• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright 2013 Advanced Micro Devices, Inc.
3  *
4  * Permission is hereby granted, free of charge, to any person obtaining a
5  * copy of this software and associated documentation files (the "Software"),
6  * to deal in the Software without restriction, including without limitation
7  * on the rights to use, copy, modify, merge, publish, distribute, sub
8  * license, and/or sell copies of the Software, and to permit persons to whom
9  * the Software is furnished to do so, subject to the following conditions:
10  *
11  * The above copyright notice and this permission notice (including the next
12  * paragraph) shall be included in all copies or substantial portions of the
13  * Software.
14  *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17  * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
18  * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
19  * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
20  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
21  * USE OR OTHER DEALINGS IN THE SOFTWARE.
22  *
23  * Authors: Marek Olšák <maraeo@gmail.com>
24  */
25 
26 /**
27  * This file contains helpers for writing commands to commands streams.
28  */
29 
30 #ifndef R600_CS_H
31 #define R600_CS_H
32 
33 #include "r600_pipe_common.h"
34 #include "r600d_common.h"
35 
36 /**
37  * Return true if there is enough memory in VRAM and GTT for the buffers
38  * added so far.
39  *
40  * \param vram      VRAM memory size not added to the buffer list yet
41  * \param gtt       GTT memory size not added to the buffer list yet
42  */
43 static inline bool
radeon_cs_memory_below_limit(struct r600_common_screen * screen,struct radeon_cmdbuf * cs,uint64_t vram,uint64_t gtt)44 radeon_cs_memory_below_limit(struct r600_common_screen *screen,
45 			     struct radeon_cmdbuf *cs,
46 			     uint64_t vram, uint64_t gtt)
47 {
48 	vram += (uint64_t)cs->used_vram_kb * 1024;
49 	gtt += (uint64_t)cs->used_gart_kb * 1024;
50 
51 	/* Anything that goes above the VRAM size should go to GTT. */
52 	if (vram > (uint64_t)screen->info.vram_size_kb * 1024)
53 		gtt += vram - (uint64_t)screen->info.vram_size_kb * 1024;
54 
55 	/* Now we just need to check if we have enough GTT. */
56 	return gtt < (uint64_t)screen->info.gart_size_kb * 1024 * 0.7;
57 }
58 
59 /**
60  * Add a buffer to the buffer list for the given command stream (CS).
61  *
62  * All buffers used by a CS must be added to the list. This tells the kernel
63  * driver which buffers are used by GPU commands. Other buffers can
64  * be swapped out (not accessible) during execution.
65  *
66  * The buffer list becomes empty after every context flush and must be
67  * rebuilt.
68  */
radeon_add_to_buffer_list(struct r600_common_context * rctx,struct r600_ring * ring,struct r600_resource * rbo,unsigned usage)69 static inline unsigned radeon_add_to_buffer_list(struct r600_common_context *rctx,
70 						 struct r600_ring *ring,
71 						 struct r600_resource *rbo,
72 						 unsigned usage)
73 {
74 	assert(usage);
75 	return rctx->ws->cs_add_buffer(
76 		&ring->cs, rbo->buf,
77 		usage | RADEON_USAGE_SYNCHRONIZED,
78 		rbo->domains) * 4;
79 }
80 
81 /**
82  * Same as above, but also checks memory usage and flushes the context
83  * accordingly.
84  *
85  * When this SHOULD NOT be used:
86  *
87  * - if r600_context_add_resource_size has been called for the buffer
88  *   followed by *_need_cs_space for checking the memory usage
89  *
90  * - if r600_need_dma_space has been called for the buffer
91  *
92  * - when emitting state packets and draw packets (because preceding packets
93  *   can't be re-emitted at that point)
94  *
95  * - if shader resource "enabled_mask" is not up-to-date or there is
96  *   a different constraint disallowing a context flush
97  */
98 static inline unsigned
radeon_add_to_buffer_list_check_mem(struct r600_common_context * rctx,struct r600_ring * ring,struct r600_resource * rbo,unsigned usage,bool check_mem)99 radeon_add_to_buffer_list_check_mem(struct r600_common_context *rctx,
100 				    struct r600_ring *ring,
101 				    struct r600_resource *rbo,
102 				    unsigned usage,
103 				    bool check_mem)
104 {
105 	if (check_mem &&
106 	    !radeon_cs_memory_below_limit(rctx->screen, &ring->cs,
107 					  rctx->vram + rbo->vram_usage,
108 					  rctx->gtt + rbo->gart_usage))
109 		ring->flush(rctx, PIPE_FLUSH_ASYNC, NULL);
110 
111 	return radeon_add_to_buffer_list(rctx, ring, rbo, usage);
112 }
113 
r600_emit_reloc(struct r600_common_context * rctx,struct r600_ring * ring,struct r600_resource * rbo,unsigned usage)114 static inline void r600_emit_reloc(struct r600_common_context *rctx,
115 				   struct r600_ring *ring, struct r600_resource *rbo,
116 				   unsigned usage)
117 {
118 	struct radeon_cmdbuf *cs = &ring->cs;
119 	bool has_vm = ((struct r600_common_screen*)rctx->b.screen)->info.r600_has_virtual_memory;
120 	unsigned reloc = radeon_add_to_buffer_list(rctx, ring, rbo, usage);
121 
122 	if (!has_vm) {
123 		radeon_emit(cs, PKT3(PKT3_NOP, 0, 0));
124 		radeon_emit(cs, reloc);
125 	}
126 }
127 
radeon_set_config_reg_seq(struct radeon_cmdbuf * cs,unsigned reg,unsigned num)128 static inline void radeon_set_config_reg_seq(struct radeon_cmdbuf *cs, unsigned reg, unsigned num)
129 {
130 	assert(reg < R600_CONTEXT_REG_OFFSET);
131 	assert(cs->current.cdw + 2 + num <= cs->current.max_dw);
132 	radeon_emit(cs, PKT3(PKT3_SET_CONFIG_REG, num, 0));
133 	radeon_emit(cs, (reg - R600_CONFIG_REG_OFFSET) >> 2);
134 }
135 
radeon_set_config_reg(struct radeon_cmdbuf * cs,unsigned reg,unsigned value)136 static inline void radeon_set_config_reg(struct radeon_cmdbuf *cs, unsigned reg, unsigned value)
137 {
138 	radeon_set_config_reg_seq(cs, reg, 1);
139 	radeon_emit(cs, value);
140 }
141 
radeon_set_context_reg_seq(struct radeon_cmdbuf * cs,unsigned reg,unsigned num)142 static inline void radeon_set_context_reg_seq(struct radeon_cmdbuf *cs, unsigned reg, unsigned num)
143 {
144 	assert(reg >= R600_CONTEXT_REG_OFFSET);
145 	assert(cs->current.cdw + 2 + num <= cs->current.max_dw);
146 	radeon_emit(cs, PKT3(PKT3_SET_CONTEXT_REG, num, 0));
147 	radeon_emit(cs, (reg - R600_CONTEXT_REG_OFFSET) >> 2);
148 }
149 
radeon_set_context_reg(struct radeon_cmdbuf * cs,unsigned reg,unsigned value)150 static inline void radeon_set_context_reg(struct radeon_cmdbuf *cs, unsigned reg, unsigned value)
151 {
152 	radeon_set_context_reg_seq(cs, reg, 1);
153 	radeon_emit(cs, value);
154 }
155 
radeon_set_context_reg_idx(struct radeon_cmdbuf * cs,unsigned reg,unsigned idx,unsigned value)156 static inline void radeon_set_context_reg_idx(struct radeon_cmdbuf *cs,
157 					      unsigned reg, unsigned idx,
158 					      unsigned value)
159 {
160 	assert(reg >= R600_CONTEXT_REG_OFFSET);
161 	assert(cs->current.cdw + 3 <= cs->current.max_dw);
162 	radeon_emit(cs, PKT3(PKT3_SET_CONTEXT_REG, 1, 0));
163 	radeon_emit(cs, (reg - R600_CONTEXT_REG_OFFSET) >> 2 | (idx << 28));
164 	radeon_emit(cs, value);
165 }
166 
radeon_set_sh_reg_seq(struct radeon_cmdbuf * cs,unsigned reg,unsigned num)167 static inline void radeon_set_sh_reg_seq(struct radeon_cmdbuf *cs, unsigned reg, unsigned num)
168 {
169 	assert(reg >= SI_SH_REG_OFFSET && reg < SI_SH_REG_END);
170 	assert(cs->current.cdw + 2 + num <= cs->current.max_dw);
171 	radeon_emit(cs, PKT3(PKT3_SET_SH_REG, num, 0));
172 	radeon_emit(cs, (reg - SI_SH_REG_OFFSET) >> 2);
173 }
174 
radeon_set_sh_reg(struct radeon_cmdbuf * cs,unsigned reg,unsigned value)175 static inline void radeon_set_sh_reg(struct radeon_cmdbuf *cs, unsigned reg, unsigned value)
176 {
177 	radeon_set_sh_reg_seq(cs, reg, 1);
178 	radeon_emit(cs, value);
179 }
180 
radeon_set_uconfig_reg_seq(struct radeon_cmdbuf * cs,unsigned reg,unsigned num)181 static inline void radeon_set_uconfig_reg_seq(struct radeon_cmdbuf *cs, unsigned reg, unsigned num)
182 {
183 	assert(reg >= CIK_UCONFIG_REG_OFFSET && reg < CIK_UCONFIG_REG_END);
184 	assert(cs->current.cdw + 2 + num <= cs->current.max_dw);
185 	radeon_emit(cs, PKT3(PKT3_SET_UCONFIG_REG, num, 0));
186 	radeon_emit(cs, (reg - CIK_UCONFIG_REG_OFFSET) >> 2);
187 }
188 
radeon_set_uconfig_reg(struct radeon_cmdbuf * cs,unsigned reg,unsigned value)189 static inline void radeon_set_uconfig_reg(struct radeon_cmdbuf *cs, unsigned reg, unsigned value)
190 {
191 	radeon_set_uconfig_reg_seq(cs, reg, 1);
192 	radeon_emit(cs, value);
193 }
194 
radeon_set_uconfig_reg_idx(struct radeon_cmdbuf * cs,unsigned reg,unsigned idx,unsigned value)195 static inline void radeon_set_uconfig_reg_idx(struct radeon_cmdbuf *cs,
196 					      unsigned reg, unsigned idx,
197 					      unsigned value)
198 {
199 	assert(reg >= CIK_UCONFIG_REG_OFFSET && reg < CIK_UCONFIG_REG_END);
200 	assert(cs->current.cdw + 3 <= cs->current.max_dw);
201 	radeon_emit(cs, PKT3(PKT3_SET_UCONFIG_REG, 1, 0));
202 	radeon_emit(cs, (reg - CIK_UCONFIG_REG_OFFSET) >> 2 | (idx << 28));
203 	radeon_emit(cs, value);
204 }
205 
206 #endif
207