1/*
2 * Copyright 2011 Christoph Bumiller
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 shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
17 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
21 */
22
23#ifndef __NV50_IR_DRIVER_H__
24#define __NV50_IR_DRIVER_H__
25
26#include "pipe/p_shader_tokens.h"
27
28#include "tgsi/tgsi_util.h"
29#include "tgsi/tgsi_parse.h"
30#include "tgsi/tgsi_scan.h"
31
32/*
33 * This struct constitutes linkage information in TGSI terminology.
34 *
35 * It is created by the code generator and handed to the pipe driver
36 * for input/output slot assignment.
37 */
38struct nv50_ir_varying
39{
40   uint8_t slot[4]; /* native slots for xyzw (addresses in 32-bit words) */
41
42   unsigned mask     : 4; /* vec4 mask */
43   unsigned linear   : 1; /* linearly interpolated if true (and not flat) */
44   unsigned flat     : 1;
45   unsigned sc       : 1; /* special colour interpolation mode (SHADE_MODEL) */
46   unsigned centroid : 1;
47   unsigned patch    : 1; /* patch constant value */
48   unsigned regular  : 1; /* driver-specific meaning (e.g. input in sreg) */
49   unsigned input    : 1; /* indicates direction of system values */
50   unsigned oread    : 1; /* true if output is read from parallel TCP */
51
52   ubyte id; /* TGSI register index */
53   ubyte sn; /* TGSI semantic name */
54   ubyte si; /* TGSI semantic index */
55};
56
57#ifdef DEBUG
58# define NV50_IR_DEBUG_BASIC     (1 << 0)
59# define NV50_IR_DEBUG_VERBOSE   (2 << 0)
60# define NV50_IR_DEBUG_REG_ALLOC (1 << 2)
61#else
62# define NV50_IR_DEBUG_BASIC     0
63# define NV50_IR_DEBUG_VERBOSE   0
64# define NV50_IR_DEBUG_REG_ALLOC 0
65#endif
66
67struct nv50_ir_prog_symbol
68{
69   uint32_t label;
70   uint32_t offset;
71};
72
73#define NVISA_GK104_CHIPSET    0xe0
74#define NVISA_GK20A_CHIPSET    0xea
75#define NVISA_GM107_CHIPSET    0x110
76#define NVISA_GM200_CHIPSET    0x120
77
78struct nv50_ir_prog_info
79{
80   uint16_t target; /* chipset (0x50, 0x84, 0xc0, ...) */
81
82   uint8_t type; /* PIPE_SHADER */
83
84   uint8_t optLevel; /* optimization level (0 to 3) */
85   uint8_t dbgFlags;
86   bool omitLineNum; /* only used for printing the prog when dbgFlags is set */
87
88   struct {
89      int16_t maxGPR;     /* may be -1 if none used */
90      int16_t maxOutput;
91      uint32_t tlsSpace;  /* required local memory per thread */
92      uint32_t smemSize;  /* required shared memory per block */
93      uint32_t *code;
94      uint32_t codeSize;
95      uint32_t instructions;
96      uint8_t sourceRep;  /* PIPE_SHADER_IR_* */
97      const void *source;
98      void *relocData;
99      void *fixupData;
100      struct nv50_ir_prog_symbol *syms;
101      uint16_t numSyms;
102   } bin;
103
104   struct nv50_ir_varying sv[PIPE_MAX_SHADER_INPUTS];
105   struct nv50_ir_varying in[PIPE_MAX_SHADER_INPUTS];
106   struct nv50_ir_varying out[PIPE_MAX_SHADER_OUTPUTS];
107   uint8_t numInputs;
108   uint8_t numOutputs;
109   uint8_t numPatchConstants; /* also included in numInputs/numOutputs */
110   uint8_t numSysVals;
111
112   struct {
113      uint32_t *buf;    /* for IMMEDIATE_ARRAY */
114      uint16_t bufSize; /* size of immediate array */
115      uint16_t count;   /* count of inline immediates */
116      uint32_t *data;   /* inline immediate data */
117      uint8_t *type;    /* for each vec4 (128 bit) */
118   } immd;
119
120   union {
121      struct {
122         uint32_t inputMask[4]; /* mask of attributes read (1 bit per scalar) */
123         bool usesDrawParameters;
124      } vp;
125      struct {
126         uint8_t inputPatchSize;
127         uint8_t outputPatchSize;
128         uint8_t partitioning;    /* PIPE_TESS_PART */
129         int8_t winding;          /* +1 (clockwise) / -1 (counter-clockwise) */
130         uint8_t domain;          /* PIPE_PRIM_{QUADS,TRIANGLES,LINES} */
131         uint8_t outputPrim;      /* PIPE_PRIM_{TRIANGLES,LINES,POINTS} */
132      } tp;
133      struct {
134         uint8_t inputPrim;
135         uint8_t outputPrim;
136         unsigned instanceCount;
137         unsigned maxVertices;
138      } gp;
139      struct {
140         unsigned numColourResults;
141         bool writesDepth;
142         bool earlyFragTests;
143         bool postDepthCoverage;
144         bool separateFragData;
145         bool usesDiscard;
146         bool persampleInvocation;
147         bool usesSampleMaskIn;
148         bool readsFramebuffer;
149         bool readsSampleLocations;
150      } fp;
151      struct {
152         uint32_t inputOffset; /* base address for user args */
153         uint32_t sharedOffset; /* reserved space in s[] */
154         uint32_t gridInfoBase;  /* base address for NTID,NCTAID */
155         uint16_t numThreads[3]; /* max number of threads */
156      } cp;
157   } prop;
158
159   uint8_t numBarriers;
160
161   struct {
162      uint8_t clipDistances;     /* number of clip distance outputs */
163      uint8_t cullDistances;     /* number of cull distance outputs */
164      int8_t genUserClip;        /* request user clip planes for ClipVertex */
165      uint8_t auxCBSlot;         /* driver constant buffer slot */
166      uint16_t ucpBase;          /* base address for UCPs */
167      uint16_t drawInfoBase;     /* base address for draw parameters */
168      uint16_t alphaRefBase;     /* base address for alpha test values */
169      uint8_t pointSize;         /* output index for PointSize */
170      uint8_t instanceId;        /* system value index of InstanceID */
171      uint8_t vertexId;          /* system value index of VertexID */
172      uint8_t edgeFlagIn;
173      uint8_t edgeFlagOut;
174      int8_t viewportId;         /* output index of ViewportIndex */
175      uint8_t fragDepth;         /* output index of FragDepth */
176      uint8_t sampleMask;        /* output index of SampleMask */
177      uint8_t backFaceColor[2];  /* input/output indices of back face colour */
178      uint8_t globalAccess;      /* 1 for read, 2 for wr, 3 for rw */
179      bool fp64;                 /* program uses fp64 math */
180      bool mul_zero_wins;        /* program wants for x*0 = 0 */
181      bool nv50styleSurfaces;    /* generate gX[] access for raw buffers */
182      uint16_t texBindBase;      /* base address for tex handles (nve4) */
183      uint16_t fbtexBindBase;    /* base address for fbtex handle (nve4) */
184      uint16_t suInfoBase;       /* base address for surface info (nve4) */
185      uint16_t bindlessBase;     /* base address for bindless image info (nve4) */
186      uint16_t bufInfoBase;      /* base address for buffer info */
187      uint16_t sampleInfoBase;   /* base address for sample positions */
188      uint8_t msInfoCBSlot;      /* cX[] used for multisample info */
189      uint16_t msInfoBase;       /* base address for multisample info */
190      uint16_t uboInfoBase;      /* base address for compute UBOs (gk104+) */
191   } io;
192
193   /* driver callback to assign input/output locations */
194   int (*assignSlots)(struct nv50_ir_prog_info *);
195
196   void *driverPriv;
197};
198
199#ifdef __cplusplus
200extern "C" {
201#endif
202
203extern int nv50_ir_generate_code(struct nv50_ir_prog_info *);
204
205extern void nv50_ir_relocate_code(void *relocData, uint32_t *code,
206                                  uint32_t codePos,
207                                  uint32_t libPos,
208                                  uint32_t dataPos);
209
210extern void
211nv50_ir_apply_fixups(void *fixupData, uint32_t *code,
212                     bool force_per_sample, bool flatshade,
213                     uint8_t alphatest);
214
215/* obtain code that will be shared among programs */
216extern void nv50_ir_get_target_library(uint32_t chipset,
217                                       const uint32_t **code, uint32_t *size);
218
219#ifdef __cplusplus
220}
221#endif
222
223#endif // __NV50_IR_DRIVER_H__
224