2.2 WCET Results
The following table lists the normal execution time and WCET of a 32 x 32 matrix multiplication task measured using the example application.
Program Segment | U54_1 Core | U54_4 Core4 | |
---|---|---|---|
Code (bytes) |
ITIM (12736) |
LIM (804961) |
DDR cached |
Data |
LIM (780002) |
LIM |
DDR cached |
Stack |
LIM |
LIM |
DDR cached |
Normal Execution Time (±deviation) in mcycles |
8407148 (±21 cycles) (14.0119 ms) |
8407167 (±4 cycles) (14.0119 ms) |
1955036 (±136 cycles) (3.258 ms) |
WCET3 |
8409198 (±61 cycles) (14.0153 ms) |
8409426 (±200 cycles) (14.0157 ms) |
1971234 (±1653 cycles) (3.285 ms) |
- All sections are combined.
- Data and stack are combined.
- WCET includes the following:
- Branch prediction trashing and branch prediction scheme (static) are enabled in core U54_4 and U54_1.
- Adversarial runs on U54_3; cache flush runs on U54_2.
- Interrupt with execution time (1000 mcycle).
- The DDR cached execution timings are given for comparison purpose only.
Note: WCET (seconds) = (1 ∕ (CPU
Core Frequency in MHz × 1000 × 1000)) × (number of execution cycles),
where
CPU Core Frequency = 600 MHz