Home | History | Annotate | Download | only in intel
      1 /*
      2  * Copyright  2009-2011 Intel Corporation
      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 #ifdef HAVE_CONFIG_H
     25 #include "config.h"
     26 #endif
     27 
     28 #include <assert.h>
     29 #include <stdint.h>
     30 #include <stdlib.h>
     31 #include <stdio.h>
     32 #include <stdbool.h>
     33 #include <stdarg.h>
     34 #include <string.h>
     35 
     36 #include "libdrm.h"
     37 #include "xf86drm.h"
     38 #include "intel_chipset.h"
     39 #include "intel_bufmgr.h"
     40 
     41 /* Struct for tracking drm_intel_decode state. */
     42 struct drm_intel_decode {
     43 	/** stdio file where the output should land.  Defaults to stdout. */
     44 	FILE *out;
     45 
     46 	/** PCI device ID. */
     47 	uint32_t devid;
     48 
     49 	/**
     50 	 * Shorthand device identifier: 3 is 915, 4 is 965, 5 is
     51 	 * Ironlake, etc.
     52 	 */
     53 	int gen;
     54 
     55 	/** GPU address of the start of the current packet. */
     56 	uint32_t hw_offset;
     57 	/** CPU virtual address of the start of the current packet. */
     58 	uint32_t *data;
     59 	/** DWORDs of remaining batchbuffer data starting from the packet. */
     60 	uint32_t count;
     61 
     62 	/** GPU address of the start of the batchbuffer data. */
     63 	uint32_t base_hw_offset;
     64 	/** CPU Virtual address of the start of the batchbuffer data. */
     65 	uint32_t *base_data;
     66 	/** Number of DWORDs of batchbuffer data. */
     67 	uint32_t base_count;
     68 
     69 	/** @{
     70 	 * GPU head and tail pointers, which will be noted in the dump, or ~0.
     71 	 */
     72 	uint32_t head, tail;
     73 	/** @} */
     74 
     75 	/**
     76 	 * Whether to dump the dwords after MI_BATCHBUFFER_END.
     77 	 *
     78 	 * This sometimes provides clues in corrupted batchbuffers,
     79 	 * and is used by the intel-gpu-tools.
     80 	 */
     81 	bool dump_past_end;
     82 
     83 	bool overflowed;
     84 };
     85 
     86 static FILE *out;
     87 static uint32_t saved_s2 = 0, saved_s4 = 0;
     88 static char saved_s2_set = 0, saved_s4_set = 0;
     89 static uint32_t head_offset = 0xffffffff;	/* undefined */
     90 static uint32_t tail_offset = 0xffffffff;	/* undefined */
     91 
     92 #ifndef ARRAY_SIZE
     93 #define ARRAY_SIZE(A) (sizeof(A)/sizeof(A[0]))
     94 #endif
     95 
     96 #define BUFFER_FAIL(_count, _len, _name) do {			\
     97     fprintf(out, "Buffer size too small in %s (%d < %d)\n",	\
     98 	    (_name), (_count), (_len));				\
     99     return _count;						\
    100 } while (0)
    101 
    102 static float int_as_float(uint32_t intval)
    103 {
    104 	union intfloat {
    105 		uint32_t i;
    106 		float f;
    107 	} uval;
    108 
    109 	uval.i = intval;
    110 	return uval.f;
    111 }
    112 
    113 static void DRM_PRINTFLIKE(3, 4)
    114 instr_out(struct drm_intel_decode *ctx, unsigned int index,
    115 	  const char *fmt, ...)
    116 {
    117 	va_list va;
    118 	const char *parseinfo;
    119 	uint32_t offset = ctx->hw_offset + index * 4;
    120 
    121 	if (index > ctx->count) {
    122 		if (!ctx->overflowed) {
    123 			fprintf(out, "ERROR: Decode attempted to continue beyond end of batchbuffer\n");
    124 			ctx->overflowed = true;
    125 		}
    126 		return;
    127 	}
    128 
    129 	if (offset == head_offset)
    130 		parseinfo = "HEAD";
    131 	else if (offset == tail_offset)
    132 		parseinfo = "TAIL";
    133 	else
    134 		parseinfo = "    ";
    135 
    136 	fprintf(out, "0x%08x: %s 0x%08x: %s", offset, parseinfo,
    137 		ctx->data[index], index == 0 ? "" : "   ");
    138 	va_start(va, fmt);
    139 	vfprintf(out, fmt, va);
    140 	va_end(va);
    141 }
    142 
    143 static int
    144 decode_MI_SET_CONTEXT(struct drm_intel_decode *ctx)
    145 {
    146 	uint32_t data = ctx->data[1];
    147 	if (ctx->gen > 7)
    148 		return 1;
    149 
    150 	instr_out(ctx, 0, "MI_SET_CONTEXT\n");
    151 	instr_out(ctx, 1, "gtt offset = 0x%x%s%s\n",
    152 		  data & ~0xfff,
    153 		  data & (1<<1)? ", Force Restore": "",
    154 		  data & (1<<0)? ", Restore Inhibit": "");
    155 
    156 	return 2;
    157 }
    158 
    159 static int
    160 decode_MI_WAIT_FOR_EVENT(struct drm_intel_decode *ctx)
    161 {
    162 	const char *cc_wait;
    163 	int cc_shift = 0;
    164 	uint32_t data = ctx->data[0];
    165 
    166 	if (ctx->gen <= 5)
    167 		cc_shift = 9;
    168 	else
    169 		cc_shift = 16;
    170 
    171 	switch ((data >> cc_shift) & 0x1f) {
    172 	case 1:
    173 		cc_wait = ", cc wait 1";
    174 		break;
    175 	case 2:
    176 		cc_wait = ", cc wait 2";
    177 		break;
    178 	case 3:
    179 		cc_wait = ", cc wait 3";
    180 		break;
    181 	case 4:
    182 		cc_wait = ", cc wait 4";
    183 		break;
    184 	case 5:
    185 		cc_wait = ", cc wait 4";
    186 		break;
    187 	default:
    188 		cc_wait = "";
    189 		break;
    190 	}
    191 
    192 	if (ctx->gen <= 5) {
    193 		instr_out(ctx, 0, "MI_WAIT_FOR_EVENT%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
    194 			  data & (1<<18)? ", pipe B start vblank wait": "",
    195 			  data & (1<<17)? ", pipe A start vblank wait": "",
    196 			  data & (1<<16)? ", overlay flip pending wait": "",
    197 			  data & (1<<14)? ", pipe B hblank wait": "",
    198 			  data & (1<<13)? ", pipe A hblank wait": "",
    199 			  cc_wait,
    200 			  data & (1<<8)? ", plane C pending flip wait": "",
    201 			  data & (1<<7)? ", pipe B vblank wait": "",
    202 			  data & (1<<6)? ", plane B pending flip wait": "",
    203 			  data & (1<<5)? ", pipe B scan line wait": "",
    204 			  data & (1<<4)? ", fbc idle wait": "",
    205 			  data & (1<<3)? ", pipe A vblank wait": "",
    206 			  data & (1<<2)? ", plane A pending flip wait": "",
    207 			  data & (1<<1)? ", plane A scan line wait": "");
    208 	} else {
    209 		instr_out(ctx, 0, "MI_WAIT_FOR_EVENT%s%s%s%s%s%s%s%s%s%s%s%s\n",
    210 			  data & (1<<20)? ", sprite C pending flip wait": "", /* ivb */
    211 			  cc_wait,
    212 			  data & (1<<13)? ", pipe B hblank wait": "",
    213 			  data & (1<<11)? ", pipe B vblank wait": "",
    214 			  data & (1<<10)? ", sprite B pending flip wait": "",
    215 			  data & (1<<9)? ", plane B pending flip wait": "",
    216 			  data & (1<<8)? ", plane B scan line wait": "",
    217 			  data & (1<<5)? ", pipe A hblank wait": "",
    218 			  data & (1<<3)? ", pipe A vblank wait": "",
    219 			  data & (1<<2)? ", sprite A pending flip wait": "",
    220 			  data & (1<<1)? ", plane A pending flip wait": "",
    221 			  data & (1<<0)? ", plane A scan line wait": "");
    222 	}
    223 
    224 	return 1;
    225 }
    226 
    227 static int
    228 decode_mi(struct drm_intel_decode *ctx)
    229 {
    230 	unsigned int opcode, len = -1;
    231 	const char *post_sync_op = "";
    232 	uint32_t *data = ctx->data;
    233 
    234 	struct {
    235 		uint32_t opcode;
    236 		int len_mask;
    237 		unsigned int min_len;
    238 		unsigned int max_len;
    239 		const char *name;
    240 		int (*func)(struct drm_intel_decode *ctx);
    241 	} opcodes_mi[] = {
    242 		{ 0x08, 0, 1, 1, "MI_ARB_ON_OFF" },
    243 		{ 0x0a, 0, 1, 1, "MI_BATCH_BUFFER_END" },
    244 		{ 0x30, 0x3f, 3, 3, "MI_BATCH_BUFFER" },
    245 		{ 0x31, 0x3f, 2, 2, "MI_BATCH_BUFFER_START" },
    246 		{ 0x14, 0x3f, 3, 3, "MI_DISPLAY_BUFFER_INFO" },
    247 		{ 0x04, 0, 1, 1, "MI_FLUSH" },
    248 		{ 0x22, 0x1f, 3, 3, "MI_LOAD_REGISTER_IMM" },
    249 		{ 0x13, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_EXCL" },
    250 		{ 0x12, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_INCL" },
    251 		{ 0x00, 0, 1, 1, "MI_NOOP" },
    252 		{ 0x11, 0x3f, 2, 2, "MI_OVERLAY_FLIP" },
    253 		{ 0x07, 0, 1, 1, "MI_REPORT_HEAD" },
    254 		{ 0x18, 0x3f, 2, 2, "MI_SET_CONTEXT", decode_MI_SET_CONTEXT },
    255 		{ 0x20, 0x3f, 3, 4, "MI_STORE_DATA_IMM" },
    256 		{ 0x21, 0x3f, 3, 4, "MI_STORE_DATA_INDEX" },
    257 		{ 0x24, 0x3f, 3, 3, "MI_STORE_REGISTER_MEM" },
    258 		{ 0x02, 0, 1, 1, "MI_USER_INTERRUPT" },
    259 		{ 0x03, 0, 1, 1, "MI_WAIT_FOR_EVENT", decode_MI_WAIT_FOR_EVENT },
    260 		{ 0x16, 0x7f, 3, 3, "MI_SEMAPHORE_MBOX" },
    261 		{ 0x26, 0x1f, 3, 4, "MI_FLUSH_DW" },
    262 		{ 0x28, 0x3f, 3, 3, "MI_REPORT_PERF_COUNT" },
    263 		{ 0x29, 0xff, 3, 3, "MI_LOAD_REGISTER_MEM" },
    264 		{ 0x0b, 0, 1, 1, "MI_SUSPEND_FLUSH"},
    265 	}, *opcode_mi = NULL;
    266 
    267 	/* check instruction length */
    268 	for (opcode = 0; opcode < sizeof(opcodes_mi) / sizeof(opcodes_mi[0]);
    269 	     opcode++) {
    270 		if ((data[0] & 0x1f800000) >> 23 == opcodes_mi[opcode].opcode) {
    271 			len = 1;
    272 			if (opcodes_mi[opcode].max_len > 1) {
    273 				len =
    274 				    (data[0] & opcodes_mi[opcode].len_mask) + 2;
    275 				if (len < opcodes_mi[opcode].min_len
    276 				    || len > opcodes_mi[opcode].max_len) {
    277 					fprintf(out,
    278 						"Bad length (%d) in %s, [%d, %d]\n",
    279 						len, opcodes_mi[opcode].name,
    280 						opcodes_mi[opcode].min_len,
    281 						opcodes_mi[opcode].max_len);
    282 				}
    283 			}
    284 			opcode_mi = &opcodes_mi[opcode];
    285 			break;
    286 		}
    287 	}
    288 
    289 	if (opcode_mi && opcode_mi->func)
    290 		return opcode_mi->func(ctx);
    291 
    292 	switch ((data[0] & 0x1f800000) >> 23) {
    293 	case 0x0a:
    294 		instr_out(ctx, 0, "MI_BATCH_BUFFER_END\n");
    295 		return -1;
    296 	case 0x16:
    297 		instr_out(ctx, 0, "MI_SEMAPHORE_MBOX%s%s%s%s %u\n",
    298 			  data[0] & (1 << 22) ? " global gtt," : "",
    299 			  data[0] & (1 << 21) ? " update semaphore," : "",
    300 			  data[0] & (1 << 20) ? " compare semaphore," : "",
    301 			  data[0] & (1 << 18) ? " use compare reg" : "",
    302 			  (data[0] & (0x3 << 16)) >> 16);
    303 		instr_out(ctx, 1, "value\n");
    304 		instr_out(ctx, 2, "address\n");
    305 		return len;
    306 	case 0x21:
    307 		instr_out(ctx, 0, "MI_STORE_DATA_INDEX%s\n",
    308 			  data[0] & (1 << 21) ? " use per-process HWS," : "");
    309 		instr_out(ctx, 1, "index\n");
    310 		instr_out(ctx, 2, "dword\n");
    311 		if (len == 4)
    312 			instr_out(ctx, 3, "upper dword\n");
    313 		return len;
    314 	case 0x00:
    315 		if (data[0] & (1 << 22))
    316 			instr_out(ctx, 0,
    317 				  "MI_NOOP write NOPID reg, val=0x%x\n",
    318 				  data[0] & ((1 << 22) - 1));
    319 		else
    320 			instr_out(ctx, 0, "MI_NOOP\n");
    321 		return len;
    322 	case 0x26:
    323 		switch (data[0] & (0x3 << 14)) {
    324 		case (0 << 14):
    325 			post_sync_op = "no write";
    326 			break;
    327 		case (1 << 14):
    328 			post_sync_op = "write data";
    329 			break;
    330 		case (2 << 14):
    331 			post_sync_op = "reserved";
    332 			break;
    333 		case (3 << 14):
    334 			post_sync_op = "write TIMESTAMP";
    335 			break;
    336 		}
    337 		instr_out(ctx, 0,
    338 			  "MI_FLUSH_DW%s%s%s%s post_sync_op='%s' %s%s\n",
    339 			  data[0] & (1 << 22) ?
    340 			  " enable protected mem (BCS-only)," : "",
    341 			  data[0] & (1 << 21) ? " store in hws," : "",
    342 			  data[0] & (1 << 18) ? " invalidate tlb," : "",
    343 			  data[0] & (1 << 17) ? " flush gfdt," : "",
    344 			  post_sync_op,
    345 			  data[0] & (1 << 8) ? " enable notify interrupt," : "",
    346 			  data[0] & (1 << 7) ?
    347 			  " invalidate video state (BCS-only)," : "");
    348 		if (data[0] & (1 << 21))
    349 			instr_out(ctx, 1, "hws index\n");
    350 		else
    351 			instr_out(ctx, 1, "address\n");
    352 		instr_out(ctx, 2, "dword\n");
    353 		if (len == 4)
    354 			instr_out(ctx, 3, "upper dword\n");
    355 		return len;
    356 	}
    357 
    358 	for (opcode = 0; opcode < sizeof(opcodes_mi) / sizeof(opcodes_mi[0]);
    359 	     opcode++) {
    360 		if ((data[0] & 0x1f800000) >> 23 == opcodes_mi[opcode].opcode) {
    361 			unsigned int i;
    362 
    363 			instr_out(ctx, 0, "%s\n",
    364 				  opcodes_mi[opcode].name);
    365 			for (i = 1; i < len; i++) {
    366 				instr_out(ctx, i, "dword %d\n", i);
    367 			}
    368 
    369 			return len;
    370 		}
    371 	}
    372 
    373 	instr_out(ctx, 0, "MI UNKNOWN\n");
    374 	return 1;
    375 }
    376 
    377 static void
    378 decode_2d_br00(struct drm_intel_decode *ctx, const char *cmd)
    379 {
    380 	instr_out(ctx, 0,
    381 		  "%s (rgb %sabled, alpha %sabled, src tile %d, dst tile %d)\n",
    382 		  cmd,
    383 		  (ctx->data[0] & (1 << 20)) ? "en" : "dis",
    384 		  (ctx->data[0] & (1 << 21)) ? "en" : "dis",
    385 		  (ctx->data[0] >> 15) & 1,
    386 		  (ctx->data[0] >> 11) & 1);
    387 }
    388 
    389 static void
    390 decode_2d_br01(struct drm_intel_decode *ctx)
    391 {
    392 	const char *format;
    393 	switch ((ctx->data[1] >> 24) & 0x3) {
    394 	case 0:
    395 		format = "8";
    396 		break;
    397 	case 1:
    398 		format = "565";
    399 		break;
    400 	case 2:
    401 		format = "1555";
    402 		break;
    403 	case 3:
    404 		format = "8888";
    405 		break;
    406 	}
    407 
    408 	instr_out(ctx, 1,
    409 		  "format %s, pitch %d, rop 0x%02x, "
    410 		  "clipping %sabled, %s%s \n",
    411 		  format,
    412 		  (short)(ctx->data[1] & 0xffff),
    413 		  (ctx->data[1] >> 16) & 0xff,
    414 		  ctx->data[1] & (1 << 30) ? "en" : "dis",
    415 		  ctx->data[1] & (1 << 31) ? "solid pattern enabled, " : "",
    416 		  ctx->data[1] & (1 << 31) ?
    417 		  "mono pattern transparency enabled, " : "");
    418 
    419 }
    420 
    421 static int
    422 decode_2d(struct drm_intel_decode *ctx)
    423 {
    424 	unsigned int opcode, len;
    425 	uint32_t *data = ctx->data;
    426 
    427 	struct {
    428 		uint32_t opcode;
    429 		unsigned int min_len;
    430 		unsigned int max_len;
    431 		const char *name;
    432 	} opcodes_2d[] = {
    433 		{ 0x40, 5, 5, "COLOR_BLT" },
    434 		{ 0x43, 6, 6, "SRC_COPY_BLT" },
    435 		{ 0x01, 8, 8, "XY_SETUP_BLT" },
    436 		{ 0x11, 9, 9, "XY_SETUP_MONO_PATTERN_SL_BLT" },
    437 		{ 0x03, 3, 3, "XY_SETUP_CLIP_BLT" },
    438 		{ 0x24, 2, 2, "XY_PIXEL_BLT" },
    439 		{ 0x25, 3, 3, "XY_SCANLINES_BLT" },
    440 		{ 0x26, 4, 4, "Y_TEXT_BLT" },
    441 		{ 0x31, 5, 134, "XY_TEXT_IMMEDIATE_BLT" },
    442 		{ 0x50, 6, 6, "XY_COLOR_BLT" },
    443 		{ 0x51, 6, 6, "XY_PAT_BLT" },
    444 		{ 0x76, 8, 8, "XY_PAT_CHROMA_BLT" },
    445 		{ 0x72, 7, 135, "XY_PAT_BLT_IMMEDIATE" },
    446 		{ 0x77, 9, 137, "XY_PAT_CHROMA_BLT_IMMEDIATE" },
    447 		{ 0x52, 9, 9, "XY_MONO_PAT_BLT" },
    448 		{ 0x59, 7, 7, "XY_MONO_PAT_FIXED_BLT" },
    449 		{ 0x53, 8, 8, "XY_SRC_COPY_BLT" },
    450 		{ 0x54, 8, 8, "XY_MONO_SRC_COPY_BLT" },
    451 		{ 0x71, 9, 137, "XY_MONO_SRC_COPY_IMMEDIATE_BLT" },
    452 		{ 0x55, 9, 9, "XY_FULL_BLT" },
    453 		{ 0x55, 9, 137, "XY_FULL_IMMEDIATE_PATTERN_BLT" },
    454 		{ 0x56, 9, 9, "XY_FULL_MONO_SRC_BLT" },
    455 		{ 0x75, 10, 138, "XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT" },
    456 		{ 0x57, 12, 12, "XY_FULL_MONO_PATTERN_BLT" },
    457 		{ 0x58, 12, 12, "XY_FULL_MONO_PATTERN_MONO_SRC_BLT"},
    458 	};
    459 
    460 	switch ((data[0] & 0x1fc00000) >> 22) {
    461 	case 0x25:
    462 		instr_out(ctx, 0,
    463 			  "XY_SCANLINES_BLT (pattern seed (%d, %d), dst tile %d)\n",
    464 			  (data[0] >> 12) & 0x8,
    465 			  (data[0] >> 8) & 0x8, (data[0] >> 11) & 1);
    466 
    467 		len = (data[0] & 0x000000ff) + 2;
    468 		if (len != 3)
    469 			fprintf(out, "Bad count in XY_SCANLINES_BLT\n");
    470 
    471 		instr_out(ctx, 1, "dest (%d,%d)\n",
    472 			  data[1] & 0xffff, data[1] >> 16);
    473 		instr_out(ctx, 2, "dest (%d,%d)\n",
    474 			  data[2] & 0xffff, data[2] >> 16);
    475 		return len;
    476 	case 0x01:
    477 		decode_2d_br00(ctx, "XY_SETUP_BLT");
    478 
    479 		len = (data[0] & 0x000000ff) + 2;
    480 		if (len != 8)
    481 			fprintf(out, "Bad count in XY_SETUP_BLT\n");
    482 
    483 		decode_2d_br01(ctx);
    484 		instr_out(ctx, 2, "cliprect (%d,%d)\n",
    485 			  data[2] & 0xffff, data[2] >> 16);
    486 		instr_out(ctx, 3, "cliprect (%d,%d)\n",
    487 			  data[3] & 0xffff, data[3] >> 16);
    488 		instr_out(ctx, 4, "setup dst offset 0x%08x\n",
    489 			  data[4]);
    490 		instr_out(ctx, 5, "setup background color\n");
    491 		instr_out(ctx, 6, "setup foreground color\n");
    492 		instr_out(ctx, 7, "color pattern offset\n");
    493 		return len;
    494 	case 0x03:
    495 		decode_2d_br00(ctx, "XY_SETUP_CLIP_BLT");
    496 
    497 		len = (data[0] & 0x000000ff) + 2;
    498 		if (len != 3)
    499 			fprintf(out, "Bad count in XY_SETUP_CLIP_BLT\n");
    500 
    501 		instr_out(ctx, 1, "cliprect (%d,%d)\n",
    502 			  data[1] & 0xffff, data[2] >> 16);
    503 		instr_out(ctx, 2, "cliprect (%d,%d)\n",
    504 			  data[2] & 0xffff, data[3] >> 16);
    505 		return len;
    506 	case 0x11:
    507 		decode_2d_br00(ctx, "XY_SETUP_MONO_PATTERN_SL_BLT");
    508 
    509 		len = (data[0] & 0x000000ff) + 2;
    510 		if (len != 9)
    511 			fprintf(out,
    512 				"Bad count in XY_SETUP_MONO_PATTERN_SL_BLT\n");
    513 
    514 		decode_2d_br01(ctx);
    515 		instr_out(ctx, 2, "cliprect (%d,%d)\n",
    516 			  data[2] & 0xffff, data[2] >> 16);
    517 		instr_out(ctx, 3, "cliprect (%d,%d)\n",
    518 			  data[3] & 0xffff, data[3] >> 16);
    519 		instr_out(ctx, 4, "setup dst offset 0x%08x\n",
    520 			  data[4]);
    521 		instr_out(ctx, 5, "setup background color\n");
    522 		instr_out(ctx, 6, "setup foreground color\n");
    523 		instr_out(ctx, 7, "mono pattern dw0\n");
    524 		instr_out(ctx, 8, "mono pattern dw1\n");
    525 		return len;
    526 	case 0x50:
    527 		decode_2d_br00(ctx, "XY_COLOR_BLT");
    528 
    529 		len = (data[0] & 0x000000ff) + 2;
    530 		if (len != 6)
    531 			fprintf(out, "Bad count in XY_COLOR_BLT\n");
    532 
    533 		decode_2d_br01(ctx);
    534 		instr_out(ctx, 2, "(%d,%d)\n",
    535 			  data[2] & 0xffff, data[2] >> 16);
    536 		instr_out(ctx, 3, "(%d,%d)\n",
    537 			  data[3] & 0xffff, data[3] >> 16);
    538 		instr_out(ctx, 4, "offset 0x%08x\n", data[4]);
    539 		instr_out(ctx, 5, "color\n");
    540 		return len;
    541 	case 0x53:
    542 		decode_2d_br00(ctx, "XY_SRC_COPY_BLT");
    543 
    544 		len = (data[0] & 0x000000ff) + 2;
    545 		if (len != 8)
    546 			fprintf(out, "Bad count in XY_SRC_COPY_BLT\n");
    547 
    548 		decode_2d_br01(ctx);
    549 		instr_out(ctx, 2, "dst (%d,%d)\n",
    550 			  data[2] & 0xffff, data[2] >> 16);
    551 		instr_out(ctx, 3, "dst (%d,%d)\n",
    552 			  data[3] & 0xffff, data[3] >> 16);
    553 		instr_out(ctx, 4, "dst offset 0x%08x\n", data[4]);
    554 		instr_out(ctx, 5, "src (%d,%d)\n",
    555 			  data[5] & 0xffff, data[5] >> 16);
    556 		instr_out(ctx, 6, "src pitch %d\n",
    557 			  (short)(data[6] & 0xffff));
    558 		instr_out(ctx, 7, "src offset 0x%08x\n", data[7]);
    559 		return len;
    560 	}
    561 
    562 	for (opcode = 0; opcode < sizeof(opcodes_2d) / sizeof(opcodes_2d[0]);
    563 	     opcode++) {
    564 		if ((data[0] & 0x1fc00000) >> 22 == opcodes_2d[opcode].opcode) {
    565 			unsigned int i;
    566 
    567 			len = 1;
    568 			instr_out(ctx, 0, "%s\n",
    569 				  opcodes_2d[opcode].name);
    570 			if (opcodes_2d[opcode].max_len > 1) {
    571 				len = (data[0] & 0x000000ff) + 2;
    572 				if (len < opcodes_2d[opcode].min_len ||
    573 				    len > opcodes_2d[opcode].max_len) {
    574 					fprintf(out, "Bad count in %s\n",
    575 						opcodes_2d[opcode].name);
    576 				}
    577 			}
    578 
    579 			for (i = 1; i < len; i++) {
    580 				instr_out(ctx, i, "dword %d\n", i);
    581 			}
    582 
    583 			return len;
    584 		}
    585 	}
    586 
    587 	instr_out(ctx, 0, "2D UNKNOWN\n");
    588 	return 1;
    589 }
    590 
    591 static int
    592 decode_3d_1c(struct drm_intel_decode *ctx)
    593 {
    594 	uint32_t *data = ctx->data;
    595 	uint32_t opcode;
    596 
    597 	opcode = (data[0] & 0x00f80000) >> 19;
    598 
    599 	switch (opcode) {
    600 	case 0x11:
    601 		instr_out(ctx, 0,
    602 			  "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE\n");
    603 		return 1;
    604 	case 0x10:
    605 		instr_out(ctx, 0, "3DSTATE_SCISSOR_ENABLE %s\n",
    606 			  data[0] & 1 ? "enabled" : "disabled");
    607 		return 1;
    608 	case 0x01:
    609 		instr_out(ctx, 0, "3DSTATE_MAP_COORD_SET_I830\n");
    610 		return 1;
    611 	case 0x0a:
    612 		instr_out(ctx, 0, "3DSTATE_MAP_CUBE_I830\n");
    613 		return 1;
    614 	case 0x05:
    615 		instr_out(ctx, 0, "3DSTATE_MAP_TEX_STREAM_I830\n");
    616 		return 1;
    617 	}
    618 
    619 	instr_out(ctx, 0, "3D UNKNOWN: 3d_1c opcode = 0x%x\n",
    620 		  opcode);
    621 	return 1;
    622 }
    623 
    624 /** Sets the string dstname to describe the destination of the PS instruction */
    625 static void
    626 i915_get_instruction_dst(uint32_t *data, int i, char *dstname, int do_mask)
    627 {
    628 	uint32_t a0 = data[i];
    629 	int dst_nr = (a0 >> 14) & 0xf;
    630 	char dstmask[8];
    631 	const char *sat;
    632 
    633 	if (do_mask) {
    634 		if (((a0 >> 10) & 0xf) == 0xf) {
    635 			dstmask[0] = 0;
    636 		} else {
    637 			int dstmask_index = 0;
    638 
    639 			dstmask[dstmask_index++] = '.';
    640 			if (a0 & (1 << 10))
    641 				dstmask[dstmask_index++] = 'x';
    642 			if (a0 & (1 << 11))
    643 				dstmask[dstmask_index++] = 'y';
    644 			if (a0 & (1 << 12))
    645 				dstmask[dstmask_index++] = 'z';
    646 			if (a0 & (1 << 13))
    647 				dstmask[dstmask_index++] = 'w';
    648 			dstmask[dstmask_index++] = 0;
    649 		}
    650 
    651 		if (a0 & (1 << 22))
    652 			sat = ".sat";
    653 		else
    654 			sat = "";
    655 	} else {
    656 		dstmask[0] = 0;
    657 		sat = "";
    658 	}
    659 
    660 	switch ((a0 >> 19) & 0x7) {
    661 	case 0:
    662 		if (dst_nr > 15)
    663 			fprintf(out, "bad destination reg R%d\n", dst_nr);
    664 		sprintf(dstname, "R%d%s%s", dst_nr, dstmask, sat);
    665 		break;
    666 	case 4:
    667 		if (dst_nr > 0)
    668 			fprintf(out, "bad destination reg oC%d\n", dst_nr);
    669 		sprintf(dstname, "oC%s%s", dstmask, sat);
    670 		break;
    671 	case 5:
    672 		if (dst_nr > 0)
    673 			fprintf(out, "bad destination reg oD%d\n", dst_nr);
    674 		sprintf(dstname, "oD%s%s", dstmask, sat);
    675 		break;
    676 	case 6:
    677 		if (dst_nr > 3)
    678 			fprintf(out, "bad destination reg U%d\n", dst_nr);
    679 		sprintf(dstname, "U%d%s%s", dst_nr, dstmask, sat);
    680 		break;
    681 	default:
    682 		sprintf(dstname, "RESERVED");
    683 		break;
    684 	}
    685 }
    686 
    687 static const char *
    688 i915_get_channel_swizzle(uint32_t select)
    689 {
    690 	switch (select & 0x7) {
    691 	case 0:
    692 		return (select & 8) ? "-x" : "x";
    693 	case 1:
    694 		return (select & 8) ? "-y" : "y";
    695 	case 2:
    696 		return (select & 8) ? "-z" : "z";
    697 	case 3:
    698 		return (select & 8) ? "-w" : "w";
    699 	case 4:
    700 		return (select & 8) ? "-0" : "0";
    701 	case 5:
    702 		return (select & 8) ? "-1" : "1";
    703 	default:
    704 		return (select & 8) ? "-bad" : "bad";
    705 	}
    706 }
    707 
    708 static void
    709 i915_get_instruction_src_name(uint32_t src_type, uint32_t src_nr, char *name)
    710 {
    711 	switch (src_type) {
    712 	case 0:
    713 		sprintf(name, "R%d", src_nr);
    714 		if (src_nr > 15)
    715 			fprintf(out, "bad src reg %s\n", name);
    716 		break;
    717 	case 1:
    718 		if (src_nr < 8)
    719 			sprintf(name, "T%d", src_nr);
    720 		else if (src_nr == 8)
    721 			sprintf(name, "DIFFUSE");
    722 		else if (src_nr == 9)
    723 			sprintf(name, "SPECULAR");
    724 		else if (src_nr == 10)
    725 			sprintf(name, "FOG");
    726 		else {
    727 			fprintf(out, "bad src reg T%d\n", src_nr);
    728 			sprintf(name, "RESERVED");
    729 		}
    730 		break;
    731 	case 2:
    732 		sprintf(name, "C%d", src_nr);
    733 		if (src_nr > 31)
    734 			fprintf(out, "bad src reg %s\n", name);
    735 		break;
    736 	case 4:
    737 		sprintf(name, "oC");
    738 		if (src_nr > 0)
    739 			fprintf(out, "bad src reg oC%d\n", src_nr);
    740 		break;
    741 	case 5:
    742 		sprintf(name, "oD");
    743 		if (src_nr > 0)
    744 			fprintf(out, "bad src reg oD%d\n", src_nr);
    745 		break;
    746 	case 6:
    747 		sprintf(name, "U%d", src_nr);
    748 		if (src_nr > 3)
    749 			fprintf(out, "bad src reg %s\n", name);
    750 		break;
    751 	default:
    752 		fprintf(out, "bad src reg type %d\n", src_type);
    753 		sprintf(name, "RESERVED");
    754 		break;
    755 	}
    756 }
    757 
    758 static void i915_get_instruction_src0(uint32_t *data, int i, char *srcname)
    759 {
    760 	uint32_t a0 = data[i];
    761 	uint32_t a1 = data[i + 1];
    762 	int src_nr = (a0 >> 2) & 0x1f;
    763 	const char *swizzle_x = i915_get_channel_swizzle((a1 >> 28) & 0xf);
    764 	const char *swizzle_y = i915_get_channel_swizzle((a1 >> 24) & 0xf);
    765 	const char *swizzle_z = i915_get_channel_swizzle((a1 >> 20) & 0xf);
    766 	const char *swizzle_w = i915_get_channel_swizzle((a1 >> 16) & 0xf);
    767 	char swizzle[100];
    768 
    769 	i915_get_instruction_src_name((a0 >> 7) & 0x7, src_nr, srcname);
    770 	sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z,
    771 		swizzle_w);
    772 	if (strcmp(swizzle, ".xyzw") != 0)
    773 		strcat(srcname, swizzle);
    774 }
    775 
    776 static void i915_get_instruction_src1(uint32_t *data, int i, char *srcname)
    777 {
    778 	uint32_t a1 = data[i + 1];
    779 	uint32_t a2 = data[i + 2];
    780 	int src_nr = (a1 >> 8) & 0x1f;
    781 	const char *swizzle_x = i915_get_channel_swizzle((a1 >> 4) & 0xf);
    782 	const char *swizzle_y = i915_get_channel_swizzle((a1 >> 0) & 0xf);
    783 	const char *swizzle_z = i915_get_channel_swizzle((a2 >> 28) & 0xf);
    784 	const char *swizzle_w = i915_get_channel_swizzle((a2 >> 24) & 0xf);
    785 	char swizzle[100];
    786 
    787 	i915_get_instruction_src_name((a1 >> 13) & 0x7, src_nr, srcname);
    788 	sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z,
    789 		swizzle_w);
    790 	if (strcmp(swizzle, ".xyzw") != 0)
    791 		strcat(srcname, swizzle);
    792 }
    793 
    794 static void i915_get_instruction_src2(uint32_t *data, int i, char *srcname)
    795 {
    796 	uint32_t a2 = data[i + 2];
    797 	int src_nr = (a2 >> 16) & 0x1f;
    798 	const char *swizzle_x = i915_get_channel_swizzle((a2 >> 12) & 0xf);
    799 	const char *swizzle_y = i915_get_channel_swizzle((a2 >> 8) & 0xf);
    800 	const char *swizzle_z = i915_get_channel_swizzle((a2 >> 4) & 0xf);
    801 	const char *swizzle_w = i915_get_channel_swizzle((a2 >> 0) & 0xf);
    802 	char swizzle[100];
    803 
    804 	i915_get_instruction_src_name((a2 >> 21) & 0x7, src_nr, srcname);
    805 	sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z,
    806 		swizzle_w);
    807 	if (strcmp(swizzle, ".xyzw") != 0)
    808 		strcat(srcname, swizzle);
    809 }
    810 
    811 static void
    812 i915_get_instruction_addr(uint32_t src_type, uint32_t src_nr, char *name)
    813 {
    814 	switch (src_type) {
    815 	case 0:
    816 		sprintf(name, "R%d", src_nr);
    817 		if (src_nr > 15)
    818 			fprintf(out, "bad src reg %s\n", name);
    819 		break;
    820 	case 1:
    821 		if (src_nr < 8)
    822 			sprintf(name, "T%d", src_nr);
    823 		else if (src_nr == 8)
    824 			sprintf(name, "DIFFUSE");
    825 		else if (src_nr == 9)
    826 			sprintf(name, "SPECULAR");
    827 		else if (src_nr == 10)
    828 			sprintf(name, "FOG");
    829 		else {
    830 			fprintf(out, "bad src reg T%d\n", src_nr);
    831 			sprintf(name, "RESERVED");
    832 		}
    833 		break;
    834 	case 4:
    835 		sprintf(name, "oC");
    836 		if (src_nr > 0)
    837 			fprintf(out, "bad src reg oC%d\n", src_nr);
    838 		break;
    839 	case 5:
    840 		sprintf(name, "oD");
    841 		if (src_nr > 0)
    842 			fprintf(out, "bad src reg oD%d\n", src_nr);
    843 		break;
    844 	default:
    845 		fprintf(out, "bad src reg type %d\n", src_type);
    846 		sprintf(name, "RESERVED");
    847 		break;
    848 	}
    849 }
    850 
    851 static void
    852 i915_decode_alu1(struct drm_intel_decode *ctx,
    853 		 int i, char *instr_prefix, const char *op_name)
    854 {
    855 	char dst[100], src0[100];
    856 
    857 	i915_get_instruction_dst(ctx->data, i, dst, 1);
    858 	i915_get_instruction_src0(ctx->data, i, src0);
    859 
    860 	instr_out(ctx, i++, "%s: %s %s, %s\n", instr_prefix,
    861 		  op_name, dst, src0);
    862 	instr_out(ctx, i++, "%s\n", instr_prefix);
    863 	instr_out(ctx, i++, "%s\n", instr_prefix);
    864 }
    865 
    866 static void
    867 i915_decode_alu2(struct drm_intel_decode *ctx,
    868 		 int i, char *instr_prefix, const char *op_name)
    869 {
    870 	char dst[100], src0[100], src1[100];
    871 
    872 	i915_get_instruction_dst(ctx->data, i, dst, 1);
    873 	i915_get_instruction_src0(ctx->data, i, src0);
    874 	i915_get_instruction_src1(ctx->data, i, src1);
    875 
    876 	instr_out(ctx, i++, "%s: %s %s, %s, %s\n", instr_prefix,
    877 		  op_name, dst, src0, src1);
    878 	instr_out(ctx, i++, "%s\n", instr_prefix);
    879 	instr_out(ctx, i++, "%s\n", instr_prefix);
    880 }
    881 
    882 static void
    883 i915_decode_alu3(struct drm_intel_decode *ctx,
    884 		 int i, char *instr_prefix, const char *op_name)
    885 {
    886 	char dst[100], src0[100], src1[100], src2[100];
    887 
    888 	i915_get_instruction_dst(ctx->data, i, dst, 1);
    889 	i915_get_instruction_src0(ctx->data, i, src0);
    890 	i915_get_instruction_src1(ctx->data, i, src1);
    891 	i915_get_instruction_src2(ctx->data, i, src2);
    892 
    893 	instr_out(ctx, i++, "%s: %s %s, %s, %s, %s\n", instr_prefix,
    894 		  op_name, dst, src0, src1, src2);
    895 	instr_out(ctx, i++, "%s\n", instr_prefix);
    896 	instr_out(ctx, i++, "%s\n", instr_prefix);
    897 }
    898 
    899 static void
    900 i915_decode_tex(struct drm_intel_decode *ctx, int i,
    901 		const char *instr_prefix, const char *tex_name)
    902 {
    903 	uint32_t t0 = ctx->data[i];
    904 	uint32_t t1 = ctx->data[i + 1];
    905 	char dst_name[100];
    906 	char addr_name[100];
    907 	int sampler_nr;
    908 
    909 	i915_get_instruction_dst(ctx->data, i, dst_name, 0);
    910 	i915_get_instruction_addr((t1 >> 24) & 0x7,
    911 				  (t1 >> 17) & 0xf, addr_name);
    912 	sampler_nr = t0 & 0xf;
    913 
    914 	instr_out(ctx, i++, "%s: %s %s, S%d, %s\n", instr_prefix,
    915 		  tex_name, dst_name, sampler_nr, addr_name);
    916 	instr_out(ctx, i++, "%s\n", instr_prefix);
    917 	instr_out(ctx, i++, "%s\n", instr_prefix);
    918 }
    919 
    920 static void
    921 i915_decode_dcl(struct drm_intel_decode *ctx, int i, char *instr_prefix)
    922 {
    923 	uint32_t d0 = ctx->data[i];
    924 	const char *sampletype;
    925 	int dcl_nr = (d0 >> 14) & 0xf;
    926 	const char *dcl_x = d0 & (1 << 10) ? "x" : "";
    927 	const char *dcl_y = d0 & (1 << 11) ? "y" : "";
    928 	const char *dcl_z = d0 & (1 << 12) ? "z" : "";
    929 	const char *dcl_w = d0 & (1 << 13) ? "w" : "";
    930 	char dcl_mask[10];
    931 
    932 	switch ((d0 >> 19) & 0x3) {
    933 	case 1:
    934 		sprintf(dcl_mask, ".%s%s%s%s", dcl_x, dcl_y, dcl_z, dcl_w);
    935 		if (strcmp(dcl_mask, ".") == 0)
    936 			fprintf(out, "bad (empty) dcl mask\n");
    937 
    938 		if (dcl_nr > 10)
    939 			fprintf(out, "bad T%d dcl register number\n", dcl_nr);
    940 		if (dcl_nr < 8) {
    941 			if (strcmp(dcl_mask, ".x") != 0 &&
    942 			    strcmp(dcl_mask, ".xy") != 0 &&
    943 			    strcmp(dcl_mask, ".xz") != 0 &&
    944 			    strcmp(dcl_mask, ".w") != 0 &&
    945 			    strcmp(dcl_mask, ".xyzw") != 0) {
    946 				fprintf(out, "bad T%d.%s dcl mask\n", dcl_nr,
    947 					dcl_mask);
    948 			}
    949 			instr_out(ctx, i++, "%s: DCL T%d%s\n",
    950 				  instr_prefix, dcl_nr, dcl_mask);
    951 		} else {
    952 			if (strcmp(dcl_mask, ".xz") == 0)
    953 				fprintf(out, "errataed bad dcl mask %s\n",
    954 					dcl_mask);
    955 			else if (strcmp(dcl_mask, ".xw") == 0)
    956 				fprintf(out, "errataed bad dcl mask %s\n",
    957 					dcl_mask);
    958 			else if (strcmp(dcl_mask, ".xzw") == 0)
    959 				fprintf(out, "errataed bad dcl mask %s\n",
    960 					dcl_mask);
    961 
    962 			if (dcl_nr == 8) {
    963 				instr_out(ctx, i++,
    964 					  "%s: DCL DIFFUSE%s\n", instr_prefix,
    965 					  dcl_mask);
    966 			} else if (dcl_nr == 9) {
    967 				instr_out(ctx, i++,
    968 					  "%s: DCL SPECULAR%s\n", instr_prefix,
    969 					  dcl_mask);
    970 			} else if (dcl_nr == 10) {
    971 				instr_out(ctx, i++,
    972 					  "%s: DCL FOG%s\n", instr_prefix,
    973 					  dcl_mask);
    974 			}
    975 		}
    976 		instr_out(ctx, i++, "%s\n", instr_prefix);
    977 		instr_out(ctx, i++, "%s\n", instr_prefix);
    978 		break;
    979 	case 3:
    980 		switch ((d0 >> 22) & 0x3) {
    981 		case 0:
    982 			sampletype = "2D";
    983 			break;
    984 		case 1:
    985 			sampletype = "CUBE";
    986 			break;
    987 		case 2:
    988 			sampletype = "3D";
    989 			break;
    990 		default:
    991 			sampletype = "RESERVED";
    992 			break;
    993 		}
    994 		if (dcl_nr > 15)
    995 			fprintf(out, "bad S%d dcl register number\n", dcl_nr);
    996 		instr_out(ctx, i++, "%s: DCL S%d %s\n",
    997 			  instr_prefix, dcl_nr, sampletype);
    998 		instr_out(ctx, i++, "%s\n", instr_prefix);
    999 		instr_out(ctx, i++, "%s\n", instr_prefix);
   1000 		break;
   1001 	default:
   1002 		instr_out(ctx, i++, "%s: DCL RESERVED%d\n",
   1003 			  instr_prefix, dcl_nr);
   1004 		instr_out(ctx, i++, "%s\n", instr_prefix);
   1005 		instr_out(ctx, i++, "%s\n", instr_prefix);
   1006 	}
   1007 }
   1008 
   1009 static void
   1010 i915_decode_instruction(struct drm_intel_decode *ctx,
   1011 			int i, char *instr_prefix)
   1012 {
   1013 	switch ((ctx->data[i] >> 24) & 0x1f) {
   1014 	case 0x0:
   1015 		instr_out(ctx, i++, "%s: NOP\n", instr_prefix);
   1016 		instr_out(ctx, i++, "%s\n", instr_prefix);
   1017 		instr_out(ctx, i++, "%s\n", instr_prefix);
   1018 		break;
   1019 	case 0x01:
   1020 		i915_decode_alu2(ctx, i, instr_prefix, "ADD");
   1021 		break;
   1022 	case 0x02:
   1023 		i915_decode_alu1(ctx, i, instr_prefix, "MOV");
   1024 		break;
   1025 	case 0x03:
   1026 		i915_decode_alu2(ctx, i, instr_prefix, "MUL");
   1027 		break;
   1028 	case 0x04:
   1029 		i915_decode_alu3(ctx, i, instr_prefix, "MAD");
   1030 		break;
   1031 	case 0x05:
   1032 		i915_decode_alu3(ctx, i, instr_prefix, "DP2ADD");
   1033 		break;
   1034 	case 0x06:
   1035 		i915_decode_alu2(ctx, i, instr_prefix, "DP3");
   1036 		break;
   1037 	case 0x07:
   1038 		i915_decode_alu2(ctx, i, instr_prefix, "DP4");
   1039 		break;
   1040 	case 0x08:
   1041 		i915_decode_alu1(ctx, i, instr_prefix, "FRC");
   1042 		break;
   1043 	case 0x09:
   1044 		i915_decode_alu1(ctx, i, instr_prefix, "RCP");
   1045 		break;
   1046 	case 0x0a:
   1047 		i915_decode_alu1(ctx, i, instr_prefix, "RSQ");
   1048 		break;
   1049 	case 0x0b:
   1050 		i915_decode_alu1(ctx, i, instr_prefix, "EXP");
   1051 		break;
   1052 	case 0x0c:
   1053 		i915_decode_alu1(ctx, i, instr_prefix, "LOG");
   1054 		break;
   1055 	case 0x0d:
   1056 		i915_decode_alu2(ctx, i, instr_prefix, "CMP");
   1057 		break;
   1058 	case 0x0e:
   1059 		i915_decode_alu2(ctx, i, instr_prefix, "MIN");
   1060 		break;
   1061 	case 0x0f:
   1062 		i915_decode_alu2(ctx, i, instr_prefix, "MAX");
   1063 		break;
   1064 	case 0x10:
   1065 		i915_decode_alu1(ctx, i, instr_prefix, "FLR");
   1066 		break;
   1067 	case 0x11:
   1068 		i915_decode_alu1(ctx, i, instr_prefix, "MOD");
   1069 		break;
   1070 	case 0x12:
   1071 		i915_decode_alu1(ctx, i, instr_prefix, "TRC");
   1072 		break;
   1073 	case 0x13:
   1074 		i915_decode_alu2(ctx, i, instr_prefix, "SGE");
   1075 		break;
   1076 	case 0x14:
   1077 		i915_decode_alu2(ctx, i, instr_prefix, "SLT");
   1078 		break;
   1079 	case 0x15:
   1080 		i915_decode_tex(ctx, i, instr_prefix, "TEXLD");
   1081 		break;
   1082 	case 0x16:
   1083 		i915_decode_tex(ctx, i, instr_prefix, "TEXLDP");
   1084 		break;
   1085 	case 0x17:
   1086 		i915_decode_tex(ctx, i, instr_prefix, "TEXLDB");
   1087 		break;
   1088 	case 0x19:
   1089 		i915_decode_dcl(ctx, i, instr_prefix);
   1090 		break;
   1091 	default:
   1092 		instr_out(ctx, i++, "%s: unknown\n", instr_prefix);
   1093 		instr_out(ctx, i++, "%s\n", instr_prefix);
   1094 		instr_out(ctx, i++, "%s\n", instr_prefix);
   1095 		break;
   1096 	}
   1097 }
   1098 
   1099 static const char *
   1100 decode_compare_func(uint32_t op)
   1101 {
   1102 	switch (op & 0x7) {
   1103 	case 0:
   1104 		return "always";
   1105 	case 1:
   1106 		return "never";
   1107 	case 2:
   1108 		return "less";
   1109 	case 3:
   1110 		return "equal";
   1111 	case 4:
   1112 		return "lequal";
   1113 	case 5:
   1114 		return "greater";
   1115 	case 6:
   1116 		return "notequal";
   1117 	case 7:
   1118 		return "gequal";
   1119 	}
   1120 	return "";
   1121 }
   1122 
   1123 static const char *
   1124 decode_stencil_op(uint32_t op)
   1125 {
   1126 	switch (op & 0x7) {
   1127 	case 0:
   1128 		return "keep";
   1129 	case 1:
   1130 		return "zero";
   1131 	case 2:
   1132 		return "replace";
   1133 	case 3:
   1134 		return "incr_sat";
   1135 	case 4:
   1136 		return "decr_sat";
   1137 	case 5:
   1138 		return "greater";
   1139 	case 6:
   1140 		return "incr";
   1141 	case 7:
   1142 		return "decr";
   1143 	}
   1144 	return "";
   1145 }
   1146 
   1147 #if 0
   1148 static const char *
   1149 decode_logic_op(uint32_t op)
   1150 {
   1151 	switch (op & 0xf) {
   1152 	case 0:
   1153 		return "clear";
   1154 	case 1:
   1155 		return "nor";
   1156 	case 2:
   1157 		return "and_inv";
   1158 	case 3:
   1159 		return "copy_inv";
   1160 	case 4:
   1161 		return "and_rvrse";
   1162 	case 5:
   1163 		return "inv";
   1164 	case 6:
   1165 		return "xor";
   1166 	case 7:
   1167 		return "nand";
   1168 	case 8:
   1169 		return "and";
   1170 	case 9:
   1171 		return "equiv";
   1172 	case 10:
   1173 		return "noop";
   1174 	case 11:
   1175 		return "or_inv";
   1176 	case 12:
   1177 		return "copy";
   1178 	case 13:
   1179 		return "or_rvrse";
   1180 	case 14:
   1181 		return "or";
   1182 	case 15:
   1183 		return "set";
   1184 	}
   1185 	return "";
   1186 }
   1187 #endif
   1188 
   1189 static const char *
   1190 decode_blend_fact(uint32_t op)
   1191 {
   1192 	switch (op & 0xf) {
   1193 	case 1:
   1194 		return "zero";
   1195 	case 2:
   1196 		return "one";
   1197 	case 3:
   1198 		return "src_colr";
   1199 	case 4:
   1200 		return "inv_src_colr";
   1201 	case 5:
   1202 		return "src_alpha";
   1203 	case 6:
   1204 		return "inv_src_alpha";
   1205 	case 7:
   1206 		return "dst_alpha";
   1207 	case 8:
   1208 		return "inv_dst_alpha";
   1209 	case 9:
   1210 		return "dst_colr";
   1211 	case 10:
   1212 		return "inv_dst_colr";
   1213 	case 11:
   1214 		return "src_alpha_sat";
   1215 	case 12:
   1216 		return "cnst_colr";
   1217 	case 13:
   1218 		return "inv_cnst_colr";
   1219 	case 14:
   1220 		return "cnst_alpha";
   1221 	case 15:
   1222 		return "inv_const_alpha";
   1223 	}
   1224 	return "";
   1225 }
   1226 
   1227 static const char *
   1228 decode_tex_coord_mode(uint32_t mode)
   1229 {
   1230 	switch (mode & 0x7) {
   1231 	case 0:
   1232 		return "wrap";
   1233 	case 1:
   1234 		return "mirror";
   1235 	case 2:
   1236 		return "clamp_edge";
   1237 	case 3:
   1238 		return "cube";
   1239 	case 4:
   1240 		return "clamp_border";
   1241 	case 5:
   1242 		return "mirror_once";
   1243 	}
   1244 	return "";
   1245 }
   1246 
   1247 static const char *
   1248 decode_sample_filter(uint32_t mode)
   1249 {
   1250 	switch (mode & 0x7) {
   1251 	case 0:
   1252 		return "nearest";
   1253 	case 1:
   1254 		return "linear";
   1255 	case 2:
   1256 		return "anisotropic";
   1257 	case 3:
   1258 		return "4x4_1";
   1259 	case 4:
   1260 		return "4x4_2";
   1261 	case 5:
   1262 		return "4x4_flat";
   1263 	case 6:
   1264 		return "6x5_mono";
   1265 	}
   1266 	return "";
   1267 }
   1268 
   1269 static int
   1270 decode_3d_1d(struct drm_intel_decode *ctx)
   1271 {
   1272 	unsigned int len, i, c, idx, word, map, sampler, instr;
   1273 	const char *format, *zformat, *type;
   1274 	uint32_t opcode;
   1275 	uint32_t *data = ctx->data;
   1276 	uint32_t devid = ctx->devid;
   1277 
   1278 	struct {
   1279 		uint32_t opcode;
   1280 		int i830_only;
   1281 		unsigned int min_len;
   1282 		unsigned int max_len;
   1283 		const char *name;
   1284 	} opcodes_3d_1d[] = {
   1285 		{ 0x86, 0, 4, 4, "3DSTATE_CHROMA_KEY" },
   1286 		{ 0x88, 0, 2, 2, "3DSTATE_CONSTANT_BLEND_COLOR" },
   1287 		{ 0x99, 0, 2, 2, "3DSTATE_DEFAULT_DIFFUSE" },
   1288 		{ 0x9a, 0, 2, 2, "3DSTATE_DEFAULT_SPECULAR" },
   1289 		{ 0x98, 0, 2, 2, "3DSTATE_DEFAULT_Z" },
   1290 		{ 0x97, 0, 2, 2, "3DSTATE_DEPTH_OFFSET_SCALE" },
   1291 		{ 0x9d, 0, 65, 65, "3DSTATE_FILTER_COEFFICIENTS_4X4" },
   1292 		{ 0x9e, 0, 4, 4, "3DSTATE_MONO_FILTER" },
   1293 		{ 0x89, 0, 4, 4, "3DSTATE_FOG_MODE" },
   1294 		{ 0x8f, 0, 2, 16, "3DSTATE_MAP_PALLETE_LOAD_32" },
   1295 		{ 0x83, 0, 2, 2, "3DSTATE_SPAN_STIPPLE" },
   1296 		{ 0x8c, 1, 2, 2, "3DSTATE_MAP_COORD_TRANSFORM_I830" },
   1297 		{ 0x8b, 1, 2, 2, "3DSTATE_MAP_VERTEX_TRANSFORM_I830" },
   1298 		{ 0x8d, 1, 3, 3, "3DSTATE_W_STATE_I830" },
   1299 		{ 0x01, 1, 2, 2, "3DSTATE_COLOR_FACTOR_I830" },
   1300 		{ 0x02, 1, 2, 2, "3DSTATE_MAP_COORD_SETBIND_I830"},
   1301 	}, *opcode_3d_1d;
   1302 
   1303 	opcode = (data[0] & 0x00ff0000) >> 16;
   1304 
   1305 	switch (opcode) {
   1306 	case 0x07:
   1307 		/* This instruction is unusual.  A 0 length means just
   1308 		 * 1 DWORD instead of 2.  The 0 length is specified in
   1309 		 * one place to be unsupported, but stated to be
   1310 		 * required in another, and 0 length LOAD_INDIRECTs
   1311 		 * appear to cause no harm at least.
   1312 		 */
   1313 		instr_out(ctx, 0, "3DSTATE_LOAD_INDIRECT\n");
   1314 		len = (data[0] & 0x000000ff) + 1;
   1315 		i = 1;
   1316 		if (data[0] & (0x01 << 8)) {
   1317 			instr_out(ctx, i++, "SIS.0\n");
   1318 			instr_out(ctx, i++, "SIS.1\n");
   1319 		}
   1320 		if (data[0] & (0x02 << 8)) {
   1321 			instr_out(ctx, i++, "DIS.0\n");
   1322 		}
   1323 		if (data[0] & (0x04 << 8)) {
   1324 			instr_out(ctx, i++, "SSB.0\n");
   1325 			instr_out(ctx, i++, "SSB.1\n");
   1326 		}
   1327 		if (data[0] & (0x08 << 8)) {
   1328 			instr_out(ctx, i++, "MSB.0\n");
   1329 			instr_out(ctx, i++, "MSB.1\n");
   1330 		}
   1331 		if (data[0] & (0x10 << 8)) {
   1332 			instr_out(ctx, i++, "PSP.0\n");
   1333 			instr_out(ctx, i++, "PSP.1\n");
   1334 		}
   1335 		if (data[0] & (0x20 << 8)) {
   1336 			instr_out(ctx, i++, "PSC.0\n");
   1337 			instr_out(ctx, i++, "PSC.1\n");
   1338 		}
   1339 		if (len != i) {
   1340 			fprintf(out, "Bad count in 3DSTATE_LOAD_INDIRECT\n");
   1341 			return len;
   1342 		}
   1343 		return len;
   1344 	case 0x04:
   1345 		instr_out(ctx, 0,
   1346 			  "3DSTATE_LOAD_STATE_IMMEDIATE_1\n");
   1347 		len = (data[0] & 0x0000000f) + 2;
   1348 		i = 1;
   1349 		for (word = 0; word <= 8; word++) {
   1350 			if (data[0] & (1 << (4 + word))) {
   1351 				/* save vertex state for decode */
   1352 				if (!IS_GEN2(devid)) {
   1353 					int tex_num;
   1354 
   1355 					if (word == 2) {
   1356 						saved_s2_set = 1;
   1357 						saved_s2 = data[i];
   1358 					}
   1359 					if (word == 4) {
   1360 						saved_s4_set = 1;
   1361 						saved_s4 = data[i];
   1362 					}
   1363 
   1364 					switch (word) {
   1365 					case 0:
   1366 						instr_out(ctx, i,
   1367 							  "S0: vbo offset: 0x%08x%s\n",
   1368 							  data[i] & (~1),
   1369 							  data[i] & 1 ?
   1370 							  ", auto cache invalidate disabled"
   1371 							  : "");
   1372 						break;
   1373 					case 1:
   1374 						instr_out(ctx, i,
   1375 							  "S1: vertex width: %i, vertex pitch: %i\n",
   1376 							  (data[i] >> 24) &
   1377 							  0x3f,
   1378 							  (data[i] >> 16) &
   1379 							  0x3f);
   1380 						break;
   1381 					case 2:
   1382 						instr_out(ctx, i,
   1383 							  "S2: texcoord formats: ");
   1384 						for (tex_num = 0;
   1385 						     tex_num < 8; tex_num++) {
   1386 							switch ((data[i] >>
   1387 								 tex_num *
   1388 								 4) & 0xf) {
   1389 							case 0:
   1390 								fprintf(out,
   1391 									"%i=2D ",
   1392 									tex_num);
   1393 								break;
   1394 							case 1:
   1395 								fprintf(out,
   1396 									"%i=3D ",
   1397 									tex_num);
   1398 								break;
   1399 							case 2:
   1400 								fprintf(out,
   1401 									"%i=4D ",
   1402 									tex_num);
   1403 								break;
   1404 							case 3:
   1405 								fprintf(out,
   1406 									"%i=1D ",
   1407 									tex_num);
   1408 								break;
   1409 							case 4:
   1410 								fprintf(out,
   1411 									"%i=2D_16 ",
   1412 									tex_num);
   1413 								break;
   1414 							case 5:
   1415 								fprintf(out,
   1416 									"%i=4D_16 ",
   1417 									tex_num);
   1418 								break;
   1419 							case 0xf:
   1420 								fprintf(out,
   1421 									"%i=NP ",
   1422 									tex_num);
   1423 								break;
   1424 							}
   1425 						}
   1426 						fprintf(out, "\n");
   1427 
   1428 						break;
   1429 					case 3:
   1430 						instr_out(ctx, i,
   1431 							  "S3: not documented\n");
   1432 						break;
   1433 					case 4:
   1434 						{
   1435 							const char *cullmode = "";
   1436 							const char *vfmt_xyzw = "";
   1437 							switch ((data[i] >> 13)
   1438 								& 0x3) {
   1439 							case 0:
   1440 								cullmode =
   1441 								    "both";
   1442 								break;
   1443 							case 1:
   1444 								cullmode =
   1445 								    "none";
   1446 								break;
   1447 							case 2:
   1448 								cullmode = "cw";
   1449 								break;
   1450 							case 3:
   1451 								cullmode =
   1452 								    "ccw";
   1453 								break;
   1454 							}
   1455 							switch (data[i] &
   1456 								(7 << 6 | 1 <<
   1457 								 2)) {
   1458 							case 1 << 6:
   1459 								vfmt_xyzw =
   1460 								    "XYZ,";
   1461 								break;
   1462 							case 2 << 6:
   1463 								vfmt_xyzw =
   1464 								    "XYZW,";
   1465 								break;
   1466 							case 3 << 6:
   1467 								vfmt_xyzw =
   1468 								    "XY,";
   1469 								break;
   1470 							case 4 << 6:
   1471 								vfmt_xyzw =
   1472 								    "XYW,";
   1473 								break;
   1474 							case 1 << 6 | 1 << 2:
   1475 								vfmt_xyzw =
   1476 								    "XYZF,";
   1477 								break;
   1478 							case 2 << 6 | 1 << 2:
   1479 								vfmt_xyzw =
   1480 								    "XYZWF,";
   1481 								break;
   1482 							case 3 << 6 | 1 << 2:
   1483 								vfmt_xyzw =
   1484 								    "XYF,";
   1485 								break;
   1486 							case 4 << 6 | 1 << 2:
   1487 								vfmt_xyzw =
   1488 								    "XYWF,";
   1489 								break;
   1490 							}
   1491 							instr_out(ctx, i,
   1492 								  "S4: point_width=%i, line_width=%.1f,"
   1493 								  "%s%s%s%s%s cullmode=%s, vfmt=%s%s%s%s%s%s "
   1494 								  "%s%s%s%s%s\n",
   1495 								  (data[i] >>
   1496 								   23) & 0x1ff,
   1497 								  ((data[i] >>
   1498 								    19) & 0xf) /
   1499 								  2.0,
   1500 								  data[i] & (0xf
   1501 									     <<
   1502 									     15)
   1503 								  ?
   1504 								  " flatshade="
   1505 								  : "",
   1506 								  data[i] & (1
   1507 									     <<
   1508 									     18)
   1509 								  ? "Alpha," :
   1510 								  "",
   1511 								  data[i] & (1
   1512 									     <<
   1513 									     17)
   1514 								  ? "Fog," : "",
   1515 								  data[i] & (1
   1516 									     <<
   1517 									     16)
   1518 								  ? "Specular,"
   1519 								  : "",
   1520 								  data[i] & (1
   1521 									     <<
   1522 									     15)
   1523 								  ? "Color," :
   1524 								  "", cullmode,
   1525 								  data[i] & (1
   1526 									     <<
   1527 									     12)
   1528 								  ?
   1529 								  "PointWidth,"
   1530 								  : "",
   1531 								  data[i] & (1
   1532 									     <<
   1533 									     11)
   1534 								  ? "SpecFog," :
   1535 								  "",
   1536 								  data[i] & (1
   1537 									     <<
   1538 									     10)
   1539 								  ? "Color," :
   1540 								  "",
   1541 								  data[i] & (1
   1542 									     <<
   1543 									     9)
   1544 								  ? "DepthOfs,"
   1545 								  : "",
   1546 								  vfmt_xyzw,
   1547 								  data[i] & (1
   1548 									     <<
   1549 									     9)
   1550 								  ? "FogParam,"
   1551 								  : "",
   1552 								  data[i] & (1
   1553 									     <<
   1554 									     5)
   1555 								  ?
   1556 								  "force default diffuse, "
   1557 								  : "",
   1558 								  data[i] & (1
   1559 									     <<
   1560 									     4)
   1561 								  ?
   1562 								  "force default specular, "
   1563 								  : "",
   1564 								  data[i] & (1
   1565 									     <<
   1566 									     3)
   1567 								  ?
   1568 								  "local depth ofs enable, "
   1569 								  : "",
   1570 								  data[i] & (1
   1571 									     <<
   1572 									     1)
   1573 								  ?
   1574 								  "point sprite enable, "
   1575 								  : "",
   1576 								  data[i] & (1
   1577 									     <<
   1578 									     0)
   1579 								  ?
   1580 								  "line AA enable, "
   1581 								  : "");
   1582 							break;
   1583 						}
   1584 					case 5:
   1585 						{
   1586 							instr_out(ctx, i,
   1587 								  "S5:%s%s%s%s%s"
   1588 								  "%s%s%s%s stencil_ref=0x%x, stencil_test=%s, "
   1589 								  "stencil_fail=%s, stencil_pass_z_fail=%s, "
   1590 								  "stencil_pass_z_pass=%s, %s%s%s%s\n",
   1591 								  data[i] & (0xf
   1592 									     <<
   1593 									     28)
   1594 								  ?
   1595 								  " write_disable="
   1596 								  : "",
   1597 								  data[i] & (1
   1598 									     <<
   1599 									     31)
   1600 								  ? "Alpha," :
   1601 								  "",
   1602 								  data[i] & (1
   1603 									     <<
   1604 									     30)
   1605 								  ? "Red," : "",
   1606 								  data[i] & (1
   1607 									     <<
   1608 									     29)
   1609 								  ? "Green," :
   1610 								  "",
   1611 								  data[i] & (1
   1612 									     <<
   1613 									     28)
   1614 								  ? "Blue," :
   1615 								  "",
   1616 								  data[i] & (1
   1617 									     <<
   1618 									     27)
   1619 								  ?
   1620 								  " force default point size,"
   1621 								  : "",
   1622 								  data[i] & (1
   1623 									     <<
   1624 									     26)
   1625 								  ?
   1626 								  " last pixel enable,"
   1627 								  : "",
   1628 								  data[i] & (1
   1629 									     <<
   1630 									     25)
   1631 								  ?
   1632 								  " global depth ofs enable,"
   1633 								  : "",
   1634 								  data[i] & (1
   1635 									     <<
   1636 									     24)
   1637 								  ?
   1638 								  " fog enable,"
   1639 								  : "",
   1640 								  (data[i] >>
   1641 								   16) & 0xff,
   1642 								  decode_compare_func
   1643 								  (data[i] >>
   1644 								   13),
   1645 								  decode_stencil_op
   1646 								  (data[i] >>
   1647 								   10),
   1648 								  decode_stencil_op
   1649 								  (data[i] >>
   1650 								   7),
   1651 								  decode_stencil_op
   1652 								  (data[i] >>
   1653 								   4),
   1654 								  data[i] & (1
   1655 									     <<
   1656 									     3)
   1657 								  ?
   1658 								  "stencil write enable, "
   1659 								  : "",
   1660 								  data[i] & (1
   1661 									     <<
   1662 									     2)
   1663 								  ?
   1664 								  "stencil test enable, "
   1665 								  : "",
   1666 								  data[i] & (1
   1667 									     <<
   1668 									     1)
   1669 								  ?
   1670 								  "color dither enable, "
   1671 								  : "",
   1672 								  data[i] & (1
   1673 									     <<
   1674 									     0)
   1675 								  ?
   1676 								  "logicop enable, "
   1677 								  : "");
   1678 						}
   1679 						break;
   1680 					case 6:
   1681 						instr_out(ctx, i,
   1682 							  "S6: %salpha_test=%s, alpha_ref=0x%x, "
   1683 							  "depth_test=%s, %ssrc_blnd_fct=%s, dst_blnd_fct=%s, "
   1684 							  "%s%stristrip_provoking_vertex=%i\n",
   1685 							  data[i] & (1 << 31) ?
   1686 							  "alpha test enable, "
   1687 							  : "",
   1688 							  decode_compare_func
   1689 							  (data[i] >> 28),
   1690 							  data[i] & (0xff <<
   1691 								     20),
   1692 							  decode_compare_func
   1693 							  (data[i] >> 16),
   1694 							  data[i] & (1 << 15) ?
   1695 							  "cbuf blend enable, "
   1696 							  : "",
   1697 							  decode_blend_fact(data
   1698 									    [i]
   1699 									    >>
   1700 									    8),
   1701 							  decode_blend_fact(data
   1702 									    [i]
   1703 									    >>
   1704 									    4),
   1705 							  data[i] & (1 << 3) ?
   1706 							  "depth write enable, "
   1707 							  : "",
   1708 							  data[i] & (1 << 2) ?
   1709 							  "cbuf write enable, "
   1710 							  : "",
   1711 							  data[i] & (0x3));
   1712 						break;
   1713 					case 7:
   1714 						instr_out(ctx, i,
   1715 							  "S7: depth offset constant: 0x%08x\n",
   1716 							  data[i]);
   1717 						break;
   1718 					}
   1719 				} else {
   1720 					instr_out(ctx, i,
   1721 						  "S%d: 0x%08x\n", word, data[i]);
   1722 				}
   1723 				i++;
   1724 			}
   1725 		}
   1726 		if (len != i) {
   1727 			fprintf(out,
   1728 				"Bad count in 3DSTATE_LOAD_STATE_IMMEDIATE_1\n");
   1729 		}
   1730 		return len;
   1731 	case 0x03:
   1732 		instr_out(ctx, 0,
   1733 			  "3DSTATE_LOAD_STATE_IMMEDIATE_2\n");
   1734 		len = (data[0] & 0x0000000f) + 2;
   1735 		i = 1;
   1736 		for (word = 6; word <= 14; word++) {
   1737 			if (data[0] & (1 << word)) {
   1738 				if (word == 6)
   1739 					instr_out(ctx, i++,
   1740 						  "TBCF\n");
   1741 				else if (word >= 7 && word <= 10) {
   1742 					instr_out(ctx, i++,
   1743 						  "TB%dC\n", word - 7);
   1744 					instr_out(ctx, i++,
   1745 						  "TB%dA\n", word - 7);
   1746 				} else if (word >= 11 && word <= 14) {
   1747 					instr_out(ctx, i,
   1748 						  "TM%dS0: offset=0x%08x, %s\n",
   1749 						  word - 11,
   1750 						  data[i] & 0xfffffffe,
   1751 						  data[i] & 1 ? "use fence" :
   1752 						  "");
   1753 					i++;
   1754 					instr_out(ctx, i,
   1755 						  "TM%dS1: height=%i, width=%i, %s\n",
   1756 						  word - 11, data[i] >> 21,
   1757 						  (data[i] >> 10) & 0x3ff,
   1758 						  data[i] & 2 ? (data[i] & 1 ?
   1759 								 "y-tiled" :
   1760 								 "x-tiled") :
   1761 						  "");
   1762 					i++;
   1763 					instr_out(ctx, i,
   1764 						  "TM%dS2: pitch=%i, \n",
   1765 						  word - 11,
   1766 						  ((data[i] >> 21) + 1) * 4);
   1767 					i++;
   1768 					instr_out(ctx, i++,
   1769 						  "TM%dS3\n", word - 11);
   1770 					instr_out(ctx, i++,
   1771 						  "TM%dS4: dflt color\n",
   1772 						  word - 11);
   1773 				}
   1774 			}
   1775 		}
   1776 		if (len != i) {
   1777 			fprintf(out,
   1778 				"Bad count in 3DSTATE_LOAD_STATE_IMMEDIATE_2\n");
   1779 		}
   1780 		return len;
   1781 	case 0x00:
   1782 		instr_out(ctx, 0, "3DSTATE_MAP_STATE\n");
   1783 		len = (data[0] & 0x0000003f) + 2;
   1784 		instr_out(ctx, 1, "mask\n");
   1785 
   1786 		i = 2;
   1787 		for (map = 0; map <= 15; map++) {
   1788 			if (data[1] & (1 << map)) {
   1789 				int width, height, pitch, dword;
   1790 				const char *tiling;
   1791 
   1792 				dword = data[i];
   1793 				instr_out(ctx, i++,
   1794 					  "map %d MS2 %s%s%s\n", map,
   1795 					  dword & (1 << 31) ?
   1796 					  "untrusted surface, " : "",
   1797 					  dword & (1 << 1) ?
   1798 					  "vertical line stride enable, " : "",
   1799 					  dword & (1 << 0) ?
   1800 					  "vertical ofs enable, " : "");
   1801 
   1802 				dword = data[i];
   1803 				width = ((dword >> 10) & ((1 << 11) - 1)) + 1;
   1804 				height = ((dword >> 21) & ((1 << 11) - 1)) + 1;
   1805 
   1806 				tiling = "none";
   1807 				if (dword & (1 << 2))
   1808 					tiling = "fenced";
   1809 				else if (dword & (1 << 1))
   1810 					tiling = dword & (1 << 0) ? "Y" : "X";
   1811 				type = " BAD";
   1812 				format = "BAD";
   1813 				switch ((dword >> 7) & 0x7) {
   1814 				case 1:
   1815 					type = "8b";
   1816 					switch ((dword >> 3) & 0xf) {
   1817 					case 0:
   1818 						format = "I";
   1819 						break;
   1820 					case 1:
   1821 						format = "L";
   1822 						break;
   1823 					case 4:
   1824 						format = "A";
   1825 						break;
   1826 					case 5:
   1827 						format = " mono";
   1828 						break;
   1829 					}
   1830 					break;
   1831 				case 2:
   1832 					type = "16b";
   1833 					switch ((dword >> 3) & 0xf) {
   1834 					case 0:
   1835 						format = " rgb565";
   1836 						break;
   1837 					case 1:
   1838 						format = " argb1555";
   1839 						break;
   1840 					case 2:
   1841 						format = " argb4444";
   1842 						break;
   1843 					case 5:
   1844 						format = " ay88";
   1845 						break;
   1846 					case 6:
   1847 						format = " bump655";
   1848 						break;
   1849 					case 7:
   1850 						format = "I";
   1851 						break;
   1852 					case 8:
   1853 						format = "L";
   1854 						break;
   1855 					case 9:
   1856 						format = "A";
   1857 						break;
   1858 					}
   1859 					break;
   1860 				case 3:
   1861 					type = "32b";
   1862 					switch ((dword >> 3) & 0xf) {
   1863 					case 0:
   1864 						format = " argb8888";
   1865 						break;
   1866 					case 1:
   1867 						format = " abgr8888";
   1868 						break;
   1869 					case 2:
   1870 						format = " xrgb8888";
   1871 						break;
   1872 					case 3:
   1873 						format = " xbgr8888";
   1874 						break;
   1875 					case 4:
   1876 						format = " qwvu8888";
   1877 						break;
   1878 					case 5:
   1879 						format = " axvu8888";
   1880 						break;
   1881 					case 6:
   1882 						format = " lxvu8888";
   1883 						break;
   1884 					case 7:
   1885 						format = " xlvu8888";
   1886 						break;
   1887 					case 8:
   1888 						format = " argb2101010";
   1889 						break;
   1890 					case 9:
   1891 						format = " abgr2101010";
   1892 						break;
   1893 					case 10:
   1894 						format = " awvu2101010";
   1895 						break;
   1896 					case 11:
   1897 						format = " gr1616";
   1898 						break;
   1899 					case 12:
   1900 						format = " vu1616";
   1901 						break;
   1902 					case 13:
   1903 						format = " xI824";
   1904 						break;
   1905 					case 14:
   1906 						format = " xA824";
   1907 						break;
   1908 					case 15:
   1909 						format = " xL824";
   1910 						break;
   1911 					}
   1912 					break;
   1913 				case 5:
   1914 					type = "422";
   1915 					switch ((dword >> 3) & 0xf) {
   1916 					case 0:
   1917 						format = " yuv_swapy";
   1918 						break;
   1919 					case 1:
   1920 						format = " yuv";
   1921 						break;
   1922 					case 2:
   1923 						format = " yuv_swapuv";
   1924 						break;
   1925 					case 3:
   1926 						format = " yuv_swapuvy";
   1927 						break;
   1928 					}
   1929 					break;
   1930 				case 6:
   1931 					type = "compressed";
   1932 					switch ((dword >> 3) & 0x7) {
   1933 					case 0:
   1934 						format = " dxt1";
   1935 						break;
   1936 					case 1:
   1937 						format = " dxt2_3";
   1938 						break;
   1939 					case 2:
   1940 						format = " dxt4_5";
   1941 						break;
   1942 					case 3:
   1943 						format = " fxt1";
   1944 						break;
   1945 					case 4:
   1946 						format = " dxt1_rb";
   1947 						break;
   1948 					}
   1949 					break;
   1950 				case 7:
   1951 					type = "4b indexed";
   1952 					switch ((dword >> 3) & 0xf) {
   1953 					case 7:
   1954 						format = " argb8888";
   1955 						break;
   1956 					}
   1957 					break;
   1958 				}
   1959 				dword = data[i];
   1960 				instr_out(ctx, i++,
   1961 					  "map %d MS3 [width=%d, height=%d, format=%s%s, tiling=%s%s]\n",
   1962 					  map, width, height, type, format,
   1963 					  tiling,
   1964 					  dword & (1 << 9) ? " palette select" :
   1965 					  "");
   1966 
   1967 				dword = data[i];
   1968 				pitch =
   1969 				    4 * (((dword >> 21) & ((1 << 11) - 1)) + 1);
   1970 				instr_out(ctx, i++,
   1971 					  "map %d MS4 [pitch=%d, max_lod=%i, vol_depth=%i, cube_face_ena=%x, %s]\n",
   1972 					  map, pitch, (dword >> 9) & 0x3f,
   1973 					  dword & 0xff, (dword >> 15) & 0x3f,
   1974 					  dword & (1 << 8) ? "miplayout legacy"
   1975 					  : "miplayout right");
   1976 			}
   1977 		}
   1978 		if (len != i) {
   1979 			fprintf(out, "Bad count in 3DSTATE_MAP_STATE\n");
   1980 			return len;
   1981 		}
   1982 		return len;
   1983 	case 0x06:
   1984 		instr_out(ctx, 0,
   1985 			  "3DSTATE_PIXEL_SHADER_CONSTANTS\n");
   1986 		len = (data[0] & 0x000000ff) + 2;
   1987 
   1988 		i = 2;
   1989 		for (c = 0; c <= 31; c++) {
   1990 			if (data[1] & (1 << c)) {
   1991 				instr_out(ctx, i, "C%d.X = %f\n", c,
   1992 					  int_as_float(data[i]));
   1993 				i++;
   1994 				instr_out(ctx, i, "C%d.Y = %f\n",
   1995 					  c, int_as_float(data[i]));
   1996 				i++;
   1997 				instr_out(ctx, i, "C%d.Z = %f\n",
   1998 					  c, int_as_float(data[i]));
   1999 				i++;
   2000 				instr_out(ctx, i, "C%d.W = %f\n",
   2001 					  c, int_as_float(data[i]));
   2002 				i++;
   2003 			}
   2004 		}
   2005 		if (len != i) {
   2006 			fprintf(out,
   2007 				"Bad count in 3DSTATE_PIXEL_SHADER_CONSTANTS\n");
   2008 		}
   2009 		return len;
   2010 	case 0x05:
   2011 		instr_out(ctx, 0, "3DSTATE_PIXEL_SHADER_PROGRAM\n");
   2012 		len = (data[0] & 0x000000ff) + 2;
   2013 		if ((len - 1) % 3 != 0 || len > 370) {
   2014 			fprintf(out,
   2015 				"Bad count in 3DSTATE_PIXEL_SHADER_PROGRAM\n");
   2016 		}
   2017 		i = 1;
   2018 		for (instr = 0; instr < (len - 1) / 3; instr++) {
   2019 			char instr_prefix[10];
   2020 
   2021 			sprintf(instr_prefix, "PS%03d", instr);
   2022 			i915_decode_instruction(ctx, i,
   2023 						instr_prefix);
   2024 			i += 3;
   2025 		}
   2026 		return len;
   2027 	case 0x01:
   2028 		if (IS_GEN2(devid))
   2029 			break;
   2030 		instr_out(ctx, 0, "3DSTATE_SAMPLER_STATE\n");
   2031 		instr_out(ctx, 1, "mask\n");
   2032 		len = (data[0] & 0x0000003f) + 2;
   2033 		i = 2;
   2034 		for (sampler = 0; sampler <= 15; sampler++) {
   2035 			if (data[1] & (1 << sampler)) {
   2036 				uint32_t dword;
   2037 				const char *mip_filter = "";
   2038 
   2039 				dword = data[i];
   2040 				switch ((dword >> 20) & 0x3) {
   2041 				case 0:
   2042 					mip_filter = "none";
   2043 					break;
   2044 				case 1:
   2045 					mip_filter = "nearest";
   2046 					break;
   2047 				case 3:
   2048 					mip_filter = "linear";
   2049 					break;
   2050 				}
   2051 				instr_out(ctx, i++,
   2052 					  "sampler %d SS2:%s%s%s "
   2053 					  "base_mip_level=%i, mip_filter=%s, mag_filter=%s, min_filter=%s "
   2054 					  "lod_bias=%.2f,%s max_aniso=%i, shadow_func=%s\n",
   2055 					  sampler,
   2056 					  dword & (1 << 31) ? " reverse gamma,"
   2057 					  : "",
   2058 					  dword & (1 << 30) ? " packed2planar,"
   2059 					  : "",
   2060 					  dword & (1 << 29) ?
   2061 					  " colorspace conversion," : "",
   2062 					  (dword >> 22) & 0x1f, mip_filter,
   2063 					  decode_sample_filter(dword >> 17),
   2064 					  decode_sample_filter(dword >> 14),
   2065 					  ((dword >> 5) & 0x1ff) / (0x10 * 1.0),
   2066 					  dword & (1 << 4) ? " shadow," : "",
   2067 					  dword & (1 << 3) ? 4 : 2,
   2068 					  decode_compare_func(dword));
   2069 				dword = data[i];
   2070 				instr_out(ctx, i++,
   2071 					  "sampler %d SS3: min_lod=%.2f,%s "
   2072 					  "tcmode_x=%s, tcmode_y=%s, tcmode_z=%s,%s texmap_idx=%i,%s\n",
   2073 					  sampler,
   2074 					  ((dword >> 24) & 0xff) / (0x10 * 1.0),
   2075 					  dword & (1 << 17) ?
   2076 					  " kill pixel enable," : "",
   2077 					  decode_tex_coord_mode(dword >> 12),
   2078 					  decode_tex_coord_mode(dword >> 9),
   2079 					  decode_tex_coord_mode(dword >> 6),
   2080 					  dword & (1 << 5) ?
   2081 					  " normalized coords," : "",
   2082 					  (dword >> 1) & 0xf,
   2083 					  dword & (1 << 0) ? " deinterlacer," :
   2084 					  "");
   2085 				dword = data[i];
   2086 				instr_out(ctx, i++,
   2087 					  "sampler %d SS4: border color\n",
   2088 					  sampler);
   2089 			}
   2090 		}
   2091 		if (len != i) {
   2092 			fprintf(out, "Bad count in 3DSTATE_SAMPLER_STATE\n");
   2093 		}
   2094 		return len;
   2095 	case 0x85:
   2096 		len = (data[0] & 0x0000000f) + 2;
   2097 
   2098 		if (len != 2)
   2099 			fprintf(out,
   2100 				"Bad count in 3DSTATE_DEST_BUFFER_VARIABLES\n");
   2101 
   2102 		instr_out(ctx, 0,
   2103 			  "3DSTATE_DEST_BUFFER_VARIABLES\n");
   2104 
   2105 		switch ((data[1] >> 8) & 0xf) {
   2106 		case 0x0:
   2107 			format = "g8";
   2108 			break;
   2109 		case 0x1:
   2110 			format = "x1r5g5b5";
   2111 			break;
   2112 		case 0x2:
   2113 			format = "r5g6b5";
   2114 			break;
   2115 		case 0x3:
   2116 			format = "a8r8g8b8";
   2117 			break;
   2118 		case 0x4:
   2119 			format = "ycrcb_swapy";
   2120 			break;
   2121 		case 0x5:
   2122 			format = "ycrcb_normal";
   2123 			break;
   2124 		case 0x6:
   2125 			format = "ycrcb_swapuv";
   2126 			break;
   2127 		case 0x7:
   2128 			format = "ycrcb_swapuvy";
   2129 			break;
   2130 		case 0x8:
   2131 			format = "a4r4g4b4";
   2132 			break;
   2133 		case 0x9:
   2134 			format = "a1r5g5b5";
   2135 			break;
   2136 		case 0xa:
   2137 			format = "a2r10g10b10";
   2138 			break;
   2139 		default:
   2140 			format = "BAD";
   2141 			break;
   2142 		}
   2143 		switch ((data[1] >> 2) & 0x3) {
   2144 		case 0x0:
   2145 			zformat = "u16";
   2146 			break;
   2147 		case 0x1:
   2148 			zformat = "f16";
   2149 			break;
   2150 		case 0x2:
   2151 			zformat = "u24x8";
   2152 			break;
   2153 		default:
   2154 			zformat = "BAD";
   2155 			break;
   2156 		}
   2157 		instr_out(ctx, 1,
   2158 			  "%s format, %s depth format, early Z %sabled\n",
   2159 			  format, zformat,
   2160 			  (data[1] & (1 << 31)) ? "en" : "dis");
   2161 		return len;
   2162 
   2163 	case 0x8e:
   2164 		{
   2165 			const char *name, *tiling;
   2166 
   2167 			len = (data[0] & 0x0000000f) + 2;
   2168 			if (len != 3)
   2169 				fprintf(out,
   2170 					"Bad count in 3DSTATE_BUFFER_INFO\n");
   2171 
   2172 			switch ((data[1] >> 24) & 0x7) {
   2173 			case 0x3:
   2174 				name = "color";
   2175 				break;
   2176 			case 0x7:
   2177 				name = "depth";
   2178 				break;
   2179 			default:
   2180 				name = "unknown";
   2181 				break;
   2182 			}
   2183 
   2184 			tiling = "none";
   2185 			if (data[1] & (1 << 23))
   2186 				tiling = "fenced";
   2187 			else if (data[1] & (1 << 22))
   2188 				tiling = data[1] & (1 << 21) ? "Y" : "X";
   2189 
   2190 			instr_out(ctx, 0, "3DSTATE_BUFFER_INFO\n");
   2191 			instr_out(ctx, 1,
   2192 				  "%s, tiling = %s, pitch=%d\n", name, tiling,
   2193 				  data[1] & 0xffff);
   2194 
   2195 			instr_out(ctx, 2, "address\n");
   2196 			return len;
   2197 		}
   2198 	case 0x81:
   2199 		len = (data[0] & 0x0000000f) + 2;
   2200 
   2201 		if (len != 3)
   2202 			fprintf(out,
   2203 				"Bad count in 3DSTATE_SCISSOR_RECTANGLE\n");
   2204 
   2205 		instr_out(ctx, 0, "3DSTATE_SCISSOR_RECTANGLE\n");
   2206 		instr_out(ctx, 1, "(%d,%d)\n",
   2207 			  data[1] & 0xffff, data[1] >> 16);
   2208 		instr_out(ctx, 2, "(%d,%d)\n",
   2209 			  data[2] & 0xffff, data[2] >> 16);
   2210 
   2211 		return len;
   2212 	case 0x80:
   2213 		len = (data[0] & 0x0000000f) + 2;
   2214 
   2215 		if (len != 5)
   2216 			fprintf(out,
   2217 				"Bad count in 3DSTATE_DRAWING_RECTANGLE\n");
   2218 
   2219 		instr_out(ctx, 0, "3DSTATE_DRAWING_RECTANGLE\n");
   2220 		instr_out(ctx, 1, "%s\n",
   2221 			  data[1] & (1 << 30) ? "depth ofs disabled " : "");
   2222 		instr_out(ctx, 2, "(%d,%d)\n",
   2223 			  data[2] & 0xffff, data[2] >> 16);
   2224 		instr_out(ctx, 3, "(%d,%d)\n",
   2225 			  data[3] & 0xffff, data[3] >> 16);
   2226 		instr_out(ctx, 4, "(%d,%d)\n",
   2227 			  data[4] & 0xffff, data[4] >> 16);
   2228 
   2229 		return len;
   2230 	case 0x9c:
   2231 		len = (data[0] & 0x0000000f) + 2;
   2232 
   2233 		if (len != 7)
   2234 			fprintf(out, "Bad count in 3DSTATE_CLEAR_PARAMETERS\n");
   2235 
   2236 		instr_out(ctx, 0, "3DSTATE_CLEAR_PARAMETERS\n");
   2237 		instr_out(ctx, 1, "prim_type=%s, clear=%s%s%s\n",
   2238 			  data[1] & (1 << 16) ? "CLEAR_RECT" : "ZONE_INIT",
   2239 			  data[1] & (1 << 2) ? "color," : "",
   2240 			  data[1] & (1 << 1) ? "depth," : "",
   2241 			  data[1] & (1 << 0) ? "stencil," : "");
   2242 		instr_out(ctx, 2, "clear color\n");
   2243 		instr_out(ctx, 3, "clear depth/stencil\n");
   2244 		instr_out(ctx, 4, "color value (rgba8888)\n");
   2245 		instr_out(ctx, 5, "depth value %f\n",
   2246 			  int_as_float(data[5]));
   2247 		instr_out(ctx, 6, "clear stencil\n");
   2248 		return len;
   2249 	}
   2250 
   2251 	for (idx = 0; idx < ARRAY_SIZE(opcodes_3d_1d); idx++) {
   2252 		opcode_3d_1d = &opcodes_3d_1d[idx];
   2253 		if (opcode_3d_1d->i830_only && !IS_GEN2(devid))
   2254 			continue;
   2255 
   2256 		if (((data[0] & 0x00ff0000) >> 16) == opcode_3d_1d->opcode) {
   2257 			len = 1;
   2258 
   2259 			instr_out(ctx, 0, "%s\n",
   2260 				  opcode_3d_1d->name);
   2261 			if (opcode_3d_1d->max_len > 1) {
   2262 				len = (data[0] & 0x0000ffff) + 2;
   2263 				if (len < opcode_3d_1d->min_len ||
   2264 				    len > opcode_3d_1d->max_len) {
   2265 					fprintf(out, "Bad count in %s\n",
   2266 						opcode_3d_1d->name);
   2267 				}
   2268 			}
   2269 
   2270 			for (i = 1; i < len; i++) {
   2271 				instr_out(ctx, i, "dword %d\n", i);
   2272 			}
   2273 
   2274 			return len;
   2275 		}
   2276 	}
   2277 
   2278 	instr_out(ctx, 0, "3D UNKNOWN: 3d_1d opcode = 0x%x\n",
   2279 		  opcode);
   2280 	return 1;
   2281 }
   2282 
   2283 static int
   2284 decode_3d_primitive(struct drm_intel_decode *ctx)
   2285 {
   2286 	uint32_t *data = ctx->data;
   2287 	uint32_t count = ctx->count;
   2288 	char immediate = (data[0] & (1 << 23)) == 0;
   2289 	unsigned int len, i, j, ret;
   2290 	const char *primtype;
   2291 	int original_s2 = saved_s2;
   2292 	int original_s4 = saved_s4;
   2293 
   2294 	switch ((data[0] >> 18) & 0xf) {
   2295 	case 0x0:
   2296 		primtype = "TRILIST";
   2297 		break;
   2298 	case 0x1:
   2299 		primtype = "TRISTRIP";
   2300 		break;
   2301 	case 0x2:
   2302 		primtype = "TRISTRIP_REVERSE";
   2303 		break;
   2304 	case 0x3:
   2305 		primtype = "TRIFAN";
   2306 		break;
   2307 	case 0x4:
   2308 		primtype = "POLYGON";
   2309 		break;
   2310 	case 0x5:
   2311 		primtype = "LINELIST";
   2312 		break;
   2313 	case 0x6:
   2314 		primtype = "LINESTRIP";
   2315 		break;
   2316 	case 0x7:
   2317 		primtype = "RECTLIST";
   2318 		break;
   2319 	case 0x8:
   2320 		primtype = "POINTLIST";
   2321 		break;
   2322 	case 0x9:
   2323 		primtype = "DIB";
   2324 		break;
   2325 	case 0xa:
   2326 		primtype = "CLEAR_RECT";
   2327 		saved_s4 = 3 << 6;
   2328 		saved_s2 = ~0;
   2329 		break;
   2330 	default:
   2331 		primtype = "unknown";
   2332 		break;
   2333 	}
   2334 
   2335 	/* XXX: 3DPRIM_DIB not supported */
   2336 	if (immediate) {
   2337 		len = (data[0] & 0x0003ffff) + 2;
   2338 		instr_out(ctx, 0, "3DPRIMITIVE inline %s\n",
   2339 			  primtype);
   2340 		if (count < len)
   2341 			BUFFER_FAIL(count, len, "3DPRIMITIVE inline");
   2342 		if (!saved_s2_set || !saved_s4_set) {
   2343 			fprintf(out, "unknown vertex format\n");
   2344 			for (i = 1; i < len; i++) {
   2345 				instr_out(ctx, i,
   2346 					  "           vertex data (%f float)\n",
   2347 					  int_as_float(data[i]));
   2348 			}
   2349 		} else {
   2350 			unsigned int vertex = 0;
   2351 			for (i = 1; i < len;) {
   2352 				unsigned int tc;
   2353 
   2354 #define VERTEX_OUT(fmt, ...) do {					\
   2355     if (i < len)							\
   2356 	instr_out(ctx, i, " V%d."fmt"\n", vertex, __VA_ARGS__); \
   2357     else								\
   2358 	fprintf(out, " missing data in V%d\n", vertex);			\
   2359     i++;								\
   2360 } while (0)
   2361 
   2362 				VERTEX_OUT("X = %f", int_as_float(data[i]));
   2363 				VERTEX_OUT("Y = %f", int_as_float(data[i]));
   2364 				switch (saved_s4 >> 6 & 0x7) {
   2365 				case 0x1:
   2366 					VERTEX_OUT("Z = %f",
   2367 						   int_as_float(data[i]));
   2368 					break;
   2369 				case 0x2:
   2370 					VERTEX_OUT("Z = %f",
   2371 						   int_as_float(data[i]));
   2372 					VERTEX_OUT("W = %f",
   2373 						   int_as_float(data[i]));
   2374 					break;
   2375 				case 0x3:
   2376 					break;
   2377 				case 0x4:
   2378 					VERTEX_OUT("W = %f",
   2379 						   int_as_float(data[i]));
   2380 					break;
   2381 				default:
   2382 					fprintf(out, "bad S4 position mask\n");
   2383 				}
   2384 
   2385 				if (saved_s4 & (1 << 10)) {
   2386 					VERTEX_OUT
   2387 					    ("color = (A=0x%02x, R=0x%02x, G=0x%02x, "
   2388 					     "B=0x%02x)", data[i] >> 24,
   2389 					     (data[i] >> 16) & 0xff,
   2390 					     (data[i] >> 8) & 0xff,
   2391 					     data[i] & 0xff);
   2392 				}
   2393 				if (saved_s4 & (1 << 11)) {
   2394 					VERTEX_OUT
   2395 					    ("spec = (A=0x%02x, R=0x%02x, G=0x%02x, "
   2396 					     "B=0x%02x)", data[i] >> 24,
   2397 					     (data[i] >> 16) & 0xff,
   2398 					     (data[i] >> 8) & 0xff,
   2399 					     data[i] & 0xff);
   2400 				}
   2401 				if (saved_s4 & (1 << 12))
   2402 					VERTEX_OUT("width = 0x%08x)", data[i]);
   2403 
   2404 				for (tc = 0; tc <= 7; tc++) {
   2405 					switch ((saved_s2 >> (tc * 4)) & 0xf) {
   2406 					case 0x0:
   2407 						VERTEX_OUT("T%d.X = %f", tc,
   2408 							   int_as_float(data
   2409 									[i]));
   2410 						VERTEX_OUT("T%d.Y = %f", tc,
   2411 							   int_as_float(data
   2412 									[i]));
   2413 						break;
   2414 					case 0x1:
   2415 						VERTEX_OUT("T%d.X = %f", tc,
   2416 							   int_as_float(data
   2417 									[i]));
   2418 						VERTEX_OUT("T%d.Y = %f", tc,
   2419 							   int_as_float(data
   2420 									[i]));
   2421 						VERTEX_OUT("T%d.Z = %f", tc,
   2422 							   int_as_float(data
   2423 									[i]));
   2424 						break;
   2425 					case 0x2:
   2426 						VERTEX_OUT("T%d.X = %f", tc,
   2427 							   int_as_float(data
   2428 									[i]));
   2429 						VERTEX_OUT("T%d.Y = %f", tc,
   2430 							   int_as_float(data
   2431 									[i]));
   2432 						VERTEX_OUT("T%d.Z = %f", tc,
   2433 							   int_as_float(data
   2434 									[i]));
   2435 						VERTEX_OUT("T%d.W = %f", tc,
   2436 							   int_as_float(data
   2437 									[i]));
   2438 						break;
   2439 					case 0x3:
   2440 						VERTEX_OUT("T%d.X = %f", tc,
   2441 							   int_as_float(data
   2442 									[i]));
   2443 						break;
   2444 					case 0x4:
   2445 						VERTEX_OUT
   2446 						    ("T%d.XY = 0x%08x half-float",
   2447 						     tc, data[i]);
   2448 						break;
   2449 					case 0x5:
   2450 						VERTEX_OUT
   2451 						    ("T%d.XY = 0x%08x half-float",
   2452 						     tc, data[i]);
   2453 						VERTEX_OUT
   2454 						    ("T%d.ZW = 0x%08x half-float",
   2455 						     tc, data[i]);
   2456 						break;
   2457 					case 0xf:
   2458 						break;
   2459 					default:
   2460 						fprintf(out,
   2461 							"bad S2.T%d format\n",
   2462 							tc);
   2463 					}
   2464 				}
   2465 				vertex++;
   2466 			}
   2467 		}
   2468 
   2469 		ret = len;
   2470 	} else {
   2471 		/* indirect vertices */
   2472 		len = data[0] & 0x0000ffff;	/* index count */
   2473 		if (data[0] & (1 << 17)) {
   2474 			/* random vertex access */
   2475 			if (count < (len + 1) / 2 + 1) {
   2476 				BUFFER_FAIL(count, (len + 1) / 2 + 1,
   2477 					    "3DPRIMITIVE random indirect");
   2478 			}
   2479 			instr_out(ctx, 0,
   2480 				  "3DPRIMITIVE random indirect %s (%d)\n",
   2481 				  primtype, len);
   2482 			if (len == 0) {
   2483 				/* vertex indices continue until 0xffff is
   2484 				 * found
   2485 				 */
   2486 				for (i = 1; i < count; i++) {
   2487 					if ((data[i] & 0xffff) == 0xffff) {
   2488 						instr_out(ctx, i,
   2489 							  "    indices: (terminator)\n");
   2490 						ret = i;
   2491 						goto out;
   2492 					} else if ((data[i] >> 16) == 0xffff) {
   2493 						instr_out(ctx, i,
   2494 							  "    indices: 0x%04x, (terminator)\n",
   2495 							  data[i] & 0xffff);
   2496 						ret = i;
   2497 						goto out;
   2498 					} else {
   2499 						instr_out(ctx, i,
   2500 							  "    indices: 0x%04x, 0x%04x\n",
   2501 							  data[i] & 0xffff,
   2502 							  data[i] >> 16);
   2503 					}
   2504 				}
   2505 				fprintf(out,
   2506 					"3DPRIMITIVE: no terminator found in index buffer\n");
   2507 				ret = count;
   2508 				goto out;
   2509 			} else {
   2510 				/* fixed size vertex index buffer */
   2511 				for (j = 1, i = 0; i < len; i += 2, j++) {
   2512 					if (i * 2 == len - 1) {
   2513 						instr_out(ctx, j,
   2514 							  "    indices: 0x%04x\n",
   2515 							  data[j] & 0xffff);
   2516 					} else {
   2517 						instr_out(ctx, j,
   2518 							  "    indices: 0x%04x, 0x%04x\n",
   2519 							  data[j] & 0xffff,
   2520 							  data[j] >> 16);
   2521 					}
   2522 				}
   2523 			}
   2524 			ret = (len + 1) / 2 + 1;
   2525 			goto out;
   2526 		} else {
   2527 			/* sequential vertex access */
   2528 			instr_out(ctx, 0,
   2529 				  "3DPRIMITIVE sequential indirect %s, %d starting from "
   2530 				  "%d\n", primtype, len, data[1] & 0xffff);
   2531 			instr_out(ctx, 1, "           start\n");
   2532 			ret = 2;
   2533 			goto out;
   2534 		}
   2535 	}
   2536 
   2537 out:
   2538 	saved_s2 = original_s2;
   2539 	saved_s4 = original_s4;
   2540 	return ret;
   2541 }
   2542 
   2543 static int
   2544 decode_3d(struct drm_intel_decode *ctx)
   2545 {
   2546 	uint32_t opcode;
   2547 	unsigned int idx;
   2548 	uint32_t *data = ctx->data;
   2549 
   2550 	struct {
   2551 		uint32_t opcode;
   2552 		unsigned int min_len;
   2553 		unsigned int max_len;
   2554 		const char *name;
   2555 	} opcodes_3d[] = {
   2556 		{ 0x06, 1, 1, "3DSTATE_ANTI_ALIASING" },
   2557 		{ 0x08, 1, 1, "3DSTATE_BACKFACE_STENCIL_OPS" },
   2558 		{ 0x09, 1, 1, "3DSTATE_BACKFACE_STENCIL_MASKS" },
   2559 		{ 0x16, 1, 1, "3DSTATE_COORD_SET_BINDINGS" },
   2560 		{ 0x15, 1, 1, "3DSTATE_FOG_COLOR" },
   2561 		{ 0x0b, 1, 1, "3DSTATE_INDEPENDENT_ALPHA_BLEND" },
   2562 		{ 0x0d, 1, 1, "3DSTATE_MODES_4" },
   2563 		{ 0x0c, 1, 1, "3DSTATE_MODES_5" },
   2564 		{ 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES"},
   2565 	}, *opcode_3d;
   2566 
   2567 	opcode = (data[0] & 0x1f000000) >> 24;
   2568 
   2569 	switch (opcode) {
   2570 	case 0x1f:
   2571 		return decode_3d_primitive(ctx);
   2572 	case 0x1d:
   2573 		return decode_3d_1d(ctx);
   2574 	case 0x1c:
   2575 		return decode_3d_1c(ctx);
   2576 	}
   2577 
   2578 	for (idx = 0; idx < ARRAY_SIZE(opcodes_3d); idx++) {
   2579 		opcode_3d = &opcodes_3d[idx];
   2580 		if (opcode == opcode_3d->opcode) {
   2581 			unsigned int len = 1, i;
   2582 
   2583 			instr_out(ctx, 0, "%s\n", opcode_3d->name);
   2584 			if (opcode_3d->max_len > 1) {
   2585 				len = (data[0] & 0xff) + 2;
   2586 				if (len < opcode_3d->min_len ||
   2587 				    len > opcode_3d->max_len) {
   2588 					fprintf(out, "Bad count in %s\n",
   2589 						opcode_3d->name);
   2590 				}
   2591 			}
   2592 
   2593 			for (i = 1; i < len; i++) {
   2594 				instr_out(ctx, i, "dword %d\n", i);
   2595 			}
   2596 			return len;
   2597 		}
   2598 	}
   2599 
   2600 	instr_out(ctx, 0, "3D UNKNOWN: 3d opcode = 0x%x\n", opcode);
   2601 	return 1;
   2602 }
   2603 
   2604 static const char *get_965_surfacetype(unsigned int surfacetype)
   2605 {
   2606 	switch (surfacetype) {
   2607 	case 0:
   2608 		return "1D";
   2609 	case 1:
   2610 		return "2D";
   2611 	case 2:
   2612 		return "3D";
   2613 	case 3:
   2614 		return "CUBE";
   2615 	case 4:
   2616 		return "BUFFER";
   2617 	case 7:
   2618 		return "NULL";
   2619 	default:
   2620 		return "unknown";
   2621 	}
   2622 }
   2623 
   2624 static const char *get_965_depthformat(unsigned int depthformat)
   2625 {
   2626 	switch (depthformat) {
   2627 	case 0:
   2628 		return "s8_z24float";
   2629 	case 1:
   2630 		return "z32float";
   2631 	case 2:
   2632 		return "z24s8";
   2633 	case 5:
   2634 		return "z16";
   2635 	default:
   2636 		return "unknown";
   2637 	}
   2638 }
   2639 
   2640 static const char *get_965_element_component(uint32_t data, int component)
   2641 {
   2642 	uint32_t component_control = (data >> (16 + (3 - component) * 4)) & 0x7;
   2643 
   2644 	switch (component_control) {
   2645 	case 0:
   2646 		return "nostore";
   2647 	case 1:
   2648 		switch (component) {
   2649 		case 0:
   2650 			return "X";
   2651 		case 1:
   2652 			return "Y";
   2653 		case 2:
   2654 			return "Z";
   2655 		case 3:
   2656 			return "W";
   2657 		default:
   2658 			return "fail";
   2659 		}
   2660 	case 2:
   2661 		return "0.0";
   2662 	case 3:
   2663 		return "1.0";
   2664 	case 4:
   2665 		return "0x1";
   2666 	case 5:
   2667 		return "VID";
   2668 	default:
   2669 		return "fail";
   2670 	}
   2671 }
   2672 
   2673 static const char *get_965_prim_type(uint32_t primtype)
   2674 {
   2675 	switch (primtype) {
   2676 	case 0x01:
   2677 		return "point list";
   2678 	case 0x02:
   2679 		return "line list";
   2680 	case 0x03:
   2681 		return "line strip";
   2682 	case 0x04:
   2683 		return "tri list";
   2684 	case 0x05:
   2685 		return "tri strip";
   2686 	case 0x06:
   2687 		return "tri fan";
   2688 	case 0x07:
   2689 		return "quad list";
   2690 	case 0x08:
   2691 		return "quad strip";
   2692 	case 0x09:
   2693 		return "line list adj";
   2694 	case 0x0a:
   2695 		return "line strip adj";
   2696 	case 0x0b:
   2697 		return "tri list adj";
   2698 	case 0x0c:
   2699 		return "tri strip adj";
   2700 	case 0x0d:
   2701 		return "tri strip reverse";
   2702 	case 0x0e:
   2703 		return "polygon";
   2704 	case 0x0f:
   2705 		return "rect list";
   2706 	case 0x10:
   2707 		return "line loop";
   2708 	case 0x11:
   2709 		return "point list bf";
   2710 	case 0x12:
   2711 		return "line strip cont";
   2712 	case 0x13:
   2713 		return "line strip bf";
   2714 	case 0x14:
   2715 		return "line strip cont bf";
   2716 	case 0x15:
   2717 		return "tri fan no stipple";
   2718 	default:
   2719 		return "fail";
   2720 	}
   2721 }
   2722 
   2723 static int
   2724 i965_decode_urb_fence(struct drm_intel_decode *ctx, int len)
   2725 {
   2726 	uint32_t vs_fence, clip_fence, gs_fence, sf_fence, vfe_fence, cs_fence;
   2727 	uint32_t *data = ctx->data;
   2728 
   2729 	if (len != 3)
   2730 		fprintf(out, "Bad count in URB_FENCE\n");
   2731 
   2732 	vs_fence = data[1] & 0x3ff;
   2733 	gs_fence = (data[1] >> 10) & 0x3ff;
   2734 	clip_fence = (data[1] >> 20) & 0x3ff;
   2735 	sf_fence = data[2] & 0x3ff;
   2736 	vfe_fence = (data[2] >> 10) & 0x3ff;
   2737 	cs_fence = (data[2] >> 20) & 0x7ff;
   2738 
   2739 	instr_out(ctx, 0, "URB_FENCE: %s%s%s%s%s%s\n",
   2740 		  (data[0] >> 13) & 1 ? "cs " : "",
   2741 		  (data[0] >> 12) & 1 ? "vfe " : "",
   2742 		  (data[0] >> 11) & 1 ? "sf " : "",
   2743 		  (data[0] >> 10) & 1 ? "clip " : "",
   2744 		  (data[0] >> 9) & 1 ? "gs " : "",
   2745 		  (data[0] >> 8) & 1 ? "vs " : "");
   2746 	instr_out(ctx, 1,
   2747 		  "vs fence: %d, clip_fence: %d, gs_fence: %d\n",
   2748 		  vs_fence, clip_fence, gs_fence);
   2749 	instr_out(ctx, 2,
   2750 		  "sf fence: %d, vfe_fence: %d, cs_fence: %d\n",
   2751 		  sf_fence, vfe_fence, cs_fence);
   2752 	if (gs_fence < vs_fence)
   2753 		fprintf(out, "gs fence < vs fence!\n");
   2754 	if (clip_fence < gs_fence)
   2755 		fprintf(out, "clip fence < gs fence!\n");
   2756 	if (sf_fence < clip_fence)
   2757 		fprintf(out, "sf fence < clip fence!\n");
   2758 	if (cs_fence < sf_fence)
   2759 		fprintf(out, "cs fence < sf fence!\n");
   2760 
   2761 	return len;
   2762 }
   2763 
   2764 static void
   2765 state_base_out(struct drm_intel_decode *ctx, unsigned int index,
   2766 	       const char *name)
   2767 {
   2768 	if (ctx->data[index] & 1) {
   2769 		instr_out(ctx, index,
   2770 			  "%s state base address 0x%08x\n", name,
   2771 			  ctx->data[index] & ~1);
   2772 	} else {
   2773 		instr_out(ctx, index, "%s state base not updated\n",
   2774 			  name);
   2775 	}
   2776 }
   2777 
   2778 static void
   2779 state_max_out(struct drm_intel_decode *ctx, unsigned int index,
   2780 	      const char *name)
   2781 {
   2782 	if (ctx->data[index] & 1) {
   2783 		if (ctx->data[index] == 1) {
   2784 			instr_out(ctx, index,
   2785 				  "%s state upper bound disabled\n", name);
   2786 		} else {
   2787 			instr_out(ctx, index,
   2788 				  "%s state upper bound 0x%08x\n", name,
   2789 				  ctx->data[index] & ~1);
   2790 		}
   2791 	} else {
   2792 		instr_out(ctx, index,
   2793 			  "%s state upper bound not updated\n", name);
   2794 	}
   2795 }
   2796 
   2797 static int
   2798 gen7_3DSTATE_VIEWPORT_STATE_POINTERS_CC(struct drm_intel_decode *ctx)
   2799 {
   2800 	instr_out(ctx, 0, "3DSTATE_VIEWPORT_STATE_POINTERS_CC\n");
   2801 	instr_out(ctx, 1, "pointer to CC viewport\n");
   2802 
   2803 	return 2;
   2804 }
   2805 
   2806 static int
   2807 gen7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP(struct drm_intel_decode *ctx)
   2808 {
   2809 	instr_out(ctx, 0, "3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP\n");
   2810 	instr_out(ctx, 1, "pointer to SF_CLIP viewport\n");
   2811 
   2812 	return 2;
   2813 }
   2814 
   2815 static int
   2816 gen7_3DSTATE_BLEND_STATE_POINTERS(struct drm_intel_decode *ctx)
   2817 {
   2818 	instr_out(ctx, 0, "3DSTATE_BLEND_STATE_POINTERS\n");
   2819 	instr_out(ctx, 1, "pointer to BLEND_STATE at 0x%08x (%s)\n",
   2820 		  ctx->data[1] & ~1,
   2821 		  (ctx->data[1] & 1) ? "changed" : "unchanged");
   2822 
   2823 	return 2;
   2824 }
   2825 
   2826 static int
   2827 gen7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS(struct drm_intel_decode *ctx)
   2828 {
   2829 	instr_out(ctx, 0, "3DSTATE_DEPTH_STENCIL_STATE_POINTERS\n");
   2830 	instr_out(ctx, 1,
   2831 		  "pointer to DEPTH_STENCIL_STATE at 0x%08x (%s)\n",
   2832 		  ctx->data[1] & ~1,
   2833 		  (ctx->data[1] & 1) ? "changed" : "unchanged");
   2834 
   2835 	return 2;
   2836 }
   2837 
   2838 static int
   2839 gen7_3DSTATE_HIER_DEPTH_BUFFER(struct drm_intel_decode *ctx)
   2840 {
   2841 	instr_out(ctx, 0, "3DSTATE_HIER_DEPTH_BUFFER\n");
   2842 	instr_out(ctx, 1, "pitch %db\n",
   2843 		  (ctx->data[1] & 0x1ffff) + 1);
   2844 	instr_out(ctx, 2, "pointer to HiZ buffer\n");
   2845 
   2846 	return 3;
   2847 }
   2848 
   2849 static int
   2850 gen6_3DSTATE_CC_STATE_POINTERS(struct drm_intel_decode *ctx)
   2851 {
   2852 	instr_out(ctx, 0, "3DSTATE_CC_STATE_POINTERS\n");
   2853 	instr_out(ctx, 1, "blend change %d\n", ctx->data[1] & 1);
   2854 	instr_out(ctx, 2, "depth stencil change %d\n",
   2855 		  ctx->data[2] & 1);
   2856 	instr_out(ctx, 3, "cc change %d\n", ctx->data[3] & 1);
   2857 
   2858 	return 4;
   2859 }
   2860 
   2861 static int
   2862 gen7_3DSTATE_CC_STATE_POINTERS(struct drm_intel_decode *ctx)
   2863 {
   2864 	instr_out(ctx, 0, "3DSTATE_CC_STATE_POINTERS\n");
   2865 	instr_out(ctx, 1, "pointer to COLOR_CALC_STATE at 0x%08x "
   2866 		  "(%s)\n",
   2867 		  ctx->data[1] & ~1,
   2868 		  (ctx->data[1] & 1) ? "changed" : "unchanged");
   2869 
   2870 	return 2;
   2871 }
   2872 
   2873 static int
   2874 gen7_3DSTATE_URB_unit(struct drm_intel_decode *ctx, const char *unit)
   2875 {
   2876     int start_kb = ((ctx->data[1] >> 25) & 0x3f) * 8;
   2877     /* the field is # of 512-bit rows - 1, we print bytes */
   2878     int entry_size = (((ctx->data[1] >> 16) & 0x1ff) + 1);
   2879     int nr_entries = ctx->data[1] & 0xffff;
   2880 
   2881     instr_out(ctx, 0, "3DSTATE_URB_%s\n", unit);
   2882     instr_out(ctx, 1,
   2883 	      "%dKB start, size=%d 64B rows, nr_entries=%d, total size %dB\n",
   2884 	      start_kb, entry_size, nr_entries, nr_entries * 64 * entry_size);
   2885 
   2886     return 2;
   2887 }
   2888 
   2889 static int
   2890 gen7_3DSTATE_URB_VS(struct drm_intel_decode *ctx)
   2891 {
   2892 	return gen7_3DSTATE_URB_unit(ctx, "VS");
   2893 }
   2894 
   2895 static int
   2896 gen7_3DSTATE_URB_HS(struct drm_intel_decode *ctx)
   2897 {
   2898 	return gen7_3DSTATE_URB_unit(ctx, "HS");
   2899 }
   2900 
   2901 static int
   2902 gen7_3DSTATE_URB_DS(struct drm_intel_decode *ctx)
   2903 {
   2904 	return gen7_3DSTATE_URB_unit(ctx, "DS");
   2905 }
   2906 
   2907 static int
   2908 gen7_3DSTATE_URB_GS(struct drm_intel_decode *ctx)
   2909 {
   2910 	return gen7_3DSTATE_URB_unit(ctx, "GS");
   2911 }
   2912 
   2913 static int
   2914 gen7_3DSTATE_CONSTANT(struct drm_intel_decode *ctx, const char *unit)
   2915 {
   2916 	int rlen[4];
   2917 
   2918 	rlen[0] = (ctx->data[1] >> 0) & 0xffff;
   2919 	rlen[1] = (ctx->data[1] >> 16) & 0xffff;
   2920 	rlen[2] = (ctx->data[2] >> 0) & 0xffff;
   2921 	rlen[3] = (ctx->data[2] >> 16) & 0xffff;
   2922 
   2923 	instr_out(ctx, 0, "3DSTATE_CONSTANT_%s\n", unit);
   2924 	instr_out(ctx, 1, "len 0 = %d, len 1 = %d\n", rlen[0], rlen[1]);
   2925 	instr_out(ctx, 2, "len 2 = %d, len 3 = %d\n", rlen[2], rlen[3]);
   2926 	instr_out(ctx, 3, "pointer to constbuf 0\n");
   2927 	instr_out(ctx, 4, "pointer to constbuf 1\n");
   2928 	instr_out(ctx, 5, "pointer to constbuf 2\n");
   2929 	instr_out(ctx, 6, "pointer to constbuf 3\n");
   2930 
   2931 	return 7;
   2932 }
   2933 
   2934 static int
   2935 gen7_3DSTATE_CONSTANT_VS(struct drm_intel_decode *ctx)
   2936 {
   2937 	return gen7_3DSTATE_CONSTANT(ctx, "VS");
   2938 }
   2939 
   2940 static int
   2941 gen7_3DSTATE_CONSTANT_GS(struct drm_intel_decode *ctx)
   2942 {
   2943 	return gen7_3DSTATE_CONSTANT(ctx, "GS");
   2944 }
   2945 
   2946 static int
   2947 gen7_3DSTATE_CONSTANT_PS(struct drm_intel_decode *ctx)
   2948 {
   2949 	return gen7_3DSTATE_CONSTANT(ctx, "PS");
   2950 }
   2951 
   2952 static int
   2953 gen7_3DSTATE_CONSTANT_DS(struct drm_intel_decode *ctx)
   2954 {
   2955 	return gen7_3DSTATE_CONSTANT(ctx, "DS");
   2956 }
   2957 
   2958 static int
   2959 gen7_3DSTATE_CONSTANT_HS(struct drm_intel_decode *ctx)
   2960 {
   2961 	return gen7_3DSTATE_CONSTANT(ctx, "HS");
   2962 }
   2963 
   2964 
   2965 static int
   2966 gen6_3DSTATE_WM(struct drm_intel_decode *ctx)
   2967 {
   2968 	instr_out(ctx, 0, "3DSTATE_WM\n");
   2969 	instr_out(ctx, 1, "kernel start pointer 0\n");
   2970 	instr_out(ctx, 2,
   2971 		  "SPF=%d, VME=%d, Sampler Count %d, "
   2972 		  "Binding table count %d\n",
   2973 		  (ctx->data[2] >> 31) & 1,
   2974 		  (ctx->data[2] >> 30) & 1,
   2975 		  (ctx->data[2] >> 27) & 7,
   2976 		  (ctx->data[2] >> 18) & 0xff);
   2977 	instr_out(ctx, 3, "scratch offset\n");
   2978 	instr_out(ctx, 4,
   2979 		  "Depth Clear %d, Depth Resolve %d, HiZ Resolve %d, "
   2980 		  "Dispatch GRF start[0] %d, start[1] %d, start[2] %d\n",
   2981 		  (ctx->data[4] & (1 << 30)) != 0,
   2982 		  (ctx->data[4] & (1 << 28)) != 0,
   2983 		  (ctx->data[4] & (1 << 27)) != 0,
   2984 		  (ctx->data[4] >> 16) & 0x7f,
   2985 		  (ctx->data[4] >> 8) & 0x7f,
   2986 		  (ctx->data[4] & 0x7f));
   2987 	instr_out(ctx, 5,
   2988 		  "MaxThreads %d, PS KillPixel %d, PS computed Z %d, "
   2989 		  "PS use sourceZ %d, Thread Dispatch %d, PS use sourceW %d, "
   2990 		  "Dispatch32 %d, Dispatch16 %d, Dispatch8 %d\n",
   2991 		  ((ctx->data[5] >> 25) & 0x7f) + 1,
   2992 		  (ctx->data[5] & (1 << 22)) != 0,
   2993 		  (ctx->data[5] & (1 << 21)) != 0,
   2994 		  (ctx->data[5] & (1 << 20)) != 0,
   2995 		  (ctx->data[5] & (1 << 19)) != 0,
   2996 		  (ctx->data[5] & (1 << 8)) != 0,
   2997 		  (ctx->data[5] & (1 << 2)) != 0,
   2998 		  (ctx->data[5] & (1 << 1)) != 0,
   2999 		  (ctx->data[5] & (1 << 0)) != 0);
   3000 	instr_out(ctx, 6,
   3001 		  "Num SF output %d, Pos XY offset %d, ZW interp mode %d , "
   3002 		  "Barycentric interp mode 0x%x, Point raster rule %d, "
   3003 		  "Multisample mode %d, "
   3004 		  "Multisample Dispatch mode %d\n",
   3005 		  (ctx->data[6] >> 20) & 0x3f,
   3006 		  (ctx->data[6] >> 18) & 3,
   3007 		  (ctx->data[6] >> 16) & 3,
   3008 		  (ctx->data[6] >> 10) & 0x3f,
   3009 		  (ctx->data[6] & (1 << 9)) != 0,
   3010 		  (ctx->data[6] >> 1) & 3,
   3011 		  (ctx->data[6] & 1));
   3012 	instr_out(ctx, 7, "kernel start pointer 1\n");
   3013 	instr_out(ctx, 8, "kernel start pointer 2\n");
   3014 
   3015 	return 9;
   3016 }
   3017 
   3018 static int
   3019 gen7_3DSTATE_WM(struct drm_intel_decode *ctx)
   3020 {
   3021 	const char *computed_depth = "";
   3022 	const char *early_depth = "";
   3023 	const char *zw_interp = "";
   3024 
   3025 	switch ((ctx->data[1] >> 23) & 0x3) {
   3026 	case 0:
   3027 		computed_depth = "";
   3028 		break;
   3029 	case 1:
   3030 		computed_depth = "computed depth";
   3031 		break;
   3032 	case 2:
   3033 		computed_depth = "computed depth >=";
   3034 		break;
   3035 	case 3:
   3036 		computed_depth = "computed depth <=";
   3037 		break;
   3038 	}
   3039 
   3040 	switch ((ctx->data[1] >> 21) & 0x3) {
   3041 	case 0:
   3042 		early_depth = "";
   3043 		break;
   3044 	case 1:
   3045 		early_depth = ", EDSC_PSEXEC";
   3046 		break;
   3047 	case 2:
   3048 		early_depth = ", EDSC_PREPS";
   3049 		break;
   3050 	case 3:
   3051 		early_depth = ", BAD EDSC";
   3052 		break;
   3053 	}
   3054 
   3055 	switch ((ctx->data[1] >> 17) & 0x3) {
   3056 	case 0:
   3057 		early_depth = "";
   3058 		break;
   3059 	case 1:
   3060 		early_depth = ", BAD ZW interp";
   3061 		break;
   3062 	case 2:
   3063 		early_depth = ", ZW centroid";
   3064 		break;
   3065 	case 3:
   3066 		early_depth = ", ZW sample";
   3067 		break;
   3068 	}
   3069 
   3070 	instr_out(ctx, 0, "3DSTATE_WM\n");
   3071 	instr_out(ctx, 1, "(%s%s%s%s%s%s)%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
   3072 		  (ctx->data[1] & (1 << 11)) ? "PP " : "",
   3073 		  (ctx->data[1] & (1 << 12)) ? "PC " : "",
   3074 		  (ctx->data[1] & (1 << 13)) ? "PS " : "",
   3075 		  (ctx->data[1] & (1 << 14)) ? "NPP " : "",
   3076 		  (ctx->data[1] & (1 << 15)) ? "NPC " : "",
   3077 		  (ctx->data[1] & (1 << 16)) ? "NPS " : "",
   3078 		  (ctx->data[1] & (1 << 30)) ? ", depth clear" : "",
   3079 		  (ctx->data[1] & (1 << 29)) ? "" : ", disabled",
   3080 		  (ctx->data[1] & (1 << 28)) ? ", depth resolve" : "",
   3081 		  (ctx->data[1] & (1 << 27)) ? ", hiz resolve" : "",
   3082 		  (ctx->data[1] & (1 << 25)) ? ", kill" : "",
   3083 		  computed_depth,
   3084 		  early_depth,
   3085 		  zw_interp,
   3086 		  (ctx->data[1] & (1 << 20)) ? ", source depth" : "",
   3087 		  (ctx->data[1] & (1 << 19)) ? ", source W" : "",
   3088 		  (ctx->data[1] & (1 << 10)) ? ", coverage" : "",
   3089 		  (ctx->data[1] & (1 << 4)) ? ", poly stipple" : "",
   3090 		  (ctx->data[1] & (1 << 3)) ? ", line stipple" : "",
   3091 		  (ctx->data[1] & (1 << 2)) ? ", point UL" : ", point UR"
   3092 		  );
   3093 	instr_out(ctx, 2, "MS\n");
   3094 
   3095 	return 3;
   3096 }
   3097 
   3098 static int
   3099 gen4_3DPRIMITIVE(struct drm_intel_decode *ctx)
   3100 {
   3101 	instr_out(ctx, 0,
   3102 		  "3DPRIMITIVE: %s %s\n",
   3103 		  get_965_prim_type((ctx->data[0] >> 10) & 0x1f),
   3104 		  (ctx->data[0] & (1 << 15)) ? "random" : "sequential");
   3105 	instr_out(ctx, 1, "vertex count\n");
   3106 	instr_out(ctx, 2, "start vertex\n");
   3107 	instr_out(ctx, 3, "instance count\n");
   3108 	instr_out(ctx, 4, "start instance\n");
   3109 	instr_out(ctx, 5, "index bias\n");
   3110 
   3111 	return 6;
   3112 }
   3113 
   3114 static int
   3115 gen7_3DPRIMITIVE(struct drm_intel_decode *ctx)
   3116 {
   3117 	bool indirect = !!(ctx->data[0] & (1 << 10));
   3118 
   3119 	instr_out(ctx, 0,
   3120 		  "3DPRIMITIVE: %s%s\n",
   3121 		  indirect ? " indirect" : "",
   3122 		  (ctx->data[0] & (1 << 8)) ? " predicated" : "");
   3123 	instr_out(ctx, 1, "%s %s\n",
   3124 		  get_965_prim_type(ctx->data[1] & 0x3f),
   3125 		  (ctx->data[1] & (1 << 8)) ? "random" : "sequential");
   3126 	instr_out(ctx, 2, indirect ? "ignored" : "vertex count\n");
   3127 	instr_out(ctx, 3, indirect ? "ignored" : "start vertex\n");
   3128 	instr_out(ctx, 4, indirect ? "ignored" : "instance count\n");
   3129 	instr_out(ctx, 5, indirect ? "ignored" : "start instance\n");
   3130 	instr_out(ctx, 6, indirect ? "ignored" : "index bias\n");
   3131 
   3132 	return 7;
   3133 }
   3134 
   3135 static int
   3136 decode_3d_965(struct drm_intel_decode *ctx)
   3137 {
   3138 	uint32_t opcode;
   3139 	unsigned int len;
   3140 	unsigned int i, j, sba_len;
   3141 	const char *desc1 = NULL;
   3142 	uint32_t *data = ctx->data;
   3143 	uint32_t devid = ctx->devid;
   3144 
   3145 	struct {
   3146 		uint32_t opcode;
   3147 		uint32_t len_mask;
   3148 		int unsigned min_len;
   3149 		int unsigned max_len;
   3150 		const char *name;
   3151 		int gen;
   3152 		int (*func)(struct drm_intel_decode *ctx);
   3153 	} opcodes_3d[] = {
   3154 		{ 0x6000, 0x00ff, 3, 3, "URB_FENCE" },
   3155 		{ 0x6001, 0xffff, 2, 2, "CS_URB_STATE" },
   3156 		{ 0x6002, 0x00ff, 2, 2, "CONSTANT_BUFFER" },
   3157 		{ 0x6101, 0xffff, 6, 10, "STATE_BASE_ADDRESS" },
   3158 		{ 0x6102, 0xffff, 2, 2, "STATE_SIP" },
   3159 		{ 0x6104, 0xffff, 1, 1, "3DSTATE_PIPELINE_SELECT" },
   3160 		{ 0x680b, 0xffff, 1, 1, "3DSTATE_VF_STATISTICS" },
   3161 		{ 0x6904, 0xffff, 1, 1, "3DSTATE_PIPELINE_SELECT" },
   3162 		{ 0x7800, 0xffff, 7, 7, "3DSTATE_PIPELINED_POINTERS" },
   3163 		{ 0x7801, 0x00ff, 4, 6, "3DSTATE_BINDING_TABLE_POINTERS" },
   3164 		{ 0x7802, 0x00ff, 4, 4, "3DSTATE_SAMPLER_STATE_POINTERS" },
   3165 		{ 0x7805, 0x00ff, 7, 7, "3DSTATE_DEPTH_BUFFER", 7 },
   3166 		{ 0x7805, 0x00ff, 3, 3, "3DSTATE_URB" },
   3167 		{ 0x7804, 0x00ff, 3, 3, "3DSTATE_CLEAR_PARAMS" },
   3168 		{ 0x7806, 0x00ff, 3, 3, "3DSTATE_STENCIL_BUFFER" },
   3169 		{ 0x790f, 0x00ff, 3, 3, "3DSTATE_HIER_DEPTH_BUFFER", 6 },
   3170 		{ 0x7807, 0x00ff, 3, 3, "3DSTATE_HIER_DEPTH_BUFFER", 7, gen7_3DSTATE_HIER_DEPTH_BUFFER },
   3171 		{ 0x7808, 0x00ff, 5, 257, "3DSTATE_VERTEX_BUFFERS" },
   3172 		{ 0x7809, 0x00ff, 3, 256, "3DSTATE_VERTEX_ELEMENTS" },
   3173 		{ 0x780a, 0x00ff, 3, 3, "3DSTATE_INDEX_BUFFER" },
   3174 		{ 0x780b, 0xffff, 1, 1, "3DSTATE_VF_STATISTICS" },
   3175 		{ 0x780d, 0x00ff, 4, 4, "3DSTATE_VIEWPORT_STATE_POINTERS" },
   3176 		{ 0x780e, 0xffff, 4, 4, NULL, 6, gen6_3DSTATE_CC_STATE_POINTERS },
   3177 		{ 0x780e, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_CC_STATE_POINTERS },
   3178 		{ 0x780f, 0x00ff, 2, 2, "3DSTATE_SCISSOR_POINTERS" },
   3179 		{ 0x7810, 0x00ff, 6, 6, "3DSTATE_VS" },
   3180 		{ 0x7811, 0x00ff, 7, 7, "3DSTATE_GS" },
   3181 		{ 0x7812, 0x00ff, 4, 4, "3DSTATE_CLIP" },
   3182 		{ 0x7813, 0x00ff, 20, 20, "3DSTATE_SF", 6 },
   3183 		{ 0x7813, 0x00ff, 7, 7, "3DSTATE_SF", 7 },
   3184 		{ 0x7814, 0x00ff, 3, 3, "3DSTATE_WM", 7, gen7_3DSTATE_WM },
   3185 		{ 0x7814, 0x00ff, 9, 9, "3DSTATE_WM", 6, gen6_3DSTATE_WM },
   3186 		{ 0x7815, 0x00ff, 5, 5, "3DSTATE_CONSTANT_VS_STATE", 6 },
   3187 		{ 0x7815, 0x00ff, 7, 7, "3DSTATE_CONSTANT_VS", 7, gen7_3DSTATE_CONSTANT_VS },
   3188 		{ 0x7816, 0x00ff, 5, 5, "3DSTATE_CONSTANT_GS_STATE", 6 },
   3189 		{ 0x7816, 0x00ff, 7, 7, "3DSTATE_CONSTANT_GS", 7, gen7_3DSTATE_CONSTANT_GS },
   3190 		{ 0x7817, 0x00ff, 5, 5, "3DSTATE_CONSTANT_PS_STATE", 6 },
   3191 		{ 0x7817, 0x00ff, 7, 7, "3DSTATE_CONSTANT_PS", 7, gen7_3DSTATE_CONSTANT_PS },
   3192 		{ 0x7818, 0xffff, 2, 2, "3DSTATE_SAMPLE_MASK" },
   3193 		{ 0x7819, 0x00ff, 7, 7, "3DSTATE_CONSTANT_HS", 7, gen7_3DSTATE_CONSTANT_HS },
   3194 		{ 0x781a, 0x00ff, 7, 7, "3DSTATE_CONSTANT_DS", 7, gen7_3DSTATE_CONSTANT_DS },
   3195 		{ 0x781b, 0x00ff, 7, 7, "3DSTATE_HS" },
   3196 		{ 0x781c, 0x00ff, 4, 4, "3DSTATE_TE" },
   3197 		{ 0x781d, 0x00ff, 6, 6, "3DSTATE_DS" },
   3198 		{ 0x781e, 0x00ff, 3, 3, "3DSTATE_STREAMOUT" },
   3199 		{ 0x781f, 0x00ff, 14, 14, "3DSTATE_SBE" },
   3200 		{ 0x7820, 0x00ff, 8, 8, "3DSTATE_PS" },
   3201 		{ 0x7821, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP },
   3202 		{ 0x7823, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_VIEWPORT_STATE_POINTERS_CC },
   3203 		{ 0x7824, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_BLEND_STATE_POINTERS },
   3204 		{ 0x7825, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS },
   3205 		{ 0x7826, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_VS" },
   3206 		{ 0x7827, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_HS" },
   3207 		{ 0x7828, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_DS" },
   3208 		{ 0x7829, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_GS" },
   3209 		{ 0x782a, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_PS" },
   3210 		{ 0x782b, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_VS" },
   3211 		{ 0x782c, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_HS" },
   3212 		{ 0x782d, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_DS" },
   3213 		{ 0x782e, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_GS" },
   3214 		{ 0x782f, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_PS" },
   3215 		{ 0x7830, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_VS },
   3216 		{ 0x7831, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_HS },
   3217 		{ 0x7832, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_DS },
   3218 		{ 0x7833, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_GS },
   3219 		{ 0x7900, 0xffff, 4, 4, "3DSTATE_DRAWING_RECTANGLE" },
   3220 		{ 0x7901, 0xffff, 5, 5, "3DSTATE_CONSTANT_COLOR" },
   3221 		{ 0x7905, 0xffff, 5, 7, "3DSTATE_DEPTH_BUFFER" },
   3222 		{ 0x7906, 0xffff, 2, 2, "3DSTATE_POLY_STIPPLE_OFFSET" },
   3223 		{ 0x7907, 0xffff, 33, 33, "3DSTATE_POLY_STIPPLE_PATTERN" },
   3224 		{ 0x7908, 0xffff, 3, 3, "3DSTATE_LINE_STIPPLE" },
   3225 		{ 0x7909, 0xffff, 2, 2, "3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP" },
   3226 		{ 0x7909, 0xffff, 2, 2, "3DSTATE_CLEAR_PARAMS" },
   3227 		{ 0x790a, 0xffff, 3, 3, "3DSTATE_AA_LINE_PARAMETERS" },
   3228 		{ 0x790b, 0xffff, 4, 4, "3DSTATE_GS_SVB_INDEX" },
   3229 		{ 0x790d, 0xffff, 3, 3, "3DSTATE_MULTISAMPLE", 6 },
   3230 		{ 0x790d, 0xffff, 4, 4, "3DSTATE_MULTISAMPLE", 7 },
   3231 		{ 0x7910, 0x00ff, 2, 2, "3DSTATE_CLEAR_PARAMS" },
   3232 		{ 0x7912, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_VS" },
   3233 		{ 0x7913, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_HS" },
   3234 		{ 0x7914, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_DS" },
   3235 		{ 0x7915, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_GS" },
   3236 		{ 0x7916, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_PS" },
   3237 		{ 0x7917, 0x00ff, 2, 2+128*2, "3DSTATE_SO_DECL_LIST" },
   3238 		{ 0x7918, 0x00ff, 4, 4, "3DSTATE_SO_BUFFER" },
   3239 		{ 0x7a00, 0x00ff, 4, 6, "PIPE_CONTROL" },
   3240 		{ 0x7b00, 0x00ff, 7, 7, NULL, 7, gen7_3DPRIMITIVE },
   3241 		{ 0x7b00, 0x00ff, 6, 6, NULL, 0, gen4_3DPRIMITIVE },
   3242 	}, *opcode_3d = NULL;
   3243 
   3244 	opcode = (data[0] & 0xffff0000) >> 16;
   3245 
   3246 	for (i = 0; i < ARRAY_SIZE(opcodes_3d); i++) {
   3247 		if (opcode != opcodes_3d[i].opcode)
   3248 			continue;
   3249 
   3250 		/* If it's marked as not our gen, skip. */
   3251 		if (opcodes_3d[i].gen && opcodes_3d[i].gen != ctx->gen)
   3252 			continue;
   3253 
   3254 		opcode_3d = &opcodes_3d[i];
   3255 		break;
   3256 	}
   3257 
   3258 	if (opcode_3d) {
   3259 		if (opcode_3d->max_len == 1)
   3260 			len = 1;
   3261 		else
   3262 			len = (data[0] & opcode_3d->len_mask) + 2;
   3263 
   3264 		if (len < opcode_3d->min_len ||
   3265 		    len > opcode_3d->max_len) {
   3266 			fprintf(out, "Bad length %d in %s, expected %d-%d\n",
   3267 				len, opcode_3d->name,
   3268 				opcode_3d->min_len, opcode_3d->max_len);
   3269 		}
   3270 	} else {
   3271 		len = (data[0] & 0x0000ffff) + 2;
   3272 	}
   3273 
   3274 	switch (opcode) {
   3275 	case 0x6000:
   3276 		return i965_decode_urb_fence(ctx, len);
   3277 	case 0x6001:
   3278 		instr_out(ctx, 0, "CS_URB_STATE\n");
   3279 		instr_out(ctx, 1,
   3280 			  "entry_size: %d [%d bytes], n_entries: %d\n",
   3281 			  (data[1] >> 4) & 0x1f,
   3282 			  (((data[1] >> 4) & 0x1f) + 1) * 64, data[1] & 0x7);
   3283 		return len;
   3284 	case 0x6002:
   3285 		instr_out(ctx, 0, "CONSTANT_BUFFER: %s\n",
   3286 			  (data[0] >> 8) & 1 ? "valid" : "invalid");
   3287 		instr_out(ctx, 1,
   3288 			  "offset: 0x%08x, length: %d bytes\n", data[1] & ~0x3f,
   3289 			  ((data[1] & 0x3f) + 1) * 64);
   3290 		return len;
   3291 	case 0x6101:
   3292 		i = 0;
   3293 		instr_out(ctx, 0, "STATE_BASE_ADDRESS\n");
   3294 		i++;
   3295 
   3296 		if (IS_GEN6(devid) || IS_GEN7(devid))
   3297 			sba_len = 10;
   3298 		else if (IS_GEN5(devid))
   3299 			sba_len = 8;
   3300 		else
   3301 			sba_len = 6;
   3302 		if (len != sba_len)
   3303 			fprintf(out, "Bad count in STATE_BASE_ADDRESS\n");
   3304 
   3305 		state_base_out(ctx, i++, "general");
   3306 		state_base_out(ctx, i++, "surface");
   3307 		if (IS_GEN6(devid) || IS_GEN7(devid))
   3308 			state_base_out(ctx, i++, "dynamic");
   3309 		state_base_out(ctx, i++, "indirect");
   3310 		if (IS_GEN5(devid) || IS_GEN6(devid) || IS_GEN7(devid))
   3311 			state_base_out(ctx, i++, "instruction");
   3312 
   3313 		state_max_out(ctx, i++, "general");
   3314 		if (IS_GEN6(devid) || IS_GEN7(devid))
   3315 			state_max_out(ctx, i++, "dynamic");
   3316 		state_max_out(ctx, i++, "indirect");
   3317 		if (IS_GEN5(devid) || IS_GEN6(devid) || IS_GEN7(devid))
   3318 			state_max_out(ctx, i++, "instruction");
   3319 
   3320 		return len;
   3321 	case 0x7800:
   3322 		instr_out(ctx, 0, "3DSTATE_PIPELINED_POINTERS\n");
   3323 		instr_out(ctx, 1, "VS state\n");
   3324 		instr_out(ctx, 2, "GS state\n");
   3325 		instr_out(ctx, 3, "Clip state\n");
   3326 		instr_out(ctx, 4, "SF state\n");
   3327 		instr_out(ctx, 5, "WM state\n");
   3328 		instr_out(ctx, 6, "CC state\n");
   3329 		return len;
   3330 	case 0x7801:
   3331 		if (len != 6 && len != 4)
   3332 			fprintf(out,
   3333 				"Bad count in 3DSTATE_BINDING_TABLE_POINTERS\n");
   3334 		if (len == 6) {
   3335 			instr_out(ctx, 0,
   3336 				  "3DSTATE_BINDING_TABLE_POINTERS\n");
   3337 			instr_out(ctx, 1, "VS binding table\n");
   3338 			instr_out(ctx, 2, "GS binding table\n");
   3339 			instr_out(ctx, 3, "Clip binding table\n");
   3340 			instr_out(ctx, 4, "SF binding table\n");
   3341 			instr_out(ctx, 5, "WM binding table\n");
   3342 		} else {
   3343 			instr_out(ctx, 0,
   3344 				  "3DSTATE_BINDING_TABLE_POINTERS: VS mod %d, "
   3345 				  "GS mod %d, PS mod %d\n",
   3346 				  (data[0] & (1 << 8)) != 0,
   3347 				  (data[0] & (1 << 9)) != 0,
   3348 				  (data[0] & (1 << 12)) != 0);
   3349 			instr_out(ctx, 1, "VS binding table\n");
   3350 			instr_out(ctx, 2, "GS binding table\n");
   3351 			instr_out(ctx, 3, "WM binding table\n");
   3352 		}
   3353 
   3354 		return len;
   3355 	case 0x7802:
   3356 		instr_out(ctx, 0,
   3357 			  "3DSTATE_SAMPLER_STATE_POINTERS: VS mod %d, "
   3358 			  "GS mod %d, PS mod %d\n", (data[0] & (1 << 8)) != 0,
   3359 			  (data[0] & (1 << 9)) != 0,
   3360 			  (data[0] & (1 << 12)) != 0);
   3361 		instr_out(ctx, 1, "VS sampler state\n");
   3362 		instr_out(ctx, 2, "GS sampler state\n");
   3363 		instr_out(ctx, 3, "WM sampler state\n");
   3364 		return len;
   3365 	case 0x7805:
   3366 		/* Actually 3DSTATE_DEPTH_BUFFER on gen7. */
   3367 		if (ctx->gen == 7)
   3368 			break;
   3369 
   3370 		instr_out(ctx, 0, "3DSTATE_URB\n");
   3371 		instr_out(ctx, 1,
   3372 			  "VS entries %d, alloc size %d (1024bit row)\n",
   3373 			  data[1] & 0xffff, ((data[1] >> 16) & 0x07f) + 1);
   3374 		instr_out(ctx, 2,
   3375 			  "GS entries %d, alloc size %d (1024bit row)\n",
   3376 			  (data[2] >> 8) & 0x3ff, (data[2] & 7) + 1);
   3377 		return len;
   3378 
   3379 	case 0x7808:
   3380 		if ((len - 1) % 4 != 0)
   3381 			fprintf(out, "Bad count in 3DSTATE_VERTEX_BUFFERS\n");
   3382 		instr_out(ctx, 0, "3DSTATE_VERTEX_BUFFERS\n");
   3383 
   3384 		for (i = 1; i < len;) {
   3385 			int idx, access;
   3386 			if (IS_GEN6(devid)) {
   3387 				idx = 26;
   3388 				access = 20;
   3389 			} else {
   3390 				idx = 27;
   3391 				access = 26;
   3392 			}
   3393 			instr_out(ctx, i,
   3394 				  "buffer %d: %s, pitch %db\n", data[i] >> idx,
   3395 				  data[i] & (1 << access) ? "random" :
   3396 				  "sequential", data[i] & 0x07ff);
   3397 			i++;
   3398 			instr_out(ctx, i++, "buffer address\n");
   3399 			instr_out(ctx, i++, "max index\n");
   3400 			instr_out(ctx, i++, "mbz\n");
   3401 		}
   3402 		return len;
   3403 
   3404 	case 0x7809:
   3405 		if ((len + 1) % 2 != 0)
   3406 			fprintf(out, "Bad count in 3DSTATE_VERTEX_ELEMENTS\n");
   3407 		instr_out(ctx, 0, "3DSTATE_VERTEX_ELEMENTS\n");
   3408 
   3409 		for (i = 1; i < len;) {
   3410 			instr_out(ctx, i,
   3411 				  "buffer %d: %svalid, type 0x%04x, "
   3412 				  "src offset 0x%04x bytes\n",
   3413 				  data[i] >> ((IS_GEN6(devid) || IS_GEN7(devid)) ? 26 : 27),
   3414 				  data[i] & (1 << ((IS_GEN6(devid) || IS_GEN7(devid)) ? 25 : 26)) ?
   3415 				  "" : "in", (data[i] >> 16) & 0x1ff,
   3416 				  data[i] & 0x07ff);
   3417 			i++;
   3418 			instr_out(ctx, i, "(%s, %s, %s, %s), "
   3419 				  "dst offset 0x%02x bytes\n",
   3420 				  get_965_element_component(data[i], 0),
   3421 				  get_965_element_component(data[i], 1),
   3422 				  get_965_element_component(data[i], 2),
   3423 				  get_965_element_component(data[i], 3),
   3424 				  (data[i] & 0xff) * 4);
   3425 			i++;
   3426 		}
   3427 		return len;
   3428 
   3429 	case 0x780d:
   3430 		instr_out(ctx, 0,
   3431 			  "3DSTATE_VIEWPORT_STATE_POINTERS\n");
   3432 		instr_out(ctx, 1, "clip\n");
   3433 		instr_out(ctx, 2, "sf\n");
   3434 		instr_out(ctx, 3, "cc\n");
   3435 		return len;
   3436 
   3437 	case 0x780a:
   3438 		instr_out(ctx, 0, "3DSTATE_INDEX_BUFFER\n");
   3439 		instr_out(ctx, 1, "beginning buffer address\n");
   3440 		instr_out(ctx, 2, "ending buffer address\n");
   3441 		return len;
   3442 
   3443 	case 0x780f:
   3444 		instr_out(ctx, 0, "3DSTATE_SCISSOR_POINTERS\n");
   3445 		instr_out(ctx, 1, "scissor rect offset\n");
   3446 		return len;
   3447 
   3448 	case 0x7810:
   3449 		instr_out(ctx, 0, "3DSTATE_VS\n");
   3450 		instr_out(ctx, 1, "kernel pointer\n");
   3451 		instr_out(ctx, 2,
   3452 			  "SPF=%d, VME=%d, Sampler Count %d, "
   3453 			  "Binding table count %d\n", (data[2] >> 31) & 1,
   3454 			  (data[2] >> 30) & 1, (data[2] >> 27) & 7,
   3455 			  (data[2] >> 18) & 0xff);
   3456 		instr_out(ctx, 3, "scratch offset\n");
   3457 		instr_out(ctx, 4,
   3458 			  "Dispatch GRF start %d, VUE read length %d, "
   3459 			  "VUE read offset %d\n", (data[4] >> 20) & 0x1f,
   3460 			  (data[4] >> 11) & 0x3f, (data[4] >> 4) & 0x3f);
   3461 		instr_out(ctx, 5,
   3462 			  "Max Threads %d, Vertex Cache %sable, "
   3463 			  "VS func %sable\n", ((data[5] >> 25) & 0x7f) + 1,
   3464 			  (data[5] & (1 << 1)) != 0 ? "dis" : "en",
   3465 			  (data[5] & 1) != 0 ? "en" : "dis");
   3466 		return len;
   3467 
   3468 	case 0x7811:
   3469 		instr_out(ctx, 0, "3DSTATE_GS\n");
   3470 		instr_out(ctx, 1, "kernel pointer\n");
   3471 		instr_out(ctx, 2,
   3472 			  "SPF=%d, VME=%d, Sampler Count %d, "
   3473 			  "Binding table count %d\n", (data[2] >> 31) & 1,
   3474 			  (data[2] >> 30) & 1, (data[2] >> 27) & 7,
   3475 			  (data[2] >> 18) & 0xff);
   3476 		instr_out(ctx, 3, "scratch offset\n");
   3477 		instr_out(ctx, 4,
   3478 			  "Dispatch GRF start %d, VUE read length %d, "
   3479 			  "VUE read offset %d\n", (data[4] & 0xf),
   3480 			  (data[4] >> 11) & 0x3f, (data[4] >> 4) & 0x3f);
   3481 		instr_out(ctx, 5,
   3482 			  "Max Threads %d, Rendering %sable\n",
   3483 			  ((data[5] >> 25) & 0x7f) + 1,
   3484 			  (data[5] & (1 << 8)) != 0 ? "en" : "dis");
   3485 		instr_out(ctx, 6,
   3486 			  "Reorder %sable, Discard Adjaceny %sable, "
   3487 			  "GS %sable\n",
   3488 			  (data[6] & (1 << 30)) != 0 ? "en" : "dis",
   3489 			  (data[6] & (1 << 29)) != 0 ? "en" : "dis",
   3490 			  (data[6] & (1 << 15)) != 0 ? "en" : "dis");
   3491 		return len;
   3492 
   3493 	case 0x7812:
   3494 		instr_out(ctx, 0, "3DSTATE_CLIP\n");
   3495 		instr_out(ctx, 1,
   3496 			  "UserClip distance cull test mask 0x%x\n",
   3497 			  data[1] & 0xff);
   3498 		instr_out(ctx, 2,
   3499 			  "Clip %sable, API mode %s, Viewport XY test %sable, "
   3500 			  "Viewport Z test %sable, Guardband test %sable, Clip mode %d, "
   3501 			  "Perspective Divide %sable, Non-Perspective Barycentric %sable, "
   3502 			  "Tri Provoking %d, Line Provoking %d, Trifan Provoking %d\n",
   3503 			  (data[2] & (1 << 31)) != 0 ? "en" : "dis",
   3504 			  (data[2] & (1 << 30)) != 0 ? "D3D" : "OGL",
   3505 			  (data[2] & (1 << 28)) != 0 ? "en" : "dis",
   3506 			  (data[2] & (1 << 27)) != 0 ? "en" : "dis",
   3507 			  (data[2] & (1 << 26)) != 0 ? "en" : "dis",
   3508 			  (data[2] >> 13) & 7,
   3509 			  (data[2] & (1 << 9)) != 0 ? "dis" : "en",
   3510 			  (data[2] & (1 << 8)) != 0 ? "en" : "dis",
   3511 			  (data[2] >> 4) & 3, (data[2] >> 2) & 3,
   3512 			  (data[2] & 3));
   3513 		instr_out(ctx, 3,
   3514 			  "Min PointWidth %d, Max PointWidth %d, "
   3515 			  "Force Zero RTAIndex %sable, Max VPIndex %d\n",
   3516 			  (data[3] >> 17) & 0x7ff, (data[3] >> 6) & 0x7ff,
   3517 			  (data[3] & (1 << 5)) != 0 ? "en" : "dis",
   3518 			  (data[3] & 0xf));
   3519 		return len;
   3520 
   3521 	case 0x7813:
   3522 		if (ctx->gen == 7)
   3523 			break;
   3524 
   3525 		instr_out(ctx, 0, "3DSTATE_SF\n");
   3526 		instr_out(ctx, 1,
   3527 			  "Attrib Out %d, Attrib Swizzle %sable, VUE read length %d, "
   3528 			  "VUE read offset %d\n", (data[1] >> 22) & 0x3f,
   3529 			  (data[1] & (1 << 21)) != 0 ? "en" : "dis",
   3530 			  (data[1] >> 11) & 0x1f, (data[1] >> 4) & 0x3f);
   3531 		instr_out(ctx, 2,
   3532 			  "Legacy Global DepthBias %sable, FrontFace fill %d, BF fill %d, "
   3533 			  "VP transform %sable, FrontWinding_%s\n",
   3534 			  (data[2] & (1 << 11)) != 0 ? "en" : "dis",
   3535 			  (data[2] >> 5) & 3, (data[2] >> 3) & 3,
   3536 			  (data[2] & (1 << 1)) != 0 ? "en" : "dis",
   3537 			  (data[2] & 1) != 0 ? "CCW" : "CW");
   3538 		instr_out(ctx, 3,
   3539 			  "AA %sable, CullMode %d, Scissor %sable, Multisample m ode %d\n",
   3540 			  (data[3] & (1 << 31)) != 0 ? "en" : "dis",
   3541 			  (data[3] >> 29) & 3,
   3542 			  (data[3] & (1 << 11)) != 0 ? "en" : "dis",
   3543 			  (data[3] >> 8) & 3);
   3544 		instr_out(ctx, 4,
   3545 			  "Last Pixel %sable, SubPixel Precision %d, Use PixelWidth %d\n",
   3546 			  (data[4] & (1 << 31)) != 0 ? "en" : "dis",
   3547 			  (data[4] & (1 << 12)) != 0 ? 4 : 8,
   3548 			  (data[4] & (1 << 11)) != 0);
   3549 		instr_out(ctx, 5,
   3550 			  "Global Depth Offset Constant %f\n",
   3551 			  *(float *)(&data[5]));
   3552 		instr_out(ctx, 6, "Global Depth Offset Scale %f\n",
   3553 			  *(float *)(&data[6]));
   3554 		instr_out(ctx, 7, "Global Depth Offset Clamp %f\n",
   3555 			  *(float *)(&data[7]));
   3556 
   3557 		for (i = 0, j = 0; i < 8; i++, j += 2)
   3558 			instr_out(ctx, i + 8,
   3559 				  "Attrib %d (Override %s%s%s%s, Const Source %d, Swizzle Select %d, "
   3560 				  "Source %d); Attrib %d (Override %s%s%s%s, Const Source %d, Swizzle Select %d, Source %d)\n",
   3561 				  j + 1,
   3562 				  (data[8 + i] & (1 << 31)) != 0 ? "W" : "",
   3563 				  (data[8 + i] & (1 << 30)) != 0 ? "Z" : "",
   3564 				  (data[8 + i] & (1 << 29)) != 0 ? "Y" : "",
   3565 				  (data[8 + i] & (1 << 28)) != 0 ? "X" : "",
   3566 				  (data[8 + i] >> 25) & 3,
   3567 				  (data[8 + i] >> 22) & 3,
   3568 				  (data[8 + i] >> 16) & 0x1f, j,
   3569 				  (data[8 + i] & (1 << 15)) != 0 ? "W" : "",
   3570 				  (data[8 + i] & (1 << 14)) != 0 ? "Z" : "",
   3571 				  (data[8 + i] & (1 << 13)) != 0 ? "Y" : "",
   3572 				  (data[8 + i] & (1 << 12)) != 0 ? "X" : "",
   3573 				  (data[8 + i] >> 9) & 3,
   3574 				  (data[8 + i] >> 6) & 3, (data[8 + i] & 0x1f));
   3575 		instr_out(ctx, 16,
   3576 			  "Point Sprite TexCoord Enable\n");
   3577 		instr_out(ctx, 17, "Const Interp Enable\n");
   3578 		instr_out(ctx, 18,
   3579 			  "Attrib 7-0 WrapShortest Enable\n");
   3580 		instr_out(ctx, 19,
   3581 			  "Attrib 15-8 WrapShortest Enable\n");
   3582 
   3583 		return len;
   3584 
   3585 	case 0x7900:
   3586 		instr_out(ctx, 0, "3DSTATE_DRAWING_RECTANGLE\n");
   3587 		instr_out(ctx, 1, "top left: %d,%d\n",
   3588 			  data[1] & 0xffff, (data[1] >> 16) & 0xffff);
   3589 		instr_out(ctx, 2, "bottom right: %d,%d\n",
   3590 			  data[2] & 0xffff, (data[2] >> 16) & 0xffff);
   3591 		instr_out(ctx, 3, "origin: %d,%d\n",
   3592 			  (int)data[3] & 0xffff, ((int)data[3] >> 16) & 0xffff);
   3593 
   3594 		return len;
   3595 
   3596 	case 0x7905:
   3597 		instr_out(ctx, 0, "3DSTATE_DEPTH_BUFFER\n");
   3598 		if (IS_GEN5(devid) || IS_GEN6(devid))
   3599 			instr_out(ctx, 1,
   3600 				  "%s, %s, pitch = %d bytes, %stiled, HiZ %d, Seperate Stencil %d\n",
   3601 				  get_965_surfacetype(data[1] >> 29),
   3602 				  get_965_depthformat((data[1] >> 18) & 0x7),
   3603 				  (data[1] & 0x0001ffff) + 1,
   3604 				  data[1] & (1 << 27) ? "" : "not ",
   3605 				  (data[1] & (1 << 22)) != 0,
   3606 				  (data[1] & (1 << 21)) != 0);
   3607 		else
   3608 			instr_out(ctx, 1,
   3609 				  "%s, %s, pitch = %d bytes, %stiled\n",
   3610 				  get_965_surfacetype(data[1] >> 29),
   3611 				  get_965_depthformat((data[1] >> 18) & 0x7),
   3612 				  (data[1] & 0x0001ffff) + 1,
   3613 				  data[1] & (1 << 27) ? "" : "not ");
   3614 		instr_out(ctx, 2, "depth offset\n");
   3615 		instr_out(ctx, 3, "%dx%d\n",
   3616 			  ((data[3] & 0x0007ffc0) >> 6) + 1,
   3617 			  ((data[3] & 0xfff80000) >> 19) + 1);
   3618 		instr_out(ctx, 4, "volume depth\n");
   3619 		if (len >= 6)
   3620 			instr_out(ctx, 5, "\n");
   3621 		if (len >= 7) {
   3622 			if (IS_GEN6(devid))
   3623 				instr_out(ctx, 6, "\n");
   3624 			else
   3625 				instr_out(ctx, 6,
   3626 					  "render target view extent\n");
   3627 		}
   3628 
   3629 		return len;
   3630 
   3631 	case 0x7a00:
   3632 		if (IS_GEN6(devid) || IS_GEN7(devid)) {
   3633 			unsigned int i;
   3634 			if (len != 4 && len != 5)
   3635 				fprintf(out, "Bad count in PIPE_CONTROL\n");
   3636 
   3637 			switch ((data[1] >> 14) & 0x3) {
   3638 			case 0:
   3639 				desc1 = "no write";
   3640 				break;
   3641 			case 1:
   3642 				desc1 = "qword write";
   3643 				break;
   3644 			case 2:
   3645 				desc1 = "PS_DEPTH_COUNT write";
   3646 				break;
   3647 			case 3:
   3648 				desc1 = "TIMESTAMP write";
   3649 				break;
   3650 			}
   3651 			instr_out(ctx, 0, "PIPE_CONTROL\n");
   3652 			instr_out(ctx, 1,
   3653 				  "%s, %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
   3654 				  desc1,
   3655 				  data[1] & (1 << 20) ? "cs stall, " : "",
   3656 				  data[1] & (1 << 19) ?
   3657 				  "global snapshot count reset, " : "",
   3658 				  data[1] & (1 << 18) ? "tlb invalidate, " : "",
   3659 				  data[1] & (1 << 17) ? "gfdt flush, " : "",
   3660 				  data[1] & (1 << 17) ? "media state clear, " :
   3661 				  "",
   3662 				  data[1] & (1 << 13) ? "depth stall, " : "",
   3663 				  data[1] & (1 << 12) ?
   3664 				  "render target cache flush, " : "",
   3665 				  data[1] & (1 << 11) ?
   3666 				  "instruction cache invalidate, " : "",
   3667 				  data[1] & (1 << 10) ?
   3668 				  "texture cache invalidate, " : "",
   3669 				  data[1] & (1 << 9) ?
   3670 				  "indirect state invalidate, " : "",
   3671 				  data[1] & (1 << 8) ? "notify irq, " : "",
   3672 				  data[1] & (1 << 7) ? "PIPE_CONTROL flush, " :
   3673 				  "",
   3674 				  data[1] & (1 << 6) ? "protect mem app_id, " :
   3675 				  "", data[1] & (1 << 5) ? "DC flush, " : "",
   3676 				  data[1] & (1 << 4) ? "vf fetch invalidate, " :
   3677 				  "",
   3678 				  data[1] & (1 << 3) ?
   3679 				  "constant cache invalidate, " : "",
   3680 				  data[1] & (1 << 2) ?
   3681 				  "state cache invalidate, " : "",
   3682 				  data[1] & (1 << 1) ? "stall at scoreboard, " :
   3683 				  "",
   3684 				  data[1] & (1 << 0) ? "depth cache flush, " :
   3685 				  "");
   3686 			if (len == 5) {
   3687 				instr_out(ctx, 2,
   3688 					  "destination address\n");
   3689 				instr_out(ctx, 3,
   3690 					  "immediate dword low\n");
   3691 				instr_out(ctx, 4,
   3692 					  "immediate dword high\n");
   3693 			} else {
   3694 				for (i = 2; i < len; i++) {
   3695 					instr_out(ctx, i, "\n");
   3696 				}
   3697 			}
   3698 			return len;
   3699 		} else {
   3700 			if (len != 4)
   3701 				fprintf(out, "Bad count in PIPE_CONTROL\n");
   3702 
   3703 			switch ((data[0] >> 14) & 0x3) {
   3704 			case 0:
   3705 				desc1 = "no write";
   3706 				break;
   3707 			case 1:
   3708 				desc1 = "qword write";
   3709 				break;
   3710 			case 2:
   3711 				desc1 = "PS_DEPTH_COUNT write";
   3712 				break;
   3713 			case 3:
   3714 				desc1 = "TIMESTAMP write";
   3715 				break;
   3716 			}
   3717 			instr_out(ctx, 0,
   3718 				  "PIPE_CONTROL: %s, %sdepth stall, %sRC write flush, "
   3719 				  "%sinst flush\n",
   3720 				  desc1,
   3721 				  data[0] & (1 << 13) ? "" : "no ",
   3722 				  data[0] & (1 << 12) ? "" : "no ",
   3723 				  data[0] & (1 << 11) ? "" : "no ");
   3724 			instr_out(ctx, 1, "destination address\n");
   3725 			instr_out(ctx, 2, "immediate dword low\n");
   3726 			instr_out(ctx, 3, "immediate dword high\n");
   3727 			return len;
   3728 		}
   3729 	}
   3730 
   3731 	if (opcode_3d) {
   3732 		if (opcode_3d->func) {
   3733 			return opcode_3d->func(ctx);
   3734 		} else {
   3735 			unsigned int i;
   3736 
   3737 			instr_out(ctx, 0, "%s\n", opcode_3d->name);
   3738 
   3739 			for (i = 1; i < len; i++) {
   3740 				instr_out(ctx, i, "dword %d\n", i);
   3741 			}
   3742 			return len;
   3743 		}
   3744 	}
   3745 
   3746 	instr_out(ctx, 0, "3D UNKNOWN: 3d_965 opcode = 0x%x\n",
   3747 		  opcode);
   3748 	return 1;
   3749 }
   3750 
   3751 static int
   3752 decode_3d_i830(struct drm_intel_decode *ctx)
   3753 {
   3754 	unsigned int idx;
   3755 	uint32_t opcode;
   3756 	uint32_t *data = ctx->data;
   3757 
   3758 	struct {
   3759 		uint32_t opcode;
   3760 		unsigned int min_len;
   3761 		unsigned int max_len;
   3762 		const char *name;
   3763 	} opcodes_3d[] = {
   3764 		{ 0x02, 1, 1, "3DSTATE_MODES_3" },
   3765 		{ 0x03, 1, 1, "3DSTATE_ENABLES_1" },
   3766 		{ 0x04, 1, 1, "3DSTATE_ENABLES_2" },
   3767 		{ 0x05, 1, 1, "3DSTATE_VFT0" },
   3768 		{ 0x06, 1, 1, "3DSTATE_AA" },
   3769 		{ 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES" },
   3770 		{ 0x08, 1, 1, "3DSTATE_MODES_1" },
   3771 		{ 0x09, 1, 1, "3DSTATE_STENCIL_TEST" },
   3772 		{ 0x0a, 1, 1, "3DSTATE_VFT1" },
   3773 		{ 0x0b, 1, 1, "3DSTATE_INDPT_ALPHA_BLEND" },
   3774 		{ 0x0c, 1, 1, "3DSTATE_MODES_5" },
   3775 		{ 0x0d, 1, 1, "3DSTATE_MAP_BLEND_OP" },
   3776 		{ 0x0e, 1, 1, "3DSTATE_MAP_BLEND_ARG" },
   3777 		{ 0x0f, 1, 1, "3DSTATE_MODES_2" },
   3778 		{ 0x15, 1, 1, "3DSTATE_FOG_COLOR" },
   3779 		{ 0x16, 1, 1, "3DSTATE_MODES_4"},
   3780 	}, *opcode_3d;
   3781 
   3782 	opcode = (data[0] & 0x1f000000) >> 24;
   3783 
   3784 	switch (opcode) {
   3785 	case 0x1f:
   3786 		return decode_3d_primitive(ctx);
   3787 	case 0x1d:
   3788 		return decode_3d_1d(ctx);
   3789 	case 0x1c:
   3790 		return decode_3d_1c(ctx);
   3791 	}
   3792 
   3793 	for (idx = 0; idx < ARRAY_SIZE(opcodes_3d); idx++) {
   3794 		opcode_3d = &opcodes_3d[idx];
   3795 		if ((data[0] & 0x1f000000) >> 24 == opcode_3d->opcode) {
   3796 			unsigned int len = 1, i;
   3797 
   3798 			instr_out(ctx, 0, "%s\n", opcode_3d->name);
   3799 			if (opcode_3d->max_len > 1) {
   3800 				len = (data[0] & 0xff) + 2;
   3801 				if (len < opcode_3d->min_len ||
   3802 				    len > opcode_3d->max_len) {
   3803 					fprintf(out, "Bad count in %s\n",
   3804 						opcode_3d->name);
   3805 				}
   3806 			}
   3807 
   3808 			for (i = 1; i < len; i++) {
   3809 				instr_out(ctx, i, "dword %d\n", i);
   3810 			}
   3811 			return len;
   3812 		}
   3813 	}
   3814 
   3815 	instr_out(ctx, 0, "3D UNKNOWN: 3d_i830 opcode = 0x%x\n",
   3816 		  opcode);
   3817 	return 1;
   3818 }
   3819 
   3820 drm_public struct drm_intel_decode *
   3821 drm_intel_decode_context_alloc(uint32_t devid)
   3822 {
   3823 	struct drm_intel_decode *ctx;
   3824 
   3825 	ctx = calloc(1, sizeof(struct drm_intel_decode));
   3826 	if (!ctx)
   3827 		return NULL;
   3828 
   3829 	ctx->devid = devid;
   3830 	ctx->out = stdout;
   3831 
   3832 	if (IS_GEN9(devid))
   3833 		ctx->gen = 9;
   3834 	else if (IS_GEN8(devid))
   3835 		ctx->gen = 8;
   3836 	else if (IS_GEN7(devid))
   3837 		ctx->gen = 7;
   3838 	else if (IS_GEN6(devid))
   3839 		ctx->gen = 6;
   3840 	else if (IS_GEN5(devid))
   3841 		ctx->gen = 5;
   3842 	else if (IS_GEN4(devid))
   3843 		ctx->gen = 4;
   3844 	else if (IS_9XX(devid))
   3845 		ctx->gen = 3;
   3846 	else {
   3847 		assert(IS_GEN2(devid));
   3848 		ctx->gen = 2;
   3849 	}
   3850 
   3851 	return ctx;
   3852 }
   3853 
   3854 drm_public void
   3855 drm_intel_decode_context_free(struct drm_intel_decode *ctx)
   3856 {
   3857 	free(ctx);
   3858 }
   3859 
   3860 drm_public void
   3861 drm_intel_decode_set_dump_past_end(struct drm_intel_decode *ctx,
   3862 				   int dump_past_end)
   3863 {
   3864 	ctx->dump_past_end = !!dump_past_end;
   3865 }
   3866 
   3867 drm_public void
   3868 drm_intel_decode_set_batch_pointer(struct drm_intel_decode *ctx,
   3869 				   void *data, uint32_t hw_offset, int count)
   3870 {
   3871 	ctx->base_data = data;
   3872 	ctx->base_hw_offset = hw_offset;
   3873 	ctx->base_count = count;
   3874 }
   3875 
   3876 drm_public void
   3877 drm_intel_decode_set_head_tail(struct drm_intel_decode *ctx,
   3878 			       uint32_t head, uint32_t tail)
   3879 {
   3880 	ctx->head = head;
   3881 	ctx->tail = tail;
   3882 }
   3883 
   3884 drm_public void
   3885 drm_intel_decode_set_output_file(struct drm_intel_decode *ctx,
   3886 				 FILE *out)
   3887 {
   3888 	ctx->out = out;
   3889 }
   3890 
   3891 /**
   3892  * Decodes an i830-i915 batch buffer, writing the output to stdout.
   3893  *
   3894  * \param data batch buffer contents
   3895  * \param count number of DWORDs to decode in the batch buffer
   3896  * \param hw_offset hardware address for the buffer
   3897  */
   3898 drm_public void
   3899 drm_intel_decode(struct drm_intel_decode *ctx)
   3900 {
   3901 	int ret;
   3902 	unsigned int index = 0;
   3903 	uint32_t devid;
   3904 	int size = ctx->base_count * 4;
   3905 	void *temp;
   3906 
   3907 	if (!ctx)
   3908 		return;
   3909 
   3910 	/* Put a scratch page full of obviously undefined data after
   3911 	 * the batchbuffer.  This lets us avoid a bunch of length
   3912 	 * checking in statically sized packets.
   3913 	 */
   3914 	temp = malloc(size + 4096);
   3915 	memcpy(temp, ctx->base_data, size);
   3916 	memset((char *)temp + size, 0xd0, 4096);
   3917 	ctx->data = temp;
   3918 
   3919 	ctx->hw_offset = ctx->base_hw_offset;
   3920 	ctx->count = ctx->base_count;
   3921 
   3922 	devid = ctx->devid;
   3923 	head_offset = ctx->head;
   3924 	tail_offset = ctx->tail;
   3925 	out = ctx->out;
   3926 
   3927 	saved_s2_set = 0;
   3928 	saved_s4_set = 1;
   3929 
   3930 	while (ctx->count > 0) {
   3931 		index = 0;
   3932 
   3933 		switch ((ctx->data[index] & 0xe0000000) >> 29) {
   3934 		case 0x0:
   3935 			ret = decode_mi(ctx);
   3936 
   3937 			/* If MI_BATCHBUFFER_END happened, then dump
   3938 			 * the rest of the output in case we some day
   3939 			 * want it in debugging, but don't decode it
   3940 			 * since it'll just confuse in the common
   3941 			 * case.
   3942 			 */
   3943 			if (ret == -1) {
   3944 				if (ctx->dump_past_end) {
   3945 					index++;
   3946 				} else {
   3947 					for (index = index + 1; index < ctx->count;
   3948 					     index++) {
   3949 						instr_out(ctx, index, "\n");
   3950 					}
   3951 				}
   3952 			} else
   3953 				index += ret;
   3954 			break;
   3955 		case 0x2:
   3956 			index += decode_2d(ctx);
   3957 			break;
   3958 		case 0x3:
   3959 			if (IS_9XX(devid) && !IS_GEN3(devid)) {
   3960 				index +=
   3961 				    decode_3d_965(ctx);
   3962 			} else if (IS_GEN3(devid)) {
   3963 				index += decode_3d(ctx);
   3964 			} else {
   3965 				index +=
   3966 				    decode_3d_i830(ctx);
   3967 			}
   3968 			break;
   3969 		default:
   3970 			instr_out(ctx, index, "UNKNOWN\n");
   3971 			index++;
   3972 			break;
   3973 		}
   3974 		fflush(out);
   3975 
   3976 		if (ctx->count < index)
   3977 			break;
   3978 
   3979 		ctx->count -= index;
   3980 		ctx->data += index;
   3981 		ctx->hw_offset += 4 * index;
   3982 	}
   3983 
   3984 	free(temp);
   3985 }
   3986