Filesystem (default vs base) Impact on Cyclictest Latencies

Headline:

Filesystem choice (default vs base) does not produce a verifiable impact on max latency above the run-to-run variability observed in the data. The per-pair Δmax values are scattered in sign and overlap with the inter-run shifts seen in the rest of the data set.
SDK test: 8 pairs split roughly evenly. Mean ΔCore1 max ≈ +4 µs (base slightly higher). 4 pairs go each way on Core1, no consistent direction.
Customer test (-i1000): 8 pairs slightly favor base FS on Core1 (6/8 negative ΔCore1 max, mean ≈ -18 µs). On Core0 the pair-to-pair swings are large (-225 to +191 µs) and dominated by other variables.
Customer -i200 -M: only 2 pairs, both show base FS worse on Core0; one shows base substantially worse on Core1 too. Insufficient data for a directional conclusion.
The OoB-OPTEE / no-DDR / non-isolated customer pair (260529_2 vs 260529_1) shows ΔCore0 = +191 µs (base much worse), but the pair on OPTEE-PRNG / no-DDR / non-isolated under the same column shows ΔCore0 = -54 µs (base better). The same factor-pair in two adjacent conditions disagrees, indicating that the swings are not driven by filesystem.

What is being compared

"default" is the full Arago distribution image used by the AM62L SDK; "base" is a stripped-down image with fewer background services running. Refer to the Linux SDK docs for more information about the filesystem.

Run-to-run noise observation

Across the filesystem A/B pairs, no trend in either Core 0 or Core 1 maximum latency rises above the run-to-run variability already visible in the rest of the data. Comparable shifts in max latency appear between runs that should be more controlled (for example pairs where the only deliberately changed variable is the OPTEE binary or the DDR settings) yet still see multi-tens-of-microsecond differences in Core 0 or Core 1 max. Filesystem choice is therefore not separable from inter-run noise on this metric.

SDK test pairs (8)

Aggregate Δmax (B − A) where A=default FS, B=base FS, so negative = base FS better:
Core0: mean Δmax0 = -17.8 µs | base FS better in 5/8 pairs, worse in 3, tied in 0.
Core1: mean Δmax1 = +4.2 µs | base FS better in 3/8 pairs, worse in 4, tied in 1.

condition (fixed)	default FS file	max0	max1	base FS file	max0	max1	Δmax0	Δmax1	plot
col=SDK test, fs=default, optee=oob, iso=n, ddr=n	260518_1	126	66	260522_2	51	122	-75	+56	histograms ↓
col=SDK test, fs=default, optee=oob, iso=n, ddr=Y	260526_4	104	107	260524_1	72	85	-32	-22	histograms ↓
col=SDK test, fs=default, optee=oob, iso=Y, ddr=n	260530_4	101	111	260530_1	108	102	+7	-9	histograms ↓
col=SDK test, fs=default, optee=oob, iso=Y, ddr=Y	260526_2	104	99	260524_3	93	84	-11	-15	histograms ↓
col=SDK test, fs=default, optee=prng, iso=n, ddr=n	260519_1	62	55	260522_4	54	63	-8	+8	histograms ↓
col=SDK test, fs=default, optee=prng, iso=n, ddr=Y	260531_4	81	68	260523_3	54	83	-27	+15	histograms ↓
col=SDK test, fs=default, optee=prng, iso=Y, ddr=n	260530_2	61	36	260531_1	62	37	+1	+1	histograms ↓
col=SDK test, fs=default, optee=prng, iso=Y, ddr=Y	260524_2	64	37	260527_1	67	37	+3	0	histograms ↓

Customer test (-i1000) pairs (8)

Aggregate Δmax (B − A) where A=default FS, B=base FS, so negative = base FS better:
Core0: mean Δmax0 = -28.4 µs | base FS better in 5/8 pairs, worse in 2, tied in 1.
Core1: mean Δmax1 = -18.0 µs | base FS better in 6/8 pairs, worse in 2, tied in 0.

condition (fixed)	default FS file	max0	max1	base FS file	max0	max1	Δmax0	Δmax1	plot
col=Customer -i1000, fs=default, optee=oob, iso=n, ddr=n	260529_2	269	78	260529_1	460	72	+191	-6	histograms ↓
col=Customer -i1000, fs=default, optee=oob, iso=n, ddr=Y	260528_4	251	59	260528_3	203	69	-48	+10	histograms ↓
col=Customer -i1000, fs=default, optee=oob, iso=Y, ddr=n	260522_3	475	266	260529_4	250	180	-225	-86	histograms ↓
col=Customer -i1000, fs=default, optee=oob, iso=Y, ddr=Y	260525_2	226	133	260525_1	240	113	+14	-20	histograms ↓
col=Customer -i1000, fs=default, optee=prng, iso=n, ddr=n	260531_2	252	72	260531_3	198	73	-54	+1	histograms ↓
col=Customer -i1000, fs=default, optee=prng, iso=n, ddr=Y	260528_1	83	63	260528_2	83	49	0	-14	histograms ↓
col=Customer -i1000, fs=default, optee=prng, iso=Y, ddr=n	260529_3	251	159	260530_3	175	149	-76	-10	histograms ↓
col=Customer -i1000, fs=default, optee=prng, iso=Y, ddr=Y	260523_2	101	135	260526_1	72	116	-29	-19	histograms ↓

Customer test (-i200 -M) pairs (2)

Aggregate Δmax (B − A) where A=default FS, B=base FS, so negative = base FS better:
Core0: mean Δmax0 = +89.0 µs | base FS better in 0/2 pairs, worse in 2, tied in 0.
Core1: mean Δmax1 = +62.0 µs | base FS better in 0/2 pairs, worse in 2, tied in 0.

condition (fixed)	default FS file	max0	max1	base FS file	max0	max1	Δmax0	Δmax1	plot
col=Customer -i200 -M, fs=default, optee=oob, iso=Y, ddr=Y	260525_4	126	98	260525_3	234	221	+108	+123	histograms ↓
col=Customer -i200 -M, fs=default, optee=prng, iso=Y, ddr=Y	260523_1	53	70	260526_3	123	71	+70	+1	histograms ↓

Aggregate trends across multiple conditions

Comparing all 16 customer-test (-i1000) runs split by FS:

Default FS Core1 max: 59, 63, 72, 78, 133, 135, 159, 266 µs. Median ≈ 105 µs.
Base FS Core1 max: 49, 69, 72, 73, 113, 116, 149, 180 µs. Median ≈ 92 µs.

The two distributions overlap heavily. The base-FS median is ~13 µs lower, which is smaller than the largest run-to-run shifts in the data. The observation is consistent with "base FS is not making it worse, may be slightly better, cannot be confirmed".

Conclusions

Filesystem is not a primary lever for the customer's Core1 max goal. The other factor reports show that core isolation OFF, DDR mods, and OPTEE-PRNG together account for almost all the achievable Core1 max reduction in the data set; FS choice is at most a small additional contributor.
If the customer wants the smallest fleet image, base FS is safe. No pair shows base FS being meaningfully worse on Core1 max in the customer (-i1000) test.

Caveats

The two -i200 -M filesystem pairs disagree with the -i1000 trend (both show base worse). With only 2 pairs and the run-to-run noise observed elsewhere, this cannot be resolved without additional -i200 -M base-FS runs.
FS pairs span up to 13 days of calendar time (e.g. 260518 vs 260530). Any systematic platform drift over that window is mixed in with the FS effect and cannot be separated without same-day pairing or controls.

Appendix — per-pair histogram overlays

Reading the plots: filled circles are the baseline (A) run, "x" markers are the changed (B) run. The top panel is Core 0, the bottom panel is Core 1, both on a log-y count axis. A bump in the A curve that vanishes in the B curve is direct visual evidence that the factor under test produced or removed a latency cluster.

SDK test

fs=default, OPTEE=oob, iso=n, DDR=n | default FS=260518_1 (max0=126, max1=66) vs base FS=260522_2 (max0=51, max1=122) ↑ back to top

fs=default, OPTEE=oob, iso=n, DDR=Y | default FS=260526_4 (max0=104, max1=107) vs base FS=260524_1 (max0=72, max1=85) ↑ back to top

fs=default, OPTEE=oob, iso=Y, DDR=n | default FS=260530_4 (max0=101, max1=111) vs base FS=260530_1 (max0=108, max1=102) ↑ back to top

fs=default, OPTEE=oob, iso=Y, DDR=Y | default FS=260526_2 (max0=104, max1=99) vs base FS=260524_3 (max0=93, max1=84) ↑ back to top

fs=default, OPTEE=prng, iso=n, DDR=n | default FS=260519_1 (max0=62, max1=55) vs base FS=260522_4 (max0=54, max1=63) ↑ back to top

fs=default, OPTEE=prng, iso=n, DDR=Y | default FS=260531_4 (max0=81, max1=68) vs base FS=260523_3 (max0=54, max1=83) ↑ back to top

fs=default, OPTEE=prng, iso=Y, DDR=n | default FS=260530_2 (max0=61, max1=36) vs base FS=260531_1 (max0=62, max1=37) ↑ back to top

fs=default, OPTEE=prng, iso=Y, DDR=Y | default FS=260524_2 (max0=64, max1=37) vs base FS=260527_1 (max0=67, max1=37) ↑ back to top

Customer test (-i1000)

fs=default, OPTEE=oob, iso=n, DDR=n | default FS=260529_2 (max0=269, max1=78) vs base FS=260529_1 (max0=460, max1=72) ↑ back to top

fs=default, OPTEE=oob, iso=n, DDR=Y | default FS=260528_4 (max0=251, max1=59) vs base FS=260528_3 (max0=203, max1=69) ↑ back to top

fs=default, OPTEE=oob, iso=Y, DDR=n | default FS=260522_3 (max0=475, max1=266) vs base FS=260529_4 (max0=250, max1=180) ↑ back to top

fs=default, OPTEE=oob, iso=Y, DDR=Y | default FS=260525_2 (max0=226, max1=133) vs base FS=260525_1 (max0=240, max1=113) ↑ back to top

fs=default, OPTEE=prng, iso=n, DDR=n | default FS=260531_2 (max0=252, max1=72) vs base FS=260531_3 (max0=198, max1=73) ↑ back to top

fs=default, OPTEE=prng, iso=n, DDR=Y | default FS=260528_1 (max0=83, max1=63) vs base FS=260528_2 (max0=83, max1=49) ↑ back to top

fs=default, OPTEE=prng, iso=Y, DDR=n | default FS=260529_3 (max0=251, max1=159) vs base FS=260530_3 (max0=175, max1=149) ↑ back to top

fs=default, OPTEE=prng, iso=Y, DDR=Y | default FS=260523_2 (max0=101, max1=135) vs base FS=260526_1 (max0=72, max1=116) ↑ back to top

Customer test (-i200 -M)

fs=default, OPTEE=oob, iso=Y, DDR=Y | default FS=260525_4 (max0=126, max1=98) vs base FS=260525_3 (max0=234, max1=221) ↑ back to top

fs=default, OPTEE=prng, iso=Y, DDR=Y | default FS=260523_1 (max0=53, max1=70) vs base FS=260526_3 (max0=123, max1=71) ↑ back to top