Skip to content

Optimize IM_NORMALIZE2F_OVER_ZERO #4091

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 5 commits into
base: master
Choose a base branch
from
+19 −3
Open

Optimize IM_NORMALIZE2F_OVER_ZERO #4091

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
38563bc
Add ImRsqrt().
wolfpld Apr 30, 2021
0da5cde
Use ImRsqrt() in place of 1.0f / ImSqrt().
wolfpld Apr 30, 2021
fd3f999
Use SSE rsqrt, if available.
wolfpld Apr 30, 2021
4fb9122
Add ImRecip() wrapper.
wolfpld May 6, 2021
81eed5e
Use ImRecip2() in IM_FIXNORMAL2F().
wolfpld May 6, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions imgui_draw.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -694,9 +694,9 @@ void ImDrawList::PrimQuadUV(const ImVec2& a, const ImVec2& b, const ImVec2& c, c

// On AddPolyline() and AddConvexPolyFilled() we intentionally avoid using ImVec2 and superfluous function calls to optimize debug/non-inlined builds.
// Those macros expects l-values.
#define IM_NORMALIZE2F_OVER_ZERO(VX,VY) do { float d2 = VX*VX + VY*VY; if (d2 > 0.0f) { float inv_len = 1.0f / ImSqrt(d2); VX *= inv_len; VY *= inv_len; } } while (0)
#define IM_NORMALIZE2F_OVER_ZERO(VX,VY) do { float d2 = VX*VX + VY*VY; if (d2 > 0.0f) { float inv_len = ImRsqrt(d2); VX *= inv_len; VY *= inv_len; } } while (0)
#define IM_FIXNORMAL2F_MAX_INVLEN2 100.0f // 500.0f (see #4053, #3366)
#define IM_FIXNORMAL2F(VX,VY) do { float d2 = VX*VX + VY*VY; if (d2 > 0.000001f) { float inv_len2 = 1.0f / d2; if (inv_len2 > IM_FIXNORMAL2F_MAX_INVLEN2) inv_len2 = IM_FIXNORMAL2F_MAX_INVLEN2; VX *= inv_len2; VY *= inv_len2; } } while (0)
#define IM_FIXNORMAL2F(VX,VY) do { float d2 = VX*VX + VY*VY; if (d2 > 0.000001f) { float inv_len2 = ImRecip(d2); if (inv_len2 > IM_FIXNORMAL2F_MAX_INVLEN2) inv_len2 = IM_FIXNORMAL2F_MAX_INVLEN2; VX *= inv_len2; VY *= inv_len2; } } while (0)

// TODO: Thickness anti-aliased lines cap are missing their AA fringe.
// We avoid using the ImVec2 math operators here to reduce cost to a minimum for debug/non-inlined builds.
Expand Down
18 changes: 17 additions & 1 deletion imgui_internal.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,6 +51,10 @@ Index of this file:
#include <math.h> // sqrtf, fabsf, fmodf, powf, floorf, ceilf, cosf, sinf
#include <limits.h> // INT_MIN, INT_MAX

#if defined __SSE__ || defined __x86_64__ || defined _M_X64
#include <immintrin.h>
#endif

// Visual Studio warnings
#ifdef _MSC_VER
#pragma warning (push)
Expand Down Expand Up @@ -390,6 +394,18 @@ static inline float ImAbs(float x) { return fabsf(x); }
static inline double ImAbs(double x) { return fabs(x); }
static inline float ImSign(float x) { return (x < 0.0f) ? -1.0f : ((x > 0.0f) ? 1.0f : 0.0f); } // Sign operator - returns -1, 0 or 1 based on sign of argument
static inline double ImSign(double x) { return (x < 0.0) ? -1.0 : ((x > 0.0) ? 1.0 : 0.0); }
#if defined __SSE__ || defined __x86_64__ || defined _M_X64
static inline float ImRsqrt(float x) { return _mm_cvtss_f32(_mm_rsqrt_ss(_mm_set_ss(x))); }
#else
static inline float ImRsqrt(float x) { return 1.0f / sqrtf(x); }
#endif
static inline double ImRsqrt(double x) { return 1.0 / sqrt(x); }
#if defined __SSE__ || defined __x86_64__ || defined _M_X64
static inline float ImRecip(float x) { return _mm_cvtss_f32(_mm_rcp_ps(_mm_set_ss(x))); }
#else
static inline float ImRecip(float x) { return 1.0f / x; }
#endif
static inline double ImRecip(double x) { return 1.0 / x; }
#endif
// - ImMin/ImMax/ImClamp/ImLerp/ImSwap are used by widgets which support variety of types: signed/unsigned int/long long float/double
// (Exceptionally using templates here but we could also redefine them for those types)
Expand All @@ -410,7 +426,7 @@ static inline ImVec4 ImLerp(const ImVec4& a, const ImVec4& b, float t)
static inline float ImSaturate(float f) { return (f < 0.0f) ? 0.0f : (f > 1.0f) ? 1.0f : f; }
static inline float ImLengthSqr(const ImVec2& lhs) { return (lhs.x * lhs.x) + (lhs.y * lhs.y); }
static inline float ImLengthSqr(const ImVec4& lhs) { return (lhs.x * lhs.x) + (lhs.y * lhs.y) + (lhs.z * lhs.z) + (lhs.w * lhs.w); }
static inline float ImInvLength(const ImVec2& lhs, float fail_value) { float d = (lhs.x * lhs.x) + (lhs.y * lhs.y); if (d > 0.0f) return 1.0f / ImSqrt(d); return fail_value; }
static inline float ImInvLength(const ImVec2& lhs, float fail_value) { float d = (lhs.x * lhs.x) + (lhs.y * lhs.y); if (d > 0.0f) return ImRsqrt(d); return fail_value; }
static inline float ImFloor(float f) { return (float)(int)(f); }
static inline float ImFloorSigned(float f) { return (float)((f >= 0 || (int)f == f) ? (int)f : (int)f - 1); } // Decent replacement for floorf()
static inline ImVec2 ImFloor(const ImVec2& v) { return ImVec2((float)(int)(v.x), (float)(int)(v.y)); }
Expand Down