implement deeper compare #6
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Unfortunately, this causes leaks when comparing nested lists that I am currently unable to resolve. |
Merged
|
@sklam debugged it and the leaks are due to a bug in Numba but fixed in https://github.com/numba/numba/pull/4183/files. Will rebase now that |
This implements the operators <, <=, > and >= for the typed list. Lists with nested Numpy arrays can no longer be compared. This doesn't work in cpython also, so it isn't too bad. The error message is a bit cryptic however as this fails during lowering. Also, we now raise a TypingError if the thing being compared to is not also a typed-list. The problem here is that if we always just return False for a type-mismatch, then both a < b and a > b will be False, which doesn't make any sense. This might be a bit over-restrictive however, but should be fine for most use-cases.
|
Tested locally and looks good. |
esc
added a commit
that referenced
this pull request
May 10, 2024
This fixes a test, where a non-thread safe container is written to
during testing
To reproduce on at least `linux-64` and `osx-arm64` (and probably
others):
```
NUMBA_THREADING_LAYER=workqueue SUBPROC_TEST=1 ./runtests.py -m 32 numba.tests.test_parfors.TestPrangeSpecific.test_tuple_hoisting
```
On `linux-64`, this can be debugged with `gdb`:
```
(gdb) bt
closure____locals______listcomp____v15____v2build__list__0=...) at <string>:4438
(gdb) f 9
closure____locals______listcomp____v15____v2build__list__0=...) at <string>:4438
4438 <string>: No such file or directory.
(gdb) i args
sched = {meminfo = 0x0, parent = 0x0, nitems = 0, itemsize = 0, data = 0x0, shape = {0}, strides = {0}}
closure____locals______listcomp____v15____v2build__list__0 = {meminfo = 0x0, parent = 0x0}
(gdb) f 8
498 mi->data = NRT_Reallocate(mi->data, size);
(gdb) i args
mi = 0x21d1660
size = 2510933856
```
On `osx-arm64` we can use `lldb`:
```
(lldb) run runtests.py -m 32 numba.tests.test_parfors.TestPrangeSpecific.test_tuple_hoisting
Process 13575 launched: '/Users/esc/miniconda3-arm64/envs/numba_3.9/bin/python3' (arm64)
Parallel: 0. Serial: 1
python3(13575,0x17025b000) malloc: Non-aligned pointer 0x600000256880 being freed (2)
python3(13575,0x17025b000) malloc: *** set a breakpoint in malloc_error_break to debug
Process 13575 stopped
* thread numba#18, stop reason = signal SIGABRT
frame #0: 0x000000019c35c704 libsystem_kernel.dylib`__pthread_kill + 8
libsystem_kernel.dylib`:
-> 0x19c35c704 <+8>: b.lo 0x19c35c724 ; <+40>
0x19c35c708 <+12>: pacibsp
0x19c35c70c <+16>: stp x29, x30, [sp, #-0x10]!
0x19c35c710 <+20>: mov x29, sp
Target 0: (python3) stopped.
(lldb) bt
* thread numba#18, stop reason = signal SIGABRT
* frame #0: 0x000000019c35c704 libsystem_kernel.dylib`__pthread_kill + 8
frame #1: 0x000000019c393c28 libsystem_pthread.dylib`pthread_kill + 288
frame #2: 0x000000019c2a1ae8 libsystem_c.dylib`abort + 180
frame #3: 0x000000019c1c2e28 libsystem_malloc.dylib`malloc_vreport + 908
frame #4: 0x000000019c1d95d4 libsystem_malloc.dylib`malloc_zone_error + 104
frame #5: 0x000000019c1ca620 libsystem_malloc.dylib`_szone_free + 628
frame #6: 0x000000019c1b87f4 libsystem_malloc.dylib`nanov2_realloc + 356
frame #7: 0x000000019c1b85a4 libsystem_malloc.dylib`malloc_zone_realloc + 112
frame numba#8: 0x000000019c1b7110 libsystem_malloc.dylib`realloc + 388
frame numba#9: 0x000000013a9ff0f8 _nrt_python.cpython-39-darwin.so`NRT_MemInfo_varsize_realloc + 60
frame numba#10: 0x000000013d4f41e0
frame numba#11: 0x000000019c393fa8 libsystem_pthread.dylib`_pthread_start + 148
```
esc
added a commit
that referenced
this pull request
May 10, 2024
This fixes a test, where a non-thread safe container is written to
during testing
To reproduce on at least `linux-64` and `osx-arm64` (and probably
others):
```
NUMBA_THREADING_LAYER=workqueue SUBPROC_TEST=1 ./runtests.py -m 32 numba.tests.test_parfors.TestPrangeSpecific.test_tuple_hoisting
```
On `linux-64`, this can be debugged with `gdb`:
```
(gdb) bt
closure____locals______listcomp____v15____v2build__list__0=...) at <string>:4438
```
On `osx-arm64` we can use `lldb`:
```
(lldb) run runtests.py -m 32 numba.tests.test_parfors.TestPrangeSpecific.test_tuple_hoisting
Process 13575 launched: '/Users/esc/miniconda3-arm64/envs/numba_3.9/bin/python3' (arm64)
Parallel: 0. Serial: 1
python3(13575,0x17025b000) malloc: Non-aligned pointer 0x600000256880 being freed (2)
python3(13575,0x17025b000) malloc: *** set a breakpoint in malloc_error_break to debug
Process 13575 stopped
* thread numba#18, stop reason = signal SIGABRT
frame #0: 0x000000019c35c704 libsystem_kernel.dylib`__pthread_kill + 8
libsystem_kernel.dylib`:
-> 0x19c35c704 <+8>: b.lo 0x19c35c724 ; <+40>
0x19c35c708 <+12>: pacibsp
0x19c35c70c <+16>: stp x29, x30, [sp, #-0x10]!
0x19c35c710 <+20>: mov x29, sp
Target 0: (python3) stopped.
(lldb) bt
* thread numba#18, stop reason = signal SIGABRT
* frame #0: 0x000000019c35c704 libsystem_kernel.dylib`__pthread_kill + 8
frame #1: 0x000000019c393c28 libsystem_pthread.dylib`pthread_kill + 288
frame #2: 0x000000019c2a1ae8 libsystem_c.dylib`abort + 180
frame #3: 0x000000019c1c2e28 libsystem_malloc.dylib`malloc_vreport + 908
frame #4: 0x000000019c1d95d4 libsystem_malloc.dylib`malloc_zone_error + 104
frame #5: 0x000000019c1ca620 libsystem_malloc.dylib`_szone_free + 628
frame #6: 0x000000019c1b87f4 libsystem_malloc.dylib`nanov2_realloc + 356
frame #7: 0x000000019c1b85a4 libsystem_malloc.dylib`malloc_zone_realloc + 112
frame numba#8: 0x000000019c1b7110 libsystem_malloc.dylib`realloc + 388
frame numba#9: 0x000000013a9ff0f8 _nrt_python.cpython-39-darwin.so`NRT_MemInfo_varsize_realloc + 60
frame numba#10: 0x000000013d4f41e0
frame numba#11: 0x000000019c393fa8 libsystem_pthread.dylib`_pthread_start + 148
```
esc
added a commit
that referenced
this pull request
May 10, 2024
This fixes a test, where a non-thread safe container is written to during testing To reproduce on at least `linux-64` and `osx-arm64` (and probably others): ``` NUMBA_THREADING_LAYER=workqueue SUBPROC_TEST=1 ./runtests.py -m 32 numba.tests.test_parfors.TestPrangeSpecific.test_tuple_hoisting ``` On `linux-64`, this can be debugged with `gdb`: ``` (gdb) bt 0 0x00007ffff7c8018b in raise () from /lib/x86_64-linux-gnu/libc.so.6 1 0x00007ffff7c5f859 in abort () from /lib/x86_64-linux-gnu/libc.so.6 2 0x00007ffff7cca3ee in ?? () from /lib/x86_64-linux-gnu/libc.so.6 3 0x00007ffff7cd247c in ?? () from /lib/x86_64-linux-gnu/libc.so.6 4 0x00007ffff7cd412c in ?? () from /lib/x86_64-linux-gnu/libc.so.6 5 0x00007ffff7cd6105 in ?? () from /lib/x86_64-linux-gnu/libc.so.6 6 0x00007ffff7cd82d6 in realloc () from /lib/x86_64-linux-gnu/libc.so.6 7 0x00007fffe706de21 in NRT_Reallocate (ptr=0x224aef0, size=2552489952) at numba/core/runtime/nrt.cpp:539 8 0x00007fffe706dcf6 in NRT_MemInfo_varsize_realloc (mi=0x1fe9580, size=2552489952) at numba/core/runtime/nrt.cpp:498 9 0x00007fffdcabde0d in _3cdynamic_3e::__numba_parfor_gufunc_0x7fffdc60eb20[abi:v19][abi:c8tJTC_2fWQAliW1xhDEoY6EEMEUOEMISPGsAQMVj4QniQ4IXKQEMXwoMGLoQDDVsQR1NHAZtvoQrhyQ_2fKR8sTqKIYOQAmjYgkW7ADge6ERATM1UUQpZoA](Array<unsigned long long, 1, C, mutable, aligned>, list_28Tuple_28DictType_5bint64_2cfloat64_5d_3civ_3dNone_3e_2c_20array_28float64_2c_201d_2c_20C_29_29_29_3civ_3dNone_3e) (sched=..., closure____locals______listcomp____v15____v2build__list__0=...) at <string>:4438 10 0x00007fffdcab624e in __gufunc__._ZN13_3cdynamic_3e36__numba_parfor_gufunc_0x7fffdc60eb20B3v19B120c8tJTC_2fWQAliW1xhDEoY6EEMEUOEMISPGsAQMVj4QniQ4IXKQEMXwoMGLoQDDVsQR1NHAZtvoQrhyQ_2fKR8sTqKIYOQAmjYgkW7ADge6ERATM1UUQpZoAE5ArrayIyLi1E1C7mutable7alignedE119list_28Tuple_28DictType_5bint64_2cfloat64_5d_3civ_3dNone_3e_2c_20array_28float64_2c_201d_2c_20C_29_29_29_3civ_3dNone_3e () 11 0x00007fffdc926a3b in thread_worker (arg=0x1bd49c0) at numba/np/ufunc/workqueue.c:567 12 0x00007ffff7f8f609 in start_thread () from /lib/x86_64-linux-gnu/libpthread.so.0 13 0x00007ffff7d5c293 in clone () from /lib/x86_64-linux-gnu/libc.so.6 ``` On `osx-arm64` we can use `lldb`: ``` (lldb) run runtests.py -m 32 numba.tests.test_parfors.TestPrangeSpecific.test_tuple_hoisting Process 13575 launched: '/Users/esc/miniconda3-arm64/envs/numba_3.9/bin/python3' (arm64) Parallel: 0. Serial: 1 python3(13575,0x17025b000) malloc: Non-aligned pointer 0x600000256880 being freed (2) python3(13575,0x17025b000) malloc: *** set a breakpoint in malloc_error_break to debug Process 13575 stopped * thread numba#18, stop reason = signal SIGABRT frame #0: 0x000000019c35c704 libsystem_kernel.dylib`__pthread_kill + 8 libsystem_kernel.dylib`: -> 0x19c35c704 <+8>: b.lo 0x19c35c724 ; <+40> 0x19c35c708 <+12>: pacibsp 0x19c35c70c <+16>: stp x29, x30, [sp, #-0x10]! 0x19c35c710 <+20>: mov x29, sp Target 0: (python3) stopped. (lldb) bt * thread numba#18, stop reason = signal SIGABRT * frame #0: 0x000000019c35c704 libsystem_kernel.dylib`__pthread_kill + 8 frame #1: 0x000000019c393c28 libsystem_pthread.dylib`pthread_kill + 288 frame #2: 0x000000019c2a1ae8 libsystem_c.dylib`abort + 180 frame #3: 0x000000019c1c2e28 libsystem_malloc.dylib`malloc_vreport + 908 frame #4: 0x000000019c1d95d4 libsystem_malloc.dylib`malloc_zone_error + 104 frame #5: 0x000000019c1ca620 libsystem_malloc.dylib`_szone_free + 628 frame #6: 0x000000019c1b87f4 libsystem_malloc.dylib`nanov2_realloc + 356 frame #7: 0x000000019c1b85a4 libsystem_malloc.dylib`malloc_zone_realloc + 112 frame numba#8: 0x000000019c1b7110 libsystem_malloc.dylib`realloc + 388 frame numba#9: 0x000000013a9ff0f8 _nrt_python.cpython-39-darwin.so`NRT_MemInfo_varsize_realloc + 60 frame numba#10: 0x000000013d4f41e0 frame numba#11: 0x000000019c393fa8 libsystem_pthread.dylib`_pthread_start + 148 ```
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This implements the operators <, <=, > and >= for the typed list. Lists
with nested Numpy arrays can no longer be compared. This doesn't work in
cpython also, so it isn't too bad. The error message is a bit cryptic
however as this fails during lowering. Also, we now raise a TypingError
if the thing being compared to is not also a typed-list. The problem
here is that if we always just return False for a type-mismatch, then
both a < b and a > b will be False, which doesn't make any sense. This
might be a bit over-restrictive however, but should be fine for most
use-cases.