GCC 16 doesn't optimize away the unused return value from mpz_class
post-increment in code like "mpz_class a = 1; a++;".
I'm using gcc 16.0.0 on Cygwin. Example code below. Example compilation:
"g++ -o gmp_test -std=c++26 -O3 -march=native -W -Wall gmp_test.cpp
-L/usr/local/lib/ -lgmpxx -lgmp"
I am not sure what options there are. I would be very happy with
[[nodiscard]] as an improvement, or a warning in GMP C++ documentation
that a++ generates a temporary copy.
Martin
#include <gmpxx.h>
void* (*alloc_func_ptr)(size_t);
size_t alloc_count = 0;
void* allocate_function(size_t alloc_size) {
alloc_count++;
return alloc_func_ptr(alloc_size);
}
int main() {
mp_get_memory_functions(&alloc_func_ptr, NULL, NULL);
mp_set_memory_functions(allocate_function, NULL, NULL);
mpz_class a = 1;
a <<= 1000000;
alloc_count = 0;
for (size_t k = 0; k < 1000000; k++) {
a++;
}
printf("Loop allocations %zu\n", alloc_count);
}
_______________________________________________
gmp-bugs mailing list
[email protected]
https://gmplib.org/mailman/listinfo/gmp-bugs
