+.size aesni_ccm64_decrypt_blocks,.-aesni_ccm64_decrypt_blocks
+___
+}\f
+######################################################################
+# void aesni_ctr32_encrypt_blocks (const void *in, void *out,
+# size_t blocks, const AES_KEY *key,
+# const char *ivec);
+#
+# Handles only complete blocks, operates on 32-bit counter and
+# does not update *ivec! (see engine/eng_aesni.c for details)
+#
+{
+my $reserved = $win64?0:-0x28;
+my ($in0,$in1,$in2,$in3)=map("%xmm$_",(8..11));
+my ($iv0,$iv1,$ivec)=("%xmm12","%xmm13","%xmm14");
+my $bswap_mask="%xmm15";
+
+$code.=<<___;
+.globl aesni_ctr32_encrypt_blocks
+.type aesni_ctr32_encrypt_blocks,\@function,5
+.align 16
+aesni_ctr32_encrypt_blocks:
+___
+$code.=<<___ if ($win64);
+ lea -0xc8(%rsp),%rsp
+ movaps %xmm6,0x20(%rsp)
+ movaps %xmm7,0x30(%rsp)
+ movaps %xmm8,0x40(%rsp)
+ movaps %xmm9,0x50(%rsp)
+ movaps %xmm10,0x60(%rsp)
+ movaps %xmm11,0x70(%rsp)
+ movaps %xmm12,0x80(%rsp)
+ movaps %xmm13,0x90(%rsp)
+ movaps %xmm14,0xa0(%rsp)
+ movaps %xmm15,0xb0(%rsp)
+.Lctr32_body:
+___
+$code.=<<___;
+ cmp \$1,$len
+ je .Lctr32_one_shortcut
+
+ movdqu ($ivp),$ivec
+ movdqa .Lbswap_mask(%rip),$bswap_mask
+ xor $rounds,$rounds
+ pextrd \$3,$ivec,$rnds_ # pull 32-bit counter
+ pinsrd \$3,$rounds,$ivec # wipe 32-bit counter
+
+ mov 240($key),$rounds # key->rounds
+ bswap $rnds_
+ pxor $iv0,$iv0 # vector of 3 32-bit counters
+ pxor $iv1,$iv1 # vector of 3 32-bit counters
+ pinsrd \$0,$rnds_,$iv0
+ lea 3($rnds_),$key_
+ pinsrd \$0,$key_,$iv1
+ inc $rnds_
+ pinsrd \$1,$rnds_,$iv0
+ inc $key_
+ pinsrd \$1,$key_,$iv1
+ inc $rnds_
+ pinsrd \$2,$rnds_,$iv0
+ inc $key_
+ pinsrd \$2,$key_,$iv1
+ movdqa $iv0,$reserved(%rsp)
+ pshufb $bswap_mask,$iv0
+ movdqa $iv1,`$reserved+0x10`(%rsp)
+ pshufb $bswap_mask,$iv1
+
+ pshufd \$`3<<6`,$iv0,$inout0 # place counter to upper dword
+ pshufd \$`2<<6`,$iv0,$inout1
+ pshufd \$`1<<6`,$iv0,$inout2
+ cmp \$6,$len
+ jb .Lctr32_tail
+ shr \$1,$rounds
+ mov $key,$key_ # backup $key
+ mov $rounds,$rnds_ # backup $rounds
+ sub \$6,$len
+ jmp .Lctr32_loop6
+
+.align 16
+.Lctr32_loop6:
+ pshufd \$`3<<6`,$iv1,$inout3
+ por $ivec,$inout0 # merge counter-less ivec
+ $movkey ($key_),$rndkey0
+ pshufd \$`2<<6`,$iv1,$inout4
+ por $ivec,$inout1
+ $movkey 16($key_),$rndkey1
+ pshufd \$`1<<6`,$iv1,$inout5
+ por $ivec,$inout2
+ por $ivec,$inout3
+ xorps $rndkey0,$inout0
+ por $ivec,$inout4
+ por $ivec,$inout5
+
+ # inline _aesni_encrypt6 and interleave last rounds
+ # with own code...
+
+ pxor $rndkey0,$inout1
+ aesenc $rndkey1,$inout0
+ lea 32($key_),$key
+ pxor $rndkey0,$inout2
+ aesenc $rndkey1,$inout1
+ movdqa .Lincrement32(%rip),$iv1
+ pxor $rndkey0,$inout3
+ aesenc $rndkey1,$inout2
+ movdqa $reserved(%rsp),$iv0
+ pxor $rndkey0,$inout4
+ aesenc $rndkey1,$inout3
+ pxor $rndkey0,$inout5
+ $movkey ($key),$rndkey0
+ dec $rounds
+ aesenc $rndkey1,$inout4
+ aesenc $rndkey1,$inout5
+ jmp .Lctr32_enc_loop6_enter
+.align 16
+.Lctr32_enc_loop6:
+ aesenc $rndkey1,$inout0
+ aesenc $rndkey1,$inout1
+ dec $rounds
+ aesenc $rndkey1,$inout2
+ aesenc $rndkey1,$inout3
+ aesenc $rndkey1,$inout4
+ aesenc $rndkey1,$inout5
+.Lctr32_enc_loop6_enter:
+ $movkey 16($key),$rndkey1
+ aesenc $rndkey0,$inout0
+ aesenc $rndkey0,$inout1
+ lea 32($key),$key
+ aesenc $rndkey0,$inout2
+ aesenc $rndkey0,$inout3
+ aesenc $rndkey0,$inout4
+ aesenc $rndkey0,$inout5
+ $movkey ($key),$rndkey0
+ jnz .Lctr32_enc_loop6
+
+ aesenc $rndkey1,$inout0
+ paddd $iv1,$iv0 # increment counter vector
+ aesenc $rndkey1,$inout1
+ paddd `$reserved+0x10`(%rsp),$iv1
+ aesenc $rndkey1,$inout2
+ movdqa $iv0,$reserved(%rsp) # save counter vector
+ aesenc $rndkey1,$inout3
+ movdqa $iv1,`$reserved+0x10`(%rsp)
+ aesenc $rndkey1,$inout4
+ pshufb $bswap_mask,$iv0 # byte swap
+ aesenc $rndkey1,$inout5
+ pshufb $bswap_mask,$iv1
+
+ aesenclast $rndkey0,$inout0
+ movups ($inp),$in0 # load input
+ aesenclast $rndkey0,$inout1
+ movups 0x10($inp),$in1
+ aesenclast $rndkey0,$inout2
+ movups 0x20($inp),$in2
+ aesenclast $rndkey0,$inout3
+ movups 0x30($inp),$in3
+ aesenclast $rndkey0,$inout4
+ movups 0x40($inp),$rndkey1
+ aesenclast $rndkey0,$inout5
+ movups 0x50($inp),$rndkey0
+ lea 0x60($inp),$inp
+
+ xorps $inout0,$in0 # xor
+ pshufd \$`3<<6`,$iv0,$inout0
+ xorps $inout1,$in1
+ pshufd \$`2<<6`,$iv0,$inout1
+ movups $in0,($out) # store output
+ xorps $inout2,$in2
+ pshufd \$`1<<6`,$iv0,$inout2
+ movups $in1,0x10($out)
+ xorps $inout3,$in3
+ movups $in2,0x20($out)
+ xorps $inout4,$rndkey1
+ movups $in3,0x30($out)
+ xorps $inout5,$rndkey0
+ movups $rndkey1,0x40($out)
+ movups $rndkey0,0x50($out)
+ lea 0x60($out),$out
+ mov $rnds_,$rounds
+ sub \$6,$len
+ jnc .Lctr32_loop6
+
+ add \$6,$len
+ jz .Lctr32_done
+ mov $key_,$key # restore $key
+ lea 1($rounds,$rounds),$rounds # restore original value
+
+.Lctr32_tail:
+ por $ivec,$inout0
+ movups ($inp),$in0
+ cmp \$2,$len
+ jb .Lctr32_one
+
+ por $ivec,$inout1
+ movups 0x10($inp),$in1
+ je .Lctr32_two
+
+ pshufd \$`3<<6`,$iv1,$inout3
+ por $ivec,$inout2
+ movups 0x20($inp),$in2
+ cmp \$4,$len
+ jb .Lctr32_three
+
+ pshufd \$`2<<6`,$iv1,$inout4
+ por $ivec,$inout3
+ movups 0x30($inp),$in3
+ je .Lctr32_four
+
+ por $ivec,$inout4
+ xorps $inout5,$inout5
+
+ call _aesni_encrypt6
+
+ movups 0x40($inp),$rndkey1
+ xorps $inout0,$in0
+ xorps $inout1,$in1
+ movups $in0,($out)
+ xorps $inout2,$in2
+ movups $in1,0x10($out)
+ xorps $inout3,$in3
+ movups $in2,0x20($out)
+ xorps $inout4,$rndkey1
+ movups $in3,0x30($out)
+ movups $rndkey1,0x40($out)
+ jmp .Lctr32_done
+
+.align 16
+.Lctr32_one_shortcut:
+ movups ($ivp),$inout0
+ movups ($inp),$in0
+ mov 240($key),$rounds # key->rounds
+.Lctr32_one:
+___
+ &aesni_generate1("enc",$key,$rounds);
+$code.=<<___;
+ xorps $inout0,$in0
+ movups $in0,($out)
+ jmp .Lctr32_done
+
+.align 16
+.Lctr32_two:
+ xorps $inout2,$inout2
+ call _aesni_encrypt3
+ xorps $inout0,$in0
+ xorps $inout1,$in1
+ movups $in0,($out)
+ movups $in1,0x10($out)
+ jmp .Lctr32_done
+
+.align 16
+.Lctr32_three:
+ call _aesni_encrypt3
+ xorps $inout0,$in0
+ xorps $inout1,$in1
+ movups $in0,($out)
+ xorps $inout2,$in2
+ movups $in1,0x10($out)
+ movups $in2,0x20($out)
+ jmp .Lctr32_done
+
+.align 16
+.Lctr32_four:
+ call _aesni_encrypt4
+ xorps $inout0,$in0
+ xorps $inout1,$in1
+ movups $in0,($out)
+ xorps $inout2,$in2
+ movups $in1,0x10($out)
+ xorps $inout3,$in3
+ movups $in2,0x20($out)
+ movups $in3,0x30($out)
+
+.Lctr32_done:
+___
+$code.=<<___ if ($win64);
+ movaps 0x20(%rsp),%xmm6
+ movaps 0x30(%rsp),%xmm7
+ movaps 0x40(%rsp),%xmm8
+ movaps 0x50(%rsp),%xmm9
+ movaps 0x60(%rsp),%xmm10
+ movaps 0x70(%rsp),%xmm11
+ movaps 0x80(%rsp),%xmm12
+ movaps 0x90(%rsp),%xmm13
+ movaps 0xa0(%rsp),%xmm14
+ movaps 0xb0(%rsp),%xmm15
+ lea 0xc8(%rsp),%rsp
+.Lctr32_ret:
+___
+$code.=<<___;
+ ret
+.size aesni_ctr32_encrypt_blocks,.-aesni_ctr32_encrypt_blocks
+___
+}
+\f
+######################################################################
+# void aesni_xts_[en|de]crypt(const char *inp,char *out,size_t len,
+# const AES_KEY *key1, const AES_KEY *key2
+# const unsigned char iv[16]);
+#
+{
+my @tweak=map("%xmm$_",(10..15));
+my ($twmask,$twres,$twtmp)=("%xmm8","%xmm9",@tweak[4]);
+my ($key2,$ivp,$len_)=("%r8","%r9","%r9");
+my $frame_size = 0x68 + ($win64?160:0);
+
+$code.=<<___;
+.globl aesni_xts_encrypt
+.type aesni_xts_encrypt,\@function,6
+.align 16
+aesni_xts_encrypt:
+ lea -$frame_size(%rsp),%rsp
+___
+$code.=<<___ if ($win64);
+ movaps %xmm6,0x60(%rsp)
+ movaps %xmm7,0x70(%rsp)
+ movaps %xmm8,0x80(%rsp)
+ movaps %xmm9,0x90(%rsp)
+ movaps %xmm10,0xa0(%rsp)
+ movaps %xmm11,0xb0(%rsp)
+ movaps %xmm12,0xc0(%rsp)
+ movaps %xmm13,0xd0(%rsp)
+ movaps %xmm14,0xe0(%rsp)
+ movaps %xmm15,0xf0(%rsp)
+.Lxts_enc_body:
+___
+$code.=<<___;
+ movups ($ivp),@tweak[5] # load clear-text tweak
+ mov 240(%r8),$rounds # key2->rounds
+ mov 240($key),$rnds_ # key1->rounds
+___
+ # generate the tweak
+ &aesni_generate1("enc",$key2,$rounds,@tweak[5]);
+$code.=<<___;
+ mov $key,$key_ # backup $key
+ mov $rnds_,$rounds # backup $rounds
+ mov $len,$len_ # backup $len
+ and \$-16,$len
+
+ movdqa .Lxts_magic(%rip),$twmask
+ pxor $twtmp,$twtmp
+ pcmpgtd @tweak[5],$twtmp # broadcast upper bits
+___
+ for ($i=0;$i<4;$i++) {
+ $code.=<<___;
+ pshufd \$0x13,$twtmp,$twres
+ pxor $twtmp,$twtmp
+ movdqa @tweak[5],@tweak[$i]
+ paddq @tweak[5],@tweak[5] # psllq 1,$tweak
+ pand $twmask,$twres # isolate carry and residue
+ pcmpgtd @tweak[5],$twtmp # broadcat upper bits
+ pxor $twres,@tweak[5]
+___
+ }
+$code.=<<___;
+ sub \$16*6,$len
+ jc .Lxts_enc_short
+
+ shr \$1,$rounds
+ sub \$1,$rounds
+ mov $rounds,$rnds_
+ jmp .Lxts_enc_grandloop
+
+.align 16
+.Lxts_enc_grandloop:
+ pshufd \$0x13,$twtmp,$twres
+ movdqa @tweak[5],@tweak[4]
+ paddq @tweak[5],@tweak[5] # psllq 1,$tweak
+ movdqu `16*0`($inp),$inout0 # load input
+ pand $twmask,$twres # isolate carry and residue
+ movdqu `16*1`($inp),$inout1
+ pxor $twres,@tweak[5]
+
+ movdqu `16*2`($inp),$inout2
+ pxor @tweak[0],$inout0 # input^=tweak
+ movdqu `16*3`($inp),$inout3
+ pxor @tweak[1],$inout1
+ movdqu `16*4`($inp),$inout4
+ pxor @tweak[2],$inout2
+ movdqu `16*5`($inp),$inout5
+ lea `16*6`($inp),$inp
+ pxor @tweak[3],$inout3
+ $movkey ($key_),$rndkey0
+ pxor @tweak[4],$inout4
+ pxor @tweak[5],$inout5
+
+ # inline _aesni_encrypt6 and interleave first and last rounds
+ # with own code...
+ $movkey 16($key_),$rndkey1
+ pxor $rndkey0,$inout0
+ pxor $rndkey0,$inout1
+ movdqa @tweak[0],`16*0`(%rsp) # put aside tweaks
+ aesenc $rndkey1,$inout0
+ lea 32($key_),$key
+ pxor $rndkey0,$inout2
+ movdqa @tweak[1],`16*1`(%rsp)
+ aesenc $rndkey1,$inout1
+ pxor $rndkey0,$inout3
+ movdqa @tweak[2],`16*2`(%rsp)
+ aesenc $rndkey1,$inout2
+ pxor $rndkey0,$inout4
+ movdqa @tweak[3],`16*3`(%rsp)
+ aesenc $rndkey1,$inout3
+ pxor $rndkey0,$inout5
+ $movkey ($key),$rndkey0
+ dec $rounds
+ movdqa @tweak[4],`16*4`(%rsp)
+ aesenc $rndkey1,$inout4
+ movdqa @tweak[5],`16*5`(%rsp)
+ aesenc $rndkey1,$inout5
+ pxor $twtmp,$twtmp
+ pcmpgtd @tweak[5],$twtmp
+ jmp .Lxts_enc_loop6_enter
+
+.align 16
+.Lxts_enc_loop6:
+ aesenc $rndkey1,$inout0
+ aesenc $rndkey1,$inout1
+ dec $rounds
+ aesenc $rndkey1,$inout2
+ aesenc $rndkey1,$inout3
+ aesenc $rndkey1,$inout4
+ aesenc $rndkey1,$inout5
+.Lxts_enc_loop6_enter:
+ $movkey 16($key),$rndkey1
+ aesenc $rndkey0,$inout0
+ aesenc $rndkey0,$inout1
+ lea 32($key),$key
+ aesenc $rndkey0,$inout2
+ aesenc $rndkey0,$inout3
+ aesenc $rndkey0,$inout4
+ aesenc $rndkey0,$inout5
+ $movkey ($key),$rndkey0
+ jnz .Lxts_enc_loop6
+
+ pshufd \$0x13,$twtmp,$twres
+ pxor $twtmp,$twtmp
+ paddq @tweak[5],@tweak[5] # psllq 1,$tweak
+ aesenc $rndkey1,$inout0
+ pand $twmask,$twres # isolate carry and residue
+ aesenc $rndkey1,$inout1
+ pcmpgtd @tweak[5],$twtmp # broadcast upper bits
+ aesenc $rndkey1,$inout2
+ pxor $twres,@tweak[5]
+ aesenc $rndkey1,$inout3
+ aesenc $rndkey1,$inout4
+ aesenc $rndkey1,$inout5
+ $movkey 16($key),$rndkey1
+
+ pshufd \$0x13,$twtmp,$twres
+ pxor $twtmp,$twtmp
+ movdqa @tweak[5],@tweak[0]
+ paddq @tweak[5],@tweak[5] # psllq 1,$tweak
+ aesenc $rndkey0,$inout0
+ pand $twmask,$twres # isolate carry and residue
+ aesenc $rndkey0,$inout1
+ pcmpgtd @tweak[5],$twtmp # broadcat upper bits
+ aesenc $rndkey0,$inout2
+ pxor $twres,@tweak[5]
+ aesenc $rndkey0,$inout3
+ aesenc $rndkey0,$inout4
+ aesenc $rndkey0,$inout5
+ $movkey 32($key),$rndkey0
+
+ pshufd \$0x13,$twtmp,$twres
+ pxor $twtmp,$twtmp
+ movdqa @tweak[5],@tweak[1]
+ paddq @tweak[5],@tweak[5] # psllq 1,$tweak
+ aesenc $rndkey1,$inout0
+ pand $twmask,$twres # isolate carry and residue
+ aesenc $rndkey1,$inout1
+ pcmpgtd @tweak[5],$twtmp # broadcat upper bits
+ aesenc $rndkey1,$inout2
+ pxor $twres,@tweak[5]
+ aesenc $rndkey1,$inout3
+ aesenc $rndkey1,$inout4
+ aesenc $rndkey1,$inout5
+
+ pshufd \$0x13,$twtmp,$twres
+ pxor $twtmp,$twtmp
+ movdqa @tweak[5],@tweak[2]
+ paddq @tweak[5],@tweak[5] # psllq 1,$tweak
+ aesenclast $rndkey0,$inout0
+ pand $twmask,$twres # isolate carry and residue
+ aesenclast $rndkey0,$inout1
+ pcmpgtd @tweak[5],$twtmp # broadcat upper bits
+ aesenclast $rndkey0,$inout2
+ pxor $twres,@tweak[5]
+ aesenclast $rndkey0,$inout3
+ aesenclast $rndkey0,$inout4
+ aesenclast $rndkey0,$inout5
+
+ pshufd \$0x13,$twtmp,$twres
+ pxor $twtmp,$twtmp
+ movdqa @tweak[5],@tweak[3]
+ paddq @tweak[5],@tweak[5] # psllq 1,$tweak
+ xorps `16*0`(%rsp),$inout0 # output^=tweak
+ pand $twmask,$twres # isolate carry and residue
+ xorps `16*1`(%rsp),$inout1
+ pcmpgtd @tweak[5],$twtmp # broadcat upper bits
+ pxor $twres,@tweak[5]
+
+ xorps `16*2`(%rsp),$inout2
+ movups $inout0,`16*0`($out) # write output
+ xorps `16*3`(%rsp),$inout3
+ movups $inout1,`16*1`($out)
+ xorps `16*4`(%rsp),$inout4
+ movups $inout2,`16*2`($out)
+ xorps `16*5`(%rsp),$inout5
+ movups $inout3,`16*3`($out)
+ mov $rnds_,$rounds # restore $rounds
+ movups $inout4,`16*4`($out)
+ movups $inout5,`16*5`($out)
+ lea `16*6`($out),$out
+ sub \$16*6,$len
+ jnc .Lxts_enc_grandloop
+
+ lea 3($rounds,$rounds),$rounds # restore original value
+ mov $key_,$key # restore $key
+ mov $rounds,$rnds_ # backup $rounds
+
+.Lxts_enc_short:
+ add \$16*6,$len
+ jz .Lxts_enc_done
+
+ cmp \$0x20,$len
+ jb .Lxts_enc_one
+ je .Lxts_enc_two
+
+ cmp \$0x40,$len
+ jb .Lxts_enc_three
+ je .Lxts_enc_four
+
+ pshufd \$0x13,$twtmp,$twres
+ movdqa @tweak[5],@tweak[4]
+ paddq @tweak[5],@tweak[5] # psllq 1,$tweak
+ movdqu ($inp),$inout0
+ pand $twmask,$twres # isolate carry and residue
+ movdqu 16*1($inp),$inout1
+ pxor $twres,@tweak[5]
+
+ movdqu 16*2($inp),$inout2
+ pxor @tweak[0],$inout0
+ movdqu 16*3($inp),$inout3
+ pxor @tweak[1],$inout1
+ movdqu 16*4($inp),$inout4
+ lea 16*5($inp),$inp
+ pxor @tweak[2],$inout2
+ pxor @tweak[3],$inout3
+ pxor @tweak[4],$inout4
+
+ call _aesni_encrypt6
+
+ xorps @tweak[0],$inout0
+ movdqa @tweak[5],@tweak[0]
+ xorps @tweak[1],$inout1
+ xorps @tweak[2],$inout2
+ movdqu $inout0,($out)
+ xorps @tweak[3],$inout3
+ movdqu $inout1,16*1($out)
+ xorps @tweak[4],$inout4
+ movdqu $inout2,16*2($out)
+ movdqu $inout3,16*3($out)
+ movdqu $inout4,16*4($out)
+ lea 16*5($out),$out
+ jmp .Lxts_enc_done
+
+.align 16
+.Lxts_enc_one:
+ movups ($inp),$inout0
+ lea 16*1($inp),$inp
+ xorps @tweak[0],$inout0
+___
+ &aesni_generate1("enc",$key,$rounds);
+$code.=<<___;
+ xorps @tweak[0],$inout0
+ movdqa @tweak[1],@tweak[0]
+ movups $inout0,($out)
+ lea 16*1($out),$out
+ jmp .Lxts_enc_done
+
+.align 16
+.Lxts_enc_two:
+ movups ($inp),$inout0
+ movups 16($inp),$inout1
+ lea 32($inp),$inp
+ xorps @tweak[0],$inout0
+ xorps @tweak[1],$inout1
+
+ call _aesni_encrypt3
+
+ xorps @tweak[0],$inout0
+ movdqa @tweak[2],@tweak[0]
+ xorps @tweak[1],$inout1
+ movups $inout0,($out)
+ movups $inout1,16*1($out)
+ lea 16*2($out),$out
+ jmp .Lxts_enc_done
+
+.align 16
+.Lxts_enc_three:
+ movups ($inp),$inout0
+ movups 16*1($inp),$inout1
+ movups 16*2($inp),$inout2
+ lea 16*3($inp),$inp
+ xorps @tweak[0],$inout0
+ xorps @tweak[1],$inout1
+ xorps @tweak[2],$inout2
+
+ call _aesni_encrypt3
+
+ xorps @tweak[0],$inout0
+ movdqa @tweak[3],@tweak[0]
+ xorps @tweak[1],$inout1
+ xorps @tweak[2],$inout2
+ movups $inout0,($out)
+ movups $inout1,16*1($out)
+ movups $inout2,16*2($out)
+ lea 16*3($out),$out
+ jmp .Lxts_enc_done
+
+.align 16
+.Lxts_enc_four:
+ movups ($inp),$inout0
+ movups 16*1($inp),$inout1
+ movups 16*2($inp),$inout2
+ xorps @tweak[0],$inout0
+ movups 16*3($inp),$inout3
+ lea 16*4($inp),$inp
+ xorps @tweak[1],$inout1
+ xorps @tweak[2],$inout2
+ xorps @tweak[3],$inout3
+
+ call _aesni_encrypt4
+
+ xorps @tweak[0],$inout0
+ movdqa @tweak[5],@tweak[0]
+ xorps @tweak[1],$inout1
+ xorps @tweak[2],$inout2
+ movups $inout0,($out)
+ xorps @tweak[3],$inout3
+ movups $inout1,16*1($out)
+ movups $inout2,16*2($out)
+ movups $inout3,16*3($out)
+ lea 16*4($out),$out
+ jmp .Lxts_enc_done
+
+.align 16
+.Lxts_enc_done:
+ and \$15,$len_
+ jz .Lxts_enc_ret
+ mov $len_,$len
+
+.Lxts_enc_steal:
+ movzb ($inp),%eax # borrow $rounds ...
+ movzb -16($out),%ecx # ... and $key
+ lea 1($inp),$inp
+ mov %al,-16($out)
+ mov %cl,0($out)
+ lea 1($out),$out
+ sub \$1,$len
+ jnz .Lxts_enc_steal
+
+ sub $len_,$out # rewind $out
+ mov $key_,$key # restore $key
+ mov $rnds_,$rounds # restore $rounds
+
+ movups -16($out),$inout0
+ xorps @tweak[0],$inout0
+___
+ &aesni_generate1("enc",$key,$rounds);
+$code.=<<___;
+ xorps @tweak[0],$inout0
+ movups $inout0,-16($out)
+
+.Lxts_enc_ret:
+___
+$code.=<<___ if ($win64);
+ movaps 0x60(%rsp),%xmm6
+ movaps 0x70(%rsp),%xmm7
+ movaps 0x80(%rsp),%xmm8
+ movaps 0x90(%rsp),%xmm9
+ movaps 0xa0(%rsp),%xmm10
+ movaps 0xb0(%rsp),%xmm11
+ movaps 0xc0(%rsp),%xmm12
+ movaps 0xd0(%rsp),%xmm13
+ movaps 0xe0(%rsp),%xmm14
+ movaps 0xf0(%rsp),%xmm15
+___
+$code.=<<___;
+ lea $frame_size(%rsp),%rsp
+.Lxts_enc_epilogue:
+ ret
+.size aesni_xts_encrypt,.-aesni_xts_encrypt