Hi,<div><br></div><div>I just noticed this using Christophe Gohlke&#39;s MKL builds of numpy:</div><div><br></div><div><div>&gt;&gt;&gt; import numpy as np</div><div>&gt;&gt;&gt; val = 2**63 + 2**62</div><div>&gt;&gt;&gt; np.float64(val)</div><div>1.3835058055282164e+19</div><div>&gt;&gt;&gt; np.float64(val).astype(np.uint64)</div><div>9223372036854775808</div></div><div><br></div><div>In general it seems that floats get clipped at 2**63 when casting to uint64.   This appears to be a bug in MSVS express 2010 (the only version I tested):</div><div><br></div><div>&lt;test_cast.c&gt;</div><div><div>#include &lt;stdio.h&gt;</div><div>#include &lt;math.h&gt;</div><div><br></div><div>int main(int argc, char* argv[]) {</div><div>    double fval = pow(2, 63) + pow(2, 11);</div><div>    double fval2;</div><div>    unsigned long long int ival = fval;</div><div>    fval2 = ival;</div><div>    printf(&quot;Float %f\n&quot;, fval);</div><div>    printf(&quot;Integer %f\n&quot;, fval2);</div><div>    printf(&quot;sizeof ulong %u\n&quot;, sizeof(unsigned long long int));</div><div>}</div></div><div>&lt;/test_cast.c&gt;</div><div><br></div><div><div>Z:\&gt;test_cast.exe</div><div>Float 9223372036854777900.000000</div><div>Integer 9223372036854775800.000000</div><div>sizeof ulong 8</div></div><div><br></div><div>I realize there&#39;s nothing much numpy can do about this, just thought I&#39;d let y&#39;all know.</div><div><br></div><div>Cheers,</div><div><br></div><div>Matthew</div>