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 - Improvements to Sections:
44 * Array sections API has been simplified, with array sections being
45 automatically delegated to CkMulticastMgr (the most efficient implementation
46 in Charm++). Changes are reflected in Chapter 14 of the manual.
48 * Group sections can now be delegated to CkMulticastMgr (improved performance
49 compared to default implementation). Note that they have to be manually
50 delegated. Documentation is in Chapter 14 of Charm++ manual.
52 * Group section reductions are now supported for delegated sections
55 * Improved performance of section creation in CkMulticastMgr.
57 * CkMulticastMgr uses the improved spanning tree strategies. See above.
59 - GPU manager now creates one instance per OS process and scales the
60 pre-allocated memory pool size according to the GPU memory size and
61 number of GPU manager instances on a physical node.
63 - Several GPU Manager API changes including:
65 * Replaced references to global variables in the GPU manager API with calls to
68 * The user is no longer required to specify a bufferID in dataInfo struct.
70 * Replaced calls to kernelSelect with direct invocation of functions passed
71 via the work request object (allows CUDA to be built with all programs).
73 - Added support for malleable jobs that can dynamically shrink and
74 expand the set of compute nodes hosting Charm++ processes.
76 - Greatly expanded and improved reduction operations:
78 * Added built-in reductions for all logical and bitwise operations
79 on integer and boolean input.
81 * Reductions over groups and chare arrays that apply commutative,
82 associative operations (e.g. MIN, MAX, SUM, AND, OR, XOR) are now
83 processed in a streaming fashion. This reduces the memory footprint of
84 reductions. User-defined reductions can opt into this mode as well.
86 * Added a new `Tuple' reducer that allows combining multiple reductions
87 of different input data and operations from a common set of source
88 objects to a single target callback.
90 * Added a new `Summary Statistics' reducer that provides count, mean,
91 and standard deviation using a numerically-stable streaming algorithm.
93 - Added a `++quiet' option to suppress charmrun and charm++ non-error
96 - Calls to chare array element entry methods with the [inline] tag now
97 avoid copying their arguments when the called method takes its
98 parameters by const&, offering a substantial reduction in overhead in
101 - Synchronous entry methods that block until completion (marked with
102 the [sync] attribute) can now return any type that defines a PUP
103 method, rather than only message types.
105 - Static (non-generated) header files are now warning-free for
106 gcc -Wall -Wextra -pedantic.
108 - Deprecated setReductionClient and CkSetReductionClient in favor of
109 explicitly passing callbacks to contribute calls.
111 - On C++ standard library implementations with support for
112 std::is_constructible (e.g. GCC libstdc++ >4.5), chare array
113 elements only need to define a constructor taking CkMigrateMessage*
114 if it will actually be migrated.
116 - The PUP serialization framework gained support for some C++11
117 library classes, including unique_ptr and unordered_map, when the
118 underlying types have PUP operators.
122 - More efficient implementations of message matching infrastructure, multiple
123 completion routines, and all varieties of reductions and gathers.
125 - Support for user-defined non-commutative reductions, MPI_IN_PLACE, cancelling
126 receive requests, MPI_THREAD_FUNNELED, PSCW synchronization for RMA, and more.
128 - Fixes to AMPI's extensions for load balancing and to Isomalloc on SMP builds.
130 - More robust derived datatype support, optimizations for truly contiguous types.
132 - ROMIO is now built on AMPI and linked in by ampicc by default.
134 Platforms and Portability
136 - The runtime system code now requires compiler support for C++11
137 R-value references and move constructors. This is not expected to be
138 incompatible with any currently supported compilers.
140 - The next feature release (anticipated to be 6.9.0 or 7.0) will require
141 full C++11 support from the compiler and standard library.
143 - Added support for IBM POWER8 systems with the PAMI communication API,
144 such as development/test platforms for the upcoming Sierra and Summit
145 supercomputers at LLNL and ORNL. Contributed by Sameer Kumar of IBM.
147 - Mac OS (darwin) builds now default to the modern libc++ standard
148 library instead of the older libstdc++.
150 - Blue Gene/Q build targets have been added for the `bgclang' compiler.
152 - Charm++ can now be built on Cray's CCE 8.5.4+.
154 - Charm++ will now build without custom configuration on Arch Linux
156 - Charmrun can automatically detect rank and node count from
157 Slurm/srun environment variables.
159 - Many obsolete architecture, network, and compiler support files have
160 been removed. These include:
162 * Sony/Toshiba/IBM Cell (including PlayStation 3)
164 * Intel IA-64 (Itanium)
165 * Intel x86-32 for Windows, Mac OS X (darwin), and Solaris
167 * Older IBM AIX/POWER configurations
168 * GCC 3 and KAI compilers
171 ================================================================================
172 What's new in Charm++ 6.7.1
173 ================================================================================
175 Changes in this release are primarily bug fixes for 6.7.0. The major exception
176 is AMPI, which has seen changes to its extension APIs and now complies with more
177 of the MPI standard. A brief list of changes follows:
181 - Startup and exit sequences are more robust
183 - Error and warning messages are generally more informative
185 - CkMulticast's set and concat reducers work correctly
189 - AMPI's extensions have been renamed to use the prefix AMPI_ instead of MPI_
190 and to generally follow MPI's naming conventions
192 - AMPI_Migrate(MPI_Info) is now used for dynamic load balancing and all fault
193 tolerance schemes (see the AMPI manual)
195 - AMPI officially supports MPI-2.2, and also implements the non-blocking
196 collectives and neighborhood collectives from MPI-3.1
198 Platforms and Portability
200 - Cray regularpages build target has been fixed
202 - Clang compiler target for BlueGene/Q systems added
204 - Comm. thread tracing for SMP mode added
206 - AMPI's compiler wrappers are easier to use with autoconf and cmake
208 ================================================================================
209 What's new in Charm++ 6.7.0
210 ================================================================================
212 Over 120 bugs fixed, spanning areas across the entire system
216 - New API for efficient formula-based distributed sparse array creation
218 - CkLoop is now built by default
220 - CBase_Foo::pup need not be called from Foo::pup in user code anymore - runtime
221 code handles this automatically
223 - Error reporting and recovery in .ci files is greatly improved, providing more
224 precise line numbers and often column information
226 - Many data races occurring under shared-memory builds (smp, multicore) were
227 fixed, facilitating use of tools like ThreadSanitizer and Helgrind
231 - Further MPI standard compliance in AMPI allows users to build and run
232 Hypre-2.10.1 on AMPI with virtualization, migration, etc.
234 - Improved AMPI Fortran2003 PUP interface 'apup', similar to C++'s STL PUP
236 Platforms and Portability
238 - Compiling Charm++ now requires support for C++11 variadic templates. In GCC,
239 this became available with version 4.3, released in 2008
241 - New machine target for multicore Linux ARM7: multicore-linux-arm7
243 - Preliminary support for POWER8 processors, in preparation for the upcoming
244 Summit and Sierra supercomputers
246 - The charmrun process launcher is now much more robust in the face of slow
247 or rate-limited connections to compute nodes
249 - PXSHM now auto-detects the node size, so the '+nodesize' is no longer needed
251 - Out-of-tree builds are now supported
255 - CommLib has been removed.
257 - CmiBool has been dropped in favor of C++'s bool
260 ================================================================================
261 What's new in Charm++ 6.6.1
262 ================================================================================
264 Changes in this release are primarily bug fixes for 6.6.0. A concise list of
265 affected components follows:
269 - Reductions with syncFT
271 - mpicxx based MPI builds
273 - Increased support for macros in CI file
275 - GNI + RDMA related communication
277 - MPI_STATUSES_IGNORE support for AMPIF
279 - Restart on different node count with chkpt
281 - Immediate msgs on multicore builds
283 ================================================================================
284 What's new in Charm++ 6.6.0
285 ================================================================================
287 - Machine target files for Cray XC systems ('gni-crayxc') have been added
289 - Interoperability with MPI code using native communication interfaces on Blue
290 Gene Q (PAMI) and Cray XE/XK/XC (uGNI) systems, in addition to the universal
291 MPI communication interface
293 - Support for partitioned jobs on all machine types, including TCP/IP and IB
294 Verbs networks using 'netlrts' and 'verbs' machine layers
296 - A substantially improved version of our asynchronous library, CkIO, for
297 parallel output of large files
299 - Narrowing the circumstances in which the runtime system will send
300 overhead-inducing ReductionStarting messages
302 - A new fully distributed load balancing strategy, DistributedLB, that produces
303 high quality results with very low latency
305 - An API for applications to feed custom per-object data to specialized load
306 balancing strategies (e.g. physical simulation coordinates)
308 - SMP builds on LRTS-based machine layers (pamilrts, gni, mpi, netlrts, verbs)
309 support tracing messages through communication threads
311 - Thread affinity mapping with +pemap now supports Intel's Hyperthreading more
314 - After restarting from a checkpoint, thread affinity will use new
315 +pemap/+commap arguments
317 - Queue order randomization options were added to assist in debugging race
318 conditions in application and runtime code
320 - The full runtime code and associated libraries can now compile under the C11
321 and C++11/14 standards.
323 - Numerous bug fixes, performance enhancements, and smaller improvements in the
324 provided runtime facilities
327 * The long-unsupported FEM library has been deprecated in favor of ParFUM
328 * The CmiBool typedefs have been deleted, as C++ bool has long been universal
329 * Future versions of the runtime system and libraries will require some degree
330 of support for C++11 features from compilers
332 ================================================================================
333 What's new in Charm++ 6.5.0
334 ================================================================================
336 - The Charm++ manual has been thoroughly revised to improve its organization,
337 comprehensiveness, and clarity, with many additional example code snippets
340 - The runtime system now includes the 'Metabalancer', which can provide
341 substantial performance improvements for applications that exhibit dynamic
342 load imbalance. It provides two primary benefits. First, it automatically
343 optimizes the frequency of load balancer invocation, to avoid work stoppage
344 when it will provide too little benefit. Second, calls to AtSync() are made
345 less synchronous, to further reduce overhead when the load balancer doesn't
346 need to run. To activate the Metabalancer, pass the option +MetaLB at
347 runtime. To get the full benefits, calls to AtSync() should be made at every
348 iteration, rather than at some arbitrary longer interval as was previously
351 - Many feature additions and usability improvements have been made in the
352 interface translator that generates code from .ci files:
353 * Charmxi now provides much better error reports, including more accurate
354 line numbers and clearer reasons for failure, including some semantic
355 problems that would otherwise appear when compiling the C++ code or even at
357 * A new SDAG construct 'case' has been added that defines a disjunction over a
358 set of 'when' clauses: only one 'when' out of a set will ever be triggered.
359 * Entry method templates are now supported. An example program can be found
360 in tests/charm++/method_templates/.
361 * SDAG keyword "atomic" has been deprecated in favor of the newly supported
362 keyword "serial". The two are synonymous, but "atomic" is now provided only
363 for backward compatibility.
364 * It is no longer necessary to call __sdag_init() in chares that contain SDAG
365 code - the generated code does this automatically. The function is left as
366 a no-op for compatibility, but may be removed in a future version.
367 * Code generated from .ci files is now primarily in .def.h files, with only
368 declarations in .decl.h. This improves debugging, speeds compilation,
369 provides clearer compiler output, and enables more complete encapsulation,
370 especially in SDAG code.
371 * Mainchare constructors are expected to take CkArgMsg*, and always have
372 been. However, charmxi would allow declarations with no argument, and
373 assume the message. This is now deprecated, and generates a warning.
375 - Projections tracing has been extended and improved in various ways
376 * The trace module can generate a record of network topology of the nodes in
377 a run for certain platforms (including Cray), which Projections can
379 * If the gzip library (libz) is available when Charm++ is compiled, traces
380 are compressed by default.
381 * If traces were flushed as a results of filled buffers during the run, a
382 warning will be printed at exit to indicate that the user should be wary of
383 interference that may have resulted.
384 * In SMP builds, it is now possible to trace message progression through the
385 communication threads. This is disabled by default to avoid overhead and
386 potential misleading interpretation.
388 - Array elements can be block-mapped at the SMP node level instead of at the
389 per-PE level (option "+useNodeBlkMapping").
391 - AMPI can now privatize global and static variables using TLS. This is
392 supported in C and C++ with __thread markings on the variable declarations
393 and definitions, and in Fortran with a patched version of the gfortran
394 compiler. To activate this feature, append '-tls' to the '-thread' option's
395 argument when you link your AMPI program.
397 - Charm can now be built to only support message priorities of a specific data
398 type. This enables an optimized message queue within the the runtime
399 system. Typical applications with medium sized compute grains may not benefit
400 noticeably when switching to the new scheduler. However, this may permit
401 further optimizations in later releases.
403 The new queue is enabled by specifying the data type of the message
404 priorities while building charm using --with-prio-type=dtype. Here, dtype can
405 be one of char, short, int, long, float, double and bitvec. Specifying bitvec
406 will permit arbitrary-length bitvector priorities, and is the current default
407 mode of operation. However, we may change this in a future release.
409 - Converse now provides a complete set of wrappers for
410 fopen/fread/fwrite/fclose to handle EINTR, which is not uncommon on the
411 increasingly-popular Lustre. They are named CmiF{open,read,write,close}, and
412 are available from C and C++ code.
414 - The utility class 'CkEntryOptions' now permits method chaining for cleaner
415 usage. This applies to all its set methods (setPriority, setQueueing,
416 setGroupDepID). Example usage can be found in examples/charm++/prio/pgm.C.
418 - When creating groups or chare arrays that depend on the previous construction
419 of another such entity on the local PE, it is now possible to declare that
420 dependence to the runtime. Creation messages whose dependence is not yet
421 satisfied will be buffered until it is.
423 - For any given chare class Foo and entry method Bar, the supporting class's
424 member CkIndex_Foo::Bar() is used to lookup/specify the entry method
425 index. This release adds a newer API for such members where the argument is a
426 function pointer of the same signature as the entry method. Those new
427 functions are used like CkIndex_Foo::idx_Bar(&Foo::Bar). This permits entry
428 point index lookup without instantiating temporary variables just to feed the
429 CkIndex_Foo::Bar() methods. In cases where Foo::Bar is overloaded, &Foo::Bar
430 must be cast to the desired type to disambiguate it.
432 - CkReduction::reducerType now have PUP methods defined; and can hence be
433 passed as parameter-marshalled arguments to entry methods.
435 - The runtime option +stacksize for controlling the allocation of user-level
436 threads' stacks now accepts shorthanded annotation such as 1M.
438 - The -optimize flag to the charmc compiler wrapper now passes more aggressive
439 options to the various underlying compilers than the previous '-O'.
441 - The charmc compiler wrapper now provides a flag -use-new-std to enable
442 support for C11 and C++11 where available. To use this in application code,
443 the runtime system must have been built with that flag as well.
445 - When using, CmiMemoryUsage(), the runtime can be instructed not to use the
446 underlying mallinfo() library call, which can be inaccurate in settings where
447 usage exceeds INT_MAX. This is accomplished by setting the environment
448 variable "MEMORYUSAGE_NO_MALLINFO".
450 - Experimental Features
451 * Initial implementation of a fast message-logging protocol. Use option
452 'mlogft' to build it.
453 * Message compression support for persistent message on Gemini machine layer.
454 * Node-level inter-PE loop/task parallelization is now supported through
456 * New temperature/CPU frequency aware load balancer
457 * Support interoperation of Charm++ and native MPI code through dynamically
458 switching control between the two
459 * API in centralized load balancers to get and set PE speed
460 * A new scheme for optimization of double in-memory checkpoint/restart.
461 * Message combining library for improved fine-grained communication
463 * Support for partitioning of allocated nodes into subsets that run
464 independent Charm++ instances but can interact with each other.
466 Platform-Specific Changes
467 -------------------------
470 * The gemini_gni network layer has been heavily tuned and optimized,
471 providing substantial improvements in performance, scalability, and
473 * The gemini_gni-crayxe machine layer supports a 'hugepages' option at build
474 time, rather than requiring manual configuration file editing.
475 * Persistent message optimizations can be used to reduce latency and
477 * Experimental support for 'urgent' sends, which are sent ahead of any other
478 outgoing messages queued for transmission.
480 - IBM Blue Gene Q: Experimental machine-layer support for the native PAMI
481 interface and MPI, with and without SMP support. This supports many new
482 systems, including LLNL's Sequoia, ALCF's Mira, and FZ Juelich's Juqueen.
484 There are three network-layer implementations for these systems: 'mpi',
485 'pami', and 'pamilrts'. The 'mpi' layer is stable, but its performance and
486 scalability suffers from the additional overhead of using MPI rather than
487 driving the interconnect directly. The 'pami' layer is well tested for NAMD,
488 but has shown instability for other applications. It is likely to be replaced
489 by the 'pamilrts' layer, which is more generally stable and seems to provide
490 the same performance, in the next release.
492 In addition to the common 'smp' option to build the runtime system with
493 shared memory support, there is an 'async' option which sometimes provides
494 better performance on SMP builds. This option passes tests on 'pamilrts', but
495 is still experimental.
497 Note: Applications that have large number of messages may crash in default
498 setup due to overflow in the low-level FIFOs. Environment variables
499 MUSPI_INJFIFOSIZE and PAMI_RGETINJFIFOSIZE can be set to avoid application
500 failures due to large number of small and large messages respectively. The
501 default value of these variable is 65536 which is sufficient for 1000
504 - Infiniband Verbs: Better support for more flavors of ibverbs libraries
507 * Experimental rendezvous protocol for better performance above some MPI
509 * Some tuning parameters ("+dynCapSend" and "+dynCapRecv") are now
510 configurable at job launch, rather than Charm++ compilation.
512 - PGI C++: Disable automatic 'using namespace std;'
514 - Charm++ now supports ARM, both non-smp and smp.
516 - Mac OS X: Compilation options to build and link correctly on newer versions
519 ================================================================================
520 What's new in Charm++ 6.4.0
521 ================================================================================
523 --------------------------------------------------------------------------------
525 --------------------------------------------------------------------------------
527 - Cray XE and XK systems using the Gemini network via either MPI
528 (mpi-crayxe) or the native uGNI (gemini_gni-crayxe)
530 - IBM Blue Gene Q, using MPI (mpi-bluegeneq) or PAMI (pami-bluegeneq)
532 - Clang, Cray, and Fujitsu compilers
534 - MPI-based machine layers can now run on >64k PEs
536 --------------------------------------------------------------------------------
538 --------------------------------------------------------------------------------
540 - Added a new [reductiontarget] attribute to enable
541 parameter-marshaled recipients of reduction messages
543 - Enabled pipelining of large messages in CkMulticast by default
545 - New load balancers added:
548 * Scotch graph partitioning based: ScotchLB and Refine and Topo variants
551 - Load balancing improvements:
553 * Allow reduced load database size using floats instead of doubles
554 * Improved hierarchical balancer
555 * Periodic balancing adapts its interval dynamically
556 * User code can request a callback when migration is complete
557 * More balancers properly consider object migratability and PE
558 availability and speed
559 * Instrumentation records multicasts
561 - Chare arrays support options that can enable some optimizations
563 - New 'completion detection' library for parallel process termination
564 detection, when the need for modularity excludes full quiescence
567 - New 'mesh streamer' library for fine-grain many-to-many collectives,
568 handling message bundling and network topology
570 - Memory pooling allocator performance and resource usage improved
573 - AMPI: More routines support MPI_IN_PLACE, and those that don't check
576 ================================================================================
577 What's new in Charm++ 6.2.1 (since 6.2.0)
578 ================================================================================
580 --------------------------------------------------------------------------------
581 New Supported Platforms:
582 --------------------------------------------------------------------------------
584 POWER7 with LAPI on Linux
586 Infiniband on PowerPC
588 --------------------------------------------------------------------------------
590 --------------------------------------------------------------------------------
592 - Better support for multicasts on groups
593 - Topology information gathering has been optimized
594 - Converse (seed) load balancers have many new optimizations applied
595 - CPU affinity can be set more easily using +pemap and +commap options
596 instead of the older +coremap
597 - HybridLB (hierarchical balancing for very large core-count systems)
598 has been substantially improved
599 - Load balancing infrastructure has further optimizations and bug fixes
600 - Object mappings can be read from a file, to allow offline
601 topology-aware placement
602 - Projections logs can be spread across multiple directories, speeding
603 up output when dealing with thousands of cores (+trace-subdirs N
604 will divide log files evenly among N subdirectories of the trace
605 root, named PROGNAME.projdir.K)
606 - AMPI now implements MPI_Issend
607 - AMPI's MPI_Alltoall uses a flooding algorithm more agressively,
608 versus pairwise exchange
609 - Virtualized ARMCI support has been extended to cover the functions
612 --------------------------------------------------------------------------------
613 Architecture-specific changes
614 --------------------------------------------------------------------------------
616 - LAPI SMP has many new optimizations applied
618 - Net builds support the use of clusters' mpiexec systems for job
619 launch, via the ++mpiexec option to charmrun
621 ================================================================================
622 What's new in Charm++ 6.2.0 (since 6.1)
623 ================================================================================
625 --------------------------------------------------------------------------------
626 New Supported Platforms:
627 --------------------------------------------------------------------------------
629 64-bit MIPS, such as SiCortex, using mpi-linux-mips64
631 Windows HPC cluster, using mpi-win32/mpi-win64
633 Mac OSX 10.6, Snow Leopard (32-bit and 64-bit).
635 --------------------------------------------------------------------------------
637 --------------------------------------------------------------------------------
640 - Smarter build/configure scripts
641 - A new interface for model-based load balancing
642 - new CPU topology API
643 - a general implementation of CmiMemoryUsage()
644 - Bug fix: Quiescence detection (QD) works with immediate messages
645 - New reduction functions implemented in Converse
646 - CCS (Converse Client-Server) can deliver message to more than one processor
647 - Added a memory-aware adaptive scheduler, which can be optionally
649 - Added preliminary support for automatic message prioritization
650 (disabled by default)
653 - Cross-array and cross-group sections
654 - Structured Dagger (SDAG): Support templated arguments properly
655 - Plain chares support checkpoint/restart (both in-memory and disk-based)
656 - Conditional packing of messages and parameters in SMP scenario
657 - Changes to the CkArrayIndex class hierarchy
658 -- sizeof() all CkArrayIndex* classes is now the same
659 -- Codes using custom array indices have to use placement-new to construct
660 their custom index. Refer example code: examples/charm++/hello/fancyarray/
661 -- *** Backward Incompatibility ***
662 CkArrayIndex[4D/5D/6D]::index are now of type int (instead of short)
663 However the data is stored as shorts. Access by casting
664 CkArrayIndexND::data() appropriately
665 -- *** Deprecated ***
666 The direct use of public data member
667 CkArrayIndexND::index (N=1..6) is deprecated. We reserve the right to
668 change/remove this variable in future releases of Charm++.
669 Instead, please access the indices via member function:
670 int CkArrayIndexND::data()
673 - Compilers renamed to avoid collision with host MPI (ampicc, ampiCC,
675 - Improved MPI standard conformance, and documentation of non-conformance
676 * Bug fixes in: MPI_Ssend, MPI_Cart_shift, MPI_Get_count
677 * Support MPI_IN_PLACE in MPI_(All)Reduce
678 * Define various missing constants
679 - Return the received message's tag in response to a non-blocking
680 wildcard receive, to support SuperLU
681 - Improved tracing for BigSim
683 Multiphase Shared Arrays (MSA)
684 - Typed handles to enforce phases
685 - Split-phase synchronization to enable message-driven execution
689 - Automatic tracing of API calls for simulation and analysis
692 - Wider support for architectures other than net- (in particular MPI layers)
693 - Improved support for large scale debugging (better scalability)
694 - Enhanced record/replay stability to handle various events, and to
695 signal unexpected messages
696 - New detailed record/replay: The full content of messages can be
697 recorded, and a single processor can be re-executed outside of the
701 - Tracing of nested entry methods
703 Automatic Performance Tuning
704 - Created an automatic tuning framework [still for experimental use only]
707 - Network-topology / node aware spanning trees used internally for and
708 lower bytes on the network and improved performance in multicasts and
709 reductions delegated to this library
712 - Improved OneTimeMulticastStrategy classes
715 - Out-of-core support, with prefetching capability
716 - Detailed tracing of MPI calls
717 - Detailed record/replay support at emulation time, capable of
718 replaying any emulated processor after obtained recorded logs.
720 --------------------------------------------------------------------------------
721 Architecture-specific changes
722 --------------------------------------------------------------------------------
725 - Can run jobs with more than 1024 PEs
728 - New charmrun option ++no-va-randomization to disable address space
729 randomization (ASLR). This is most useful for running AMPI with
733 - Default to using ampicxx instead of mpiCC
736 - The +p option now has the same semantics as in other smp builds
739 - Support for VSX in SIMD abstraction API
742 - Compilers and options have been updated to the latest ones
745 - Added routines for measuring performance counters on BG/P.
746 - Updated to support latest DCMF driver version. On ANL's Intrepid, you may
747 need to set BGP_INSTALL=/bgsys/drivers/V1R4M1_460_2009-091110P/ppc in your
748 environment. This is the default on ANL's Surveyor.
751 - cputopology information is now available on XT3/4/5
754 - Bug fix: plug memory leaks that caused failures in long runs
755 - Optimized to reduce startup delays
758 - Support for SMP (experimental)
761 ================================================================================
762 Note that changes from 5.9, 6.0, and 6.1 are not documented here. A partial list
763 can be found on the charm download page, or by reading through version control
766 ================================================================================
767 What's New since Charm++ 5.4 release 1
768 ================================================================================
770 --------------------------------------------------------------------------------
771 New Supported Platforms:
772 --------------------------------------------------------------------------------
773 1. Charm++ ported to IA64 Itanium running Win2K and Linux, Charm++ also support
774 Intel C/C++ compilers;
776 2. Charm++ ported to Power Macintosh powerpc running Darwin;
778 3. Charm++ ported to Myrinet networking with GM API;
780 --------------------------------------------------------------------------------
781 Summary of New Features:
782 --------------------------------------------------------------------------------
784 Structured Dagger is a coordination language built on top of CHARM++.
785 Structured Dagger allows easy expression of dependences among messages and
786 computations and also among computations within the same object using
787 when-blocks and various structured constructs.
789 2. Entry functions support parameter marshalling
790 Now you can declare and invoke remote entry functions using parameter
791 marshalling instead of defining messages.
793 3. Easier running - standalone mode
794 For net-* version running locally, you can now run Charm programs without
795 charmrun. Running a node program directly from command line is now the
796 same as "charmrun +p1 <program>"; for SMP version, you can also specify
797 multiple (local) processors, as in "program +p2".
800 --------------------------------------------------------------------------------
802 --------------------------------------------------------------------------------
803 1. "build" changed for compilation of Charm++
804 To build Charm++ from scratch, we now take additional command line options
805 to compile with addon features and using different compilers other than gcc.
806 For example, to build Linux IA64 with Myrinet support, type command:
807 ./build net-linux-ia64 gm
810 ******* Old Change histories *******
813 ================================================================================
814 What's New in Charm++ 5.4 release 1 since 5.0
815 ================================================================================
817 --------------------------------------------------------------------------------
818 New Supported Platforms:
819 --------------------------------------------------------------------------------
821 1. Win9x/2000/NT: with Visual C++ or Cygwin gcc/g++, you can compile and run
822 Charm++ programs on all Win32 platforms.
824 2. Scyld Beowulf: Charm++ has been ported to the Linux-based Scyld Beowulf
825 operating system. For more information on Scyld, see <http://www.scyld.com>
827 3. MPI with VMI: Charm++ has been ported to NCSA's Virtual Machine Interface,
828 which is an efficient messaging library for heterogeneous cluster
832 --------------------------------------------------------------------------------
833 Summary of New Features:
834 --------------------------------------------------------------------------------
835 1. Dynamic Load balancing:
836 Chare migration is supported in the new release. Migration-based dynamic
837 load balancing framework with various load balancing strategies library has
841 Charm++ array is supported. You can now create an array of Chare objects
842 and use array index to refer the Charm++ array elements. A reduction
843 library on top of Chare array has been implemented and included.
846 Projections, a Java application for Charm++ program performance analysis and
847 visualization, has been included and distributed in the new release. Two
848 trace modes are available: trace-projections and trace-summary. Trace-summary
849 is a light-weight trace library compared to trace-projections.
852 AMPI is a load-balancing based library for porting legacy MPI applications
853 to Charm++. With few changes in the original MPI code to AMPI, the new
854 legacy MPI application on Charm++ will gain from Charm++'s adptive
855 load balancing ability.
858 "Charmrun" is now available on all platforms, with a uniform command line
859 syntax. You can forget the difference between net-* versions and MPI versions,
860 and run charm++ application with this same charmrun command syntax.
861 ++local option is added in charmrun for net-* version, it provides
862 simple local use of Charm and no longer require the ability to
863 "rsh localhost" or a nodelist file in order to run charm only on the local
864 machine. This is especially attractive when you run Charm++ on Windows.
867 Many new libraries have been added in this release. They include:
868 1) master-slave library: for writing manager-worker paradigm programs.
869 2) receiver library: provide asynchronous communication mode for chare array.
870 3) f90charm: provides Fortran90 bindings for Charm++ Array.
871 4) BlueGene: a Charm++/Converse emulator for IBM proposed Blue Gene.
873 --------------------------------------------------------------------------------
875 --------------------------------------------------------------------------------
876 1. message declaration syntax in .ci file:
877 The message declaration syntax for packed/varsize messages has been changed.
878 The packed/varsize keywords are eliminated, and you can specify the actual
879 actual varsize arrays in the interface file and have the translator generate
880 alloc, pack and unpack.
883 Here is the detailed list of Changes:
885 --------------------------------------------------------------------------------
887 --------------------------------------------------------------------------------
889 10/06/1999 rbrunner Added migration-based dynamic load balancing
891 11/15/1999 olawlor Added reduction support foe Charm++ arrays
892 02/06/2000 milind Added AMPI, an implementation of MPI with
893 dynamic load balancing
894 02/18/2000 paranjpy New platforms supported: net-win32, and net-win32-smp
895 04/04/2000 olawlor Added arbitrarily indexed Charm++ arrays.
896 Also, added translator support for new arrays.
897 04/15/2000 olawlor Added "puppers" for packing and unpacking
899 06/14/2000 milind Added the threaded FEM framework.
901 --------------------------------------------------------------------------------
903 --------------------------------------------------------------------------------
905 10/09/1999 rbrunner Added packlib, a library for C and C++ to
906 pack-unpack data to/from Charm++ messages.
907 10/13/1999 gzheng New LB strategy: RefineLB
908 10/13/1999 paranjpy New LB Strategy: Heap
909 10/14/1999 milind New LB Strategy: Metis
910 10/19/1999 olawlor New test program for testing LB strategies.
911 10/21/1999 gzheng New trace mode: trace-summary
912 10/28/1999 milind New supported platform: net-sol-x86
913 10/29/1999 milind Added runtime checks for ChareID assignment.
914 11/10/1999 rbrunner Added Neighborhood base strategy for LB
916 11/15/1999 olawlor conv-host now reads in a startup file
918 11/15/1999 olawlor New test program for testing array reductions.
919 11/16/1999 rbrunner Added processor-speed checking functions to
921 11/19/1999 milind Mapped SIGUSR to a Ccd condtion handler
922 11/22/1999 rbrunner New LB strategy: WSLB
923 11/29/1999 ruiliu Modified Metis LB strategy to deal with
924 different processor speeds
925 12/16/1999 rbrunner New LB strategy: GreedyRef
926 12/16/1999 rbrunner New LB strategy: RandRef
927 12/21/1999 skumar2 New LB strategy: CommLB
928 01/03/2000 rbrunner New LB strategy: RecBisectBfLB
929 01/08/2000 skumar2 New LB strategy: Comm1LB, with varying processor
931 01/18/2000 milind Modified SM library syntax, and added a test
933 01/19/2000 gzheng Added irecv, a library to simplify conversion
934 of message-passing programs to Charm++
935 02/20/2000 olawlor Added preliminary broadcast support to Charm++
937 02/23/2000 paranjpy Added converse-level quiescence detection
938 03/02/2000 milind Added ++server-port option to pre-specify
940 03/10/2000 wilmarth Random seed-based load balancer now uses
941 bit-vector for active PEs.
942 03/21/2000 gzheng Added support for marking user-defined events
944 03/28/2000 wilmarth Added CMK_TRUECRASH. Very helpful for
945 post-mortem debugging of Charm++ programs on
947 03/31/2000 jdesouza Added Fortran90 support to the Charm++
948 interface translator.
949 03/09/2000 milind Added support for -LANG and -rpath options
950 in charmc for Origin2000.
951 04/28/2000 milind Added prioritized converse threads.
952 05/01/2000 milind Added test programs for TeMPO, AMPI and irecv.
953 05/04/2000 milind New supported platform: mpi-sp.
954 05/04/2000 gzheng Added irecv pingpong program.
955 05/17/2000 olawlor Each chare, group and array element now has to
956 have migration constructor.
957 05/24/2000 milind Added Jacobi3D programs for irecv and AMPI both.
958 05/24/2000 milind Made migratable an optional attribute of
959 chares, groups, and nodegroups.
960 Arrays are by default migratable.
961 05/29/2000 paranjpy Added pup methods to arrays, reductions etc
963 06/13/2000 milind Made CtvInitialize idempotent. That is, it
964 can be called by any number of threads now,
965 only the first one will actually do
967 06/20/2000 milind Added a simple test program for the FEM
969 07/06/2000 milind Imported Metis 4.0 sources in the CVS tree.
970 Also added code to make metis libraries and
971 executables to Makefile.
972 07/07/2000 milind Added more meaningfull error messages using
973 perror in addition to a cryptic error codes in
975 07/10/2000 milind fem and femf are now recognized as "languages"
977 07/10/2000 saboo Added the derived datatypes library.
978 07/13/2000 milind Added +idle_timeout functionality. It takes a
979 commandline parameter denoting milliseconds of
980 maximum consecutive idle time allowed per
982 07/14/2000 milind Added group multicast. Added
983 CkSendMsgBranchMulti, CldEnqueueMulti, and
984 translator changes to support it.
985 07/14/2000 milind SUPER_INSTALL now takes "-*" arguments prior
986 to the target, that will be passed to make as
987 "makeflags". This makes it easy to suppress
988 make's output of commands etc (with the -s
989 flag). As a result of this, several Makefiles
991 07/18/2000 milind Added support for using "dbx" on suns as
993 07/19/2000 milind Added ability to tracemode projections which
994 produces binary trace files. Use flag
995 +binary-trace on the command line.
996 07/26/2000 milind Separated AMPI from TeMPO.
997 07/28/2000 milind Added test programs to test reduce, alltoall
998 and allreduce functionality of AMPI.
999 08/02/2000 milind Added an option to let the user specify which
1000 "xterm" to use. For example, on some systems
1001 (CDE), only dtterm is installed. So, by
1002 putting ++xterm dtterm on the conv-host
1003 commandline, one can use dtterm when ++in-xterm
1004 option is specified on conv-host commandline.
1005 08/14/2000 milind FEM Framework: Added capabilities to handle
1006 esoteric meshes to standalone offline programs.
1007 Makefile now produces gmap and fgmap programs,
1008 which are used for this purpose. They convert
1009 the mesh to a graph before partitioning it
1011 08/24/2000 milind Added the 2D crack propagation program as a
1012 test program for FEM framework.
1013 08/25/2000 milind Initial implementation of isomalloc-based
1014 threads. This implementation uses a fixed
1015 stack size for all threads (can be set at
1017 08/26/2000 milind Added a macro CtvAccessOther that lets you
1018 get/set a Ctv variable of any thread. It
1019 should be invoked as CtvAccessOther(thread,
1020 varname); Added CthGetData function to each of
1021 the threads implementation. This function is
1022 used in the CtvAccessOther macro.
1023 08/27/2000 milind FEM Framework: Separated mesh to graph
1024 conversion capability into a separate program.
1025 This way, the generated graph can be partitioned
1027 09/04/2000 milind Added the class static readonly variables to
1029 09/05/2000 milind FEM Framework: A very fast O(n) algorithm for
1030 mesh2graph , uses more memory, but the tradeoff
1031 was worth it. Coded by Karthik Mahesh, minor
1032 optimizations by Milind.
1033 09/05/2000 milind Added a barebones charm kernel scheduling
1034 overhead measurement program.
1035 09/15/2000 milind Added pup support for AMPI and FEM framework.
1036 09/20/2000 olawlor Added capability to have an array of base type
1037 where individual element could be of derived
1039 10/03/2000 gzheng New supported platform: net-linux-axp
1040 10/05/2000 skumar2 Added program littleMD to the test suite.
1041 10/07/2000 skumar2 New job scheduler (Faucets projects).
1042 10/15/2000 milind Improved support for Fortran90 in charmc.
1043 11/04/2000 jdesouza Made the Faucets scheduler multi-threaded.
1044 11/05/2000 olawlor FEM Framework: supports multiple element types,
1045 mesh re-assembly, etc.
1046 11/15/2000 gzheng New platform support: net-cygwin
1047 11/18/2000 gzheng conv-host no longer needs /bin/csh to start
1049 CMK_CONV_HOST_CSH_UNAVAILABLE to 1 to use
1051 11/25/2000 milind Finished experimental implementation of
1052 converse-threads based on co-operative pthreads.
1053 11/25/2000 milind Added a benchmark suite of all pingpongs in
1055 11/28/2000 milind Removed deletion of _idx at the end of every
1056 send or doneInserting call. Instead now it is
1057 in the destructor of the proxy. This allows us
1058 to cache proxies, when proxy creation becomes
1060 11/28/2000 olawlor Added "seek blocks" to puppers. This should
1061 allow out-of-order pup'ing without the ugliness
1062 of getBuf; and in a way that works with all
1064 11/29/2000 olawlor Simplified and regularized command-line-argument
1066 11/29/2000 milind AMPI: Added multiple-communicators capability.
1067 12/05/2000 gzheng Now /bin/sh is default shell to fork node
1068 program on remote machines.
1069 12/13/2000 olawlor Added charmrun wrapper for poe on mpi-sp.
1070 12/14/2000 milind Added bluegene emulator sources and test
1071 programs. Added "bluegene" as a language known
1072 to charmc. Makefile now has a target called
1073 bluegene. Added preliminary bluegene
1074 documentation. (copied from Arun's webpage.)
1075 12/15/2000 gzheng f90charm addition to Makefile and charmc. Also,
1076 added fixed size arrays support to f90charm. A
1077 test program f90charm/hello is checked in.
1078 12/17/2000 milind Added rtest test program. Contributed by jim to
1079 test Converse message transmission.
1080 12/20/2000 olawlor Added charmconfig script. Enables automatic
1081 determination of C++ compiler properties,
1082 replacing the verbose and error-prone
1083 conv-mach.h entries for CMK_BOOL,
1084 CMK_STL_USE_DOT_H, CMK_CPP_CAST_OK, ...
1085 12/20/2000 olawlor Charm++ Arrays optimizations: Key and object
1086 now variable-length fields, instead of pointers.
1087 This extra flexibility lets us save many
1088 dynamic allocations in the array framework.
1089 12/20/2000 olawlor Added PUP::able support-- dynamic type
1090 identification, allocation, and deletion.
1091 Allows you to write: p(objPtr); and
1092 objPointer will be properly identified,
1093 allocated, packed, and deallocated (depending
1094 on the PUP::er). Requires you to register any
1095 such classes with DECLARE_PUPable and
1097 12/20/2000 olawlor Arrays optimizations: Made CkArrayIndex
1098 fixed-size. This significantly improves
1099 messaging speed (7 us instead of 10 us
1100 roundtrip). Move spring cleaning check into a
1101 CcdCallFnAfter, which gains more speed (down to
1103 12/20/2000 olawlor More optimizations: Minor speed tweaks--
1104 conv-ccs.c uses hashtable for handler lookup;
1105 conv-conds skips timer test until needed;
1106 convcore.c scheduler loop optmizations (no
1107 superfluous EndIdle calls); threads.c
1108 CMK_OPTIMIZE-> no mprotect.
1109 12/20/2000 olawlor More Optimizations: Minor speed tweaks-- ck.C
1110 groups cldEnqueue skip; init.h defines
1111 CkLocalBranch inline; and supporting changes.
1112 12/22/2000 gzheng IA64 support for Converse user level threads.
1113 01/02/2001 olawlor CCS: Minor update-- enabled CcsProbe, cleaned
1114 up superflous debug messages in server, added
1115 Java interface (originally written for
1117 01/09/2001 gzheng charmconfig converted to autoconf style, need
1118 to change configure.in and conv-autoconfig.h.in,
1119 and run autoconf to get configure and copy to
1120 charmconfig. added fortran subroutine name
1121 test and get libpthread.a
1122 01/10/2001 milind Added telnet method of getting libpthread.a
1123 from charm webserver.
1124 01/11/2001 olawlor Moved projections files here from
1125 CVSROOT/projections-java. Added fast Java
1126 versions of the .log file input routines in
1127 LogReader, LogLoader, LogAnalyzer, and
1128 UsageCalc. Added "U.java" user interface
1129 utility file, allowing times to be input in
1130 seconds, milliseconds, or microseconds,
1131 instead of just microseconds.
1132 01/15/2001 gzheng add +trace-root to specify the directory to
1133 put log files in. this is need in Scyld cluster
1134 where there is no NFS mounting and no i/o
1135 access to home directory sharing on nodes.
1136 01/15/2001 milind Made AMPI into a f90 module instead of
1137 'ampif.h' inclusion. AMPI f90 bindings are
1138 now more inclusive. Fixed argc,argv handling
1139 bugs in ArgsInfo message. Fixed a bug in pup
1140 that caused thread not to be sized, but was
1141 packed nevertheless. Moved irecv to waitall
1142 instead of at in ampi_start. Made
1143 AMPI_COMM_WORLD to be 0, because it clashed
1144 with wildcard(-1). AMPI_COMM_UNIVERSE is now
1145 handled properly in the AMPI module.
1146 C/C++ data members are NOT visible to
1148 01/18/2001 gzheng New supported platform: net-linux-scyld
1149 01/20/2001 olawlor Moved array index field from CMessage_* to the
1150 Ck envelope itself. This is the right thing
1151 to do, because any message may be sent to/from
1152 an array element. To reduce the wasted space
1153 in a message, a union is used to overlay the
1154 fields for the various possible message types.
1155 01/29/2001 olawlor Freed charmrun on net-* version from using
1156 remote shell to fork off processes. One can now
1157 use a daemon provided in the distribution.
1158 02/07/2001 olawlor Added debugging support to puppers.
1159 02/13/2000 gzheng Added ++local option to charmrun to start node
1160 program locally without any daemon; fix the
1161 hang program if you type wrong pgm name in
1162 scyld version, and redirect all output to
1163 /dev/null, otherwise all node program can send
1164 its output to console in scyld. Also implemented ++local in net-win32 version.
1165 02/26/2000 milind Changed the varsize syntax. Now one can specify
1166 actual varsize arrays in the interface file
1167 and have the translator generate alloc, pack
1170 --------------------------------------------------------------------------------
1172 --------------------------------------------------------------------------------
1174 10/29/1999 milind Replaced jmemcpy by memcpy in net versions, as
1175 it was causing a bit to flip (bug reported
1177 10/29/1999 milind Fixed multiline macros in all header files.
1178 02/05/2000 milind Fixed linking errors by getting the order of
1179 libraries right from the charmc command-line.
1180 02/18/2000 paranjpy Fixed Charm++ initialization bug on SMPs.
1181 02/21/2000 milind Fixed a context-switching bug in mipspro version
1183 02/25/2000 milind Charm++ interface translator was segfaulting
1184 on interface file errors. Fixed that. Also,
1185 added linenumbers to error messages.
1186 03/02/2000 milind Made CCS work on SMPs.
1187 03/07/2000 milind Made ConverseInit consistent with the manual on
1189 04/18/2000 milind Fixed a bug in CkWaitFuture, which was caching
1190 a variable locally, while it was changed by
1192 05/04/2000 paranjpy Fixed argv deletion bug on net-win32-smp.
1193 06/08/2000 milind sp3 version: changed optimization flags, which
1194 where power2 processor-specific.
1195 06/20/2000 milind mpi-* versions: Fixed ConverseExit since it was
1196 not obeying the following statement in the MPI
1197 standard: The user must ensure that all pending
1198 communications involving a process completes
1199 before the process calls MPI_FINALIZE.
1200 07/05/2000 milind Fixed a nasty bug in charmc in the -cp option.
1201 It used to append the name provided to -o flag
1202 to the directory provided to the -cp flag.
1203 Thus, -o ../pgm -cp ../bin options meant that
1204 the pgm would be copied to ../bin/.., which is
1205 not the expected behavior. This fix correctly
1206 copies pgm to ../bin.
1207 07/07/2000 milind Removed variable arg_myhome, as it was not
1208 being used anywhere, and also, setting it was
1209 causing problems of env var HOME was not set.
1210 07/27/2000 milind thishandle for the arrayelement was not being
1211 correctly set. Bug was reported by Neelam.
1212 08/26/2000 milind Origin2000: Changed the page alignment to
1213 reflect the mmap alignment. The mmap man page
1214 specifically states that it is not the same as
1216 09/02/2000 milind Fixed a bug in code generated for threaded
1217 (void) entry methods of array elements. The
1218 dummy message that is passed to that method in
1219 a thread has to be deleted before calling the
1220 object method, because upon object method's
1221 return, the thread might have migrated.
1222 09/03/2000 olawlor Minor fix-fixes: 1.) Change to LBObjid hash
1223 function would fail for >4-int object indices.
1224 Replaced with proper function, which also
1225 preserves the 1-int case. 2.) Array element
1226 sends must go via the message queue to prevent
1227 stack build-up for deep single-processor call
1228 chains. These might happen, e.g., in a driver
1229 element calling itself for the main time loop.
1230 Messages are now properly noted as sent, then
1231 wait through the queue for delivery. This
1232 entailed minor reorganization of the message
1234 09/21/2000 olawlor Tiny SMP thread fix-- registrations of a
1235 thread-private variable now reserve space on
1236 calls after the first. This wastes space for
1237 multiple CthInitialize's-- it's a quick hack to
1238 get threads working again on SMP versions.
1239 10/16/2000 olawlor A few CCS fixes: -Added split-phase reply
1240 (delay reply indefinitely) -Cleaned up error
1241 handling -Pass user data as "void *" instead of
1243 11/03/2000 wilmarth Removed 0 size array allocation in Charm++
1244 quiescence detection.
1245 11/20/2000 gzheng Rewrote part of Fiber thread, including a bug
1246 fix for a the non thread-safe function, and a
1247 different fiber free strategy.
1248 11/29/2000 gzheng The LB init procedure tried to allocate
1249 65536*160 as initial size, which is 10M memory
1250 for communication table, which is too big.
1251 Cut it down to roughly 1M, and it can expand
1253 12/05/2000 gzheng In many cases, conv-host exits without print
1254 out the error message from remote shell. try
1255 to fix it by calling sync to flush the pipe
1257 12/10/2000 milind net-linux: Made static linking the default
1258 option because dynamic linking runtime causes
1259 isomalloc threads to crash.
1260 12/18/2000 milind Increased portability of isomalloc threads by
1261 removing dependence on alloca.
1262 12/28/2000 milind Fixed ctrl-getone abort bug on SMP.
1263 12/28/2000 milind Made _groupTable a pointer on which a
1264 constructor is explicitly called. Since it
1265 was a Cpv variable, its constructor was not
1266 called by default in case of an SMP version.
1267 12/29/2000 olawlor Prevent infinite copy constructor recursion on
1269 01/10/2001 olawlor Added "explicit" keyword to remove ambiguity
1270 for KCC, which was confused by the private
1271 PUP::er(int) "cast" constructor and the operator
1272 |(PUP::er &p,T &t) into rejecting all operator|
1273 (int,int) as ambiguous.
1274 2001/01/17 gzheng fix the charmconfig bug on paragon-red: the
1275 failure testing of fortran won't stop the
1277 01/20/2001 olawlor Arrays reduction: Fixed bug-- reduction may end
1278 because all contributors migrate away.
1279 01/29/2001 olawlor Fix heap-corrupting bug-- call ->init() on
1280 nodeGroupTable, which sets the "pending"
1281 message queue to NULL. This prevents a nasty
1282 delete-unitialized-data bug later on. Also
1283 delayed queue creation until messages actually
1286 --------------------------------------------------------------------------------
1287 Documentation Changes:
1288 --------------------------------------------------------------------------------
1290 01/31/2000 milind Installation manual: Fixed bugs pointed out by
1292 02/28/2000 wilmarth Added a new look Charm++ manual.
1293 06/20/2000 milind Added pdflatex support to generate PDF versions
1294 of manuals from LaTeX sources.
1295 12/05/2000 milind Added Orion's FEM manual. Converted from HTML.
1296 12/10/2000 milind Added pplmanual.sty for all manuals.
1297 12/17/2000 milind Added master-slave library documentation to
1299 12/21/2000 saboo Added DDT documentation.
1300 01/02/2001 olawlor Updated for new CCS version.
1302 --------------------------------------------------------------------------------
1304 --------------------------------------------------------------------------------
1306 10/24/1999 olawlor charmc is changed to Bourne shell script
1307 instead of csh. All conv-mach.csh are
1308 replaced by conv-mach.sh.
1309 10/25/1999 olawlor SUPER_INSTALL is converted to use bourne shell.
1310 10/28/1999 milind All Makefiles now take OPTS commandline
1312 01/16/2000 olawlor Simplified Charm++ interface translator.
1313 02/23/2000 ruiliu Changed rand() calls from all over the codes
1314 to the new Converse random number generator.
1315 02/26/2000 milind Simplified the converse scheduler loop by
1316 combining the maxmsgs and poll modes.
1317 08/31/2000 milind Imported system documentation into the CVS tree.
1318 Also added super_install target for docs with
1319 necessary Makefile modifications.
1320 09/08/2000 olawlor Made soft links use relative pathnames instead
1321 of absolute. This lets you move a charm++
1322 installation without having to recompile
1324 09/11/2000 olawlor Grouped commonly needed code in the new util
1325 directory. Also, added pup_c a C wrapper for
1327 09/11/2000 olawlor Slightly reorganized header structure. Now no
1328 headers should need to be listed twice (once in
1329 ALLHEADERS, again in CKHEADERS). Now headers
1330 are soft-linked instead of copied. This makes
1331 development much easier. Added support for the
1332 new Common/util directory.
1333 09/21/2000 olawlor Major reorganization of net-* codes. Now all
1334 the TCP socket routines are in separate files.
1335 Also combined windoes NT code with unix codes.
1336 09/21/2000 olawlor Major rewrite of CCS-- underlying protocol is
1337 now binary (send/recv binary data everywhere);
1338 conv-host forwards requests to nodes; and
1339 source has been significantly re-arranged.
1340 (especially if NODE_0_IS_CONVHOST).
1341 11/22/2000 milind Removed IDL translator from distribution.
1342 12/01/2000 olawlor Renamed conv-host charmrun; added test for
1343 script conv-host. Also added charmrun for most
1345 12/17/2000 milind Moved List related data structures into
1346 cklists.h in util. Removed most of the redundant
1347 list implementations.
1348 12/20/2000 gzheng SUPER_INSTALL: format the output of list of
1349 versions and make the help page fit into one
1351 12/24/2000 milind Added test-{charm,converse,ampi,fem} targets to
1353 12/28/2000 milind net-sol-smp now uses pthreads.
1354 01/29/2001 olawlor Merged windowsNT and unix build procedures by
1355 basing the Windows build on cygwin. Added
1356 scripts to deal with unix and windows