Reorganize source tree: kernel/, v3/, v4/ split and board infrastructure

Source tree reorganization:
- Move StarForth v3 engine to v3/ (src/, include/, Makefile)
- Move kernel to kernel/ (src/, include/, linker/, Makefile)
- Create v4/ skeleton for F18-ISA golden model (DECOMPOSITION.md, JUSTIFICATION.md)
- Move FABRIC-0..4.md to docs/fabric/
- Move ONTOLOGY.md and ROADMAP.md to docs/

Board infrastructure:
- Add boards/ser5/, boards/raspi/, boards/milkv/, boards/zynq7020/
- Each board has board.mk (ISA, CPU flags, boot recipe) and README.md
- Root Makefile becomes thin dispatcher: boot_image, all, clean, docs take TARGET
- make boot_image TARGET=SER5|RASPI|MILKV builds one GPT/MBR image per board
- ZYNQ7020 target exists but stops with clear error (ARMv7 port not built yet)
- scripts/mkdiskimage.sh builds disk images for all boards

Docs pipeline:
- docs/book/ with LaTeX master (main.tex) and Makefile
- pandoc converts Markdown to LaTeX at build time
- Two Lua filters: table-widths.lua (wide tables wrap), code-breaks.lua (inline code breaks)
- make docs builds single PDF (754 pages, 0 missing characters)
- make docs TARGET=<board> adds board appendix
- build/docs/<book|board>/meta.tex stamps git commit into PDF

Bug fixes:
- 42 include paths that only worked by accident now use correct relative paths
- clang-18 hardcode replaced with configurable CC variable (fixed aarch64 build)
- Pi 5: kernel_2712.img linked at 0x80000, .bss zeroed, memory reserved
- Doxyfile, .clang-tidy, README.md, Kconfig paths updated

Verified:
- Hosted v3 build passes 1012 tests, 0 failures
- SER5 image boots in QEMU (OVMF), POST passes, K exact (65536 = Q48_ONE)
- Milk-V image boots in QEMU (OpenSBI + U-Boot + bootefi), POST passes
- make clean TARGET=<board> removes only that board and its ISA objects
- make all builds all boards, hosted v3, and docs in one run

Co-authored-by: Junie <junie@jetbrains.com>
This commit is contained in:
rajames
2026-10-01 15:40:09 -04:00
co-authored by Junie
parent 3c709c115b
commit a8b70e88d3
630 changed files with 2646 additions and 2036 deletions
+288
View File
@@ -0,0 +1,288 @@
/*
StarForth — Steady-State Virtual Machine Runtime
Copyright (c) 2023–2025 Robert A. James
All rights reserved.
Licensed under the StarForth License, Version 1.0
*/
/**
* doe_log.c — DoE CSV logger for the LithosAnanke kernel
*
* Emits CSV rows to the serial log once per heartbeat tick. Each row is
* prefixed with a cyan [HADES][DOE ] tag so it can be extracted cleanly:
*
* grep -aP '\[HADES\]\[DOE \]' qemu-amd64-*.log
*
* Columns (in order):
* 0a. vm_name string — the emitting VM's own registry name
* (reverse lookup on vm->stadium_vm_id,
* "?" if unregistered at this tick).
* Added 2026-09-16 for the multiuser/
* multitasking DoE (FABRIC-3.md
* §XXXIII): every prior column already
* described *a* VM's state each row,
* but nothing said *which* VM emitted
* it, making concurrent-VM rows
* indistinguishable by source.
* 0b. vm_id_hex string — the emitting VM's own stadium_vm_id
* as "hi-lo" hex, same format
* PARITY:BIRTH/PARITY:KILL already
* print. Names get reused across a
* trial's own KILL+rebirth cycles;
* this is the one column unique to a
* single birth instance.
* 1. tick_number uint32 — monotonic heartbeat counter
* 2. elapsed_ns uint64 — ns since run start
* 3. tick_interval_ns uint64 — actual interval from prior tick
* 4. cache_hits_delta uint32 — hot-words cache hits this tick (TODO: 0)
* 5. bucket_hits_delta uint32 — bucket hits this tick (TODO: 0)
* 6. word_executions_delta uint32 — words executed this tick (TODO: 0)
* 7. hot_word_count uint64 — words with heat >= threshold
* 8. avg_word_heat_q48 uint64 — mean execution heat as raw Q48.16 integer
* 9. window_width uint32 — rolling window effective size
* 10. actual_window_size uint32 — true analysis width: min(total_executions, effective_window_size)
* 11. predicted_label_hits uint32 — ANOVA early-exit confirmations per tick (L8 validation signal)
* 12. jitter_bits uint64 — estimated_jitter_ns IEEE 754 raw bits (union-punned)
* 13. apic_ticks uint64 — APIC timer monotonic tick count
* 14. time_trust_q48 uint64 — TIME-TRUST in Q48.16 format
* 15. variance_q48 uint64 — timing variance in Q48.16 format
* 16. hera_heat_q48 uint64 — VM fleet physics: Hera's execution_heat_q48
* 17. hermes_heat_q48 uint64 — VM fleet physics: Hermes's execution_heat_q48
* 18. artemis_heat_q48 uint64 — VM fleet physics: Artemis's execution_heat_q48
* 19. fleet_k_q48 uint64 — "K": vm_physics_fleet_heat_sum() over ALL live
* VMs (not just the 3 named Tripod members above --
* this is the genuine conservation invariant, which
* should read Q48_ONE (65536) whenever conservation
* holds). Requested 2026-09-12 alongside the 3x9
* factorial analysis in experiments/std79-doe/ --
* columns 16-18 alone cannot reconstruct this since
* they omit every identity VM's own heat.
* 20. fleet_conserved uint32 — vm_physics_conserved() as 0/1: whether
* |fleet_k_q48 - Q48_ONE| < VM_PHYSICS_EPSILON_Q48
* at the moment this row was captured.
* 21. switch_count_cumulative uint64 — Stage 3 preemptive-switch count, cumulative
* since boot (capsule_vm_switch_signal.c).
* 22. switch_current_slot uint32 — which registered switch-signal participant is
* executing at this tick (0=Hera,1=Hermes,
* 2=Artemis by current registration order;
* 0xFFFFFFFF = none/unregistered at this moment).
* 23. switch_hera_readiness uint32 — Hera's raw readiness accumulator (slot 0).
* 24. switch_hermes_readiness uint32 — Hermes's raw readiness accumulator (slot 1).
* 25. switch_artemis_readiness uint32 — Artemis's raw readiness accumulator (slot 2).
* 26. switch_ticks_since uint32 — ticks elapsed since the last real switch
* (capsule_vm_switch_signal.c); empirically
* verifies the fixed 50-tick threshold under
* real DoE load rather than assuming it.
*
* Columns 21-26 (FABRIC-3.md §XXVIII Stage 3 follow-on, 2026-09-13): added
* alongside Stage 3's timer-driven preemptive context switching -- read-only
* exposure of state the switch-signal module already tracked internally for its
* own decision-making; no new measurement logic, just made observable. Slot
* readiness columns are hardcoded to the 3 known Tripod slots (same convention
* as columns 16-18) rather than a dynamic per-slot column set, for the same
* reason: this is a logging schema for the currently-known fleet, not the
* switch-signal mechanism itself (which stays slot-count-agnostic, headroom for
* up to 8 participants -- see capsule_vm_switch_signal.h).
*
* Note: avg_word_heat and estimated_jitter_ns are stored as double in the
* snapshot but the freestanding snprintf has no %%f support. avg_word_heat
* is emitted as raw Q48.16 (multiply by 65536); jitter is union-punned to its
* IEEE 754 uint64 bit pattern to avoid UB from out-of-range cast.
*
* Columns 16-18 (VM-FLEET-ATTRACTOR-DESIGN-20260705.md, Phase 3): looked up
* by name each tick rather than by a fixed vm_id, since kill/rebirth (e.g.
* TRIPOD-TEST's K-soak) assigns a fresh vm_id on every rebirth. Fixed to the
* three known Tripod VMs, not a dynamic per-VM column set -- this is a
* logging schema for a specific known fleet, not the physics mechanism
* itself (which stays genuinely VM-count-agnostic). Reads 0 for any VM not
* currently found in the registry (e.g. Hermes momentarily during its
* kill/rebirth window), which is a legitimate observation, not an error.
*/
#include "starkernel/doe_log.h"
#include "starkernel/console.h"
#include "starkernel/timer.h"
#include "starkernel/capsule_birth.h"
#include "starkernel/capsule_vm_physics.h"
#include "starkernel/capsule_vm_switch_signal.h" /* FABRIC-3.md §XXVIII Stage 3 follow-on */
#include "freestanding/stdio.h"
#include "word_registry.h"
int g_doe_log_enabled = 0; /* off by default -- HB-ON to enable, per instruction */
#define DOE_PREFIX "\x1b[36m[HADES][DOE ]\x1b[0m "
#define DOE_BUF_SIZE 512
static const char *doe_header =
DOE_PREFIX
"vm_name,vm_id_hex,"
"tick_number,elapsed_ns,tick_interval_ns,"
"cache_hits_delta,bucket_hits_delta,word_executions_delta,"
"hot_word_count,avg_word_heat_q48,"
"window_width,actual_window_size,predicted_label_hits,jitter_bits,"
"apic_ticks,time_trust_q48,variance_q48,vm_call_depth_max,"
"hera_heat_q48,hermes_heat_q48,artemis_heat_q48,"
"fleet_k_q48,fleet_conserved,"
"switch_count_cumulative,switch_current_slot,"
"switch_hera_readiness,switch_hermes_readiness,switch_artemis_readiness,"
"switch_ticks_since";
/* Looks up a Tripod VM's fleet heat by name; 0 if not currently registered
* (e.g. between KILL and rebirth). */
static uint64_t doe_log_heat_by_name(const char *name)
{
VMRegistryEntry entry;
if (capsule_vm_find_by_name_nocase(name, &entry) != 0) return 0;
return vm_physics_heat_of(entry.vm_id);
}
void doe_log_tick_row(VM *vm, const HeartbeatTickSnapshot *snap)
{
if (!g_doe_log_enabled)
return;
if (!snap)
return;
/* Print header once before the first data row */
static int header_printed = 0;
if (!header_printed) {
console_puts(doe_header);
console_puts("\r\n");
header_printed = 1;
}
/* Pull APIC timer trust state */
const TimeTrustState *ts = heartbeat_state();
uint64_t apic_ticks = ts ? ts->ticks : 0;
uint64_t time_trust_q48 = ts ? (uint64_t)ts->trust : 0;
uint64_t variance_q48 = ts ? (uint64_t)ts->variance : 0;
/* Convert doubles to integer-safe representations (freestanding snprintf has no %f).
* Use union punning for jitter to avoid UB from out-of-range double→int64 cast. */
uint64_t avg_heat_q48 = (uint64_t)(snap->avg_word_heat * 65536.0);
union { double d; uint64_t u; } jitter_bits;
jitter_bits.d = snap->estimated_jitter_ns;
uint64_t jitter_raw = jitter_bits.u;
/* Format the 20-column CSV row into a local buffer.
* Use %llu (unsigned long long) for all uint64_t fields — %lu is unreliable
* for values > 2^32 on aarch64 due to mixed-width varargs ABI behaviour. */
unsigned vm_call_depth_max = (vm && vm->call_stack_max > 0) ? (unsigned)vm->call_stack_max : 0;
/* Identity columns (2026-09-16, FABRIC-3.md §XXXIII): reverse-lookup
* this row's own emitting VM by its stadium_vm_id -- name can be
* reused across a KILL+rebirth cycle, so vm_id_hex is what actually
* disambiguates one birth instance from the next. "?" name / all-zero
* id if vm is NULL or not (yet) registered, never a crash. */
const char *vm_name = "?";
char vm_id_hex[34] = "0000000000000000-0000000000000000";
if (vm) {
VMRegistryEntry self_entry;
if (capsule_vm_registry_get(vm->stadium_vm_id, &self_entry) == 0) {
vm_name = self_entry.name;
}
/* Hand-rolled hex, not snprintf("%016llx", ...) -- found live
* (2026-09-16) that this freestanding snprintf doesn't support
* width+ll+x and silently prints the literal format string
* instead of substituting. Same nibble-table idiom vm_uuid.c's
* own vm_uuid_format() and MINT-SCRATCH-EMIT already use. */
{
static const char hexd[] = "0123456789abcdef";
uint64_t halves[2] = { vm->stadium_vm_id.hi, vm->stadium_vm_id.lo };
int pos = 0, h;
for (h = 0; h < 2; h++) {
int i;
for (i = 15; i >= 0; i--) {
vm_id_hex[pos++] = hexd[(halves[h] >> (i * 4)) & 0xF];
}
if (h == 0) vm_id_hex[pos++] = '-';
}
vm_id_hex[pos] = '\0';
}
}
uint64_t hera_heat_q48 = doe_log_heat_by_name("Hera");
uint64_t hermes_heat_q48 = doe_log_heat_by_name("Hermes");
uint64_t artemis_heat_q48 = doe_log_heat_by_name("Artemis");
uint64_t fleet_k_q48 = vm_physics_fleet_heat_sum();
unsigned fleet_conserved = vm_physics_conserved() ? 1u : 0u;
/* FABRIC-3.md §XXVIII Stage 3 follow-on (2026-09-13): switch-signal
* exposure, looked up by name each tick (same convention as the heat
* columns above -- registration order is not assumed fixed). */
uint64_t switch_count = sk_vm_switch_signal_switch_count();
int switch_slot = sk_vm_switch_signal_current_slot();
unsigned switch_current_slot = (switch_slot >= 0) ? (unsigned)switch_slot : 0xFFFFFFFFu;
uint32_t switch_ticks_since = sk_vm_switch_signal_ticks_since_switch();
uint32_t hera_readiness = 0, hermes_readiness = 0, artemis_readiness = 0;
{
VMRegistryEntry e;
if (capsule_vm_find_by_name_nocase("Hera", &e) == 0)
hera_readiness = sk_vm_switch_signal_readiness_of(e.vm_id);
if (capsule_vm_find_by_name_nocase("Hermes", &e) == 0)
hermes_readiness = sk_vm_switch_signal_readiness_of(e.vm_id);
if (capsule_vm_find_by_name_nocase("Artemis", &e) == 0)
artemis_readiness = sk_vm_switch_signal_readiness_of(e.vm_id);
}
char buf[DOE_BUF_SIZE];
snprintf(buf, sizeof(buf),
"%s,%s,"
"%u,%llu,%llu,%u,%u,%u,%llu,%llu,%u,%u,%u,%llu,%llu,%llu,%llu,%u,%llu,%llu,%llu,%llu,%u,"
"%llu,%u,%u,%u,%u,%u",
vm_name,
vm_id_hex,
snap->tick_number,
(unsigned long long)snap->elapsed_ns,
(unsigned long long)snap->tick_interval_ns,
snap->cache_hits_delta,
snap->bucket_hits_delta,
snap->word_executions_delta,
(unsigned long long)snap->hot_word_count,
(unsigned long long)avg_heat_q48,
snap->window_width,
snap->actual_window_size,
snap->predicted_label_hits,
(unsigned long long)jitter_raw,
(unsigned long long)apic_ticks,
(unsigned long long)time_trust_q48,
(unsigned long long)variance_q48,
vm_call_depth_max,
(unsigned long long)hera_heat_q48,
(unsigned long long)hermes_heat_q48,
(unsigned long long)artemis_heat_q48,
(unsigned long long)fleet_k_q48,
fleet_conserved,
(unsigned long long)switch_count,
switch_current_slot,
hera_readiness,
hermes_readiness,
artemis_readiness,
switch_ticks_since);
console_puts(DOE_PREFIX);
console_puts(buf);
console_puts("\r\n");
}
/* HB-ON ( -- ): enable per-tick DoE instrumentation (g_doe_log_enabled=1). */
static void doe_word_hb_on(VM *vm)
{
(void)vm;
g_doe_log_enabled = 1;
}
/* HB-OFF ( -- ): disable per-tick DoE instrumentation (g_doe_log_enabled=0). */
static void doe_word_hb_off(VM *vm)
{
(void)vm;
g_doe_log_enabled = 0;
}
void register_doe_log_words(VM *vm)
{
register_word(vm, "HB-ON", doe_word_hb_on);
register_word(vm, "HB-OFF", doe_word_hb_off);
}