U3memcpy.S 9.13 KB
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422
/* U3memcpy.S: UltraSparc-III optimized memcpy.
 *
 * Copyright (C) 1999, 2000, 2004 David S. Miller (davem@redhat.com)
 */

#ifdef __KERNEL__
#include <asm/visasm.h>
#include <asm/asi.h>
#define GLOBAL_SPARE	%g7
#else
#define ASI_BLK_P 0xf0
#define FPRS_FEF  0x04
#ifdef MEMCPY_DEBUG
#define VISEntryHalf rd %fprs, %o5; wr %g0, FPRS_FEF, %fprs; \
		     clr %g1; clr %g2; clr %g3; subcc %g0, %g0, %g0;
#define VISExitHalf and %o5, FPRS_FEF, %o5; wr %o5, 0x0, %fprs
#else
#define VISEntryHalf rd %fprs, %o5; wr %g0, FPRS_FEF, %fprs
#define VISExitHalf and %o5, FPRS_FEF, %o5; wr %o5, 0x0, %fprs
#endif
#define GLOBAL_SPARE	%g5
#endif

#ifndef EX_LD
#define EX_LD(x)	x
#endif

#ifndef EX_ST
#define EX_ST(x)	x
#endif

#ifndef EX_RETVAL
#define EX_RETVAL(x)	x
#endif

#ifndef LOAD
#define LOAD(type,addr,dest)	type [addr], dest
#endif

#ifndef STORE
#define STORE(type,src,addr)	type src, [addr]
#endif

#ifndef STORE_BLK
#define STORE_BLK(src,addr)	stda src, [addr] ASI_BLK_P
#endif

#ifndef FUNC_NAME
#define FUNC_NAME	U3memcpy
#endif

#ifndef PREAMBLE
#define PREAMBLE
#endif

#ifndef XCC
#define XCC xcc
#endif

	.register	%g2,#scratch
	.register	%g3,#scratch

	/* Special/non-trivial issues of this code:
	 *
	 * 1) %o5 is preserved from VISEntryHalf to VISExitHalf
	 * 2) Only low 32 FPU registers are used so that only the
	 *    lower half of the FPU register set is dirtied by this
	 *    code.  This is especially important in the kernel.
	 * 3) This code never prefetches cachelines past the end
	 *    of the source buffer.
	 */

	.text
	.align		64

	/* The cheetah's flexible spine, oversized liver, enlarged heart,
	 * slender muscular body, and claws make it the swiftest hunter
	 * in Africa and the fastest animal on land.  Can reach speeds
	 * of up to 2.4GB per second.
	 */

	.globl	FUNC_NAME
	.type	FUNC_NAME,#function
FUNC_NAME:	/* %o0=dst, %o1=src, %o2=len */
	srlx		%o2, 31, %g2
	cmp		%g2, 0
	tne		%xcc, 5
	PREAMBLE
	mov		%o0, %o4
	cmp		%o2, 0
	be,pn		%XCC, 85f
	 or		%o0, %o1, %o3
	cmp		%o2, 16
	blu,a,pn	%XCC, 80f
	 or		%o3, %o2, %o3

	cmp		%o2, (3 * 64)
	blu,pt		%XCC, 70f
	 andcc		%o3, 0x7, %g0

	/* Clobbers o5/g1/g2/g3/g7/icc/xcc.  We must preserve
	 * o5 from here until we hit VISExitHalf.
	 */
	VISEntryHalf

	/* Is 'dst' already aligned on an 64-byte boundary? */
	andcc		%o0, 0x3f, %g2
	be,pt		%XCC, 2f

	/* Compute abs((dst & 0x3f) - 0x40) into %g2.  This is the number
	 * of bytes to copy to make 'dst' 64-byte aligned.  We pre-
	 * subtract this from 'len'.
	 */
	 sub		%o0, %o1, GLOBAL_SPARE
	sub		%g2, 0x40, %g2
	sub		%g0, %g2, %g2
	sub		%o2, %g2, %o2
	andcc		%g2, 0x7, %g1
	be,pt		%icc, 2f
	 and		%g2, 0x38, %g2

1:	subcc		%g1, 0x1, %g1
	EX_LD(LOAD(ldub, %o1 + 0x00, %o3))
	EX_ST(STORE(stb, %o3, %o1 + GLOBAL_SPARE))
	bgu,pt		%XCC, 1b
	 add		%o1, 0x1, %o1

	add		%o1, GLOBAL_SPARE, %o0

2:	cmp		%g2, 0x0
	and		%o1, 0x7, %g1
	be,pt		%icc, 3f
	 alignaddr	%o1, %g0, %o1

	EX_LD(LOAD(ldd, %o1, %f4))
1:	EX_LD(LOAD(ldd, %o1 + 0x8, %f6))
	add		%o1, 0x8, %o1
	subcc		%g2, 0x8, %g2
	faligndata	%f4, %f6, %f0
	EX_ST(STORE(std, %f0, %o0))
	be,pn		%icc, 3f
	 add		%o0, 0x8, %o0

	EX_LD(LOAD(ldd, %o1 + 0x8, %f4))
	add		%o1, 0x8, %o1
	subcc		%g2, 0x8, %g2
	faligndata	%f6, %f4, %f2
	EX_ST(STORE(std, %f2, %o0))
	bne,pt		%icc, 1b
	 add		%o0, 0x8, %o0

3:	LOAD(prefetch, %o1 + 0x000, #one_read)
	LOAD(prefetch, %o1 + 0x040, #one_read)
	andn		%o2, (0x40 - 1), GLOBAL_SPARE
	LOAD(prefetch, %o1 + 0x080, #one_read)
	LOAD(prefetch, %o1 + 0x0c0, #one_read)
	LOAD(prefetch, %o1 + 0x100, #one_read)
	EX_LD(LOAD(ldd, %o1 + 0x000, %f0))
	LOAD(prefetch, %o1 + 0x140, #one_read)
	EX_LD(LOAD(ldd, %o1 + 0x008, %f2))
	LOAD(prefetch, %o1 + 0x180, #one_read)
	EX_LD(LOAD(ldd, %o1 + 0x010, %f4))
	LOAD(prefetch, %o1 + 0x1c0, #one_read)
	faligndata	%f0, %f2, %f16
	EX_LD(LOAD(ldd, %o1 + 0x018, %f6))
	faligndata	%f2, %f4, %f18
	EX_LD(LOAD(ldd, %o1 + 0x020, %f8))
	faligndata	%f4, %f6, %f20
	EX_LD(LOAD(ldd, %o1 + 0x028, %f10))
	faligndata	%f6, %f8, %f22

	EX_LD(LOAD(ldd, %o1 + 0x030, %f12))
	faligndata	%f8, %f10, %f24
	EX_LD(LOAD(ldd, %o1 + 0x038, %f14))
	faligndata	%f10, %f12, %f26
	EX_LD(LOAD(ldd, %o1 + 0x040, %f0))

	subcc		GLOBAL_SPARE, 0x80, GLOBAL_SPARE
	add		%o1, 0x40, %o1
	bgu,pt		%XCC, 1f
	 srl		GLOBAL_SPARE, 6, %o3
	ba,pt		%xcc, 2f
	 nop

	.align		64
1:
	EX_LD(LOAD(ldd, %o1 + 0x008, %f2))
	faligndata	%f12, %f14, %f28
	EX_LD(LOAD(ldd, %o1 + 0x010, %f4))
	faligndata	%f14, %f0, %f30
	EX_ST(STORE_BLK(%f16, %o0))
	EX_LD(LOAD(ldd, %o1 + 0x018, %f6))
	faligndata	%f0, %f2, %f16
	add		%o0, 0x40, %o0

	EX_LD(LOAD(ldd, %o1 + 0x020, %f8))
	faligndata	%f2, %f4, %f18
	EX_LD(LOAD(ldd, %o1 + 0x028, %f10))
	faligndata	%f4, %f6, %f20
	EX_LD(LOAD(ldd, %o1 + 0x030, %f12))
	subcc		%o3, 0x01, %o3
	faligndata	%f6, %f8, %f22
	EX_LD(LOAD(ldd, %o1 + 0x038, %f14))

	faligndata	%f8, %f10, %f24
	EX_LD(LOAD(ldd, %o1 + 0x040, %f0))
	LOAD(prefetch, %o1 + 0x1c0, #one_read)
	faligndata	%f10, %f12, %f26
	bg,pt		%XCC, 1b
	 add		%o1, 0x40, %o1

	/* Finally we copy the last full 64-byte block. */
2:
	EX_LD(LOAD(ldd, %o1 + 0x008, %f2))
	faligndata	%f12, %f14, %f28
	EX_LD(LOAD(ldd, %o1 + 0x010, %f4))
	faligndata	%f14, %f0, %f30
	EX_ST(STORE_BLK(%f16, %o0))
	EX_LD(LOAD(ldd, %o1 + 0x018, %f6))
	faligndata	%f0, %f2, %f16
	EX_LD(LOAD(ldd, %o1 + 0x020, %f8))
	faligndata	%f2, %f4, %f18
	EX_LD(LOAD(ldd, %o1 + 0x028, %f10))
	faligndata	%f4, %f6, %f20
	EX_LD(LOAD(ldd, %o1 + 0x030, %f12))
	faligndata	%f6, %f8, %f22
	EX_LD(LOAD(ldd, %o1 + 0x038, %f14))
	faligndata	%f8, %f10, %f24
	cmp		%g1, 0
	be,pt		%XCC, 1f
	 add		%o0, 0x40, %o0
	EX_LD(LOAD(ldd, %o1 + 0x040, %f0))
1:	faligndata	%f10, %f12, %f26
	faligndata	%f12, %f14, %f28
	faligndata	%f14, %f0, %f30
	EX_ST(STORE_BLK(%f16, %o0))
	add		%o0, 0x40, %o0
	add		%o1, 0x40, %o1
	membar		#Sync

	/* Now we copy the (len modulo 64) bytes at the end.
	 * Note how we borrow the %f0 loaded above.
	 *
	 * Also notice how this code is careful not to perform a
	 * load past the end of the src buffer.
	 */
	and		%o2, 0x3f, %o2
	andcc		%o2, 0x38, %g2
	be,pn		%XCC, 2f
	 subcc		%g2, 0x8, %g2
	be,pn		%XCC, 2f
	 cmp		%g1, 0

	sub		%o2, %g2, %o2
	be,a,pt		%XCC, 1f
	 EX_LD(LOAD(ldd, %o1 + 0x00, %f0))

1:	EX_LD(LOAD(ldd, %o1 + 0x08, %f2))
	add		%o1, 0x8, %o1
	subcc		%g2, 0x8, %g2
	faligndata	%f0, %f2, %f8
	EX_ST(STORE(std, %f8, %o0))
	be,pn		%XCC, 2f
	 add		%o0, 0x8, %o0
	EX_LD(LOAD(ldd, %o1 + 0x08, %f0))
	add		%o1, 0x8, %o1
	subcc		%g2, 0x8, %g2
	faligndata	%f2, %f0, %f8
	EX_ST(STORE(std, %f8, %o0))
	bne,pn		%XCC, 1b
	 add		%o0, 0x8, %o0

	/* If anything is left, we copy it one byte at a time.
	 * Note that %g1 is (src & 0x3) saved above before the
	 * alignaddr was performed.
	 */
2:
	cmp		%o2, 0
	add		%o1, %g1, %o1
	VISExitHalf
	be,pn		%XCC, 85f
	 sub		%o0, %o1, %o3

	andcc		%g1, 0x7, %g0
	bne,pn		%icc, 90f
	 andcc		%o2, 0x8, %g0
	be,pt		%icc, 1f
	 nop
	EX_LD(LOAD(ldx, %o1, %o5))
	EX_ST(STORE(stx, %o5, %o1 + %o3))
	add		%o1, 0x8, %o1

1:	andcc		%o2, 0x4, %g0
	be,pt		%icc, 1f
	 nop
	EX_LD(LOAD(lduw, %o1, %o5))
	EX_ST(STORE(stw, %o5, %o1 + %o3))
	add		%o1, 0x4, %o1

1:	andcc		%o2, 0x2, %g0
	be,pt		%icc, 1f
	 nop
	EX_LD(LOAD(lduh, %o1, %o5))
	EX_ST(STORE(sth, %o5, %o1 + %o3))
	add		%o1, 0x2, %o1

1:	andcc		%o2, 0x1, %g0
	be,pt		%icc, 85f
	 nop
	EX_LD(LOAD(ldub, %o1, %o5))
	ba,pt		%xcc, 85f
	 EX_ST(STORE(stb, %o5, %o1 + %o3))

	.align		64
70: /* 16 < len <= 64 */
	bne,pn		%XCC, 75f
	 sub		%o0, %o1, %o3

72:
	andn		%o2, 0xf, GLOBAL_SPARE
	and		%o2, 0xf, %o2
1:	subcc		GLOBAL_SPARE, 0x10, GLOBAL_SPARE
	EX_LD(LOAD(ldx, %o1 + 0x00, %o5))
	EX_LD(LOAD(ldx, %o1 + 0x08, %g1))
	EX_ST(STORE(stx, %o5, %o1 + %o3))
	add		%o1, 0x8, %o1
	EX_ST(STORE(stx, %g1, %o1 + %o3))
	bgu,pt		%XCC, 1b
	 add		%o1, 0x8, %o1
73:	andcc		%o2, 0x8, %g0
	be,pt		%XCC, 1f
	 nop
	sub		%o2, 0x8, %o2
	EX_LD(LOAD(ldx, %o1, %o5))
	EX_ST(STORE(stx, %o5, %o1 + %o3))
	add		%o1, 0x8, %o1
1:	andcc		%o2, 0x4, %g0
	be,pt		%XCC, 1f
	 nop
	sub		%o2, 0x4, %o2
	EX_LD(LOAD(lduw, %o1, %o5))
	EX_ST(STORE(stw, %o5, %o1 + %o3))
	add		%o1, 0x4, %o1
1:	cmp		%o2, 0
	be,pt		%XCC, 85f
	 nop
	ba,pt		%xcc, 90f
	 nop

75:
	andcc		%o0, 0x7, %g1
	sub		%g1, 0x8, %g1
	be,pn		%icc, 2f
	 sub		%g0, %g1, %g1
	sub		%o2, %g1, %o2

1:	subcc		%g1, 1, %g1
	EX_LD(LOAD(ldub, %o1, %o5))
	EX_ST(STORE(stb, %o5, %o1 + %o3))
	bgu,pt		%icc, 1b
	 add		%o1, 1, %o1

2:	add		%o1, %o3, %o0
	andcc		%o1, 0x7, %g1
	bne,pt		%icc, 8f
	 sll		%g1, 3, %g1

	cmp		%o2, 16
	bgeu,pt		%icc, 72b
	 nop
	ba,a,pt		%xcc, 73b

8:	mov		64, %o3
	andn		%o1, 0x7, %o1
	EX_LD(LOAD(ldx, %o1, %g2))
	sub		%o3, %g1, %o3
	andn		%o2, 0x7, GLOBAL_SPARE
	sllx		%g2, %g1, %g2
1:	EX_LD(LOAD(ldx, %o1 + 0x8, %g3))
	subcc		GLOBAL_SPARE, 0x8, GLOBAL_SPARE
	add		%o1, 0x8, %o1
	srlx		%g3, %o3, %o5
	or		%o5, %g2, %o5
	EX_ST(STORE(stx, %o5, %o0))
	add		%o0, 0x8, %o0
	bgu,pt		%icc, 1b
	 sllx		%g3, %g1, %g2

	srl		%g1, 3, %g1
	andcc		%o2, 0x7, %o2
	be,pn		%icc, 85f
	 add		%o1, %g1, %o1
	ba,pt		%xcc, 90f
	 sub		%o0, %o1, %o3

	.align		64
80: /* 0 < len <= 16 */
	andcc		%o3, 0x3, %g0
	bne,pn		%XCC, 90f
	 sub		%o0, %o1, %o3

1:
	subcc		%o2, 4, %o2
	EX_LD(LOAD(lduw, %o1, %g1))
	EX_ST(STORE(stw, %g1, %o1 + %o3))
	bgu,pt		%XCC, 1b
	 add		%o1, 4, %o1

85:	retl
	 mov		EX_RETVAL(%o4), %o0

	.align		32
90:
	subcc		%o2, 1, %o2
	EX_LD(LOAD(ldub, %o1, %g1))
	EX_ST(STORE(stb, %g1, %o1 + %o3))
	bgu,pt		%XCC, 90b
	 add		%o1, 1, %o1
	retl
	 mov		EX_RETVAL(%o4), %o0

	.size		FUNC_NAME, .-FUNC_NAME