|
@@ -100,9 +100,11 @@ ENTRY(clmul_ghash_mul)
|
|
|
movups (%rdi), DATA
|
|
|
movups (%rsi), SHASH
|
|
|
movaps .Lbswap_mask, BSWAP
|
|
|
- pshufb BSWAP, DATA
|
|
|
+ # pshufb BSWAP, DATA
|
|
|
+ .byte 0x66, 0x0f, 0x38, 0x00, 0xc5
|
|
|
call __clmul_gf128mul_ble
|
|
|
- pshufb BSWAP, DATA
|
|
|
+ # pshufb BSWAP, DATA
|
|
|
+ .byte 0x66, 0x0f, 0x38, 0x00, 0xc5
|
|
|
movups DATA, (%rdi)
|
|
|
ret
|
|
|
|
|
@@ -116,18 +118,21 @@ ENTRY(clmul_ghash_update)
|
|
|
movaps .Lbswap_mask, BSWAP
|
|
|
movups (%rdi), DATA
|
|
|
movups (%rcx), SHASH
|
|
|
- pshufb BSWAP, DATA
|
|
|
+ # pshufb BSWAP, DATA
|
|
|
+ .byte 0x66, 0x0f, 0x38, 0x00, 0xc5
|
|
|
.align 4
|
|
|
.Lupdate_loop:
|
|
|
movups (%rsi), IN1
|
|
|
- pshufb BSWAP, IN1
|
|
|
+ # pshufb BSWAP, IN1
|
|
|
+ .byte 0x66, 0x0f, 0x38, 0x00, 0xf5
|
|
|
pxor IN1, DATA
|
|
|
call __clmul_gf128mul_ble
|
|
|
sub $16, %rdx
|
|
|
add $16, %rsi
|
|
|
cmp $16, %rdx
|
|
|
jge .Lupdate_loop
|
|
|
- pshufb BSWAP, DATA
|
|
|
+ # pshufb BSWAP, DATA
|
|
|
+ .byte 0x66, 0x0f, 0x38, 0x00, 0xc5
|
|
|
movups DATA, (%rdi)
|
|
|
.Lupdate_just_ret:
|
|
|
ret
|
|
@@ -140,7 +145,8 @@ ENTRY(clmul_ghash_update)
|
|
|
ENTRY(clmul_ghash_setkey)
|
|
|
movaps .Lbswap_mask, BSWAP
|
|
|
movups (%rsi), %xmm0
|
|
|
- pshufb BSWAP, %xmm0
|
|
|
+ # pshufb BSWAP, %xmm0
|
|
|
+ .byte 0x66, 0x0f, 0x38, 0x00, 0xc5
|
|
|
movaps %xmm0, %xmm1
|
|
|
psllq $1, %xmm0
|
|
|
psrlq $63, %xmm1
|