2 * Copyright © 2016 Intel Corporation
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:
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
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
34 #include <sys/types.h>
40 #include "intel_aub.h"
41 #include "gen_disasm.h"
43 /* Below is the only command missing from intel_aub.h in libdrm
44 * So, reuse intel_aub.h from libdrm and #define the
45 * AUB_MI_BATCH_BUFFER_END as below
47 #define AUB_MI_BATCH_BUFFER_END (0x0500 << 16)
50 #define HEADER CSI "37;44m"
51 #define NORMAL CSI "0m"
52 #define CLEAR_TO_EOL CSI "0K"
56 static bool option_full_decode
= true;
57 static bool option_print_offsets
= true;
58 static enum { COLOR_AUTO
, COLOR_ALWAYS
, COLOR_NEVER
} option_color
;
62 struct gen_disasm
*disasm
;
64 uint64_t gtt_size
, gtt_end
;
66 uint64_t general_state_base
;
67 uint64_t surface_state_base
;
68 uint64_t dynamic_state_base
;
69 uint64_t instruction_base
;
70 uint64_t instruction_bound
;
72 static inline uint32_t
73 field(uint32_t value
, int start
, int end
)
77 mask
= ~0U >> (31 - end
+ start
);
79 return (value
>> start
) & mask
;
82 struct brw_instruction
;
85 valid_offset(uint32_t offset
)
87 return offset
< gtt_end
;
91 print_dword_val(struct gen_field_iterator
*iter
, uint64_t offset
, int *dword_num
)
99 f
= iter
->group
->fields
[iter
->i
- 1];
100 v
.dw
= iter
->p
[f
->start
/ 32];
102 if (*dword_num
!= (f
->start
/ 32)) {
103 printf("0x%08lx: 0x%08x : Dword %d\n",(offset
+ 4 * (f
->start
/ 32)), v
.dw
, f
->start
/ 32);
104 *dword_num
= (f
->start
/ 32);
109 print_iterator_values(struct gen_field_iterator
*iter
, int *idx
)
112 if (strstr(iter
->value
,"struct") == NULL
) {
113 printf(" %s: %s\n", iter
->name
, iter
->value
);
115 token
= strtok(iter
->value
, " ");
117 token
= strtok(NULL
, " ");
118 *idx
= atoi(strtok(NULL
, ">"));
122 printf(" %s:<struct %s>\n", iter
->name
, token
);
128 decode_structure(struct gen_spec
*spec
, struct gen_group
*strct
, const uint32_t *p
)
130 struct gen_field_iterator iter
;
132 int idx
= 0, dword_num
= 0;
135 if (option_print_offsets
)
136 offset
= (void *) p
- gtt
;
140 gen_field_iterator_init(&iter
, strct
, p
);
141 while (gen_field_iterator_next(&iter
)) {
143 print_dword_val(&iter
, offset
, &dword_num
);
144 token
= print_iterator_values(&iter
, &idx
);
146 struct gen_group
*struct_val
= gen_spec_find_struct(spec
, token
);
147 decode_structure(spec
, struct_val
, &p
[idx
]);
154 handle_struct_decode(struct gen_spec
*spec
, char *struct_name
, uint32_t *p
)
156 if (struct_name
== NULL
)
158 struct gen_group
*struct_val
= gen_spec_find_struct(spec
, struct_name
);
159 decode_structure(spec
, struct_val
, p
);
163 dump_binding_table(struct gen_spec
*spec
, uint32_t offset
)
165 uint32_t *pointers
, i
;
167 struct gen_group
*surface_state
;
169 surface_state
= gen_spec_find_struct(spec
, "RENDER_SURFACE_STATE");
170 if (surface_state
== NULL
) {
171 printf("did not find RENDER_SURFACE_STATE info\n");
175 start
= surface_state_base
+ offset
;
176 pointers
= gtt
+ start
;
177 for (i
= 0; i
< 16; i
++) {
178 if (pointers
[i
] == 0)
180 start
= pointers
[i
] + surface_state_base
;
181 if (!valid_offset(start
)) {
182 printf("pointer %u: %08x <not valid>\n",
186 printf("pointer %u: %08x\n", i
, pointers
[i
]);
189 decode_structure(spec
, surface_state
, gtt
+ start
);
194 handle_3dstate_index_buffer(struct gen_spec
*spec
, uint32_t *p
)
197 uint32_t length
, i
, type
, size
;
200 type
= (p
[1] >> 8) & 3;
202 length
= p
[4] / size
;
208 for (i
= 0; i
< length
; i
++) {
211 printf("%3d ", ((uint8_t *)start
)[i
]);
214 printf("%3d ", ((uint16_t *)start
)[i
]);
217 printf("%3d ", ((uint32_t *)start
)[i
]);
221 if (length
< p
[4] / size
)
227 static inline uint64_t
228 get_qword(uint32_t *p
)
230 return ((uint64_t) p
[1] << 32) | p
[0];
234 handle_state_base_address(struct gen_spec
*spec
, uint32_t *p
)
236 uint64_t mask
= ~((1 << 12) - 1);
238 if (gen_spec_get_gen(spec
) >= gen_make_gen(8,0)) {
240 general_state_base
= get_qword(&p
[1]) & mask
;
242 surface_state_base
= get_qword(&p
[4]) & mask
;
244 dynamic_state_base
= get_qword(&p
[6]) & mask
;
246 instruction_base
= get_qword(&p
[10]) & mask
;
248 instruction_bound
= p
[15] & mask
;
251 surface_state_base
= p
[2] & mask
;
253 dynamic_state_base
= p
[3] & mask
;
255 instruction_base
= p
[5] & mask
;
257 instruction_bound
= p
[9] & mask
;
262 dump_samplers(struct gen_spec
*spec
, uint32_t offset
)
266 struct gen_group
*sampler_state
;
268 sampler_state
= gen_spec_find_struct(spec
, "SAMPLER_STATE");
270 start
= dynamic_state_base
+ offset
;
271 for (i
= 0; i
< 4; i
++) {
272 printf("sampler state %d\n", i
);
273 decode_structure(spec
, sampler_state
, gtt
+ start
+ i
* 16);
278 handle_media_interface_descriptor_load(struct gen_spec
*spec
, uint32_t *p
)
280 int i
, length
= p
[2] / 32;
281 struct gen_group
*descriptor_structure
;
282 uint32_t *descriptors
;
284 struct brw_instruction
*insns
;
286 descriptor_structure
=
287 gen_spec_find_struct(spec
, "INTERFACE_DESCRIPTOR_DATA");
288 if (descriptor_structure
== NULL
) {
289 printf("did not find INTERFACE_DESCRIPTOR_DATA info\n");
293 start
= dynamic_state_base
+ p
[3];
294 descriptors
= gtt
+ start
;
295 for (i
= 0; i
< length
; i
++, descriptors
+= 8) {
296 printf("descriptor %u: %08x\n", i
, *descriptors
);
297 decode_structure(spec
, descriptor_structure
, descriptors
);
299 start
= instruction_base
+ descriptors
[0];
300 if (!valid_offset(start
)) {
301 printf("kernel: %08lx <not valid>\n", start
);
304 printf("kernel: %08lx\n", start
);
307 insns
= (struct brw_instruction
*) (gtt
+ start
);
308 gen_disasm_disassemble(disasm
, insns
, 0, 8192, stdout
);
310 dump_samplers(spec
, descriptors
[3] & ~0x1f);
311 dump_binding_table(spec
, descriptors
[4] & ~0x1f);
315 /* Heuristic to determine whether a uint32_t is probably actually a float
316 * (http://stackoverflow.com/a/2953466)
320 probably_float(uint32_t bits
)
322 int exp
= ((bits
& 0x7f800000U
) >> 23) - 127;
323 uint32_t mant
= bits
& 0x007fffff;
326 if (exp
== -127 && mant
== 0)
329 /* +- 1 billionth to 1 billion */
330 if (-30 <= exp
&& exp
<= 30)
333 /* some value with only a few binary digits */
334 if ((mant
& 0x0000ffff) == 0)
341 handle_3dstate_vertex_buffers(struct gen_spec
*spec
, uint32_t *p
)
343 uint32_t *end
, *s
, *dw
, *dwend
;
345 int n
, i
, count
, stride
;
347 end
= (p
[0] & 0xff) + p
+ 2;
348 for (s
= &p
[1], n
= 0; s
< end
; s
+= 4, n
++) {
349 if (gen_spec_get_gen(spec
) >= gen_make_gen(8, 0)) {
350 offset
= *(uint64_t *) &s
[1];
351 dwend
= gtt
+ offset
+ s
[3];
354 dwend
= gtt
+ s
[2] + 1;
357 stride
= field(s
[0], 0, 11);
359 printf("vertex buffer %d, size %d\n", n
, s
[3]);
360 for (dw
= gtt
+ offset
, i
= 0; dw
< dwend
&& i
< 256; dw
++) {
361 if (count
== 0 && count
% (8 * 4) == 0)
364 if (probably_float(*dw
))
365 printf(" %8.2f", *(float *) dw
);
367 printf(" 0x%08x", *dw
);
372 if (count
== stride
) {
375 } else if (count
% (8 * 4) == 0) {
381 if (count
> 0 && count
% (8 * 4) != 0)
387 handle_3dstate_vs(struct gen_spec
*spec
, uint32_t *p
)
390 struct brw_instruction
*insns
;
393 if (gen_spec_get_gen(spec
) >= gen_make_gen(8, 0)) {
394 start
= get_qword(&p
[1]);
395 vs_enable
= p
[7] & 1;
398 vs_enable
= p
[5] & 1;
402 printf("instruction_base %08lx, start %08lx\n",
403 instruction_base
, start
);
405 insns
= (struct brw_instruction
*) (gtt
+ instruction_base
+ start
);
406 gen_disasm_disassemble(disasm
, insns
, 0, 8192, stdout
);
411 handle_3dstate_hs(struct gen_spec
*spec
, uint32_t *p
)
414 struct brw_instruction
*insns
;
417 if (gen_spec_get_gen(spec
) >= gen_make_gen(8, 0)) {
418 start
= get_qword(&p
[4]);
423 hs_enable
= p
[2] & 0x80000000;
426 printf("instruction_base %08lx, start %08lx\n",
427 instruction_base
, start
);
429 insns
= (struct brw_instruction
*) (gtt
+ instruction_base
+ start
);
430 gen_disasm_disassemble(disasm
, insns
, 0, 8192, stdout
);
435 handle_3dstate_constant(struct gen_spec
*spec
, uint32_t *p
)
441 for (i
= 0; i
< 4; i
++) {
442 length
= (p
[1 + i
/ 2] >> (i
& 1) * 16) & 0xffff;
443 f
= (float *) (gtt
+ p
[3 + i
* 2] + dynamic_state_base
);
445 for (j
= 0; j
< length
* 8; j
++) {
446 if (probably_float(dw
[j
]))
447 printf(" %04.3f", f
[j
]);
449 printf(" 0x%08x", dw
[j
]);
458 handle_3dstate_ps(struct gen_spec
*spec
, uint32_t *p
)
460 uint32_t mask
= ~((1 << 6) - 1);
462 struct brw_instruction
*insns
;
463 static const char unused
[] = "unused";
464 static const char *pixel_type
[3] = {"8 pixel", "16 pixel", "32 pixel"};
465 const char *k0
, *k1
, *k2
;
466 uint32_t k_mask
, k1_offset
, k2_offset
;
468 if (gen_spec_get_gen(spec
) >= gen_make_gen(8, 0)) {
479 #define DISPATCH_16 2
480 #define DISPATCH_32 4
493 case DISPATCH_8
| DISPATCH_16
:
503 case DISPATCH_16
| DISPATCH_32
:
508 case DISPATCH_8
| DISPATCH_16
| DISPATCH_32
:
520 start
= instruction_base
+ (p
[1] & mask
);
521 printf(" Kernel[0] %s\n", k0
);
523 insns
= (struct brw_instruction
*) (gtt
+ start
);
524 gen_disasm_disassemble(disasm
, insns
, 0, 8192, stdout
);
527 start
= instruction_base
+ (p
[k1_offset
] & mask
);
528 printf(" Kernel[1] %s\n", k1
);
530 insns
= (struct brw_instruction
*) (gtt
+ start
);
531 gen_disasm_disassemble(disasm
, insns
, 0, 8192, stdout
);
534 start
= instruction_base
+ (p
[k2_offset
] & mask
);
535 printf(" Kernel[2] %s\n", k2
);
537 insns
= (struct brw_instruction
*) (gtt
+ start
);
538 gen_disasm_disassemble(disasm
, insns
, 0, 8192, stdout
);
543 handle_3dstate_binding_table_pointers(struct gen_spec
*spec
, uint32_t *p
)
545 dump_binding_table(spec
, p
[1]);
549 handle_3dstate_sampler_state_pointers(struct gen_spec
*spec
, uint32_t *p
)
551 dump_samplers(spec
, p
[1]);
555 handle_3dstate_viewport_state_pointers_cc(struct gen_spec
*spec
, uint32_t *p
)
558 struct gen_group
*cc_viewport
;
560 cc_viewport
= gen_spec_find_struct(spec
, "CC_VIEWPORT");
562 start
= dynamic_state_base
+ (p
[1] & ~0x1fu
);
563 for (uint32_t i
= 0; i
< 4; i
++) {
564 printf("viewport %d\n", i
);
565 decode_structure(spec
, cc_viewport
, gtt
+ start
+ i
* 8);
570 handle_3dstate_viewport_state_pointers_sf_clip(struct gen_spec
*spec
, uint32_t *p
)
573 struct gen_group
*sf_clip_viewport
;
575 sf_clip_viewport
= gen_spec_find_struct(spec
, "SF_CLIP_VIEWPORT");
577 start
= dynamic_state_base
+ (p
[1] & ~0x3fu
);
578 for (uint32_t i
= 0; i
< 4; i
++) {
579 printf("viewport %d\n", i
);
580 decode_structure(spec
, sf_clip_viewport
, gtt
+ start
+ i
* 64);
585 handle_3dstate_blend_state_pointers(struct gen_spec
*spec
, uint32_t *p
)
588 struct gen_group
*blend_state
;
590 blend_state
= gen_spec_find_struct(spec
, "BLEND_STATE");
592 start
= dynamic_state_base
+ (p
[1] & ~0x3fu
);
593 decode_structure(spec
, blend_state
, gtt
+ start
);
597 handle_3dstate_cc_state_pointers(struct gen_spec
*spec
, uint32_t *p
)
600 struct gen_group
*cc_state
;
602 cc_state
= gen_spec_find_struct(spec
, "COLOR_CALC_STATE");
604 start
= dynamic_state_base
+ (p
[1] & ~0x3fu
);
605 decode_structure(spec
, cc_state
, gtt
+ start
);
609 handle_3dstate_scissor_state_pointers(struct gen_spec
*spec
, uint32_t *p
)
612 struct gen_group
*scissor_rect
;
614 scissor_rect
= gen_spec_find_struct(spec
, "SCISSOR_RECT");
616 start
= dynamic_state_base
+ (p
[1] & ~0x1fu
);
617 decode_structure(spec
, scissor_rect
, gtt
+ start
);
620 #define ARRAY_LENGTH(a) (sizeof (a) / sizeof (a)[0])
622 #define STATE_BASE_ADDRESS 0x61010000
624 #define MEDIA_INTERFACE_DESCRIPTOR_LOAD 0x70020000
626 #define _3DSTATE_INDEX_BUFFER 0x780a0000
627 #define _3DSTATE_VERTEX_BUFFERS 0x78080000
629 #define _3DSTATE_VS 0x78100000
630 #define _3DSTATE_GS 0x78110000
631 #define _3DSTATE_HS 0x781b0000
632 #define _3DSTATE_DS 0x781d0000
634 #define _3DSTATE_CONSTANT_VS 0x78150000
635 #define _3DSTATE_CONSTANT_GS 0x78160000
636 #define _3DSTATE_CONSTANT_PS 0x78170000
637 #define _3DSTATE_CONSTANT_HS 0x78190000
638 #define _3DSTATE_CONSTANT_DS 0x781A0000
640 #define _3DSTATE_PS 0x78200000
642 #define _3DSTATE_BINDING_TABLE_POINTERS_VS 0x78260000
643 #define _3DSTATE_BINDING_TABLE_POINTERS_HS 0x78270000
644 #define _3DSTATE_BINDING_TABLE_POINTERS_DS 0x78280000
645 #define _3DSTATE_BINDING_TABLE_POINTERS_GS 0x78290000
646 #define _3DSTATE_BINDING_TABLE_POINTERS_PS 0x782a0000
648 #define _3DSTATE_SAMPLER_STATE_POINTERS_VS 0x782b0000
649 #define _3DSTATE_SAMPLER_STATE_POINTERS_GS 0x782e0000
650 #define _3DSTATE_SAMPLER_STATE_POINTERS_PS 0x782f0000
652 #define _3DSTATE_VIEWPORT_STATE_POINTERS_CC 0x78230000
653 #define _3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP 0x78210000
654 #define _3DSTATE_BLEND_STATE_POINTERS 0x78240000
655 #define _3DSTATE_CC_STATE_POINTERS 0x780e0000
656 #define _3DSTATE_SCISSOR_STATE_POINTERS 0x780f0000
658 struct custom_handler
{
660 void (*handle
)(struct gen_spec
*spec
, uint32_t *p
);
661 } custom_handlers
[] = {
662 { STATE_BASE_ADDRESS
, handle_state_base_address
},
663 { MEDIA_INTERFACE_DESCRIPTOR_LOAD
, handle_media_interface_descriptor_load
},
664 { _3DSTATE_VERTEX_BUFFERS
, handle_3dstate_vertex_buffers
},
665 { _3DSTATE_INDEX_BUFFER
, handle_3dstate_index_buffer
},
666 { _3DSTATE_VS
, handle_3dstate_vs
},
667 { _3DSTATE_GS
, handle_3dstate_vs
},
668 { _3DSTATE_DS
, handle_3dstate_vs
},
669 { _3DSTATE_HS
, handle_3dstate_hs
},
670 { _3DSTATE_CONSTANT_VS
, handle_3dstate_constant
},
671 { _3DSTATE_CONSTANT_GS
, handle_3dstate_constant
},
672 { _3DSTATE_CONSTANT_PS
, handle_3dstate_constant
},
673 { _3DSTATE_CONSTANT_HS
, handle_3dstate_constant
},
674 { _3DSTATE_CONSTANT_DS
, handle_3dstate_constant
},
675 { _3DSTATE_PS
, handle_3dstate_ps
},
677 { _3DSTATE_BINDING_TABLE_POINTERS_VS
, handle_3dstate_binding_table_pointers
},
678 { _3DSTATE_BINDING_TABLE_POINTERS_HS
, handle_3dstate_binding_table_pointers
},
679 { _3DSTATE_BINDING_TABLE_POINTERS_DS
, handle_3dstate_binding_table_pointers
},
680 { _3DSTATE_BINDING_TABLE_POINTERS_GS
, handle_3dstate_binding_table_pointers
},
681 { _3DSTATE_BINDING_TABLE_POINTERS_PS
, handle_3dstate_binding_table_pointers
},
683 { _3DSTATE_SAMPLER_STATE_POINTERS_VS
, handle_3dstate_sampler_state_pointers
},
684 { _3DSTATE_SAMPLER_STATE_POINTERS_GS
, handle_3dstate_sampler_state_pointers
},
685 { _3DSTATE_SAMPLER_STATE_POINTERS_PS
, handle_3dstate_sampler_state_pointers
},
687 { _3DSTATE_VIEWPORT_STATE_POINTERS_CC
, handle_3dstate_viewport_state_pointers_cc
},
688 { _3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP
, handle_3dstate_viewport_state_pointers_sf_clip
},
689 { _3DSTATE_BLEND_STATE_POINTERS
, handle_3dstate_blend_state_pointers
},
690 { _3DSTATE_CC_STATE_POINTERS
, handle_3dstate_cc_state_pointers
},
691 { _3DSTATE_SCISSOR_STATE_POINTERS
, handle_3dstate_scissor_state_pointers
}
695 parse_commands(struct gen_spec
*spec
, uint32_t *cmds
, int size
, int engine
)
697 uint32_t *p
, *end
= cmds
+ size
/ 4;
698 unsigned int length
, i
;
699 struct gen_group
*inst
;
701 for (p
= cmds
; p
< end
; p
+= length
) {
702 inst
= gen_spec_find_instruction(spec
, p
);
704 printf("unknown instruction %08x\n", p
[0]);
705 length
= (p
[0] & 0xff) + 2;
708 length
= gen_group_get_length(inst
, p
);
710 const char *color
, *reset_color
= CLEAR_TO_EOL NORMAL
;
713 if (option_full_decode
)
718 if (option_color
== COLOR_NEVER
) {
723 if (option_print_offsets
)
724 offset
= (void *) p
- gtt
;
728 printf("%s0x%08lx: 0x%08x: %s%s\n",
730 gen_group_get_name(inst
), reset_color
);
732 if (option_full_decode
) {
733 struct gen_field_iterator iter
;
735 int idx
= 0, dword_num
= 0;
736 gen_field_iterator_init(&iter
, inst
, p
);
737 while (gen_field_iterator_next(&iter
)) {
739 print_dword_val(&iter
, offset
, &dword_num
);
741 token
= print_iterator_values(&iter
, &idx
);
743 printf("0x%08lx: 0x%08x : Dword %d\n",(offset
+4*idx
), p
[idx
], idx
);
744 handle_struct_decode(spec
,token
, &p
[idx
]);
749 for (i
= 0; i
< ARRAY_LENGTH(custom_handlers
); i
++) {
750 if (gen_group_get_opcode(inst
) ==
751 custom_handlers
[i
].opcode
)
752 custom_handlers
[i
].handle(spec
, p
);
756 if ((p
[0] & 0xffff0000) == AUB_MI_BATCH_BUFFER_START
) {
758 if (gen_spec_get_gen(spec
) >= gen_make_gen(8,0))
759 start
= get_qword(&p
[1]);
763 parse_commands(spec
, gtt
+ start
, 1 << 20, engine
);
764 } else if ((p
[0] & 0xffff0000) == AUB_MI_BATCH_BUFFER_END
) {
770 #define GEN_ENGINE_RENDER 1
771 #define GEN_ENGINE_BLITTER 2
774 handle_trace_block(struct gen_spec
*spec
, uint32_t *p
)
776 int operation
= p
[1] & AUB_TRACE_OPERATION_MASK
;
777 int type
= p
[1] & AUB_TRACE_TYPE_MASK
;
778 int address_space
= p
[1] & AUB_TRACE_ADDRESS_SPACE_MASK
;
779 uint64_t offset
= p
[3];
780 uint32_t size
= p
[4];
781 int header_length
= p
[0] & 0xffff;
782 uint32_t *data
= p
+ header_length
+ 2;
783 int engine
= GEN_ENGINE_RENDER
;
785 if (gen_spec_get_gen(spec
) >= gen_make_gen(8,0))
786 offset
+= (uint64_t) p
[5] << 32;
789 case AUB_TRACE_OP_DATA_WRITE
:
790 if (address_space
!= AUB_TRACE_MEMTYPE_GTT
)
792 if (gtt_size
< offset
+ size
)
793 error(EXIT_FAILURE
, errno
, "overflow gtt space");
794 memcpy((char *) gtt
+ offset
, data
, size
);
795 if (gtt_end
< offset
+ size
)
796 gtt_end
= offset
+ size
;
798 case AUB_TRACE_OP_COMMAND_WRITE
:
800 case AUB_TRACE_TYPE_RING_PRB0
:
801 engine
= GEN_ENGINE_RENDER
;
803 case AUB_TRACE_TYPE_RING_PRB2
:
804 engine
= GEN_ENGINE_BLITTER
;
807 printf("command write to unknown ring %d\n", type
);
811 parse_commands(spec
, data
, size
, engine
);
821 uint32_t *map
, *end
, *cursor
;
824 static struct aub_file
*
825 aub_file_open(const char *filename
)
827 struct aub_file
*file
;
829 file
= malloc(sizeof *file
);
830 file
->filename
= strdup(filename
);
831 file
->fd
= open(file
->filename
, O_RDONLY
);
833 error(EXIT_FAILURE
, errno
, "open %s failed", file
->filename
);
835 if (fstat(file
->fd
, &file
->sb
) == -1)
836 error(EXIT_FAILURE
, errno
, "stat failed");
838 file
->map
= mmap(NULL
, file
->sb
.st_size
,
839 PROT_READ
, MAP_SHARED
, file
->fd
, 0);
840 if (file
->map
== MAP_FAILED
)
841 error(EXIT_FAILURE
, errno
, "mmap failed");
843 file
->cursor
= file
->map
;
844 file
->end
= file
->map
+ file
->sb
.st_size
/ 4;
846 /* mmap a terabyte for our gtt space. */
847 gtt_size
= 1ul << 40;
848 gtt
= mmap(NULL
, gtt_size
, PROT_READ
| PROT_WRITE
,
849 MAP_PRIVATE
| MAP_ANONYMOUS
| MAP_NORESERVE
, -1, 0);
850 if (gtt
== MAP_FAILED
)
851 error(EXIT_FAILURE
, errno
, "failed to alloc gtt space");
856 #define TYPE(dw) (((dw) >> 29) & 7)
857 #define OPCODE(dw) (((dw) >> 23) & 0x3f)
858 #define SUBOPCODE(dw) (((dw) >> 16) & 0x7f)
860 #define MAKE_HEADER(type, opcode, subopcode) \
861 (((type) << 29) | ((opcode) << 23) | ((subopcode) << 16))
865 /* Classic AUB opcodes */
866 #define OPCODE_AUB 0x01
867 #define SUBOPCODE_HEADER 0x05
868 #define SUBOPCODE_BLOCK 0x41
869 #define SUBOPCODE_BMP 0x1e
871 /* Newer version AUB opcode*/
872 #define OPCODE_NEW_AUB 0x2e
873 #define SUBOPCODE_VERSION 0x00
874 #define SUBOPCODE_REG_WRITE 0x03
875 #define SUBOPCODE_MEM_POLL 0x05
876 #define SUBOPCODE_MEM_WRITE 0x06
878 #define MAKE_GEN(major, minor) ( ((major) << 8) | (minor) )
884 { "bwr", MAKE_GEN(4, 0) },
885 { "cln", MAKE_GEN(4, 0) },
886 { "blc", MAKE_GEN(4, 0) },
887 { "ctg", MAKE_GEN(4, 0) },
888 { "el", MAKE_GEN(4, 0) },
889 { "il", MAKE_GEN(4, 0) },
890 { "sbr", MAKE_GEN(6, 0) },
891 { "ivb", MAKE_GEN(7, 0) },
892 { "lrb2", MAKE_GEN(0, 0) },
893 { "hsw", MAKE_GEN(7, 5) },
894 { "vlv", MAKE_GEN(7, 0) },
895 { "bdw", MAKE_GEN(8, 0) },
896 { "skl", MAKE_GEN(9, 0) },
897 { "chv", MAKE_GEN(8, 0) },
898 { "bxt", MAKE_GEN(9, 0) }
902 aub_file_decode_batch(struct aub_file
*file
, struct gen_spec
*spec
)
904 uint32_t *p
, h
, device
, data_type
;
905 int header_length
, payload_size
, bias
;
909 header_length
= h
& 0xffff;
919 printf("unknown opcode %d at %ld/%ld\n",
920 OPCODE(h
), file
->cursor
- file
->map
,
921 file
->end
- file
->map
);
922 file
->cursor
= file
->end
;
927 switch (h
& 0xffff0000) {
928 case MAKE_HEADER(TYPE_AUB
, OPCODE_AUB
, SUBOPCODE_HEADER
):
929 payload_size
= p
[12];
931 case MAKE_HEADER(TYPE_AUB
, OPCODE_AUB
, SUBOPCODE_BLOCK
):
933 handle_trace_block(spec
, p
);
935 case MAKE_HEADER(TYPE_AUB
, OPCODE_AUB
, SUBOPCODE_BMP
):
938 case MAKE_HEADER(TYPE_AUB
, OPCODE_NEW_AUB
, SUBOPCODE_VERSION
):
939 printf("version block: dw1 %08x\n", p
[1]);
940 device
= (p
[1] >> 8) & 0xff;
941 printf(" device %s\n", device_map
[device
].name
);
943 case MAKE_HEADER(TYPE_AUB
, OPCODE_NEW_AUB
, SUBOPCODE_REG_WRITE
):
944 printf("register write block: (dwords %d)\n", h
& 0xffff);
945 printf(" reg 0x%x, data 0x%x\n", p
[1], p
[5]);
947 case MAKE_HEADER(TYPE_AUB
, OPCODE_NEW_AUB
, SUBOPCODE_MEM_WRITE
):
948 printf("memory write block (dwords %d):\n", h
& 0xffff);
949 printf(" address 0x%lx\n", *(uint64_t *) &p
[1]);
950 data_type
= (p
[3] >> 20) & 0xff;
952 printf(" data type 0x%x\n", data_type
);
953 printf(" address space 0x%x\n", (p
[3] >> 28) & 0xf);
955 case MAKE_HEADER(TYPE_AUB
, OPCODE_NEW_AUB
, SUBOPCODE_MEM_POLL
):
956 printf("memory poll block (dwords %d):\n", h
& 0xffff);
959 printf("unknown block type=0x%x, opcode=0x%x, "
960 "subopcode=0x%x (%08x)\n", TYPE(h
), OPCODE(h
), SUBOPCODE(h
), h
);
963 file
->cursor
= p
+ header_length
+ bias
+ payload_size
/ 4;
967 aub_file_more_stuff(struct aub_file
*file
)
969 return file
->cursor
< file
->end
;
991 execlp("less", "less", "-rFi", NULL
);
1000 print_help(FILE *file
)
1003 "Usage: %s [OPTION]... FILE\n"
1004 "Decode aub file contents.\n\n"
1005 "A valid --gen option must be provided.\n\n"
1006 " --help display this help and exit\n"
1007 " --gen=platform decode for given platform (ivb, byt, hsw, bdw, chv, skl, kbl or bxt)\n"
1008 " --headers decode only command headers\n"
1009 " --color[=WHEN] colorize the output; WHEN can be 'auto' (default\n"
1010 " if omitted), 'always', or 'never'\n"
1011 " --no-pager don't launch pager\n"
1012 " --no-offsets don't print instruction offsets\n",
1013 basename(program_invocation_name
));
1017 is_prefix(const char *arg
, const char *prefix
, const char **value
)
1019 int l
= strlen(prefix
);
1021 if (strncmp(arg
, prefix
, l
) == 0 && (arg
[l
] == '\0' || arg
[l
] == '=')) {
1023 *value
= arg
+ l
+ 1;
1033 int main(int argc
, char *argv
[])
1035 struct gen_spec
*spec
;
1036 struct aub_file
*file
;
1038 bool found_arg_gen
= false, pager
= true;
1039 int gen_major
, gen_minor
;
1041 char gen_file
[256], gen_val
[24];
1048 for (i
= 1; i
< argc
; ++i
) {
1049 if (strcmp(argv
[i
], "--no-pager") == 0) {
1051 } else if (strcmp(argv
[i
], "--no-offsets") == 0) {
1052 option_print_offsets
= false;
1053 } else if (is_prefix(argv
[i
], "--gen", &value
)) {
1055 error(EXIT_FAILURE
, 0, "option '--gen' requires an argument\n");
1056 found_arg_gen
= true;
1059 snprintf(gen_val
, sizeof(gen_val
), "%s", value
);
1060 } else if (strcmp(argv
[i
], "--headers") == 0) {
1061 option_full_decode
= false;
1062 } else if (is_prefix(argv
[i
], "--color", &value
)) {
1063 if (value
== NULL
|| strcmp(value
, "always") == 0)
1064 option_color
= COLOR_ALWAYS
;
1065 else if (strcmp(value
, "never") == 0)
1066 option_color
= COLOR_NEVER
;
1067 else if (strcmp(value
, "auto") == 0)
1068 option_color
= COLOR_AUTO
;
1070 error(EXIT_FAILURE
, 0, "invalid value for --color: %s", value
);
1071 } else if (strcmp(argv
[i
], "--help") == 0) {
1075 if (argv
[i
][0] == '-') {
1076 fprintf(stderr
, "unknown option %s\n", argv
[i
]);
1083 if (!found_arg_gen
) {
1084 fprintf(stderr
, "argument --gen is required\n");
1088 if (strstr(gen_val
,"ivb") != NULL
) {
1089 /* Intel(R) Ivybridge Mobile GT2 */
1093 } else if (strstr(gen_val
,"hsw") != NULL
) {
1094 /* Intel(R) Haswell Mobile GT2 */
1098 } else if (strstr(gen_val
,"byt") != NULL
) {
1099 /* Intel(R) Bay Trail */
1103 } else if (strstr(gen_val
,"bdw") != NULL
) {
1104 /* Intel(R) HD Graphics 5500 (Broadwell GT2) */
1108 } else if (strstr(gen_val
,"chv") != NULL
) {
1109 /* Intel(R) HD Graphics (Cherryview) */
1113 } else if (strstr(gen_val
,"skl") != NULL
) {
1114 /* Intel(R) HD Graphics 530 (Skylake GT2) */
1118 } else if (strstr(gen_val
,"kbl") != NULL
) {
1119 /* Intel(R) Kabylake GT2 */
1123 } else if (strstr(gen_val
,"bxt") != NULL
) {
1124 /* Intel(R) HD Graphics (Broxton) */
1129 error(EXIT_FAILURE
, 0, "can't parse gen: %s, expected ivb, byt, hsw, bdw, chv, skl, kbl or bxt\n", gen_val
);
1132 /* Do this before we redirect stdout to pager. */
1133 if (option_color
== COLOR_AUTO
)
1134 option_color
= isatty(1) ? COLOR_ALWAYS
: COLOR_NEVER
;
1136 if (isatty(1) && pager
)
1140 snprintf(gen_file
, sizeof(gen_file
), "../genxml/gen%d%d.xml", gen_major
, gen_minor
);
1142 snprintf(gen_file
, sizeof(gen_file
), "../genxml/gen%d.xml", gen_major
);
1144 spec
= gen_spec_load(gen_file
);
1145 disasm
= gen_disasm_create(pci_id
);
1147 if (argv
[i
] == NULL
) {
1151 file
= aub_file_open(argv
[i
]);
1154 while (aub_file_more_stuff(file
))
1155 aub_file_decode_batch(file
, spec
);
1158 /* close the stdout which is opened to write the output */
1163 return EXIT_SUCCESS
;