Index MIDX
🗂️ INDEX_MIDX (RegionId.INDEX_MIDX, ID: 12)
Memory layout for the index entry slot table (48 bytes fixed size, v6).
Overview¶
| Property | Value |
|---|---|
| Bundle | runtime.bundle |
| Memory Shape | RecordMemory |
| Layout Class | IndexEntryLayout |
| Store Class | IndexEntryMemory |
| Record Stride | 48 bytes |
| Cache-Line Aligned | No (48B, 8B aligned) |
Wire Diagram (48 Bytes)¶
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| idPoolOffset (8B) |
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| idPoolLength (4B) | typeOrdinal (4B) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| offset (8B) |
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| graphSlot (4B) | textOffset (8B) ... |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| ... textOffset (cont'd) | textLength (4B) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| colocatedPartition (4B) | reserved (4B) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Field Specifications¶
| Offset | Size | Field Name | Type | Description |
|---|---|---|---|---|
0x00 | 8 | idPoolOffset | long | Offset of the ID string in IDPL |
0x08 | 4 | idPoolLength | int | Length of the ID string |
0x0C | 4 | typeOrdinal | int | Memory type ordinal |
0x10 | 8 | offset | long | Offset in the record region |
0x18 | 4 | graphSlot | int | Semantic-HNSW / Hebbian node slot |
0x1C | 8 | textOffset | long | Offset of the textual representation |
0x24 | 4 | textLength | int | Length of the textual representation |
0x28 | 4 | colocatedPartition | int | DISK partition the record lives in (v6) |
0x2C | 4 | reserved | int | Must be 0 (for 8-byte alignment) |
Access Patterns & Concurrency¶
Fixed-size contiguous records. Used in conjunction with IDPL to provide an index mapping memory UUIDs to internal structures.
Cognitive Role¶
Acts as the master directory of all memory records, allowing fast lookup by ID and providing structural offsets for text and graph representations.
Related¶
- [Issue #443 Phase 2 (v6 changes)]