<html>
<head>
<style><!--
.hmmessage P
{
margin:0px;
padding:0px
}
body.hmmessage
{
font-size: 10pt;
font-family:Verdana
}
--></style>
</head>
<body class='hmmessage'>
(verified that empty records are of zero size; in C++ that's a not always possible optimization, they often have a filler byte, in order to generate unique addresses in places.)<BR> <BR>
<HR id=stopSpelling>
From: jay.krell@cornell.edu<BR>To: hosking@cs.purdue.edu; m3devel@elegosoft.com<BR>Date: Wed, 16 Dec 2009 15:36:24 +0000<BR>Subject: [M3devel] pthread/Interix<BR><BR>
<STYLE>
.ExternalClass .ecxhmmessage P
{padding:0px;}
.ExternalClass body.ecxhmmessage
{font-size:10pt;font-family:Verdana;}
</STYLE>
[again, expect a bounce, but the mail does get through]<BR> <BR> <BR>Tony, for Interix, suspend/resume thread can be achieved writing to a file.<BR>Getting a suspended thread's registers can be gotten reading from a file.<BR>(I believe the same thing works on Solaris.)<BR>It appears to be intended for debuggers.<BR>I can see that it works well.<BR>I doubt that a thread's context is on a signal handler's stack.<BR>I propose:<BR> <BR> <BR> ActState = { Starting, Started, Stopping, Stopped };<BR> REVEAL Activation = UNTRACED BRANDED REF RECORD<BR> frame: ADDRESS := NIL; (* exception handling support *)<BR> handle: pthread_t := NIL; (* LL = activeMu; thread handle *)<BR> statusFile: int := -1; (* for Interix *)<BR> controlFile: int := -1; (* for Interix *)<BR> (* C code knows the structure above this point *)<BR>...<BR> END;<BR><BR> <BR>ThreadPThread.h:<BR> <BR> <BR>typedef struct _Activation_t {<BR> void* frame;<BR> void* pthread;<BR> int statusFile; /* for Interix */<BR> int controlFile; /* for Interix */<BR>} Activation_t;<BR><BR> <BR>and then whereever we pass act.handle off to C code, pass act instead.<BR> <BR> <BR>frame is kept first so that it is offset 0 in case that gives codegen efficiencies<BR>on the often used push/popframe.<BR>The other three are moved up to the top to reduce maintainence of ThreadPThread.h.<BR> <BR> <BR>I assume "branding" doesn't introduce data in the record.<BR>I'd have to check the codegen to verify frame is at offset 0.<BR> <BR> <BR>#ifdef code will open the files.<BR>Modula-3 code can check if the files are >= 0 and close them,<BR>or it can be a call out to #ifdefed C.<BR> <BR> <BR>An alternative is that wherever we pass act.handle, also pass act.statusFile and act.controlFile.<BR>(as VAR for initialization).<BR>That is less fragile and less extensible.<BR>I'm fine with that approach too.<BR> <BR> <BR>As well though, I was wondering, maybe we should have:<BR> <BR> <BR>common/InterixThreadState.i3<BR> T = RECORD END; (* zero size? *)<BR> <BR> <BR>interix/InterixThreadState.i3<BR> T = RECORD statusFile, controlFile := -1; END;<BR> <BR> <BR>and then put InterixThreadState.T in Activation.<BR>That way saving the other platforms from spending the space.<BR> Just two integers, not much, but in principle..<BR> (We also sometimes use up 8 bytes for the pthread_t when 4 would suffice,<BR> on 64bit platforms with pthread_t == int, I think HP-UX is like that at least.)<BR> <BR> <BR>I could also open/close the files for every suspend/resume/processStopped,<BR>though that seems too lame.<BR> <BR> <BR>Another idea, that I don't really like, is to never optimize and<BR>ignore the registers and just use the stack.<BR>I rarely optimize, but depending on that seems bad.<BR> <BR> <BR>Your "safe points" proposal would also make this work, though<BR>I'd rather not wait for that.<BR> <BR> <BR>?<BR> <BR>Ok to add the two integers all the time?<BR>Ok to either depend on the record layout in C, or to always pass the integers around?<BR> <BR> <BR>Thanks,<BR> - Jay<BR> </body>
</html>