<html>
<head>
<base href="https://bugs.llvm.org/">
</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 - Unrolled loop generates worse code"
href="https://bugs.llvm.org/show_bug.cgi?id=49955">49955</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>Unrolled loop generates worse code
</td>
</tr>
<tr>
<th>Product</th>
<td>clang
</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>enhancement
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>-New Bugs
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedclangbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>davidfromonline@gmail.com
</td>
</tr>
<tr>
<th>CC</th>
<td>htmldeveloper@gmail.com, llvm-bugs@lists.llvm.org, neeilans@live.com, richard-llvm@metafoo.co.uk
</td>
</tr></table>
<p>
<div>
<pre>The following code
```
#include <vector>
void f() {
auto v = std::vector<int>();
v.reserve(2);
#if defined LOOP
for (int n = 0; n != 2; ++n) {
v.push_back(0);
}
#else
v.push_back(0);
v.push_back(0);
#endif
}
```
When compiled with `clang++ -stdlib=libc++ -O3 -DLOOP` (the loop version)
compiles down to just
```
f(): # @f()
ret
```
But when compiled with `clang++ -stdlib=libc++ -O3` (the unrolled version)
generates
```
f(): # @f()
push rbx
sub rsp, 32
xorps xmm0, xmm0
movaps xmmword ptr [rsp], xmm0
mov qword ptr [rsp + 16], 0
mov edi, 8
call operator new(unsigned long)
mov rcx, rax
add rcx, 8
mov qword ptr [rsp], rax
mov qword ptr [rsp + 8], rax
mov qword ptr [rsp + 16], rcx
mov qword ptr [rax], 0
mov rdi, rax
add rsp, 32
pop rbx
jmp operator delete(void*) # TAILCALL
mov rbx, rax
mov rdi, qword ptr [rsp]
test rdi, rdi
je .LBB0_4
mov qword ptr [rsp + 8], rdi
call operator delete(void*)
.LBB0_4: #
%_ZNSt3__16vectorIiNS_9allocatorIiEEED2Ev.exit69
mov rdi, rbx
call _Unwind_Resume@PLT
```
See it live: <a href="https://godbolt.org/z/3MY3YMME6">https://godbolt.org/z/3MY3YMME6</a>
A similar (but less dramatic) example can be seen in the following much reduced
code:
```
int buffer[3];
struct vector {
int * ptr = buffer;
long size = 0;
long capacity = 0;
};
auto a() {
auto v = vector();
#if defined LOOP
for (int n = 0; n != 3; ++n) {
v.ptr[n] = 0;
}
v.size = 3;
#elif defined LOOP_ASSIGN_INSIDE
for (int n = 0; n != 3; ++n) {
v.ptr[n] = 0;
v.size = 3;
}
#else
v.ptr[0] = 0;
v.ptr[1] = 0;
v.ptr[2] = 0;
v.size = 3;
#endif
return v;
}
```
When compiled with `clang++ -O3 -DLOOP` this gives
```
a(): # @a()
mov rax, rdi
mov qword ptr [rdi], offset buffer
mov qword ptr [rdi + 16], 0
mov dword ptr [rip + buffer+8], 0
mov qword ptr [rip + buffer], 0
mov qword ptr [rdi + 8], 3
ret
buffer:
.zero 12
```
When compiled with `clang++ -O3 -DLOOP_ASSIGN_INSIDE` gives
```
.LCPI0_0:
.long 3 # 0x3
.long 0 # 0x0
.long 0 # 0x0
.long 0 # 0x0
a(): # @a()
mov rax, rdi
mov qword ptr [rdi], offset buffer
movaps xmm0, xmmword ptr [rip + .LCPI0_0] # xmm0 = [3,0,0,0]
movups xmmword ptr [rdi + 8], xmm0
mov dword ptr [rip + buffer+8], 0
mov qword ptr [rip + buffer], 0
ret
buffer:
.zero 12
```
When compiled with `clang++ -O3` gives
```
.LCPI0_0:
.long 3 # 0x3
.long 0 # 0x0
.long 0 # 0x0
.long 0 # 0x0
a(): # @a()
mov rax, rdi
mov qword ptr [rdi], offset buffer
mov qword ptr [rip + buffer], 0
mov dword ptr [rip + buffer+8], 0
movaps xmm0, xmmword ptr [rip + .LCPI0_0] # xmm0 = [3,0,0,0]
movups xmmword ptr [rdi + 8], xmm0
ret
buffer:
.zero 12
```
See it live <a href="https://godbolt.org/z/8EG5db4fc">https://godbolt.org/z/8EG5db4fc</a></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>