Warning, /FWCore/MessageService/doc/single-thread.txt is written in an unsupported language. File is not indexed.
0001 Single-threaded MessageLogger design plan
0002 -----------------------------------------
0003
0004
0005 The key points of change are on either side of the queue:
0006
0007 On the left:
0008 ------------
0009
0010 MessageLoggerQ needs to change every time it does b.commit() to a call
0011 to something in the scribe. Clearly, it needs to know whetehter this is single
0012 or multiple threaded to make this choice.
0013
0014 The change is to substitute runCommand(opcode, operand) for the sequence
0015 of creating, filling and committing the buffer for the SingleConsuerQ.
0016
0017 There will be some commands with other slight changes as well.
0018
0019 On the right:
0020 -------------
0021
0022 MessageLoggerScribe::run()
0023
0024 do {
0025 MessageLoggerQ::consume(opcode, operand); // grab next work item from Q
0026
0027 has to become something that gets invoked directly,
0028 and when the break happens, the code needs to return to the
0029 calling routine. The way to do this is to break off the meat
0030 of that internal do loop, into a call to runCommand(opcode, operand).
0031
0032 The remainder of changes on the "right" of the fence are in particulars
0033 of the specific commands.
0034
0035 B) Notify_all() is now not going to be needed in the case of single-threaded.
0036 That implies that the scribe ought to know whether it is single-treaded or
0037 not.
0038
0039
0040 On the fence:
0041 -------------
0042
0043 The issue is that one must not create a separate message logger thread if
0044 single-thread is desired.
0045
0046 We first have to find out which option is wanted: Here, we "parse" the command
0047 line directly, because the setup of the boost parser allows for an exception,
0048 which we want handled by the logger.
0049
0050 We create a SingleThreadMSPresence to substitute for the MessageServicePresence.
0051
0052 That has been tested with the plug-in mechanism, by making it identical to the
0053 service it replaces, and using it against the unit tests. Note that I had to
0054 include the new one in FWCore/MessageService/plugins/Module.cc also.
0055
0056
0057
0058 At this point I have started changing SingleThreadMSPresence but am at a snag:
0059 MessageLoggerQ wants to call MessageLoggerScribe::runCommand, but that would
0060 introduce a dependency of MessageLogger on MessageService.
0061
0062
0063 Resolution of the snag:
0064
0065 When SingleThreadMSPresence is created, it was going to have to instantiate a
0066 MessageLoggerScribe. However, now MessageLoggerScribe puhlicly inherits from
0067 AbstractMessageLoggerScribe, which is defined in the MessageLogger pacakge.
0068 A pointer to that AbstractMessageLoggerScribe, which has a virtual method
0069 runCommand(opcode, operand) is provided to MessageLoggerQ. So now, we
0070 can do scribePointer->runCommand(opcode, operand) an in the implementation
0071 level, it is actually pointing to a true MessageLoggerScribe!
0072
0073
0074 TODO:
0075
0076 SingleThreadMSPresence ctor
0077 OK create the (concrete) MessageLoggerScribe(true)
0078 OK supply the pointer as an AbstractMLscribe to MessageLoggerQ
0079 OK Determine if we need to instantiate MessageLoggerQ early in the
0080 initializer list -- I don't think so, but...
0081
0082 MessageLoggerQ
0083 OK Have place for the mlscribe_ptr
0084 OK Obtain mlscribe_ptr
0085 change commit sequence to mlscribe-ptr->runCommand
0086 change each of the specialized handshake commands (in tandem with MLS)
0087
0088 MessageLoggerScribe
0089 Deal with wind-down issues - may not need code changes, but...
0090 change each of the specialized handshake commands (in tandem with MLQ)
0091
0092 MessageDrop
0093 Inform MessageDrop about singleThread status
0094 When single thread, don't use thread-specific singletons.
0095
0096 Testing
0097
0098
0099 Do we need to pre-instantiate MessageLoggerQ early in the SingleThreadMSPresence
0100 initializer list? No, because before it is ever used, it will be instantiated
0101 by the code trying to use it. There is no worry about interthread coherence
0102 in the single-thread mode.
0103
0104 MessageDrop is also going to have to change, since it is a thread-specific
0105 static. This in turn means it needsto be invformed and so forth.
