LithosAnanake/experiments/bare_metal/runs/doe-aarch64-20260803-221007.csv at daf8f82f3b096514d2ac1438395b048190825c81

Files

T

Robert Allan JamesandClaude Sonnet 5 43aa0a2a36 aarch64: arm the ARM Generic Timer -- CNTP/CNTHP TVAL+CTL, per-tick re-arm

Punch list §25 item 0.7 complete. This is what finally makes items 0.5 and
0.6 provably work end to end.

apic_timer_start(): writes CNTP_TVAL_EL0 (or CNTHP_TVAL_EL2 at EL2 --
aarch64_current_el(), same EL-aware discipline as 0.4-0.6) to
s_timer_period_tsc, then CTL.ENABLE=1/IMASK=0, followed by an ISB. The ISB
is not decorative: confirmed against Linux's own arch_timer_reg_write_cp15()
(arch/arm64/include/asm/arch_timer.h) that only the *control* register write
needs synchronising before the enable/mask state is guaranteed visible to
the interrupt pipeline -- TVAL/CVAL writes do not carry the same
requirement, which is why apic_timer_rearm() omits it.

TVAL is architecturally 32-bit but MSR-to-system-register is always a
64-bit instruction form -- passing a uint32_t operand directly failed to
build (-Wasm-operand-widths). Fixed by truncating to 32 bits ourselves then
zero-extending back to 64 for the operand, which supplies explicit,
provably-correct zeros in the RES0 upper field rather than depending on
unverified hardware behaviour -- Linux's own driver never exercises this
path (it always uses the 64-bit CVAL form instead), so there was no local
source to confirm the alternative against.

apic_timer_rearm() (new): re-writes TVAL only, no ISB needed. TVAL is
relative to "now," not an absolute deadline like riscv64's SBI interface
(item 0.3), so there is no drift-correction bookkeeping -- each write means
"N ticks from this instant." Wired into aarch64_irq_handler() and called
*first*, before heartbeat_tick(), matching riscv64_timer_rearm()'s ordering
discipline exactly: the ARM Generic Timer does not auto-reload, so a return
path that skips this leaves the interrupt condition latched, which the GIC
would redeliver the instant it's EOI'd -- a real storm, the same class of
failure item 0.6's verification investigated (and that time found absent,
because nothing was armed yet).

Verified: builds clean; every generated instruction checked against
disassembly, not just reviewed by eye (both EL branches, correct TVAL/CTL
register names, single shared ISB in apic_timer_start(), no ISB in
apic_timer_rearm()). Boots to ok> with no regression, dict_hash
0x3d4e1daf289da94f unchanged.

Rate measured directly against real wall-clock time via QEMU's own -d int
trap trace (same method as riscv64's item 0.3), two independent windows:
1,090 interrupts over 11.05 s (98.679 Hz) and 4,031 over 40.88 s (98.614 Hz)
-- consistent across both, so this is a real, small, systematic bias
(~1.3-1.4% slow), not measurement noise from polling granularity, which
would have shrunk with the longer window and did not. Attributed to genuine
per-interrupt service latency: TVAL is rewritten mid-ISR, so the trampoline
save/restore, GICC_IAR read, EL branch and GICC_EOIR write all lengthen the
effective period slightly versus the nominal 10 ms, inherent to any
relative-countdown re-arm scheme. Reported as measured, not smoothed over.
The interrupt sustained continuously across both windows with no stall and
no storm, which is the primary evidence re-arm-every-tick is correct;
the small rate bias is overhead, not a defect.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

2026-08-03 22:14:52 -04:00

4.0 KiB

Raw Blame History

1	tick_number	elapsed_ns	tick_interval_ns	word_executions_delta	hot_word_count	avg_word_heat_q48	window_width	actual_window_size	time_trust_q48	hera_heat_q48
2	1	10000	10000	183	4	45	731	731	65536	65536
3	2	20000	10000	173	5	44	582	582	65536	65536
4	3	30000	10000	184	6	60	582	582	65536	65536
5	4	40000	10000	174	8	71	4096	0	65536	65536
6	5	50000	10000	150	9	78	582	582	65536	65536
7	6	60000	10000	159	11	73	4096	0	65536	65536
8	7	70000	10000	188	15	71	685	685	65536	65536
9	8	80000	10000	214	16	83	685	685	65536	65536
10	9	90000	10000	216	19	82	4096	0	65536	65536
11	10	100000	10000	222	21	84	567	567	65536	65536
12	11	110000	10000	207	23	86	731	731	65536	65536
13	12	120000	10000	189	26	83	4096	0	65536	65536
14	13	130000	10000	221	28	80	731	731	65536	65536
15	14	140000	10000	204	29	83	685	685	65536	65536
16	15	150000	10000	203	30	82	731	731	65536	65536
17	16	160000	10000	201	32	83	567	567	65536	65536
18	17	170000	10000	190	34	86	731	731	65536	65536
19	18	180000	10000	170	35	73	582	582	65536	65536
20	19	190000	10000	154	28	44	731	731	65536	65536
21	20	200000	10000	154	28	41	567	567	65536	65536
22	21	210000	10000	153	32	39	4096	0	65536	65536
23	22	220000	10000	167	35	42	582	582	65536	65536
24	23	230000	10000	176	37	41	685	685	65536	65536
25	24	240000	10000	175	37	39	731	731	65536	65536
26	25	250000	10000	196	38	37	731	731	65536	65536
27	26	260000	10000	192	41	37	4096	0	65536	65536
28	27	270000	10000	158	40	38	685	685	65536	65536
29	28	280000	10000	158	40	35	4096	0	65536	65536
30	29	290000	10000	164	41	37	4096	0	65536	65536
31	30	300000	10000	154	43	39	685	685	65536	65536
32	31	310000	10000	172	46	35	582	582	65536	65536
33	32	320000	10000	180	49	34	567	567	65536	65536
34	33	330000	10000	145	51	35	567	567	65536	65536
35	34	340000	10000	203	53	37	582	582	65536	65536
36	35	350000	10000	154	55	39	731	731	65536	65536
37	36	360000	10000	144	57	37	4096	0	65536	65536
38	37	370000	10000	144	57	35	4096	0	65536	65536
39	38	380000	10000	132	57	34	4096	0	65536	65536
40	39	390000	10000	180	57	32	4096	0	65536	65536
41	40	400000	10000	220	63	29	4096	0	65536	65536
42	41	410000	10000	215	64	28	4096	24	65536	65536
43	42	420000	10000	208	68	29	4096	24	65536	65536
44	43	430000	10000	177	72	27	4096	24	65536	65536
45	44	440000	10000	199	77	28	4096	24	65536	65536
46	45	450000	10000	223	82	26	4096	30	65536	65536
47	46	460000	10000	202	85	24	4096	67	65536	65536
48	47	470000	10000	218	91	25	4096	198	65536	65536
49	48	480000	10000	221	92	26	4096	350	65536	65536
50	49	490000	10000	225	97	26	4096	528	65536	65536
51	50	500000	10000	214	100	27	4096	694	65536	65536
52	51	510000	10000	227	102	28	4096	883	65536	65536
53	52	520000	10000	208	102	29	4096	1043	65536	65536
54	53	530000	10000	207	102	30	4096	1202	65536	65536
55	54	540000	10000	200	102	31	4096	1319	65536	65536
56	55	550000	10000	205	103	32	4096	1464	65536	65536
57	56	560000	10000	211	103	32	4096	1620	65536	65536
58	57	570000	10000	193	110	33	4096	1712	65536	65536
59	58	580000	10000	190	5	67	4096	1775	65536	65536

4.0 KiB Raw Blame History

4.0 KiB

Raw Blame History