From 40a25a2a4885112e70ae335be2d404d6d3609309 Mon Sep 17 00:00:00 2001
From: Jonathan Wilkes <jon.w.wilkes@gmail.com>
Date: Wed, 23 May 2018 13:18:42 -0400
Subject: [PATCH] use uint32_t to ensure alignment in union, improve comments
---
pd/src/m_pd.h | 26 ++++++++++++++++++++------
1 file changed, 20 insertions(+), 6 deletions(-)
diff --git a/pd/src/m_pd.h b/pd/src/m_pd.h
index be41e2173..fa41e85bd 100644
--- a/pd/src/m_pd.h
+++ b/pd/src/m_pd.h
@@ -772,10 +772,11 @@ defined, there is a "te_xpix" field in objects, not a "te_xpos" as before: */
typedef union
{
t_float f;
- unsigned int ui;
+ uint32_t ui;
}t_bigorsmall32;
-static inline int PD_BADFLOAT(t_float f) // test for NANs, infs and denormals
+/* Test strictly for NANs and infs */
+static inline int PD_BADFLOAT(t_float f)
{
t_bigorsmall32 pun;
pun.f = f;
@@ -783,7 +784,13 @@ static inline int PD_BADFLOAT(t_float f) // test for NANs, infs and denormals
return((pun.ui == 0) | (pun.ui == 0x7f800000));
}
-static inline int PD_BIGORSMALL(t_float f) // > abs(2^64) or < abs(2^-64)
+/* Test to find unusually large or small normal values, in
+ addition to denormals, NANs and infs:
+ f > abs(2^65) or < abs(2^-63)
+
+ This is useful for catching extreme values in, say, a filter,
+ then bashing to zero before ever calculating a denormal. */
+static inline int PD_BIGORSMALL(t_float f)
{
t_bigorsmall32 pun;
pun.f = f;
@@ -795,10 +802,11 @@ static inline int PD_BIGORSMALL(t_float f) // > abs(2^64) or < abs(2^-64)
typedef union
{
t_float f;
- unsigned int ui[2];
+ uint32_t ui[2];
}t_bigorsmall64;
-static inline int PD_BADFLOAT(t_float f) // test for NANs, infs and denormals
+/* Test for NANs and infs */
+static inline int PD_BADFLOAT(t_float f)
{
t_bigorsmall64 pun;
pun.f = f;
@@ -806,7 +814,13 @@ static inline int PD_BADFLOAT(t_float f) // test for NANs, infs and denormals
return((pun.ui[1] == 0) | (pun.ui[1] == 0x7ff00000));
}
-static inline int PD_BIGORSMALL(t_float f) // > abs(2^512) or < abs(2^-512)
+/* Test to find unusually large or small normal values, in
+ addition to denormals, NANs and infs:
+ f > abs(2^513) or < abs(2^-511)
+
+ This is useful for catching extreme values in, say, a filter,
+ then bashing to zero before ever calculating a denormal. */
+static inline int PD_BIGORSMALL(t_float f)
{
t_bigorsmall64 pun;
pun.f = f;
--
GitLab