add double compute kernel

2025-12-26 17:16:12 +08:00
parent e5539bcc6d
commit c6d53d97ea
16 changed files with 821 additions and 300 deletions
@@ -2,8 +2,8 @@

 // Hybrid implementation: SIMD for Abs, Scalar for Max
 // Assumes Column-Major layout for both MAT_KERNEL content and Kernel sequence
-static void pivot_float(MAT_KERNEL(float) const *source, size_t col_begin, size_t col_length,
-                        size_t *restrict ans)
+INTERNAL void SN_DECL pivot_float(MAT_KERNEL(float) const *source, size_t col_begin,
+                                  size_t col_length, size_t *restrict ans)
 {
        float max_v[4] = { -1.0f, -1.0f, -1.0f, -1.0f };