Merge branch 'master' into vec2_mat2

include/cglm/simd/arm.h

@@ -13,6 +13,12 @@
 #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) {
@@ -25,6 +31,22 @@ glmm_hadd(float32x4_t v) {
 #endif
 }
 
+static inline
+float
+glmm_hmin(float32x4_t v) {
+  v = vpmin_f32(vget_low_f32(v), vget_high_f32(v));
+  v = vpmin_f32(v, v);
+  return vget_lane_f32(v, 0);
+}
+
+static inline
+float
+glmm_hmax(float32x4_t v) {
+  v = vpmax_f32(vget_low_f32(v), vget_high_f32(v));
+  v = vpmax_f32(v, v);
+  return vget_lane_f32(v, 0);
+}
+
 static inline
 float
 glmm_dot(float32x4_t a, float32x4_t b) {
@@ -43,5 +65,17 @@ 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 */

include/cglm/simd/x86.h

@@ -42,6 +42,12 @@
 #  endif
 #endif
 
+static inline
+__m128
+glmm_abs(__m128 x) {
+  return _mm_andnot_ps(_mm_set1_ps(-0.0f), x);
+}
+
 static inline
 __m128
 glmm_vhadds(__m128 v) {
@@ -68,6 +74,38 @@ glmm_hadd(__m128 v) {
   return _mm_cvtss_f32(glmm_vhadds(v));
 }
 
+static inline
+__m128
+glmm_vhmin(__m128 v) {
+  __m128 x0, x1, x2;
+  x0 = _mm_movehl_ps(v, v);     /* [2, 3, 2, 3] */
+  x1 = _mm_min_ps(x0, v);       /* [0|2, 1|3, 2|2, 3|3] */
+  x2 = glmm_shuff1x(x1, 1);     /* [1|3, 1|3, 1|3, 1|3] */
+  return _mm_min_ss(x1, x2);
+}
+
+static inline
+float
+glmm_hmin(__m128 v) {
+  return _mm_cvtss_f32(glmm_vhmin(v));
+}
+
+static inline
+__m128
+glmm_vhmax(__m128 v) {
+  __m128 x0, x1, x2;
+  x0 = _mm_movehl_ps(v, v);     /* [2, 3, 2, 3] */
+  x1 = _mm_max_ps(x0, v);       /* [0|2, 1|3, 2|2, 3|3] */
+  x2 = glmm_shuff1x(x1, 1);     /* [1|3, 1|3, 1|3, 1|3] */
+  return _mm_max_ss(x1, x2);
+}
+
+static inline
+float
+glmm_hmax(__m128 v) {
+  return _mm_cvtss_f32(glmm_vhmax(v));
+}
+
 static inline
 __m128
 glmm_vdots(__m128 a, __m128 b) {
@@ -119,6 +157,18 @@ glmm_norm2(__m128 a) {
   return _mm_cvtss_f32(glmm_vhadds(_mm_mul_ps(a, a)));
 }
 
+static inline
+float
+glmm_norm_one(__m128 a) {
+  return _mm_cvtss_f32(glmm_vhadds(glmm_abs(a)));
+}
+
+static inline
+float
+glmm_norm_inf(__m128 a) {
+  return _mm_cvtss_f32(glmm_vhmax(glmm_abs(a)));
+}
+
 static inline
 __m128
 glmm_load3(float v[3]) {