<html>
<head>
<style><!--
.hmmessage P
{
margin:0px;
padding:0px
}
body.hmmessage
{
font-size: 12pt;
font-family:Calibri
}
--></style></head>
<body class='hmmessage'><div dir='ltr'><div>Ok either, way, but here is another point.</div><div><br></div><div><br></div><div>In some of our code we do:</div><div><br></div><div><br></div><div>CONST DEBUG = FALSE;</div><div><br></div><div>IF FALSE THEN</div><div> xx</div><div>END;</div><div><br></div><div><br></div><div><br></div><div>It will tend to be optimized.</div><div>Furthermore, syscalls -- even if not optimized,</div><div>IF FALSE is going to be way faster than a syscall.</div><div><br></div><div><br></div><div>Also, gettimeofday shouldn't necessarily be a syscall.</div><div><br></div><div><br></div><div>But/and there are/should-be faster ways of getting time for benchmarking.</div><div><br></div><div><br></div><div>On Windows, GetTickCount and GetTickCount64 just read a value out of memory.</div><div>Even GetSystemTimeAsFileTime is just a read of a 64bit value from memory.</div><div>(The memory is read only, at the same address in all processes, and updated by</div><div>the kernel.)</div><div>x86 has rdtsc, a 64bit hardware cycle counter readable in one instruction.</div><div><br></div><div><br></div><div><br></div><div><br></div><div>We should consider researching analogs to rdtsc across mips/powerpc/sparc/arm and</div><div>provide something in m3core. I think.</div><div>Add it to the backend interface likely, for inlining.</div><div>or maybe it is worth a function call, then just use #ifdefed C.</div><div><br></div><div><br></div><div><br></div><div>Maybe Posix provides something that tends to be like this?</div><div><br></div><div><br></div><div>Mika, how did you find this?</div><div>Control-c every so often and look at callstack?</div><div>Or a lot of stepping?</div><div><br></div><div><br></div><div> - Jay</div><div><br></div><br><div>: Thu, 9 Jan 2014 15:01:55 -0600<br>> From: rodney_bates@lcwb.coop<br>> To: mika@async.caltech.edu<br>> CC: m3devel@elegosoft.com<br>> Subject: Re: [M3devel] performance problems in CM3 Text operations on AMD64_LINUX<br>> <br>> <br>> <br>> On 01/04/2014 12:55 PM, mika@async.caltech.edu wrote:<br>> > I have this program that generates Modula-3 code from XML.<br>> ><br>> > It uses Wr.PutText a lot, on small strings.<br>> ><br>> > It runs 10 times faster under PM3 on FreeBSD/i386 on a 2 GHz old Pentium<br>> > III type computer than under CM3 (at head) on Debian/amd64 on a 4 GHz<br>> > Core i7 4960X.<br>> ><br>> > (176)rover:~/t/calarm/fix/gen/src>time ../FreeBSD4/genfix -o ../../fix42/src -T ../templates -L fix42 FIX42.xml<br>> > 0.419u 0.089s 0:00.70 70.0% 471+27608k 0+546io 0pf+0w<br>> ><br>> > (29)truffles:~/t/calarm/fix/gen/src>time ../AMD64_LINUX/genfix -o ../../fix42/src -T ../templates -L fix42 FIX42.xml<br>> > 1.952u 0.232s 0:05.40 40.3% 0+0k 0+11504io 0pf+0w<br>> ><br>> > I believe the issue is that Text.m3 is making a ton of system calls!<br>> ><br>> > Here's a typical traceback:<br>> ><br>> > (gdb) where<br>> > #0 0x00002aaaaaacb770 in ?? ()<br>> > #1 0x00002aaaaaacba1b in gettimeofday ()<br>> > #2 0x00002aaaac468f8a in gettimeofday () at ../sysdeps/unix/sysv/linux/x86_64/gettimeofday.S:37<br>> > #3 0x000000000048c10c in TimePosix__Now () at ../src/time/POSIX/TimePosixC.c:50<br>> > #4 0x000000000048b4d2 in Time__Now () at ../src/time/POSIX/TimePosix.m3:14<br>> > #5 0x000000000049cbcf in TextStats__NoteFinished (M3_ESffkp_o=<error reading variable>) at ../src/text/TextStats.m3:64<br>> > #6 0x00000000004921de in Text__SetChars (M3_CKMnXU_a=<error reading variable>, M3_Bd56fi_t=<error reading variable>,<br>> > M3_Cwb5VA_start=<error reading variable>) at ../src/text/Text.m3:512<br>> > #7 0x0000000000441a26 in UnsafeWr__FastPutText (M3_BxxOH1_wr=<error reading variable>, M3_Bd56fi_t=<error reading variable>) at ../src/rw/Wr.m3:100<br>> > #8 0x000000000044192f in Wr__PutText (M3_BxxOH1_wr=<error reading variable>, M3_Bd56fi_t=<error reading variable>) at ../src/rw/Wr.m3:89<br>> > #9 0x000000000041627f in M3Syntax__TextSetPutList (M3_CT76zs_s=<error reading variable>, M3_BxxOH1_wr=<error reading variable>,<br>> > M3_AicXUJ_quoted=<error reading variable>) at ../src/M3Syntax.m3:129<br>> > #10 0x00000000004169cf in M3Syntax__EnumToText (M3_Bqnwti_e=<error reading variable>, M3_ACGgGY_env=<error reading variable>) at ../src/M3Syntax.m3:173<br>> > #11 0x000000000041791a in M3Syntax__UnstrToText (M3_AIPC2K_u=<error reading variable>, M3_ACGgGY_env=<error reading variable>) at ../src/M3Syntax.m3:295<br>> > #12 0x0000000000414ad4 in M3File__DumpInterface (M3_CjySNq_t=<error reading variable>, M3_Bd56fi_dirPath=<error reading variable>,<br>> > M3_AicXUJ_updateMakefile=<error reading variable>) at ../src/M3File.m3:78<br>> > #13 0x000000000040f3f9 in Fields__Do__MakeFieldsFiles.1150 () at ../AMD64_LINUX/Fields.m3:526<br>> > #14 0x000000000040a713 in Fields__Do (M3_DsEykq_xml=<error reading variable>, M3_Bd56fi_versionString=<error reading variable>,<br>> > M3_Bd56fi_beginStringV=<error reading variable>, M3_D9m5ya_fieldTbl=<error reading variable>, M3_Bd56fi_Dir=<error reading variable>,<br>> > M3_Bd56fi_SrcDir=<error reading variable>, M3_AZx9O5_binaryFields=<error reading variable>, M3_CebKnt_specialFields=<error reading variable>,<br>> > M3_CT76zs_checkedTypes=<error reading variable>) at ../AMD64_LINUX/Fields.m3:538<br>> > #15 0x0000000000413948 in Main_M3 (M3_AcxOUs_mode=<error reading variable>) at ../AMD64_LINUX/Main.m3:431<br>> > #16 0x00000000004742ad in RTLinker__RunMainBody (M3_DjPxE3_m=<error reading variable>) at ../src/runtime/common/RTLinker.m3:408<br>> > #17 0x0000000000473638 in RTLinker__AddUnitI (M3_DjPxE3_m=<error reading variable>) at ../src/runtime/common/RTLinker.m3:115<br>> > #18 0x00000000004736cc in RTLinker__AddUnit (M3_DjPxE5_b=<error reading variable>) at ../src/runtime/common/RTLinker.m3:124<br>> > #19 0x0000000000405df8 in main (argc=8, argv=0x7fffffffe2b8, envp=0x7fffffffe300) at _m3main.c:22<br>> > (gdb)<br>> ><br>> > I believe all the TextStats operations are killing the performance. Can we remove them?<br>> ><br>> > Here is SetChars:<br>> ><br>> > PROCEDURE SetChars (VAR a: ARRAY OF CHAR; t: T; start: CARDINAL) =<br>> > BEGIN<br>> > TextStats.NoteGround (TextStats.Op.SetChars);<br>> > TextStats.NoteGround (TextStats.Op.get_chars);<br>> > t.get_chars (a, start);<br>> > TextStats.NoteFinished (TextStats.Op.get_chars);<br>> > TextStats.NoteFinished (TextStats.Op.SetChars);<br>> > END SetChars;<br>> ><br>> > I'm not sure how best to do this. Removing them without killing functionality might not be<br>> > easy. Looks like a case for conditional compilation.<br>> <br>> Now the instrumentation calls are commented out with (*47 ... 74*), which could be reinstated<br>> by string search & replace "(*47" -> "(*47*)", etc., and removed again similarly. These are<br>> only needed for performance study of the algorithms, which likely won't be done often.<br>> <br>> ><br>> > Maybe generate the .m3s from some other source, with and without the TextStats, depending on<br>> > a setting in m3makefile?<br>> ><br>> > Second best would be to have a boolean flag in Text.m3 so the TextStats never get called.<br>> ><br>> > Mika<br>> ><br>> ><br>> <br></div> </div></body>
</html>