summaryrefslogtreecommitdiff
path: root/src/gallium/drivers/nvfx/nvfx_transfer.c
blob: 7cb47a20f645818ecb046b24a258051b3e312c4f (plain)
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
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
#include "pipe/p_state.h"
#include "pipe/p_defines.h"
#include "util/u_inlines.h"
#include "util/u_format.h"
#include "util/u_memory.h"
#include "util/u_math.h"
#include "util/u_staging.h"
#include "nvfx_context.h"
#include "nvfx_screen.h"
#include "nvfx_state.h"
#include "nvfx_resource.h"
#include "nvfx_transfer.h"

struct nvfx_staging_transfer
{
	struct util_staging_transfer base;

	unsigned offset;
	unsigned map_count;
};

struct pipe_transfer *
nvfx_transfer_new(struct pipe_context *pipe,
			  struct pipe_resource *pt,
			  struct pipe_subresource sr,
			  unsigned usage,
			  const struct pipe_box *box)
{
        if((usage & (PIPE_TRANSFER_UNSYNCHRONIZED | PIPE_TRANSFER_DONTBLOCK)) == PIPE_TRANSFER_DONTBLOCK)
        {
                struct nouveau_bo* bo = ((struct nvfx_resource*)pt)->bo;
                if(bo && nouveau_bo_busy(bo, NOUVEAU_BO_WR))
                        return NULL;
        }

	if(pt->target == PIPE_BUFFER)
	{
		// it would be nice if we could avoid all this ridiculous overhead...
		struct pipe_transfer* tx;
		struct nvfx_buffer* buffer = nvfx_buffer(pt);

		tx = CALLOC_STRUCT(pipe_transfer);
		if (!tx)
			return NULL;

		pipe_resource_reference(&tx->resource, pt);
		tx->sr = sr;
		tx->usage = usage;
		tx->box = *box;

		tx->slice_stride = tx->stride = util_format_get_stride(pt->format, box->width);
		tx->data = buffer->data + util_format_get_stride(pt->format, box->x);

		return tx;
	}
	else
	{
	        struct nvfx_staging_transfer* tx;
	        bool direct = !nvfx_resource_on_gpu(pt) && pt->flags & NVFX_RESOURCE_FLAG_LINEAR;

	        tx = CALLOC_STRUCT(nvfx_staging_transfer);
	        if(!tx)
	        	return NULL;

	        util_staging_transfer_init(pipe, pt, sr, usage, box, direct, &tx->base);

		if(direct)
		{
			tx->base.base.stride = nvfx_subresource_pitch(pt, sr.level);
			tx->base.base.slice_stride = tx->base.base.stride * u_minify(pt->height0, sr.level);
			tx->offset = nvfx_subresource_offset(pt, sr.face, sr.level, box->z)
				+ util_format_get_2d_size(pt->format, tx->base.base.stride, box->y)
				+ util_format_get_stride(pt->format, box->x);
		}
		else
		{
			tx->base.base.stride = nvfx_subresource_pitch(tx->base.staging_resource, 0);
			tx->base.base.slice_stride = tx->base.base.stride * tx->base.staging_resource->height0;
			tx->offset = 0;
		}

		assert(tx->base.base.stride);

		return &tx->base.base;
	}
}

static void nvfx_buffer_dirty_interval(struct nvfx_buffer* buffer, unsigned begin, unsigned size, boolean unsynchronized)
{
	struct nvfx_screen* screen = nvfx_screen(buffer->base.base.screen);
	buffer->last_update_static = buffer->bytes_to_draw_until_static < 0;
	if(buffer->dirty_begin == buffer->dirty_end)
	{
		buffer->dirty_begin = begin;
		buffer->dirty_end = begin + size;
		buffer->dirty_unsynchronized = unsynchronized;
	}
	else
	{
		buffer->dirty_begin = MIN2(buffer->dirty_begin, begin);
		buffer->dirty_end = MAX2(buffer->dirty_end, begin + size);
		buffer->dirty_unsynchronized &= unsynchronized;
	}

	if(unsynchronized)
	{
		// TODO: revisit this, it doesn't seem quite right
		//printf("UNSYNC UPDATE %p %u %u\n", buffer, begin, size);
		buffer->bytes_to_draw_until_static += size * screen->static_reuse_threshold;
	}
	else
		buffer->bytes_to_draw_until_static = buffer->size * screen->static_reuse_threshold;
}

static void nvfx_transfer_flush_region( struct pipe_context *pipe,
				      struct pipe_transfer *ptx,
				      const struct pipe_box *box)
{
	if(ptx->resource->target == PIPE_BUFFER && (ptx->usage & PIPE_TRANSFER_FLUSH_EXPLICIT))
	{
		struct nvfx_buffer* buffer = nvfx_buffer(ptx->resource);
		nvfx_buffer_dirty_interval(buffer,
				(uint8_t*)ptx->data - buffer->data + util_format_get_stride(buffer->base.base.format, box->x),
				util_format_get_stride(buffer->base.base.format, box->width),
				!!(ptx->usage & PIPE_TRANSFER_UNSYNCHRONIZED));
	}
}

static void
nvfx_transfer_destroy(struct pipe_context *pipe, struct pipe_transfer *ptx)
{
	if(ptx->resource->target == PIPE_BUFFER)
	{
		struct nvfx_buffer* buffer = nvfx_buffer(ptx->resource);
		if((ptx->usage & (PIPE_TRANSFER_WRITE | PIPE_TRANSFER_FLUSH_EXPLICIT)) == PIPE_TRANSFER_WRITE)
			nvfx_buffer_dirty_interval(buffer,
				(uint8_t*)ptx->data - buffer->data,
				ptx->stride,
				!!(ptx->usage & PIPE_TRANSFER_UNSYNCHRONIZED));
		pipe_resource_reference(&ptx->resource, 0);
		FREE(ptx);
	}
	else
	{
		struct nouveau_channel* chan = nvfx_context(pipe)->screen->base.channel;
		util_staging_transfer_destroy(pipe, ptx);

		FIRE_RING(chan);
	}
}

void *
nvfx_transfer_map(struct pipe_context *pipe, struct pipe_transfer *ptx)
{
	if(ptx->resource->target == PIPE_BUFFER)
		return ptx->data;
	else
	{
		struct nvfx_staging_transfer *tx = (struct nvfx_staging_transfer *)ptx;
		if(!ptx->data)
		{
			struct nvfx_miptree *mt = (struct nvfx_miptree *)tx->base.staging_resource;
			uint8_t *map = nouveau_screen_bo_map(pipe->screen, mt->base.bo, nouveau_screen_transfer_flags(ptx->usage));
			ptx->data = map + tx->offset;
		}

		++tx->map_count;
		return ptx->data;
	}
}

void
nvfx_transfer_unmap(struct pipe_context *pipe, struct pipe_transfer *ptx)
{
	if(ptx->resource->target != PIPE_BUFFER)
	{
		struct nvfx_staging_transfer *tx = (struct nvfx_staging_transfer *)ptx;
		struct nvfx_miptree *mt = (struct nvfx_miptree *)tx->base.staging_resource;

		if(!--tx->map_count)
		{
			nouveau_screen_bo_unmap(pipe->screen, mt->base.bo);
			ptx->data = 0;
		}
	}
}

static void nvfx_transfer_inline_write( struct pipe_context *pipe,
				      struct pipe_resource *pr,
				      struct pipe_subresource sr,
				      unsigned usage,
				      const struct pipe_box *box,
				      const void *data,
				      unsigned stride,
				      unsigned slice_stride)
{
	if(pr->target != PIPE_BUFFER)
	{
		u_default_transfer_inline_write(pipe, pr, sr, usage, box, data, stride, slice_stride);
	}
	else
	{
		struct nvfx_buffer* buffer = nvfx_buffer(pr);
		unsigned begin = util_format_get_stride(pr->format, box->x);
		unsigned size = util_format_get_stride(pr->format, box->width);
		memcpy(buffer->data + begin, data, size);
		nvfx_buffer_dirty_interval(buffer, begin, size,
				!!(pr->flags & PIPE_TRANSFER_UNSYNCHRONIZED));
	}
}

void
nvfx_init_transfer_functions(struct pipe_context *pipe)
{
	pipe->get_transfer = nvfx_transfer_new;
	pipe->transfer_map = nvfx_transfer_map;
	pipe->transfer_flush_region = nvfx_transfer_flush_region;
	pipe->transfer_unmap = nvfx_transfer_unmap;
	pipe->transfer_destroy = nvfx_transfer_destroy;
	pipe->transfer_inline_write = nvfx_transfer_inline_write;
}