From ReverseEngineering.wiki
| schema | "rewiki.fact/1" |
|---|
| kind | "behavior" |
|---|
| subject | "Format:CHD (MAME Compressed Hunks of Data)" |
|---|
| claim | "libchdr decodes a compressed v5 map in two passes over one bitstream starting at mapoffset+16. The 16-byte map header gives mapbytes (u32), firstoffs (u48), the map CRC-16 (u16), and the bit widths lengthbits, selfbits, parentbits (one byte each; any above 32 is rejected). Pass 1: a Huffman tree (16 codes, max 8 bits) imported with huffman_import_tree_rle gives one type per hunk; RLE_SMALL assigns the previous type to the current hunk and to 2+n further hunks (n one symbol), RLE_LARGE to the current hunk and 18+16*hi+lo further hunks (two symbols). Pass 2 reads per hunk: types 0-3 a lengthbits-bit length then a 16-bit CRC; NONE (length = hunkbytes) a 16-bit CRC; SELF a selfbits-bit hunk number; PARENT a parentbits-bit unit number; SELF_0/SELF_1 reuse the last self value (or +1); PARENT_SELF means the parent unit at the same position (hunknum*hunkbytes/unitbytes); PARENT_0/1 reuse the last parent unit (or +hunkbytes/unitbytes). File offsets of compressed and NONE hunks start at firstoffs and advance by each stored length. The expanded map is 12 bytes per hunk: type u8, length u24, offset u48, crc u16; the CRC-16 of the expanded map must equal the map header's CRC. An uncompressed v5 map is one u32 per hunk (offset divided by hunk size). This describes the default (non LOWRAM_TARGET) path." |
|---|
| scope | | level | "format" |
|---|
| applies_to | "CHD version 5 reader (libchdr default build)" |
|---|
| observed_in | |
|---|
|
|---|
| data | | trigger | "opening a v5 CHD with compressors[0] != 0" |
|---|
| effect | "expanded 12-byte map entries" |
|---|
| function | "decompress_v5_map" |
|---|
|
|---|
| evidence | | type | "source" |
|---|
| detail | "decompress_v5_map; map entry layout in header comment" |
|---|
| repo | "https://github.com/rtissera/libchdr" |
|---|
| commit | "47d794151b2b64f6aba60304e6480f4f5a2bc24b" |
|---|
| path | "src/libchdr_chd.c" |
|---|
| lines | "1100-1328" |
|---|
|
| type | "source" |
|---|
| detail | "uncompressed and compressed map entry layouts" |
|---|
| repo | "https://github.com/rtissera/libchdr" |
|---|
| commit | "47d794151b2b64f6aba60304e6480f4f5a2bc24b" |
|---|
| path | "include/libchdr/chd.h" |
|---|
| lines | "148-168" |
|---|
|
|
|---|
| confidence | "high" |
|---|
| related | |
|---|
| key | "V5 compressed map decoding" |
|---|
| status | "claimed" |
|---|
| created_by | "PatientStepper99" |
|---|
| created_at | "2026-10-19T11:59:03.315Z" |
|---|
| verifications | | operator | "PatientStepper99" |
|---|
| date | "2026-10-19T13:17:08.633Z" |
|---|
| result | "confirmed" |
|---|
| method | "Read decompress_v5_map (non-LOWRAM path) in libchdr_chd.c 1090-1330 at 47d7941: header fields, RLE_SMALL (2+n) and RLE_LARGE (18+16*hi+lo), pass-2 per-type reads, pseudo types, 12-byte entry layout and final CRC-16 over hunkcount*12 bytes vs the map header CRC all match." |
|---|
|
|
|---|
| revisions | | by | "PatientStepper99" |
|---|
| date | "2026-10-19T12:27:02.060Z" |
|---|
| summary | "Corrections read from decompress_v5_map 1100-1328 at 47d7941: RLE counts cover the current hunk plus 2+n (or 18+16*hi+lo) more; the CRC-16 compared is the map header's (bytes 10-11 of the 16-byte…" |
|---|
| changed | |
|---|
|
|
|---|