<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 --- - Lowering operations to compiler intrinsics with reference parameters generates bad code"
href="https://llvm.org/bugs/show_bug.cgi?id=31363">31363</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>Lowering operations to compiler intrinsics with reference parameters generates bad code
</td>
</tr>
<tr>
<th>Product</th>
<td>libraries
</td>
</tr>
<tr>
<th>Version</th>
<td>3.9
</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>simonas+llvm.org@kazlauskas.me
</td>
</tr>
<tr>
<th>CC</th>
<td>llvm-bugs@lists.llvm.org
</td>
</tr>
<tr>
<th>Classification</th>
<td>Unclassified
</td>
</tr></table>
<p>
<div>
<pre>Consider this snippet of LLVM IR:
```llvm
target datalayout = "e-m:w-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-pc-windows-msvc"
; Function Attrs: nounwind uwtable
define void @test3(i128* noalias nocapture sret dereferenceable(16), i128*
noalias readonly nocapture dereferenceable(16), i128* noalias readonly
nocapture dereferenceable(16)) unnamed_addr #0 {
start:
%3 = load i128, i128* %1
%4 = load i128, i128* %2
%5 = udiv i128 %3, %4
store i128 %5, i128* %0
ret void
}
attributes #0 = { nounwind uwtable }
```
This generates following assembly:
```asm
test3:
push rsi
.seh_pushreg 6
sub rsp, 64
.seh_stackalloc 64
.seh_endprologue
mov rsi, rcx
mov r9, qword ptr [rdx]
mov rcx, qword ptr [rdx + 8]
mov rdx, qword ptr [r8]
mov rax, qword ptr [r8 + 8]
mov qword ptr [rsp + 56], rcx
mov qword ptr [rsp + 48], r9
mov qword ptr [rsp + 40], rax
mov qword ptr [rsp + 32], rdx
lea rcx, [rsp + 48]
lea rdx, [rsp + 32]
call __udivti3
pshufd xmm1, xmm0, 78 # xmm1 = xmm0[2,3,0,1]
movq qword ptr [rsi], xmm0
movq qword ptr [rsi + 8], xmm1
mov rax, rsi
add rsp, 64
pop rsi
ret
.seh_handlerdata
.text
.seh_endproc
```
noting that LLVM thinks __udivti3 looks like this when targetting windows:
```
declare i128 @__udivti3(i128* readonly nocapture dereferenceable(16), i128*
readonly nocapture dereferenceable(16));
```
its easy to see that there’s way too much stack trashing going on and the
assembly instead could be just this:
```
test3:
push rsi
.seh_pushreg 6
sub rsp, 32
.seh_stackalloc 32
.seh_endprologue
mov rsi, rcx
mov rcx, rdx
mov rdx, r8
call __udivti3
pshufd xmm1, xmm0, 78 # xmm1 = xmm0[2,3,0,1]
movq qword ptr [rsi], xmm0
movq qword ptr [rsi + 8], xmm1
mov rax, rsi
add rsp, 32
pop rsi
ret
.seh_handlerdata
.text
.seh_endproc
```
which only does write to the sret parameter, as one would expect from a well
optimised code.</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>