<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 --- - Two Addr Instruction pass could be improved further"
href="http://llvm.org/bugs/show_bug.cgi?id=22689">22689</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>Two Addr Instruction pass could be improved further
</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>Linux
</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>Common Code Generator Code
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>wmi@google.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>A problem was identified in TwoAddrInstruction pass and a patch was propsed in
<a href="http://reviews.llvm.org/D7806">http://reviews.llvm.org/D7806</a>.
However, there are other cases that the patch cannot handle currently. Like the
following one:
1.c:
int M, total;
int A[100], B[100], C[100];
volatile int g;
void goo(int);
void foo() {
int i, a, b, x, y, z;
a = 3;
for (i = 0; i < M; i++) {
b = i;
z = g;
y = b * a;
x = y + z;
a = g;
goo(b);
total += x;
}
}
~/workarea/llvm-r230041/build/bin/clang -O2 -fno-vectorize -fno-unroll-loops -S
1.c
kernel loop in 1.s:
.LBB0_2: # %for.body
movl g(%rip), %eax
imull %ebx, %ebp
addl %eax, %ebp
movl g(%rip), %r14d
movl %ebx, %edi
callq goo
addl %ebp, total(%rip)
incl %ebx
cmpl M(%rip), %ebx
movl %r14d, %ebp
jl .LBB0_2
The reg copy insn "movl %r14d, %ebp" could be removed if we switch the
positions of "z = g;" and "y = b * a;" in 1.c
2.c:
int M, total;
int A[100], B[100], C[100];
volatile int g;
void goo(int);
void foo() {
int i, a, b, x, y, z;
a = 3;
for (i = 0; i < M; i++) {
b = i;
y = b * a; // here is the difference with 1.c
z = g; // here is the difference with 1.c
x = y + z;
a = g;
goo(b);
total += x;
}
}
~/workarea/llvm-r230041/build/bin/clang -O2 -fno-vectorize -fno-unroll-loops -S
2.c
kernel loop in 2.s:
.LBB0_2: # %for.body
imull %ebx, %ebp
movl g(%rip), %r14d
addl %ebp, %r14d
movl g(%rip), %ebp
movl %ebx, %edi
callq goo
addl %r14d, total(%rip)
incl %ebx
cmpl M(%rip), %ebx
jl .LBB0_2
The bug is better to be fixed incrementally, so filed this bug to track the
problem.</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>