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
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
|
From 705f105de150240594945703df70f82d5ab861ce Mon Sep 17 00:00:00 2001
From: Sanjay R Mehta <sanju.mehta@amd.com>
Date: Tue, 6 Dec 2016 19:33:01 +0530
Subject: [PATCH 05/10] implement raster configuration for gfx v8
This patch is to implement the raster configuration and harvested
configuration of gfx v8.
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Signed-off-by: Sanjay R Mehta <sanju.mehta@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 168 +++++++++++++++++++++++++++++++++-
drivers/gpu/drm/amd/amdgpu/vid.h | 37 ++++++++
2 files changed, 204 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index c5a3d04..20ac07f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -3484,13 +3484,163 @@ static u32 gfx_v8_0_get_rb_active_bitmap(struct amdgpu_device *adev)
return (~data) & mask;
}
+static void
+gfx_v8_0_raster_config(struct amdgpu_device *adev, u32 *rconf, u32 *rconf1)
+{
+ switch (adev->asic_type) {
+ case CHIP_FIJI:
+ *rconf |= RB_MAP_PKR0(2) | RB_MAP_PKR1(2) |
+ RB_XSEL2(1) | PKR_MAP(2) |
+ PKR_XSEL(1) | PKR_YSEL(1) |
+ SE_MAP(2) | SE_XSEL(2) | SE_YSEL(3);
+ *rconf1 |= SE_PAIR_MAP(2) | SE_PAIR_XSEL(3) |
+ SE_PAIR_YSEL(2);
+ break;
+ case CHIP_TONGA:
+ case CHIP_POLARIS10:
+ *rconf |= RB_MAP_PKR0(2) | RB_XSEL2(1) | SE_MAP(2) |
+ SE_XSEL(1) | SE_YSEL(1);
+ *rconf1 |= SE_PAIR_MAP(2) | SE_PAIR_XSEL(2) |
+ SE_PAIR_YSEL(2);
+ break;
+ case CHIP_TOPAZ:
+ case CHIP_CARRIZO:
+ *rconf |= RB_MAP_PKR0(2);
+ *rconf1 |= 0x0;
+ break;
+ case CHIP_POLARIS11:
+ *rconf |= RB_MAP_PKR0(2) | RB_XSEL2(1) | SE_MAP(2) |
+ SE_XSEL(1) | SE_YSEL(1);
+ *rconf1 |= 0x0;
+ break;
+ case CHIP_STONEY:
+ *rconf |= 0x0;
+ *rconf1 |= 0x0;
+ break;
+ default:
+ DRM_ERROR("unknown asic: 0x%x\n", adev->asic_type);
+ break;
+ }
+}
+
+static void
+gfx_v8_0_write_harvested_raster_configs(struct amdgpu_device *adev,
+ u32 raster_config, u32 raster_config_1,
+ unsigned rb_mask, unsigned num_rb)
+{
+ unsigned sh_per_se = max_t(unsigned, adev->gfx.config.max_sh_per_se, 1);
+ unsigned num_se = max_t(unsigned, adev->gfx.config.max_shader_engines, 1);
+ unsigned rb_per_pkr = min_t(unsigned, num_rb / num_se / sh_per_se, 2);
+ unsigned rb_per_se = num_rb / num_se;
+ unsigned se_mask[4];
+ unsigned se;
+
+ se_mask[0] = ((1 << rb_per_se) - 1) & rb_mask;
+ se_mask[1] = (se_mask[0] << rb_per_se) & rb_mask;
+ se_mask[2] = (se_mask[1] << rb_per_se) & rb_mask;
+ se_mask[3] = (se_mask[2] << rb_per_se) & rb_mask;
+
+ WARN_ON(!(num_se == 1 || num_se == 2 || num_se == 4));
+ WARN_ON(!(sh_per_se == 1 || sh_per_se == 2));
+ WARN_ON(!(rb_per_pkr == 1 || rb_per_pkr == 2));
+
+ if ((num_se > 2) && ((!se_mask[0] && !se_mask[1]) ||
+ (!se_mask[2] && !se_mask[3]))) {
+ raster_config_1 &= ~SE_PAIR_MAP_MASK;
+
+ if (!se_mask[0] && !se_mask[1]) {
+ raster_config_1 |=
+ SE_PAIR_MAP(RASTER_CONFIG_SE_PAIR_MAP_3);
+ } else {
+ raster_config_1 |=
+ SE_PAIR_MAP(RASTER_CONFIG_SE_PAIR_MAP_0);
+ }
+ }
+
+ for (se = 0; se < num_se; se++) {
+ unsigned raster_config_se = raster_config;
+ unsigned pkr0_mask = ((1 << rb_per_pkr) - 1) << (se * rb_per_se);
+ unsigned pkr1_mask = pkr0_mask << rb_per_pkr;
+ int idx = (se / 2) * 2;
+
+ if ((num_se > 1) && (!se_mask[idx] || !se_mask[idx + 1])) {
+ raster_config_se &= ~SE_MAP_MASK;
+
+ if (!se_mask[idx]) {
+ raster_config_se |= SE_MAP(RASTER_CONFIG_SE_MAP_3);
+ } else {
+ raster_config_se |= SE_MAP(RASTER_CONFIG_SE_MAP_0);
+ }
+ }
+
+ pkr0_mask &= rb_mask;
+ pkr1_mask &= rb_mask;
+ if (rb_per_se > 2 && (!pkr0_mask || !pkr1_mask)) {
+ raster_config_se &= ~PKR_MAP_MASK;
+
+ if (!pkr0_mask) {
+ raster_config_se |= PKR_MAP(RASTER_CONFIG_PKR_MAP_3);
+ } else {
+ raster_config_se |= PKR_MAP(RASTER_CONFIG_PKR_MAP_0);
+ }
+ }
+
+ if (rb_per_se >= 2) {
+ unsigned rb0_mask = 1 << (se * rb_per_se);
+ unsigned rb1_mask = rb0_mask << 1;
+
+ rb0_mask &= rb_mask;
+ rb1_mask &= rb_mask;
+ if (!rb0_mask || !rb1_mask) {
+ raster_config_se &= ~RB_MAP_PKR0_MASK;
+
+ if (!rb0_mask) {
+ raster_config_se |=
+ RB_MAP_PKR0(RASTER_CONFIG_RB_MAP_3);
+ } else {
+ raster_config_se |=
+ RB_MAP_PKR0(RASTER_CONFIG_RB_MAP_0);
+ }
+ }
+
+ if (rb_per_se > 2) {
+ rb0_mask = 1 << (se * rb_per_se + rb_per_pkr);
+ rb1_mask = rb0_mask << 1;
+ rb0_mask &= rb_mask;
+ rb1_mask &= rb_mask;
+ if (!rb0_mask || !rb1_mask) {
+ raster_config_se &= ~RB_MAP_PKR1_MASK;
+
+ if (!rb0_mask) {
+ raster_config_se |=
+ RB_MAP_PKR1(RASTER_CONFIG_RB_MAP_3);
+ } else {
+ raster_config_se |=
+ RB_MAP_PKR1(RASTER_CONFIG_RB_MAP_0);
+ }
+ }
+ }
+ }
+
+ /* GRBM_GFX_INDEX has a different offset on VI */
+ gfx_v8_0_select_se_sh(adev, se, 0xffffffff);
+ WREG32(mmPA_SC_RASTER_CONFIG, raster_config_se);
+ WREG32(mmPA_SC_RASTER_CONFIG_1, raster_config_1);
+ }
+
+ /* GRBM_GFX_INDEX has a different offset on VI */
+ gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+}
+
static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
{
int i, j;
u32 data;
+ u32 raster_config = 0, raster_config_1 = 0;
u32 active_rbs = 0;
u32 rb_bitmap_width_per_sh = adev->gfx.config.max_backends_per_se /
adev->gfx.config.max_sh_per_se;
+ unsigned num_rb_pipes;
mutex_lock(&adev->grbm_idx_mutex);
for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
@@ -3502,10 +3652,26 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
}
}
gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
- mutex_unlock(&adev->grbm_idx_mutex);
adev->gfx.config.backend_enable_mask = active_rbs;
adev->gfx.config.num_rbs = hweight32(active_rbs);
+
+ num_rb_pipes = min_t(unsigned, adev->gfx.config.max_backends_per_se *
+ adev->gfx.config.max_shader_engines, 16);
+
+ gfx_v8_0_raster_config(adev, &raster_config, &raster_config_1);
+
+ if (!adev->gfx.config.backend_enable_mask ||
+ adev->gfx.config.num_rbs >= num_rb_pipes) {
+ WREG32(mmPA_SC_RASTER_CONFIG, raster_config);
+ WREG32(mmPA_SC_RASTER_CONFIG_1, raster_config_1);
+ } else {
+ gfx_v8_0_write_harvested_raster_configs(adev, raster_config, raster_config_1,
+ adev->gfx.config.backend_enable_mask,
+ num_rb_pipes);
+ }
+
+ mutex_unlock(&adev->grbm_idx_mutex);
}
/**
diff --git a/drivers/gpu/drm/amd/amdgpu/vid.h b/drivers/gpu/drm/amd/amdgpu/vid.h
index 3bf7172..4bd2bfd 100644
--- a/drivers/gpu/drm/amd/amdgpu/vid.h
+++ b/drivers/gpu/drm/amd/amdgpu/vid.h
@@ -368,4 +368,41 @@
#define VCE_CMD_IB_AUTO 0x00000005
#define VCE_CMD_SEMAPHORE 0x00000006
+
+/* mmPA_SC_RASTER_CONFIG mask */
+#define RB_MAP_PKR0(x) ((x) << 0)
+#define RB_MAP_PKR0_MASK (0x3 << 0)
+#define RB_MAP_PKR1(x) ((x) << 2)
+#define RB_MAP_PKR1_MASK (0x3 << 2)
+#define RB_XSEL2(x) ((x) << 4)
+#define RB_XSEL2_MASK (0x3 << 4)
+#define RB_XSEL (1 << 6)
+#define RB_YSEL (1 << 7)
+#define PKR_MAP(x) ((x) << 8)
+#define PKR_MAP_MASK (0x3 << 8)
+#define PKR_XSEL(x) ((x) << 10)
+#define PKR_XSEL_MASK (0x3 << 10)
+#define PKR_YSEL(x) ((x) << 12)
+#define PKR_YSEL_MASK (0x3 << 12)
+#define SC_MAP(x) ((x) << 16)
+#define SC_MAP_MASK (0x3 << 16)
+#define SC_XSEL(x) ((x) << 18)
+#define SC_XSEL_MASK (0x3 << 18)
+#define SC_YSEL(x) ((x) << 20)
+#define SC_YSEL_MASK (0x3 << 20)
+#define SE_MAP(x) ((x) << 24)
+#define SE_MAP_MASK (0x3 << 24)
+#define SE_XSEL(x) ((x) << 26)
+#define SE_XSEL_MASK (0x3 << 26)
+#define SE_YSEL(x) ((x) << 28)
+#define SE_YSEL_MASK (0x3 << 28)
+
+/* mmPA_SC_RASTER_CONFIG_1 mask */
+#define SE_PAIR_MAP(x) ((x) << 0)
+#define SE_PAIR_MAP_MASK (0x3 << 0)
+#define SE_PAIR_XSEL(x) ((x) << 2)
+#define SE_PAIR_XSEL_MASK (0x3 << 2)
+#define SE_PAIR_YSEL(x) ((x) << 4)
+#define SE_PAIR_YSEL_MASK (0x3 << 4)
+
#endif
--
2.7.4
|
