[llvm-bugs] [Bug 28980] New: memcpy without llvm.mem.parallel_loop_access metadata in loop annotated with #pragma clang loop vectorize(assume_safety)
via llvm-bugs
llvm-bugs at lists.llvm.org
Mon Aug 15 04:15:37 PDT 2016
https://llvm.org/bugs/show_bug.cgi?id=28980
Bug ID: 28980
Summary: memcpy without llvm.mem.parallel_loop_access metadata
in loop annotated with #pragma clang loop
vectorize(assume_safety)
Product: clang
Version: trunk
Hardware: All
OS: Linux
Status: NEW
Severity: normal
Priority: P
Component: LLVM Codegen
Assignee: unassignedclangbugs at nondot.org
Reporter: dorit.nuzman at intel.com
CC: llvm-bugs at lists.llvm.org
Classification: Unclassified
Created attachment 16949
--> https://llvm.org/bugs/attachment.cgi?id=16949&action=edit
Example .cpp and output of clang for that .cpp where memcpy is missing the
mem_par metadata
Clang doesn't attach a llvm.mem.parallel_loop_access metadata for memcpy
instructions in a loop that is annotated with
#pragma clang loop vectorize(assume_safety):
Example1:
#pragma clang loop vectorize(assume_safety)
for (long i = 0; i < size; i+=2) {
memcpy(&(out[i]), &(out[i+size]), 2);
}
Example2: Here the struct copies result in a memcpy:
void test(Complex *out, long size)
{
long m = size / 2;
#pragma clang loop vectorize(assume_safety)
for (long offset = 0; offset < m; ++offset) {
Complex t0 = out[offset];
Complex t1 = out[offset+m];
out[offset] = t0 + t1;
out[offset+m] = t0 - t1;
}
}
.c and .ll created by clang attached.
Generated the .ll like so:
clang++ -S -emit-llvm -mllvm -disable-llvm-optzns [-g -m64 -static -mfpmath=sse
-march=core-avx2]
--
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/20160815/f9fcd057/attachment.html>
More information about the llvm-bugs
mailing list