Lines Matching +full:3 +full:- +full:5
1 /* SPDX-License-Identifier: GPL-2.0 */
3 * arch/alpha/lib/ev6-memset.S
8 * 21264 version contributed by Rick Gorton <rick.gorton@alpha-processor.com>
13 * ftp.digital.com/pub/Digital/info/semiconductor/literature/dsc-library.html
15 * E - either cluster
16 * U - upper subcluster; U0 - subcluster U0; U1 - subcluster U1
17 * L - lower subcluster; L0 - subcluster L0; L1 - subcluster L1
41 .align 5
48 * undertake a major re-write to interleave the constant materialization
49 * with other parts of the fall-through code. This is important, even
60 insbl $1,2,$3 # U : 0000000000ch0000
61 insbl $1,3,$4 # U : 00000000ch000000
63 or $3,$4,$3 # E : 00000000chch0000
64 inswl $17,4,$5 # U : 0000chch00000000
68 or $17,$3,$17 # E : 00000000chchchch
69 or $2,$5,$2 # E : chchchch00000000
71 and $16,7,$3 # E : Target addr misalignment
76 beq $3,aligned_b # U : target is 0mod8
82 bis $16,$16,$5 # E : Save the address
84 subq $3,8,$3 # E : Invert (for addressing uses)
86 addq $18,$3,$18 # E : $18 is new count ($3 is negative)
88 subq $16,$3,$16 # E : $16 is new aligned destination
92 stq_u $1,0($5) # L : Store result
103 sra $18,3,$3 # U : Number of remaining quads to write
105 bis $16,$16,$5 # E : Save dest address
106 beq $3,no_quad_b # U : tail stuff only
113 * $5 A copy of $16
116 * $3 Number quads to write
120 subq $3, 16, $4 # E : Only try to unroll if > 128 bytes
136 stq $17, 0($5) # L :
137 subq $3, 1, $3 # E : For consistency later
139 addq $5, 8, $4 # E : Initial wh64 address (filler instruction)
143 addq $5, 8, $5 # E : Inc address
148 * $3 - number quads left to go
149 * $5 - target address (aligned 0mod64)
150 * $17 - mask of stuff to store
162 subq $3, 24, $2 # E : For determining future wh64 addresses
163 stq $17, 0($5) # L :
166 addq $5, 128, $4 # E : speculative target of next wh64
167 stq $17, 8($5) # L :
168 stq $17, 16($5) # L :
169 addq $5, 64, $7 # E : Fallback address for wh64 (== next trip addr)
171 stq $17, 24($5) # L :
172 stq $17, 32($5) # L :
176 stq $17, 40($5) # L :
177 stq $17, 48($5) # L :
178 subq $3, 16, $2 # E : Repeat the loop at least once more?
181 stq $17, 56($5) # L :
182 addq $5, 64, $5 # E :
183 subq $3, 8, $3 # E :
189 beq $3, no_quad_b # U : Might have finished already
197 stq $17,0($5) # L :
198 subq $3,1,$3 # E : Decrement number quads left
199 addq $5,8,$5 # E : Inc address
200 bne $3,loop_b # U : more?
208 ldq $7,0($5) # L :
213 stq $1,0($5) # L : And back to memory
253 and $16,7,$3 # E : Target addr misalignment
254 beq $3,aligned # U : target is 0mod8
260 bis $16,$16,$5 # E : Save the address
262 subq $3,8,$3 # E : Invert (for addressing uses)
264 addq $18,$3,$18 # E : $18 is new count ($3 is negative)
266 subq $16,$3,$16 # E : $16 is new aligned destination
270 stq_u $1,0($5) # L : Store result
281 sra $18,3,$3 # U : Number of remaining quads to write
283 bis $16,$16,$5 # E : Save dest address
284 beq $3,no_quad # U : tail stuff only
291 * $5 A copy of $16
294 * $3 Number quads to write
298 subq $3, 16, $4 # E : Only try to unroll if > 128 bytes
314 stq $17, 0($5) # L :
315 subq $3, 1, $3 # E : For consistency later
317 addq $5, 8, $4 # E : Initial wh64 address (filler instruction)
321 addq $5, 8, $5 # E : Inc address
326 * $3 - number quads left to go
327 * $5 - target address (aligned 0mod64)
328 * $17 - mask of stuff to store
340 subq $3, 24, $2 # E : For determining future wh64 addresses
341 stq $17, 0($5) # L :
344 addq $5, 128, $4 # E : speculative target of next wh64
345 stq $17, 8($5) # L :
346 stq $17, 16($5) # L :
347 addq $5, 64, $7 # E : Fallback address for wh64 (== next trip addr)
349 stq $17, 24($5) # L :
350 stq $17, 32($5) # L :
354 stq $17, 40($5) # L :
355 stq $17, 48($5) # L :
356 subq $3, 16, $2 # E : Repeat the loop at least once more?
359 stq $17, 56($5) # L :
360 addq $5, 64, $5 # E :
361 subq $3, 8, $3 # E :
367 beq $3, no_quad # U : Might have finished already
375 stq $17,0($5) # L :
376 subq $3,1,$3 # E : Decrement number quads left
377 addq $5,8,$5 # E : Inc address
378 bne $3,loop # U : more?
386 ldq $7,0($5) # L :
391 stq $1,0($5) # L : And back to memory
417 .align 5
424 inswl $17,0,$5 # U : 000000000000c1c2
430 inswl $17,4,$3 # U : 0000c1c200000000
434 or $2,$5,$2 # E : 00000000c1c2c1c2
435 or $3,$4,$17 # E : c1c2c1c200000000
437 and $16,7,$3 # E : Target addr misalignment
442 beq $3,aligned_w # U : target is 0mod8
448 bis $16,$16,$5 # E : Save the address
450 subq $3,8,$3 # E : Invert (for addressing uses)
452 addq $18,$3,$18 # E : $18 is new count ($3 is negative)
454 subq $16,$3,$16 # E : $16 is new aligned destination
458 stq_u $1,0($5) # L : Store result
469 sra $18,3,$3 # U : Number of remaining quads to write
471 bis $16,$16,$5 # E : Save dest address
472 beq $3,no_quad_w # U : tail stuff only
479 * $5 A copy of $16
482 * $3 Number quads to write
486 subq $3, 16, $4 # E : Only try to unroll if > 128 bytes
502 stq $17, 0($5) # L :
503 subq $3, 1, $3 # E : For consistency later
505 addq $5, 8, $4 # E : Initial wh64 address (filler instruction)
509 addq $5, 8, $5 # E : Inc address
514 * $3 - number quads left to go
515 * $5 - target address (aligned 0mod64)
516 * $17 - mask of stuff to store
528 subq $3, 24, $2 # E : For determining future wh64 addresses
529 stq $17, 0($5) # L :
532 addq $5, 128, $4 # E : speculative target of next wh64
533 stq $17, 8($5) # L :
534 stq $17, 16($5) # L :
535 addq $5, 64, $7 # E : Fallback address for wh64 (== next trip addr)
537 stq $17, 24($5) # L :
538 stq $17, 32($5) # L :
542 stq $17, 40($5) # L :
543 stq $17, 48($5) # L :
544 subq $3, 16, $2 # E : Repeat the loop at least once more?
547 stq $17, 56($5) # L :
548 addq $5, 64, $5 # E :
549 subq $3, 8, $3 # E :
555 beq $3, no_quad_w # U : Might have finished already
563 stq $17,0($5) # L :
564 subq $3,1,$3 # E : Decrement number quads left
565 addq $5,8,$5 # E : Inc address
566 bne $3,loop_w # U : more?
574 ldq $7,0($5) # L :
579 stq $1,0($5) # L : And back to memory