• forkit (13/13) Jan 04 2022 strange things happen when I compile (on windows) the code below
• H. S. Teoh (14/20) Jan 04 2022 I bet you it's the same problem I found a couple of years ago, where the
• H. S. Teoh (19/22) Jan 04 2022 Or the compiler would run out of memory before it gets to optimizing
• forkit (9/12) Jan 04 2022 yes, this sounds like it might be it.
• forkit (3/19) Jan 05 2022 Well, in my case, it was nice to see that the oom reaper thread
• H. S. Teoh (9/11) Jan 05 2022 I'm well-acquainted with the OOM reaper; it and dmd are good friends,

forkit <forkit gmail.com> writes:

strange things happen when I compile (on windows) the code below 
with:

dmd -m64
(compilation just crashes - no error message at all)

or

ldc2 -m64
(compilation works, but memory usage during compilation goes 
completely wild! upto 22GB, then down to 7GB, then finally 
completes.)


// ----

module test;
void main(){ static char[2147483646] arr; }

// ----

"H. S. Teoh" <hsteoh quickfur.ath.cx> writes:

On Tue, Jan 04, 2022 at 10:53:50PM +0000, forkit via Digitalmars-d-learn wrote:
[...]
 // ----
 
 module test;
 void main(){ static char[2147483646] arr; }
 
 // ----

I bet you it's the same problem I found a couple of years ago, where the
compiler translates the above code to something that individually
initializes every element of the array. I.e., there would be 2.1 billion
assignments, one for each element in the array.  Of course, afterwards
the optimizer would merge them into something saner, but while the
compiler is unfolding all those assignments, its memory usage would
obviously skyrocket.

Try setting `arr = void;` to see if it makes a difference. If it does,
it's probably the same problem.


T

-- 
Skill without imagination is craftsmanship and gives us many useful objects
such as wickerwork picnic baskets.  Imagination without skill gives us modern
art. -- Tom Stoppard

"H. S. Teoh" <hsteoh quickfur.ath.cx> writes:

On Tue, Jan 04, 2022 at 03:34:41PM -0800, H. S. Teoh wrote:
[...]
 Of course, afterwards the optimizer would merge them into something
 saner, but while the compiler is unfolding all those assignments, its
 memory usage would obviously skyrocket.

Or the compiler would run out of memory before it gets to optimizing
away those assignments, so it would just outright crash. I ran your code
on my computer and got this:

	uncaught exception
	core.exception.AssertError src/dmd/common/outbuffer.d(204): OutBuffer: out of
memory.
	----------------
	??:? _d_assert_msg [0x5616e935648e]
	??:? _ZN9OutBuffer7reserveEm [0x5616e9350bda]
	??:? nothrow  trusted void dmd.common.outbuffer.OutBuffer.position(ulong,
ulong) [0x5616e935129c]
	??:? _Z11ElfObj_termPKc [0x5616e933bcd6]
	Aborted

Which seems to confirm my suspicions.


T

-- 
They say that "guns don't kill people, people kill people." Well I think
the gun helps. If you just stood there and yelled BANG, I don't think
you'd kill too many people. -- Eddie Izzard, Dressed to Kill

forkit <forkit gmail.com> writes:

On Tuesday, 4 January 2022 at 23:37:57 UTC, H. S. Teoh wrote:
 ... .. .

 Which seems to confirm my suspicions.


 T

yes, this sounds like it might be it.

I tried using = void;

.. results:

dmd -m64 -> just results in an 'out of memory' message (but at 
least i get a message this time)

ldc2 -m64 -> again, compiles ok, but this time memory usage 
during compilation never exceeds about 2.1GB (compared to 22GB 
when not using =void;

forkit <forkit gmail.com> writes:

On Tuesday, 4 January 2022 at 23:37:57 UTC, H. S. Teoh wrote:
 Or the compiler would run out of memory before it gets to 
 optimizing away those assignments, so it would just outright 
 crash. I ran your code on my computer and got this:

 	uncaught exception
 	core.exception.AssertError src/dmd/common/outbuffer.d(204): 
 OutBuffer: out of memory.
 	----------------
 	??:? _d_assert_msg [0x5616e935648e]
 	??:? _ZN9OutBuffer7reserveEm [0x5616e9350bda]
 	??:? nothrow  trusted void 
 dmd.common.outbuffer.OutBuffer.position(ulong, ulong) 
 [0x5616e935129c]
 	??:? _Z11ElfObj_termPKc [0x5616e933bcd6]
 	Aborted

 Which seems to confirm my suspicions.


 T

Well, in my case, it was nice to see that the oom reaper thread 
is working correctly  ;-)

"H. S. Teoh" <hsteoh quickfur.ath.cx> writes:

On Wed, Jan 05, 2022 at 08:00:45AM +0000, forkit via Digitalmars-d-learn wrote:
[...]
 Well, in my case, it was nice to see that the oom reaper thread is
 working correctly  ;-)

I'm well-acquainted with the OOM reaper; it and dmd are good friends,
and love to throw parties esp. when CTFE and Recursive Templates are in
town.

:-P


T

-- 
What doesn't kill me makes me stranger.