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
|
From 5c3422b0b135b46c8dca9c1d909c1ae84f3561bd Mon Sep 17 00:00:00 2001
From: "monk.liu" <monk.liu@amd.com>
Date: Wed, 23 Sep 2015 13:49:58 +0800
Subject: [PATCH 0563/1050] drm/amdgpu: sync ce and me with SWITCH_BUFFER(2)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
we used to adopt wait_reg_mem to let CE wait before DE finish page
updating, but from Tonga+, CE doesn't support wait_reg_mem package so
this logic no longer works.
so here is another approach to do same thing:
Insert two of SWITCH_BUFFER at both front and end of vm_flush can
guarantee that CE not go further to process IB_const before vm_flush
done.
Insert two of SWITCH_BUFFER also works on CI, so remove legency method
to sync CE and ME
v2:
Insert double SWITCH_BUFFER at front of vm flush as well.
Signed-off-by: monk.liu <monk.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 --
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 55 +++++++--------------------------
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 58 ++++++++---------------------------
3 files changed, 23 insertions(+), 92 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 57b427f..6647fb2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -1202,8 +1202,6 @@ struct amdgpu_gfx {
struct amdgpu_irq_src priv_inst_irq;
/* gfx status */
uint32_t gfx_current_status;
- /* sync signal for const engine */
- unsigned ce_sync_offs;
/* ce ram size*/
unsigned ce_ram_size;
};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
index 392ec10..e992bf2 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
@@ -3610,41 +3610,6 @@ static int gfx_v7_0_cp_resume(struct amdgpu_device *adev)
return 0;
}
-static void gfx_v7_0_ce_sync_me(struct amdgpu_ring *ring)
-{
- struct amdgpu_device *adev = ring->adev;
- u64 gpu_addr = adev->wb.gpu_addr + adev->gfx.ce_sync_offs * 4;
-
- /* instruct DE to set a magic number */
- amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
- amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
- WRITE_DATA_DST_SEL(5)));
- amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
- amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
- amdgpu_ring_write(ring, 1);
-
- /* let CE wait till condition satisfied */
- amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
- amdgpu_ring_write(ring, (WAIT_REG_MEM_OPERATION(0) | /* wait */
- WAIT_REG_MEM_MEM_SPACE(1) | /* memory */
- WAIT_REG_MEM_FUNCTION(3) | /* == */
- WAIT_REG_MEM_ENGINE(2))); /* ce */
- amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
- amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
- amdgpu_ring_write(ring, 1);
- amdgpu_ring_write(ring, 0xffffffff);
- amdgpu_ring_write(ring, 4); /* poll interval */
-
- /* instruct CE to reset wb of ce_sync to zero */
- amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
- amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) |
- WRITE_DATA_DST_SEL(5) |
- WR_CONFIRM));
- amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
- amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
- amdgpu_ring_write(ring, 0);
-}
-
/*
* vm
* VMID 0 is the physical GPU addresses as used by the kernel.
@@ -3663,6 +3628,13 @@ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned vm_id, uint64_t pd_addr)
{
int usepfp = (ring->type == AMDGPU_RING_TYPE_GFX);
+ if (usepfp) {
+ /* synce CE with ME to prevent CE fetch CEIB before context switch done */
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
+ }
amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) |
@@ -3703,7 +3675,10 @@ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
amdgpu_ring_write(ring, 0x0);
/* synce CE with ME to prevent CE fetch CEIB before context switch done */
- gfx_v7_0_ce_sync_me(ring);
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
}
}
@@ -4805,12 +4780,6 @@ static int gfx_v7_0_sw_init(void *handle)
return r;
}
- r = amdgpu_wb_get(adev, &adev->gfx.ce_sync_offs);
- if (r) {
- DRM_ERROR("(%d) gfx.ce_sync_offs wb alloc failed\n", r);
- return r;
- }
-
for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
ring = &adev->gfx.gfx_ring[i];
ring->ring_obj = NULL;
@@ -4889,8 +4858,6 @@ static int gfx_v7_0_sw_fini(void *handle)
for (i = 0; i < adev->gfx.num_compute_rings; i++)
amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
- amdgpu_wb_free(adev, adev->gfx.ce_sync_offs);
-
gfx_v7_0_cp_compute_fini(adev);
gfx_v7_0_rlc_fini(adev);
gfx_v7_0_mec_fini(adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index 78e5900..cb4f68f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -940,12 +940,6 @@ static int gfx_v8_0_sw_init(void *handle)
return r;
}
- r = amdgpu_wb_get(adev, &adev->gfx.ce_sync_offs);
- if (r) {
- DRM_ERROR("(%d) gfx.ce_sync_offs wb alloc failed\n", r);
- return r;
- }
-
/* set up the gfx ring */
for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
ring = &adev->gfx.gfx_ring[i];
@@ -1033,8 +1027,6 @@ static int gfx_v8_0_sw_fini(void *handle)
for (i = 0; i < adev->gfx.num_compute_rings; i++)
amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
- amdgpu_wb_free(adev, adev->gfx.ce_sync_offs);
-
gfx_v8_0_mec_fini(adev);
return 0;
@@ -4006,41 +3998,6 @@ static bool gfx_v8_0_ring_emit_semaphore(struct amdgpu_ring *ring,
return true;
}
-static void gfx_v8_0_ce_sync_me(struct amdgpu_ring *ring)
-{
- struct amdgpu_device *adev = ring->adev;
- u64 gpu_addr = adev->wb.gpu_addr + adev->gfx.ce_sync_offs * 4;
-
- /* instruct DE to set a magic number */
- amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
- amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
- WRITE_DATA_DST_SEL(5)));
- amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
- amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
- amdgpu_ring_write(ring, 1);
-
- /* let CE wait till condition satisfied */
- amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
- amdgpu_ring_write(ring, (WAIT_REG_MEM_OPERATION(0) | /* wait */
- WAIT_REG_MEM_MEM_SPACE(1) | /* memory */
- WAIT_REG_MEM_FUNCTION(3) | /* == */
- WAIT_REG_MEM_ENGINE(2))); /* ce */
- amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
- amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
- amdgpu_ring_write(ring, 1);
- amdgpu_ring_write(ring, 0xffffffff);
- amdgpu_ring_write(ring, 4); /* poll interval */
-
- /* instruct CE to reset wb of ce_sync to zero */
- amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
- amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) |
- WRITE_DATA_DST_SEL(5) |
- WR_CONFIRM));
- amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
- amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
- amdgpu_ring_write(ring, 0);
-}
-
static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned vm_id, uint64_t pd_addr)
{
@@ -4057,6 +4014,14 @@ static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
amdgpu_ring_write(ring, 0xffffffff);
amdgpu_ring_write(ring, 4); /* poll interval */
+ if (usepfp) {
+ /* synce CE with ME to prevent CE fetch CEIB before context switch done */
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
+ }
+
amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) |
WRITE_DATA_DST_SEL(0)) |
@@ -4096,9 +4061,10 @@ static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
/* sync PFP to ME, otherwise we might get invalid PFP reads */
amdgpu_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
amdgpu_ring_write(ring, 0x0);
-
- /* synce CE with ME to prevent CE fetch CEIB before context switch done */
- gfx_v8_0_ce_sync_me(ring);
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
}
}
--
1.9.1
|
