From c95739757316a50ada1888cb5380b9d8aa8f0a59 Mon Sep 17 00:00:00 2001 From: Christoph Oelckers Date: Tue, 1 Sep 2020 19:47:01 +0200 Subject: [PATCH] - undid the workarounds for undefined-ness of negative shifts. For any mainstream platform that is totally irrelevant and besides, in C++20 it will be well defined and all current compilers on the relevant platforms treat it accordingly. This is not worth deoptimizing the code. --- source/common/utility/m_fixed.h | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/source/common/utility/m_fixed.h b/source/common/utility/m_fixed.h index 1381fdedb..3bdd055f0 100644 --- a/source/common/utility/m_fixed.h +++ b/source/common/utility/m_fixed.h @@ -55,8 +55,7 @@ inline fixed_t FloatToFixed(double f) inline constexpr fixed_t IntToFixed(int32_t f) { - // Negative shifts are undefined, so multiply instead of shifting left. - return f * FRACUNIT; + return f << FRACBITS; } inline constexpr double FixedToFloat(fixed_t f) @@ -66,7 +65,7 @@ inline constexpr double FixedToFloat(fixed_t f) inline constexpr int32_t FixedToInt(fixed_t f) { - return xs_CRoundToInt(FixedToFloat(f)); + return (f + FRACUNIT/2) >> FRACBITS; } inline constexpr unsigned FloatToAngle(double f)