1/*
2 * Copyright 2012 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:
24 *      Christian König <christian.koenig@amd.com>
25 */
26
27#include "util/u_memory.h"
28#include "radeonsi_pipe.h"
29#include "radeonsi_pm4.h"
30#include "sid.h"
31#include "r600_hw_context_priv.h"
32
33#define NUMBER_OF_STATES (sizeof(union si_state) / sizeof(struct si_pm4_state *))
34
35void si_pm4_cmd_begin(struct si_pm4_state *state, unsigned opcode)
36{
37	state->last_opcode = opcode;
38	state->last_pm4 = state->ndw++;
39}
40
41void si_pm4_cmd_add(struct si_pm4_state *state, uint32_t dw)
42{
43	state->pm4[state->ndw++] = dw;
44}
45
46void si_pm4_cmd_end(struct si_pm4_state *state, bool predicate)
47{
48	unsigned count;
49	count = state->ndw - state->last_pm4 - 2;
50	state->pm4[state->last_pm4] = PKT3(state->last_opcode,
51					   count, predicate);
52
53	assert(state->ndw <= SI_PM4_MAX_DW);
54}
55
56void si_pm4_set_reg(struct si_pm4_state *state, unsigned reg, uint32_t val)
57{
58	unsigned opcode;
59
60	if (reg >= SI_CONFIG_REG_OFFSET && reg <= SI_CONFIG_REG_END) {
61		opcode = PKT3_SET_CONFIG_REG;
62		reg -= SI_CONFIG_REG_OFFSET;
63
64	} else if (reg >= SI_SH_REG_OFFSET && reg <= SI_SH_REG_END) {
65		opcode = PKT3_SET_SH_REG;
66		reg -= SI_SH_REG_OFFSET;
67
68	} else if (reg >= SI_CONTEXT_REG_OFFSET && reg <= SI_CONTEXT_REG_END) {
69		opcode = PKT3_SET_CONTEXT_REG;
70		reg -= SI_CONTEXT_REG_OFFSET;
71	} else {
72		R600_ERR("Invalid register offset %08x!\n", reg);
73		return;
74	}
75
76	reg >>= 2;
77
78	if (opcode != state->last_opcode || reg != (state->last_reg + 1)) {
79		si_pm4_cmd_begin(state, opcode);
80		si_pm4_cmd_add(state, reg);
81	}
82
83	state->last_reg = reg;
84	si_pm4_cmd_add(state, val);
85	si_pm4_cmd_end(state, false);
86}
87
88void si_pm4_add_bo(struct si_pm4_state *state,
89                   struct si_resource *bo,
90                   enum radeon_bo_usage usage)
91{
92	unsigned idx = state->nbo++;
93	assert(idx < SI_PM4_MAX_BO);
94
95	si_resource_reference(&state->bo[idx], bo);
96	state->bo_usage[idx] = usage;
97}
98
99void si_pm4_sh_data_begin(struct si_pm4_state *state)
100{
101	si_pm4_cmd_begin(state, PKT3_NOP);
102}
103
104void si_pm4_sh_data_add(struct si_pm4_state *state, uint32_t dw)
105{
106	si_pm4_cmd_add(state, dw);
107}
108
109void si_pm4_sh_data_end(struct si_pm4_state *state, unsigned reg)
110{
111	unsigned offs = state->last_pm4 + 1;
112
113	/* Bail if no data was added */
114	if (state->ndw == offs) {
115		state->ndw--;
116		return;
117	}
118
119	si_pm4_cmd_end(state, false);
120
121	si_pm4_cmd_begin(state, PKT3_SET_SH_REG_OFFSET);
122	si_pm4_cmd_add(state, (reg - SI_SH_REG_OFFSET) >> 2);
123	state->relocs[state->nrelocs++] = state->ndw;
124	si_pm4_cmd_add(state, offs << 2);
125	si_pm4_cmd_add(state, 0);
126	si_pm4_cmd_end(state, false);
127}
128
129void si_pm4_inval_shader_cache(struct si_pm4_state *state)
130{
131	state->cp_coher_cntl |= S_0085F0_SH_ICACHE_ACTION_ENA(1);
132	state->cp_coher_cntl |= S_0085F0_SH_KCACHE_ACTION_ENA(1);
133}
134
135void si_pm4_inval_texture_cache(struct si_pm4_state *state)
136{
137	state->cp_coher_cntl |= S_0085F0_TC_ACTION_ENA(1);
138}
139
140void si_pm4_inval_vertex_cache(struct si_pm4_state *state)
141{
142        /* Some GPUs don't have the vertex cache and must use the texture cache instead. */
143	state->cp_coher_cntl |= S_0085F0_TC_ACTION_ENA(1);
144}
145
146void si_pm4_inval_fb_cache(struct si_pm4_state *state, unsigned nr_cbufs)
147{
148	state->cp_coher_cntl |= S_0085F0_CB_ACTION_ENA(1);
149	state->cp_coher_cntl |= ((1 << nr_cbufs) - 1) << S_0085F0_CB0_DEST_BASE_ENA_SHIFT;
150}
151
152void si_pm4_inval_zsbuf_cache(struct si_pm4_state *state)
153{
154	state->cp_coher_cntl |= S_0085F0_DB_ACTION_ENA(1) | S_0085F0_DB_DEST_BASE_ENA(1);
155}
156
157void si_pm4_free_state(struct r600_context *rctx,
158		       struct si_pm4_state *state,
159		       unsigned idx)
160{
161	if (state == NULL)
162		return;
163
164	if (idx != ~0 && rctx->emitted.array[idx] == state) {
165		rctx->emitted.array[idx] = NULL;
166	}
167
168	for (int i = 0; i < state->nbo; ++i) {
169		si_resource_reference(&state->bo[i], NULL);
170	}
171	FREE(state);
172}
173
174uint32_t si_pm4_sync_flags(struct r600_context *rctx)
175{
176	uint32_t cp_coher_cntl = 0;
177
178	for (int i = 0; i < NUMBER_OF_STATES; ++i) {
179		struct si_pm4_state *state = rctx->queued.array[i];
180
181		if (!state || rctx->emitted.array[i] == state)
182			continue;
183
184		cp_coher_cntl |= state->cp_coher_cntl;
185	}
186	return cp_coher_cntl;
187}
188
189unsigned si_pm4_dirty_dw(struct r600_context *rctx)
190{
191	unsigned count = 0;
192
193	for (int i = 0; i < NUMBER_OF_STATES; ++i) {
194		struct si_pm4_state *state = rctx->queued.array[i];
195
196		if (!state || rctx->emitted.array[i] == state)
197			continue;
198
199		count += state->ndw;
200	}
201
202	return count;
203}
204
205void si_pm4_emit(struct r600_context *rctx, struct si_pm4_state *state)
206{
207	struct radeon_winsys_cs *cs = rctx->cs;
208	for (int i = 0; i < state->nbo; ++i) {
209		r600_context_bo_reloc(rctx, state->bo[i],
210				      state->bo_usage[i]);
211	}
212
213	memcpy(&cs->buf[cs->cdw], state->pm4, state->ndw * 4);
214
215	for (int i = 0; i < state->nrelocs; ++i) {
216		cs->buf[cs->cdw + state->relocs[i]] += cs->cdw << 2;
217	}
218
219	cs->cdw += state->ndw;
220}
221
222void si_pm4_emit_dirty(struct r600_context *rctx)
223{
224	for (int i = 0; i < NUMBER_OF_STATES; ++i) {
225		struct si_pm4_state *state = rctx->queued.array[i];
226
227		if (!state || rctx->emitted.array[i] == state)
228			continue;
229
230		si_pm4_emit(rctx, state);
231		rctx->emitted.array[i] = state;
232	}
233}
234
235void si_pm4_reset_emitted(struct r600_context *rctx)
236{
237	memset(&rctx->emitted, 0, sizeof(rctx->emitted));
238}
239