rocPRIM
Namespaces | Classes | Macros | Typedefs
Utilsmodule
Collaboration diagram for Utilsmodule:

Namespaces

 detail
 Deprecated: Configuration of device-level scan primitives.
 

Classes

class  double_buffer< T >
 This class provides an convenient way to do double buffering. More...
 
class  future_value< T, Iter >
 Allows passing values that are not yet known at launch time as paramters to device algorithms. More...
 
struct  key_value_pair< Key_, Value_ >
 Convenience struct that allows you to store key-value pairs as a composite entity. More...
 
struct  empty_type
 Empty type used as a placeholder, usually used to flag that given template parameter should not be used. More...
 
struct  empty_binary_op
 Binary operator that takes two instances of empty_type, usually used as nop replacement for the HIP-CPU back-end. More...
 

Macros

#define __AMDGCN_WAVEFRONT_SIZE   64
 

Typedefs

using half = ::__half
 Half-precision floating point type.
 
using bfloat16 = ::hip_bfloat16
 bfloat16 floating point type
 
using lane_mask_type = unsigned long long int
 The lane_mask_type is an integer that contains one bit per thread. More...
 
using native_half = _Float16
 Native half-precision floating point type.
 
using native_bfloat16 = bfloat16
 native bfloat16 type
 

Detailed Description

Typedef Documentation

◆ lane_mask_type

using lane_mask_type = unsigned long long int

The lane_mask_type is an integer that contains one bit per thread.

The total number of bits is equal to the total number of threads in a warp. Used to for warp-level operations.

Note
This is defined only on the device side.