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
|
/*
* GP10B GPU GR
*
* Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
* version 2, as published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*/
#include "gk20a/gk20a.h" /* FERMI and MAXWELL classes defined here */
#include "gk20a/gr_gk20a.h"
#include "gm20b/gr_gm20b.h" /* for MAXWELL classes */
#include "gp10b/gr_gp10b.h"
#include "hw_gr_gp10b.h"
#include "hw_proj_gp10b.h"
bool gr_gp10b_is_valid_class(struct gk20a *g, u32 class_num)
{
bool valid = false;
switch (class_num) {
case PASCAL_COMPUTE_A:
case PASCAL_A:
case PASCAL_DMA_COPY_A:
valid = true;
break;
case MAXWELL_COMPUTE_B:
case MAXWELL_B:
case FERMI_TWOD_A:
case KEPLER_DMA_COPY_A:
case MAXWELL_DMA_COPY_A:
valid = true;
break;
default:
break;
}
gk20a_dbg_info("class=0x%x valid=%d", class_num, valid);
return valid;
}
int gr_gp10b_commit_global_cb_manager(struct gk20a *g,
struct channel_gk20a *c, bool patch)
{
struct gr_gk20a *gr = &g->gr;
struct channel_ctx_gk20a *ch_ctx = NULL;
u32 attrib_offset_in_chunk = 0;
u32 alpha_offset_in_chunk = 0;
u32 pd_ab_max_output;
u32 gpc_index, ppc_index;
u32 temp;
u32 cbm_cfg_size1, cbm_cfg_size2;
gk20a_dbg_fn("");
if (patch) {
int err;
ch_ctx = &c->ch_ctx;
err = gr_gk20a_ctx_patch_write_begin(g, ch_ctx);
if (err)
return err;
}
gr_gk20a_ctx_patch_write(g, ch_ctx, gr_ds_tga_constraintlogic_beta_r(),
gr->attrib_cb_default_size, patch);
gr_gk20a_ctx_patch_write(g, ch_ctx, gr_ds_tga_constraintlogic_alpha_r(),
gr->alpha_cb_default_size, patch);
pd_ab_max_output = (gr->alpha_cb_default_size *
gr_gpc0_ppc0_cbm_beta_cb_size_v_granularity_v()) /
gr_pd_ab_dist_cfg1_max_output_granularity_v();
gr_gk20a_ctx_patch_write(g, ch_ctx, gr_pd_ab_dist_cfg1_r(),
gr_pd_ab_dist_cfg1_max_output_f(pd_ab_max_output) |
gr_pd_ab_dist_cfg1_max_batches_init_f(), patch);
alpha_offset_in_chunk = attrib_offset_in_chunk +
gr->tpc_count * gr->attrib_cb_size;
for (gpc_index = 0; gpc_index < gr->gpc_count; gpc_index++) {
temp = proj_gpc_stride_v() * gpc_index;
for (ppc_index = 0; ppc_index < gr->gpc_ppc_count[gpc_index];
ppc_index++) {
cbm_cfg_size1 = gr->attrib_cb_default_size *
gr->pes_tpc_count[ppc_index][gpc_index];
cbm_cfg_size2 = gr->alpha_cb_default_size *
gr->pes_tpc_count[ppc_index][gpc_index];
gr_gk20a_ctx_patch_write(g, ch_ctx,
gr_gpc0_ppc0_cbm_beta_cb_size_r() + temp +
proj_ppc_in_gpc_stride_v() * ppc_index,
cbm_cfg_size1, patch);
gr_gk20a_ctx_patch_write(g, ch_ctx,
gr_gpc0_ppc0_cbm_beta_cb_offset_r() + temp +
proj_ppc_in_gpc_stride_v() * ppc_index,
attrib_offset_in_chunk, patch);
attrib_offset_in_chunk += gr->attrib_cb_size *
gr->pes_tpc_count[ppc_index][gpc_index];
gr_gk20a_ctx_patch_write(g, ch_ctx,
gr_gpc0_ppc0_cbm_alpha_cb_size_r() + temp +
proj_ppc_in_gpc_stride_v() * ppc_index,
cbm_cfg_size2, patch);
gr_gk20a_ctx_patch_write(g, ch_ctx,
gr_gpc0_ppc0_cbm_alpha_cb_offset_r() + temp +
proj_ppc_in_gpc_stride_v() * ppc_index,
alpha_offset_in_chunk, patch);
alpha_offset_in_chunk += gr->alpha_cb_size *
gr->pes_tpc_count[ppc_index][gpc_index];
gr_gk20a_ctx_patch_write(g, ch_ctx,
gr_gpc0_ppc0_cbm_beta_steady_state_cb_size_r() + temp +
proj_ppc_in_gpc_stride_v() * ppc_index,
gr->alpha_cb_default_size * gr->pes_tpc_count[ppc_index][gpc_index],
patch);
gr_gk20a_ctx_patch_write(g, ch_ctx,
gr_gpcs_swdx_tc_beta_cb_size_r(ppc_index + gpc_index),
gr_gpcs_swdx_tc_beta_cb_size_v_f(cbm_cfg_size1),
patch);
}
}
if (patch)
gr_gk20a_ctx_patch_write_end(g, ch_ctx);
return 0;
}
void gr_gp10b_commit_global_pagepool(struct gk20a *g,
struct channel_ctx_gk20a *ch_ctx,
u64 addr, u32 size, bool patch)
{
gr_gk20a_ctx_patch_write(g, ch_ctx, gr_scc_pagepool_base_r(),
gr_scc_pagepool_base_addr_39_8_f(addr), patch);
gr_gk20a_ctx_patch_write(g, ch_ctx, gr_scc_pagepool_r(),
gr_scc_pagepool_total_pages_f(size) |
gr_scc_pagepool_valid_true_f(), patch);
gr_gk20a_ctx_patch_write(g, ch_ctx, gr_gpcs_gcc_pagepool_base_r(),
gr_gpcs_gcc_pagepool_base_addr_39_8_f(addr), patch);
gr_gk20a_ctx_patch_write(g, ch_ctx, gr_gpcs_gcc_pagepool_r(),
gr_gpcs_gcc_pagepool_total_pages_f(size), patch);
}
void gp10b_init_gr(struct gpu_ops *gops)
{
gm20b_init_gr(gops);
gops->gr.is_valid_class = gr_gp10b_is_valid_class;
gops->gr.commit_global_cb_manager = gr_gp10b_commit_global_cb_manager;
gops->gr.commit_global_pagepool = gr_gp10b_commit_global_pagepool;
}
|
