Lines Matching refs:v0
134 ushr v0.16b, v7.16b, #4 // vpsrlb $4, %xmm0, %xmm0
137 tbl v2.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm3, %xmm2
138 eor v0.16b, v1.16b, v16.16b // vpxor %xmm5, %xmm1, %xmm0
139 eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0
148 tbl v0.16b, {v24.16b}, v3.16b // vpshufb %xmm3, %xmm12, %xmm0 # 0 = sb1t
151 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A
154 tbl v3.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm3 # 0 = B
156 tbl v0.16b, {v0.16b}, v4.16b // vpshufb %xmm4, %xmm0, %xmm0 # 3 = D
159 eor v0.16b, v0.16b, v3.16b // vpxor %xmm3, %xmm0, %xmm0 # 3 = 2A+B+D
161 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = 2A+3B+C+D
166 and v1.16b, v0.16b, v17.16b // vpand %xmm0, %xmm9, %xmm1 # 0 = k
167 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i
169 eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j
170 tbl v3.16b, {v18.16b}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
177 eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo
187 tbl v0.16b, {v23.16b}, v3.16b // vpshufb %xmm3, %xmm0, %xmm0 # 0 = sb1t
189 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A
190 tbl v0.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm0
205 st1 {v0.16b}, [x1]
221 ushr v0.16b, v14.16b, #4 // vpsrlb $4, %xmm0, %xmm0
227 tbl v2.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm3, %xmm2
229 eor v0.16b, v1.16b, v16.16b // vpxor %xmm5, %xmm1, %xmm0
231 eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0
242 tbl v0.16b, {v24.16b}, v3.16b // vpshufb %xmm3, %xmm12, %xmm0 # 0 = sb1t
248 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A
253 tbl v3.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm3 # 0 = B
257 tbl v0.16b, {v0.16b}, v4.16b // vpshufb %xmm4, %xmm0, %xmm0 # 3 = D
263 eor v0.16b, v0.16b, v3.16b // vpxor %xmm3, %xmm0, %xmm0 # 3 = 2A+B+D
266 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = 2A+3B+C+D
272 and v1.16b, v0.16b, v17.16b // vpand %xmm0, %xmm9, %xmm1 # 0 = k
273 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i
278 eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j
280 tbl v3.16b, {v18.16b},v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
294 eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo
306 tbl v0.16b, {v23.16b}, v3.16b // vpshufb %xmm3, %xmm0, %xmm0 # 0 = sb1t
310 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A
312 tbl v0.16b, {v0.16b},v1.16b // vpshufb %xmm1, %xmm0, %xmm0
351 ushr v0.16b, v7.16b, #4 // vpsrlb $4, %xmm0, %xmm0
355 tbl v0.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm1, %xmm0
357 eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0
369 eor v0.16b, v4.16b, v16.16b // vpxor %xmm4, %xmm0, %xmm0
371 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
375 tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
377 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
379 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
383 tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
385 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
387 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
391 tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
393 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
395 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
400 and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k
401 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i
403 eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j
404 tbl v3.16b, {v18.16b}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
411 eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo
422 eor v0.16b, v1.16b, v4.16b // vpxor %xmm4, %xmm1, %xmm0 # 0 = A
423 tbl v0.16b, {v0.16b}, v2.16b // vpshufb %xmm2, %xmm0, %xmm0
438 st1 {v0.16b}, [x1]
462 ushr v0.16b, v14.16b, #4 // vpsrlb $4, %xmm0, %xmm0
469 tbl v0.16b, {v21.16b},v0.16b // vpshufb %xmm0, %xmm1, %xmm0
473 eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0
488 eor v0.16b, v4.16b, v16.16b // vpxor %xmm4, %xmm0, %xmm0
491 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
497 tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
501 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
504 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
510 tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
514 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
517 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
523 tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
527 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
530 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
536 and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k
537 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i
542 eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j
544 tbl v3.16b, {v18.16b},v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
558 eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo
573 eor v0.16b, v1.16b, v4.16b // vpxor %xmm4, %xmm1, %xmm0 # 0 = A
575 tbl v0.16b, {v0.16b},v2.16b // vpshufb %xmm2, %xmm0, %xmm0
611 ld1 {v0.16b}, [x0],#16 // vmovdqu (%rdi), %xmm0 # load key (unaligned)
614 mov v3.16b, v0.16b // vmovdqa %xmm0, %xmm3
616 mov v7.16b, v0.16b // vmovdqa %xmm0, %xmm7
623 st1 {v0.2d}, [x2] // vmovdqu %xmm0, (%rdx)
675 ld1 {v0.16b}, [x0] // vmovdqu 8(%rdi),%xmm0 # load key part 2 (very unaligned)
677 mov v6.16b, v0.16b // vmovdqa %xmm0, %xmm6 # save short part
685 ext v0.16b, v6.16b, v0.16b, #8 // vpalignr $8,%xmm6,%xmm0,%xmm0
707 ld1 {v0.16b}, [x0] // vmovdqu 16(%rdi),%xmm0 # load key part 2 (unaligned)
714 mov v6.16b, v0.16b // vmovdqa %xmm0, %xmm6 # save cur_lo in xmm6
722 dup v0.4s, v0.s[3] // vpshufd $0xFF, %xmm0, %xmm0
751 tbl v0.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm0 # output permute
756 eor v0.16b, v0.16b, v16.16b // vpxor .Lk_s63(%rip), %xmm0, %xmm0
758 st1 {v0.2d}, [x2] // vmovdqu %xmm0, (%rdx) # save last key
761 eor v0.16b, v0.16b, v0.16b // vpxor %xmm0, %xmm0, %xmm0
792 dup v0.4s, v7.s[3]
794 ins v0.s[0], v7.s[2] // vpshufd $0xFE, %xmm7, %xmm0 # b a _ _ -> b b b a
797 eor v6.16b, v6.16b, v0.16b // vpxor %xmm0, %xmm6, %xmm6 # -> b+c+d b+c b a
798 mov v0.16b, v6.16b // vmovdqa %xmm6, %xmm0
831 dup v0.4s, v0.s[3] // vpshufd $0xFF, %xmm0, %xmm0
832 ext v0.16b, v0.16b, v0.16b, #1 // vpalignr $1, %xmm0, %xmm0, %xmm0
844 and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k
845 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i
848 eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j
849 tbl v3.16b, {v18.16b}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
857 eor v2.16b, v2.16b, v0.16b // vpxor %xmm0, %xmm2, %xmm2 # 3 = jo
863 eor v0.16b, v1.16b, v7.16b // vpxor %xmm7, %xmm1, %xmm0
880 and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1
881 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0
885 tbl v0.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm1, %xmm0
886 eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0
916 mov v4.16b, v0.16b // vmovdqa %xmm0, %xmm4 # save xmm0 for later
921 eor v4.16b, v0.16b, v16.16b // vpxor .Lk_s63(%rip), %xmm0, %xmm4
1045 ld1 {v0.16b}, [x4] // load ivec
1052 eor v7.16b, v7.16b, v0.16b // xor with ivec
1054 st1 {v0.16b}, [x1],#16 // save output
1058 st1 {v0.16b}, [x4] // write ivec
1086 eor v0.16b, v0.16b, v6.16b // xor with ivec
1088 st1 {v0.16b}, [x1], #16
1096 eor v0.16b, v0.16b, v6.16b // xor with ivec
1099 st1 {v0.16b,v1.16b}, [x1], #32
1134 st1 {v0.16b}, [x1],#16
1142 st1 {v0.16b,v1.16b}, [x1], #32
1176 st1 {v0.16b}, [x1],#16
1184 st1 {v0.16b,v1.16b}, [x1], #32