On 19.03.2013 11:38, Zoltán Herczeg wrote:
> I am sorry, I misunderstood your question. It is about compilation
> not running the pattern.
Correct. The increase is on compilation, not on running the pattern.
> On my 64 bit Linux, the stack size reported by valgrind is:
I can not test on Linux. Win64 runs fine because it handles the stack
differently. The problem is with Win32 only which sets the stack size on
application start.
> r1239: 827264 r1295: 892576
>
> This is definitely an increase (by about 8%), but not serious
> increase. The binary is compiled by GCC and -O2 optimization level.
My compiler and environment are very different from yours. In particular
I do not have alloca(). I replaced the only call to alloca() by char
array[SLJIT_MAX_LOCAL_SIZE]; according to your recommendation a long
time ago.
I realize that this uses more stack than alloca() but so far it was
within reasonable bounds. The recent increase is quite huge in comparison.
