[LLVMbugs] [Bug 17679] New: [Vectorizer] Implement interleaved stride vectorization

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=17679

            Bug ID: 17679
           Summary: [Vectorizer] Implement interleaved stride
                    vectorization
           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

After implementing the non-interleaved vectorization, detect interleaved
patterns, such as:

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

Where you can unroll this into:

for (i..N/3) {
 a[3*i] = b[3*i] + I;
 a[3*i+3] = b[3*i+3] + I;
 a[3*i+6] = b[3*i+6] + I;
 a[3*i+9] = b[3*i+9] + I;

 a[3*i+1] = b[3*i+1] + J;
 a[3*i+4] = b[3*i+4] + J;
 a[3*i+7] = b[3*i+7] + J;
 a[3*i+10] = b[3*i+10] + J;

 a[3*i+2] = b[3*i+2] + K;
 a[3*i+5] = b[3*i+5] + K;
 a[3*i+8] = b[3*i+8] + K;
 a[3*i+11] = b[3*i+11] + K;
}

And use interleaved reads/writes, for example ARM's VLDN/VSTN.

-- 
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/20131024/1a9095ef/attachment.html>