RISC-V 552 Processor

Performance Benchmarks

The following performance metrics were obtained by running assembly programs from the ./non_proc_files/benchmark folder on our RISC-V processor implementation.

Baseline Without Cache (CPI)

Benchmark	CPI
image_filter	6.456857
hash_lookup	7.612327
bubblesort_large	7.952474
matmul_4x4	6.702153

Baseline With Write-Through Cache (CPI)

Benchmark	CPI
image_filter	1.398761
hash_lookup	1.816820
bubblesort_large	1.623710
matmul_4x4	1.496691

Baseline With Write-Through and 2-Bit Saturating Branch Predictor (CPI)

Benchmark	CPI
image_filter	1.398761
hash_lookup	1.834101
bubblesort_large	1.617658
matmul_4x4	1.403582

Baseline With Write-Through and Prefetching (CPI)

Benchmark	CPI
image_filter	1.396640
hash_lookup	1.687212
bubblesort_large	1.616590
matmul_4x4	1.490962

Baseline With Write-Back Cache (CPI)

Benchmark	CPI
image_filter	1.394243
hash_lookup	1.761521
bubblesort_large	1.572090
matmul_4x4	1.494633

Baseline With Write-Back Cache and 2-Way/32-Byte (CPI)

Benchmark	CPI
image_filter	1.395792
hash_lookup	1.717742
bubblesort_large	1.579566
matmul_4x4	1.496301

Branch Prediction Accuracy

Predictor Type	image_filter	hash_lookup	bubblesort_large	matmul_4x4
Always Not-Taken	97.60%	84.31%	73.50%	51.57%
2-Bit Saturating (Dynamic)	97.60%	78.83%	82.28%	94.00%

Memory Traffic (Bytes) - Data Cache Only

Cache Policy	image_filter	hash_lookup	bubblesort_large	matmul_4x4
Write-Back	400	288	336	272
Write-Through	488	360	420	328

Total Cache Misses (Data + Instruction)

Cache Configuration	image_filter	hash_lookup	bubblesort_large	matmul_4x4
2-way/16-byte	58	50	71	53
2-way/32-byte	30	27	36	27
4-way/16-byte	58	50	71	53

Instruction Cache Misses - Prefetching Impact

Configuration	image_filter	hash_lookup	bubblesort_large	matmul_4x4
With Prefetching	1	3	1	1
Without Prefetching	46	40	29	36