<html>
<head>
<base href="http://llvm.org/bugs/" />
</head>
<body><table border="1" cellspacing="0" cellpadding="8">
<tr>
<th>Bug ID</th>
<td><a class="bz_bug_link
bz_status_NEW "
title="NEW --- - loop increment/compare/branch should be removed for vectorized and unrolled code"
href="http://llvm.org/bugs/show_bug.cgi?id=17803">17803</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>loop increment/compare/branch should be removed for vectorized and unrolled code
</td>
</tr>
<tr>
<th>Product</th>
<td>libraries
</td>
</tr>
<tr>
<th>Version</th>
<td>trunk
</td>
</tr>
<tr>
<th>Hardware</th>
<td>PC
</td>
</tr>
<tr>
<th>OS</th>
<td>All
</td>
</tr>
<tr>
<th>Status</th>
<td>NEW
</td>
</tr>
<tr>
<th>Severity</th>
<td>normal
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>Loop Optimizer
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>kkhoo@perfwizard.com
</td>
</tr>
<tr>
<th>CC</th>
<td>llvmbugs@cs.uiuc.edu
</td>
</tr>
<tr>
<th>Classification</th>
<td>Unclassified
</td>
</tr></table>
<p>
<div>
<pre>For this fixed-length loop:
$ cat xor.c
void foo(unsigned char *a) {
for (int i = 0; i < 128; i++) {
a[i] = 0x36 ^ a[i];
}
}
LLVM is unrolling and vectorizing, but fails to remove the loop
comparison/branch. The unrolled portion of the code will never loop:
$ ./clang -O3 -fomit-frame-pointer -o - -S xor.c -ffp-contract=fast
-march=core-avx2
.section __TEXT,__text,regular,pure_instructions
.section __TEXT,__const
.align 5
LCPI0_0:
.space 32,54
.section __TEXT,__text,regular,pure_instructions
.globl _foo
.align 4, 0x90
_foo: ## @foo
.cfi_startproc
## BB#0: ## %entry
xorl %eax, %eax
vmovaps LCPI0_0(%rip), %ymm0
.align 4, 0x90
LBB0_1: ## %vector.body
## =>This Inner Loop Header: Depth=1
vmovups (%rdi,%rax), %ymm1
vmovups 32(%rdi,%rax), %ymm2
vxorps %ymm0, %ymm1, %ymm1
vxorps %ymm0, %ymm2, %ymm2
vxorps 64(%rdi,%rax), %ymm0, %ymm3
vmovups %ymm1, (%rdi,%rax)
vmovups %ymm2, 32(%rdi,%rax)
vmovups %ymm3, 64(%rdi,%rax)
addq $96, %rax <--- %rax = 96 the first time through
cmpq $96, %rax <--- always true
jne LBB0_1
## BB#2: ## %for.body.preheader
xorl %eax, %eax
.align 4, 0x90
LBB0_3: ## %for.body
## =>This Inner Loop Header: Depth=1
xorb $54, 96(%rdi,%rax)
incq %rax
cmpq $32, %rax
jne LBB0_3
## BB#4: ## %for.end
vzeroupper
ret
This is with:
$ ./clang -v
clang version 3.4 (trunk 193529) (llvm/trunk 193527)
Target: x86_64-apple-darwin11.4.2
Thread model: posix</pre>
</div>
</p>
<hr>
<span>You are receiving this mail because:</span>
<ul>
<li>You are on the CC list for the bug.</li>
</ul>
</body>
</html>