Struct microbench::CmdCudaCopy

pub(crate) struct CmdCudaCopy {
    pub(crate) copy_direction: Vec<CopyMethod>,
    pub(crate) size: Vec<usize>,
    pub(crate) device_id: u16,
    pub(crate) mem_type: ArgMemType,
    pub(crate) mem_location: u16,
    pub(crate) page_type: ArgPageType,
    pub(crate) repeat: u32,
}

Fields

copy_direction: Vec<CopyMethod>

Direction in which to execute the copy

size: Vec<usize>

Size of buffer to copy (MB)

device_id: u16

Execute on GPU (See CUDA device list)

mem_type: ArgMemType

Memory type with which to allocate data

mem_location: u16

Allocate memory on NUMA node (See numactl -H)

page_type: ArgPageType

Page type with with to allocate memory

repeat: u32

Number of times to repeat benchmark

Trait Implementations

impl StructOpt for CmdCudaCopy

fn clap<'a, 'b>() -> App<'a, 'b>

Returns clap::App corresponding to the struct.

fn from_clap(matches: &ArgMatches<'_>) -> Self

Builds the struct from clap::ArgMatches. It’s guaranteed to succeed if matches originates from an App generated by [StructOpt::clap] called on the same type, otherwise it must panic.

fn from_args() -> Self

Builds the struct from the command line arguments (std::env::args_os). Calls clap::Error::exit on failure, printing the error message and aborting the program.

fn from_args_safe() -> Result<Self, Error>

Builds the struct from the command line arguments (std::env::args_os). Unlike [StructOpt::from_args], returns clap::Error on failure instead of aborting the program, so calling .exit is up to you.

fn from_iter<I>(iter: I) -> Self where
    I: IntoIterator,
    <I as IntoIterator>::Item: Into<OsString>,
    <I as IntoIterator>::Item: Clone, 

Gets the struct from any iterator such as a Vec of your making. Print the error message and quit the program in case of failure.

fn from_iter_safe<I>(iter: I) -> Result<Self, Error> where
    I: IntoIterator,
    <I as IntoIterator>::Item: Into<OsString>,
    <I as IntoIterator>::Item: Clone, 

Gets the struct from any iterator such as a Vec of your making.