Special Data Types
Some data types (such as 24s) are not supported by the HMPP. You can use the Convert function to convert the data types to those supported by the HMPP for subsequent processing.
For 24-bit signed data, each vector element consists of three consecutive 8-bit unsigned character bytes (uint8_t) and is stored in little-endian byte order (the lower eight bits of the vector element are stored in the lower address). The sign is represented by the first bit of the eight most significant bits. These data can be converted to 32-bit signed integer (32s) or 32-bit floating-point (32f) type by using the Convert function in the HMPP.
Table 1 describes the user-defined data ranges in the HMPP.
Macro Definition |
Maximum Value |
Description |
|---|---|---|
UINT24_MAX |
16777215 |
The corresponding data type is not defined and is stored in the variable of the uint32_t type. |
INT24_MAX |
8388607 |
The corresponding data type is not defined and is stored in the variable of the int32_t type. |
INT24_MIN |
-8388608 |
The corresponding data type is not defined and is stored in the variable of the int32_t type. |