renamed some flags for consistency
obey casing rules for symbol names ( case sensitive for unix like )
deleted the setenv code
proper stacking of symbols values
some non symbols related mods ( cosmetics ??? )
tested on ...
Yosemite(CLANG), Centos-6(GCC), Centos-7 (CLANG and GCC)
( reinstalling windows XP under VMware takes longer than planned )
Make changes necessary to successfully build without OPTION_SHARED_DEVICES defined.
Define new HQA Build Scenarios 12 & 13 for building with/without OPTION_SHARED_DEVICES.
Note: this is the first (i.e. part 2a) of Part 2 of the locking/threading rework effort: hthreads restructuring. Part 2b will be the hthreads.c code necessary to detect threads exiting while still holding locks (i.e. abandoned locks).
featall.h current #undef's some OPTION_NO_LINE... vstore/dat build options (so that they are always inlined). Prior to this commit, attempting to #define the options (so the functions are NOT inlined) didn't work. This commit corrects this problem so they may be #define'd or #undef'ed as desired.
Additionally, some new HQA scenarios have been defined related to inlining so we can determine which setting we should choose before removing the build option altogether.
FishHang was useful during development of fthreads but has not been needed for so many years and so has FINALLY been deleted. Sorry it took me so long to do it.
build option in featall.h and report in version.c which one
that Hercules was built with.
git-svn-id: file:///home/jj/hercules.svn/trunk@7513 956126f8-22a0-4046-8f4a-272fa8102e63
Default is to #define it (i.e. to generate support for it)
I decided to make it optional so you could disable it if you want since
I noticed it *does* slow down power-on when you have lots of dasd since
each dasd is essentially being opened twice.
git-svn-id: file:///home/jj/hercules.svn/trunk@7477 956126f8-22a0-4046-8f4a-272fa8102e63
I did some profiling with the package perf of Ingo Molnar. After
investigation of the most called functions I implemented the following
features:
The features uses the two-byte opcode table. The second byte can be part
of the instruction format and is used to jump to optimized routines. In
this patch to routines with zero x2.
Optimized BC and BRC
Normally a value of 0x80 is shifted by psw.cc and then masked at the
instruction. The optimized instructions can calculate on psw.cc
directly.
Optimized ST, CL, L, LG and STG
These are RX and RXY instructions, but most of the time the X2 value is
zero. The optimized instructions don't have to calculate and use X2.
run_cpu and run_sie.
I did profiling and testing on those 2. The current setting runs the
fastest on core i7.
Have fun!
Bernard van der Helm
git-svn-id: file:///home/jj/hercules.svn/trunk@7187 956126f8-22a0-4046-8f4a-272fa8102e63
(originally introduced in SVN #6966), default to "msglevel debug" for
DEBUG builds, 'OPTION_DEBUG_MESSAGES' build option is now obsolete.
git-svn-id: file:///home/jj/hercules.svn/trunk@7151 956126f8-22a0-4046-8f4a-272fa8102e63
