<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 --- - psABI inconsistency when passing vector registers?"
href="https://llvm.org/bugs/show_bug.cgi?id=29092">29092</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>psABI inconsistency when passing vector registers?
</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>Backend: X86
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>wenzel.jakob@epfl.ch
</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>Clang trunk does not pass vector types through registers when they are
contained in an aggregate data type. Consider the following simple program:
------------------------------
C code:
#include <immintrin.h>
typedef struct __attribute__((aligned(16))) {
__m128 x, y;
} Structure;
__m128 test1(__m128 x, __m128 y) { return y; }
__m128 test2(Structure t) { return t.y; }
------------------------------
Assembly:
$ clang++ test.cpp -msse4.2 -I include -std=c++14 -o - -S -fomit-frame-pointer
-O3
__Z5test1Dv4_fS_:
vmovaps %xmm1, %xmm0
retq
__Z5test29Structure:
vmovaps 24(%rsp), %xmm0 <------ copy from memory :(
retq
------------------------------
The psABI spec
(<a href="https://github.com/hjl-tools/x86-psABI/blob/master/low-level-sys-info.tex#L506">https://github.com/hjl-tools/x86-psABI/blob/master/low-level-sys-info.tex#L506</a>)
has the following to say about this:
Each field of an object is classified *recursively* (emphasis added) so that
always two fields are considered. The resulting class is calculated according
to the classes of the fields in the eightbyte [...].
(a)-(e) [... cases which don't apply ...]
(f) Otherwise class SSE is used.
Intuitively (and according to the spec), I would have expected that "x" is
assigned to %xmm0 and "y" is assigned to %xmm1.
I'm wondering if I am misreading the ABI here, or if Clang is in violation?</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>