Correlate infrastructure prelude classes by profile span

This commit is contained in:
Jan Petykiewicz 2026-04-18 14:42:01 -07:00
commit d998c08219
3 changed files with 387 additions and 0 deletions

View file

@ -1950,6 +1950,9 @@ pub struct SmpSavePlacedStructureDynamicSideBufferPayloadEnvelopeSummary {
#[serde(default)]
pub fixed_policy_summary: Option<SmpSavePlacedStructureDynamicSideBufferFixedPolicySummary>,
#[serde(default)]
pub name_prelude_candidate_summary:
Option<SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSummary>,
#[serde(default)]
pub sample_rows: Vec<SmpSavePlacedStructureDynamicSideBufferPayloadEnvelopeSample>,
}
@ -2049,6 +2052,70 @@ pub struct SmpSavePlacedStructureDynamicSideBufferPayloadEnvelopeSample {
pub profile_chunk_len_to_next_name_or_end: Option<usize>,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSummary {
pub row_count_with_candidate_window: usize,
pub unique_candidate_pattern_count: usize,
#[serde(default)]
pub dominant_child_count_candidate: Option<u16>,
pub dominant_child_count_candidate_count: usize,
#[serde(default)]
pub dominant_saved_primary_child_byte_candidate: Option<u8>,
#[serde(default)]
pub dominant_saved_primary_child_byte_candidate_hex: Option<String>,
pub dominant_saved_primary_child_byte_candidate_count: usize,
#[serde(default)]
pub dominant_candidate_pattern:
Option<SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidatePattern>,
#[serde(default)]
pub profile_span_correlations:
Vec<SmpSavePlacedStructureDynamicSideBufferNamePreludeProfileSpanCorrelation>,
#[serde(default)]
pub sample_rows: Vec<SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSample>,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidatePattern {
pub child_count_candidate: u16,
pub child_count_candidate_hex: String,
pub saved_primary_child_byte_candidate: u8,
pub saved_primary_child_byte_candidate_hex: String,
pub count: usize,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSample {
pub sample_index: usize,
pub name_tag_relative_offset: usize,
#[serde(default)]
pub primary_name: Option<String>,
#[serde(default)]
pub secondary_name: Option<String>,
pub child_count_candidate: u16,
pub child_count_candidate_hex: String,
pub saved_primary_child_byte_candidate: u8,
pub saved_primary_child_byte_candidate_hex: String,
#[serde(default)]
pub previous_profile_chunk_len_to_next_name_or_end: Option<usize>,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct SmpSavePlacedStructureDynamicSideBufferNamePreludeProfileSpanCorrelation {
pub previous_profile_chunk_len_to_next_name_or_end: usize,
pub row_count: usize,
#[serde(default)]
pub dominant_child_count_candidate: Option<u16>,
pub dominant_child_count_candidate_count: usize,
#[serde(default)]
pub dominant_saved_primary_child_byte_candidate: Option<u8>,
#[serde(default)]
pub dominant_saved_primary_child_byte_candidate_hex: Option<String>,
pub dominant_saved_primary_child_byte_candidate_count: usize,
#[serde(default)]
pub dominant_candidate_pattern:
Option<SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidatePattern>,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct SmpSavePlacedStructureDynamicSideBufferLiveEntryPreludeSummary {
pub live_entry_directory_row_count: usize,
@ -4247,6 +4314,72 @@ fn build_infrastructure_asset_trace_report(
.map(|summary| summary.dominant_trailing_word_count)
.unwrap_or_default()
),
side_buffer
.and_then(|probe| probe.payload_envelope_summary.as_ref())
.and_then(|summary| summary.name_prelude_candidate_summary.as_ref())
.and_then(|summary| {
summary
.profile_span_correlations
.iter()
.find(|row| row.previous_profile_chunk_len_to_next_name_or_end == 3)
})
.map(|correlation| {
format!(
"current save-side probe now also shows the short 0x03-byte post-profile gaps collapsing cleanly to the next-record prelude: dominant candidate pattern is {}/{} x{} across {} rows",
correlation
.dominant_candidate_pattern
.as_ref()
.map(|pattern| pattern.child_count_candidate_hex.as_str())
.unwrap_or("0x0000"),
correlation
.dominant_candidate_pattern
.as_ref()
.map(|pattern| pattern.saved_primary_child_byte_candidate_hex.as_str())
.unwrap_or("0x00"),
correlation
.dominant_candidate_pattern
.as_ref()
.map(|pattern| pattern.count)
.unwrap_or_default(),
correlation.row_count
)
})
.unwrap_or_else(|| {
"no grounded 0x03-byte post-profile span correlation was available for the prelude candidates".to_string()
}),
side_buffer
.and_then(|probe| probe.payload_envelope_summary.as_ref())
.and_then(|summary| summary.name_prelude_candidate_summary.as_ref())
.and_then(|summary| {
summary
.profile_span_correlations
.iter()
.find(|row| row.previous_profile_chunk_len_to_next_name_or_end == 0)
})
.map(|correlation| {
format!(
"the zero-length post-profile class is now a separate grounded outlier: dominant candidate pattern is {}/{} x{} across {} rows",
correlation
.dominant_candidate_pattern
.as_ref()
.map(|pattern| pattern.child_count_candidate_hex.as_str())
.unwrap_or("0x0000"),
correlation
.dominant_candidate_pattern
.as_ref()
.map(|pattern| pattern.saved_primary_child_byte_candidate_hex.as_str())
.unwrap_or("0x00"),
correlation
.dominant_candidate_pattern
.as_ref()
.map(|pattern| pattern.count)
.unwrap_or_default(),
correlation.row_count
)
})
.unwrap_or_else(|| {
"no grounded zero-length post-profile span correlation was available for the prelude candidates".to_string()
}),
"direct disassembly now also shows 0x530720 publishing the first fixed-triplet lane into [this+0x1e2/+0x1e6/+0x1ea], while 0x52e8b0 publishes the second fixed-triplet lane into [this+0x4b/+0x4f/+0x53] and sets bit 0x02".to_string(),
"direct disassembly now also shows the outer owner at 0x0048dcf0 reading one u16 child count through 0x531150 into the stream prelude, zeroing [this+0x08], and conditionally reading one saved primary-child byte before the per-child callback loop runs".to_string(),
side_buffer
@ -12789,6 +12922,224 @@ fn parse_save_placed_structure_dynamic_side_buffer_probe(
)
.collect(),
});
let name_prelude_candidate_rows = embedded_name_rows
.iter()
.enumerate()
.filter_map(|(row_index, row)| {
let candidate_offset = row.name_tag_relative_offset.checked_sub(3)?;
let child_count_candidate = read_u16_at(records_payload, candidate_offset)?;
let saved_primary_child_byte_candidate =
read_u8_at(records_payload, candidate_offset + 2)?;
Some((
row.name_tag_relative_offset,
row.primary_name.clone(),
row.secondary_name.clone(),
child_count_candidate,
saved_primary_child_byte_candidate,
row_index
.checked_sub(1)
.and_then(|previous_index| {
payload_envelope_rows
.get(previous_index)
.and_then(|row| row.profile_chunk_len_to_next_name_or_end)
}),
))
})
.collect::<Vec<_>>();
let mut name_prelude_candidate_pattern_counts = BTreeMap::<(u16, u8), usize>::new();
let mut name_prelude_child_count_counts = BTreeMap::<u16, usize>::new();
let mut name_prelude_saved_primary_counts = BTreeMap::<u8, usize>::new();
for (_, _, _, child_count_candidate, saved_primary_child_byte_candidate, _) in
&name_prelude_candidate_rows
{
*name_prelude_candidate_pattern_counts
.entry((*child_count_candidate, *saved_primary_child_byte_candidate))
.or_default() += 1;
*name_prelude_child_count_counts
.entry(*child_count_candidate)
.or_default() += 1;
*name_prelude_saved_primary_counts
.entry(*saved_primary_child_byte_candidate)
.or_default() += 1;
}
let dominant_name_prelude_candidate_pattern = name_prelude_candidate_pattern_counts
.iter()
.max_by(|(left_key, left_count), (right_key, right_count)| {
left_count
.cmp(right_count)
.then_with(|| right_key.cmp(left_key))
})
.map(
|((child_count_candidate, saved_primary_child_byte_candidate), count)| {
SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidatePattern {
child_count_candidate: *child_count_candidate,
child_count_candidate_hex: format!("0x{child_count_candidate:04x}"),
saved_primary_child_byte_candidate: *saved_primary_child_byte_candidate,
saved_primary_child_byte_candidate_hex: format!(
"0x{saved_primary_child_byte_candidate:02x}"
),
count: *count,
}
},
);
let dominant_name_prelude_child_count = name_prelude_child_count_counts
.iter()
.max_by(|(left_key, left_count), (right_key, right_count)| {
left_count
.cmp(right_count)
.then_with(|| right_key.cmp(left_key))
})
.map(|(value, count)| (*value, *count));
let dominant_name_prelude_saved_primary = name_prelude_saved_primary_counts
.iter()
.max_by(|(left_key, left_count), (right_key, right_count)| {
left_count
.cmp(right_count)
.then_with(|| right_key.cmp(left_key))
})
.map(|(value, count)| (*value, *count));
let mut name_prelude_profile_span_groups =
BTreeMap::<usize, Vec<(u16, u8)>>::new();
for (_, _, _, child_count_candidate, saved_primary_child_byte_candidate, previous_span) in
&name_prelude_candidate_rows
{
if let Some(previous_span) = previous_span {
name_prelude_profile_span_groups
.entry(*previous_span)
.or_default()
.push((*child_count_candidate, *saved_primary_child_byte_candidate));
}
}
let profile_span_correlations = name_prelude_profile_span_groups
.into_iter()
.map(|(previous_profile_chunk_len_to_next_name_or_end, rows)| {
let mut pattern_counts = BTreeMap::<(u16, u8), usize>::new();
let mut child_count_counts = BTreeMap::<u16, usize>::new();
let mut saved_primary_counts = BTreeMap::<u8, usize>::new();
for (child_count_candidate, saved_primary_child_byte_candidate) in &rows {
*pattern_counts
.entry((*child_count_candidate, *saved_primary_child_byte_candidate))
.or_default() += 1;
*child_count_counts.entry(*child_count_candidate).or_default() += 1;
*saved_primary_counts
.entry(*saved_primary_child_byte_candidate)
.or_default() += 1;
}
let dominant_pattern = pattern_counts
.iter()
.max_by(|(left_key, left_count), (right_key, right_count)| {
left_count
.cmp(right_count)
.then_with(|| right_key.cmp(left_key))
})
.map(
|((child_count_candidate, saved_primary_child_byte_candidate), count)| {
SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidatePattern {
child_count_candidate: *child_count_candidate,
child_count_candidate_hex: format!(
"0x{child_count_candidate:04x}"
),
saved_primary_child_byte_candidate:
*saved_primary_child_byte_candidate,
saved_primary_child_byte_candidate_hex: format!(
"0x{saved_primary_child_byte_candidate:02x}"
),
count: *count,
}
},
);
let dominant_child_count = child_count_counts
.iter()
.max_by(|(left_key, left_count), (right_key, right_count)| {
left_count
.cmp(right_count)
.then_with(|| right_key.cmp(left_key))
})
.map(|(value, count)| (*value, *count));
let dominant_saved_primary = saved_primary_counts
.iter()
.max_by(|(left_key, left_count), (right_key, right_count)| {
left_count
.cmp(right_count)
.then_with(|| right_key.cmp(left_key))
})
.map(|(value, count)| (*value, *count));
SmpSavePlacedStructureDynamicSideBufferNamePreludeProfileSpanCorrelation {
previous_profile_chunk_len_to_next_name_or_end,
row_count: rows.len(),
dominant_child_count_candidate: dominant_child_count.map(|(value, _)| value),
dominant_child_count_candidate_count: dominant_child_count
.map(|(_, count)| count)
.unwrap_or_default(),
dominant_saved_primary_child_byte_candidate: dominant_saved_primary
.map(|(value, _)| value),
dominant_saved_primary_child_byte_candidate_hex: dominant_saved_primary
.map(|(value, _)| format!("0x{value:02x}")),
dominant_saved_primary_child_byte_candidate_count: dominant_saved_primary
.map(|(_, count)| count)
.unwrap_or_default(),
dominant_candidate_pattern: dominant_pattern,
}
})
.take(8)
.collect::<Vec<_>>();
let name_prelude_candidate_summary =
Some(SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSummary {
row_count_with_candidate_window: name_prelude_candidate_rows.len(),
unique_candidate_pattern_count: name_prelude_candidate_pattern_counts.len(),
dominant_child_count_candidate: dominant_name_prelude_child_count
.map(|(value, _)| value),
dominant_child_count_candidate_count: dominant_name_prelude_child_count
.map(|(_, count)| count)
.unwrap_or_default(),
dominant_saved_primary_child_byte_candidate:
dominant_name_prelude_saved_primary.map(|(value, _)| value),
dominant_saved_primary_child_byte_candidate_hex:
dominant_name_prelude_saved_primary
.map(|(value, _)| format!("0x{value:02x}")),
dominant_saved_primary_child_byte_candidate_count:
dominant_name_prelude_saved_primary
.map(|(_, count)| count)
.unwrap_or_default(),
dominant_candidate_pattern: dominant_name_prelude_candidate_pattern.clone(),
profile_span_correlations,
sample_rows: name_prelude_candidate_rows
.iter()
.take(8)
.enumerate()
.map(
|(
sample_index,
(
name_tag_relative_offset,
primary_name,
secondary_name,
child_count_candidate,
saved_primary_child_byte_candidate,
previous_profile_chunk_len_to_next_name_or_end,
),
)| {
SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSample {
sample_index,
name_tag_relative_offset: *name_tag_relative_offset,
primary_name: primary_name.clone(),
secondary_name: secondary_name.clone(),
child_count_candidate: *child_count_candidate,
child_count_candidate_hex: format!(
"0x{child_count_candidate:04x}"
),
saved_primary_child_byte_candidate:
*saved_primary_child_byte_candidate,
saved_primary_child_byte_candidate_hex: format!(
"0x{saved_primary_child_byte_candidate:02x}"
),
previous_profile_chunk_len_to_next_name_or_end:
*previous_profile_chunk_len_to_next_name_or_end,
}
},
)
.collect(),
});
let payload_envelope_summary = Some(
SmpSavePlacedStructureDynamicSideBufferPayloadEnvelopeSummary {
row_count_with_policy_tag_before_next_name,
@ -12807,6 +13158,7 @@ fn parse_save_placed_structure_dynamic_side_buffer_probe(
.unwrap_or_default(),
short_profile_flag_pair_summary: short_profile_flag_pair_summary.clone(),
fixed_policy_summary: fixed_policy_summary.clone(),
name_prelude_candidate_summary: name_prelude_candidate_summary.clone(),
sample_rows: payload_envelope_rows
.iter()
.take(8)
@ -23290,6 +23642,30 @@ mod tests {
sample_rows: Vec::new(),
},
),
name_prelude_candidate_summary: Some(
SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSummary {
row_count_with_candidate_window: 118,
unique_candidate_pattern_count: 2,
dominant_child_count_candidate: Some(1),
dominant_child_count_candidate_count: 110,
dominant_saved_primary_child_byte_candidate: Some(0xff),
dominant_saved_primary_child_byte_candidate_hex: Some(
"0xff".to_string(),
),
dominant_saved_primary_child_byte_candidate_count: 110,
dominant_candidate_pattern: Some(
SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidatePattern {
child_count_candidate: 1,
child_count_candidate_hex: "0x0001".to_string(),
saved_primary_child_byte_candidate: 0xff,
saved_primary_child_byte_candidate_hex: "0xff".to_string(),
count: 110,
},
),
profile_span_correlations: Vec::new(),
sample_rows: Vec::new(),
},
),
sample_rows: Vec::new(),
},
),

View file

@ -2975,6 +2975,11 @@ The low helper strip beneath that shared family is tighter now too: `0x0052ecd0`
the save-side probe be looking for above the fixed rows?”; it is the narrower partitioning
problem of how the observed `0x55f3`-to-next-`0x55f1` gaps divide between the two `0x52ebd0`
flag bytes and the next records `u16 + byte` prelude.
The widened prelude correlation narrows that again on grounded `q.gms`: the `0x03` post-profile
gap class now collapses cleanly to the next-record prelude `0x0001 / 0xff` across `17/17` rows,
while the zero-length class is a separate `0x0055 / 0x00` outlier across `18/18` rows and the
`0x06` class is the only large mixed frontier left. So the next infrastructure pass should focus
on classifying the mixed `0x06` rows instead of re-proving the pure-prelude `0x03` class.
The child loader family is explicit now too: local `.rdata` at `0x005cfd00` proves the
`Infrastructure` child vtable uses the shared tagged callback strip directly, with
`+0x40 = 0x00455fc0`, `+0x48 = 0x00455870`, and `+0x4c = 0x00455930`. So the remaining

View file

@ -125,6 +125,12 @@ Working rule:
looking for above the fixed rows?”; it is the narrower partitioning problem of how the observed
`0x55f3`-to-next-`0x55f1` gaps divide between the two `0x52ebd0` flag bytes and the next
records `u16 + byte` prelude.
- The widened prelude correlation closes part of that partitioning too: grounded `q.gms` rows with
a `0x03` post-profile gap now collapse cleanly to the next-record prelude pattern
`0x0001 / 0xff` across `17/17` rows, while the zero-length class is a separate grounded outlier
with dominant pattern `0x0055 / 0x00` across `18/18` rows and the `0x06` class remains the only
large mixed frontier. So the next infrastructure slice should focus on classifying the mixed
`0x06` rows, not on rediscovering the already-grounded pure-prelude `0x03` rows.
- Reconstruct the save-side region record body on top of the newly corrected non-direct tagged
region seam (`0x5209/0x520a/0x520b`, stride hint `0x06`, `Marker09` record stems) now that the
`0x55f3` payload is known to be fully consumed by the embedded profile collection on grounded