The SamplingPruner implements "normal" sampling with equivalence-class
reuse. Unlike the FESamplingPruner, the SamplingPruner implements
uniform fault-space sampling that counts multiple hits of an
equivalence class.
This change modifies the database schema, more specifically it adds
the "weight" column to the fspgroup table. Update existing databases
with this query:
ALTER TABLE fspgroup ADD COLUMN weight INT UNSIGNED;
Change-Id: I668fc9b25fc4d79a60aa1ef8d69cdf5fa076cc6d
SumTree::get now non-intrusively picks an element and returns a
reference to it, SumTree::remove removes and returns a copy. The
former is needed for sampling with replacement.
Change-Id: Iefef2fdf0b7df6ea7a9949f2588528ec9e86bb7a
In the sampling step, the --no-weighting switch disables the
equivalence-class weighting by using a weight of one instead of the
equivalence-class size. This is usually not a good idea, and should
only be used for demonstration purposes, or if the fault model
requires weight-less sampling.
Change-Id: Id903d1924c6ecbcd217815aa5ce9271560130071
The --use-known-results switch simulates sampling (with fault
expansion, FESamplingPruner) by reusing results from a previous
campaign covering the full fault space (that used the "basic" pruner).
The pruner only creates entries in the "fspgroup" table that refer to
already existing pilots and corresponding results.
This switch is not for normal Fail* use, but only for experimenting
with the FESamplingPruner.
Change-Id: I1bf561d93f55918d243c5306551a1c6b48027198
The project version number is now not user-editable anymore, and
custom values get overwritten by force. Without this change,
CMakeCache.txt stays at the version number it was first instantiated
from.
Change-Id: If9ad1549937dad98db9a0f0eda16ef752e7d74aa
This change fixes several file-descriptor leaks in the DwarfReader
implementation. The patch is taken from Richard Hellwig's not yet
merged change I161f626d12ca7f2b7b9d13ba9cbc254eb55692f1. I did not
apply any white-space cleanups to prevent conflicts when Richard's
change will be merged later.
Change-Id: Icd9c1bdeeab39e77900e2ce88b756a8cf7ade96a
This script removes dangling rows from the database, for example
'trace' entries with a variant_id not mentioned in the 'variants'
table, or result rows referencing a nonexistent 'fsppilot' entry.
IOW, this script enforces referential integrity as it would be
maintained by foreign key constraints (that can only be used with
InnoDB tables).
Change-Id: I4dce1e46277d470f8c3eca31447ca71f63c6353f
This one omitted mappings to the first static instruction belonging to a
high-level source-code line.
Change-Id: I14427021e61f27a8b029fb56fad2ba813652422a
Using "$x .= $y" instead of "$x = $x . $y" is actually an extreme
speedup for long $x. I had no idea PHP's "compiler" was that bad and
doesn't optimize this.
Change-Id: I39aec5f14f45b75a2467d8074b5ea2ffe5d4b856
The dbg_methods, dbg_variables and dbg_stacktrace tables are not
needed by the current VisualFAIL implementation.
Change-Id: I5f685c7d8100fe57fb2a23e1f8161770e64c9464
Luckily this didn't work as advertised, as the global connection
variable wasn't imported into the functions' scopes.
Change-Id: I080a32c6418deb3da3578987ec8a181b047608ad
Since the AspectC++ daily builds switched to the clang frontend, Fail*
doesn't build anymore. Unless that changes, use the 1.2 release.
Change-Id: Ia7817c396dfabd3a10f0cf0ea6a8d4c1fb95365d
Up to now, the sanity checks only tested equivalence classes
determined by dynamic instructions (instr1/instr2), although the
actual timing boundaries (time1/time2) have been the really relevant
information for quite some time now. This change extends the checks
to time1/time2.
Change-Id: I53d1ed10684ff09f6b9f1245ef842dd0d85f2655
The sanity check testing whether the fault space is rectangular is
unnecessarily slow, as it joins the "trace" table without any reason,
and includes all variants instead of the currently imported one.
Change-Id: Icfe948290ec595209868952fc1639c979bd78d83
Up to now, source code lines and mappings from static instruction
addresses to lines could be linked to the wrong file if
dbg_filename.path contained duplicates. This is unfortunately the
normal case when importing multiple variants of the same program into
the database.
Change-Id: I57e71379584d7b01177606192b3aa644846225db
This change makes all C++-based tools in tools/ abort when they
encounter an unknown commandline parameter (both option or
non-option). This has already caused some confusion, as in some cases
unexpected behaviour can be the result. For example, "prune-trace -t
mytrace.tc -d database" up to now ignored the "-t" parameter, took
"mytrace.tc" as the first non-option parameter (and ignored it); as no
option parameter may follow the non-option parameters, all other
options were ignored as well.
Change-Id: Ia0812a518c4760fa28ed54979c81f43fa7aa096e
This change sets the option parser to GNU mode, that is, allows
further option parameters after non-option parameters, e.g.:
dump-trace foo.tc -s
instead of
dump-trace -s foo.tc
As Fail* currently works on GNU platforms only, this behavior is the
one presumably expected from users.
Change-Id: I9c55eaf4560cde81ebd0b94214201c8ad02c2b74
This change fixes the Protobuf->MySQL bridge for strings, which were
corrupted in rare cases, especially with debug builds of the
DatabaseCampaign. String columns in result tables from any campaign
up to this point may contain corrupted data.
The core reason for the corruption was that the TypeBridge_string
bound a temporary (a nameless local variable) to the prepared
statement. This temporary is destroyed before the subsequent call to
mysql_stmt_bind_param(), and the string within can only be referenced
successfully if it has not been overwritten yet. The solution is to
copy the string to a bridge-internal variable.
Although it might seem that TypeBridge_enum has the same problem, the
protobuf library seems to return references to internal string
constants when retrieving the enum values.
Change-Id: Id127e6b3333d7c304d688e45de9bea44bbc610b0
This fixes a "funny" conflict between the SumTreeTest and
weather-monitor's experiment.hpp.
Note to self: Preprocessor macros are evil.
Change-Id: I3f8c95fe086357db77110c0c53d3120ca839f30a
By using the AliasedRegistry, "prune-trace --help" (and
"prune-trace -p '?'") now lists all available Pruners to the user.
Change-Id: Ib5e3d00aabc37e6d48d804d2d709812af3f7efb2
With this change, prune-trace checks for existing fsppilot/fspgroup
entries for each variant to be pruned, and skips the variant in this
case. This safety measure can be switched off with --overwrite.
Change-Id: I7e758a9853a25685ca176cf1a1810523753cdd4a
This change removes an unnecessary "#ifndef __puma" from the LLVM
disassembler test code and fixes compilation with the latest AspectC++
binaries.
Change-Id: Ibe835a4a6df69255555c668985f15b9cf8fb82b4
Up to now, the IOPortListener only caught invocations of INB/OUTB with
the port number in DX and 8-bit input/output (from/to AL), but none of
the various other methods to access the I/O port range (e.g., by 8-bit
immediate port numbers). This change is supposed to catch all other
cases, too.
Change-Id: I1180cd9c1d59df600067739817adab684b18a608
This change allows DatabaseCampaign users to take advantage of the
improved variant selection methods in the Database class (multiple
uses of --variant/--benchmark possible, plus
--exclude-variant/--exclude-benchmark switches).
Change-Id: Idb1ca04538ff7601b3648cd9ba766aa8690fff6b
The FESamplingPruner implements the fault-expansion variance reduction
technique (FE-VRT) as described in: Smith, D. Todd and Johnson, Barry W. and
Andrianos, Nikos and Profeta, III, Joseph A., "A variance-reduction technique
via fault-expansion for fault-coverage estimation" (1997), 366--374.
Change-Id: I04a0c9bb2622974278bd8c73793e51451119e650
If no --variant / --benchmark is specified, it's more reasonable to
prune or run *all* variants/benchmarks (using the wildcard "%")
instead of defaulting to "none"/"none". The trivial case with only
one single variant/benchmark (which may still be "none"/"none" if
import-trace's default is used) is still covered by this new default
behavior.
Change-Id: I0e9001137d5e052183dd74211e2edbcfab749528
This change moves prune-trace's --variants-exclude / --benchmarks-exclude
capabilities to Database::get_variants() to make it available to all users.
Change-Id: Icbc6bb1a3ae7c846d2de40b881f47a9cc1ed7bbf
Otherwise it's not possible to keep a "basic" and a "sampling" pruned
version of one variant in the same database.
Change-Id: Ic71eb27ea16df23e2289cbf9f96ae10209745791
The SumTree implements an efficient tree data structure for
"roulette-wheel" sampling, or "sampling with fault expansion", i.e.,
sampling of trace entries / pilots without replacement and with a
picking probability proportional to the entries' sizes.
For every sample, the naive approach picks a random number between 0
and the sum of all entry sizes minus one. It then iterates over all
entries and sums their sizes until the sum exceeds the random number.
The current entry gets picked. The main disadvantage is the linear
complexity, which gets unpleasant for millions of entries.
The core idea behind the SumTree implementation is to maintain the
size sum of groups of entries, kept in "buckets". Thereby, a bucket
can be quickly jumped over. To keep bucket sizes (and thereby linear
search times) bounded, more bucket hierarchy levels are introduced
when a defined bucket size limit is reached.
Note that the current implementation is built for a pure growth phase
(when the tree gets filled with pilots from the database), followed by
a sampling phase when the tree gets emptied. It does not handle a
mixed add/remove case very smartly, although it should remain
functional.
Change-Id: If05e9700bc84761b5bc31006402641e7112b3a72
The compiler should be able to completely optimize away side-effect
free usage of this logger. Can be used as a drop-in replacement for
Loggers to silence logging output for known-good code without having
to remove the corresponding "LOG << ..." code.
Change-Id: Ifb276223f61686773dd6108aafd567e99c88b223
This enables commonly used event sources and other backend features to
avoid newcomers not getting their first experiments to work. The
performance impact is not really a problem, but for optimal throughput
unused features can still be disabled.
Change-Id: Id507474eab4647137ce82cb8674fe23f7102975a
gethostbyname() doesn't set errno but h_errno. Thus, we need to call
herror() instead of perror() to print an appropriate error message.
(Thanks, Björn.)
Change-Id: I8fd4bdd4af41774dd290151c5ad37090d006f423
This change makes the RegisterImporter continue importing if it
encounters an instruction pointer that is not part of the disassembled
ELF binary (and, thus, cannot be disassembled). This is OK if we
don't want to inject into registers used by these instructions.
Change-Id: Ia9b5e7f789367f8386d63f235451dae5d399610d
The disassembled memory region's end (variable "End") is exclusive
now. Up to now, the two branches defining this variable disagreed on
inclusiveness, leading to an infinite loop in one case.
Change-Id: I055fc240f6ec2f4a1d1937e48617c86612cff5c5
Since commit c142818, DatabaseCampaignMessage.proto is generated into
the binary directory, which must be announced to the protobuf
compiler.
Change-Id: I0c3c4c6525df687012330573979dc897dfa8d701
Enabling both CONFIG_FAST_BREAKPOINTS / CONFIG_FAST_WATCHPOINTS and
CONFIG_EVENT_BREAKPOINTS / CONFIG_EVENT_BREAKPOINTS_RANGE /
CONFIG_EVENT_MEMREAD / CONFIG_EVENT_MEMWRITE led to an (unnecessary)
include cycle.
Change-Id: I8144e3e72da69b98e21a844a4bfded1b77bdce07
This fixes a gcc 4.7 warning: comparison between signed and unsigned
integer expressions [-Wsign-compare].
Change-Id: If440567c2767494f456e8e68df29e2f8d1582955
BochsController::getTimerTicksPerSecond() only works reliably when the
simulation is already running (e.g., after a restore()). This broke
timeout conditions for the very first experiment in a FailBochs instance.
Change-Id: Ice5f0aa0c6759f2d9341ad4f21d5c346307b4c12
This change fixes another Bochs crash in the floppy controller,
triggered by one out of 670 million experiments in my current
campaign.
Change-Id: I8a2ff78c9d2c8fca12eefb97f508bc213373bbfa
This change makes the DatabaseCampaign load all pilot_ids from the result
table in memory instead of LEFT JOINing them for each variant. This vastly
improves campaign speed (possibly making commit 5567c59 superfluous) at the
cost of slightly increased startup time for half-completed (large)
campaigns.
By exploiting the generally continuous nature of pilot IDs and using a
boost::icl::interval_map, the additional memory requirements are
insignificant.
Change-Id: I1e744fb9ca33efea77a2a785cea3c94106f360df
When no variants matching the command line parameters were found, the
campaign printed an uninitialized sent_pilots count.
Change-Id: Ib1d70ae86f02059daeb9a62567d6c83802e4986e
The fix for tools/convert-trace/Gem5Converter.cc repairs a real issue
which caused a runtime warning to never be displayed.
Change-Id: I5d94acdbffc99d5f2a78f047062c011c6f8ab4e3
This change updates the documentation on how to manually build LLVM
3.3 or 3.4, and removes the forced -fno-rtti compiler flag from
util/llvmdisassembler/: The flag breaks compilation with (not yet
released) AspectC++ 2.0, which complains about dynamic_cast (used in
some active aspect headers that should not affect llvmdisassembler at
all) being unusable with -fno-rtti. It's probably not a good idea to
compile only a part of Fail* with -fno-rtti anyways.
Note that the Debian and Ubuntu LLVM packages are built with
REQUIRES_RTTI=1, too.
Change-Id: I9891cf074d4201df786f7f5a9b96033e18832562
This crippled version of the weather-monitor experiment is not needed
anymore, as gem5's restore() works now.
Change-Id: Icc3bced6b04241d55f2acc46f341cf553b6dbd08
As the eCos kernel tests do not sufficiently check for failures, a
(compared to the golden run) measurably *reduced* simulation time can be
used as an indication for SDC.
Change-Id: I303f6c3f7ef56cb691c1f226eff74d113dd16629
Horst Schirmeier