#include <kernel_dot.h>
|
| template<typename Expr1 , typename Expr2 > |
| static FORCE_INLINE T | dot (const Expr1 &expr1, my_size_t base1, my_size_t stride1, const Expr2 &expr2, my_size_t base2, my_size_t stride2, my_size_t len) noexcept |
| | Dispatch dot product based on stride values.
|
| |
| template<typename Expr1 , typename Expr2 > |
| static FORCE_INLINE T | naive_dot_physical (const Expr1 &expr1, my_size_t base1, my_size_t stride1, const Expr2 &expr2, my_size_t base2, my_size_t stride2, my_size_t len) noexcept |
| | Naive scalar dot product for testing/validation.
|
| |
◆ Helpers
template<typename T ,
my_size_t Bits, typename Arch >
template<typename T ,
my_size_t Bits, typename Arch >
◆ dot()
template<typename T ,
my_size_t Bits, typename Arch >
template<typename Expr1 , typename Expr2 >
Dispatch dot product based on stride values.
- Parameters
-
| base1 | Physical offset of first fiber's start |
| stride1 | Physical stride along contraction axis (1 = contiguous) |
| base2 | Physical offset of second fiber's start |
| stride2 | Physical stride along contraction axis (1 = contiguous) |
| len | Number of elements along contraction axis (logical dim) |
◆ naive_dot_physical()
template<typename T ,
my_size_t Bits, typename Arch >
template<typename Expr1 , typename Expr2 >
Naive scalar dot product for testing/validation.
Accesses physical memory directly via data_.data(). Only used in tests to verify SIMD dot results.
◆ simdWidth
template<typename T ,
my_size_t Bits, typename Arch >
The documentation for this struct was generated from the following file: