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 Olk <maraeo (at) 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 "amd/common/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 44 radeon_cs_memory_below_limit(struct r600_common_screen *screen, 45 struct radeon_winsys_cs *cs, 46 uint64_t vram, uint64_t gtt) 47 { 48 vram += cs->used_vram; 49 gtt += cs->used_gart; 50 51 /* Anything that goes above the VRAM size should go to GTT. */ 52 if (vram > screen->info.vram_size) 53 gtt += vram - screen->info.vram_size; 54 55 /* Now we just need to check if we have enough GTT. */ 56 return gtt < screen->info.gart_size * 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 */ 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 enum radeon_bo_usage usage, 73 enum radeon_bo_priority priority) 74 { 75 assert(usage); 76 return rctx->ws->cs_add_buffer( 77 ring->cs, rbo->buf, 78 (enum radeon_bo_usage)(usage | RADEON_USAGE_SYNCHRONIZED), 79 rbo->domains, priority) * 4; 80 } 81 82 /** 83 * Same as above, but also checks memory usage and flushes the context 84 * accordingly. 85 * 86 * When this SHOULD NOT be used: 87 * 88 * - if r600_context_add_resource_size has been called for the buffer 89 * followed by *_need_cs_space for checking the memory usage 90 * 91 * - if r600_need_dma_space has been called for the buffer 92 * 93 * - when emitting state packets and draw packets (because preceding packets 94 * can't be re-emitted at that point) 95 * 96 * - if shader resource "enabled_mask" is not up-to-date or there is 97 * a different constraint disallowing a context flush 98 */ 99 static inline unsigned 100 radeon_add_to_buffer_list_check_mem(struct r600_common_context *rctx, 101 struct r600_ring *ring, 102 struct r600_resource *rbo, 103 enum radeon_bo_usage usage, 104 enum radeon_bo_priority priority, 105 bool check_mem) 106 { 107 if (check_mem && 108 !radeon_cs_memory_below_limit(rctx->screen, ring->cs, 109 rctx->vram + rbo->vram_usage, 110 rctx->gtt + rbo->gart_usage)) 111 ring->flush(rctx, RADEON_FLUSH_ASYNC, NULL); 112 113 return radeon_add_to_buffer_list(rctx, ring, rbo, usage, priority); 114 } 115 116 static inline void r600_emit_reloc(struct r600_common_context *rctx, 117 struct r600_ring *ring, struct r600_resource *rbo, 118 enum radeon_bo_usage usage, 119 enum radeon_bo_priority priority) 120 { 121 struct radeon_winsys_cs *cs = ring->cs; 122 bool has_vm = ((struct r600_common_screen*)rctx->b.screen)->info.has_virtual_memory; 123 unsigned reloc = radeon_add_to_buffer_list(rctx, ring, rbo, usage, priority); 124 125 if (!has_vm) { 126 radeon_emit(cs, PKT3(PKT3_NOP, 0, 0)); 127 radeon_emit(cs, reloc); 128 } 129 } 130 131 static inline void radeon_set_config_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num) 132 { 133 assert(reg < R600_CONTEXT_REG_OFFSET); 134 assert(cs->current.cdw + 2 + num <= cs->current.max_dw); 135 radeon_emit(cs, PKT3(PKT3_SET_CONFIG_REG, num, 0)); 136 radeon_emit(cs, (reg - R600_CONFIG_REG_OFFSET) >> 2); 137 } 138 139 static inline void radeon_set_config_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value) 140 { 141 radeon_set_config_reg_seq(cs, reg, 1); 142 radeon_emit(cs, value); 143 } 144 145 static inline void radeon_set_context_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num) 146 { 147 assert(reg >= R600_CONTEXT_REG_OFFSET); 148 assert(cs->current.cdw + 2 + num <= cs->current.max_dw); 149 radeon_emit(cs, PKT3(PKT3_SET_CONTEXT_REG, num, 0)); 150 radeon_emit(cs, (reg - R600_CONTEXT_REG_OFFSET) >> 2); 151 } 152 153 static inline void radeon_set_context_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value) 154 { 155 radeon_set_context_reg_seq(cs, reg, 1); 156 radeon_emit(cs, value); 157 } 158 159 static inline void radeon_set_context_reg_idx(struct radeon_winsys_cs *cs, 160 unsigned reg, unsigned idx, 161 unsigned value) 162 { 163 assert(reg >= R600_CONTEXT_REG_OFFSET); 164 assert(cs->current.cdw + 3 <= cs->current.max_dw); 165 radeon_emit(cs, PKT3(PKT3_SET_CONTEXT_REG, 1, 0)); 166 radeon_emit(cs, (reg - R600_CONTEXT_REG_OFFSET) >> 2 | (idx << 28)); 167 radeon_emit(cs, value); 168 } 169 170 static inline void radeon_set_sh_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num) 171 { 172 assert(reg >= SI_SH_REG_OFFSET && reg < SI_SH_REG_END); 173 assert(cs->current.cdw + 2 + num <= cs->current.max_dw); 174 radeon_emit(cs, PKT3(PKT3_SET_SH_REG, num, 0)); 175 radeon_emit(cs, (reg - SI_SH_REG_OFFSET) >> 2); 176 } 177 178 static inline void radeon_set_sh_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value) 179 { 180 radeon_set_sh_reg_seq(cs, reg, 1); 181 radeon_emit(cs, value); 182 } 183 184 static inline void radeon_set_uconfig_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num) 185 { 186 assert(reg >= CIK_UCONFIG_REG_OFFSET && reg < CIK_UCONFIG_REG_END); 187 assert(cs->current.cdw + 2 + num <= cs->current.max_dw); 188 radeon_emit(cs, PKT3(PKT3_SET_UCONFIG_REG, num, 0)); 189 radeon_emit(cs, (reg - CIK_UCONFIG_REG_OFFSET) >> 2); 190 } 191 192 static inline void radeon_set_uconfig_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value) 193 { 194 radeon_set_uconfig_reg_seq(cs, reg, 1); 195 radeon_emit(cs, value); 196 } 197 198 static inline void radeon_set_uconfig_reg_idx(struct radeon_winsys_cs *cs, 199 unsigned reg, unsigned idx, 200 unsigned value) 201 { 202 assert(reg >= CIK_UCONFIG_REG_OFFSET && reg < CIK_UCONFIG_REG_END); 203 assert(cs->current.cdw + 3 <= cs->current.max_dw); 204 radeon_emit(cs, PKT3(PKT3_SET_UCONFIG_REG, 1, 0)); 205 radeon_emit(cs, (reg - CIK_UCONFIG_REG_OFFSET) >> 2 | (idx << 28)); 206 radeon_emit(cs, value); 207 } 208 209 #endif 210