| 1 | | log2(double x) on Tiger is supposed to return a double, not a float. So as it is broken on i386 I saw the uint32_t overflowing the int32 on i386 Tiger being the issue, as a uint32_t does not fit into 32bits. |
| 2 | | |
| 3 | | But perhaps it uses a float internally and then does a poor internal conversion to a double. log2(4096) is exactly 12, so that would be really really bad if so, but ... it's Tiger! |
| | 1 | log2(double x) on Tiger is supposed to return a double, but perhaps it uses a float internally and then does a poor internal conversion to a double. |
