1/*
2 * Copyright © 2019 Raspberry Pi
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 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * 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 NONINFRINGEMENT.  IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
21 * IN THE SOFTWARE.
22 */
23
24#ifndef V3DV_CL_H
25#define V3DV_CL_H
26
27#include "broadcom/cle/v3d_packet_helpers.h"
28
29#include "list.h"
30
31struct v3dv_bo;
32struct v3dv_job;
33struct v3dv_cl;
34
35void v3dv_job_add_bo(struct v3dv_job *job, struct v3dv_bo *bo);
36
37/**
38 * Undefined structure, used for typechecking that you're passing the pointers
39 * to these functions correctly.
40 */
41struct v3dv_cl_out;
42
43/** A reference to a BO used in the CL packing functions */
44struct v3dv_cl_reloc {
45   struct v3dv_bo *bo;
46   uint32_t offset;
47};
48
49static inline void
50pack_emit_reloc(void *cl, const void *reloc) {}
51
52#define __gen_user_data struct v3dv_cl
53#define __gen_address_type struct v3dv_cl_reloc
54#define __gen_address_offset(reloc) (((reloc)->bo ? (reloc)->bo->offset : 0) + \
55                                     (reloc)->offset)
56#define __gen_emit_reloc cl_pack_emit_reloc
57#define __gen_unpack_address(cl, s, e) __unpack_address(cl, s, e)
58
59struct v3dv_cl {
60   void *base;
61   struct v3dv_job *job;
62   struct v3dv_cl_out *next;
63   struct v3dv_bo *bo;
64   uint32_t size;
65   struct list_head bo_list;
66};
67
68static inline struct v3dv_cl_reloc
69__unpack_address(const uint8_t *cl, uint32_t s, uint32_t e)
70{
71    struct v3dv_cl_reloc reloc =
72            { NULL, __gen_unpack_uint(cl, s, e) << (31 - (e - s)) };
73    return reloc;
74}
75
76static inline uint32_t
77v3dv_cl_offset(struct v3dv_cl *cl)
78{
79   return (char *)cl->next - (char *)cl->base;
80}
81
82static inline struct v3dv_cl_reloc
83v3dv_cl_address(struct v3dv_bo *bo, uint32_t offset)
84{
85   struct v3dv_cl_reloc reloc = {
86      .bo = bo,
87      .offset = offset,
88   };
89   return reloc;
90}
91
92static inline struct v3dv_cl_reloc
93v3dv_cl_get_address(struct v3dv_cl *cl)
94{
95   return (struct v3dv_cl_reloc){ .bo = cl->bo, .offset = v3dv_cl_offset(cl) };
96}
97
98void v3dv_cl_init(struct v3dv_job *job, struct v3dv_cl *cl);
99void v3dv_cl_destroy(struct v3dv_cl *cl);
100
101static inline struct v3dv_cl_out *
102cl_start(struct v3dv_cl *cl)
103{
104   return cl->next;
105}
106
107static inline void
108cl_end(struct v3dv_cl *cl, struct v3dv_cl_out *next)
109{
110   cl->next = next;
111   assert(v3dv_cl_offset(cl) <= cl->size);
112}
113
114static inline void
115cl_advance(struct v3dv_cl_out **cl, uint32_t n)
116{
117   (*cl) = (struct v3dv_cl_out *)((char *)(*cl) + n);
118}
119
120static inline void
121cl_aligned_u32(struct v3dv_cl_out **cl, uint32_t n)
122{
123   *(uint32_t *)(*cl) = n;
124   cl_advance(cl, 4);
125}
126
127static inline void
128cl_aligned_f(struct v3dv_cl_out **cl, float f)
129{
130   cl_aligned_u32(cl, fui(f));
131}
132
133static inline void
134cl_aligned_reloc(struct v3dv_cl *cl,
135                 struct v3dv_cl_out **cl_out,
136                 struct v3dv_bo *bo,
137                 uint32_t offset)
138{
139   cl_aligned_u32(cl_out, bo->offset + offset);
140   v3dv_job_add_bo(cl->job, bo);
141}
142
143uint32_t v3dv_cl_ensure_space(struct v3dv_cl *cl, uint32_t space, uint32_t alignment);
144void v3dv_cl_ensure_space_with_branch(struct v3dv_cl *cl, uint32_t space);
145
146/* We redefine ALIGN as a macro as we want to use cl_aligned_packet_length for
147 * struct fields
148 */
149#define ALIGN(value, alignment)                           \
150        (((value) + (alignment) - 1) & ~((alignment) - 1))
151
152#define cl_packet_header(packet) V3DX(packet ## _header)
153#define cl_packet_length(packet) V3DX(packet ## _length)
154#define cl_aligned_packet_length(packet, alignment) ALIGN(cl_packet_length(packet), alignment)
155#define cl_packet_pack(packet)   V3DX(packet ## _pack)
156#define cl_packet_struct(packet) V3DX(packet)
157
158/* Macro for setting up an emit of a CL struct.  A temporary unpacked struct
159 * is created, which you get to set fields in of the form:
160 *
161 * cl_emit(bcl, FLAT_SHADE_FLAGS, flags) {
162 *     .flags.flat_shade_flags = 1 << 2,
163 * }
164 *
165 * or default values only can be emitted with just:
166 *
167 * cl_emit(bcl, FLAT_SHADE_FLAGS, flags);
168 *
169 * The trick here is that we make a for loop that will execute the body
170 * (either the block or the ';' after the macro invocation) exactly once.
171 */
172#define cl_emit(cl, packet, name)                                \
173        for (struct cl_packet_struct(packet) name = {            \
174                cl_packet_header(packet)                         \
175        },                                                       \
176        *_loop_terminate = &name;                                \
177        __builtin_expect(_loop_terminate != NULL, 1);            \
178        ({                                                       \
179                struct v3dv_cl_out *cl_out = cl_start(cl);        \
180                cl_packet_pack(packet)(cl, (uint8_t *)cl_out, &name); \
181                cl_advance(&cl_out, cl_packet_length(packet));   \
182                cl_end(cl, cl_out);                              \
183                _loop_terminate = NULL;                          \
184        }))                                                      \
185
186#define cl_emit_with_prepacked(cl, packet, prepacked, name)      \
187        for (struct cl_packet_struct(packet) name = {            \
188                cl_packet_header(packet)                         \
189        },                                                       \
190        *_loop_terminate = &name;                                \
191        __builtin_expect(_loop_terminate != NULL, 1);            \
192        ({                                                       \
193                struct v3dv_cl_out *cl_out = cl_start(cl);        \
194                uint8_t packed[cl_packet_length(packet)];         \
195                cl_packet_pack(packet)(cl, packed, &name);       \
196                for (int _i = 0; _i < cl_packet_length(packet); _i++) \
197                        ((uint8_t *)cl_out)[_i] = packed[_i] | (prepacked)[_i]; \
198                cl_advance(&cl_out, cl_packet_length(packet));   \
199                cl_end(cl, cl_out);                              \
200                _loop_terminate = NULL;                          \
201        }))                                                      \
202
203/**
204 * Helper function called by the XML-generated pack functions for filling in
205 * an address field in shader records.
206 *
207 * Since we have a private address space as of V3D, our BOs can have lifelong
208 * offsets, and all the kernel needs to know is which BOs need to be paged in
209 * for this exec.
210 */
211static inline void
212cl_pack_emit_reloc(struct v3dv_cl *cl, const struct v3dv_cl_reloc *reloc)
213{
214        if (reloc->bo)
215                v3dv_job_add_bo(cl->job, reloc->bo);
216}
217
218#define cl_emit_prepacked_sized(cl, packet, size) do {                \
219        memcpy((cl)->next, packet, size);             \
220        cl_advance(&(cl)->next, size);                \
221} while (0)
222
223#define cl_emit_prepacked(cl, packet) \
224        cl_emit_prepacked_sized(cl, packet, sizeof(*(packet)))
225
226#define v3dvx_pack(packed, packet, name)                         \
227        for (struct cl_packet_struct(packet) name = {            \
228                cl_packet_header(packet)                         \
229        },                                                       \
230        *_loop_terminate = &name;                                \
231        __builtin_expect(_loop_terminate != NULL, 1);            \
232        ({                                                       \
233                cl_packet_pack(packet)(NULL, (uint8_t *)packed, &name); \
234                VG(VALGRIND_CHECK_MEM_IS_DEFINED((uint8_t *)packed, \
235                                                 cl_packet_length(packet))); \
236                _loop_terminate = NULL;                          \
237        }))                                                      \
238
239#endif /* V3DV_CL_H */
240