On Mon, 10 Mar 2025, Martin Storsjö wrote:
On Wed, 5 Mar 2025, Jacek Caban wrote:
---
mingw-w64-crt/math/truncl.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/mingw-w64-crt/math/truncl.c b/mingw-w64-crt/math/truncl.c
index 3b47e53d1..d1b296661 100644
--- a/mingw-w64-crt/math/truncl.c
+++ b/mingw-w64-crt/math/truncl.c
@@ -3,7 +3,7 @@
* This file is part of the mingw-w64 runtime package.
* No warranty is given; refer to the file DISCLAIMER.PD within this
package.
*/
-#include <fenv.h>
+
#include <math.h>
long double
@@ -16,8 +16,7 @@ truncl (long double _x)
unsigned short saved_cw;
unsigned short tmp_cw;
__asm__ __volatile__ ("fnstcw %0;" : "=m" (saved_cw)); /* save FPU
control word */
- tmp_cw = (saved_cw & ~(FE_TONEAREST | FE_DOWNWARD | FE_UPWARD |
FE_TOWARDZERO))
- | FE_TOWARDZERO;
+ tmp_cw = saved_cw | 0xc00; /* round towards zero */
I don't quite understand this change.
As far as I can see, the existing code does this, after expanding macros and
merging constants:
tmp_cw = (saved_cw & ~ 0x300) | 0x300;
And this is, practically speaking, the same as just tmp_cw = saved_cw |
0x300; - because FE_TOWARDZERO happens to be a value with all bits set.
Sorry, this was based on looking at the headers after patch 2/2.
When looking at the right version of fenv.h it does indeed look like a
no-op.
But it would be good to include a motivation in the commit message.
// Martin
_______________________________________________
Mingw-w64-public mailing list
Mingw-w64-public@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mingw-w64-public
