From fb0f12f937ee9d943f9adafa3d350aff65105215 Mon Sep 17 00:00:00 2001
From: Delyan Angelov <delian66@gmail.com>
Date: Mon, 4 Sep 2023 21:09:01 +0300
Subject: [PATCH] math: add more C. fn overrides, for the default C backend
 (speed up examples/path_tracing.v) (#19271)

---
 bench/vectors/vectors.cs  | 10 ++++------
 bench/vectors/vectors.v   | 13 +++++++------
 vlib/math/exp.c.v         | 25 +++++++++++++++++++++++++
 vlib/math/log.c.v         | 17 +++++++++++++++++
 vlib/math/log.v           |  2 +-
 vlib/math/pow.c.v         | 10 +++++++++-
 vlib/math/pow.v           |  2 +-
 vlib/math/sin.c.v         | 16 ++++++++++++++++
 vlib/math/sqrt.c.v        |  7 +++++++
 vlib/math/square.v        | 13 +++++++++++++
 vlib/math/square_test.v   | 21 +++++++++++++++++++++
 vlib/math/unsafe.v        |  7 +++++++
 vlib/math/vec/vec2_test.v |  5 +++--
 13 files changed, 131 insertions(+), 17 deletions(-)
 create mode 100644 vlib/math/exp.c.v
 create mode 100644 vlib/math/log.c.v
 create mode 100644 vlib/math/square.v
 create mode 100644 vlib/math/square_test.v

diff --git a/bench/vectors/vectors.cs b/bench/vectors/vectors.cs
index 691997497d..40a49cc1b5 100644
--- a/bench/vectors/vectors.cs
+++ b/bench/vectors/vectors.cs
@@ -62,13 +62,11 @@ internal static class Program
                 var differenceY = position.Y - otherPosition.Y;
                 var differenceZ = position.Z - otherPosition.Z;
                 
-                var distance = Math.Sqrt(
-                    differenceX * differenceX +
+                var distance = differenceX * differenceX +
                     differenceY * differenceY +
-                    differenceZ * differenceZ
-                );
+                    differenceZ * differenceZ;
 
-                if (distance <= cohesionDistance)
+                if (distance <= cohesionDistance * cohesionDistance)
                 {
                     closeBoids.Add(otherBoidIndex);
                 }
@@ -188,4 +186,4 @@ internal static class Program
         Console.WriteLine(positionSum.ToString());
         Console.WriteLine(velocitySum.ToString());
     }
-}
\ No newline at end of file
+}
diff --git a/bench/vectors/vectors.v b/bench/vectors/vectors.v
index 0504ff0506..92c8b1d7c7 100644
--- a/bench/vectors/vectors.v
+++ b/bench/vectors/vectors.v
@@ -44,11 +44,10 @@ fn main() {
 			difference_y := position.y - other_position.y
 			difference_z := position.z - other_position.z
 
-			distance := math.sqrt(difference_x * difference_x +
-			difference_y * difference_y +
-			difference_z * difference_z)
+			distance := difference_x * difference_x + difference_y * difference_y +
+				difference_z * difference_z
 
-			if distance <= cohesion_distance {
+			if distance <= cohesion_distance * cohesion_distance {
 				close_boids_ids << other_boid_index
 			}
 		}
@@ -77,8 +76,10 @@ fn main() {
 				z: position.z - close_boid_position.z
 			}
 
-			difference_magnitude := math.sqrt(difference_from_closest.x * difference_from_closest.x +
-			difference_from_closest.y * difference_from_closest.y + difference_from_closest.z * difference_from_closest.z)
+			difference_magnitude := math.sqrt(
+				difference_from_closest.x * difference_from_closest.x +
+				difference_from_closest.y * difference_from_closest.y +
+				difference_from_closest.z * difference_from_closest.z)
 
 			if difference_magnitude <= separation_distance {
 				separation = Vector{
diff --git a/vlib/math/exp.c.v b/vlib/math/exp.c.v
new file mode 100644
index 0000000000..02596302e5
--- /dev/null
+++ b/vlib/math/exp.c.v
@@ -0,0 +1,25 @@
+module math
+
+fn C.exp(x f64) f64
+
+fn C.exp2(x f64) f64
+
+fn C.ldexp(x f64, exp int) f64
+
+// exp returns e**x, the base-e exponential of x.
+[inline]
+pub fn exp(x f64) f64 {
+	return C.exp(x)
+}
+
+// exp2 returns 2**x, the base-2 exponential of x.
+[inline]
+pub fn exp2(x f64) f64 {
+	return C.exp2(x)
+}
+
+// ldexp calculates frac*(2**exp).
+[inline]
+pub fn ldexp(frac f64, exp int) f64 {
+	return C.ldexp(frac, exp)
+}
diff --git a/vlib/math/log.c.v b/vlib/math/log.c.v
new file mode 100644
index 0000000000..17e04a2aec
--- /dev/null
+++ b/vlib/math/log.c.v
@@ -0,0 +1,17 @@
+module math
+
+fn C.log(x f64) f64
+
+fn C.logf(x f32) f32
+
+// log returns the natural logarithm of x (float64)
+[inline]
+pub fn log(a f64) f64 {
+	return C.log(a)
+}
+
+// log returns the natural logarithm of x (float32)
+[inline]
+pub fn logf(a f32) f32 {
+	return C.logf(a)
+}
diff --git a/vlib/math/log.v b/vlib/math/log.v
index da7eaedaa5..1d3d5f0a9c 100644
--- a/vlib/math/log.v
+++ b/vlib/math/log.v
@@ -77,7 +77,7 @@ fn ilog_b_(x_ f64) int {
 	return int((f64_bits(x) >> shift) & mask) - bias + exp
 }
 
-// log returns the logarithm of x
+// log returns the natural logarithm of x
 //
 // Method :
 //   1. Argument Reduction: find k and f such that
diff --git a/vlib/math/pow.c.v b/vlib/math/pow.c.v
index 475620cc92..beaa0c026b 100644
--- a/vlib/math/pow.c.v
+++ b/vlib/math/pow.c.v
@@ -1,8 +1,16 @@
 module math
 
+fn C.pow(x f64, y f64) f64
+
 fn C.powf(x f32, y f32) f32
 
-// powf returns base raised to the provided power. (float32)
+// pow returns the base x, raised to the provided power y. (float64)
+[inline]
+pub fn pow(x f64, y f64) f64 {
+	return C.pow(x, y)
+}
+
+// powf returns the base a, raised to the provided power b. (float32)
 [inline]
 pub fn powf(a f32, b f32) f32 {
 	return C.powf(a, b)
diff --git a/vlib/math/pow.v b/vlib/math/pow.v
index f045362f75..ddb06076e6 100644
--- a/vlib/math/pow.v
+++ b/vlib/math/pow.v
@@ -70,7 +70,7 @@ pub fn powi(a i64, b i64) i64 {
 	return v
 }
 
-// pow returns base raised to the provided power.
+// pow returns the base x, raised to the provided power y. (float64)
 //
 // todo(playXE): make this function work on JS backend, probably problem of JS codegen that it does not work.
 pub fn pow(x f64, y f64) f64 {
diff --git a/vlib/math/sin.c.v b/vlib/math/sin.c.v
index ef313af18e..6fffa2bc77 100644
--- a/vlib/math/sin.c.v
+++ b/vlib/math/sin.c.v
@@ -1,9 +1,25 @@
 module math
 
+fn C.cos(x f64) f64
+
+fn C.sin(x f64) f64
+
 fn C.cosf(x f32) f32
 
 fn C.sinf(x f32) f32
 
+// cos calculates cosine in radians (float64)
+[inline]
+pub fn cos(a f64) f64 {
+	return C.cos(a)
+}
+
+// sin calculates sine in radians (float64)
+[inline]
+pub fn sin(a f64) f64 {
+	return C.sin(a)
+}
+
 // cosf calculates cosine in radians (float32)
 [inline]
 pub fn cosf(a f32) f32 {
diff --git a/vlib/math/sqrt.c.v b/vlib/math/sqrt.c.v
index 02766657ff..62382e650b 100644
--- a/vlib/math/sqrt.c.v
+++ b/vlib/math/sqrt.c.v
@@ -1,7 +1,14 @@
 module math
 
+fn C.sqrt(x f64) f64
 fn C.sqrtf(x f32) f32
 
+// sqrt calculates square-root of the provided value. (float64)
+[inline]
+pub fn sqrt(a f64) f64 {
+	return C.sqrt(a)
+}
+
 // sqrtf calculates square-root of the provided value. (float32)
 [inline]
 pub fn sqrtf(a f32) f32 {
diff --git a/vlib/math/square.v b/vlib/math/square.v
new file mode 100644
index 0000000000..a5ace7ad07
--- /dev/null
+++ b/vlib/math/square.v
@@ -0,0 +1,13 @@
+module math
+
+// square returns the square of the argument x, i.e. x * x
+[inline]
+pub fn square[T](x T) T {
+	return x * x
+}
+
+// cube returns the cube of the argument x, i.e. x * x * x
+[inline]
+pub fn cube[T](x T) T {
+	return x * x * x
+}
diff --git a/vlib/math/square_test.v b/vlib/math/square_test.v
new file mode 100644
index 0000000000..c0ef187967
--- /dev/null
+++ b/vlib/math/square_test.v
@@ -0,0 +1,21 @@
+import math
+
+fn test_square() {
+	assert math.square(0) == 0
+	assert math.square(1) == 1
+	assert math.square(-1) == 1
+	assert math.square(5) == 25
+	assert math.square(u8(3)) == 9
+	assert math.square(u32(10)) == 100
+	assert math.square(-7) == 49
+}
+
+fn test_cube() {
+	assert math.cube(0) == 0
+	assert math.cube(1) == 1
+	assert math.cube(-1) == -1
+	assert math.cube(5) == 125
+	assert math.cube(u8(3)) == 27
+	assert math.cube(u32(10)) == 1000
+	assert math.cube(-7) == -343
+}
diff --git a/vlib/math/unsafe.v b/vlib/math/unsafe.v
index efbb5f0b63..d304299a1c 100644
--- a/vlib/math/unsafe.v
+++ b/vlib/math/unsafe.v
@@ -6,6 +6,7 @@ module math
 // f32_bits returns the IEEE 754 binary representation of f,
 // with the sign bit of f and the result in the same bit position.
 // f32_bits(f32_from_bits(x)) == x.
+[inline]
 pub fn f32_bits(f f32) u32 {
 	p := *unsafe { &u32(&f) }
 	return p
@@ -15,6 +16,7 @@ pub fn f32_bits(f f32) u32 {
 // to the IEEE 754 binary representation b, with the sign bit of b
 // and the result in the same bit position.
 // f32_from_bits(f32_bits(x)) == x.
+[inline]
 pub fn f32_from_bits(b u32) f32 {
 	p := *unsafe { &f32(&b) }
 	return p
@@ -23,6 +25,7 @@ pub fn f32_from_bits(b u32) f32 {
 // f64_bits returns the IEEE 754 binary representation of f,
 // with the sign bit of f and the result in the same bit position,
 // and f64_bits(f64_from_bits(x)) == x.
+[inline]
 pub fn f64_bits(f f64) u64 {
 	p := *unsafe { &u64(&f) }
 	return p
@@ -32,12 +35,14 @@ pub fn f64_bits(f f64) u64 {
 // to the IEEE 754 binary representation b, with the sign bit of b
 // and the result in the same bit position.
 // f64_from_bits(f64_bits(x)) == x.
+[inline]
 pub fn f64_from_bits(b u64) f64 {
 	p := *unsafe { &f64(&b) }
 	return p
 }
 
 // with_set_low_word sets low word of `f` to `lo`
+[inline]
 pub fn with_set_low_word(f f64, lo u32) f64 {
 	mut tmp := f64_bits(f)
 	tmp &= 0xffffffff_00000000
@@ -46,6 +51,7 @@ pub fn with_set_low_word(f f64, lo u32) f64 {
 }
 
 // with_set_high_word sets high word of `f` to `lo`
+[inline]
 pub fn with_set_high_word(f f64, hi u32) f64 {
 	mut tmp := f64_bits(f)
 	tmp &= 0x00000000_ffffffff
@@ -54,6 +60,7 @@ pub fn with_set_high_word(f f64, hi u32) f64 {
 }
 
 // get_high_word returns high part of the word of `f`.
+[inline]
 pub fn get_high_word(f f64) u32 {
 	return u32(f64_bits(f) >> 32)
 }
diff --git a/vlib/math/vec/vec2_test.v b/vlib/math/vec/vec2_test.v
index f11d790b19..b20adbabfb 100644
--- a/vlib/math/vec/vec2_test.v
+++ b/vlib/math/vec/vec2_test.v
@@ -1,3 +1,4 @@
+import math
 import math.vec
 
 fn test_vec2_int() {
@@ -72,11 +73,11 @@ fn test_vec2_f64_utils_1() {
 
 fn test_vec2_f64_utils_2() {
 	mut v1 := vec.vec2(4.0, 4.0)
-	assert v1.unit().magnitude() == 1
+	assert math.veryclose(v1.unit().magnitude(), 1)
 	v2 := v1.mul_scalar(0.5)
 	assert v2.x == 2
 	assert v2.y == 2
-	assert v2.unit().magnitude() == 1
+	assert math.veryclose(v2.unit().magnitude(), 1)
 
 	invv2 := v2.inv()
 	assert invv2.x == 0.5