tzh
2024-08-22 c7d0944258c7d0943aa7b2211498fd612971ce27
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
<html>
<!-- Copyright (C) 1988-2017 Free Software Foundation, Inc.
 
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3 or
any later version published by the Free Software Foundation; with the
Invariant Sections being "Funding Free Software", the Front-Cover
Texts being (a) (see below), and with the Back-Cover Texts being (b)
(see below).  A copy of the license is included in the section entitled
"GNU Free Documentation License".
 
(a) The FSF's Front-Cover Text is:
 
A GNU Manual
 
(b) The FSF's Back-Cover Text is:
 
You have freedom to copy and modify this GNU Manual, like GNU
     software.  Copies published by the Free Software Foundation raise
     funds for GNU development. -->
<!-- Created by GNU Texinfo 5.2, http://www.gnu.org/software/texinfo/ -->
<head>
<title>Using the GNU Compiler Collection (GCC): Nvidia PTX Options</title>
 
<meta name="description" content="Using the GNU Compiler Collection (GCC): Nvidia PTX Options">
<meta name="keywords" content="Using the GNU Compiler Collection (GCC): Nvidia PTX Options">
<meta name="resource-type" content="document">
<meta name="distribution" content="global">
<meta name="Generator" content="makeinfo">
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<link href="index.html#Top" rel="start" title="Top">
<link href="Option-Index.html#Option-Index" rel="index" title="Option Index">
<link href="index.html#SEC_Contents" rel="contents" title="Table of Contents">
<link href="Submodel-Options.html#Submodel-Options" rel="up" title="Submodel Options">
<link href="PDP_002d11-Options.html#PDP_002d11-Options" rel="next" title="PDP-11 Options">
<link href="Nios-II-Options.html#Nios-II-Options" rel="prev" title="Nios II Options">
<style type="text/css">
<!--
a.summary-letter {text-decoration: none}
blockquote.smallquotation {font-size: smaller}
div.display {margin-left: 3.2em}
div.example {margin-left: 3.2em}
div.indentedblock {margin-left: 3.2em}
div.lisp {margin-left: 3.2em}
div.smalldisplay {margin-left: 3.2em}
div.smallexample {margin-left: 3.2em}
div.smallindentedblock {margin-left: 3.2em; font-size: smaller}
div.smalllisp {margin-left: 3.2em}
kbd {font-style:oblique}
pre.display {font-family: inherit}
pre.format {font-family: inherit}
pre.menu-comment {font-family: serif}
pre.menu-preformatted {font-family: serif}
pre.smalldisplay {font-family: inherit; font-size: smaller}
pre.smallexample {font-size: smaller}
pre.smallformat {font-family: inherit; font-size: smaller}
pre.smalllisp {font-size: smaller}
span.nocodebreak {white-space:nowrap}
span.nolinebreak {white-space:nowrap}
span.roman {font-family:serif; font-weight:normal}
span.sansserif {font-family:sans-serif; font-weight:normal}
ul.no-bullet {list-style: none}
-->
</style>
 
 
</head>
 
<body lang="en" bgcolor="#FFFFFF" text="#000000" link="#0000FF" vlink="#800080" alink="#FF0000">
<a name="Nvidia-PTX-Options"></a>
<div class="header">
<p>
Next: <a href="PDP_002d11-Options.html#PDP_002d11-Options" accesskey="n" rel="next">PDP-11 Options</a>, Previous: <a href="Nios-II-Options.html#Nios-II-Options" accesskey="p" rel="prev">Nios II Options</a>, Up: <a href="Submodel-Options.html#Submodel-Options" accesskey="u" rel="up">Submodel Options</a> &nbsp; [<a href="index.html#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="Option-Index.html#Option-Index" title="Index" rel="index">Index</a>]</p>
</div>
<hr>
<a name="Nvidia-PTX-Options-1"></a>
<h4 class="subsection">3.18.33 Nvidia PTX Options</h4>
<a name="index-Nvidia-PTX-options"></a>
<a name="index-nvptx-options"></a>
 
<p>These options are defined for Nvidia PTX:
</p>
<dl compact="compact">
<dt><code>-m32</code></dt>
<dt><code>-m64</code></dt>
<dd><a name="index-m32"></a>
<a name="index-m64"></a>
<p>Generate code for 32-bit or 64-bit ABI.
</p>
</dd>
<dt><code>-mmainkernel</code></dt>
<dd><a name="index-mmainkernel"></a>
<p>Link in code for a __main kernel.  This is for stand-alone instead of
offloading execution.
</p>
</dd>
<dt><code>-moptimize</code></dt>
<dd><a name="index-moptimize"></a>
<p>Apply partitioned execution optimizations.  This is the default when any
level of optimization is selected.
</p>
</dd>
<dt><code>-msoft-stack</code></dt>
<dd><a name="index-msoft_002dstack"></a>
<p>Generate code that does not use <code>.local</code> memory
directly for stack storage. Instead, a per-warp stack pointer is
maintained explicitly. This enables variable-length stack allocation (with
variable-length arrays or <code>alloca</code>), and when global memory is used for
underlying storage, makes it possible to access automatic variables from other
threads, or with atomic instructions. This code generation variant is used
for OpenMP offloading, but the option is exposed on its own for the purpose
of testing the compiler; to generate code suitable for linking into programs
using OpenMP offloading, use option <samp>-mgomp</samp>.
</p>
</dd>
<dt><code>-muniform-simt</code></dt>
<dd><a name="index-muniform_002dsimt"></a>
<p>Switch to code generation variant that allows to execute all threads in each
warp, while maintaining memory state and side effects as if only one thread
in each warp was active outside of OpenMP SIMD regions.  All atomic operations
and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
current lane index equals the master lane index), and the register being
assigned is copied via a shuffle instruction from the master lane.  Outside of
SIMD regions lane 0 is the master; inside, each thread sees itself as the
master.  Shared memory array <code>int __nvptx_uni[]</code> stores all-zeros or
all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
regions).  Each thread can bitwise-and the bitmask at position <code>tid.y</code>
with current lane index to compute the master lane index.
</p>
</dd>
<dt><code>-mgomp</code></dt>
<dd><a name="index-mgomp"></a>
<p>Generate code for use in OpenMP offloading: enables <samp>-msoft-stack</samp> and
<samp>-muniform-simt</samp> options, and selects corresponding multilib variant.
</p>
</dd>
</dl>
 
<hr>
<div class="header">
<p>
Next: <a href="PDP_002d11-Options.html#PDP_002d11-Options" accesskey="n" rel="next">PDP-11 Options</a>, Previous: <a href="Nios-II-Options.html#Nios-II-Options" accesskey="p" rel="prev">Nios II Options</a>, Up: <a href="Submodel-Options.html#Submodel-Options" accesskey="u" rel="up">Submodel Options</a> &nbsp; [<a href="index.html#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="Option-Index.html#Option-Index" title="Index" rel="index">Index</a>]</p>
</div>
 
 
 
</body>
</html>