^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1) /* SPDX-License-Identifier: GPL-2.0 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3) * arch/alpha/lib/strrchr.S
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) * Contributed by Richard Henderson (rth@tamu.edu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) * Return the address of the last occurrence of a given character
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7) * within a null-terminated string, or null if it is not found.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9) #include <asm/export.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) #include <asm/regdef.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12) .set noreorder
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13) .set noat
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15) .align 3
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) .ent strrchr
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) .globl strrchr
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18) strrchr:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) .frame sp, 0, ra
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) .prologue 0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) zapnot a1, 1, a1 # e0 : zero extend our test character
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23) mov zero, t6 # .. e1 : t6 is last match aligned addr
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) sll a1, 8, t5 # e0 : replicate our test character
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25) mov zero, t8 # .. e1 : t8 is last match byte compare mask
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26) or t5, a1, a1 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) ldq_u t0, 0(a0) # .. e1 : load first quadword
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) sll a1, 16, t5 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29) andnot a0, 7, v0 # .. e1 : align source addr
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) or t5, a1, a1 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31) lda t4, -1 # .. e1 : build garbage mask
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32) sll a1, 32, t5 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33) cmpbge zero, t0, t1 # .. e1 : bits set iff byte == zero
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) mskqh t4, a0, t4 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35) or t5, a1, a1 # .. e1 : character replication complete
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36) xor t0, a1, t2 # e0 : make bytes == c zero
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) cmpbge zero, t4, t4 # .. e1 : bits set iff byte is garbage
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) cmpbge zero, t2, t3 # e0 : bits set iff byte == c
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39) andnot t1, t4, t1 # .. e1 : clear garbage from null test
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) andnot t3, t4, t3 # e0 : clear garbage from char test
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41) bne t1, $eos # .. e1 : did we already hit the terminator?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43) /* Character search main loop */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44) $loop:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) ldq t0, 8(v0) # e0 : load next quadword
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) cmovne t3, v0, t6 # .. e1 : save previous comparisons match
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) cmovne t3, t3, t8 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48) addq v0, 8, v0 # .. e1 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49) xor t0, a1, t2 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) cmpbge zero, t0, t1 # .. e1 : bits set iff byte == zero
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51) cmpbge zero, t2, t3 # e0 : bits set iff byte == c
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) beq t1, $loop # .. e1 : if we havnt seen a null, loop
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54) /* Mask out character matches after terminator */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55) $eos:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) negq t1, t4 # e0 : isolate first null byte match
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) and t1, t4, t4 # e1 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) subq t4, 1, t5 # e0 : build a mask of the bytes up to...
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59) or t4, t5, t4 # e1 : ... and including the null
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) and t3, t4, t3 # e0 : mask out char matches after null
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62) cmovne t3, t3, t8 # .. e1 : save it, if match found
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) cmovne t3, v0, t6 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) /* Locate the address of the last matched character */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67) /* Retain the early exit for the ev4 -- the ev5 mispredict penalty
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68) is 5 cycles -- the same as just falling through. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69) beq t8, $retnull # .. e1 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) and t8, 0xf0, t2 # e0 : binary search for the high bit set
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72) cmovne t2, t2, t8 # .. e1 (zdb)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) cmovne t2, 4, t2 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74) and t8, 0xcc, t1 # .. e1 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75) cmovne t1, t1, t8 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) cmovne t1, 2, t1 # .. e1 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77) and t8, 0xaa, t0 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78) cmovne t0, 1, t0 # .. e1 (zdb)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79) addq t2, t1, t1 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80) addq t6, t0, v0 # .. e1 : add our aligned base ptr to the mix
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81) addq v0, t1, v0 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) ret # .. e1 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84) $retnull:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85) mov zero, v0 # e0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86) ret # .. e1 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88) .end strrchr
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89) EXPORT_SYMBOL(strrchr)