mtgp32_params_fast_t Struct Reference#
#include <rocrand_mtgp32.h>
Public Attributes | |
int | mexp |
int | pos |
int | sh1 |
int | sh2 |
uint32_t | tbl [16] |
uint32_t | tmp_tbl [16] |
uint32_t | flt_tmp_tbl [16] |
uint32_t | mask |
unsigned char | poly_sha1 [21] |
Detailed Description
MTGP32 parameters. Some element is redundant to keep structure simple.
pos is a pick up position which is selected to have good performance on graphic processors. 3 < pos < Q, where Q is a maximum number such that the size of status array - Q is a power of 2. For example, when mexp is 44497, size of 32-bit status array is 696, and Q is 184, then pos is between 4 and 183. This means 512 parallel calculations is allowed when mexp is 44497.
poly_sha1 is SHA1 digest of the characteristic polynomial of state transition function. SHA1 is calculated based on printing form of the polynomial. This is important when we use parameters generated by the dynamic creator which
mask This is a mask to make the dimension of state space have just Mersenne Prime. This is redundant.
Member Data Documentation
◆ flt_tmp_tbl
uint32_t mtgp32_params_fast_t::flt_tmp_tbl[16] |
A small matrix for tempering and converting to float.
◆ mask
uint32_t mtgp32_params_fast_t::mask |
This is a mask for state space.
◆ mexp
int mtgp32_params_fast_t::mexp |
Mersenne exponent. This is redundant.
◆ poly_sha1
unsigned char mtgp32_params_fast_t::poly_sha1[21] |
SHA1 digest.
◆ pos
int mtgp32_params_fast_t::pos |
pick up position.
◆ sh1
int mtgp32_params_fast_t::sh1 |
Shift value 1. 0 < sh1 < 32.
◆ sh2
int mtgp32_params_fast_t::sh2 |
Shift value 2. 0 < sh2 < 32.
◆ tbl
uint32_t mtgp32_params_fast_t::tbl[16] |
A small matrix.
◆ tmp_tbl
uint32_t mtgp32_params_fast_t::tmp_tbl[16] |
A small matrix for tempering.
The documentation for this struct was generated from the following file:
- /home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-rocrand/checkouts/docs-6.3.1/library/include/rocrand/rocrand_mtgp32.h