<html><head><base href="x-msg://48/"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; "><br><div><div>On Oct 18, 2011, at 10:00 AM, Gabriel Rodríguez wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><span class="Apple-style-span" style="border-collapse: separate; font-family: Helvetica; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; orphans: 2; text-align: -webkit-auto; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-border-horizontal-spacing: 0px; -webkit-border-vertical-spacing: 0px; -webkit-text-decorations-in-effect: none; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px; font-size: medium; "><div><div style="font-family: 'Times New Roman'; font-size: 12pt; color: rgb(0, 0, 0); "><font class="Apple-style-span" size="3">Dear All,</font><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; "><br></div><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; ">As of late I am having a hard time getting my head around how array accesses</div><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; ">are translated by Clang into LLVM IR: the often misunderstood GEP instruction.</div><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; ">I am trying to reverse-engineer array accesses to discover the number of dimensions</div><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; ">and actual indexes of the original, and I am beginning to wonder whether this is</div><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; ">possible at all. To illustrate (some of) my troubles, consider the following code and</div><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; ">the LLVM IR for both 32 and 64 bit memory addresses:</div><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; "><br></div><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; ">--</div><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; ">original C:</div><div style="color: rgb(0, 0, 0); font-family: 'Times New Roman'; font-size: 12pt; "><br></div><div><div>#define N 1000</div><div><br></div><div>int main(int argc, char **argv)</div><div>{</div><div> int i, k;</div><div> float aux, A[N][N];</div></div></div></div></span></blockquote><div><br></div><div>See "How does VLA addressing work with GEPs?" in the GEP FAQ. In short, you have to</div><div>reverse-engineer, and the ScalarEvolution library can help you with that.</div><div><br></div><div>Also, one thing not mentioned in the FAQ is that if you want to assume that the dimensions</div><div>are independent (in other words, that the inner dimension is never over-indexed), you</div><div>have to prove it for yourself. Even though overindexing may be prohibited at the source level,</div><div>it's valid at the LLVM IR level, and some LLVM optimizations do use it. ScalarEvolution</div><div>can help with this as well, though it doesn't do everything.</div><div><br></div><div>Dan</div><div><br></div></div></body></html>