hardware::use_unrolled_loop_poly_mul Struct Reference

Implementation of polynomial_mul_rev<w>(rev_u) function which uses a simple loop over i: if(w>>i&1)result^=rev_u<<(32-i), which is equivalent to w * flip_at_32(rev_u), which in turn is equivalent to rev(rev(w) * rev_u),. More...

Inheritance diagram for hardware::use_unrolled_loop_poly_mul:

Classes
struct	Polynomial_mul_rev_step_executor

Static Public Member Functions
template<uint32_t w>
static uint64_t	polynomial_mul_rev (uint32_t rev_u)
Static Public Member Functions inherited from hardware::crc32_impl
static uint32_t	update (uint32_t crc, unsigned char data)
static uint32_t	update (uint32_t crc, uint16_t data)
static uint32_t	update (uint32_t crc, uint32_t data)
static uint64_t	update (uint64_t crc, uint64_t data)

Detailed Description

Implementation of polynomial_mul_rev<w>(rev_u) function which uses a simple loop over i: if(w>>i&1)result^=rev_u<<(32-i), which is equivalent to w * flip_at_32(rev_u), which in turn is equivalent to rev(rev(w) * rev_u),.

See also

flip_at_32 for explanation why this holds,

use_pclmul for explanation of what polynomial_mul_rev is computing. This implementation is to be used when hardware accelerated polynomial multiplication is not available. It tries to unroll the simple loop, so just the few xors and shifts for non-zero bits of w are emitted.

Member Function Documentation

polynomial_mul_rev()

template<uint32_t w>

static uint64_t hardware::use_unrolled_loop_poly_mul::polynomial_mul_rev ( uint32_t  rev_u )

inlinestatic

---

The documentation for this struct was generated from the following file:

• storage/innobase/ut/crc32.cc