qfloat16(float) constructor: explicit cast on aarch64 to avoid warning

The warning was

global/qfloat16.h: In constructor ‘qfloat16::qfloat16(float)’:
global/qfloat16.h:124:18: error: conversion to ‘__fp16’ from ‘float’ may alter its value [-Werror=float-conversion]
     __fp16 f16 = f;
                  ^
cc1plus: all warnings being treated as errors

The warning was added by fb5976038162d93d60c7f76376bbb4df38e83ba9.

Change-Id: I489348c4d5d672bfa5d4db99c837696a2a69a27e
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This commit is contained in:
Shawn Rutledge 2017-11-17 10:35:58 +01:00 committed by Liang Qi
parent c4885b21d6
commit a4113d0c64

View File

@ -121,7 +121,7 @@ inline qfloat16::qfloat16(float f) Q_DECL_NOTHROW
__m128i packhalf = _mm_cvtps_ph(packsingle, 0);
b16 = _mm_extract_epi16(packhalf, 0);
#elif defined (__ARM_FP16_FORMAT_IEEE)
__fp16 f16 = f;
__fp16 f16 = __fp16(f);
memcpy(&b16, &f16, sizeof(quint16));
#else
quint32 u;