d12b6e375b
vm: iterate_callstack_reversed in slot_visitor
2011-12-14 09:56:47 -08:00
e116bc497b
vm: reversed callstack iteration
2011-12-14 09:56:47 -08:00
e2fafaed9c
WIP verify_callstack function
...
dumps callstack in reverse order so i can visually inspect that it matches the old forward walking logic
2011-12-14 09:56:47 -08:00
ef38688e87
vm: method to calculate frame size for address
...
If we're before the prolog or or in a leaf procedure, the stack frame is really a leaf frame created by the signal handler, with a known fixed size, instead of the real stack frame (if any) with a different size.
2011-12-14 09:56:47 -08:00
5b9aee2c91
vm: don't suffer null return addresses
2011-12-14 09:56:46 -08:00
1911905c6b
vm: frame size for signal handler code blocks
2011-12-14 09:56:46 -08:00
0b94018d9f
vm: initialize all_blocks set after initial fixup
2011-12-14 09:56:46 -08:00
3868895b56
vm: make all_blocks a set of cells
...
gcc was doing bad things assuming the pointer given to upper_bound was aligned
2011-12-14 09:56:46 -08:00
36d1e0a78c
vm: verifier shouldn't check for leftovers
...
It doesn't work after things have been freed
2011-12-14 09:56:45 -08:00
d6227c182b
vm: pack frame size into code block header bits
2011-12-14 09:56:45 -08:00
e21f316583
vm: store stack frame size in code blocks
...
Change modify-code-heap primitive so it takes a sixth element in each array for the frame size.
2011-12-14 09:56:45 -08:00
bd7e6b33f8
vm: erase code blocks from all_blocks during sweep
...
Fixes #437
2011-12-13 12:28:09 -08:00
083b856604
vm: reinitialize all_blocks from scratch after gc
...
Faster and more likely to be correct than trying to fixup the existing set
2011-12-12 19:05:48 -08:00
4bbb2f5b72
vm: only verify all_blocks set if it's invalid
...
Verifying on every sample is way too slow to be practical when debugging bootstrap.
2011-12-12 19:04:56 -08:00
4870ac8f1d
vm: add compact-gc command to factorbug
2011-12-12 19:03:42 -08:00
dea0bedb99
vm: fix msvc warning
2011-11-29 12:19:17 -08:00
26226d460e
vm: factor out embedded image detection code
2011-11-27 14:39:21 -08:00
cca526df73
vm: support self-executing image file
2011-11-27 12:37:54 -08:00
88d31b166b
vm: don't die if foreign thread gets async signal
...
Fixes #425
2011-11-23 11:26:42 -08:00
065a4d4353
vm: remove useless forward declare
2011-11-23 11:11:28 -08:00
b6246f2c00
vm: quell warnings from vars only used by asserts
2011-11-23 11:11:28 -08:00
4df56f31f2
vm: clean up baroque jit safepoint logic
2011-11-23 11:11:27 -08:00
15c4b7aa57
vm: verify code heap's all_blocks set when DEBUG
2011-11-23 11:11:27 -08:00
c2c2023e34
vm: FACTOR_ASSERT only if FACTOR_DEBUG
2011-11-23 11:11:27 -08:00
0a1a252cdb
vm: replace assert with FACTOR_ASSERT
...
Factor is finally a real C++ project and has a custom assert macro. Assertion failures were still getting caught as exceptions and causing failure loops. Write our own macro that calls factor::abort on failure.
2011-11-23 11:11:26 -08:00
6c0ecf631a
vm: charge samples collected in prolog to parent
...
Fixes #379
2011-11-23 11:11:26 -08:00
81911b9f87
vm: initialize all_blocks map from image
...
um, code blocks from the image need to go in the set too
2011-11-23 11:11:26 -08:00
2705fc67cb
vm: keep a set of all code_blocks in the code_heap
...
This way looking up code blocks by return address is an O(log n) set lookup rather than an O(n) linear scan.
2011-11-23 11:11:26 -08:00
44b608bb22
vm: emit safepoints before VM compiler prologs
...
For #379
2011-11-23 11:11:25 -08:00
5e99a7bbbf
vm: disable SIGINT handler while in factorbug
...
Fixes #410
2011-11-15 18:50:42 -08:00
c4885ac082
vm: fashion police
2011-11-12 23:05:44 -08:00
c0a44000cc
vm: set stdin_loop cancellation state
...
Make it asynchronous so it doesn't wait till the next syscall to cancel.
2011-11-12 23:05:44 -08:00
3fc82282f7
vm: close_console before exiting anywhere
...
Add a factor_vm::exit function that does close_console() before ::exit-ing
2011-11-12 23:05:44 -08:00
db19efe1b3
vm: close the console when exiting
...
This prevents an endless loop caused by the nvidia drivers on linux.
See http://www.nvnews.net/vbulletin/showthread.php?t=164619
2011-11-12 23:05:43 -08:00
8e7baef1a0
vm: typo in os-windows.hpp
2011-11-12 19:23:44 -08:00
e0dc236725
cpu.*.bootstrap: clear faulting flag directly
...
No reason to go through a safepoint in unwind-native-frames really.
2011-11-12 13:02:57 -08:00
e24400679f
vm: more defense against multi-faulting
...
* Clear faulting_p from a safepoint rather than inside general_error, because jumping into unwind-native-frames could blow up.
* Handle multiple faults from fatal_error by breakpointing. Is there anything else we can safely do at that point?
* Verify memory protection faults in the top half of the signal handlers because signal dispatch could fault. Treat memory faults during gc or fep as fatal errors.
* Add a function factor_vm::abort() that restores the default SIGABRT handler and ::abort()s. Use it from fatal_error() so we get useful context from gdb and so the user gets feedback from the system crash reporter that Factor blew up and didn't just disappear.
* In factorbug(), don't proceed with .s .r .c if it would be unsafe to do so.
* Don't pile on signals if we've already called fatal_error().
2011-11-12 13:02:57 -08:00
2c5014e74e
vm: "edges" command for factorbug
2011-11-11 10:12:36 -08:00
c31530caec
vm: strip out call-counting profiler
...
This makes the separate "code" and "entry_point" fields in word and quotation redundant, so also remove them to reclaim an additional cell per word and quotation object, which should help with #318 .
2011-11-10 16:01:07 -08:00
139c9ca38c
init signals or console after stage2 init
...
A signal during early bootstrap is certainly an error, and the environment isn't there yet for Factor's signal handlers to work, so it makes no sense to install Factor's signal handlers beforehand. They'll just crash.
2011-11-10 16:01:07 -08:00
9e4a51c55f
vm: don't SIGQUIT on unix or Ctrl-Break on windows
...
This gives us an escape hatch for when things go way south and ^C can't be handled. Also unmask SIGQUIT and SIGTERM from the stdin_loop thread so the thread doesn't keep the process alive after those signals.
2011-11-10 16:01:07 -08:00
2e9552d98a
vm: hand SIGALRM to signal pipe if not sampling
2011-11-08 23:37:22 -08:00
c0094186a1
vm: fix build on windows
2011-11-08 23:17:51 -08:00
5f3cd4079c
vm: make signal pipe nonblocking
...
On the off chance the Factor end of the signal handler mechanism isn't listening, we don't want to tie up the VM because the pipe's buffer got full.
2011-11-08 22:07:22 -08:00
820fd6417d
vm: don't send foreign samples to outer space
2011-11-08 14:00:33 -08:00
90609cc5d8
vm: don't pass this to safepoint constructor
...
It's bad juju and MSVC calls us out on it.
2011-11-08 11:06:00 -08:00
fc7c175c9f
new module unix.signals: app-level signal handlers
...
example:
IN: scratchpad USING: unix.ffi unix.signals ;
Loading resource:basis/unix/signals/signals.factor
IN: scratchpad [ V{ "Me not that kind of orc!" "Me busy, leave me alone!" "Work work" "Zug zug" } pop print flush ] SIGINFO add-signal-handler
IN: scratchpad load: 0.60 cmd: factor 41541 running 0.28u 0.16s
Zug zug
load: 0.71 cmd: factor 41541 running 0.28u 0.16s
Work work
load: 0.71 cmd: factor 41541 running 0.28u 0.16s
Me busy, leave me alone!
load: 0.73 cmd: factor 41541 running 0.28u 0.16s
Me not that kind of orc!
2011-11-08 10:21:49 -08:00
0c58fd5c26
vm: more commentary on lock_console() stuff
2011-11-08 08:52:52 -08:00
5f811c0573
vm: don't mask SIGTTIN from stdin_loop
...
Fixes #360
2011-11-08 08:42:48 -08:00
f9aa0419c4
vm: interrupt stdin_loop when entering fep
...
The stdin_loop thread will keep trying to consume input unless we stop it by sending it a signal. Use SIGUSR2 to stop the read syscall and a mutex to hold up the loop while the fep is active.
2011-11-08 08:42:48 -08:00
04c6b2c393
vm: write signals to pipe immediately
...
The whole point is to wake up the mx without waiting for the next safepoint derp
2011-11-08 08:42:47 -08:00
0e3c315099
vm: move bignum.cpp includes to master.hpp
...
That way they get PCHed
2011-11-08 08:42:47 -08:00
18e01f7770
vm: enqueue signals for FEP signals too
...
Although SIGINT still can't interrupt the current Factor thread in a sane way, this will at least wake up the run loop when waiting for input and fix #348 when implemented at the application level.
2011-11-08 08:42:47 -08:00
b46fd5e655
GNUmakefile: use PCH so compiler goes faster
...
Also add more accurate dependencies on headers so "make" works right when headers change
2011-11-08 08:42:46 -08:00
72d335f2b5
vm: write async signals to pipe for multiplexer
...
also factor out safepoint logic into its own file
2011-11-08 08:42:46 -08:00
212d012cb1
vm: switch atomic::add to atomic::fetch_add
...
Win32 only provides fetch-and-add, not add-and-fetch, and we don't really care. Rename the functions so the order of operations is clear.
2011-11-07 17:58:53 -08:00
ec19171220
vm: readable feps for more types
2011-11-06 20:26:52 -08:00
f4479ee1e6
vm: print type names instead of numbers in fep
2011-11-06 20:26:51 -08:00
f0d4fcc587
switch back to gcc from clang since it goes faster
2011-11-04 00:09:30 -07:00
94eb922998
makefile: add macosx-x86-fat target
2011-11-03 20:19:37 -07:00
3cad288f90
use gcc-4.2 on OS X 10.5, or clang on Xcode 4
2011-11-03 12:18:07 -07:00
6bb46a3f1e
vm: always generate safepoints in jit
...
Even if there's no stack frame we still need to safepoint before leaving the function. Fixes #332 .
2011-11-03 11:57:43 -07:00
92445d77e2
vm: restore sigaltstack setup on unix
...
#ifdef __OpenBSD__ is different from #ifndef __OpenBSD
2011-11-03 11:57:43 -07:00
d28845a1ef
vm: reserve some more special-objects
...
Need a jit-safepoint object to deal with #332 , and ffi-signal-handler words for immediate signal handling during FFI calls
2011-11-02 22:16:08 -07:00
2d87b68a56
vm: sample pc on windows
2011-11-02 21:18:31 -07:00
1833beea52
vm: fix MSVC warnings
2011-11-02 20:26:53 -07:00
1b38835207
remove lingering mentions of macosx-ppc
2011-11-02 18:40:27 -07:00
8cf18d1a82
remove BSD,solaris,etc. misc platform support code
2011-11-02 17:13:51 -07:00
82a62181ce
vm: count samples during unoptimized compiler
2011-11-02 13:23:21 -07:00
e0c68d5a3f
vm: fix scrub-league assert abuse
2011-11-02 13:23:21 -07:00
2fc6b0b9be
vm: forgot to add atomic.hpp
2011-11-02 13:23:20 -07:00
73845fdad9
vm: separate 'current_gc_p' from 'current_gc'
...
The 'current_gc' handle doesn't need to be volatile, just the state of GC-ness
2011-11-02 13:23:19 -07:00
4d39e59054
vm: add atomic::load and atomic::store functions
...
Abstract out the fence-and-load and store-and-fence patterns
2011-11-02 13:23:19 -07:00
be8db5aed3
vm: sample thread object rather than raw context
2011-11-02 13:23:18 -07:00
7da8a9776f
vm: sample code block owners instead of blocks
2011-11-02 13:23:17 -07:00
e880095da3
vm: fix data roots in get-samples primitive
2011-11-02 13:23:17 -07:00
4037c981eb
vm: die if we fault in a fault
2011-11-02 13:23:16 -07:00
29fa23c40a
vm: quell MSVC warnings
2011-11-02 13:23:16 -07:00
30ed9b535b
vm: parameterize sampling rate
2011-11-02 13:23:15 -07:00
799053661e
vm: better organize atomics and sample counter
2011-11-02 13:23:15 -07:00
89ea69b61b
vm: fix comments mangled by find & replace
2011-11-02 13:23:14 -07:00
27ac4c60f4
vm: factor out sample counting code
2011-11-02 13:23:13 -07:00
baebb317e4
whitespace
2011-11-02 13:23:13 -07:00
34ea20beca
vm: only show factorbug help the first time
2011-11-02 13:23:13 -07:00
1a665f3eac
vm: more accurate sample timing on windows
2011-11-02 13:23:12 -07:00
d9afe92c97
get sampler working under windows
2011-11-02 13:23:12 -07:00
252915a37c
fix platform dependency in leaf-signal-handler
2011-11-02 13:23:12 -07:00
abaa051768
vm: fix build on windows
2011-11-02 13:23:12 -07:00
1351e30e52
vm: sampling profiler timer for windows
...
Spawn a thread and have it spin on the performance counter, triggering safepoints on the main thread every sample time
2011-11-02 13:23:11 -07:00
125c46c10f
vm: stub out windows sampling timer implementation
...
AFAIK Windows has no equivalent to setitimer. We'll probably need to use a thread that periodically pauses and samples the main VM thread.
2011-11-02 13:23:10 -07:00
f5cdfaeb2a
vm: count samples from foreign threads
...
On OS X it appears the mach exception thread and libdispatch queue threads occasionally get the SIGALRM from the itimer. Count those as foreign_thread_samples instead of letting the signal kill the process.
2011-11-02 13:23:10 -07:00
27d7e4af53
vm: actually collect samples and add primitives...
2011-11-02 13:23:10 -07:00
e10c3183c8
vm: add primitives to lift sample data to factor
2011-11-02 13:23:09 -07:00
3dd17537eb
vm: write back fixed-up code address to samples
2011-11-02 13:23:09 -07:00
e3e2788414
vm: walk code roots from sample callstacks
2011-11-02 13:23:08 -07:00
1089816b21
vm: release memory used by sampler vectors
2011-11-02 13:23:08 -07:00
40bc8459db
vm: bottom_frame method for contexts
...
ctx->callstack_bottom - 1 -- the fuck does that mean?!
2011-11-02 13:23:07 -07:00
be5c7476d7
vm: don't allocate callstack samples on data heap
...
Invoking a GC during a safepoint has some complications, so stuff the callstack samples in a vector instead.
2011-11-02 13:23:07 -07:00
b341b0ce13
vm: only show factorbug basic commands at first
...
"help" brings up the entire command list.
2011-11-02 13:23:07 -07:00
80fb7788d9
vm: groundwork for sampling profiler
...
Set up the state necessary to collect samples. We still need to add GC support for walking the sample set, and the compiler needs to register GC roots before safepoints as well. We also need primitives to expose the data to Factor for reporting.
2011-11-02 13:23:06 -07:00
Joe Groff
Philipp Brüschweiler