cglm 0.1.0 → 0.1.3

data/ext/cglm/cglm-0.6.2/include/cglm/simd/arm.h

@@ -0,0 +1,83 @@
+/*
+ * Copyright (c), Recep Aslantas.
+ *
+ * MIT License (MIT), http://opensource.org/licenses/MIT
+ * Full license can be found in the LICENSE file
+ */
+
+#ifndef cglm_simd_arm_h
+#define cglm_simd_arm_h
+#include "intrin.h"
+#ifdef CGLM_SIMD_ARM
+
+#define glmm_load(p)      vld1q_f32(p)
+#define glmm_store(p, a)  vst1q_f32(p, a)
+
+static inline
+float32x4_t
+glmm_abs(float32x4_t v) {
+  return vabsq_f32(v);
+}
+
+static inline
+float
+glmm_hadd(float32x4_t v) {
+#if defined(__aarch64__)
+  return vaddvq_f32(v);
+#else
+  v = vaddq_f32(v, vrev64q_f32(v));
+  v = vaddq_f32(v, vcombine_f32(vget_high_f32(v), vget_low_f32(v)));
+  return vgetq_lane_f32(v, 0);
+#endif
+}
+
+static inline
+float
+glmm_hmin(float32x4_t v) {
+  float32x2_t t;
+  t = vpmin_f32(vget_low_f32(v), vget_high_f32(v));
+  t = vpmin_f32(t, t);
+  return vget_lane_f32(t, 0);
+}
+
+static inline
+float
+glmm_hmax(float32x4_t v) {
+  float32x2_t t;
+  t = vpmax_f32(vget_low_f32(v), vget_high_f32(v));
+  t = vpmax_f32(t, t);
+  return vget_lane_f32(t, 0);
+}
+
+static inline
+float
+glmm_dot(float32x4_t a, float32x4_t b) {
+  return glmm_hadd(vmulq_f32(a, b));
+}
+
+static inline
+float
+glmm_norm(float32x4_t a) {
+  return sqrtf(glmm_dot(a, a));
+}
+
+static inline
+float
+glmm_norm2(float32x4_t a) {
+  return glmm_dot(a, a);
+}
+
+static inline
+float
+glmm_norm_one(float32x4_t a) {
+  return glmm_hadd(glmm_abs(a));
+}
+
+static inline
+float
+glmm_norm_inf(float32x4_t a) {
+  return glmm_hmax(glmm_abs(a));
+}
+
+#endif
+#endif /* cglm_simd_arm_h */

data/ext/cglm/cglm-0.6.2/include/cglm/simd/avx/affine.h

@@ -0,0 +1,66 @@
+/*
+ * Copyright (c), Recep Aslantas.
+ *
+ * MIT License (MIT), http://opensource.org/licenses/MIT
+ * Full license can be found in the LICENSE file
+ */
+
+#ifndef cglm_affine_mat_avx_h
+#define cglm_affine_mat_avx_h
+#ifdef __AVX__
+
+#include "../../common.h"
+#include "../intrin.h"
+
+#include <immintrin.h>
+
+CGLM_INLINE
+void
+glm_mul_avx(mat4 m1, mat4 m2, mat4 dest) {
+  /* D = R * L (Column-Major) */
+
+  __m256 y0, y1, y2, y3, y4, y5, y6, y7, y8, y9;
+
+  y0 = glmm_load256(m2[0]); /* h g f e d c b a */
+  y1 = glmm_load256(m2[2]); /* p o n m l k j i */
+
+  y2 = glmm_load256(m1[0]); /* h g f e d c b a */
+  y3 = glmm_load256(m1[2]); /* p o n m l k j i */
+
+  /* 0x03: 0b00000011 */
+  y4 = _mm256_permute2f128_ps(y2, y2, 0x03); /* d c b a h g f e */
+  y5 = _mm256_permute2f128_ps(y3, y3, 0x03); /* l k j i p o n m */
+
+  /* f f f f a a a a */
+  /* h h h h c c c c */
+  /* e e e e b b b b */
+  /* g g g g d d d d */
+  y6 = _mm256_permutevar_ps(y0, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0));
+  y7 = _mm256_permutevar_ps(y0, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2));
+  y8 = _mm256_permutevar_ps(y0, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1));
+  y9 = _mm256_permutevar_ps(y0, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3));
+
+  glmm_store256(dest[0],
+                _mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),
+                                            _mm256_mul_ps(y3, y7)),
+                              _mm256_add_ps(_mm256_mul_ps(y4, y8),
+                                            _mm256_mul_ps(y5, y9))));
+
+  /* n n n n i i i i */
+  /* p p p p k k k k */
+  /* m m m m j j j j */
+  /* o o o o l l l l */
+  y6 = _mm256_permutevar_ps(y1, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0));
+  y7 = _mm256_permutevar_ps(y1, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2));
+  y8 = _mm256_permutevar_ps(y1, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1));
+  y9 = _mm256_permutevar_ps(y1, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3));
+
+  glmm_store256(dest[2],
+                _mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),
+                                            _mm256_mul_ps(y3, y7)),
+                              _mm256_add_ps(_mm256_mul_ps(y4, y8),
+                                            _mm256_mul_ps(y5, y9))));
+}
+
+#endif
+#endif /* cglm_affine_mat_avx_h */

data/ext/cglm/cglm-0.6.2/include/cglm/simd/avx/mat4.h

@@ -0,0 +1,66 @@
+/*
+ * Copyright (c), Recep Aslantas.
+ *
+ * MIT License (MIT), http://opensource.org/licenses/MIT
+ * Full license can be found in the LICENSE file
+ */
+
+#ifndef cglm_mat_simd_avx_h
+#define cglm_mat_simd_avx_h
+#ifdef __AVX__
+
+#include "../../common.h"
+#include "../intrin.h"
+
+#include <immintrin.h>
+
+CGLM_INLINE
+void
+glm_mat4_mul_avx(mat4 m1, mat4 m2, mat4 dest) {
+  /* D = R * L (Column-Major) */
+
+  __m256 y0, y1, y2, y3, y4, y5, y6, y7, y8, y9;
+
+  y0 = glmm_load256(m2[0]); /* h g f e d c b a */
+  y1 = glmm_load256(m2[2]); /* p o n m l k j i */
+
+  y2 = glmm_load256(m1[0]); /* h g f e d c b a */
+  y3 = glmm_load256(m1[2]); /* p o n m l k j i */
+
+  /* 0x03: 0b00000011 */
+  y4 = _mm256_permute2f128_ps(y2, y2, 0x03); /* d c b a h g f e */
+  y5 = _mm256_permute2f128_ps(y3, y3, 0x03); /* l k j i p o n m */
+
+  /* f f f f a a a a */
+  /* h h h h c c c c */
+  /* e e e e b b b b */
+  /* g g g g d d d d */
+  y6 = _mm256_permutevar_ps(y0, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0));
+  y7 = _mm256_permutevar_ps(y0, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2));
+  y8 = _mm256_permutevar_ps(y0, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1));
+  y9 = _mm256_permutevar_ps(y0, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3));
+
+  glmm_store256(dest[0],
+                _mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),
+                                            _mm256_mul_ps(y3, y7)),
+                              _mm256_add_ps(_mm256_mul_ps(y4, y8),
+                                            _mm256_mul_ps(y5, y9))));
+
+  /* n n n n i i i i */
+  /* p p p p k k k k */
+  /* m m m m j j j j */
+  /* o o o o l l l l */
+  y6 = _mm256_permutevar_ps(y1, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0));
+  y7 = _mm256_permutevar_ps(y1, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2));
+  y8 = _mm256_permutevar_ps(y1, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1));
+  y9 = _mm256_permutevar_ps(y1, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3));
+
+  glmm_store256(dest[2],
+                _mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),
+                                            _mm256_mul_ps(y3, y7)),
+                              _mm256_add_ps(_mm256_mul_ps(y4, y8),
+                                            _mm256_mul_ps(y5, y9))));
+}
+
+#endif
+#endif /* cglm_mat_simd_avx_h */

data/ext/cglm/cglm-0.6.2/include/cglm/simd/intrin.h

@@ -0,0 +1,90 @@
+/*
+ * Copyright (c), Recep Aslantas.
+ *
+ * MIT License (MIT), http://opensource.org/licenses/MIT
+ * Full license can be found in the LICENSE file
+ */
+
+#ifndef cglm_intrin_h
+#define cglm_intrin_h
+
+#if defined( _MSC_VER )
+#  if (defined(_M_AMD64) || defined(_M_X64)) || _M_IX86_FP == 2
+#    ifndef __SSE2__
+#      define __SSE2__
+#    endif
+#  elif _M_IX86_FP == 1
+#    ifndef __SSE__
+#      define __SSE__
+#    endif
+#  endif
+/* do not use alignment for older visual studio versions */
+#  if _MSC_VER < 1913     /* Visual Studio 2017 version 15.6 */
+#    define CGLM_ALL_UNALIGNED
+#  endif
+#endif
+
+#if defined( __SSE__ ) || defined( __SSE2__ )
+#  include <xmmintrin.h>
+#  include <emmintrin.h>
+#  define CGLM_SSE_FP 1
+#  ifndef CGLM_SIMD_x86
+#    define CGLM_SIMD_x86
+#  endif
+#endif
+
+#if defined(__SSE3__)
+#  include <x86intrin.h>
+#  ifndef CGLM_SIMD_x86
+#    define CGLM_SIMD_x86
+#  endif
+#endif
+
+#if defined(__SSE4_1__)
+#  include <smmintrin.h>
+#  ifndef CGLM_SIMD_x86
+#    define CGLM_SIMD_x86
+#  endif
+#endif
+
+#if defined(__SSE4_2__)
+#  include <nmmintrin.h>
+#  ifndef CGLM_SIMD_x86
+#    define CGLM_SIMD_x86
+#  endif
+#endif
+
+#ifdef __AVX__
+#  include <immintrin.h>
+#  define CGLM_AVX_FP 1
+#  ifndef CGLM_SIMD_x86
+#    define CGLM_SIMD_x86
+#  endif
+#endif
+
+/* ARM Neon */
+#if defined(__ARM_NEON)
+#  include <arm_neon.h>
+#  if defined(__ARM_NEON_FP)
+#    define CGLM_NEON_FP 1
+#    ifndef CGLM_SIMD_ARM
+#      define CGLM_SIMD_ARM
+#    endif
+#  endif
+#endif
+
+#if defined(CGLM_SIMD_x86) || defined(CGLM_NEON_FP)
+#  ifndef CGLM_SIMD
+#    define CGLM_SIMD
+#  endif
+#endif
+
+#if defined(CGLM_SIMD_x86)
+#  include "x86.h"
+#endif
+
+#if defined(CGLM_SIMD_ARM)
+#  include "arm.h"
+#endif
+
+#endif /* cglm_intrin_h */

data/ext/cglm/cglm-0.6.2/include/cglm/simd/neon/mat4.h

@@ -0,0 +1,57 @@
+/*
+ * Copyright (c), Recep Aslantas.
+ *
+ * MIT License (MIT), http://opensource.org/licenses/MIT
+ * Full license can be found in the LICENSE file
+ */
+
+#ifndef cglm_mat4_neon_h
+#define cglm_mat4_neon_h
+#if defined(__ARM_NEON_FP)
+
+#include "../../common.h"
+#include "../intrin.h"
+
+CGLM_INLINE
+void
+glm_mat4_mul_neon(mat4 m1, mat4 m2, mat4 dest) {
+  /* D = R * L (Column-Major) */
+  float32x4_t l0, l1, l2, l3, r, d0, d1, d2, d3;
+
+  l0 = vld1q_f32(m2[0]);
+  l1 = vld1q_f32(m2[1]);
+  l2 = vld1q_f32(m2[2]);
+  l3 = vld1q_f32(m2[3]);
+
+  r  = vld1q_f32(m1[0]);
+  d0 = vmulq_lane_f32(r, vget_low_f32(l0), 0);
+  d1 = vmulq_lane_f32(r, vget_low_f32(l1), 0);
+  d2 = vmulq_lane_f32(r, vget_low_f32(l2), 0);
+  d3 = vmulq_lane_f32(r, vget_low_f32(l3), 0);
+
+  r  = vld1q_f32(m1[1]);
+  d0 = vmlaq_lane_f32(d0, r, vget_low_f32(l0), 1);
+  d1 = vmlaq_lane_f32(d1, r, vget_low_f32(l1), 1);
+  d2 = vmlaq_lane_f32(d2, r, vget_low_f32(l2), 1);
+  d3 = vmlaq_lane_f32(d3, r, vget_low_f32(l3), 1);
+
+  r  = vld1q_f32(m1[2]);
+  d0 = vmlaq_lane_f32(d0, r, vget_high_f32(l0), 0);
+  d1 = vmlaq_lane_f32(d1, r, vget_high_f32(l1), 0);
+  d2 = vmlaq_lane_f32(d2, r, vget_high_f32(l2), 0);
+  d3 = vmlaq_lane_f32(d3, r, vget_high_f32(l3), 0);
+
+  r  = vld1q_f32(m1[3]);
+  d0 = vmlaq_lane_f32(d0, r, vget_high_f32(l0), 1);
+  d1 = vmlaq_lane_f32(d1, r, vget_high_f32(l1), 1);
+  d2 = vmlaq_lane_f32(d2, r, vget_high_f32(l2), 1);
+  d3 = vmlaq_lane_f32(d3, r, vget_high_f32(l3), 1);
+
+  vst1q_f32(dest[0], d0);
+  vst1q_f32(dest[1], d1);
+  vst1q_f32(dest[2], d2);
+  vst1q_f32(dest[3], d3);
+}
+
+#endif
+#endif /* cglm_mat4_neon_h */

data/ext/cglm/cglm-0.6.2/include/cglm/simd/sse2/affine.h

@@ -0,0 +1,111 @@
+/*
+ * Copyright (c), Recep Aslantas.
+ *
+ * MIT License (MIT), http://opensource.org/licenses/MIT
+ * Full license can be found in the LICENSE file
+ */
+
+#ifndef cglm_affine_mat_sse2_h
+#define cglm_affine_mat_sse2_h
+#if defined( __SSE__ ) || defined( __SSE2__ )
+
+#include "../../common.h"
+#include "../intrin.h"
+
+CGLM_INLINE
+void
+glm_mul_sse2(mat4 m1, mat4 m2, mat4 dest) {
+  /* D = R * L (Column-Major) */
+  __m128 l0, l1, l2, l3, r;
+
+  l0 = glmm_load(m1[0]);
+  l1 = glmm_load(m1[1]);
+  l2 = glmm_load(m1[2]);
+  l3 = glmm_load(m1[3]);
+
+  r = glmm_load(m2[0]);
+  glmm_store(dest[0],
+             _mm_add_ps(_mm_add_ps(_mm_mul_ps(glmm_shuff1x(r, 0), l0),
+                                   _mm_mul_ps(glmm_shuff1x(r, 1), l1)),
+                        _mm_mul_ps(glmm_shuff1x(r, 2), l2)));
+
+  r = glmm_load(m2[1]);
+  glmm_store(dest[1],
+             _mm_add_ps(_mm_add_ps(_mm_mul_ps(glmm_shuff1x(r, 0), l0),
+                                   _mm_mul_ps(glmm_shuff1x(r, 1), l1)),
+                        _mm_mul_ps(glmm_shuff1x(r, 2), l2)));
+
+  r = glmm_load(m2[2]);
+  glmm_store(dest[2],
+             _mm_add_ps(_mm_add_ps(_mm_mul_ps(glmm_shuff1x(r, 0), l0),
+                                   _mm_mul_ps(glmm_shuff1x(r, 1), l1)),
+                        _mm_mul_ps(glmm_shuff1x(r, 2), l2)));
+
+  r = glmm_load(m2[3]);
+  glmm_store(dest[3],
+             _mm_add_ps(_mm_add_ps(_mm_mul_ps(glmm_shuff1x(r, 0), l0),
+                                   _mm_mul_ps(glmm_shuff1x(r, 1), l1)),
+                        _mm_add_ps(_mm_mul_ps(glmm_shuff1x(r, 2), l2),
+                                   _mm_mul_ps(glmm_shuff1x(r, 3), l3))));
+}
+
+CGLM_INLINE
+void
+glm_mul_rot_sse2(mat4 m1, mat4 m2, mat4 dest) {
+  /* D = R * L (Column-Major) */
+  __m128 l0, l1, l2, l3, r;
+
+  l0 = glmm_load(m1[0]);
+  l1 = glmm_load(m1[1]);
+  l2 = glmm_load(m1[2]);
+  l3 = glmm_load(m1[3]);
+
+  r = glmm_load(m2[0]);
+  glmm_store(dest[0],
+             _mm_add_ps(_mm_add_ps(_mm_mul_ps(glmm_shuff1x(r, 0), l0),
+                                   _mm_mul_ps(glmm_shuff1x(r, 1), l1)),
+                        _mm_mul_ps(glmm_shuff1x(r, 2), l2)));
+
+  r = glmm_load(m2[1]);
+  glmm_store(dest[1],
+             _mm_add_ps(_mm_add_ps(_mm_mul_ps(glmm_shuff1x(r, 0), l0),
+                                   _mm_mul_ps(glmm_shuff1x(r, 1), l1)),
+                        _mm_mul_ps(glmm_shuff1x(r, 2), l2)));
+
+  r = glmm_load(m2[2]);
+  glmm_store(dest[2],
+             _mm_add_ps(_mm_add_ps(_mm_mul_ps(glmm_shuff1x(r, 0), l0),
+                                   _mm_mul_ps(glmm_shuff1x(r, 1), l1)),
+                        _mm_mul_ps(glmm_shuff1x(r, 2), l2)));
+
+  glmm_store(dest[3], l3);
+}
+
+CGLM_INLINE
+void
+glm_inv_tr_sse2(mat4 mat) {
+  __m128 r0, r1, r2, r3, x0, x1;
+
+  r0 = glmm_load(mat[0]);
+  r1 = glmm_load(mat[1]);
+  r2 = glmm_load(mat[2]);
+  r3 = glmm_load(mat[3]);
+  x1 = _mm_set_ps(1.0f, 0.0f, 0.0f, 0.0f);
+
+  _MM_TRANSPOSE4_PS(r0, r1, r2, x1);
+
+  x0 = _mm_add_ps(_mm_mul_ps(r0, glmm_shuff1(r3, 0, 0, 0, 0)),
+                  _mm_mul_ps(r1, glmm_shuff1(r3, 1, 1, 1, 1)));
+  x0 = _mm_add_ps(x0, _mm_mul_ps(r2, glmm_shuff1(r3, 2, 2, 2, 2)));
+  x0 = _mm_xor_ps(x0, _mm_set1_ps(-0.f));
+
+  x0 = _mm_add_ps(x0, x1);
+
+  glmm_store(mat[0], r0);
+  glmm_store(mat[1], r1);
+  glmm_store(mat[2], r2);
+  glmm_store(mat[3], x0);
+}
+
+#endif
+#endif /* cglm_affine_mat_sse2_h */

data/ext/cglm/cglm-0.6.2/include/cglm/simd/sse2/mat3.h

@@ -0,0 +1,59 @@
+/*
+ * Copyright (c), Recep Aslantas.
+ *
+ * MIT License (MIT), http://opensource.org/licenses/MIT
+ * Full license can be found in the LICENSE file
+ */
+
+#ifndef cglm_mat3_sse_h
+#define cglm_mat3_sse_h
+#if defined( __SSE__ ) || defined( __SSE2__ )
+
+#include "../../common.h"
+#include "../intrin.h"
+
+CGLM_INLINE
+void
+glm_mat3_mul_sse2(mat3 m1, mat3 m2, mat3 dest) {
+  __m128 l0, l1, l2;
+  __m128 r0, r1, r2;
+  __m128 x0, x1, x2;
+
+  l0 = _mm_loadu_ps(m1[0]);
+  l1 = _mm_loadu_ps(&m1[1][1]);
+  l2 = _mm_set1_ps(m1[2][2]);
+
+  r0 = _mm_loadu_ps(m2[0]);
+  r1 = _mm_loadu_ps(&m2[1][1]);
+  r2 = _mm_set1_ps(m2[2][2]);
+
+  x1 = glmm_shuff2(l0, l1, 1, 0, 3, 3, 0, 3, 2, 0);
+  x2 = glmm_shuff2(l1, l2, 0, 0, 3, 2, 0, 2, 1, 0);
+
+  x0 = _mm_add_ps(_mm_mul_ps(glmm_shuff1(l0, 0, 2, 1, 0),
+                             glmm_shuff1(r0, 3, 0, 0, 0)),
+                  _mm_mul_ps(x1, glmm_shuff2(r0, r1, 0, 0, 1, 1, 2, 0, 0, 0)));
+
+  x0 = _mm_add_ps(x0,
+                  _mm_mul_ps(x2, glmm_shuff2(r0, r1, 1, 1, 2, 2, 2, 0, 0, 0)));
+
+  _mm_storeu_ps(dest[0], x0);
+
+  x0 = _mm_add_ps(_mm_mul_ps(glmm_shuff1(l0, 1, 0, 2, 1),
+                             _mm_shuffle_ps(r0, r1, _MM_SHUFFLE(2, 2, 3, 3))),
+                  _mm_mul_ps(glmm_shuff1(x1, 1, 0, 2, 1),
+                             glmm_shuff1(r1, 3, 3, 0, 0)));
+
+  x0 = _mm_add_ps(x0,
+                  _mm_mul_ps(glmm_shuff1(x2, 1, 0, 2, 1),
+                             _mm_shuffle_ps(r1, r2, _MM_SHUFFLE(0, 0, 1, 1))));
+
+  _mm_storeu_ps(&dest[1][1], x0);
+
+  dest[2][2] = m1[0][2] * m2[2][0]
+             + m1[1][2] * m2[2][1]
+             + m1[2][2] * m2[2][2];
+}
+
+#endif
+#endif /* cglm_mat3_sse_h */