Files
RP412/RP_L4/RPL4SPOOLSTATS.cpp
T
CydandClaude Fable 5 7d485c9672 The wire keeps what it could not send
Cyd asked for an analysis of the networking stack and what would make the
simulation feel better over the internet. The analysis found something
more urgent than latency: the transport has been losing data silently
since the arcade, and nothing in the game could see it happen.

Every send result was discarded - L4NET, the console, all of it. On the
1ms arcade LAN the socket buffer never filled, so it never mattered. Over
the internet it matters twice. A peer stalled in its own 10-30 second
mission load stops reading, its window closes, and our nonblocking send
starts answering would-block, which threw the message away; or worse,
answering a PARTIAL count, and since framing on that stream is recovered
purely from each message's length prefix, the bytes that never followed
sheared it for good. Both are reachable in an ordinary race, because
every race has a load in it.

So sends go through a bounded per-connection queue now. What the wire
will not take is kept, byte-exact, and retried at three flush points -
before the render (the present blocks on vsync, and this frame's state
should be travelling while it does), at the top of the receive pump, and
before a connect sequence. Nothing is ever dropped from the middle: these
are reliable ordered messages carrying entity creation, damage and race
control, so a queue that overflows its 256K declares the connection dead
and lets the disconnect path run rather than quietly desyncing the
stream. RP412NETSENDQ=0 restores the old behaviour and still logs what it
would have lost, which is the honest way to A/B it. On Steam the same
queue finally surfaces k_EResultLimitExceeded, which the old code
collapsed into -1 and discarded - that was backpressure, unlogged.

The receive side gained the check the release build never had. The length
prefix is untrusted input; Verify() compiles away in release, so a
corrupt one went to memmove as a negative, or copied 4096 bytes of
assembled packet into a 1600-byte stack buffer, or named a size the pad
could never complete and wedged the connection forever. It is now
validated against the same bounds the sender works to, and a stream that
fails them is dropped like any other lost peer.

And a fry that never ends: drop zones are map entities dealt round-robin
at load, ownership transfer is not implemented, so a leaver's pads stay
in the DropZones group. The respawn request dispatched to one goes to a
host that is gone - dropped at the send, the 'no host N in the table'
path - and the two-second retry re-dispatches to the same dead owner
forever. The pad scan now skips zones whose owner has left, and
re-validates one assigned earlier before reusing it.

The rest is measurement, because the symptoms this work exists to chase
are all reported in prose and none of them are in any log. Sixteen logs
from the six-player night contain zero player-facing latency lines. A
race now ends with a NetLog summary: per remote pod, how many updates
arrived and how evenly (median and p95 out of a log2 histogram), the
widest gap, how many gaps were long enough to mean a quiet sender versus
short enough to mean OUR loop stalled, how often its motion snapped
instead of blending, and how far arriving updates moved it. Per peer,
whether the clock alignment ever had to step mid-race - which is the
input for deciding if it needs slewing, rather than guessing. The mission
t0 tick goes in the log too, alongside the console's per-pod RunMission
send ticks, because nothing has ever measured how far apart the machines
actually start; the clockwork doors inherit that skew directly.

RP412NETSTATS adds the transport's own view - per connection: messages,
bytes, wire writes, partials, refusals, how much sat queued - and on
Steam the first read this codebase has ever taken of GetConnectionRealTime
Status. Ping, quality, pending and unacked bytes, and one route
description per connection at teardown. The API was vendored and never
called; there was no RTT number anywhere in the game.

Finally, rpl4opt -spoolstats reads any recording offline. The data was
already in every spool ever made and nothing read it that way: the
recorder restamps each packet with local arrival time while the update
records inside keep the sender's sim-grid stamp, so the difference is
clock offset plus one-way delay, and the same running-minimum estimator
the game runs live separates them. It prints delay above the per-host
minimum, and decomposes each entity's gaps into sender pacing versus
delivery jitter - which no live counter can do. It lives in the game exe
rather than RPL4TOOL because the tool is deliberately not /Zp1 and would
misread every struct in the file.

Verified on the two-pod loopback harness: mesh up, egg fed, 60s raced,
stopped on command, scores collected, and both summaries reading exactly
what a pair of PARKED pods should read - heartbeat cadence, one snap per
heartbeat, sub-quarter-metre corrections, no clock steps. The t0 ticks
and the netclock offsets agree with each other to the two seconds the
pods launched apart.

The latency tier is deliberately NOT here. TCP_NODELAY, the Steam
NoNagle flag, per-frame coalescing and the pre-sim receive drain are all
scoped and all wait on this build's numbers, because the point of
shipping measurement first is to find out whether the thing we would fix
is the thing that hurts. Nagle is still on. Interest management is still
inert. The wire format is untouched, so this build and the last one still
race each other.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-13 12:59:25 -05:00

527 lines
14 KiB
C++

#include "..\munga_l4\mungal4.h"
#pragma hdrstop
#include <stdio.h>
#include "..\munga\app.h"
#include "..\munga\network.h"
#include "..\munga\interest.h"
#include "..\munga\entity.h"
#include "..\munga\simulate.h"
//########################################################################
// rpl4opt -spoolstats <file.spl> - offline latency analysis of a spool
//
// Every Live Cam / pod recording already contains a complete one-way-
// delay dataset and nothing ever read it that way: the recorder restamps
// each packet's header with LOCAL ARRIVAL time (SPOOLER.cpp), while the
// UpdateRecords inside still carry the SENDER's sim-grid timestamp
// untouched. Per sender,
//
// sample = arrival - senderStamp = clockOffset + oneWayDelay
//
// and the running minimum of sample converges on the offset from above -
// the same estimator RP412NETCLOCK runs live (SIMULATE.cpp). Subtracting
// it back out gives one-way delay ABOVE THE MINIMUM per packet: jitter
// and queueing, which is what the dead reckoner actually fights. The
// minimum itself (the flat latency floor) is unobservable without a
// synchronized clock and is absorbed into the offset - every figure
// printed here is delay above that floor, not absolute delay.
//
// Also decomposed per entity: gaps on the SENDER timeline (send pacing)
// versus gaps on the ARRIVAL timeline (network + receive-loop jitter).
// The send side of the tick story lives in the first, the delivery side
// in the second, and no live counter can separate them.
//
// This must live in the game exe, not RPL4TOOL: the engine compiles
// /Zp1 and RPL4TOOL deliberately does not, so the tool would misread
// every struct in the file.
//
// Caveats printed with the output:
// - arrival stamps are quantized by the receive drain (post-render,
// so roughly frame cadence)
// - sender stamps sit on the sender's fixed-step grid (20 ms at 50 Hz)
// - a racer's spool never contains its own pod; a Live Cam hears all
//########################################################################
namespace
{
enum
{
statsMaxHosts = 32,
statsMaxEntities = 128,
statsWindow = 128, // mirrors netClockWindow (SIMULATE.cpp)
statsBuckets = 16
};
//
// log2-millisecond histogram, same shape as Entity::NetRaceStats -
// bucket 0 is sub-millisecond, bucket b covers [2^(b-1), 2^b) ms.
//
struct Histogram
{
unsigned long buckets[statsBuckets];
unsigned long total;
long worstMs;
void
Count(long ms)
{
if (ms > worstMs)
{
worstMs = ms;
}
unsigned long value = (ms > 0) ? (unsigned long) ms : 0;
int bucket = 0;
while (value != 0 && bucket < statsBuckets - 1)
{
value >>= 1;
++bucket;
}
buckets[bucket]++;
total++;
}
unsigned long
PercentileMs(unsigned long percent) const
{
if (total == 0)
{
return 0;
}
unsigned long target = (total * percent + 99) / 100;
unsigned long seen = 0;
for (int b = 0; b < statsBuckets; ++b)
{
seen += buckets[b];
if (seen >= target)
{
return (b == 0) ? 0 : (1UL << (b - 1));
}
}
return 1UL << (statsBuckets - 2);
}
};
//
// Per sender host: packet accounting and the netclock replica.
//
struct HostStats
{
HostID host;
Logical inUse;
unsigned long packetCount;
unsigned long updateMessageCount;
unsigned long byteCount;
Logical settled;
long offsetTicks;
long windowMinTicks;
int windowCount;
Histogram owdAboveMin;
};
//
// Per entity: one entry per replicated entity seen in the spool,
// with the sender-vs-arrival gap decomposition.
//
struct EntityStats
{
Logical inUse;
int entityHost; // EntityID host part
int entityLocal; // EntityID local part
HostID fromHost;
unsigned long updateCount;
long lastSenderTicks;
long lastArrivalTicks;
Logical haveLast;
Histogram senderGaps;
Histogram arrivalGaps;
};
HostStats gHosts[statsMaxHosts];
EntityStats gEntities[statsMaxEntities];
HostStats *
FindHost(HostID host)
{
HostStats *free_slot = NULL;
for (int i = 0; i < statsMaxHosts; ++i)
{
if (gHosts[i].inUse)
{
if (gHosts[i].host == host)
{
return &gHosts[i];
}
}
else if (free_slot == NULL)
{
free_slot = &gHosts[i];
}
}
if (free_slot != NULL)
{
memset(free_slot, 0, sizeof(*free_slot));
free_slot->inUse = True;
free_slot->host = host;
}
return free_slot;
}
EntityStats *
FindEntity(int entity_host, int entity_local, HostID from_host)
{
EntityStats *free_slot = NULL;
for (int i = 0; i < statsMaxEntities; ++i)
{
if (gEntities[i].inUse)
{
if (gEntities[i].entityHost == entity_host &&
gEntities[i].entityLocal == entity_local)
{
return &gEntities[i];
}
}
else if (free_slot == NULL)
{
free_slot = &gEntities[i];
}
}
if (free_slot != NULL)
{
memset(free_slot, 0, sizeof(*free_slot));
free_slot->inUse = True;
free_slot->entityHost = entity_host;
free_slot->entityLocal = entity_local;
free_slot->fromHost = from_host;
}
return free_slot;
}
//
// The netclock replica: returns delay-above-minimum for one sample,
// updating the host's running-minimum offset exactly the way the
// live estimator does (immediate adopt of a shorter path, window
// close follows drift).
//
long
OwdAboveMin(HostStats *host, long sample)
{
if (!host->settled)
{
host->settled = True;
host->offsetTicks = sample;
host->windowMinTicks = sample;
host->windowCount = 0;
return 0;
}
if (sample < host->windowMinTicks)
{
host->windowMinTicks = sample;
}
if (sample < host->offsetTicks)
{
host->offsetTicks = sample;
}
if (++host->windowCount >= statsWindow)
{
host->offsetTicks = host->windowMinTicks;
host->windowMinTicks = sample;
host->windowCount = 0;
}
long above = sample - host->offsetTicks;
return (above > 0) ? above : 0;
}
//
// Is there a plausible packet chain at this offset? The same check
// playback runs against a wrong egg, run over a few packets in a
// row - used to find where the host-pair table ends without needing
// the egg at all.
//
Logical
PacketChainPlausible(
const unsigned char *buffer,
size_t size,
size_t offset)
{
int checked = 0;
while (checked < 5)
{
if (offset == size)
{
return (checked > 0) ? True : False;
}
if (offset + sizeof(NetworkPacket) > size)
{
// a truncated tail is normal (the buffer filled mid-write)
return (checked > 0) ? True : False;
}
const NetworkPacket *packet = (const NetworkPacket *) (buffer + offset);
size_t length = packet->messageData.messageLength;
if (length < sizeof(Receiver::Message) || length > 65536)
{
return False;
}
if (offset + length + sizeof(NetworkPacketHeader) > size)
{
return (checked > 0) ? True : False;
}
offset += length + sizeof(NetworkPacketHeader);
++checked;
}
return True;
}
}
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
//
// Returns a process exit code: 0 analysis written, 1 something wrong.
//
int
RPL4SpoolStats_Run(const char *path)
{
printf("SpoolStats: %s\n", path);
FILE *file = fopen(path, "rb");
if (file == NULL)
{
printf("SpoolStats: cannot open the file\n");
return 1;
}
fseek(file, 0, SEEK_END);
long file_size = ftell(file);
fseek(file, 0, SEEK_SET);
if (file_size <= 0)
{
printf("SpoolStats: empty file\n");
fclose(file);
return 1;
}
unsigned char *buffer = (unsigned char *) malloc((size_t) file_size);
if (buffer == NULL ||
fread(buffer, 1, (size_t) file_size, file) != (size_t) file_size)
{
printf("SpoolStats: could not read %ld bytes\n", file_size);
fclose(file);
free(buffer);
return 1;
}
fclose(file);
size_t size = (size_t) file_size;
//
// Skip the header: ApplicationID + resource major revision, then one
// (Logical remote, HostID) pair per host named in the egg. The pair
// count is not recorded, so find it by trying each count and asking
// whether a plausible packet chain starts where the table would end.
//
size_t table_start = sizeof(ApplicationID) + sizeof(int);
size_t pair_size = sizeof(Logical) + sizeof(HostID);
size_t packets_start = 0;
int host_pairs = -1;
for (int try_count = 0; try_count <= 16; ++try_count)
{
size_t offset = table_start + (size_t) try_count * pair_size;
if (offset > size)
{
break;
}
//
// Each pair must read as a Logical 0/1 to be believable.
//
Logical pairs_believable = True;
for (int p = 0; p < try_count; ++p)
{
long remote = *(const long *) (buffer + table_start + (size_t) p * pair_size);
if (remote != 0 && remote != 1)
{
pairs_believable = False;
break;
}
}
if (pairs_believable && PacketChainPlausible(buffer, size, offset))
{
packets_start = offset;
host_pairs = try_count;
break;
}
}
if (host_pairs < 0)
{
printf("SpoolStats: no plausible packet chain found - not a spool?\n");
free(buffer);
return 1;
}
printf("SpoolStats: header holds %d host pair(s), %lu bytes of packets\n",
host_pairs, (unsigned long) (size - packets_start));
memset(gHosts, 0, sizeof(gHosts));
memset(gEntities, 0, sizeof(gEntities));
char csv_path[512];
_snprintf(csv_path, sizeof(csv_path) - 1, "%s.netstats.csv", path);
csv_path[sizeof(csv_path) - 1] = 0;
FILE *csv = fopen(csv_path, "w");
if (csv != NULL)
{
fprintf(csv, "fromHost,entityHost,entityLocal,senderMs,arrivalMs,owdAboveMinMs\n");
}
//
// The packet walk.
//
size_t offset = packets_start;
unsigned long packet_total = 0;
long first_arrival = 0;
long last_arrival = 0;
Logical have_arrival = False;
while (offset + sizeof(NetworkPacket) <= size)
{
NetworkPacket *packet = (NetworkPacket *) (buffer + offset);
size_t length = packet->messageData.messageLength;
if (length < sizeof(Receiver::Message) || length > 65536 ||
offset + length + sizeof(NetworkPacketHeader) > size)
{
break; // truncated tail (buffer filled mid-write)
}
long arrival = packet->timeStamp.ticks;
if (!have_arrival)
{
first_arrival = arrival;
have_arrival = True;
}
last_arrival = arrival;
HostStats *host = FindHost(packet->fromHost);
if (host != NULL)
{
host->packetCount++;
host->byteCount += (unsigned long) (length + sizeof(NetworkPacketHeader));
}
++packet_total;
//
// Entity update messages carry the sender sim-grid stamps.
//
if (host != NULL &&
packet->clientID == NetworkClient::InterestManagerClientID &&
packet->messageData.messageID ==
InterestManager::EntityUpdateReplicantsMessageID &&
length >= sizeof(Entity::Message) + sizeof(Simulation::UpdateRecord))
{
host->updateMessageCount++;
Entity::Message *entity_message =
(Entity::Message *) &packet->messageData;
//
// EntityID is (hostID, localID), both 32-bit under /Zp1;
// its members are private and the analyzer only needs the
// two numbers, so read them as the pair they are on disk.
//
int entity_id[2];
memcpy(entity_id, &entity_message->entityID, sizeof(entity_id));
//
// First record's stamp speaks for the message - every record
// in it was written at the same lastPerformance.
//
Simulation::UpdateRecord *record =
(Simulation::UpdateRecord *)
((unsigned char *) entity_message + sizeof(Entity::Message));
long sender_ticks = record->timeStamp.ticks;
long sample = arrival - sender_ticks;
long above = OwdAboveMin(host, sample);
host->owdAboveMin.Count(above);
EntityStats *entity =
FindEntity(entity_id[0], entity_id[1], packet->fromHost);
if (entity != NULL)
{
entity->updateCount++;
if (entity->haveLast)
{
entity->senderGaps.Count(sender_ticks - entity->lastSenderTicks);
entity->arrivalGaps.Count(arrival - entity->lastArrivalTicks);
}
entity->lastSenderTicks = sender_ticks;
entity->lastArrivalTicks = arrival;
entity->haveLast = True;
}
if (csv != NULL)
{
fprintf(csv, "%d,%d,%d,%ld,%ld,%ld\n",
(int) packet->fromHost, entity_id[0], entity_id[1],
sender_ticks, arrival, above);
}
}
offset += length + sizeof(NetworkPacketHeader);
}
//
// The report.
//
printf("SpoolStats: %lu packets spanning %.1f s of arrivals\n",
packet_total,
have_arrival ? (last_arrival - first_arrival) / 1000.0f : 0.0f);
printf("SpoolStats: figures are delay ABOVE the per-host minimum - the\n");
printf("SpoolStats: flat latency floor is absorbed into the clock offset\n");
printf("SpoolStats: and is unobservable without a synchronized clock.\n");
printf("SpoolStats: arrivals are quantized by the receive drain (~frame\n");
printf("SpoolStats: cadence); sender stamps sit on the 20 ms sim grid.\n");
int h;
for (h = 0; h < statsMaxHosts; ++h)
{
HostStats *host = &gHosts[h];
if (!host->inUse)
{
continue;
}
printf("SpoolStats: host %d: %lu packets %lu B, %lu update msgs, "
"owd-above-min med/p95 ~%lu/~%lu ms, worst %ld ms, offset %ld ms\n",
(int) host->host,
host->packetCount,
host->byteCount,
host->updateMessageCount,
host->owdAboveMin.PercentileMs(50),
host->owdAboveMin.PercentileMs(95),
host->owdAboveMin.worstMs,
host->offsetTicks);
}
for (int e = 0; e < statsMaxEntities; ++e)
{
EntityStats *entity = &gEntities[e];
if (!entity->inUse || entity->updateCount < 2)
{
continue;
}
printf("SpoolStats: entity %d:%d (host %d): %lu updates, "
"sender gaps med/p95 ~%lu/~%lu ms, arrival gaps med/p95 ~%lu/~%lu ms, "
"worst %ld/%ld ms\n",
entity->entityHost, entity->entityLocal, (int) entity->fromHost,
entity->updateCount,
entity->senderGaps.PercentileMs(50),
entity->senderGaps.PercentileMs(95),
entity->arrivalGaps.PercentileMs(50),
entity->arrivalGaps.PercentileMs(95),
entity->senderGaps.worstMs,
entity->arrivalGaps.worstMs);
}
if (csv != NULL)
{
fclose(csv);
printf("SpoolStats: per-update rows written to %s\n", csv_path);
}
free(buffer);
return 0;
}