make style

This commit is contained in:
Philippe Teuwen committed 2026-05-21 20:54:44 +02:00
1 parent a7be3148db
commit d078e8a62b
44 files changed
+1127 -1162

No files matched your search

+61 -88
View File
@@ -2,9 +2,8 @@
#ifndef VEC_FN_ADD
#define VEC_FN_ADD
static inline union vec
vec_add (const union vec a, const union vec b)
{
return (union vec) { .gcc.f = a.gcc.f + b.gcc.f };
vec_add(const union vec a, const union vec b) {
return (union vec) { .gcc.f = a.gcc.f + b.gcc.f };
}
#endif
@@ -12,9 +11,8 @@ vec_add (const union vec a, const union vec b)
#ifndef VEC_FN_IADD
#define VEC_FN_IADD
static inline union vec
vec_iadd (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.i + b.gcc.i };
vec_iadd(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.i + b.gcc.i };
}
#endif
@@ -22,9 +20,8 @@ vec_iadd (const union vec a, const union vec b)
#ifndef VEC_FN_UADD
#define VEC_FN_UADD
static inline union vec
vec_uadd (const union vec a, const union vec b)
{
return (union vec) { .gcc.u = a.gcc.u + b.gcc.u };
vec_uadd(const union vec a, const union vec b) {
return (union vec) { .gcc.u = a.gcc.u + b.gcc.u };
}
#endif
@@ -32,9 +29,8 @@ vec_uadd (const union vec a, const union vec b)
#ifndef VEC_FN_SUB
#define VEC_FN_SUB
static inline union vec
vec_sub (const union vec a, const union vec b)
{
return (union vec) { .gcc.f = a.gcc.f - b.gcc.f };
vec_sub(const union vec a, const union vec b) {
return (union vec) { .gcc.f = a.gcc.f - b.gcc.f };
}
#endif
@@ -42,9 +38,8 @@ vec_sub (const union vec a, const union vec b)
#ifndef VEC_FN_ISUB
#define VEC_FN_ISUB
static inline union vec
vec_isub (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.i - b.gcc.i };
vec_isub(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.i - b.gcc.i };
}
#endif
@@ -52,9 +47,8 @@ vec_isub (const union vec a, const union vec b)
#ifndef VEC_FN_USUB
#define VEC_FN_USUB
static inline union vec
vec_usub (const union vec a, const union vec b)
{
return (union vec) { .gcc.u = a.gcc.u - b.gcc.u };
vec_usub(const union vec a, const union vec b) {
return (union vec) { .gcc.u = a.gcc.u - b.gcc.u };
}
#endif
@@ -62,9 +56,8 @@ vec_usub (const union vec a, const union vec b)
#ifndef VEC_FN_MUL
#define VEC_FN_MUL
static inline union vec
vec_mul (const union vec a, const union vec b)
{
return (union vec) { .gcc.f = a.gcc.f * b.gcc.f };
vec_mul(const union vec a, const union vec b) {
return (union vec) { .gcc.f = a.gcc.f * b.gcc.f };
}
#endif
@@ -72,9 +65,8 @@ vec_mul (const union vec a, const union vec b)
#ifndef VEC_FN_DIV
#define VEC_FN_DIV
static inline union vec
vec_div (const union vec a, const union vec b)
{
return (union vec) { .gcc.f = a.gcc.f / b.gcc.f };
vec_div(const union vec a, const union vec b) {
return (union vec) { .gcc.f = a.gcc.f / b.gcc.f };
}
#endif
@@ -82,9 +74,8 @@ vec_div (const union vec a, const union vec b)
#ifndef VEC_FN_AND
#define VEC_FN_AND
static inline union vec
vec_and (const union vec a, const union vec b)
{
return (union vec) { .gcc.u = a.gcc.u & b.gcc.u };
vec_and(const union vec a, const union vec b) {
return (union vec) { .gcc.u = a.gcc.u & b.gcc.u };
}
#endif
@@ -92,9 +83,8 @@ vec_and (const union vec a, const union vec b)
#ifndef VEC_FN_OR
#define VEC_FN_OR
static inline union vec
vec_or (const union vec a, const union vec b)
{
return (union vec) { .gcc.u = a.gcc.u | b.gcc.u };
vec_or(const union vec a, const union vec b) {
return (union vec) { .gcc.u = a.gcc.u | b.gcc.u };
}
#endif
@@ -102,9 +92,8 @@ vec_or (const union vec a, const union vec b)
#ifndef VEC_FN_XOR
#define VEC_FN_XOR
static inline union vec
vec_xor (const union vec a, const union vec b)
{
return (union vec) { .gcc.u = a.gcc.u ^ b.gcc.u };
vec_xor(const union vec a, const union vec b) {
return (union vec) { .gcc.u = a.gcc.u ^ b.gcc.u };
}
#endif
@@ -112,32 +101,30 @@ vec_xor (const union vec a, const union vec b)
#ifndef VEC_FN_NOT
#define VEC_FN_NOT
static inline union vec
vec_not (const union vec v)
{
return (union vec) { .gcc.u = ~v.gcc.u };
vec_not(const union vec v) {
return (union vec) { .gcc.u = ~v.gcc.u };
}
#endif
// GCC added support for vector comparisons in version 4.8:
#ifdef __GNUC__
#if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8)
#define VEC_GCC_COMPARE
#endif
#if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8)
#define VEC_GCC_COMPARE
#endif
#else
// Assume we're using clang - it's supported vector comparisons since at
// least version 3.0, possibly earlier:
#define VEC_GCC_COMPARE
// Assume we're using clang - it's supported vector comparisons since at
// least version 3.0, possibly earlier:
#define VEC_GCC_COMPARE
#endif
// Return float version of a == b:
#ifdef VEC_GCC_COMPARE
#ifndef VEC_FN_EQ
#ifndef VEC_FN_EQ
#define VEC_FN_EQ
static inline union vec
vec_eq (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.f == b.gcc.f };
vec_eq(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.f == b.gcc.f };
}
#endif
#endif
@@ -147,9 +134,8 @@ vec_eq (const union vec a, const union vec b)
#ifndef VEC_FN_IEQ
#define VEC_FN_IEQ
static inline union vec
vec_ieq (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.i == b.gcc.i };
vec_ieq(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.i == b.gcc.i };
}
#endif
#endif
@@ -159,9 +145,8 @@ vec_ieq (const union vec a, const union vec b)
#ifndef VEC_FN_UEQ
#define VEC_FN_UEQ
static inline union vec
vec_ueq (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.u == b.gcc.u };
vec_ueq(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.u == b.gcc.u };
}
#endif
#endif
@@ -171,9 +156,8 @@ vec_ueq (const union vec a, const union vec b)
#ifndef VEC_FN_LT
#define VEC_FN_LT
static inline union vec
vec_lt (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.f < b.gcc.f };
vec_lt(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.f < b.gcc.f };
}
#endif
#endif
@@ -183,9 +167,8 @@ vec_lt (const union vec a, const union vec b)
#ifndef VEC_FN_ILT
#define VEC_FN_ILT
static inline union vec
vec_ilt (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.i < b.gcc.i };
vec_ilt(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.i < b.gcc.i };
}
#endif
#endif
@@ -195,9 +178,8 @@ vec_ilt (const union vec a, const union vec b)
#ifndef VEC_FN_ULT
#define VEC_FN_ULT
static inline union vec
vec_ult (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.u < b.gcc.u };
vec_ult(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.u < b.gcc.u };
}
#endif
#endif
@@ -207,9 +189,8 @@ vec_ult (const union vec a, const union vec b)
#ifndef VEC_FN_LE
#define VEC_FN_LE
static inline union vec
vec_le (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.f <= b.gcc.f };
vec_le(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.f <= b.gcc.f };
}
#endif
#endif
@@ -219,9 +200,8 @@ vec_le (const union vec a, const union vec b)
#ifndef VEC_FN_ILE
#define VEC_FN_ILE
static inline union vec
vec_ile (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.i <= b.gcc.i };
vec_ile(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.i <= b.gcc.i };
}
#endif
#endif
@@ -231,9 +211,8 @@ vec_ile (const union vec a, const union vec b)
#ifndef VEC_FN_ULE
#define VEC_FN_ULE
static inline union vec
vec_ule (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.u <= b.gcc.u };
vec_ule(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.u <= b.gcc.u };
}
#endif
#endif
@@ -243,9 +222,8 @@ vec_ule (const union vec a, const union vec b)
#ifndef VEC_FN_GT
#define VEC_FN_GT
static inline union vec
vec_gt (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.f > b.gcc.f };
vec_gt(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.f > b.gcc.f };
}
#endif
#endif
@@ -255,9 +233,8 @@ vec_gt (const union vec a, const union vec b)
#ifndef VEC_FN_IGT
#define VEC_FN_IGT
static inline union vec
vec_igt (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.i > b.gcc.i };
vec_igt(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.i > b.gcc.i };
}
#endif
#endif
@@ -267,9 +244,8 @@ vec_igt (const union vec a, const union vec b)
#ifndef VEC_FN_UGT
#define VEC_FN_UGT
static inline union vec
vec_ugt (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.u > b.gcc.u };
vec_ugt(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.u > b.gcc.u };
}
#endif
#endif
@@ -279,9 +255,8 @@ vec_ugt (const union vec a, const union vec b)
#ifndef VEC_FN_GE
#define VEC_FN_GE
static inline union vec
vec_ge (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.f >= b.gcc.f };
vec_ge(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.f >= b.gcc.f };
}
#endif
#endif
@@ -291,9 +266,8 @@ vec_ge (const union vec a, const union vec b)
#ifndef VEC_FN_IGE
#define VEC_FN_IGE
static inline union vec
vec_ige (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.i >= b.gcc.i };
vec_ige(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.i >= b.gcc.i };
}
#endif
#endif
@@ -303,9 +277,8 @@ vec_ige (const union vec a, const union vec b)
#ifndef VEC_FN_UGE
#define VEC_FN_UGE
static inline union vec
vec_uge (const union vec a, const union vec b)
{
return (union vec) { .gcc.i = a.gcc.u >= b.gcc.u };
vec_uge(const union vec a, const union vec b) {
return (union vec) { .gcc.i = a.gcc.u >= b.gcc.u };
}
#endif
#endif
+220 -260
View File
@@ -2,9 +2,8 @@
#ifndef VEC_FN_
#define VEC_FN_
static inline union vec
vec (const float x, const float y, const float z, const float w)
{
return (union vec) { .x = x, .y = y, .z = z, .w = w };
vec(const float x, const float y, const float z, const float w) {
return (union vec) { .x = x, .y = y, .z = z, .w = w };
}
#endif
@@ -12,9 +11,8 @@ vec (const float x, const float y, const float z, const float w)
#ifndef VEC_FN_I
#define VEC_FN_I
static inline union vec
vec_i (const int32_t x, const int32_t y, const int32_t z, const int32_t w)
{
return (union vec) { .xi = x, .yi = y, .zi = z, .wi = w };
vec_i(const int32_t x, const int32_t y, const int32_t z, const int32_t w) {
return (union vec) { .xi = x, .yi = y, .zi = z, .wi = w };
}
#endif
@@ -22,9 +20,8 @@ vec_i (const int32_t x, const int32_t y, const int32_t z, const int32_t w)
#ifndef VEC_FN_U
#define VEC_FN_U
static inline union vec
vec_u (const uint32_t x, const uint32_t y, const uint32_t z, const uint32_t w)
{
return (union vec) { .xu = x, .yu = y, .zu = z, .wu = w };
vec_u(const uint32_t x, const uint32_t y, const uint32_t z, const uint32_t w) {
return (union vec) { .xu = x, .yu = y, .zu = z, .wu = w };
}
#endif
@@ -32,9 +29,8 @@ vec_u (const uint32_t x, const uint32_t y, const uint32_t z, const uint32_t w)
#ifndef VEC_FN_1
#define VEC_FN_1
static inline union vec
vec_1 (const float val)
{
return vec(val, val, val, val);
vec_1(const float val) {
return vec(val, val, val, val);
}
#endif
@@ -42,9 +38,8 @@ vec_1 (const float val)
#ifndef VEC_FN_I1
#define VEC_FN_I1
static inline union vec
vec_i1 (const int32_t val)
{
return vec_i(val, val, val, val);
vec_i1(const int32_t val) {
return vec_i(val, val, val, val);
}
#endif
@@ -52,9 +47,8 @@ vec_i1 (const int32_t val)
#ifndef VEC_FN_U1
#define VEC_FN_U1
static inline union vec
vec_u1 (const uint32_t val)
{
return vec_u(val, val, val, val);
vec_u1(const uint32_t val) {
return vec_u(val, val, val, val);
}
#endif
@@ -62,9 +56,8 @@ vec_u1 (const uint32_t val)
#ifndef VEC_FN_ZERO
#define VEC_FN_ZERO
static inline union vec
vec_zero (void)
{
return vec_1(0.0f);
vec_zero(void) {
return vec_1(0.0f);
}
#endif
@@ -72,9 +65,8 @@ vec_zero (void)
#ifndef VEC_FN_IZERO
#define VEC_FN_IZERO
static inline union vec
vec_izero (void)
{
return vec_i1(0);
vec_izero(void) {
return vec_i1(0);
}
#endif
@@ -82,9 +74,8 @@ vec_izero (void)
#ifndef VEC_FN_UZERO
#define VEC_FN_UZERO
static inline union vec
vec_uzero (void)
{
return vec_u1(0U);
vec_uzero(void) {
return vec_u1(0U);
}
#endif
@@ -92,9 +83,8 @@ vec_uzero (void)
#ifndef VEC_FN_TO_INT
#define VEC_FN_TO_INT
static inline union vec
vec_to_int (const union vec v)
{
return (union vec) { .xi = v.x, .yi = v.y, .zi = v.z, .wi = v.w };
vec_to_int(const union vec v) {
return (union vec) { .xi = v.x, .yi = v.y, .zi = v.z, .wi = v.w };
}
#endif
@@ -102,9 +92,8 @@ vec_to_int (const union vec v)
#ifndef VEC_FN_TO_FLOAT
#define VEC_FN_TO_FLOAT
static inline union vec
vec_to_float (const union vec v)
{
return (union vec) { .x = v.xi, .y = v.yi, .z = v.zi, .w = v.wi };
vec_to_float(const union vec v) {
return (union vec) { .x = v.xi, .y = v.yi, .z = v.zi, .w = v.wi };
}
#endif
@@ -112,14 +101,13 @@ vec_to_float (const union vec v)
#ifndef VEC_FN_ADD
#define VEC_FN_ADD
static inline union vec
vec_add (const union vec a, const union vec b)
{
return (union vec) {
.x = a.x + b.x,
.y = a.y + b.y,
.z = a.z + b.z,
.w = a.w + b.w,
};
vec_add(const union vec a, const union vec b) {
return (union vec) {
.x = a.x + b.x,
.y = a.y + b.y,
.z = a.z + b.z,
.w = a.w + b.w,
};
}
#endif
@@ -127,14 +115,13 @@ vec_add (const union vec a, const union vec b)
#ifndef VEC_FN_IADD
#define VEC_FN_IADD
static inline union vec
vec_iadd (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xi + b.xi,
.yi = a.yi + b.yi,
.zi = a.zi + b.zi,
.wi = a.wi + b.wi,
};
vec_iadd(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xi + b.xi,
.yi = a.yi + b.yi,
.zi = a.zi + b.zi,
.wi = a.wi + b.wi,
};
}
#endif
@@ -142,14 +129,13 @@ vec_iadd (const union vec a, const union vec b)
#ifndef VEC_FN_UADD
#define VEC_FN_UADD
static inline union vec
vec_uadd (const union vec a, const union vec b)
{
return (union vec) {
.xu = a.xu + b.xu,
.yu = a.yu + b.yu,
.zu = a.zu + b.zu,
.wu = a.wu + b.wu,
};
vec_uadd(const union vec a, const union vec b) {
return (union vec) {
.xu = a.xu + b.xu,
.yu = a.yu + b.yu,
.zu = a.zu + b.zu,
.wu = a.wu + b.wu,
};
}
#endif
@@ -157,14 +143,13 @@ vec_uadd (const union vec a, const union vec b)
#ifndef VEC_FN_SUB
#define VEC_FN_SUB
static inline union vec
vec_sub (const union vec a, const union vec b)
{
return (union vec) {
.x = a.x - b.x,
.y = a.y - b.y,
.z = a.z - b.z,
.w = a.w - b.w,
};
vec_sub(const union vec a, const union vec b) {
return (union vec) {
.x = a.x - b.x,
.y = a.y - b.y,
.z = a.z - b.z,
.w = a.w - b.w,
};
}
#endif
@@ -172,14 +157,13 @@ vec_sub (const union vec a, const union vec b)
#ifndef VEC_FN_ISUB
#define VEC_FN_ISUB
static inline union vec
vec_isub (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xi - b.xi,
.yi = a.yi - b.yi,
.zi = a.zi - b.zi,
.wi = a.wi - b.wi,
};
vec_isub(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xi - b.xi,
.yi = a.yi - b.yi,
.zi = a.zi - b.zi,
.wi = a.wi - b.wi,
};
}
#endif
@@ -187,14 +171,13 @@ vec_isub (const union vec a, const union vec b)
#ifndef VEC_FN_USUB
#define VEC_FN_USUB
static inline union vec
vec_usub (const union vec a, const union vec b)
{
return (union vec) {
.xu = a.xu - b.xu,
.yu = a.yu - b.yu,
.zu = a.zu - b.zu,
.wu = a.wu - b.wu,
};
vec_usub(const union vec a, const union vec b) {
return (union vec) {
.xu = a.xu - b.xu,
.yu = a.yu - b.yu,
.zu = a.zu - b.zu,
.wu = a.wu - b.wu,
};
}
#endif
@@ -202,14 +185,13 @@ vec_usub (const union vec a, const union vec b)
#ifndef VEC_FN_MUL
#define VEC_FN_MUL
static inline union vec
vec_mul (const union vec a, const union vec b)
{
return (union vec) {
.x = a.x * b.x,
.y = a.y * b.y,
.z = a.z * b.z,
.w = a.w * b.w,
};
vec_mul(const union vec a, const union vec b) {
return (union vec) {
.x = a.x * b.x,
.y = a.y * b.y,
.z = a.z * b.z,
.w = a.w * b.w,
};
}
#endif
@@ -217,14 +199,13 @@ vec_mul (const union vec a, const union vec b)
#ifndef VEC_FN_DIV
#define VEC_FN_DIV
static inline union vec
vec_div (const union vec a, const union vec b)
{
return (union vec) {
.x = a.x / b.x,
.y = a.y / b.y,
.z = a.z / b.z,
.w = a.w / b.w,
};
vec_div(const union vec a, const union vec b) {
return (union vec) {
.x = a.x / b.x,
.y = a.y / b.y,
.z = a.z / b.z,
.w = a.w / b.w,
};
}
#endif
@@ -232,14 +213,13 @@ vec_div (const union vec a, const union vec b)
#ifndef VEC_FN_AND
#define VEC_FN_AND
static inline union vec
vec_and (const union vec a, const union vec b)
{
return (union vec) {
.xu = a.xu & b.xu,
.yu = a.yu & b.yu,
.zu = a.zu & b.zu,
.wu = a.wu & b.wu,
};
vec_and(const union vec a, const union vec b) {
return (union vec) {
.xu = a.xu & b.xu,
.yu = a.yu & b.yu,
.zu = a.zu & b.zu,
.wu = a.wu & b.wu,
};
}
#endif
@@ -247,14 +227,13 @@ vec_and (const union vec a, const union vec b)
#ifndef VEC_FN_OR
#define VEC_FN_OR
static inline union vec
vec_or (const union vec a, const union vec b)
{
return (union vec) {
.xu = a.xu | b.xu,
.yu = a.yu | b.yu,
.zu = a.zu | b.zu,
.wu = a.wu | b.wu,
};
vec_or(const union vec a, const union vec b) {
return (union vec) {
.xu = a.xu | b.xu,
.yu = a.yu | b.yu,
.zu = a.zu | b.zu,
.wu = a.wu | b.wu,
};
}
#endif
@@ -262,14 +241,13 @@ vec_or (const union vec a, const union vec b)
#ifndef VEC_FN_XOR
#define VEC_FN_XOR
static inline union vec
vec_xor (const union vec a, const union vec b)
{
return (union vec) {
.xu = a.xu ^ b.xu,
.yu = a.yu ^ b.yu,
.zu = a.zu ^ b.zu,
.wu = a.wu ^ b.wu,
};
vec_xor(const union vec a, const union vec b) {
return (union vec) {
.xu = a.xu ^ b.xu,
.yu = a.yu ^ b.yu,
.zu = a.zu ^ b.zu,
.wu = a.wu ^ b.wu,
};
}
#endif
@@ -277,14 +255,13 @@ vec_xor (const union vec a, const union vec b)
#ifndef VEC_FN_NOT
#define VEC_FN_NOT
static inline union vec
vec_not (const union vec v)
{
return (union vec) {
.xu = ~v.xu,
.yu = ~v.yu,
.zu = ~v.zu,
.wu = ~v.wu,
};
vec_not(const union vec v) {
return (union vec) {
.xu = ~v.xu,
.yu = ~v.yu,
.zu = ~v.zu,
.wu = ~v.wu,
};
}
#endif
@@ -292,14 +269,13 @@ vec_not (const union vec v)
#ifndef VEC_FN_EQ
#define VEC_FN_EQ
static inline union vec
vec_eq (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.x == b.x ? -1 : 0,
.yi = a.y == b.y ? -1 : 0,
.zi = a.z == b.z ? -1 : 0,
.wi = a.w == b.w ? -1 : 0,
};
vec_eq(const union vec a, const union vec b) {
return (union vec) {
.xi = a.x == b.x ? -1 : 0,
.yi = a.y == b.y ? -1 : 0,
.zi = a.z == b.z ? -1 : 0,
.wi = a.w == b.w ? -1 : 0,
};
}
#endif
@@ -307,14 +283,13 @@ vec_eq (const union vec a, const union vec b)
#ifndef VEC_FN_IEQ
#define VEC_FN_IEQ
static inline union vec
vec_ieq (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xi == b.xi ? -1 : 0,
.yi = a.yi == b.yi ? -1 : 0,
.zi = a.zi == b.zi ? -1 : 0,
.wi = a.wi == b.wi ? -1 : 0,
};
vec_ieq(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xi == b.xi ? -1 : 0,
.yi = a.yi == b.yi ? -1 : 0,
.zi = a.zi == b.zi ? -1 : 0,
.wi = a.wi == b.wi ? -1 : 0,
};
}
#endif
@@ -322,14 +297,13 @@ vec_ieq (const union vec a, const union vec b)
#ifndef VEC_FN_UEQ
#define VEC_FN_UEQ
static inline union vec
vec_ueq (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xu == b.xu ? -1 : 0,
.yi = a.yu == b.yu ? -1 : 0,
.zi = a.zu == b.zu ? -1 : 0,
.wi = a.wu == b.wu ? -1 : 0,
};
vec_ueq(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xu == b.xu ? -1 : 0,
.yi = a.yu == b.yu ? -1 : 0,
.zi = a.zu == b.zu ? -1 : 0,
.wi = a.wu == b.wu ? -1 : 0,
};
}
#endif
@@ -337,14 +311,13 @@ vec_ueq (const union vec a, const union vec b)
#ifndef VEC_FN_LT
#define VEC_FN_LT
static inline union vec
vec_lt (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.x < b.x ? -1 : 0,
.yi = a.y < b.y ? -1 : 0,
.zi = a.z < b.z ? -1 : 0,
.wi = a.w < b.w ? -1 : 0,
};
vec_lt(const union vec a, const union vec b) {
return (union vec) {
.xi = a.x < b.x ? -1 : 0,
.yi = a.y < b.y ? -1 : 0,
.zi = a.z < b.z ? -1 : 0,
.wi = a.w < b.w ? -1 : 0,
};
}
#endif
@@ -352,14 +325,13 @@ vec_lt (const union vec a, const union vec b)
#ifndef VEC_FN_ILT
#define VEC_FN_ILT
static inline union vec
vec_ilt (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xi < b.xi ? -1 : 0,
.yi = a.yi < b.yi ? -1 : 0,
.zi = a.zi < b.zi ? -1 : 0,
.wi = a.wi < b.wi ? -1 : 0,
};
vec_ilt(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xi < b.xi ? -1 : 0,
.yi = a.yi < b.yi ? -1 : 0,
.zi = a.zi < b.zi ? -1 : 0,
.wi = a.wi < b.wi ? -1 : 0,
};
}
#endif
@@ -367,14 +339,13 @@ vec_ilt (const union vec a, const union vec b)
#ifndef VEC_FN_ULT
#define VEC_FN_ULT
static inline union vec
vec_ult (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xu < b.xu ? -1 : 0,
.yi = a.yu < b.yu ? -1 : 0,
.zi = a.zu < b.zu ? -1 : 0,
.wi = a.wu < b.wu ? -1 : 0,
};
vec_ult(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xu < b.xu ? -1 : 0,
.yi = a.yu < b.yu ? -1 : 0,
.zi = a.zu < b.zu ? -1 : 0,
.wi = a.wu < b.wu ? -1 : 0,
};
}
#endif
@@ -382,14 +353,13 @@ vec_ult (const union vec a, const union vec b)
#ifndef VEC_FN_LE
#define VEC_FN_LE
static inline union vec
vec_le (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.x <= b.x ? -1 : 0,
.yi = a.y <= b.y ? -1 : 0,
.zi = a.z <= b.z ? -1 : 0,
.wi = a.w <= b.w ? -1 : 0,
};
vec_le(const union vec a, const union vec b) {
return (union vec) {
.xi = a.x <= b.x ? -1 : 0,
.yi = a.y <= b.y ? -1 : 0,
.zi = a.z <= b.z ? -1 : 0,
.wi = a.w <= b.w ? -1 : 0,
};
}
#endif
@@ -397,14 +367,13 @@ vec_le (const union vec a, const union vec b)
#ifndef VEC_FN_ILE
#define VEC_FN_ILE
static inline union vec
vec_ile (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xi <= b.xi ? -1 : 0,
.yi = a.yi <= b.yi ? -1 : 0,
.zi = a.zi <= b.zi ? -1 : 0,
.wi = a.wi <= b.wi ? -1 : 0,
};
vec_ile(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xi <= b.xi ? -1 : 0,
.yi = a.yi <= b.yi ? -1 : 0,
.zi = a.zi <= b.zi ? -1 : 0,
.wi = a.wi <= b.wi ? -1 : 0,
};
}
#endif
@@ -412,14 +381,13 @@ vec_ile (const union vec a, const union vec b)
#ifndef VEC_FN_ULE
#define VEC_FN_ULE
static inline union vec
vec_ule (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xu <= b.xu ? -1 : 0,
.yi = a.yu <= b.yu ? -1 : 0,
.zi = a.zu <= b.zu ? -1 : 0,
.wi = a.wu <= b.wu ? -1 : 0,
};
vec_ule(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xu <= b.xu ? -1 : 0,
.yi = a.yu <= b.yu ? -1 : 0,
.zi = a.zu <= b.zu ? -1 : 0,
.wi = a.wu <= b.wu ? -1 : 0,
};
}
#endif
@@ -427,14 +395,13 @@ vec_ule (const union vec a, const union vec b)
#ifndef VEC_FN_GT
#define VEC_FN_GT
static inline union vec
vec_gt (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.x > b.x ? -1 : 0,
.yi = a.y > b.y ? -1 : 0,
.zi = a.z > b.z ? -1 : 0,
.wi = a.w > b.w ? -1 : 0,
};
vec_gt(const union vec a, const union vec b) {
return (union vec) {
.xi = a.x > b.x ? -1 : 0,
.yi = a.y > b.y ? -1 : 0,
.zi = a.z > b.z ? -1 : 0,
.wi = a.w > b.w ? -1 : 0,
};
}
#endif
@@ -442,14 +409,13 @@ vec_gt (const union vec a, const union vec b)
#ifndef VEC_FN_IGT
#define VEC_FN_IGT
static inline union vec
vec_igt (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xi > b.xi ? -1 : 0,
.yi = a.yi > b.yi ? -1 : 0,
.zi = a.zi > b.zi ? -1 : 0,
.wi = a.wi > b.wi ? -1 : 0,
};
vec_igt(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xi > b.xi ? -1 : 0,
.yi = a.yi > b.yi ? -1 : 0,
.zi = a.zi > b.zi ? -1 : 0,
.wi = a.wi > b.wi ? -1 : 0,
};
}
#endif
@@ -457,14 +423,13 @@ vec_igt (const union vec a, const union vec b)
#ifndef VEC_FN_UGT
#define VEC_FN_UGT
static inline union vec
vec_ugt (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xu > b.xu ? -1 : 0,
.yi = a.yu > b.yu ? -1 : 0,
.zi = a.zu > b.zu ? -1 : 0,
.wi = a.wu > b.wu ? -1 : 0,
};
vec_ugt(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xu > b.xu ? -1 : 0,
.yi = a.yu > b.yu ? -1 : 0,
.zi = a.zu > b.zu ? -1 : 0,
.wi = a.wu > b.wu ? -1 : 0,
};
}
#endif
@@ -472,14 +437,13 @@ vec_ugt (const union vec a, const union vec b)
#ifndef VEC_FN_GE
#define VEC_FN_GE
static inline union vec
vec_ge (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.x >= b.x ? -1 : 0,
.yi = a.y >= b.y ? -1 : 0,
.zi = a.z >= b.z ? -1 : 0,
.wi = a.w >= b.w ? -1 : 0,
};
vec_ge(const union vec a, const union vec b) {
return (union vec) {
.xi = a.x >= b.x ? -1 : 0,
.yi = a.y >= b.y ? -1 : 0,
.zi = a.z >= b.z ? -1 : 0,
.wi = a.w >= b.w ? -1 : 0,
};
}
#endif
@@ -487,14 +451,13 @@ vec_ge (const union vec a, const union vec b)
#ifndef VEC_FN_IGE
#define VEC_FN_IGE
static inline union vec
vec_ige (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xi >= b.xi ? -1 : 0,
.yi = a.yi >= b.yi ? -1 : 0,
.zi = a.zi >= b.zi ? -1 : 0,
.wi = a.wi >= b.wi ? -1 : 0,
};
vec_ige(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xi >= b.xi ? -1 : 0,
.yi = a.yi >= b.yi ? -1 : 0,
.zi = a.zi >= b.zi ? -1 : 0,
.wi = a.wi >= b.wi ? -1 : 0,
};
}
#endif
@@ -502,14 +465,13 @@ vec_ige (const union vec a, const union vec b)
#ifndef VEC_FN_UGE
#define VEC_FN_UGE
static inline union vec
vec_uge (const union vec a, const union vec b)
{
return (union vec) {
.xi = a.xu >= b.xu ? -1 : 0,
.yi = a.yu >= b.yu ? -1 : 0,
.zi = a.zu >= b.zu ? -1 : 0,
.wi = a.wu >= b.wu ? -1 : 0,
};
vec_uge(const union vec a, const union vec b) {
return (union vec) {
.xi = a.xu >= b.xu ? -1 : 0,
.yi = a.yu >= b.yu ? -1 : 0,
.zi = a.zu >= b.zu ? -1 : 0,
.wi = a.wu >= b.wu ? -1 : 0,
};
}
#endif
@@ -517,9 +479,8 @@ vec_uge (const union vec a, const union vec b)
#ifndef VEC_FN_DOT
#define VEC_FN_DOT
static inline float
vec_dot (const union vec a, const union vec b)
{
return a.x * b.x + a.y * b.y + a.z * b.z;
vec_dot(const union vec a, const union vec b) {
return a.x * b.x + a.y * b.y + a.z * b.z;
}
#endif
@@ -527,13 +488,12 @@ vec_dot (const union vec a, const union vec b)
#ifndef VEC_FN_CROSS
#define VEC_FN_CROSS
static inline union vec
vec_cross (const union vec a, const union vec b)
{
return (union vec) {
.x = a.y * b.z - a.z * b.y,
.y = a.z * b.x - a.x * b.z,
.z = a.x * b.y - a.y * b.x,
.w = 0.0f,
};
vec_cross(const union vec a, const union vec b) {
return (union vec) {
.x = a.y * b.z - a.z * b.y,
.y = a.z * b.x - a.x * b.z,
.z = a.x * b.y - a.y * b.x,
.w = 0.0f,
};
}
#endif
+55 -57
View File
@@ -2,34 +2,33 @@
#ifndef VEC_FN_DIV
#define VEC_FN_DIV
static inline union vec
vec_div (const union vec a, const union vec b)
{
vec_div(const union vec a, const union vec b) {
#ifdef __aarch64__
float32x4_t ret;
float32x4_t ret;
__asm__ (
"fdiv %0.4s,%1.4s,%2.4s \n\t"
__asm__(
"fdiv %0.4s,%1.4s,%2.4s \n\t"
: "=&w" (ret) // Outputs: ret=v0
: "w" (a.neon.f), "w" (b.neon.f) // Inputs: a=v1, b=v2
);
: "=&w"(ret) // Outputs: ret=v0
: "w"(a.neon.f), "w"(b.neon.f) // Inputs: a=v1, b=v2
);
#else
float32x4_t ret, tmp;
float32x4_t ret, tmp;
__asm__ (
"vrecpeq.f32 %q0,%q3 \n\t"
"vrecpsq.f32 %q1,%q3,%q0 \n\t"
"vmulq.f32 %q0,%q1,%q0 \n\t"
"vrecpsq.f32 %q1,%q3,%q0 \n\t"
"vmulq.f32 %q0,%q1,%q0 \n\t"
"vmulq.f32 %q0,%q2,%q0 \n\t"
__asm__(
"vrecpeq.f32 %q0,%q3 \n\t"
"vrecpsq.f32 %q1,%q3,%q0 \n\t"
"vmulq.f32 %q0,%q1,%q0 \n\t"
"vrecpsq.f32 %q1,%q3,%q0 \n\t"
"vmulq.f32 %q0,%q1,%q0 \n\t"
"vmulq.f32 %q0,%q2,%q0 \n\t"
: "=&w" (ret), "=&w" (tmp) // Outputs: ret=q0, tmp=q1
: "w" (a.neon.f), "w" (b.neon.f) // Inputs: a=q2, b=q3
);
: "=&w"(ret), "=&w"(tmp) // Outputs: ret=q0, tmp=q1
: "w"(a.neon.f), "w"(b.neon.f) // Inputs: a=q2, b=q3
);
#endif
return (union vec) { .neon.f = ret };
return (union vec) { .neon.f = ret };
}
#endif
@@ -37,54 +36,53 @@ vec_div (const union vec a, const union vec b)
#ifndef VEC_FN_CROSS
#define VEC_FN_CROSS
static inline union vec
vec_cross (const union vec a, const union vec b)
{
// Make read/write copies of the inputs:
float32x4_t atmp = a.neon.f, btmp = b.neon.f;
vec_cross(const union vec a, const union vec b) {
// Make read/write copies of the inputs:
float32x4_t atmp = a.neon.f, btmp = b.neon.f;
#ifdef __aarch64__
float32x4_t ret, t1, t2;
float32x4_t ret, t1, t2;
__asm__ (
"zip1 %0.4s,%3.4s,%4.4s \n\t" // v0 := a0 b0 a1 b1
"zip2 %1.4s,%3.4s,%4.4s \n\t" // v1 := a2 b2 a3 b3
"trn1 %2.4s,%3.4s,%4.4s \n\t" // v2 := a0 b0 a2 b2
"trn2 %3.4s,%3.4s,%4.4s \n\t" // v3 := a1 b1 a3 b3
"ext %0.16b,%0.16b,%0.16b,#8 \n\t" // v0 := a1 b1 a0 b0
"ext %2.16b,%2.16b,%2.16b,#8 \n\t" // v2 := a2 b2 a0 b0
"trn2 %4.4s,%0.4s,%1.4s \n\t" // v4 := b1 b2 b0 b3 = byzxw
"trn1 %0.4s,%0.4s,%1.4s \n\t" // v0 := a1 a2 a0 a3 = ayzxw
"uzp1 %1.4s,%2.4s,%3.4s \n\t" // v1 := a2 a0 a1 a3 = azxyw
"uzp2 %2.4s,%2.4s,%3.4s \n\t" // v2 := b2 b0 b1 b3 = bzxyw
__asm__(
"zip1 %0.4s,%3.4s,%4.4s \n\t" // v0 := a0 b0 a1 b1
"zip2 %1.4s,%3.4s,%4.4s \n\t" // v1 := a2 b2 a3 b3
"trn1 %2.4s,%3.4s,%4.4s \n\t" // v2 := a0 b0 a2 b2
"trn2 %3.4s,%3.4s,%4.4s \n\t" // v3 := a1 b1 a3 b3
"ext %0.16b,%0.16b,%0.16b,#8 \n\t" // v0 := a1 b1 a0 b0
"ext %2.16b,%2.16b,%2.16b,#8 \n\t" // v2 := a2 b2 a0 b0
"trn2 %4.4s,%0.4s,%1.4s \n\t" // v4 := b1 b2 b0 b3 = byzxw
"trn1 %0.4s,%0.4s,%1.4s \n\t" // v0 := a1 a2 a0 a3 = ayzxw
"uzp1 %1.4s,%2.4s,%3.4s \n\t" // v1 := a2 a0 a1 a3 = azxyw
"uzp2 %2.4s,%2.4s,%3.4s \n\t" // v2 := b2 b0 b1 b3 = bzxyw
"fmul %0.4s,%0.4s,%2.4s \n\t" // v0 := ayzxw * bzxyw
"fmls %0.4s,%1.4s,%4.4s \n\t" // v0 := ayzxw * bzxyw - azxyw * byzxw
"fmul %0.4s,%0.4s,%2.4s \n\t" // v0 := ayzxw * bzxyw
"fmls %0.4s,%1.4s,%4.4s \n\t" // v0 := ayzxw * bzxyw - azxyw * byzxw
: "=&w" (ret), "=&w" (t1), "=&w" (t2), // Outputs: ret=v0, t1=v1, t2=v2
"+w" (atmp), "+w" (btmp) // Outputs (reused inputs): a=v3, b=v4
);
: "=&w"(ret), "=&w"(t1), "=&w"(t2), // Outputs: ret=v0, t1=v1, t2=v2
"+w"(atmp), "+w"(btmp) // Outputs (reused inputs): a=v3, b=v4
);
#else
float32x4_t ret, tmp;
float32x4_t ret, tmp;
__asm__ (
"vext.8 %q0,%q2,%q2,#4 \n\t" // q0 := ayzwx
"vext.8 %q1,%q3,%q3,#4 \n\t" // q1 := byzwx
"vrev64.32 %f0,%f0 \n\t" // q0 := ayzxw
"vrev64.32 %f1,%f1 \n\t" // q1 := byzxw
__asm__(
"vext.8 %q0,%q2,%q2,#4 \n\t" // q0 := ayzwx
"vext.8 %q1,%q3,%q3,#4 \n\t" // q1 := byzwx
"vrev64.32 %f0,%f0 \n\t" // q0 := ayzxw
"vrev64.32 %f1,%f1 \n\t" // q1 := byzxw
"vtrn.32 %e2,%f2 \n\t" // q2 := axzyw
"vtrn.32 %e3,%f3 \n\t" // q3 := bxzyw
"vrev64.32 %e2,%e2 \n\t" // q2 := azxyw
"vrev64.32 %e3,%e3 \n\t" // q3 := bzxyw
"vtrn.32 %e2,%f2 \n\t" // q2 := axzyw
"vtrn.32 %e3,%f3 \n\t" // q3 := bxzyw
"vrev64.32 %e2,%e2 \n\t" // q2 := azxyw
"vrev64.32 %e3,%e3 \n\t" // q3 := bzxyw
"vmulq.f32 %q0,%q0,%q3 \n\t" // q0 := ayzxw * bzxyw
"vmlsq.f32 %q0,%q2,%q1 \n\t" // q0 := ayzxw * bzxyw - azxyw * byzxw
"vmulq.f32 %q0,%q0,%q3 \n\t" // q0 := ayzxw * bzxyw
"vmlsq.f32 %q0,%q2,%q1 \n\t" // q0 := ayzxw * bzxyw - azxyw * byzxw
: "=&w" (ret), "=&w" (tmp), // Outputs: ret=q0, tmp=q1
"+w" (atmp), "+w" (btmp) // Outputs (reused inputs): a=q2, b=q3
);
: "=&w"(ret), "=&w"(tmp), // Outputs: ret=q0, tmp=q1
"+w"(atmp), "+w"(btmp) // Outputs (reused inputs): a=q2, b=q3
);
#endif
return (union vec) { .neon.f = ret };
return (union vec) { .neon.f = ret };
}
#endif
+112 -152
View File
@@ -2,15 +2,14 @@
#ifndef VEC_FN_
#define VEC_FN_
static inline union vec
vec (const float x, const float y, const float z, const float w)
{
union {
float i[2];
uint64_t u;
} lo = { { x, y } }
, hi = { { z, w } };
vec(const float x, const float y, const float z, const float w) {
union {
float i[2];
uint64_t u;
} lo = { { x, y } }
, hi = { { z, w } };
return (union vec) { .neon.f = vcombine_f32(vcreate_f32(lo.u), vcreate_f32(hi.u)) };
return (union vec) { .neon.f = vcombine_f32(vcreate_f32(lo.u), vcreate_f32(hi.u)) };
}
#endif
@@ -18,15 +17,14 @@ vec (const float x, const float y, const float z, const float w)
#ifndef VEC_FN_I
#define VEC_FN_I
static inline union vec
vec_i (const int32_t x, const int32_t y, const int32_t z, const int32_t w)
{
union {
int32_t i[2];
uint64_t u;
} lo = { { x, y } }
, hi = { { z, w } };
vec_i(const int32_t x, const int32_t y, const int32_t z, const int32_t w) {
union {
int32_t i[2];
uint64_t u;
} lo = { { x, y } }
, hi = { { z, w } };
return (union vec) { .neon.i = vcombine_s32(vcreate_s32(lo.u), vcreate_s32(hi.u)) };
return (union vec) { .neon.i = vcombine_s32(vcreate_s32(lo.u), vcreate_s32(hi.u)) };
}
#endif
@@ -34,15 +32,14 @@ vec_i (const int32_t x, const int32_t y, const int32_t z, const int32_t w)
#ifndef VEC_FN_U
#define VEC_FN_U
static inline union vec
vec_u (const uint32_t x, const uint32_t y, const uint32_t z, const uint32_t w)
{
union {
uint32_t i[2];
uint64_t u;
} lo = { { x, y } }
, hi = { { z, w } };
vec_u(const uint32_t x, const uint32_t y, const uint32_t z, const uint32_t w) {
union {
uint32_t i[2];
uint64_t u;
} lo = { { x, y } }
, hi = { { z, w } };
return (union vec) { .neon.u = vcombine_u32(vcreate_u32(lo.u), vcreate_u32(hi.u)) };
return (union vec) { .neon.u = vcombine_u32(vcreate_u32(lo.u), vcreate_u32(hi.u)) };
}
#endif
@@ -50,9 +47,8 @@ vec_u (const uint32_t x, const uint32_t y, const uint32_t z, const uint32_t w)
#ifndef VEC_FN_1
#define VEC_FN_1
static inline union vec
vec_1 (const float val)
{
return (union vec) { .neon.f = vdupq_n_f32(val) };
vec_1(const float val) {
return (union vec) { .neon.f = vdupq_n_f32(val) };
}
#endif
@@ -60,9 +56,8 @@ vec_1 (const float val)
#ifndef VEC_FN_I1
#define VEC_FN_I1
static inline union vec
vec_i1 (const int32_t val)
{
return (union vec) { .neon.i = vdupq_n_s32(val) };
vec_i1(const int32_t val) {
return (union vec) { .neon.i = vdupq_n_s32(val) };
}
#endif
@@ -70,9 +65,8 @@ vec_i1 (const int32_t val)
#ifndef VEC_FN_U1
#define VEC_FN_U1
static inline union vec
vec_u1 (const uint32_t val)
{
return (union vec) { .neon.u = vdupq_n_u32(val) };
vec_u1(const uint32_t val) {
return (union vec) { .neon.u = vdupq_n_u32(val) };
}
#endif
@@ -80,9 +74,8 @@ vec_u1 (const uint32_t val)
#ifndef VEC_FN_ZERO
#define VEC_FN_ZERO
static inline union vec
vec_zero (void)
{
return vec_1(0.0f);
vec_zero(void) {
return vec_1(0.0f);
}
#endif
@@ -90,9 +83,8 @@ vec_zero (void)
#ifndef VEC_FN_IZERO
#define VEC_FN_IZERO
static inline union vec
vec_izero (void)
{
return vec_i1(0);
vec_izero(void) {
return vec_i1(0);
}
#endif
@@ -100,9 +92,8 @@ vec_izero (void)
#ifndef VEC_FN_UZERO
#define VEC_FN_UZERO
static inline union vec
vec_uzero (void)
{
return vec_u1(0U);
vec_uzero(void) {
return vec_u1(0U);
}
#endif
@@ -110,9 +101,8 @@ vec_uzero (void)
#ifndef VEC_FN_TO_INT
#define VEC_FN_TO_INT
static inline union vec
vec_to_int (const union vec v)
{
return (union vec) { .neon.i = vcvtq_s32_f32(v.neon.f) };
vec_to_int(const union vec v) {
return (union vec) { .neon.i = vcvtq_s32_f32(v.neon.f) };
}
#endif
@@ -120,9 +110,8 @@ vec_to_int (const union vec v)
#ifndef VEC_FN_TO_FLOAT
#define VEC_FN_TO_FLOAT
static inline union vec
vec_to_float (const union vec v)
{
return (union vec) { .neon.f = vcvtq_f32_s32(v.neon.i) };
vec_to_float(const union vec v) {
return (union vec) { .neon.f = vcvtq_f32_s32(v.neon.i) };
}
#endif
@@ -130,9 +119,8 @@ vec_to_float (const union vec v)
#ifndef VEC_FN_ADD
#define VEC_FN_ADD
static inline union vec
vec_add (const union vec a, const union vec b)
{
return (union vec) { .neon.f = vaddq_f32(a.neon.f, b.neon.f) };
vec_add(const union vec a, const union vec b) {
return (union vec) { .neon.f = vaddq_f32(a.neon.f, b.neon.f) };
}
#endif
@@ -140,9 +128,8 @@ vec_add (const union vec a, const union vec b)
#ifndef VEC_FN_IADD
#define VEC_FN_IADD
static inline union vec
vec_iadd (const union vec a, const union vec b)
{
return (union vec) { .neon.i = vaddq_s32(a.neon.i, b.neon.i) };
vec_iadd(const union vec a, const union vec b) {
return (union vec) { .neon.i = vaddq_s32(a.neon.i, b.neon.i) };
}
#endif
@@ -150,9 +137,8 @@ vec_iadd (const union vec a, const union vec b)
#ifndef VEC_FN_UADD
#define VEC_FN_UADD
static inline union vec
vec_uadd (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vaddq_u32(a.neon.u, b.neon.u) };
vec_uadd(const union vec a, const union vec b) {
return (union vec) { .neon.u = vaddq_u32(a.neon.u, b.neon.u) };
}
#endif
@@ -160,9 +146,8 @@ vec_uadd (const union vec a, const union vec b)
#ifndef VEC_FN_SUB
#define VEC_FN_SUB
static inline union vec
vec_sub (const union vec a, const union vec b)
{
return (union vec) { .neon.f = vsubq_f32(a.neon.f, b.neon.f) };
vec_sub(const union vec a, const union vec b) {
return (union vec) { .neon.f = vsubq_f32(a.neon.f, b.neon.f) };
}
#endif
@@ -170,9 +155,8 @@ vec_sub (const union vec a, const union vec b)
#ifndef VEC_FN_ISUB
#define VEC_FN_ISUB
static inline union vec
vec_isub (const union vec a, const union vec b)
{
return (union vec) { .neon.i = vsubq_s32(a.neon.i, b.neon.i) };
vec_isub(const union vec a, const union vec b) {
return (union vec) { .neon.i = vsubq_s32(a.neon.i, b.neon.i) };
}
#endif
@@ -180,9 +164,8 @@ vec_isub (const union vec a, const union vec b)
#ifndef VEC_FN_USUB
#define VEC_FN_USUB
static inline union vec
vec_usub (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vsubq_u32(a.neon.u, b.neon.u) };
vec_usub(const union vec a, const union vec b) {
return (union vec) { .neon.u = vsubq_u32(a.neon.u, b.neon.u) };
}
#endif
@@ -190,9 +173,8 @@ vec_usub (const union vec a, const union vec b)
#ifndef VEC_FN_MUL
#define VEC_FN_MUL
static inline union vec
vec_mul (const union vec a, const union vec b)
{
return (union vec) { .neon.f = vmulq_f32(a.neon.f, b.neon.f) };
vec_mul(const union vec a, const union vec b) {
return (union vec) { .neon.f = vmulq_f32(a.neon.f, b.neon.f) };
}
#endif
@@ -200,17 +182,16 @@ vec_mul (const union vec a, const union vec b)
#ifndef VEC_FN_DIV
#define VEC_FN_DIV
static inline union vec
vec_div (const union vec a, const union vec b)
{
// Estimate reciprocal of b:
float32x4_t recp = vrecpeq_f32(b.neon.f);
vec_div(const union vec a, const union vec b) {
// Estimate reciprocal of b:
float32x4_t recp = vrecpeq_f32(b.neon.f);
// Refine the estimate using Newton-Raphson:
recp = vmulq_f32(vrecpsq_f32(b.neon.f, recp), recp);
recp = vmulq_f32(vrecpsq_f32(b.neon.f, recp), recp);
// Refine the estimate using Newton-Raphson:
recp = vmulq_f32(vrecpsq_f32(b.neon.f, recp), recp);
recp = vmulq_f32(vrecpsq_f32(b.neon.f, recp), recp);
// Multiply by the reciprocal to approximate a / b:
return (union vec) { .neon.f = vmulq_f32(a.neon.f, recp) };
// Multiply by the reciprocal to approximate a / b:
return (union vec) { .neon.f = vmulq_f32(a.neon.f, recp) };
}
#endif
@@ -218,9 +199,8 @@ vec_div (const union vec a, const union vec b)
#ifndef VEC_FN_AND
#define VEC_FN_AND
static inline union vec
vec_and (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vandq_u32(a.neon.u, b.neon.u) };
vec_and(const union vec a, const union vec b) {
return (union vec) { .neon.u = vandq_u32(a.neon.u, b.neon.u) };
}
#endif
@@ -228,9 +208,8 @@ vec_and (const union vec a, const union vec b)
#ifndef VEC_FN_OR
#define VEC_FN_OR
static inline union vec
vec_or (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vorrq_u32(a.neon.u, b.neon.u) };
vec_or(const union vec a, const union vec b) {
return (union vec) { .neon.u = vorrq_u32(a.neon.u, b.neon.u) };
}
#endif
@@ -238,9 +217,8 @@ vec_or (const union vec a, const union vec b)
#ifndef VEC_FN_XOR
#define VEC_FN_XOR
static inline union vec
vec_xor (const union vec a, const union vec b)
{
return (union vec) { .neon.u = veorq_u32(a.neon.u, b.neon.u) };
vec_xor(const union vec a, const union vec b) {
return (union vec) { .neon.u = veorq_u32(a.neon.u, b.neon.u) };
}
#endif
@@ -248,9 +226,8 @@ vec_xor (const union vec a, const union vec b)
#ifndef VEC_FN_NOT
#define VEC_FN_NOT
static inline union vec
vec_not (const union vec v)
{
return (union vec) { .neon.u = vmvnq_u32(v.neon.u) };
vec_not(const union vec v) {
return (union vec) { .neon.u = vmvnq_u32(v.neon.u) };
}
#endif
@@ -258,9 +235,8 @@ vec_not (const union vec v)
#ifndef VEC_FN_EQ
#define VEC_FN_EQ
static inline union vec
vec_eq (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vceqq_f32(a.neon.f, b.neon.f) };
vec_eq(const union vec a, const union vec b) {
return (union vec) { .neon.u = vceqq_f32(a.neon.f, b.neon.f) };
}
#endif
@@ -268,9 +244,8 @@ vec_eq (const union vec a, const union vec b)
#ifndef VEC_FN_IEQ
#define VEC_FN_IEQ
static inline union vec
vec_ieq (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vceqq_s32(a.neon.i, b.neon.i) };
vec_ieq(const union vec a, const union vec b) {
return (union vec) { .neon.u = vceqq_s32(a.neon.i, b.neon.i) };
}
#endif
@@ -278,9 +253,8 @@ vec_ieq (const union vec a, const union vec b)
#ifndef VEC_FN_UEQ
#define VEC_FN_UEQ
static inline union vec
vec_ueq (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vceqq_u32(a.neon.u, b.neon.u) };
vec_ueq(const union vec a, const union vec b) {
return (union vec) { .neon.u = vceqq_u32(a.neon.u, b.neon.u) };
}
#endif
@@ -288,9 +262,8 @@ vec_ueq (const union vec a, const union vec b)
#ifndef VEC_FN_LT
#define VEC_FN_LT
static inline union vec
vec_lt (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcltq_f32(a.neon.f, b.neon.f) };
vec_lt(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcltq_f32(a.neon.f, b.neon.f) };
}
#endif
@@ -298,9 +271,8 @@ vec_lt (const union vec a, const union vec b)
#ifndef VEC_FN_ILT
#define VEC_FN_ILT
static inline union vec
vec_ilt (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcltq_s32(a.neon.i, b.neon.i) };
vec_ilt(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcltq_s32(a.neon.i, b.neon.i) };
}
#endif
@@ -308,9 +280,8 @@ vec_ilt (const union vec a, const union vec b)
#ifndef VEC_FN_ULT
#define VEC_FN_ULT
static inline union vec
vec_ult (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcltq_u32(a.neon.u, b.neon.u) };
vec_ult(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcltq_u32(a.neon.u, b.neon.u) };
}
#endif
@@ -318,9 +289,8 @@ vec_ult (const union vec a, const union vec b)
#ifndef VEC_FN_LE
#define VEC_FN_LE
static inline union vec
vec_le (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcleq_f32(a.neon.f, b.neon.f) };
vec_le(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcleq_f32(a.neon.f, b.neon.f) };
}
#endif
@@ -328,9 +298,8 @@ vec_le (const union vec a, const union vec b)
#ifndef VEC_FN_ILE
#define VEC_FN_ILE
static inline union vec
vec_ile (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcleq_s32(a.neon.i, b.neon.i) };
vec_ile(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcleq_s32(a.neon.i, b.neon.i) };
}
#endif
@@ -338,9 +307,8 @@ vec_ile (const union vec a, const union vec b)
#ifndef VEC_FN_ULE
#define VEC_FN_ULE
static inline union vec
vec_ule (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcleq_u32(a.neon.u, b.neon.u) };
vec_ule(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcleq_u32(a.neon.u, b.neon.u) };
}
#endif
@@ -348,9 +316,8 @@ vec_ule (const union vec a, const union vec b)
#ifndef VEC_FN_GT
#define VEC_FN_GT
static inline union vec
vec_gt (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcgtq_f32(a.neon.f, b.neon.f) };
vec_gt(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcgtq_f32(a.neon.f, b.neon.f) };
}
#endif
@@ -358,9 +325,8 @@ vec_gt (const union vec a, const union vec b)
#ifndef VEC_FN_IGT
#define VEC_FN_IGT
static inline union vec
vec_igt (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcgtq_s32(a.neon.i, b.neon.i) };
vec_igt(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcgtq_s32(a.neon.i, b.neon.i) };
}
#endif
@@ -368,9 +334,8 @@ vec_igt (const union vec a, const union vec b)
#ifndef VEC_FN_UGT
#define VEC_FN_UGT
static inline union vec
vec_ugt (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcgtq_u32(a.neon.u, b.neon.u) };
vec_ugt(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcgtq_u32(a.neon.u, b.neon.u) };
}
#endif
@@ -378,9 +343,8 @@ vec_ugt (const union vec a, const union vec b)
#ifndef VEC_FN_GE
#define VEC_FN_GE
static inline union vec
vec_ge (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcgeq_f32(a.neon.f, b.neon.f) };
vec_ge(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcgeq_f32(a.neon.f, b.neon.f) };
}
#endif
@@ -388,9 +352,8 @@ vec_ge (const union vec a, const union vec b)
#ifndef VEC_FN_IGE
#define VEC_FN_IGE
static inline union vec
vec_ige (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcgeq_s32(a.neon.i, b.neon.i) };
vec_ige(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcgeq_s32(a.neon.i, b.neon.i) };
}
#endif
@@ -398,9 +361,8 @@ vec_ige (const union vec a, const union vec b)
#ifndef VEC_FN_UGE
#define VEC_FN_UGE
static inline union vec
vec_uge (const union vec a, const union vec b)
{
return (union vec) { .neon.u = vcgeq_u32(a.neon.u, b.neon.u) };
vec_uge(const union vec a, const union vec b) {
return (union vec) { .neon.u = vcgeq_u32(a.neon.u, b.neon.u) };
}
#endif
@@ -408,12 +370,11 @@ vec_uge (const union vec a, const union vec b)
#ifndef VEC_FN_DOT
#define VEC_FN_DOT
static inline float
vec_dot (const union vec a, const union vec b)
{
const float32x4_t mul = vmulq_f32(a.neon.f, b.neon.f);
return vgetq_lane_f32(mul, 0)
+ vgetq_lane_f32(mul, 1)
+ vgetq_lane_f32(mul, 2);
vec_dot(const union vec a, const union vec b) {
const float32x4_t mul = vmulq_f32(a.neon.f, b.neon.f);
return vgetq_lane_f32(mul, 0)
+ vgetq_lane_f32(mul, 1)
+ vgetq_lane_f32(mul, 2);
}
#endif
@@ -421,20 +382,19 @@ vec_dot (const union vec a, const union vec b)
#ifndef VEC_FN_CROSS
#define VEC_FN_CROSS
static inline union vec
vec_cross (const union vec a, const union vec b)
{
const float32x4_t ayzwx = (float32x4_t) vextq_u32((uint32x4_t) a.neon.f, (uint32x4_t) a.neon.f, 1);
const float32x4_t byzwx = (float32x4_t) vextq_u32((uint32x4_t) b.neon.f, (uint32x4_t) b.neon.f, 1);
vec_cross(const union vec a, const union vec b) {
const float32x4_t ayzwx = (float32x4_t) vextq_u32((uint32x4_t) a.neon.f, (uint32x4_t) a.neon.f, 1);
const float32x4_t byzwx = (float32x4_t) vextq_u32((uint32x4_t) b.neon.f, (uint32x4_t) b.neon.f, 1);
const float32x4_t ayzxw = vcombine_f32(vget_low_f32(ayzwx), vrev64_f32(vget_high_f32(ayzwx)));
const float32x4_t byzxw = vcombine_f32(vget_low_f32(byzwx), vrev64_f32(vget_high_f32(byzwx)));
const float32x4_t ayzxw = vcombine_f32(vget_low_f32(ayzwx), vrev64_f32(vget_high_f32(ayzwx)));
const float32x4_t byzxw = vcombine_f32(vget_low_f32(byzwx), vrev64_f32(vget_high_f32(byzwx)));
const float32x2x2_t azxwy_pair = vtrn_f32(vget_high_f32(a.neon.f), vget_low_f32(a.neon.f));
const float32x2x2_t bzxwy_pair = vtrn_f32(vget_high_f32(b.neon.f), vget_low_f32(b.neon.f));
const float32x2x2_t azxwy_pair = vtrn_f32(vget_high_f32(a.neon.f), vget_low_f32(a.neon.f));
const float32x2x2_t bzxwy_pair = vtrn_f32(vget_high_f32(b.neon.f), vget_low_f32(b.neon.f));
const float32x4_t azxyw = vcombine_f32(azxwy_pair.val[0], vrev64_f32(azxwy_pair.val[1]));
const float32x4_t bzxyw = vcombine_f32(bzxwy_pair.val[0], vrev64_f32(bzxwy_pair.val[1]));
const float32x4_t azxyw = vcombine_f32(azxwy_pair.val[0], vrev64_f32(azxwy_pair.val[1]));
const float32x4_t bzxyw = vcombine_f32(bzxwy_pair.val[0], vrev64_f32(bzxwy_pair.val[1]));
return (union vec) { .neon.f = vmlsq_f32(vmulq_f32(ayzxw, bzxyw), byzxw, azxyw) };
return (union vec) { .neon.f = vmlsq_f32(vmulq_f32(ayzxw, bzxyw), byzxw, azxyw) };
}
#endif
+114 -146
View File
@@ -2,9 +2,8 @@
#ifndef VEC_FN_
#define VEC_FN_
static inline union vec
vec (const float x, const float y, const float z, const float w)
{
return (union vec) { .sse.f = _mm_setr_ps(x, y, z, w) };
vec(const float x, const float y, const float z, const float w) {
return (union vec) { .sse.f = _mm_setr_ps(x, y, z, w) };
}
#endif
@@ -12,9 +11,8 @@ vec (const float x, const float y, const float z, const float w)
#ifndef VEC_FN_I
#define VEC_FN_I
static inline union vec
vec_i (const int32_t x, const int32_t y, const int32_t z, const int32_t w)
{
return (union vec) { .sse.i = _mm_setr_epi32(x, y, z, w) };
vec_i(const int32_t x, const int32_t y, const int32_t z, const int32_t w) {
return (union vec) { .sse.i = _mm_setr_epi32(x, y, z, w) };
}
#endif
@@ -22,9 +20,8 @@ vec_i (const int32_t x, const int32_t y, const int32_t z, const int32_t w)
#ifndef VEC_FN_U
#define VEC_FN_U
static inline union vec
vec_u (const uint32_t x, const uint32_t y, const uint32_t z, const uint32_t w)
{
return vec_i((int32_t) x, (int32_t) y, (int32_t) z, (int32_t) w);
vec_u(const uint32_t x, const uint32_t y, const uint32_t z, const uint32_t w) {
return vec_i((int32_t) x, (int32_t) y, (int32_t) z, (int32_t) w);
}
#endif
@@ -32,9 +29,8 @@ vec_u (const uint32_t x, const uint32_t y, const uint32_t z, const uint32_t w)
#ifndef VEC_FN_1
#define VEC_FN_1
static inline union vec
vec_1 (const float val)
{
return (union vec) { .sse.f = _mm_set1_ps(val) };
vec_1(const float val) {
return (union vec) { .sse.f = _mm_set1_ps(val) };
}
#endif
@@ -42,9 +38,8 @@ vec_1 (const float val)
#ifndef VEC_FN_I1
#define VEC_FN_I1
static inline union vec
vec_i1 (const int32_t val)
{
return (union vec) { .sse.i = _mm_set1_epi32(val) };
vec_i1(const int32_t val) {
return (union vec) { .sse.i = _mm_set1_epi32(val) };
}
#endif
@@ -52,9 +47,8 @@ vec_i1 (const int32_t val)
#ifndef VEC_FN_U1
#define VEC_FN_U1
static inline union vec
vec_u1 (const uint32_t val)
{
return vec_i1((int32_t) val);
vec_u1(const uint32_t val) {
return vec_i1((int32_t) val);
}
#endif
@@ -62,9 +56,8 @@ vec_u1 (const uint32_t val)
#ifndef VEC_FN_ZERO
#define VEC_FN_ZERO
static inline union vec
vec_zero (void)
{
return (union vec) { .sse.f = _mm_setzero_ps() };
vec_zero(void) {
return (union vec) { .sse.f = _mm_setzero_ps() };
}
#endif
@@ -72,9 +65,8 @@ vec_zero (void)
#ifndef VEC_FN_IZERO
#define VEC_FN_IZERO
static inline union vec
vec_izero (void)
{
return (union vec) { .sse.i = _mm_setzero_si128() };
vec_izero(void) {
return (union vec) { .sse.i = _mm_setzero_si128() };
}
#endif
@@ -82,9 +74,8 @@ vec_izero (void)
#ifndef VEC_FN_UZERO
#define VEC_FN_UZERO
static inline union vec
vec_uzero (void)
{
return vec_izero();
vec_uzero(void) {
return vec_izero();
}
#endif
@@ -92,9 +83,8 @@ vec_uzero (void)
#ifndef VEC_FN_TO_INT
#define VEC_FN_TO_INT
static inline union vec
vec_to_int (const union vec v)
{
return (union vec) { .sse.i = _mm_cvttps_epi32(v.sse.f) };
vec_to_int(const union vec v) {
return (union vec) { .sse.i = _mm_cvttps_epi32(v.sse.f) };
}
#endif
@@ -102,9 +92,8 @@ vec_to_int (const union vec v)
#ifndef VEC_FN_TO_FLOAT
#define VEC_FN_TO_FLOAT
static inline union vec
vec_to_float (const union vec v)
{
return (union vec) { .sse.f = _mm_cvtepi32_ps(v.sse.i) };
vec_to_float(const union vec v) {
return (union vec) { .sse.f = _mm_cvtepi32_ps(v.sse.i) };
}
#endif
@@ -112,9 +101,8 @@ vec_to_float (const union vec v)
#ifndef VEC_FN_ADD
#define VEC_FN_ADD
static inline union vec
vec_add (const union vec a, const union vec b)
{
return (union vec) { .sse.f = _mm_add_ps(a.sse.f, b.sse.f) };
vec_add(const union vec a, const union vec b) {
return (union vec) { .sse.f = _mm_add_ps(a.sse.f, b.sse.f) };
}
#endif
@@ -122,9 +110,8 @@ vec_add (const union vec a, const union vec b)
#ifndef VEC_FN_IADD
#define VEC_FN_IADD
static inline union vec
vec_iadd (const union vec a, const union vec b)
{
return (union vec) { .sse.i = _mm_add_epi32(a.sse.i, b.sse.i) };
vec_iadd(const union vec a, const union vec b) {
return (union vec) { .sse.i = _mm_add_epi32(a.sse.i, b.sse.i) };
}
#endif
@@ -132,9 +119,8 @@ vec_iadd (const union vec a, const union vec b)
#ifndef VEC_FN_UADD
#define VEC_FN_UADD
static inline union vec
vec_uadd (const union vec a, const union vec b)
{
return vec_iadd(a, b);
vec_uadd(const union vec a, const union vec b) {
return vec_iadd(a, b);
}
#endif
@@ -142,9 +128,8 @@ vec_uadd (const union vec a, const union vec b)
#ifndef VEC_FN_SUB
#define VEC_FN_SUB
static inline union vec
vec_sub (const union vec a, const union vec b)
{
return (union vec) { .sse.f = _mm_sub_ps(a.sse.f, b.sse.f) };
vec_sub(const union vec a, const union vec b) {
return (union vec) { .sse.f = _mm_sub_ps(a.sse.f, b.sse.f) };
}
#endif
@@ -152,9 +137,8 @@ vec_sub (const union vec a, const union vec b)
#ifndef VEC_FN_ISUB
#define VEC_FN_ISUB
static inline union vec
vec_isub (const union vec a, const union vec b)
{
return (union vec) { .sse.i = _mm_sub_epi32(a.sse.i, b.sse.i) };
vec_isub(const union vec a, const union vec b) {
return (union vec) { .sse.i = _mm_sub_epi32(a.sse.i, b.sse.i) };
}
#endif
@@ -162,9 +146,8 @@ vec_isub (const union vec a, const union vec b)
#ifndef VEC_FN_USUB
#define VEC_FN_USUB
static inline union vec
vec_usub (const union vec a, const union vec b)
{
return vec_isub(a, b);
vec_usub(const union vec a, const union vec b) {
return vec_isub(a, b);
}
#endif
@@ -172,9 +155,8 @@ vec_usub (const union vec a, const union vec b)
#ifndef VEC_FN_MUL
#define VEC_FN_MUL
static inline union vec
vec_mul (const union vec a, const union vec b)
{
return (union vec) { .sse.f = _mm_mul_ps(a.sse.f, b.sse.f) };
vec_mul(const union vec a, const union vec b) {
return (union vec) { .sse.f = _mm_mul_ps(a.sse.f, b.sse.f) };
}
#endif
@@ -182,9 +164,8 @@ vec_mul (const union vec a, const union vec b)
#ifndef VEC_FN_DIV
#define VEC_FN_DIV
static inline union vec
vec_div (const union vec a, const union vec b)
{
return (union vec) { .sse.f = _mm_div_ps(a.sse.f, b.sse.f) };
vec_div(const union vec a, const union vec b) {
return (union vec) { .sse.f = _mm_div_ps(a.sse.f, b.sse.f) };
}
#endif
@@ -192,9 +173,8 @@ vec_div (const union vec a, const union vec b)
#ifndef VEC_FN_AND
#define VEC_FN_AND
static inline union vec
vec_and (const union vec a, const union vec b)
{
return (union vec) { .sse.u = _mm_and_si128(a.sse.u, b.sse.u) };
vec_and(const union vec a, const union vec b) {
return (union vec) { .sse.u = _mm_and_si128(a.sse.u, b.sse.u) };
}
#endif
@@ -202,9 +182,8 @@ vec_and (const union vec a, const union vec b)
#ifndef VEC_FN_OR
#define VEC_FN_OR
static inline union vec
vec_or (const union vec a, const union vec b)
{
return (union vec) { .sse.u = _mm_or_si128(a.sse.u, b.sse.u) };
vec_or(const union vec a, const union vec b) {
return (union vec) { .sse.u = _mm_or_si128(a.sse.u, b.sse.u) };
}
#endif
@@ -212,9 +191,8 @@ vec_or (const union vec a, const union vec b)
#ifndef VEC_FN_XOR
#define VEC_FN_XOR
static inline union vec
vec_xor (const union vec a, const union vec b)
{
return (union vec) { .sse.u = _mm_xor_si128(a.sse.u, b.sse.u) };
vec_xor(const union vec a, const union vec b) {
return (union vec) { .sse.u = _mm_xor_si128(a.sse.u, b.sse.u) };
}
#endif
@@ -222,11 +200,10 @@ vec_xor (const union vec a, const union vec b)
#ifndef VEC_FN_NOT
#define VEC_FN_NOT
static inline union vec
vec_not (const union vec v)
{
const __m128i allset = _mm_set1_epi32(INT32_C(0xFFFFFFFF));
vec_not(const union vec v) {
const __m128i allset = _mm_set1_epi32(INT32_C(0xFFFFFFFF));
return (union vec) { .sse.u = _mm_xor_si128(v.sse.u, allset) };
return (union vec) { .sse.u = _mm_xor_si128(v.sse.u, allset) };
}
#endif
@@ -234,9 +211,8 @@ vec_not (const union vec v)
#ifndef VEC_FN_EQ
#define VEC_FN_EQ
static inline union vec
vec_eq (const union vec a, const union vec b)
{
return (union vec) { .sse.f = _mm_cmpeq_ps(a.sse.f, b.sse.f) };
vec_eq(const union vec a, const union vec b) {
return (union vec) { .sse.f = _mm_cmpeq_ps(a.sse.f, b.sse.f) };
}
#endif
@@ -244,9 +220,8 @@ vec_eq (const union vec a, const union vec b)
#ifndef VEC_FN_IEQ
#define VEC_FN_IEQ
static inline union vec
vec_ieq (const union vec a, const union vec b)
{
return (union vec) { .sse.i = _mm_cmpeq_epi32(a.sse.i, b.sse.i) };
vec_ieq(const union vec a, const union vec b) {
return (union vec) { .sse.i = _mm_cmpeq_epi32(a.sse.i, b.sse.i) };
}
#endif
@@ -254,9 +229,8 @@ vec_ieq (const union vec a, const union vec b)
#ifndef VEC_FN_UEQ
#define VEC_FN_UEQ
static inline union vec
vec_ueq (const union vec a, const union vec b)
{
return (union vec) { .sse.u = _mm_cmpeq_epi32(a.sse.u, b.sse.u) };
vec_ueq(const union vec a, const union vec b) {
return (union vec) { .sse.u = _mm_cmpeq_epi32(a.sse.u, b.sse.u) };
}
#endif
@@ -264,9 +238,8 @@ vec_ueq (const union vec a, const union vec b)
#ifndef VEC_FN_LT
#define VEC_FN_LT
static inline union vec
vec_lt (const union vec a, const union vec b)
{
return (union vec) { .sse.f = _mm_cmplt_ps(a.sse.f, b.sse.f) };
vec_lt(const union vec a, const union vec b) {
return (union vec) { .sse.f = _mm_cmplt_ps(a.sse.f, b.sse.f) };
}
#endif
@@ -274,9 +247,8 @@ vec_lt (const union vec a, const union vec b)
#ifndef VEC_FN_ILT
#define VEC_FN_ILT
static inline union vec
vec_ilt (const union vec a, const union vec b)
{
return (union vec) { .sse.i = _mm_cmplt_epi32(a.sse.i, b.sse.i) };
vec_ilt(const union vec a, const union vec b) {
return (union vec) { .sse.i = _mm_cmplt_epi32(a.sse.i, b.sse.i) };
}
#endif
@@ -284,13 +256,14 @@ vec_ilt (const union vec a, const union vec b)
#ifndef VEC_FN_ULT
#define VEC_FN_ULT
static inline union vec
vec_ult (const union vec a, const union vec b)
{
const __m128i bias = _mm_set1_epi32(INT32_C(0x80000000));
vec_ult(const union vec a, const union vec b) {
const __m128i bias = _mm_set1_epi32(INT32_C(0x80000000));
return (union vec) { .sse.i = _mm_cmplt_epi32(
_mm_sub_epi32(a.sse.i, bias),
_mm_sub_epi32(b.sse.i, bias)) };
return (union vec) {
.sse.i = _mm_cmplt_epi32(
_mm_sub_epi32(a.sse.i, bias),
_mm_sub_epi32(b.sse.i, bias))
};
}
#endif
@@ -298,9 +271,8 @@ vec_ult (const union vec a, const union vec b)
#ifndef VEC_FN_LE
#define VEC_FN_LE
static inline union vec
vec_le (const union vec a, const union vec b)
{
return (union vec) { .sse.f = _mm_cmple_ps(a.sse.f, b.sse.f) };
vec_le(const union vec a, const union vec b) {
return (union vec) { .sse.f = _mm_cmple_ps(a.sse.f, b.sse.f) };
}
#endif
@@ -308,11 +280,12 @@ vec_le (const union vec a, const union vec b)
#ifndef VEC_FN_ILE
#define VEC_FN_ILE
static inline union vec
vec_ile (const union vec a, const union vec b)
{
return (union vec) { .sse.i = _mm_or_si128(
_mm_cmplt_epi32(a.sse.i, b.sse.i),
_mm_cmpeq_epi32(a.sse.i, b.sse.i)) };
vec_ile(const union vec a, const union vec b) {
return (union vec) {
.sse.i = _mm_or_si128(
_mm_cmplt_epi32(a.sse.i, b.sse.i),
_mm_cmpeq_epi32(a.sse.i, b.sse.i))
};
}
#endif
@@ -320,9 +293,8 @@ vec_ile (const union vec a, const union vec b)
#ifndef VEC_FN_ULE
#define VEC_FN_ULE
static inline union vec
vec_ule (const union vec a, const union vec b)
{
return vec_or(vec_ult(a, b), vec_ueq(a, b));
vec_ule(const union vec a, const union vec b) {
return vec_or(vec_ult(a, b), vec_ueq(a, b));
}
#endif
@@ -330,9 +302,8 @@ vec_ule (const union vec a, const union vec b)
#ifndef VEC_FN_GT
#define VEC_FN_GT
static inline union vec
vec_gt (const union vec a, const union vec b)
{
return (union vec) { .sse.f = _mm_cmpgt_ps(a.sse.f, b.sse.f) };
vec_gt(const union vec a, const union vec b) {
return (union vec) { .sse.f = _mm_cmpgt_ps(a.sse.f, b.sse.f) };
}
#endif
@@ -340,9 +311,8 @@ vec_gt (const union vec a, const union vec b)
#ifndef VEC_FN_IGT
#define VEC_FN_IGT
static inline union vec
vec_igt (const union vec a, const union vec b)
{
return (union vec) { .sse.i = _mm_cmpgt_epi32(a.sse.i, b.sse.i) };
vec_igt(const union vec a, const union vec b) {
return (union vec) { .sse.i = _mm_cmpgt_epi32(a.sse.i, b.sse.i) };
}
#endif
@@ -350,13 +320,14 @@ vec_igt (const union vec a, const union vec b)
#ifndef VEC_FN_UGT
#define VEC_FN_UGT
static inline union vec
vec_ugt (const union vec a, const union vec b)
{
const __m128i bias = _mm_set1_epi32(INT32_C(0x80000000));
vec_ugt(const union vec a, const union vec b) {
const __m128i bias = _mm_set1_epi32(INT32_C(0x80000000));
return (union vec) { .sse.i = _mm_cmpgt_epi32(
_mm_sub_epi32(a.sse.i, bias),
_mm_sub_epi32(b.sse.i, bias)) };
return (union vec) {
.sse.i = _mm_cmpgt_epi32(
_mm_sub_epi32(a.sse.i, bias),
_mm_sub_epi32(b.sse.i, bias))
};
}
#endif
@@ -364,9 +335,8 @@ vec_ugt (const union vec a, const union vec b)
#ifndef VEC_FN_GE
#define VEC_FN_GE
static inline union vec
vec_ge (const union vec a, const union vec b)
{
return (union vec) { .sse.f = _mm_cmpge_ps(a.sse.f, b.sse.f) };
vec_ge(const union vec a, const union vec b) {
return (union vec) { .sse.f = _mm_cmpge_ps(a.sse.f, b.sse.f) };
}
#endif
@@ -374,11 +344,12 @@ vec_ge (const union vec a, const union vec b)
#ifndef VEC_FN_IGE
#define VEC_FN_IGE
static inline union vec
vec_ige (const union vec a, const union vec b)
{
return (union vec) { .sse.i = _mm_or_si128(
_mm_cmpgt_epi32(a.sse.i, b.sse.i),
_mm_cmpeq_epi32(a.sse.i, b.sse.i)) };
vec_ige(const union vec a, const union vec b) {
return (union vec) {
.sse.i = _mm_or_si128(
_mm_cmpgt_epi32(a.sse.i, b.sse.i),
_mm_cmpeq_epi32(a.sse.i, b.sse.i))
};
}
#endif
@@ -386,9 +357,8 @@ vec_ige (const union vec a, const union vec b)
#ifndef VEC_FN_UGE
#define VEC_FN_UGE
static inline union vec
vec_uge (const union vec a, const union vec b)
{
return vec_or(vec_ugt(a, b), vec_ueq(a, b));
vec_uge(const union vec a, const union vec b) {
return vec_or(vec_ugt(a, b), vec_ueq(a, b));
}
#endif
@@ -396,15 +366,14 @@ vec_uge (const union vec a, const union vec b)
#ifndef VEC_FN_DOT
#define VEC_FN_DOT
static inline float
vec_dot (const union vec a, const union vec b)
{
const __m128 mulx = _mm_mul_ps(a.sse.f, b.sse.f);
const __m128 muly = _mm_shuffle_ps(mulx, mulx, _MM_SHUFFLE(0, 0, 0, 1));
const __m128 mulz = _mm_shuffle_ps(mulx, mulx, _MM_SHUFFLE(0, 0, 0, 2));
const __m128 sum = _mm_add_ps(mulx, _mm_add_ps(muly, mulz));
vec_dot(const union vec a, const union vec b) {
const __m128 mulx = _mm_mul_ps(a.sse.f, b.sse.f);
const __m128 muly = _mm_shuffle_ps(mulx, mulx, _MM_SHUFFLE(0, 0, 0, 1));
const __m128 mulz = _mm_shuffle_ps(mulx, mulx, _MM_SHUFFLE(0, 0, 0, 2));
const __m128 sum = _mm_add_ps(mulx, _mm_add_ps(muly, mulz));
// Extract first element:
return ((union vec) { .xi = _mm_cvtsi128_si32(_mm_castps_si128(sum)) }).x;
// Extract first element:
return ((union vec) { .xi = _mm_cvtsi128_si32(_mm_castps_si128(sum)) }).x;
}
#endif
@@ -412,22 +381,21 @@ vec_dot (const union vec a, const union vec b)
#ifndef VEC_FN_CROSS
#define VEC_FN_CROSS
static inline union vec
vec_cross (const union vec a, const union vec b)
{
// Shuffle a:
const __m128 ayzx = _mm_shuffle_ps(a.sse.f, a.sse.f, _MM_SHUFFLE(0, 0, 2, 1));
const __m128 azxy = _mm_shuffle_ps(a.sse.f, a.sse.f, _MM_SHUFFLE(0, 1, 0, 2));
vec_cross(const union vec a, const union vec b) {
// Shuffle a:
const __m128 ayzx = _mm_shuffle_ps(a.sse.f, a.sse.f, _MM_SHUFFLE(0, 0, 2, 1));
const __m128 azxy = _mm_shuffle_ps(a.sse.f, a.sse.f, _MM_SHUFFLE(0, 1, 0, 2));
// Shuffle b:
const __m128 byzx = _mm_shuffle_ps(b.sse.f, b.sse.f, _MM_SHUFFLE(0, 0, 2, 1));
const __m128 bzxy = _mm_shuffle_ps(b.sse.f, b.sse.f, _MM_SHUFFLE(0, 1, 0, 2));
// Shuffle b:
const __m128 byzx = _mm_shuffle_ps(b.sse.f, b.sse.f, _MM_SHUFFLE(0, 0, 2, 1));
const __m128 bzxy = _mm_shuffle_ps(b.sse.f, b.sse.f, _MM_SHUFFLE(0, 1, 0, 2));
// (ay * bz), (az * bx), (ax * by), (ax * bx):
const __m128 mul1 = _mm_mul_ps(ayzx, bzxy);
// (ay * bz), (az * bx), (ax * by), (ax * bx):
const __m128 mul1 = _mm_mul_ps(ayzx, bzxy);
// (az * by), (ax * bz), (ay * bx), (ax * bx):
const __m128 mul2 = _mm_mul_ps(azxy, byzx);
// (az * by), (ax * bz), (ay * bx), (ax * bx):
const __m128 mul2 = _mm_mul_ps(azxy, byzx);
return (union vec) { .sse.f = _mm_sub_ps(mul1, mul2) };
return (union vec) { .sse.f = _mm_sub_ps(mul1, mul2) };
}
#endif
+4 -5
View File
@@ -2,11 +2,10 @@
#ifndef VEC_FN_DOT
#define VEC_FN_DOT
static inline float
vec_dot (const union vec a, const union vec b)
{
const __m128 dot = _mm_dp_ps(a.sse.f, b.sse.f, 0x7F);
vec_dot(const union vec a, const union vec b) {
const __m128 dot = _mm_dp_ps(a.sse.f, b.sse.f, 0x7F);
// Extract first element:
return ((union vec) { .xi = _mm_cvtsi128_si32(_mm_castps_si128(dot)) }).x;
// Extract first element:
return ((union vec) { .xi = _mm_cvtsi128_si32(_mm_castps_si128(dot)) }).x;
}
#endif
+75 -75
View File
@@ -4,106 +4,106 @@
// GCC-style vector extensions work with Clang >= 3 or GCC >= 4.0:
#if !defined(VEC_SUPPRESS_GCC)
#if __GNUC__ >= 4 || __clang_major__ >= 3
#define VEC_GCC
#endif
#if __GNUC__ >= 4 || __clang_major__ >= 3
#define VEC_GCC
#endif
#endif
// For testing purposes, allow the user to bypass any platform-specific support
// and to use only the generic functions:
#if !defined(VEC_SUPPRESS_HW)
// Check for SSE support:
#if defined(__SSE2__)
#define VEC_SSE2
#include <emmintrin.h>
#endif
// Check for SSE support:
#if defined(__SSE2__)
#define VEC_SSE2
#include <emmintrin.h>
#endif
// Check for SSE4 support:
#if defined(__SSE4_1__)
#define VEC_SSE41
#include <nmmintrin.h>
#endif
// Check for SSE4 support:
#if defined(__SSE4_1__)
#define VEC_SSE41
#include <nmmintrin.h>
#endif
// Check for ARM NEON support:
#if defined(__ARM_NEON__) || defined(__ARM_NEON)
#define VEC_NEON
#include <arm_neon.h>
// Check for ARM NEON support:
#if defined(__ARM_NEON__) || defined(__ARM_NEON)
#define VEC_NEON
#include <arm_neon.h>
// GCC and Clang support inline assembly:
#if defined(__GNUC__) || defined(__clang_major__)
#define VEC_NEON_ASM
#endif
#endif
// GCC and Clang support inline assembly:
#if defined(__GNUC__) || defined(__clang_major__)
#define VEC_NEON_ASM
#endif
#endif
#endif
// Define architecture-independent vector types.
// This syntax is valid for GCC and LLVM/Clang:
#if defined(VEC_GCC)
typedef float vec4f __attribute__ ((vector_size (16), aligned (16)));
typedef int32_t vec4i __attribute__ ((vector_size (16), aligned (16)));
typedef uint32_t vec4u __attribute__ ((vector_size (16), aligned (16)));
typedef float vec4f __attribute__((vector_size(16), aligned(16)));
typedef int32_t vec4i __attribute__((vector_size(16), aligned(16)));
typedef uint32_t vec4u __attribute__((vector_size(16), aligned(16)));
#endif
// Union to represent a 4x32-bit (128-bit) vector in various ways:
union vec {
// Named member representation in float form, assuming a 3D vector:
struct {
float x;
float y;
float z;
float w;
};
// Named member representation in float form, assuming a 3D vector:
struct {
float x;
float y;
float z;
float w;
};
// Named member representation in signed integer form:
struct {
int32_t xi;
int32_t yi;
int32_t zi;
int32_t wi;
};
// Named member representation in signed integer form:
struct {
int32_t xi;
int32_t yi;
int32_t zi;
int32_t wi;
};
// Named member representation in unsigned integer form:
struct {
uint32_t xu;
uint32_t yu;
uint32_t zu;
uint32_t wu;
};
// Named member representation in unsigned integer form:
struct {
uint32_t xu;
uint32_t yu;
uint32_t zu;
uint32_t wu;
};
// Representation as array of separate elements:
union {
float f[4];
int32_t i[4];
uint32_t u[4];
} elem;
// Representation as array of separate elements:
union {
float f[4];
int32_t i[4];
uint32_t u[4];
} elem;
// Representation as GCC-style vector extensions:
// Representation as GCC-style vector extensions:
#if defined(VEC_GCC)
union {
vec4f f;
vec4i i;
vec4u u;
} gcc;
union {
vec4f f;
vec4i i;
vec4u u;
} gcc;
#endif
// SSE representation:
// SSE representation:
#if defined(VEC_SSE2)
union {
__m128 f;
__m128i i;
__m128i u;
} sse;
union {
__m128 f;
__m128i i;
__m128i u;
} sse;
#endif
// NEON representation:
// NEON representation:
#if defined(VEC_NEON)
union {
float32x4_t f;
int32x4_t i;
uint32x4_t u;
} neon;
union {
float32x4_t f;
int32x4_t i;
uint32x4_t u;
} neon;
#endif
};
@@ -124,23 +124,23 @@ union vec {
// are available.
#if defined(VEC_SSE41)
#include "private/sse41.h"
#include "private/sse41.h"
#endif
#if defined(VEC_SSE2)
#include "private/sse2.h"
#include "private/sse2.h"
#endif
#if defined(VEC_NEON_ASM)
#include "private/neon_asm.h"
#include "private/neon_asm.h"
#endif
#if defined(VEC_NEON)
#include "private/neon_intrin.h"
#include "private/neon_intrin.h"
#endif
#if defined(VEC_GCC)
#include "private/gcc.h"
#include "private/gcc.h"
#endif
#include "private/generic.h"