messages
679 - New detailed record/replay: The full content of messages can be
680 recorded, and a single processor can be re-executed outside of the
684 - Tracing of nested entry methods
686 Automatic Performance Tuning
687 - Created an automatic tuning framework [still for experimental use only]
690 - Network-topology / node aware spanning trees used internally for and
691 lower bytes on the network and improved performance in multicasts and
692 reductions delegated to this library
695 - Improved OneTimeMulticastStrategy classes
698 - Out-of-core support, with prefetching capability
699 - Detailed tracing of MPI calls
700 - Detailed record/replay support at emulation time, capable of
701 replaying any emulated processor after obtained recorded logs.
703 --------------------------------------------------------------------------------
704 Architecture-specific changes
705 --------------------------------------------------------------------------------
708 - Can run jobs with more than 1024 PEs
711 - New charmrun option ++no-va-randomization to disable address space
712 randomization (ASLR). This is most useful for running AMPI with
716 - Default to using ampicxx instead of mpiCC
719 - The +p option now has the same semantics as in other smp builds
722 - Support for VSX in SIMD abstraction API
725 - Compilers and options have been updated to the latest ones
728 - Added routines for measuring performance counters on BG/P.
729 - Updated to support latest DCMF driver version. On ANL's Intrepid, you may
730 need to set BGP_INSTALL=/bgsys/drivers/V1R4M1_460_2009-091110P/ppc in your
731 environment. This is the default on ANL's Surveyor.
734 - cputopology information is now available on XT3/4/5
737 - Bug fix: plug memory leaks that caused failures in long runs
738 - Optimized to reduce startup delays
741 - Support for SMP (experimental)
744 ================================================================================
745 Note that changes from 5.9, 6.0, and 6.1 are not documented here. A partial list
746 can be found on the charm download page, or by reading through version control
749 ================================================================================
750 What's New since Charm++ 5.4 release 1
751 ================================================================================
753 --------------------------------------------------------------------------------
754 New Supported Platforms:
755 --------------------------------------------------------------------------------
756 1. Charm++ ported to IA64 Itanium running Win2K and Linux, Charm++ also support
757 Intel C/C++ compilers;
759 2. Charm++ ported to Power Macintosh powerpc running Darwin;
761 3. Charm++ ported to Myrinet networking with GM API;
763 --------------------------------------------------------------------------------
764 Summary of New Features:
765 --------------------------------------------------------------------------------
767 Structured Dagger is a coordination language built on top of CHARM++.
768 Structured Dagger allows easy expression of dependences among messages and
769 computations and also among computations within the same object using
770 when-blocks and various structured constructs.
772 2. Entry functions support parameter marshalling
773 Now you can declare and invoke remote entry functions using parameter
774 marshalling instead of defining messages.
776 3. Easier running - standalone mode
777 For net-* version running locally, you can now run Charm programs without
778 charmrun. Running a node program directly from command line is now the
779 same as "charmrun +p1 <program>"; for SMP version, you can also specify
780 multiple (local) processors, as in "program +p2".
783 --------------------------------------------------------------------------------
785 --------------------------------------------------------------------------------
786 1. "build" changed for compilation of Charm++
787 To build Charm++ from scratch, we now take additional command line options
788 to compile with addon features and using different compilers other than gcc.
789 For example, to build Linux IA64 with Myrinet support, type command:
790 ./build net-linux-ia64 gm
793 ******* Old Change histories *******
796 ================================================================================
797 What's New in Charm++ 5.4 release 1 since 5.0
798 ================================================================================
800 --------------------------------------------------------------------------------
801 New Supported Platforms:
802 --------------------------------------------------------------------------------
804 1. Win9x/2000/NT: with Visual C++ or Cygwin gcc/g++, you can compile and run
805 Charm++ programs on all Win32 platforms.
807 2. Scyld Beowulf: Charm++ has been ported to the Linux-based Scyld Beowulf
808 operating system. For more information on Scyld, see <http://www.scyld.com>
810 3. MPI with VMI: Charm++ has been ported to NCSA's Virtual Machine Interface,
811 which is an efficient messaging library for heterogeneous cluster
815 --------------------------------------------------------------------------------
816 Summary of New Features:
817 --------------------------------------------------------------------------------
818 1. Dynamic Load balancing:
819 Chare migration is supported in the new release. Migration-based dynamic
820 load balancing framework with various load balancing strategies library has
824 Charm++ array is supported. You can now create an array of Chare objects
825 and use array index to refer the Charm++ array elements. A reduction
826 library on top of Chare array has been implemented and included.
829 Projections, a Java application for Charm++ program performance analysis and
830 visualization, has been included and distributed in the new release. Two
831 trace modes are available: trace-projections and trace-summary. Trace-summary
832 is a light-weight trace library compared to trace-projections.
835 AMPI is a load-balancing based library for porting legacy MPI applications
836 to Charm++. With few changes in the original MPI code to AMPI, the new
837 legacy MPI application on Charm++ will gain from Charm++'s adptive
838 load balancing ability.
841 "Charmrun" is now available on all platforms, with a uniform command line
842 syntax. You can forget the difference between net-* versions and MPI versions,
843 and run charm++ application with this same charmrun command syntax.
844 ++local option is added in charmrun for net-* version, it provides
845 simple local use of Charm and no longer require the ability to
846 "rsh localhost" or a nodelist file in order to run charm only on the local
847 machine. This is especially attractive when you run Charm++ on Windows.
850 Many new libraries have been added in this release. They include:
851 1) master-slave library: for writing manager-worker paradigm programs.
852 2) receiver library: provide asynchronous communication mode for chare array.
853 3) f90charm: provides Fortran90 bindings for Charm++ Array.
854 4) BlueGene: a Charm++/Converse emulator for IBM proposed Blue Gene.
856 --------------------------------------------------------------------------------
858 --------------------------------------------------------------------------------
859 1. message declaration syntax in .ci file:
860 The message declaration syntax for packed/varsize messages has been changed.
861 The packed/varsize keywords are eliminated, and you can specify the actual
862 actual varsize arrays in the interface file and have the translator generate
863 alloc, pack and unpack.
866 Here is the detailed list of Changes:
868 --------------------------------------------------------------------------------
870 --------------------------------------------------------------------------------
872 10/06/1999 rbrunner Added migration-based dynamic load balancing
874 11/15/1999 olawlor Added reduction support foe Charm++ arrays
875 02/06/2000 milind Added AMPI, an implementation of MPI with
876 dynamic load balancing
877 02/18/2000 paranjpy New platforms supported: net-win32, and net-win32-smp
878 04/04/2000 olawlor Added arbitrarily indexed Charm++ arrays.
879 Also, added translator support for new arrays.
880 04/15/2000 olawlor Added "puppers" for packing and unpacking
882 06/14/2000 milind Added the threaded FEM framework.
884 --------------------------------------------------------------------------------
886 --------------------------------------------------------------------------------
888 10/09/1999 rbrunner Added packlib, a library for C and C++ to
889 pack-unpack data to/from Charm++ messages.
890 10/13/1999 gzheng New LB strategy: RefineLB
891 10/13/1999 paranjpy New LB Strategy: Heap
892 10/14/1999 milind New LB Strategy: Metis
893 10/19/1999 olawlor New test program for testing LB strategies.
894 10/21/1999 gzheng New trace mode: trace-summary
895 10/28/1999 milind New supported platform: net-sol-x86
896 10/29/1999 milind Added runtime checks for ChareID assignment.
897 11/10/1999 rbrunner Added Neighborhood base strategy for LB
899 11/15/1999 olawlor conv-host now reads in a startup file
901 11/15/1999 olawlor New test program for testing array reductions.
902 11/16/1999 rbrunner Added processor-speed checking functions to
904 11/19/1999 milind Mapped SIGUSR to a Ccd condtion handler
905 11/22/1999 rbrunner New LB strategy: WSLB
906 11/29/1999 ruiliu Modified Metis LB strategy to deal with
907 different processor speeds
908 12/16/1999 rbrunner New LB strategy: GreedyRef
909 12/16/1999 rbrunner New LB strategy: RandRef
910 12/21/1999 skumar2 New LB strategy: CommLB
911 01/03/2000 rbrunner New LB strategy: RecBisectBfLB
912 01/08/2000 skumar2 New LB strategy: Comm1LB, with varying processor
914 01/18/2000 milind Modified SM library syntax, and added a test
916 01/19/2000 gzheng Added irecv, a library to simplify conversion
917 of message-passing programs to Charm++
918 02/20/2000 olawlor Added preliminary broadcast support to Charm++
920 02/23/2000 paranjpy Added converse-level quiescence detection
921 03/02/2000 milind Added ++server-port option to pre-specify
923 03/10/2000 wilmarth Random seed-based load balancer now uses
924 bit-vector for active PEs.
925 03/21/2000 gzheng Added support for marking user-defined events
927 03/28/2000 wilmarth Added CMK_TRUECRASH. Very helpful for
928 post-mortem debugging of Charm++ programs on
930 03/31/2000 jdesouza Added Fortran90 support to the Charm++
931 interface translator.
932 03/09/2000 milind Added support for -LANG and -rpath options
933 in charmc for Origin2000.
934 04/28/2000 milind Added prioritized converse threads.
935 05/01/2000 milind Added test programs for TeMPO, AMPI and irecv.
936 05/04/2000 milind New supported platform: mpi-sp.
937 05/04/2000 gzheng Added irecv pingpong program.
938 05/17/2000 olawlor Each chare, group and array element now has to
939 have migration constructor.
940 05/24/2000 milind Added Jacobi3D programs for irecv and AMPI both.
941 05/24/2000 milind Made migratable an optional attribute of
942 chares, groups, and nodegroups.
943 Arrays are by default migratable.
944 05/29/2000 paranjpy Added pup methods to arrays, reductions etc
946 06/13/2000 milind Made CtvInitialize idempotent. That is, it
947 can be called by any number of threads now,
948 only the first one will actually do
950 06/20/2000 milind Added a simple test program for the FEM
952 07/06/2000 milind Imported Metis 4.0 sources in the CVS tree.
953 Also added code to make metis libraries and
954 executables to Makefile.
955 07/07/2000 milind Added more meaningfull error messages using
956 perror in addition to a cryptic error codes in
958 07/10/2000 milind fem and femf are now recognized as "languages"
960 07/10/2000 saboo Added the derived datatypes library.
961 07/13/2000 milind Added +idle_timeout functionality. It takes a
962 commandline parameter denoting milliseconds of
963 maximum consecutive idle time allowed per
965 07/14/2000 milind Added group multicast. Added
966 CkSendMsgBranchMulti, CldEnqueueMulti, and
967 translator changes to support it.
968 07/14/2000 milind SUPER_INSTALL now takes "-*" arguments prior
969 to the target, that will be passed to make as
970 "makeflags". This makes it easy to suppress
971 make's output of commands etc (with the -s
972 flag). As a result of this, several Makefiles
974 07/18/2000 milind Added support for using "dbx" on suns as
976 07/19/2000 milind Added ability to tracemode projections which
977 produces binary trace files. Use flag
978 +binary-trace on the command line.
979 07/26/2000 milind Separated AMPI from TeMPO.
980 07/28/2000 milind Added test programs to test reduce, alltoall
981 and allreduce functionality of AMPI.
982 08/02/2000 milind Added an option to let the user specify which
983 "xterm" to use. For example, on some systems
984 (CDE), only dtterm is installed. So, by
985 putting ++xterm dtterm on the conv-host
986 commandline, one can use dtterm when ++in-xterm
987 option is specified on conv-host commandline.
988 08/14/2000 milind FEM Framework: Added capabilities to handle
989 esoteric meshes to standalone offline programs.
990 Makefile now produces gmap and fgmap programs,
991 which are used for this purpose. They convert
992 the mesh to a graph before partitioning it
994 08/24/2000 milind Added the 2D crack propagation program as a
995 test program for FEM framework.
996 08/25/2000 milind Initial implementation of isomalloc-based
997 threads. This implementation uses a fixed
998 stack size for all threads (can be set at
1000 08/26/2000 milind Added a macro CtvAccessOther that lets you
1001 get/set a Ctv variable of any thread. It
1002 should be invoked as CtvAccessOther(thread,
1003 varname); Added CthGetData function to each of
1004 the threads implementation. This function is
1005 used in the CtvAccessOther macro.
1006 08/27/2000 milind FEM Framework: Separated mesh to graph
1007 conversion capability into a separate program.
1008 This way, the generated graph can be partitioned
1010 09/04/2000 milind Added the class static readonly variables to
1012 09/05/2000 milind FEM Framework: A very fast O(n) algorithm for
1013 mesh2graph , uses more memory, but the tradeoff
1014 was worth it. Coded by Karthik Mahesh, minor
1015 optimizations by Milind.
1016 09/05/2000 milind Added a barebones charm kernel scheduling
1017 overhead measurement program.
1018 09/15/2000 milind Added pup support for AMPI and FEM framework.
1019 09/20/2000 olawlor Added capability to have an array of base type
1020 where individual element could be of derived
1022 10/03/2000 gzheng New supported platform: net-linux-axp
1023 10/05/2000 skumar2 Added program littleMD to the test suite.
1024 10/07/2000 skumar2 New job scheduler (Faucets projects).
1025 10/15/2000 milind Improved support for Fortran90 in charmc.
1026 11/04/2000 jdesouza Made the Faucets scheduler multi-threaded.
1027 11/05/2000 olawlor FEM Framework: supports multiple element types,
1028 mesh re-assembly, etc.
1029 11/15/2000 gzheng New platform support: net-cygwin
1030 11/18/2000 gzheng conv-host no longer needs /bin/csh to start
1032 CMK_CONV_HOST_CSH_UNAVAILABLE to 1 to use
1034 11/25/2000 milind Finished experimental implementation of
1035 converse-threads based on co-operative pthreads.
1036 11/25/2000 milind Added a benchmark suite of all pingpongs in
1038 11/28/2000 milind Removed deletion of _idx at the end of every
1039 send or doneInserting call. Instead now it is
1040 in the destructor of the proxy. This allows us
1041 to cache proxies, when proxy creation becomes
1043 11/28/2000 olawlor Added "seek blocks" to puppers. This should
1044 allow out-of-order pup'ing without the ugliness
1045 of getBuf; and in a way that works with all
1047 11/29/2000 olawlor Simplified and regularized command-line-argument
1049 11/29/2000 milind AMPI: Added multiple-communicators capability.
1050 12/05/2000 gzheng Now /bin/sh is default shell to fork node
1051 program on remote machines.
1052 12/13/2000 olawlor Added charmrun wrapper for poe on mpi-sp.
1053 12/14/2000 milind Added bluegene emulator sources and test
1054 programs. Added "bluegene" as a language known
1055 to charmc. Makefile now has a target called
1056 bluegene. Added preliminary bluegene
1057 documentation. (copied from Arun's webpage.)
1058 12/15/2000 gzheng f90charm addition to Makefile and charmc. Also,
1059 added fixed size arrays support to f90charm. A
1060 test program f90charm/hello is checked in.
1061 12/17/2000 milind Added rtest test program. Contributed by jim to
1062 test Converse message transmission.
1063 12/20/2000 olawlor Added charmconfig script. Enables automatic
1064 determination of C++ compiler properties,
1065 replacing the verbose and error-prone
1066 conv-mach.h entries for CMK_BOOL,
1067 CMK_STL_USE_DOT_H, CMK_CPP_CAST_OK, ...
1068 12/20/2000 olawlor Charm++ Arrays optimizations: Key and object
1069 now variable-length fields, instead of pointers.
1070 This extra flexibility lets us save many
1071 dynamic allocations in the array framework.
1072 12/20/2000 olawlor Added PUP::able support-- dynamic type
1073 identification, allocation, and deletion.
1074 Allows you to write: p(objPtr); and
1075 objPointer will be properly identified,
1076 allocated, packed, and deallocated (depending
1077 on the PUP::er). Requires you to register any
1078 such classes with DECLARE_PUPable and
1080 12/20/2000 olawlor Arrays optimizations: Made CkArrayIndex
1081 fixed-size. This significantly improves
1082 messaging speed (7 us instead of 10 us
1083 roundtrip). Move spring cleaning check into a
1084 CcdCallFnAfter, which gains more speed (down to
1086 12/20/2000 olawlor More optimizations: Minor speed tweaks--
1087 conv-ccs.c uses hashtable for handler lookup;
1088 conv-conds skips timer test until needed;
1089 convcore.c scheduler loop optmizations (no
1090 superfluous EndIdle calls); threads.c
1091 CMK_OPTIMIZE-> no mprotect.
1092 12/20/2000 olawlor More Optimizations: Minor speed tweaks-- ck.C
1093 groups cldEnqueue skip; init.h defines
1094 CkLocalBranch inline; and supporting changes.
1095 12/22/2000 gzheng IA64 support for Converse user level threads.
1096 01/02/2001 olawlor CCS: Minor update-- enabled CcsProbe, cleaned
1097 up superflous debug messages in server, added
1098 Java interface (originally written for
1100 01/09/2001 gzheng charmconfig converted to autoconf style, need
1101 to change configure.in and conv-autoconfig.h.in,
1102 and run autoconf to get configure and copy to
1103 charmconfig. added fortran subroutine name
1104 test and get libpthread.a
1105 01/10/2001 milind Added telnet method of getting libpthread.a
1106 from charm webserver.
1107 01/11/2001 olawlor Moved projections files here from
1108 CVSROOT/projections-java. Added fast Java
1109 versions of the .log file input routines in
1110 LogReader, LogLoader, LogAnalyzer, and
1111 UsageCalc. Added "U.java" user interface
1112 utility file, allowing times to be input in
1113 seconds, milliseconds, or microseconds,
1114 instead of just microseconds.
1115 01/15/2001 gzheng add +trace-root to specify the directory to
1116 put log files in. this is need in Scyld cluster
1117 where there is no NFS mounting and no i/o
1118 access to home directory sharing on nodes.
1119 01/15/2001 milind Made AMPI into a f90 module instead of
1120 'ampif.h' inclusion. AMPI f90 bindings are
1121 now more inclusive. Fixed argc,argv handling
1122 bugs in ArgsInfo message. Fixed a bug in pup
1123 that caused thread not to be sized, but was
1124 packed nevertheless. Moved irecv to waitall
1125 instead of at in ampi_start. Made
1126 AMPI_COMM_WORLD to be 0, because it clashed
1127 with wildcard(-1). AMPI_COMM_UNIVERSE is now
1128 handled properly in the AMPI module.
1129 C/C++ data members are NOT visible to
1131 01/18/2001 gzheng New supported platform: net-linux-scyld
1132 01/20/2001 olawlor Moved array index field from CMessage_* to the
1133 Ck envelope itself. This is the right thing
1134 to do, because any message may be sent to/from
1135 an array element. To reduce the wasted space
1136 in a message, a union is used to overlay the
1137 fields for the various possible message types.
1138 01/29/2001 olawlor Freed charmrun on net-* version from using
1139 remote shell to fork off processes. One can now
1140 use a daemon provided in the distribution.
1141 02/07/2001 olawlor Added debugging support to puppers.
1142 02/13/2000 gzheng Added ++local option to charmrun to start node
1143 program locally without any daemon; fix the
1144 hang program if you type wrong pgm name in
1145 scyld version, and redirect all output to
1146 /dev/null, otherwise all node program can send
1147 its output to console in scyld. Also implemented ++local in net-win32 version.
1148 02/26/2000 milind Changed the varsize syntax. Now one can specify
1149 actual varsize arrays in the interface file
1150 and have the translator generate alloc, pack
1153 --------------------------------------------------------------------------------
1155 --------------------------------------------------------------------------------
1157 10/29/1999 milind Replaced jmemcpy by memcpy in net versions, as
1158 it was causing a bit to flip (bug reported
1160 10/29/1999 milind Fixed multiline macros in all header files.
1161 02/05/2000 milind Fixed linking errors by getting the order of
1162 libraries right from the charmc command-line.
1163 02/18/2000 paranjpy Fixed Charm++ initialization bug on SMPs.
1164 02/21/2000 milind Fixed a context-switching bug in mipspro version
1166 02/25/2000 milind Charm++ interface translator was segfaulting
1167 on interface file errors. Fixed that. Also,
1168 added linenumbers to error messages.
1169 03/02/2000 milind Made CCS work on SMPs.
1170 03/07/2000 milind Made ConverseInit consistent with the manual on
1172 04/18/2000 milind Fixed a bug in CkWaitFuture, which was caching
1173 a variable locally, while it was changed by
1175 05/04/2000 paranjpy Fixed argv deletion bug on net-win32-smp.
1176 06/08/2000 milind sp3 version: changed optimization flags, which
1177 where power2 processor-specific.
1178 06/20/2000 milind mpi-* versions: Fixed ConverseExit since it was
1179 not obeying the following statement in the MPI
1180 standard: The user must ensure that all pending
1181 communications involving a process completes
1182 before the process calls MPI_FINALIZE.
1183 07/05/2000 milind Fixed a nasty bug in charmc in the -cp option.
1184 It used to append the name provided to -o flag
1185 to the directory provided to the -cp flag.
1186 Thus, -o ../pgm -cp ../bin options meant that
1187 the pgm would be copied to ../bin/.., which is
1188 not the expected behavior. This fix correctly
1189 copies pgm to ../bin.
1190 07/07/2000 milind Removed variable arg_myhome, as it was not
1191 being used anywhere, and also, setting it was
1192 causing problems of env var HOME was not set.
1193 07/27/2000 milind thishandle for the arrayelement was not being
1194 correctly set. Bug was reported by Neelam.
1195 08/26/2000 milind Origin2000: Changed the page alignment to
1196 reflect the mmap alignment. The mmap man page
1197 specifically states that it is not the same as
1199 09/02/2000 milind Fixed a bug in code generated for threaded
1200 (void) entry methods of array elements. The
1201 dummy message that is passed to that method in
1202 a thread has to be deleted before calling the
1203 object method, because upon object method's
1204 return, the thread might have migrated.
1205 09/03/2000 olawlor Minor fix-fixes: 1.) Change to LBObjid hash
1206 function would fail for >4-int object indices.
1207 Replaced with proper function, which also
1208 preserves the 1-int case. 2.) Array element
1209 sends must go via the message queue to prevent
1210 stack build-up for deep single-processor call
1211 chains. These might happen, e.g., in a driver
1212 element calling itself for the main time loop.
1213 Messages are now properly noted as sent, then
1214 wait through the queue for delivery. This
1215 entailed minor reorganization of the message
1217 09/21/2000 olawlor Tiny SMP thread fix-- registrations of a
1218 thread-private variable now reserve space on
1219 calls after the first. This wastes space for
1220 multiple CthInitialize's-- it's a quick hack to
1221 get threads working again on SMP versions.
1222 10/16/2000 olawlor A few CCS fixes: -Added split-phase reply
1223 (delay reply indefinitely) -Cleaned up error
1224 handling -Pass user data as "void *" instead of
1226 11/03/2000 wilmarth Removed 0 size array allocation in Charm++
1227 quiescence detection.
1228 11/20/2000 gzheng Rewrote part of Fiber thread, including a bug
1229 fix for a the non thread-safe function, and a
1230 different fiber free strategy.
1231 11/29/2000 gzheng The LB init procedure tried to allocate
1232 65536*160 as initial size, which is 10M memory
1233 for communication table, which is too big.
1234 Cut it down to roughly 1M, and it can expand
1236 12/05/2000 gzheng In many cases, conv-host exits without print
1237 out the error message from remote shell. try
1238 to fix it by calling sync to flush the pipe
1240 12/10/2000 milind net-linux: Made static linking the default
1241 option because dynamic linking runtime causes
1242 isomalloc threads to crash.
1243 12/18/2000 milind Increased portability of isomalloc threads by
1244 removing dependence on alloca.
1245 12/28/2000 milind Fixed ctrl-getone abort bug on SMP.
1246 12/28/2000 milind Made _groupTable a pointer on which a
1247 constructor is explicitly called. Since it
1248 was a Cpv variable, its constructor was not
1249 called by default in case of an SMP version.
1250 12/29/2000 olawlor Prevent infinite copy constructor recursion on
1252 01/10/2001 olawlor Added "explicit" keyword to remove ambiguity
1253 for KCC, which was confused by the private
1254 PUP::er(int) "cast" constructor and the operator
1255 |(PUP::er &p,T &t) into rejecting all operator|
1256 (int,int) as ambiguous.
1257 2001/01/17 gzheng fix the charmconfig bug on paragon-red: the
1258 failure testing of fortran won't stop the
1260 01/20/2001 olawlor Arrays reduction: Fixed bug-- reduction may end
1261 because all contributors migrate away.
1262 01/29/2001 olawlor Fix heap-corrupting bug-- call ->init() on
1263 nodeGroupTable, which sets the "pending"
1264 message queue to NULL. This prevents a nasty
1265 delete-unitialized-data bug later on. Also
1266 delayed queue creation until messages actually
1269 --------------------------------------------------------------------------------
1270 Documentation Changes:
1271 --------------------------------------------------------------------------------
1273 01/31/2000 milind Installation manual: Fixed bugs pointed out by
1275 02/28/2000 wilmarth Added a new look Charm++ manual.
1276 06/20/2000 milind Added pdflatex support to generate PDF versions
1277 of manuals from LaTeX sources.
1278 12/05/2000 milind Added Orion's FEM manual. Converted from HTML.
1279 12/10/2000 milind Added pplmanual.sty for all manuals.
1280 12/17/2000 milind Added master-slave library documentation to
1282 12/21/2000 saboo Added DDT documentation.
1283 01/02/2001 olawlor Updated for new CCS version.
1285 --------------------------------------------------------------------------------
1287 --------------------------------------------------------------------------------
1289 10/24/1999 olawlor charmc is changed to Bourne shell script
1290 instead of csh. All conv-mach.csh are
1291 replaced by conv-mach.sh.
1292 10/25/1999 olawlor SUPER_INSTALL is converted to use bourne shell.
1293 10/28/1999 milind All Makefiles now take OPTS commandline
1295 01/16/2000 olawlor Simplified Charm++ interface translator.
1296 02/23/2000 ruiliu Changed rand() calls from all over the codes
1297 to the new Converse random number generator.
1298 02/26/2000 milind Simplified the converse scheduler loop by
1299 combining the maxmsgs and poll modes.
1300 08/31/2000 milind Imported system documentation into the CVS tree.
1301 Also added super_install target for docs with
1302 necessary Makefile modifications.
1303 09/08/2000 olawlor Made soft links use relative pathnames instead
1304 of absolute. This lets you move a charm++
1305 installation without having to recompile
1307 09/11/2000 olawlor Grouped commonly needed code in the new util
1308 directory. Also, added pup_c a C wrapper for
1310 09/11/2000 olawlor Slightly reorganized header structure. Now no
1311 headers should need to be listed twice (once in
1312 ALLHEADERS, again in CKHEADERS). Now headers
1313 are soft-linked instead of copied. This makes
1314 development much easier. Added support for the
1315 new Common/util directory.
1316 09/21/2000 olawlor Major reorganization of net-* codes. Now all
1317 the TCP socket routines are in separate files.
1318 Also combined windoes NT code with unix codes.
1319 09/21/2000 olawlor Major rewrite of CCS-- underlying protocol is
1320 now binary (send/recv binary data everywhere);
1321 conv-host forwards requests to nodes; and
1322 source has been significantly re-arranged.
1323 (especially if NODE_0_IS_CONVHOST).
1324 11/22/2000 milind Removed IDL translator from distribution.
1325 12/01/2000 olawlor Renamed conv-host charmrun; added test for
1326 script conv-host. Also added charmrun for most
1328 12/17/2000 milind Moved List related data structures into
1329 cklists.h in util. Removed most of the redundant
1330 list implementations.
1331 12/20/2000 gzheng SUPER_INSTALL: format the output of list of
1332 versions and make the help page fit into one
1334 12/24/2000 milind Added test-{charm,converse,ampi,fem} targets to
1336 12/28/2000 milind net-sol-smp now uses pthreads.
1337 01/29/2001 olawlor Merged windowsNT and unix build procedures by
1338 basing the Windows build on cygwin. Added
1339 scripts to deal with unix and windows