You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
qt6-base/debian/patches/use_float16_only_with_sse2....

40 lines
1.7 KiB

Description: use _Float16 only when SSE2 is enabled
The GCC documentation [1] says: “On x86 targets with SSE2 enabled, GCC
supports half-precision (16-bit) floating point via the _Float16 type”.
.
On non-SSE2 x86 (such as Debian i386 baseline [2]), __FLT16_MAX__ is
defined starting with GCC 14 [3], however any non-trivial use of the
_Float16 type results in an error:
.
error: operation not permitted on type _Float16 without option -msse2
.
which makes some packages fail to build on i386 architecture [4].
.
[1]: https://gcc.gnu.org/onlinedocs/gcc/Half-Precision.html
[2]: https://wiki.debian.org/ArchitectureSpecificsMemo#i386-1
[3]: https://gcc.gnu.org/g:9a19fa8b616f83474c35cc5b34a3865073ced829
[4]: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1076986
Author: Dmitry Shachnev <mitya57@debian.org>
Forwarded: https://codereview.qt-project.org/c/qt/qtbase/+/579205
Last-Update: 2024-08-01
--- a/src/corelib/global/qtypes.h
+++ b/src/corelib/global/qtypes.h
@@ -263,13 +263,12 @@ using NativeFloat16Type = std::float16_t
// disabled due to https://github.com/llvm/llvm-project/issues/56963
# define QFLOAT16_IS_NATIVE 1
using NativeFloat16Type = decltype(__FLT16_MAX__);
-#elif defined(Q_CC_GNU_ONLY) && defined(__FLT16_MAX__)
+#elif defined(Q_CC_GNU_ONLY) && defined(__FLT16_MAX__) && defined(__ARM_FP16_FORMAT_IEEE)
# define QFLOAT16_IS_NATIVE 1
-# ifdef __ARM_FP16_FORMAT_IEEE
using NativeFloat16Type = __fp16;
-# else
+#elif defined(Q_CC_GNU_ONLY) && defined(__FLT16_MAX__) && defined(__SSE2__)
+# define QFLOAT16_IS_NATIVE 1
using NativeFloat16Type = _Float16;
-# endif
#else
# define QFLOAT16_IS_NATIVE 0
using NativeFloat16Type = void;