Lines Matching refs:v0
152 ushr v0.16b, v7.16b, #4 // vpsrlb $4, %xmm0, %xmm0
155 tbl v2.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm3, %xmm2
156 eor v0.16b, v1.16b, v16.16b // vpxor %xmm5, %xmm1, %xmm0
157 eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0
166 tbl v0.16b, {v24.16b}, v3.16b // vpshufb %xmm3, %xmm12, %xmm0 # 0 = sb1t
169 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A
172 tbl v3.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm3 # 0 = B
174 tbl v0.16b, {v0.16b}, v4.16b // vpshufb %xmm4, %xmm0, %xmm0 # 3 = D
177 eor v0.16b, v0.16b, v3.16b // vpxor %xmm3, %xmm0, %xmm0 # 3 = 2A+B+D
179 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = 2A+3B+C+D
184 and v1.16b, v0.16b, v17.16b // vpand %xmm0, %xmm9, %xmm1 # 0 = k
185 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i
187 eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j
188 tbl v3.16b, {v18.16b}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
195 eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo
205 tbl v0.16b, {v23.16b}, v3.16b // vpshufb %xmm3, %xmm0, %xmm0 # 0 = sb1t
207 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A
208 tbl v0.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm0
223 st1 {v0.16b}, [x1]
239 ushr v0.16b, v14.16b, #4 // vpsrlb $4, %xmm0, %xmm0
245 tbl v2.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm3, %xmm2
247 eor v0.16b, v1.16b, v16.16b // vpxor %xmm5, %xmm1, %xmm0
249 eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0
260 tbl v0.16b, {v24.16b}, v3.16b // vpshufb %xmm3, %xmm12, %xmm0 # 0 = sb1t
266 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A
271 tbl v3.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm3 # 0 = B
275 tbl v0.16b, {v0.16b}, v4.16b // vpshufb %xmm4, %xmm0, %xmm0 # 3 = D
281 eor v0.16b, v0.16b, v3.16b // vpxor %xmm3, %xmm0, %xmm0 # 3 = 2A+B+D
284 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = 2A+3B+C+D
290 and v1.16b, v0.16b, v17.16b // vpand %xmm0, %xmm9, %xmm1 # 0 = k
291 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i
296 eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j
298 tbl v3.16b, {v18.16b},v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
312 eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo
324 tbl v0.16b, {v23.16b}, v3.16b // vpshufb %xmm3, %xmm0, %xmm0 # 0 = sb1t
328 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A
330 tbl v0.16b, {v0.16b},v1.16b // vpshufb %xmm1, %xmm0, %xmm0
373 ushr v0.16b, v7.16b, #4 // vpsrlb $4, %xmm0, %xmm0
377 tbl v0.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm1, %xmm0
379 eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0
391 eor v0.16b, v4.16b, v16.16b // vpxor %xmm4, %xmm0, %xmm0
393 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
397 tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
399 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
401 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
405 tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
407 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
409 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
413 tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
415 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
417 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
422 and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k
423 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i
425 eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j
426 tbl v3.16b, {v18.16b}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
433 eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo
444 eor v0.16b, v1.16b, v4.16b // vpxor %xmm4, %xmm1, %xmm0 # 0 = A
445 tbl v0.16b, {v0.16b}, v2.16b // vpshufb %xmm2, %xmm0, %xmm0
460 st1 {v0.16b}, [x1]
485 ushr v0.16b, v14.16b, #4 // vpsrlb $4, %xmm0, %xmm0
492 tbl v0.16b, {v21.16b},v0.16b // vpshufb %xmm0, %xmm1, %xmm0
496 eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0
511 eor v0.16b, v4.16b, v16.16b // vpxor %xmm4, %xmm0, %xmm0
514 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
520 tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
524 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
527 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
533 tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
537 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
540 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
546 tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch
550 eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
553 eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
559 and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k
560 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i
565 eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j
567 tbl v3.16b, {v18.16b},v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
581 eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo
596 eor v0.16b, v1.16b, v4.16b // vpxor %xmm4, %xmm1, %xmm0 # 0 = A
598 tbl v0.16b, {v0.16b},v2.16b // vpshufb %xmm2, %xmm0, %xmm0
637 ld1 {v0.16b}, [x0],#16 // vmovdqu (%rdi), %xmm0 # load key (unaligned)
640 mov v3.16b, v0.16b // vmovdqa %xmm0, %xmm3
642 mov v7.16b, v0.16b // vmovdqa %xmm0, %xmm7
651 st1 {v0.2d}, [x2] // vmovdqu %xmm0, (%rdx)
703 ld1 {v0.16b}, [x0] // vmovdqu 8(%rdi),%xmm0 # load key part 2 (very unaligned)
705 mov v6.16b, v0.16b // vmovdqa %xmm0, %xmm6 # save short part
713 ext v0.16b, v6.16b, v0.16b, #8 // vpalignr $8,%xmm6,%xmm0,%xmm0
735 ld1 {v0.16b}, [x0] // vmovdqu 16(%rdi),%xmm0 # load key part 2 (unaligned)
742 mov v6.16b, v0.16b // vmovdqa %xmm0, %xmm6 # save cur_lo in xmm6
750 dup v0.4s, v0.s[3] // vpshufd $0xFF, %xmm0, %xmm0
782 tbl v0.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm0 # output permute
787 eor v0.16b, v0.16b, v16.16b // vpxor Lk_s63(%rip), %xmm0, %xmm0
789 st1 {v0.2d}, [x2] // vmovdqu %xmm0, (%rdx) # save last key
792 eor v0.16b, v0.16b, v0.16b // vpxor %xmm0, %xmm0, %xmm0
822 dup v0.4s, v7.s[3]
824 ins v0.s[0], v7.s[2] // vpshufd $0xFE, %xmm7, %xmm0 # b a _ _ -> b b b a
827 eor v6.16b, v6.16b, v0.16b // vpxor %xmm0, %xmm6, %xmm6 # -> b+c+d b+c b a
828 mov v0.16b, v6.16b // vmovdqa %xmm6, %xmm0
861 dup v0.4s, v0.s[3] // vpshufd $0xFF, %xmm0, %xmm0
862 ext v0.16b, v0.16b, v0.16b, #1 // vpalignr $1, %xmm0, %xmm0, %xmm0
874 and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k
875 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i
878 eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j
879 tbl v3.16b, {v18.16b}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
887 eor v2.16b, v2.16b, v0.16b // vpxor %xmm0, %xmm2, %xmm2 # 3 = jo
893 eor v0.16b, v1.16b, v7.16b // vpxor %xmm7, %xmm1, %xmm0
910 and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1
911 ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0
915 tbl v0.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm1, %xmm0
916 eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0
946 mov v4.16b, v0.16b // vmovdqa %xmm0, %xmm4 # save xmm0 for later
951 eor v4.16b, v0.16b, v16.16b // vpxor Lk_s63(%rip), %xmm0, %xmm4
1073 ld1 {v0.16b}, [x4] // load ivec
1080 eor v7.16b, v7.16b, v0.16b // xor with ivec
1082 st1 {v0.16b}, [x1],#16 // save output
1086 st1 {v0.16b}, [x4] // write ivec
1112 eor v0.16b, v0.16b, v6.16b // xor with ivec
1114 st1 {v0.16b}, [x1], #16
1122 eor v0.16b, v0.16b, v6.16b // xor with ivec
1125 st1 {v0.16b,v1.16b}, [x1], #32
1171 eor v0.16b, v0.16b, v6.16b // XOR input and result
1172 st1 {v0.16b}, [x1], #16
1192 eor v0.16b, v0.16b, v6.16b // XOR input and result
1194 st1 {v0.16b,v1.16b}, [x1], #32