Changeset 516e780 in mainline for uspace/lib/softfloat

Timestamp:

2018-08-31T11:55:41Z (7 years ago)

Author:

GitHub <noreply@…>

Branches:

lfn, master, serial, ticket/834-toolchain-update, topic/msim-upgrade, topic/simplify-dev-export

Children:

fa86fff

Parents:

7f7d642

git-author:

Jiří Zárevúcky <zarevucky.jiri@…> (2018-08-31 11:55:41)

git-committer:

GitHub <noreply@…> (2018-08-31 11:55:41)

Message:

Strip down libmath. (#45)

libmath is mostly unused (except for trunc(), sin() and cos()), and most functions in it are either very imprecise or downright broken. Additionally, it is implemented in manner that conflicts with C standard. Instead of trying to fix all the shortcomings while maintaining unused functionality, I'm opting to simply remove most of it and only keep the parts that are currently necessary.

Later readdition of the removed functions is possible, but there needs to be a reliable way to evaluate their quality first.

Location:

uspace/lib/softfloat

Files:

• Makefile (modified) (1 diff)
• comparison.c (modified) (2 diffs)
• comparison.h (modified) (2 diffs)
• mathtypes.h (moved) (moved from uspace/lib/math/include/mathtypes.h )

uspace/lib/softfloat/Makefile

@@ -30 +30 @@
 USPACE_PREFIX = ../..
 LIBRARY = libsoftfloat
-EXTRA_CFLAGS += $(LIBMATH_INCLUDES_FLAGS)
 
 SOURCES = \

uspace/lib/softfloat/comparison.c

@@ -681 +681 @@
 }
 
+int __aeabi_fcmpun(float32_t a, float32_t b)
+{
+        float32_u ua;
+        ua.val = a;
+
+        float32_u ub;
+        ub.val = b;
+
+        // TODO: sigNaNs
+        return is_float32_nan(ua.data) || is_float32_nan(ub.data);
+}
+
 #endif
 
@@ -920 +932 @@
 }
 
+int __aeabi_dcmpun(float64_t a, float64_t b)
+{
+        float64_u ua;
+        ua.val = a;
+
+        float64_u ub;
+        ub.val = b;
+
+        // TODO: sigNaNs
+        return is_float64_nan(ua.data) || is_float64_nan(ub.data);
+}
+
 #endif
 

uspace/lib/softfloat/comparison.h

@@ -93 +93 @@
 extern int __aeabi_fcmple(float32_t, float32_t);
 extern int __aeabi_fcmpeq(float32_t, float32_t);
+extern int __aeabi_fcmpun(float32_t, float32_t);
 #endif
 
@@ -109 +110 @@
 extern int __aeabi_dcmpge(float64_t, float64_t);
 extern int __aeabi_dcmple(float64_t, float64_t);
+extern int __aeabi_dcmpun(float64_t, float64_t);
 #endif