<html>
<head>
<base href="https://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 --- - Optimize memset for AVX2"
href="https://llvm.org/bugs/show_bug.cgi?id=27100">27100</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>Optimize memset for AVX2
</td>
</tr>
<tr>
<th>Product</th>
<td>new-bugs
</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>new bugs
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>hjl.tools@gmail.com
</td>
</tr>
<tr>
<th>CC</th>
<td>david.l.kreitzer@intel.com, llvm-bugs@lists.llvm.org, zia.ansari@intel.com
</td>
</tr>
<tr>
<th>Classification</th>
<td>Unclassified
</td>
</tr></table>
<p>
<div>
<pre>For
[hjl@gnu-6 memcpy-3]$ cat c.i
extern char *src, *dst;
void
foo3 (int x)
{
__builtin_memset (dst, x, 64);
}
void
foo4 (int x)
{
__builtin_memset (dst, x, 32);
}
void
foo5 (int x)
{
__builtin_memset (dst, x, 16);
}
[hjl@gnu-6 memcpy-3]$
clang trunk -O2 -mavx2 generates
[hjl@gnu-6 memcpy-3]$ cat c.s
.text
.file "c.i"
.globl foo3
.p2align 4, 0x90
.type foo3,@function
foo3: # @foo3
.cfi_startproc
# BB#0:
movl dst(%rip), %eax
movzbl %dil, %ecx
movabsq $72340172838076673, %rdx # imm = 0x101010101010101
imulq %rcx, %rdx
movq %rdx, 56(%eax)
movq %rdx, 48(%eax)
movq %rdx, 40(%eax)
movq %rdx, 32(%eax)
movq %rdx, 24(%eax)
movq %rdx, 16(%eax)
movq %rdx, 8(%eax)
movq %rdx, (%eax)
retq
.Lfunc_end0:
.size foo3, .Lfunc_end0-foo3
.cfi_endproc
.globl foo4
.p2align 4, 0x90
.type foo4,@function
foo4: # @foo4
.cfi_startproc
# BB#0:
movl dst(%rip), %eax
movzbl %dil, %ecx
movabsq $72340172838076673, %rdx # imm = 0x101010101010101
imulq %rcx, %rdx
movq %rdx, 24(%eax)
movq %rdx, 16(%eax)
movq %rdx, 8(%eax)
movq %rdx, (%eax)
retq
.Lfunc_end1:
.size foo4, .Lfunc_end1-foo4
.cfi_endproc
.globl foo5
.p2align 4, 0x90
.type foo5,@function
foo5: # @foo5
.cfi_startproc
# BB#0:
movl dst(%rip), %eax
movzbl %dil, %ecx
movabsq $72340172838076673, %rdx # imm = 0x101010101010101
imulq %rcx, %rdx
movq %rdx, 8(%eax)
movq %rdx, (%eax)
retq
.Lfunc_end2:
.size foo5, .Lfunc_end2-foo5
.cfi_endproc
Will
.text
.p2align 4,,15
.globl foo3
.type foo3, @function
foo3:
.LFB0:
.cfi_startproc
movq dst(%rip), %rax
vmovd %edi, %xmm0
pushq %rbp
.cfi_def_cfa_offset 16
.cfi_offset 6, -16
vpbroadcastb %xmm0, %ymm0
vmovdqu %ymm0, (%rax)
movq %rsp, %rbp
.cfi_def_cfa_register 6
vmovdqu %ymm0, 32(%rax)
vzeroupper
popq %rbp
.cfi_def_cfa 7, 8
ret
.cfi_endproc
.LFE0:
.size foo3, .-foo3
.p2align 4,,15
.globl foo4
.type foo4, @function
foo4:
.LFB1:
.cfi_startproc
movq dst(%rip), %rax
vmovd %edi, %xmm0
pushq %rbp
.cfi_def_cfa_offset 16
.cfi_offset 6, -16
vpbroadcastb %xmm0, %ymm0
vmovdqu %ymm0, (%rax)
movq %rsp, %rbp
.cfi_def_cfa_register 6
vzeroupper
popq %rbp
.cfi_def_cfa 7, 8
ret
.cfi_endproc
.LFE1:
.size foo4, .-foo4
.p2align 4,,15
.globl foo5
.type foo5, @function
foo5:
.LFB2:
.cfi_startproc
movq dst(%rip), %rax
vmovd %edi, %xmm0
vpbroadcastb %xmm0, %xmm0
vmovdqu %xmm0, (%rax)
ret
.cfi_endproc
.LFE2:
.size foo5, .-foo5
be better?</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>