Visible to Intel only — GUID: upo1615859866123
Ixiasoft
Visible to Intel only — GUID: upo1615859866123
Ixiasoft
12.1.1. Math Function Hardware Implementation Summary
The following table summarizes the default hardware implementations for math functions for different data type and what combinations of data type and math function support hardware implementation control.
Operand 1 |
Operand 2 |
Addition | Subtraction | Multiplication | Division | Dot |
---|---|---|---|---|---|---|
float |
float |
(D)/S | (D)/S | D | S+D | D |
float |
(D)/S | (D)/S | S | <const numerator> D |
D | |
double |
double |
S | S | S+D | S+D | S+D |
double |
S | S | S | <const numerator> D |
S | |
int/ac_int |
int/ac_int |
S | S | S+D | S | S+D |
int/ac_int |
S | S | S | <const numerator> D |
S | |
ac_fixed |
ac_fixed |
S | S | S+D | S+D | S+D |
ac_fixed |
S | S | S | <const numerator> D |
S | |
hls_float |
hls_float |
The hls_float<8,23> data type uses the same implementation as float. Other hls_float data types use the same implementation as double. |
||||
hls_float |
- S
- Math function is implemented only in soft logic with ALMs.
- D
- Math function is implemented only using DSPs.
- S+D
- Math function is implemented with a combination of soft logic (ALMs) and DSPs.
- (D)/S
-
Bold text indicates that hardware implementation controls are available for that data type and math function.
Bold-italic text in parentheses indicates the default hardware implementation in the absence of applying any hardware implementation controls.