From 3363cc81a814cb99b71595eef9e597bdfe2b30c5 Mon Sep 17 00:00:00 2001
From: Andreas Kloeckner <inform@tiker.net>
Date: Mon, 22 Jun 2015 01:27:21 -0500
Subject: [PATCH] Make doctest output predictable
---
doc/tutorial.rst | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/doc/tutorial.rst b/doc/tutorial.rst
index 5ed456705..995af8f87 100644
--- a/doc/tutorial.rst
+++ b/doc/tutorial.rst
@@ -849,8 +849,8 @@ variable, as one might do in C or another programming language:
... "{ [i]: 0<=i<n }",
... """
... <float32> a_temp = sin(a[i])
- ... out1[i] = a_temp
- ... out2[i] = sqrt(1-a_temp*a_temp)
+ ... out1[i] = a_temp {id=out1}
+ ... out2[i] = sqrt(1-a_temp*a_temp) {dep=out1}
... """)
The angle brackets ``<>`` denote the creation of a temporary. The name of
@@ -861,6 +861,9 @@ the conventional :mod:`numpy` scalar types (:class:`numpy.int16`,
:class:`numpy.complex128`) will work. (Yes, :mod:`loopy` supports and
generates correct code for complex arithmetic.)
+(If you're wondering, the dependencies above were added to make the doctest
+produce predictable output.)
+
The generated code places this variable into what OpenCL calls 'private'
memory, local to each work item.
@@ -876,8 +879,8 @@ memory, local to each work item.
for (int i = 0; i <= -1 + n; ++i)
{
a_temp = sin(a[i]);
- out2[i] = sqrt(1.0f + -1.0f * a_temp * a_temp);
out1[i] = a_temp;
+ out2[i] = sqrt(1.0f + -1.0f * a_temp * a_temp);
}
}
--
GitLab