summaryrefslogtreecommitdiffstats
path: root/src/mesa/drivers/dri/i965/brw_eu.c
blob: 4efc1f45a60a77c4ee0c79d9a5c0f9a57d91928d (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
/*
 Copyright (C) Intel Corp.  2006.  All Rights Reserved.
 Intel funded Tungsten Graphics to
 develop this 3D driver.

 Permission is hereby granted, free of charge, to any person obtaining
 a copy of this software and associated documentation files (the
 "Software"), to deal in the Software without restriction, including
 without limitation the rights to use, copy, modify, merge, publish,
 distribute, sublicense, and/or sell copies of the Software, and to
 permit persons to whom the Software is furnished to do so, subject to
 the following conditions:

 The above copyright notice and this permission notice (including the
 next paragraph) shall be included in all copies or substantial
 portions of the Software.

 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

 **********************************************************************/
 /*
  * Authors:
  *   Keith Whitwell <keithw@vmware.com>
  */


#include "brw_context.h"
#include "brw_defines.h"
#include "brw_eu.h"

#include "glsl/ralloc.h"

/**
 * Converts a BRW_REGISTER_TYPE_* enum to a short string (F, UD, and so on).
 *
 * This is different than reg_encoding from brw_disasm.c in that it operates
 * on the abstract enum values, rather than the generation-specific encoding.
 */
const char *
brw_reg_type_letters(unsigned type)
{
   const char *names[] = {
      [BRW_REGISTER_TYPE_UD] = "UD",
      [BRW_REGISTER_TYPE_D]  = "D",
      [BRW_REGISTER_TYPE_UW] = "UW",
      [BRW_REGISTER_TYPE_W]  = "W",
      [BRW_REGISTER_TYPE_F]  = "F",
      [BRW_REGISTER_TYPE_UB] = "UB",
      [BRW_REGISTER_TYPE_B]  = "B",
      [BRW_REGISTER_TYPE_UV] = "UV",
      [BRW_REGISTER_TYPE_V]  = "V",
      [BRW_REGISTER_TYPE_VF] = "VF",
      [BRW_REGISTER_TYPE_DF] = "DF",
      [BRW_REGISTER_TYPE_HF] = "HF",
      [BRW_REGISTER_TYPE_UQ] = "UQ",
      [BRW_REGISTER_TYPE_Q]  = "Q",
   };
   assert(type <= BRW_REGISTER_TYPE_UQ);
   return names[type];
}

/* Returns the corresponding conditional mod for swapping src0 and
 * src1 in e.g. CMP.
 */
uint32_t
brw_swap_cmod(uint32_t cmod)
{
   switch (cmod) {
   case BRW_CONDITIONAL_Z:
   case BRW_CONDITIONAL_NZ:
      return cmod;
   case BRW_CONDITIONAL_G:
      return BRW_CONDITIONAL_L;
   case BRW_CONDITIONAL_GE:
      return BRW_CONDITIONAL_LE;
   case BRW_CONDITIONAL_L:
      return BRW_CONDITIONAL_G;
   case BRW_CONDITIONAL_LE:
      return BRW_CONDITIONAL_GE;
   default:
      return ~0;
   }
}


/* How does predicate control work when execution_size != 8?  Do I
 * need to test/set for 0xffff when execution_size is 16?
 */
void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value )
{
   p->current->header.predicate_control = BRW_PREDICATE_NONE;

   if (value != 0xff) {
      if (value != p->flag_value) {
	 brw_push_insn_state(p);
	 brw_MOV(p, brw_flag_reg(0, 0), brw_imm_uw(value));
	 p->flag_value = value;
	 brw_pop_insn_state(p);
      }

      p->current->header.predicate_control = BRW_PREDICATE_NORMAL;
   }
}

void brw_set_predicate_control( struct brw_compile *p, unsigned pc )
{
   p->current->header.predicate_control = pc;
}

void brw_set_predicate_inverse(struct brw_compile *p, bool predicate_inverse)
{
   p->current->header.predicate_inverse = predicate_inverse;
}

void brw_set_conditionalmod( struct brw_compile *p, unsigned conditional )
{
   p->current->header.destreg__conditionalmod = conditional;
}

void brw_set_flag_reg(struct brw_compile *p, int reg, int subreg)
{
   p->current->bits2.da1.flag_reg_nr = reg;
   p->current->bits2.da1.flag_subreg_nr = subreg;
}

void brw_set_access_mode( struct brw_compile *p, unsigned access_mode )
{
   p->current->header.access_mode = access_mode;
}

void
brw_set_compression_control(struct brw_compile *p,
			    enum brw_compression compression_control)
{
   p->compressed = (compression_control == BRW_COMPRESSION_COMPRESSED);

   if (p->brw->gen >= 6) {
      /* Since we don't use the SIMD32 support in gen6, we translate
       * the pre-gen6 compression control here.
       */
      switch (compression_control) {
      case BRW_COMPRESSION_NONE:
	 /* This is the "use the first set of bits of dmask/vmask/arf
	  * according to execsize" option.
	  */
	 p->current->header.compression_control = GEN6_COMPRESSION_1Q;
	 break;
      case BRW_COMPRESSION_2NDHALF:
	 /* For SIMD8, this is "use the second set of 8 bits." */
	 p->current->header.compression_control = GEN6_COMPRESSION_2Q;
	 break;
      case BRW_COMPRESSION_COMPRESSED:
	 /* For SIMD16 instruction compression, use the first set of 16 bits
	  * since we don't do SIMD32 dispatch.
	  */
	 p->current->header.compression_control = GEN6_COMPRESSION_1H;
	 break;
      default:
	 assert(!"not reached");
	 p->current->header.compression_control = GEN6_COMPRESSION_1H;
	 break;
      }
   } else {
      p->current->header.compression_control = compression_control;
   }
}

void brw_set_mask_control( struct brw_compile *p, unsigned value )
{
   p->current->header.mask_control = value;
}

void brw_set_saturate( struct brw_compile *p, bool enable )
{
   p->current->header.saturate = enable;
}

void brw_set_acc_write_control(struct brw_compile *p, unsigned value)
{
   if (p->brw->gen >= 6)
      p->current->header.acc_wr_control = value;
}

void brw_push_insn_state( struct brw_compile *p )
{
   assert(p->current != &p->stack[BRW_EU_MAX_INSN_STACK-1]);
   memcpy(p->current+1, p->current, sizeof(struct brw_instruction));
   p->compressed_stack[p->current - p->stack] = p->compressed;
   p->current++;
}

void brw_pop_insn_state( struct brw_compile *p )
{
   assert(p->current != p->stack);
   p->current--;
   p->compressed = p->compressed_stack[p->current - p->stack];
}


/***********************************************************************
 */
void
brw_init_compile(struct brw_context *brw, struct brw_compile *p, void *mem_ctx)
{
   memset(p, 0, sizeof(*p));

   p->brw = brw;
   /*
    * Set the initial instruction store array size to 1024, if found that
    * isn't enough, then it will double the store size at brw_next_insn()
    * until out of memory.
    */
   p->store_size = 1024;
   p->store = rzalloc_array(mem_ctx, struct brw_instruction, p->store_size);
   p->nr_insn = 0;
   p->current = p->stack;
   p->compressed = false;
   memset(p->current, 0, sizeof(p->current[0]));

   p->mem_ctx = mem_ctx;

   /* Some defaults?
    */
   brw_set_mask_control(p, BRW_MASK_ENABLE); /* what does this do? */
   brw_set_saturate(p, 0);
   brw_set_compression_control(p, BRW_COMPRESSION_NONE);
   brw_set_predicate_control_flag_value(p, 0xff);

   /* Set up control flow stack */
   p->if_stack_depth = 0;
   p->if_stack_array_size = 16;
   p->if_stack = rzalloc_array(mem_ctx, int, p->if_stack_array_size);

   p->loop_stack_depth = 0;
   p->loop_stack_array_size = 16;
   p->loop_stack = rzalloc_array(mem_ctx, int, p->loop_stack_array_size);
   p->if_depth_in_loop = rzalloc_array(mem_ctx, int, p->loop_stack_array_size);

   brw_init_compaction_tables(brw);
}


const unsigned *brw_get_program( struct brw_compile *p,
			       unsigned *sz )
{
   *sz = p->next_insn_offset;
   return (const unsigned *)p->store;
}

void
brw_disassemble(struct brw_context *brw,
                void *assembly, int start, int end, FILE *out)
{
   bool dump_hex = false;

   for (int offset = start; offset < end;) {
      struct brw_instruction *insn = assembly + offset;
      struct brw_instruction uncompacted;
      bool compacted = insn->header.cmpt_control;
      fprintf(out, "0x%08x: ", offset);

      if (insn->header.cmpt_control) {
	 struct brw_compact_instruction *compacted = (void *)insn;
	 if (dump_hex) {
	    fprintf(out, "0x%08x 0x%08x                       ",
		    ((uint32_t *)insn)[1],
		    ((uint32_t *)insn)[0]);
	 }

	 brw_uncompact_instruction(brw, &uncompacted, compacted);
	 insn = &uncompacted;
	 offset += 8;
      } else {
	 if (dump_hex) {
	    fprintf(out, "0x%08x 0x%08x 0x%08x 0x%08x ",
		    ((uint32_t *)insn)[3],
		    ((uint32_t *)insn)[2],
		    ((uint32_t *)insn)[1],
		    ((uint32_t *)insn)[0]);
	 }
	 offset += 16;
      }

      brw_disassemble_inst(out, insn, brw->gen, compacted);
   }
}