Replies: 5 comments 3 replies
-
Good new ! Following a discussion here : On OSX, with :
GDL> TEST_LOOPS
|
Beta Was this translation helpful? Give feedback.
-
@alaingdl very good find! But then with the fsanitize=address flag set, at exit the program GDL should report a list of leaks. And, it should run slower than without the fsanitize option since asan adds tests and make it slower ? |
Beta Was this translation helpful? Give feedback.
-
fsanitize is just a doctor, as Valgrind. However, as it replaces the memory allocation code by its own, and it was suggested in #1635 that the slowness could come from allocation of objects, perhaps this replacement is beneficial. |
Beta Was this translation helpful? Give feedback.
-
@alaingdl under linux there is abolutely no difference in the list of memory leaks (using #1762 to get clearer outputs, and
and
to be tried on OSX of course. |
Beta Was this translation helpful? Give feedback.
-
test_loops.pro
)Hi
The problem "discovered" on M2 processors + OSX is in fact general on OSX, including intel x86_64
(see also : #1755 #1498 )
No relation to OpenMP so far
On OSX, GDL is now compiled with Clang.
I succeed to compile an old version of GDL with GCC.
Same kind of bad preformances :(
best regards
Alain
*** My laptop (u2204 old cpu, last GDL)
*** a very good (16 cores) OSX with x86_64 proc (OSX 12.3.1; 16 cores)
Beta Was this translation helpful? Give feedback.
All reactions