to do it:
in .ini file set binned_fns=foo null bar foo
what this says is, track foo when it is called by null (i.e. anything). bin bar only when it is called by foo.
essentially, if you have a path of functions to track, the 0th, 2nd, 4th fn listed are the fns to track, and the 1st, 3rd, 5th are their respective callers.
base/statistics.hh:
turn it back to FS_MEASURE since we already have a build in place
kern/tru64/tru64_events.cc:
remove FS_MEASURE #defs. add more DPRINTF's. manage an anomaly that happens when tracking idle_thread.
kern/tru64/tru64_events.hh:
remove FS_MEASURE #defs
kern/tru64/tru64_system.cc:
make DumpState print all the time, but only with DPRINTF. add a new parameter to tru64System a vector<string> binned_fns, to read in from .ini file. now all this binning stuff is dynamically generated.
kern/tru64/tru64_system.hh:
remove all static binning stuff, add support for dynamic
sim/system.cc:
change nonPath bin name to Kernel, remove FS_MEASURE
sim/system.hh:
change nonPath to Kernel
--HG--
extra : convert_revision : 9ee813c0a64273bab4125815b7bc8145c5897ec1
code.
base/statistics.hh:
We're getting rid of FS_MEASURE, but for now, we're going
to still use a compile time flag to turn on and off binning
of statistics. (The flag is STATS_BINNING)
cpu/exec_context.cc:
cpu/exec_context.hh:
kern/tru64/tru64_system.cc:
get rid of FS_MEASURE
cpu/simple_cpu/simple_cpu.cc:
yank the function call tracking code out of the cpu and move
it into the software context class itself.
kern/tru64/tru64_system.hh:
get rid of FS_MEASURE
move all of the tacking stuff to the same place.
sim/system.hh:
cleanup
--HG--
extra : convert_revision : 73d3843afe1b3ba0d5445421c39c1148d3f4e7c0
- Make the MemoryController use address ranges (via Range) instead
of an address and a mask
base/remote_gdb.cc:
reflect name change
dev/alpha_access.h:
better include
dev/alpha_console.cc:
- FunctionalMemory no longer takes care of mapping my address into
the proper address space. It must be done locally.
- the memory controller no longer uses a mask, but a size, and the
size is determined by the device, not the .ini file
- fix up address calculations to reflect the removal of a mask
- PhysicalMemory::getSize() -> PhysicalMemory::size()
dev/alpha_console.hh:
- FunctionalMemory no longer takes care of mapping my address into
the proper address space. It must be done locally.
- the memory controller no longer uses a mask, but a size, and the
size is determined by the device, not the .ini file
- fix up address calculations to reflect the removal of a mask
- get rid of MmapDevice and inherit from FunctionalMemory
--HG--
extra : convert_revision : e3a65c9debf6f899632d62c70781cbdc2826616b
base/range.hh:
Add a constructor that takes the beginning and end as arguments
size returns T not a bool
quick make_range() function that is a shortcut for making a range
kinda like make_pair()
quick formatting fix
--HG--
extra : convert_revision : 94b1d462710e6fb55e72e1da2ad8c46993af1ef7
to be consistent with the way that the stl works. It also makes
lots of other stuff easier. (Maybe those guys were smart?)
Overload the various comparison operators so that you can test
for overlapping of ranges, etc.
base/range.hh:
Totally rework the Range class. Now the range is from [start, end)
to be consistent with the way that the stl works. It also makes
lots of other stuff easier. (Maybe those guys were smart?)
Overload the various comparison operators so that you can test
for overlapping of ranges, etc.
make parse function private and offer operator= instead
isValid -> valid
and for you erik, I add comments
test/Makefile:
add range.o
test/rangetest.cc:
better tests
--HG--
extra : convert_revision : dd58720aa3d02f20b03e933f2eaa3320c82bb27a
The last change only caught the ones with types that
started with capitals. This pass catches the rest
(mostly STL and uint*_t types).
base/cprintf_formats.hh:
cpu/simple_cpu/simple_cpu.cc:
sim/serialize.cc:
sim/serialize.hh:
Change "foo_t& foo" to "foo_t &foo".
--HG--
extra : convert_revision : fc7f7425db2aef33e490f952b5ce74c8c36d0d41
base/kgdb.h:
Remove flags that aren't used
base/remote_gdb.cc:
Better debugging:
- Give each class a name() function so that the trace infrastructure
knows the correct object name.
- Make the remote debugger capable of detach.
- Split out the RGDB trace flag into a bunch of specific flags.
Remove dead code
Add a new trap type
base/remote_gdb.hh:
Add a name() to the various objects for the trace system
base/trace.hh:
don't need a using directive
add DPRINTFNR: debug printf, no flag, raw output
kern/tru64/tru64_system.cc:
use the INT trap type instead of IF
--HG--
extra : convert_revision : 25e610216c6f43d5d328651bba915f71bade059e
This primarily to be internally consistent (sometimes we used one,
sometimes the other, even within the same line of code!).
I picked the latter to be symmetric with "Foo *foo".
base/cprintf_formats.hh:
base/range.hh:
base/refcnt.hh:
base/res_list.hh:
base/statistics.hh:
base/str.hh:
cpu/exec_context.hh:
cpu/simple_cpu/simple_cpu.cc:
cpu/simple_cpu/simple_cpu.hh:
sim/serialize.cc:
sim/serialize.hh:
sim/syscall_emul.hh:
Change "Foo& foo" declarations to "Foo &foo".
--HG--
extra : convert_revision : ca1b0e85a578b539214bda3b8d61ac23792f2e87
not while printing out the data. This allows the data
to be dumped more than once.
base/cprintf.hh:
need a destructor
--HG--
extra : convert_revision : 235e9fe24488ac4c0ae1b562ef9fa6e0bd1e899c
statStream catchall that we had before)
Also provide an optional output directory that multiple simulator output
files can be written to.
Make most output files use the output directory
base/misc.cc:
send warnings to the outputStream as well
base/trace.cc:
dprintf_stream defaults to cerr
dev/console.cc:
use the output directory for the console output if it exists
dev/etherdump.cc:
dump to the output directory if it exists
sim/builder.cc:
sim/builder.hh:
move constructor and destructor to .cc file
use a function to get the stream that the builder dumps its
output to, and create a separate file in the output directory
if able
sim/main.cc:
statStream -> outputStream
sim/serialize.cc:
dump checkpoints to the output directory if specified
sim/universe.cc:
provide an output stream for simulator output. (This takes place of the
statStream catchall that we had before)
Also provide an optional output directory that multiple simulator output
files can be written to.
--HG--
extra : convert_revision : 03abce20edbbf7ec19c9ddd8d69ec8485c383532
arch/alpha/isa_desc:
don't say warn: Warning:
base/misc.cc:
avoid printing two newlines in a row
sim/main.cc:
print out a message just before we enter the event queue
--HG--
extra : convert_revision : 2a824d4b67661903fc739a0fb0759aa91d72382c
but did not test... however the only thing this
affects in libelf is the header, so I don't expect
any functional problems.
base/loader/elf_object.cc:
Set LIBELF_LINUX to 0 to build on zax.
--HG--
extra : convert_revision : d024b33deff6fc8ea6f1d465f76dc8d9d63254ab
to find and remove stats from the files so we can put less
burden on the python interpreter.
base/statistics.cc:
Manually insert newlines into the python code so that we now have one
stat per line. (Make it so we can use grep -v to remove stats)
test/stattest.cc:
update to reflect changes in how python is accessed
--HG--
extra : convert_revision : 554edcf9c795b33d00d3d15554447c8accebebfa
base/loader/elf_object.cc:
- Use program header instead of section headers for loading.
The old code doesn't work for Alpha/Linux user binaries.
- Deal properly with single-segment ELF executables (e.g. kernels)...
haven't tested this but it looks like it matches the functionality
of the patch the L4::Pistachio guys sent.
- Factor load symbol code into one common function.
- Some formatting to match style conventions.
- Only include libelf stuff in .cc file, not .hh.
base/loader/elf_object.hh:
- Use program header instead of section headers for loading.
The old code doesn't work for Alpha/Linux user binaries.
- Factor load symbol code into one common function.
- Only include libelf stuff in .cc file, not .hh.
--HG--
extra : convert_revision : 782062d09d43083b01e65cd645c941f49c27640a
base/cprintf.cc:
Fix bug where a call with a format string with no specifiers
but one or more arguments would crash. Old code tried to print
extra args using last available format, but in this case there
was no format, hence the problem. Now we just print "<extra arg>"
for each extra arg.
Also reorganized code a bit to make scope of fmt variable
match the scope in which you can be confident it's meaningful.
base/cprintf.hh:
Print specific error message instead of calling format_invalid().
base/cprintf_formats.hh:
Clear Format object in constructor.
Use specific error messages instead of format_invalid() function.
--HG--
extra : convert_revision : 87d5e902174e3eb2583131d056742af166540db0
machine.
base/loader/elf_object.hh:
put #defines for libelf that normally arn't defined so that -Wundef
doesn't cause problems.
--HG--
extra : convert_revision : fa2de6aa94c7ddbaa59990db4473402b0df7d93a
and setup filed had to be modified to build libelf. The ext repository
MUST exist in the directory below m5!
base/loader/elf_object.cc:
base/loader/elf_object.hh:
now uses libelf to read elf binaries
--HG--
extra : convert_revision : b70511f5379edf35d639f3bb64b786cb0bd180af
fix some bugs
base/statistics.cc:
- Move the python dump stream stuff into the statistics package
because it needs more control over the file when there
are multiple dumps
- add a subname to the output
- make all of the stats a list instead of variable arguments to
get around the 255 argument limit
- the description needs to be triple quoted
- avoid errors for formulas that don't have any root element
set up.
base/statistics.hh:
- get rid of mode_python and just separate python dumping from
regular stats dumping
- fix a huge bug that made a Formula use a VectorData instead
of a FormulaData
--HG--
extra : convert_revision : 7303cff3ccdcc3d306ab17375219fc7fecac7e5e
base/statistics.cc:
base/statistics.hh:
- add python output support to the statistics package
- each statistic type has a python() member function that takes
a Python object to which the stat will output it's python
representation
- add getStatData hack so that the StatData pointer can be looked
up by the proxies with their opaque pointer to the stat they're
proxying for. This is necessary because the proxy really proxies
for the bin and not the stat. Be nice to figure out how to get
rid of it. The hack is used so that the str() function of a
proxy can properly name itself.
- To print formula stats, every stat has a str() function that
converts that stat to a string that python can execute to get
a value.
test/Makefile:
add python stuff
test/stattest.cc:
add more tests and test python support
--HG--
extra : convert_revision : 513814ab0a125606897f2c57dccdf22879032ef9
python dumping takes a name for the output data
base/statistics.cc:
base/statistics.hh:
initial basic hooks into the stats package to output python code
--HG--
extra : convert_revision : 37f52dc03df50aa90346dc9ca341f961e7e855e8
base/statistics.hh:
- use a typedef to get at bin_t
- clean up template declaration
- public access to typedefs
--HG--
extra : convert_revision : 0650c5a1cba51c7af04790498af0d354884be0ee
base/statistics.cc:
- give an internal name to statistics that aren't printable
- speed up reset a tad
- Make sure that the original value of the current bin is restored properly
- check for the case where there are no subnames so we don't have an error
- check to make sure that the formula's root is set before accessing it
base/statistics.hh:
- clean up includes a bit
- deal with the case where there are no subnames
- get rid of remnants of FS_MEASURE stuff
--HG--
extra : convert_revision : 1e4338fc2b225647df47641b04c13b7046b2e114
- make init and print StatFlags
- default_mode -> DefaultMode
- the display mode is no longer needed as part of the stat
pass it into the display functions so the mode can be determined
from dump to dump
- get rid of old simplescalar bug and make the distribution min_val
work correctly
- get rid of the DisplayMode parameter to distribution updates since
it is no longer needed
--HG--
extra : convert_revision : 386f69f4d12fb91faf60690aaab08542e009e555
- Factor out the formatted output into a display function
- Make the stats package support tracing
- Clean up output of binned data
base/statistics.cc:
- Factor out the formatted output into a display function and
pass in the mode
- Make the stats package support tracing
- Clean up output of binned data
- Get rid of the list of binned stats and the map of bins
since it is no longer needed for output
base/statistics.hh:
- Allow the stats package user to determine in which mode to dump
--HG--
extra : convert_revision : a1d5fc728f4dc22a851d58ea89d44018a94a64c7
checking in places to make sure stuff is behaving properly.
base/statistics.cc:
separate the per stat check from the general stats check.
We always want the general stuff to happen
base/statistics.hh:
- separate the per stat check from the general stats check.
we always want the general stuff to happen
- make every stat check that its bin is at least initialized
- set the vector2d x and y coordinates in init to prevent an
uninitialized memory access
test/Makefile:
don't need sim_time.o to test stats
test/stattest.cc:
don't make x and y the same on the 2d test so that we make
sure that the two dimensions are correct
--HG--
extra : convert_revision : 81320325056ac1c09f6842474fb6ee3bcc030a8e
base/cprintf.cc:
Get rid of unnecessary stream state check.
Bug is fixed now, and I'm not sure this did much anyway.
cpu/exetrace.cc:
Get rid of unnecessary flush() call.
--HG--
extra : convert_revision : 107dcdc578b2b0ad1652ee52ea0a33b0f3cd4e39
arch/alpha/alpha_tru64_process.cc:
So, I don't know why linux uses an off_t here.
I'm also not sure why linux defines an off_t to be a long
Let's just use long here since it works for linux, and that's
what bsd does
base/inifile.cc:
correct #include for OpenBSD
dev/disk_image.cc:
the correct type for this is streampos
--HG--
extra : convert_revision : f3ac3a3b8515d66e07ffb9780d8a9e387297b6a0
Most notable is that the rundiff trace-output lossage is fixed!
base/pollevent.cc:
Use SA_RESTART to keep blocked write() calls from failing
with EINTR when signals happen.
base/trace.hh:
Make TRACING_ON depend on NDEBUG not being set (instead of
DEBUG being set) so m5.opt/m5.fast do what we want.
util/rundiff:
Print header lines showing file/cmd names.
--HG--
extra : convert_revision : 4966814ad65aedb6d62a5a62a6ad3e3f326d61b8
base/statistics.hh:
Make a couple of typedefs public rather than protected.
g++ 3.3 complains otherwise.
--HG--
extra : convert_revision : 6a8e4edb0c39474c0e0c7bc474ad0f22b9fb8505
disable FastAlloc. Gcc 3.3.x appears to have a bug that
breaks FastAlloc when compiled with optimization.
base/fast_alloc.cc:
base/fast_alloc.hh:
#define NO_FAST_ALLOC makes FastAlloc a no-op.
--HG--
extra : convert_revision : bf4bb8e963331c6782f2cfd475a811aa520e31b5
base/str.hh:
this should really be inline
base/time.cc:
base/time.hh:
clean up how the time class works. Export only one, and let
people calculate their own elapsed times, etc.
sim/main.cc:
sim/sim_time.X -> base/time.X
--HG--
rename : sim/sim_time.cc => base/time.cc
rename : sim/sim_time.hh => base/time.hh
extra : convert_revision : f3888fe3a1fdd1022084c282b58407c631a6d9a0
- make stuff compile with NDEBUG
- remove the TRACE setup option since it isn't useful anymore
base/statistics.cc:
Compile with -DNDEBUG
--HG--
extra : convert_revision : a75b3881be1513abb5915b6463c6cdc1f23a62e1
