Experiment Log

Each bitstream will produce 4 results: UnixBench, NPB and also performance counters will be measured in each case:

Baseline bitstream

This bitstream is the bitstream at the end of the COMPAS 2025 tutorial with performance counters enabled at: ~/cva6/core/include/cv64a6_imafdch_sv39_config_pkg.sv

#from
localparam CVA6ConfigPerfCounterEn = 0;

#to
localparam CVA6ConfigPerfCounterEn = 1;

Project Summary

Implementation Utilization

✅ Reading report from: ./bitstreams/reports/utilization/SoC_wrapper_utilization_placed_base.rpt

Implementation Timing

--- Timing Summary ---

Metric	Value (ns)
WNS (Setup)	1.218
WHS (Hold)	0.051

File locations

Bitstream

./bitstreams/baseline.bit

Implementation utilization report

./bitstreams/reports/utilization/SoC_wrapper_utilization_placed_base.rpt

Implementation timing report

./bitstreams/reports/timing/SoC_wrapper_timing_summary_routed_base.rpt

Implementation power report

./bitstreams/reports/power/SoC_wrapper_power_routed_base.rpt

UnixBench

Experiment metadata ./runs/UnixBench/unixbench_19700101_064547_base.txt

✅ Loading data from: ./runs/UnixBench/unixbench_19700101_064547_base.txt

=== UnixBench Results (Detailed) ===

	Label	Category	IPC	Instructions	Cycles	Time_s	Score	Unit
0	Arith Overhead	Integer / ALU	0.388300	96160360	247619804	4.952396	22607716.000000	lps
1	Arith Register	Integer / ALU	0.568500	140705146	247489519	4.949790	658341.000000	lps
2	Arith Short	Integer / ALU	0.569000	140859170	247565240	4.951305	659170.000000	lps
3	Arith Int	Integer / ALU	0.568700	140769091	247527031	4.950541	658720.000000	lps
4	Dhrystone	Integer / ALU	0.447600	81389671	181828082	3.636562	259837.000000	lps
5	Arith Double	Integer / ALU	0.568800	140675372	247333099	4.946662	658256.000000	lps
6	Whetstone	Float / FPU	0.410800	103499392	251932322	5.038646	19.449000	MWIPS
7	Sys Mix	System / OS	0.174900	43254593	247366131	4.947323	22651.000000	lps
8	Sys GetPID	System / OS	0.225900	55509267	245775079	4.915502	205011.000000	lps
9	Sys Exec	System / OS	0.119100	2620492	21995437	0.439909	20.000000	lps
10	Pipe_Throughput	System / OS	0.140400	34742090	247521360	4.950427	8840.000000	lps
11	Context_Switching	System / OS	0.116900	14197502	121473123	2.429462	856.000000	lps

✅ Data exported to unixbench_results_final.csv

NAS Parallel Benchmark

Experiment metadata ./runs/NPB/NPB_19700101_000506_base.txt

=== NAS Parallel Benchmark Results ===

	Label	Category	IPC	Instructions	Cycles	Time_s	Score	Score Unit
0	EP	Compute Bound	0.297000	2091350139	7042643975	284.540000	0.120000	Mop/s
1	IS	Memory Bound	0.213100	7566437	35513805	1.440000	0.450000	Mop/s
2	CG	Memory Bound	0.167400	355825955	2125041156	85.900000	0.780000	Mop/s
3	MG	Memory Bound	0.192400	33959879	176499365	7.130000	1.070000	Mop/s
4	FT	Mixed / Streaming	0.159500	586665770	3678664319	148.680000	1.190000	Mop/s
5	BT	Mixed / Streaming	0.209400	549503500	2624049654	106.010000	2.150000	Mop/s
6	SP	Mixed / Streaming	0.181200	357293866	1971849827	79.680000	1.210000	Mop/s
7	LU	Mixed / Streaming	0.229800	200392992	872027385	35.360000	2.890000	Mop/s

1st Iteration

This bitstream increases all the parameters to the maximum to test the limits of the PYNQ-Z2 FPGA where the experiments take place.

#from
localparam CVA6ConfigBTBEntries = 16;
localparam CVA6ConfigDcacheByteSize = 4096;
localparam CVA6ConfigDcacheSetAssoc = 4;

localparam CVA6ConfigRASDepth = 2;
localparam CVA6ConfigBTBEntries = 16;
localparam CVA6ConfigBHTEntries = 16;
#to
localparam CVA6ConfigBTBEntries = 32;
localparam CVA6ConfigDcacheByteSize = 8192;
localparam CVA6ConfigDcacheSetAssoc = 2;

localparam CVA6ConfigRASDepth = 4;
localparam CVA6ConfigBTBEntries = 32;
localparam CVA6ConfigBHTEntries = 32;

Project Summary

Implementation Utilization

✅ Reading report from: ./bitstreams/reports/utilization/SoC_wrapper_utilization_placed_1st.rpt

Implementation Timing

--- Timing Summary ---

Metric	Value (ns)
WNS (Setup)	1.351
WHS (Hold)	0.056

File locations

Bitstream

./bitstreams/1st.bit

Implementation utilization report

./bitstreams/reports/utilization/SoC_wrapper_utilization_placed_1st.rpt

Implementation timing report

./bitstreams/reports/timing/SoC_wrapper_timing_summary_routed_1st.rpt

Implementation power report

./bitstreams/reports/power/SoC_wrapper_power_routed_1st.rpt

UnixBench

Experiment metadata ./runs/UnixBench/unixbench_19700101_003833_1st.txt

✅ Loading data from: ./runs/UnixBench/unixbench_19700101_003833_1st.txt

=== UnixBench Results (Detailed) ===

	Label	Category	IPC	Instructions	Cycles	Time_s	Score	Unit
0	Arith Overhead	Integer / ALU	0.391500	95753191	244578185	4.891564	22495175.000000	lps
1	Arith Register	Integer / ALU	0.573100	141364718	246647777	4.932956	661449.000000	lps
2	Arith Short	Integer / ALU	0.573100	140696214	245510428	4.910209	658204.000000	lps
3	Arith Int	Integer / ALU	0.573600	141335408	246388097	4.927762	661411.000000	lps
4	Dhrystone	Integer / ALU	0.452300	83175606	183896069	3.677921	265552.000000	lps
5	Arith Double	Integer / ALU	0.573400	141352529	246535206	4.930704	661348.000000	lps
6	Whetstone	Float / FPU	0.417900	103434562	247524633	4.950493	19.731000	MWIPS
7	Sys Mix	System / OS	0.180400	44336566	245778092	4.915562	23285.000000	lps
8	Sys GetPID	System / OS	0.215500	53320540	247461857	4.949237	195993.000000	lps
9	Sys Exec	System / OS	0.124600	2604873	20913995	0.418280	21.000000	lps
10	Pipe_Throughput	System / OS	0.152200	37684057	247551644	4.951033	9751.000000	lps
11	Context_Switching	System / OS	0.124700	14407118	115529727	2.310595	1009.000000	lps

NAS Parallel Benchmark

Experiment metadata ./runs/NPB/NPB_19700101_002355_1st.txt

=== NAS Parallel Benchmark Results ===

	Label	Category	IPC	Instructions	Cycles	Time_s	Score	Score Unit
0	EP	Compute Bound	0.312400	2082531833	6665771571	269.340000	0.120000	Mop/s
1	IS	Memory Bound	0.249300	7445551	29870149	1.220000	0.540000	Mop/s
2	CG	Memory Bound	0.179700	350957451	1953553106	79.030000	0.840000	Mop/s
3	MG	Memory Bound	0.202500	33901924	167451164	6.970000	1.090000	Mop/s
4	FT	Mixed / Streaming	0.165900	582655330	3513083635	141.860000	1.250000	Mop/s
5	BT	Mixed / Streaming	0.212300	548700198	2584271595	104.490000	2.190000	Mop/s
6	SP	Mixed / Streaming	0.188500	355398830	1885536072	76.360000	1.270000	Mop/s
7	LU	Mixed / Streaming	0.241600	199234007	824743421	33.420000	3.060000	Mop/s

2nd Iteration

#from
  localparam CVA6ConfigIcacheByteSize = 4096;
  localparam CVA6ConfigDcacheSetAssoc = 2;
#to
  localparam CVA6ConfigIcacheByteSize = 8192;
  localparam CVA6ConfigDcacheSetAssoc = 4;

Project Summary

Implementation Utilization

✅ Reading report from: ./bitstreams/reports/utilization/SoC_wrapper_utilization_placed_2nd.rpt

Implementation Timing

--- Timing Summary ---

Metric	Value (ns)
WNS (Setup)	1.496
WHS (Hold)	0.051

File locations

Bitstream

./bitstreams/2nd.bit

Implementation utilization report

./bitstreams/reports/utilization/SoC_wrapper_utilization_placed_2nd.rpt

Implementation timing report

./bitstreams/reports/timing/SoC_wrapper_timing_summary_routed_2nd.rpt

Implementation power report

./bitstreams/reports/power/SoC_wrapper_power_routed_2nd.rpt

UnixBench

Experiment metadata ./runs/UnixBench/unixbench_19700101_003833_1st.txt

✅ Loading data from: ./runs/UnixBench/unixbench_19700101_001958_2nd.txt

=== UnixBench Results (Detailed) ===

	Label	Category	IPC	Instructions	Cycles	Time_s	Score	Unit
0	Arith Overhead	Integer / ALU	0.408700	101317858	247886038	4.957721	23890762.000000	lps
1	Arith Register	Integer / ALU	0.597800	147624586	246961875	4.939237	691936.000000	lps
2	Arith Short	Integer / ALU	0.598900	148504569	247979271	4.959585	696400.000000	lps
3	Arith Int	Integer / ALU	0.598700	148399622	247887814	4.957756	695912.000000	lps
4	Dhrystone	Integer / ALU	0.471300	87991678	186683112	3.733662	281643.000000	lps
5	Arith Double	Integer / ALU	0.598800	148435454	247898996	4.957980	696012.000000	lps
6	Whetstone	Float / FPU	0.436200	108567539	248878043	4.977561	20.754000	MWIPS
7	Sys Mix	System / OS	0.207900	51521386	247796049	4.955921	27655.000000	lps
8	Sys GetPID	System / OS	0.249300	61428770	246364031	4.927281	229367.000000	lps
9	Sys Exec	System / OS	0.132100	2800643	21208683	0.424174	23.000000	lps
10	Pipe_Throughput	System / OS	0.195200	48358031	247792073	4.955841	13081.000000	lps
11	Context_Switching	System / OS	0.135800	15766359	116069269	2.321385	1216.000000	lps

NAS Parallel Benchmark

Experiment metadata ./runs/NPB/NPB_19700101_002355_1st.txt

=== NAS Parallel Benchmark Results ===

	Label	Category	IPC	Instructions	Cycles	Time_s	Score	Score Unit
0	EP	Compute Bound	0.325500	2075695973	6377044336	257.330000	0.130000	Mop/s
1	IS	Memory Bound	0.254800	7414347	29097247	1.170000	0.560000	Mop/s
2	CG	Memory Bound	0.186900	349903827	1872230866	75.600000	0.880000	Mop/s
3	MG	Memory Bound	0.214800	33482205	155861090	6.290000	1.210000	Mop/s
4	FT	Mixed / Streaming	0.171700	579361103	3373637514	136.150000	1.300000	Mop/s
5	BT	Mixed / Streaming	0.221700	545606453	2460613665	99.280000	2.300000	Mop/s
6	SP	Mixed / Streaming	0.199100	352486732	1770307322	71.440000	1.350000	Mop/s
7	LU	Mixed / Streaming	0.251000	198250642	789748912	31.860000	3.210000	Mop/s