[LLVMbugs] [Bug 17677] New: [Vectorizer] Implement stride access

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=17677

            Bug ID: 17677
           Summary: [Vectorizer] Implement stride access
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: Loop Optimizer
          Assignee: unassignedbugs at nondot.org
          Reporter: renato.golin at linaro.org
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

Implement stride access detection for induction variables with Step > 1, with
cost-model support for the (currently unclear) benefits, and make sure that the
vectorization happens later on.

The basic loop in which this should work is:

for (i..N/3) {
  a[3*i] = b[3*i];
  a[3*i+1] = b[3*i+1];
  a[3*i+2] = b[3*i+2];
}

With S = sizeof(a[0]);

Since the loop stride is { 3 * S } and the evolutions are:

{a, +, 3S}
{a+S, +, 3S}
{a+2S, +, 3S}

The reads/writes don't overlap, and they're sequential, so we can vectorize.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20131024/57cce4e1/attachment.html>