Fixed Ubi alt IMA decoding [Splinter Cell: Essentials (PSP)]

This commit is contained in:
NicknineTheEagle 2020-05-27 23:37:25 +03:00
parent 1a0913d90d
commit b0b7b20bf4
2 changed files with 17 additions and 11 deletions

View File

@ -11,7 +11,7 @@
* - expand type: IMA style or variations; low or high nibble first
*/
static const int ADPCMTable[89] = {
static const int ADPCMTable[90] = {
7, 8, 9, 10, 11, 12, 13, 14,
16, 17, 19, 21, 23, 25, 28, 31,
34, 37, 41, 45, 50, 55, 60, 66,
@ -23,7 +23,9 @@ static const int ADPCMTable[89] = {
3327, 3660, 4026, 4428, 4871, 5358, 5894, 6484,
7132, 7845, 8630, 9493, 10442, 11487, 12635, 13899,
15289, 16818, 18500, 20350, 22385, 24623, 27086, 29794,
32767
32767,
0 /* garbage value for Ubisoft IMA (see blocked_ubi_sce.c) */
};
static const int IMA_IndexTable[16] = {
@ -1054,10 +1056,14 @@ void decode_ubi_ima(VGMSTREAMCHANNEL * stream, sample_t * outbuf, int channelspa
if (has_header) {
first_sample -= 10; //todo fix hack (needed to adjust nibble offset below)
}
if (step_index < 0) step_index=0;
if (step_index > 88) step_index=88;
if (step_index < 0) step_index = 0;
if (step_index > 88) step_index = 88;
} else {
if (step_index < 0) step_index = 0;
if (step_index > 89) step_index = 89;
}
for (i = first_sample; i < first_sample + samples_to_do; i++, sample_count += channelspacing) {
off_t byte_offset = channelspacing == 1 ?

View File

@ -44,13 +44,13 @@ void block_update_ubi_sce(off_t block_offset, VGMSTREAM* vgmstream) {
vgmstream->ch[i].adpcm_step_index = read_32bitLE(block_offset + header_size * i + 0x04, sf);
vgmstream->ch[i].adpcm_history1_32 = read_32bitLE(block_offset + header_size * i + 0x08, sf);
/* TODO figure out
* First step seems to always be a special value for the decoder, unsure of meaning.
* 0 = too quiet and max = 88 = waveform starts a bit off and clicky. First hist is usually +-1,
* other frames look, fine not sure what are they aiming for.
*/
/* First step is always 0x500, not sure if it's a bug or a feature but the game just takes it as is and
* ends up reading 0 from out-of-bounds memory area which causes a pop at the start. Yikes.
* It gets clampled later so the rest of the sound plays ok.
* We put 89 here as our special index which contains 0 to simulate this.
*/
if (vgmstream->ch[i].adpcm_step_index == 0x500) {
vgmstream->ch[i].adpcm_step_index = 88;
vgmstream->ch[i].adpcm_step_index = 89;
}
}
}