1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
|
/*
* PSn00bSDK GPU library (image and VRAM transfer functions)
* (C) 2022-2023 spicyjpeg - MPL licensed
*
* TODO: MoveImage() is currently commented out as it won't trigger a DMA IRQ,
* making it unusable as a draw queue command. A way around this (perhaps using
* the GPU IRQ?) shall be found.
*/
#include <stdint.h>
#include <assert.h>
#include <psxetc.h>
#include <psxgpu.h>
#include <hwregs_c.h>
#define QUEUE_LENGTH 16
#define DMA_CHUNK_LENGTH 16
/* Internal globals */
// LoadImage() and StoreImage() run asynchronously but may be called with a
// pointer to a RECT struct in the stack, which might no longer be valid by the
// time the transfer is actually started. This buffer is used to store a copy
// of all RECTs passed to LoadImage()/StoreImage() as a workaround.
static RECT _saved_rects[QUEUE_LENGTH];
static int _next_saved_rect = 0;
/* Private utilities */
static void _dma_transfer(const RECT *rect, uint32_t *data, int write) {
size_t length = rect->w * rect->h;
if (length % 2)
_sdk_log("can't transfer an odd number of pixels\n");
length /= 2;
if ((length >= DMA_CHUNK_LENGTH) && (length % DMA_CHUNK_LENGTH)) {
_sdk_log("transfer data length (%d) is not a multiple of %d, rounding\n", length, DMA_CHUNK_LENGTH);
length += DMA_CHUNK_LENGTH - 1;
}
while (!(GPU_GP1 & (1 << 26)))
__asm__ volatile("");
GPU_GP1 = 0x04000000; // Disable DMA request
GPU_GP0 = 0x01000000; // Flush cache
GPU_GP0 = write ? 0xa0000000 : 0xc0000000;
//GPU_GP0 = rect->x | (rect->y << 16);
GPU_GP0 = *((const uint32_t *) &(rect->x));
//GPU_GP0 = rect->w | (rect->h << 16);
GPU_GP0 = *((const uint32_t *) &(rect->w));
// Enable DMA request, route to GP0 (2) or from GPU_READ (3)
GPU_GP1 = 0x04000002 | (write ^ 1);
while (DMA_CHCR(DMA_GPU) & (1 << 24))
__asm__ volatile("");
DMA_MADR(DMA_GPU) = (uint32_t) data;
if (length < DMA_CHUNK_LENGTH)
DMA_BCR(DMA_GPU) = 0x00010000 | length;
else
DMA_BCR(DMA_GPU) = DMA_CHUNK_LENGTH |
((length / DMA_CHUNK_LENGTH) << 16);
DMA_CHCR(DMA_GPU) = 0x01000200 | write;
}
/* VRAM transfer API */
int LoadImage(const RECT *rect, const uint32_t *data) {
int index = _next_saved_rect;
_saved_rects[index] = *rect;
_next_saved_rect = (index + 1) % QUEUE_LENGTH;
return EnqueueDrawOp(
(void *) &_dma_transfer,
(uint32_t) &_saved_rects[index],
(uint32_t) data,
1
);
}
int StoreImage(const RECT *rect, uint32_t *data) {
int index = _next_saved_rect;
_saved_rects[index] = *rect;
_next_saved_rect = (index + 1) % QUEUE_LENGTH;
return EnqueueDrawOp(
(void *) &_dma_transfer,
(uint32_t) &_saved_rects[index],
(uint32_t) data,
0
);
}
int MoveImage(const RECT *rect, int x, int y) {
return EnqueueDrawOp((void *) &MoveImage2, (uint32_t) rect, x, y);
}
void LoadImage2(const RECT *rect, const uint32_t *data) {
_dma_transfer(rect, (uint32_t *) data, 1);
}
void StoreImage2(const RECT *rect, uint32_t *data) {
_dma_transfer(rect, data, 0);
}
void MoveImage2(const RECT *rect, int x, int y) {
while (!(GPU_GP1 & (1 << 26)))
__asm__ volatile("");
GPU_GP0 = 0x80000000;
//GPU_GP0 = rect->x | (rect->y << 16);
GPU_GP0 = *((const uint32_t *) &(rect->x));
GPU_GP0 = (x & 0xffff) | (y << 16);
//GPU_GP0 = rect->w | (rect->h << 16);
GPU_GP0 = *((const uint32_t *) &(rect->w));
// As no DMA transfer is performed by this command, the GPU IRQ is used
// instead of the DMA IRQ to trigger the draw queue callback.
GPU_GP0 = 0x1f000000;
}
/* .TIM image parsers */
// This is the only libgs function PSn00bSDK is ever going to implement. The
// difference from GetTimInfo() is that it copies RECTs rather than merely
// returning pointers to them, which become useless once the .TIM file is
// unloaded from main RAM.
int GsGetTimInfo(const uint32_t *tim, GsIMAGE *info) {
if ((*(tim++) & 0xffff) != 0x0010)
return 1;
info->pmode = *(tim++);
if (info->pmode & 8) {
const uint32_t *palette_end = tim;
palette_end += *(tim++) / 4;
*((uint32_t *) &(info->cx)) = *(tim++);
*((uint32_t *) &(info->cw)) = *(tim++);
info->clut = (uint32_t *) tim;
tim = palette_end;
} else {
info->clut = 0;
}
tim++;
*((uint32_t *) &(info->px)) = *(tim++);
*((uint32_t *) &(info->pw)) = *(tim++);
info->pixel = (uint32_t *) tim;
return 0;
}
int GetTimInfo(const uint32_t *tim, TIM_IMAGE *info) {
if ((*(tim++) & 0xffff) != 0x0010)
return 1;
info->mode = *(tim++);
if (info->mode & 8) {
const uint32_t *palette_end = tim;
palette_end += *(tim++) / 4;
info->crect = (RECT *) tim;
info->caddr = (uint32_t *) &tim[2];
tim = palette_end;
} else {
info->caddr = 0;
}
tim++;
info->prect = (RECT *) tim;
info->paddr = (uint32_t *) &tim[2];
return 0;
}
|
