Open
Description
Given the following program:
#include <cmath>
#include <cstdio>
void fn5() {
static constexpr int the_value_count = 64;
float the_large_array_1[the_value_count]{};
for (int ivalue = 0; ivalue < the_value_count; ++ivalue) {
the_large_array_1[ivalue] = std::sin(ivalue);
}
for (int ivalue = 0; ivalue < the_value_count; ++ivalue) {
float ref = std::sin(static_cast<float>(ivalue));
if (ref != the_large_array_1[ivalue])
printf("%d %f %f\n", ivalue, static_cast<double>(ref),
static_cast<double>(the_large_array_1[ivalue]));
}
}
int main() { fn5(); }I get the output:
ASM generation compiler returned: 0
Execution build compiler returned: 0
Program returned: 0
2 0.909297 -0.166667
3 0.141120 -0.158529
6 -0.279415 -0.160633
7 0.656987 -0.153471
10 -0.544021 -0.150545
11 -0.999990 -0.165169
14 0.990607 -0.164014
15 0.650288 -0.165107
18 -0.750987 -0.165957
19 0.149877 -0.153296
22 -0.008851 -0.155979
23 -0.846220 -0.158669
26 0.762558 -0.156295
27 0.956376 -0.166520
30 -0.988032 -0.166041
31 -0.404038 -0.162333
34 0.529083 -0.163847
35 -0.428183 -0.147580
38 0.296369 -0.150162
39 0.963795 -0.162641
42 -0.916522 -0.160875
43 -0.831775 -0.166455
46 0.901788 -0.166664
47 0.123573 -0.158245
50 -0.262375 -0.160386
51 0.670229 -0.153819
54 -0.558789 -0.150925
55 -0.999755 -0.165291
58 0.992873 -0.164176
59 0.636738 -0.164977
62 -0.739181 -0.165869
63 0.167356 -0.152942
Check it out at:
https://godbolt.org/z/ad55fr3b1
The program output is erroneous, the sin value is completely wrong. The last two columns should be similar.
The behavior changes if we merge the two loops.