summaryrefslogtreecommitdiffstats
path: root/security/nss/lib/freebl/intel-gcm-x86-masm.asm
diff options
context:
space:
mode:
Diffstat (limited to 'security/nss/lib/freebl/intel-gcm-x86-masm.asm')
-rw-r--r--security/nss/lib/freebl/intel-gcm-x86-masm.asm1207
1 files changed, 1207 insertions, 0 deletions
diff --git a/security/nss/lib/freebl/intel-gcm-x86-masm.asm b/security/nss/lib/freebl/intel-gcm-x86-masm.asm
new file mode 100644
index 0000000000..32f4257884
--- /dev/null
+++ b/security/nss/lib/freebl/intel-gcm-x86-masm.asm
@@ -0,0 +1,1207 @@
+; LICENSE:
+; This submission to NSS is to be made available under the terms of the
+; Mozilla Public License, v. 2.0. You can obtain one at http:
+; //mozilla.org/MPL/2.0/.
+;###############################################################################
+; Copyright(c) 2014, Intel Corp.
+; Developers and authors:
+; Shay Gueron and Vlad Krasnov
+; Intel Corporation, Israel Development Centre, Haifa, Israel
+; Please send feedback directly to crypto.feedback.alias@intel.com
+
+
+.MODEL FLAT, C
+.XMM
+
+.DATA
+ALIGN 16
+Lone dq 1,0
+Ltwo dq 2,0
+Lbswap_mask db 15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
+Lshuff_mask dq 0f0f0f0f0f0f0f0fh, 0f0f0f0f0f0f0f0fh
+Lpoly dq 01h, 0c200000000000000h
+
+.CODE
+
+
+GFMUL MACRO DST, SRC1, SRC2, TMP1, TMP2, TMP3, TMP4
+ vpclmulqdq TMP1, SRC2, SRC1, 0h
+ vpclmulqdq TMP4, SRC2, SRC1, 011h
+
+ vpshufd TMP2, SRC2, 78
+ vpshufd TMP3, SRC1, 78
+ vpxor TMP2, TMP2, SRC2
+ vpxor TMP3, TMP3, SRC1
+
+ vpclmulqdq TMP2, TMP2, TMP3, 0h
+ vpxor TMP2, TMP2, TMP1
+ vpxor TMP2, TMP2, TMP4
+
+ vpslldq TMP3, TMP2, 8
+ vpsrldq TMP2, TMP2, 8
+
+ vpxor TMP1, TMP1, TMP3
+ vpxor TMP4, TMP4, TMP2
+
+ vpclmulqdq TMP2, TMP1, [Lpoly], 010h
+ vpshufd TMP3, TMP1, 78
+ vpxor TMP1, TMP2, TMP3
+
+ vpclmulqdq TMP2, TMP1, [Lpoly], 010h
+ vpshufd TMP3, TMP1, 78
+ vpxor TMP1, TMP2, TMP3
+
+ vpxor DST, TMP1, TMP4
+
+ ENDM
+
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+;
+; Generates the final GCM tag
+; void intel_aes_gcmTAG(unsigned char Htbl[16*16],
+; unsigned char *Tp,
+; unsigned int Mlen,
+; unsigned int Alen,
+; unsigned char* X0,
+; unsigned char* TAG);
+;
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+
+ALIGN 16
+intel_aes_gcmTAG PROC
+
+Htbl textequ <eax>
+Tp textequ <ecx>
+X0 textequ <edx>
+TAG textequ <ebx>
+
+T textequ <xmm0>
+TMP0 textequ <xmm1>
+
+ push ebx
+
+ mov Htbl, [esp + 2*4 + 0*4]
+ mov Tp, [esp + 2*4 + 1*4]
+ mov X0, [esp + 2*4 + 4*4]
+ mov TAG, [esp + 2*4 + 5*4]
+
+ vzeroupper
+ vmovdqu T, XMMWORD PTR[Tp]
+
+ vpxor TMP0, TMP0, TMP0
+ vpinsrd TMP0, TMP0, DWORD PTR[esp + 2*4 + 2*4], 0
+ vpinsrd TMP0, TMP0, DWORD PTR[esp + 2*4 + 3*4], 2
+ vpsllq TMP0, TMP0, 3
+
+ vpxor T, T, TMP0
+ vmovdqu TMP0, XMMWORD PTR[Htbl]
+ GFMUL T, T, TMP0, xmm2, xmm3, xmm4, xmm5
+
+ vpshufb T, T, [Lbswap_mask]
+ vpxor T, T, [X0]
+ vmovdqu XMMWORD PTR[TAG], T
+ vzeroupper
+
+ pop ebx
+
+ ret
+
+intel_aes_gcmTAG ENDP
+
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+;
+; Generates the H table
+; void intel_aes_gcmINIT(unsigned char Htbl[16*16], unsigned char *KS, int NR);
+;
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+
+ALIGN 16
+intel_aes_gcmINIT PROC
+
+Htbl textequ <eax>
+KS textequ <ecx>
+NR textequ <edx>
+
+T textequ <xmm0>
+TMP0 textequ <xmm1>
+
+ mov Htbl, [esp + 4*1 + 0*4]
+ mov KS, [esp + 4*1 + 1*4]
+ mov NR, [esp + 4*1 + 2*4]
+
+ vzeroupper
+ ; AES-ENC(0)
+ vmovdqu T, XMMWORD PTR[KS]
+ lea KS, [16 + KS]
+ dec NR
+Lenc_loop:
+ vaesenc T, T, [KS]
+ lea KS, [16 + KS]
+ dec NR
+ jnz Lenc_loop
+
+ vaesenclast T, T, [KS]
+ vpshufb T, T, [Lbswap_mask]
+
+ ;Calculate H` = GFMUL(H, 2)
+ vpsrad xmm3, T, 31
+ vpshufd xmm3, xmm3, 0ffh
+ vpand xmm5, xmm3, [Lpoly]
+ vpsrld xmm3, T, 31
+ vpslld xmm4, T, 1
+ vpslldq xmm3, xmm3, 4
+ vpxor T, xmm4, xmm3
+ vpxor T, T, xmm5
+
+ vmovdqu TMP0, T
+ vmovdqu XMMWORD PTR[Htbl + 0*16], T
+
+ vpshufd xmm2, T, 78
+ vpxor xmm2, xmm2, T
+ vmovdqu XMMWORD PTR[Htbl + 8*16 + 0*16], xmm2
+
+ i = 1
+ WHILE i LT 8
+ GFMUL T, T, TMP0, xmm2, xmm3, xmm4, xmm5
+ vmovdqu XMMWORD PTR[Htbl + i*16], T
+ vpshufd xmm2, T, 78
+ vpxor xmm2, xmm2, T
+ vmovdqu XMMWORD PTR[Htbl + 8*16 + i*16], xmm2
+ i = i+1
+ ENDM
+ vzeroupper
+ ret
+intel_aes_gcmINIT ENDP
+
+
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+;
+; Authenticate only
+; void intel_aes_gcmAAD(unsigned char Htbl[16*16], unsigned char *AAD, unsigned int Alen, unsigned char *Tp);
+;
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+
+ALIGN 16
+intel_aes_gcmAAD PROC
+
+Htbl textequ <eax>
+inp textequ <ecx>
+len textequ <edx>
+Tp textequ <ebx>
+hlp0 textequ <esi>
+
+DATA textequ <xmm0>
+T textequ <xmm1>
+TMP0 textequ <xmm2>
+TMP1 textequ <xmm3>
+TMP2 textequ <xmm4>
+TMP3 textequ <xmm5>
+TMP4 textequ <xmm6>
+Xhi textequ <xmm7>
+
+KARATSUBA_AAD MACRO i
+ vpclmulqdq TMP3, DATA, [Htbl + i*16], 0h
+ vpxor TMP0, TMP0, TMP3
+ vpclmulqdq TMP3, DATA, [Htbl + i*16], 011h
+ vpxor TMP1, TMP1, TMP3
+ vpshufd TMP3, DATA, 78
+ vpxor TMP3, TMP3, DATA
+ vpclmulqdq TMP3, TMP3, [Htbl + 8*16 + i*16], 0h
+ vpxor TMP2, TMP2, TMP3
+ENDM
+
+ cmp DWORD PTR[esp + 1*3 + 2*4], 0
+ jnz LbeginAAD
+ ret
+
+LbeginAAD:
+ push ebx
+ push esi
+
+ mov Htbl, [esp + 4*3 + 0*4]
+ mov inp, [esp + 4*3 + 1*4]
+ mov len, [esp + 4*3 + 2*4]
+ mov Tp, [esp + 4*3 + 3*4]
+
+ vzeroupper
+
+ vpxor Xhi, Xhi, Xhi
+
+ vmovdqu T, XMMWORD PTR[Tp]
+ ;we hash 8 block each iteration, if the total amount of blocks is not a multiple of 8, we hash the first n%8 blocks first
+ mov hlp0, len
+ and hlp0, 128-1
+ jz Lmod_loop
+
+ and len, -128
+ sub hlp0, 16
+
+ ; Prefix block
+ vmovdqu DATA, XMMWORD PTR[inp]
+ vpshufb DATA, DATA, [Lbswap_mask]
+ vpxor DATA, DATA, T
+
+ vpclmulqdq TMP0, DATA, XMMWORD PTR[Htbl + hlp0], 0h
+ vpclmulqdq TMP1, DATA, XMMWORD PTR[Htbl + hlp0], 011h
+ vpshufd TMP3, DATA, 78
+ vpxor TMP3, TMP3, DATA
+ vpclmulqdq TMP2, TMP3, XMMWORD PTR[Htbl + 8*16 + hlp0], 0h
+
+ lea inp, [inp+16]
+ test hlp0, hlp0
+ jnz Lpre_loop
+ jmp Lred1
+
+ ;hash remaining prefix bocks (up to 7 total prefix blocks)
+Lpre_loop:
+
+ sub hlp0, 16
+
+ vmovdqu DATA, XMMWORD PTR[inp]
+ vpshufb DATA, DATA, [Lbswap_mask]
+
+ vpclmulqdq TMP3, DATA, XMMWORD PTR[Htbl + hlp0], 0h
+ vpxor TMP0, TMP0, TMP3
+ vpclmulqdq TMP3, DATA, XMMWORD PTR[Htbl + hlp0], 011h
+ vpxor TMP1, TMP1, TMP3
+ vpshufd TMP3, DATA, 78
+ vpxor TMP3, TMP3, DATA
+ vpclmulqdq TMP3, TMP3, XMMWORD PTR[Htbl + 8*16 + hlp0], 0h
+ vpxor TMP2, TMP2, TMP3
+
+ test hlp0, hlp0
+ lea inp, [inp+16]
+ jnz Lpre_loop
+
+Lred1:
+
+ vpxor TMP2, TMP2, TMP0
+ vpxor TMP2, TMP2, TMP1
+ vpsrldq TMP3, TMP2, 8
+ vpslldq TMP2, TMP2, 8
+
+ vpxor Xhi, TMP1, TMP3
+ vpxor T, TMP0, TMP2
+
+Lmod_loop:
+
+ sub len, 16*8
+ jb Ldone
+ ; Block #0
+ vmovdqu DATA, XMMWORD PTR[inp + 16*7]
+ vpshufb DATA, DATA, XMMWORD PTR[Lbswap_mask]
+
+ vpclmulqdq TMP0, DATA, XMMWORD PTR[Htbl + 0*16], 0h
+ vpclmulqdq TMP1, DATA, XMMWORD PTR[Htbl + 0*16], 011h
+ vpshufd TMP3, DATA, 78
+ vpxor TMP3, TMP3, DATA
+ vpclmulqdq TMP2, TMP3, XMMWORD PTR[Htbl + 8*16 + 0*16], 0h
+
+ ; Block #1
+ vmovdqu DATA, XMMWORD PTR[inp + 16*6]
+ vpshufb DATA, DATA, [Lbswap_mask]
+ KARATSUBA_AAD 1
+
+ ; Block #2
+ vmovdqu DATA, XMMWORD PTR[inp + 16*5]
+ vpshufb DATA, DATA, [Lbswap_mask]
+
+ vpclmulqdq TMP4, T, [Lpoly], 010h ;reduction stage 1a
+ vpalignr T, T, T, 8
+
+ KARATSUBA_AAD 2
+
+ vpxor T, T, TMP4 ;reduction stage 1b
+
+ ; Block #3
+ vmovdqu DATA, XMMWORD PTR[inp + 16*4]
+ vpshufb DATA, DATA, [Lbswap_mask]
+ KARATSUBA_AAD 3
+ ; Block #4
+ vmovdqu DATA, XMMWORD PTR[inp + 16*3]
+ vpshufb DATA, DATA, [Lbswap_mask]
+
+ vpclmulqdq TMP4, T, [Lpoly], 010h ;reduction stage 2a
+ vpalignr T, T, T, 8
+
+ KARATSUBA_AAD 4
+
+ vpxor T, T, TMP4 ;reduction stage 2b
+ ; Block #5
+ vmovdqu DATA, XMMWORD PTR[inp + 16*2]
+ vpshufb DATA, DATA, [Lbswap_mask]
+ KARATSUBA_AAD 5
+
+ vpxor T, T, Xhi ;reduction finalize
+ ; Block #6
+ vmovdqu DATA, XMMWORD PTR[inp + 16*1]
+ vpshufb DATA, DATA, [Lbswap_mask]
+ KARATSUBA_AAD 6
+ ; Block #7
+ vmovdqu DATA, XMMWORD PTR[inp + 16*0]
+ vpshufb DATA, DATA, [Lbswap_mask]
+ vpxor DATA, DATA, T
+ KARATSUBA_AAD 7
+ ; Aggregated 8 blocks, now karatsuba fixup
+ vpxor TMP2, TMP2, TMP0
+ vpxor TMP2, TMP2, TMP1
+ vpsrldq TMP3, TMP2, 8
+ vpslldq TMP2, TMP2, 8
+
+ vpxor Xhi, TMP1, TMP3
+ vpxor T, TMP0, TMP2
+
+ lea inp, [inp + 16*8]
+ jmp Lmod_loop
+
+Ldone:
+ vpclmulqdq TMP4, T, [Lpoly], 010h
+ vpalignr T, T, T, 8
+ vpxor T, T, TMP4
+
+ vpclmulqdq TMP4, T, [Lpoly], 010h
+ vpalignr T, T, T, 8
+ vpxor T, T, TMP4
+
+ vpxor T, T, Xhi
+ vmovdqu XMMWORD PTR[Tp], T
+ vzeroupper
+
+ pop esi
+ pop ebx
+ ret
+
+intel_aes_gcmAAD ENDP
+
+
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+;
+; Encrypt and Authenticate
+; void intel_aes_gcmENC(unsigned char* PT, unsigned char* CT, void *Gctx, unsigned int len);
+;
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+
+ALIGN 16
+intel_aes_gcmENC PROC
+
+PT textequ <eax>
+CT textequ <ecx>
+Htbl textequ <edx>
+Gctx textequ <edx>
+len textequ <DWORD PTR[ebp + 5*4 + 3*4]>
+KS textequ <esi>
+NR textequ <DWORD PTR[244+KS]>
+
+aluCTR textequ <ebx>
+aluTMP textequ <edi>
+
+T textequ <XMMWORD PTR[16*16 + 1*16 + Gctx]>
+TMP0 textequ <xmm1>
+TMP1 textequ <xmm2>
+TMP2 textequ <xmm3>
+TMP3 textequ <xmm4>
+TMP4 textequ <xmm5>
+TMP5 textequ <xmm6>
+
+CTR0 textequ <xmm0>
+CTR1 textequ <xmm1>
+CTR2 textequ <xmm2>
+CTR3 textequ <xmm3>
+CTR4 textequ <xmm4>
+CTR5 textequ <xmm5>
+CTR6 textequ <xmm6>
+
+ROUND MACRO i
+ vmovdqu xmm7, XMMWORD PTR[i*16 + KS]
+ vaesenc CTR0, CTR0, xmm7
+ vaesenc CTR1, CTR1, xmm7
+ vaesenc CTR2, CTR2, xmm7
+ vaesenc CTR3, CTR3, xmm7
+ vaesenc CTR4, CTR4, xmm7
+ vaesenc CTR5, CTR5, xmm7
+ vaesenc CTR6, CTR6, xmm7
+ENDM
+
+KARATSUBA MACRO i
+ vpshufd TMP4, TMP5, 78
+ vpxor TMP4, TMP4, TMP5
+ vpclmulqdq TMP3, TMP4, XMMWORD PTR[i*16 + 8*16 + Htbl], 000h
+ vpxor TMP0, TMP0, TMP3
+ vmovdqu TMP4, XMMWORD PTR[i*16 + Htbl]
+ vpclmulqdq TMP3, TMP5, TMP4, 011h
+ vpxor TMP1, TMP1, TMP3
+ vpclmulqdq TMP3, TMP5, TMP4, 000h
+ vpxor TMP2, TMP2, TMP3
+ENDM
+
+NEXTCTR MACRO i
+ add aluCTR, 1
+ mov aluTMP, aluCTR
+ bswap aluTMP
+ xor aluTMP, [3*4 + KS]
+ mov [3*4 + 8*16 + i*16 + esp], aluTMP
+ENDM
+
+ cmp DWORD PTR[1*4 + 3*4 + esp], 0
+ jne LbeginENC
+ ret
+
+LbeginENC:
+
+ vzeroupper
+ push ebp
+ push ebx
+ push esi
+ push edi
+
+ mov ebp, esp
+ sub esp, 16*16
+ and esp, -16
+
+ mov PT, [ebp + 5*4 + 0*4]
+ mov CT, [ebp + 5*4 + 1*4]
+ mov Gctx, [ebp + 5*4 + 2*4]
+
+ mov KS, [16*16 + 3*16 + Gctx]
+
+ mov aluCTR, [16*16 + 2*16 + 3*4 + Gctx]
+ bswap aluCTR
+
+
+ vmovdqu TMP0, XMMWORD PTR[0*16 + KS]
+ vpxor TMP0, TMP0, XMMWORD PTR[16*16 + 2*16 + Gctx]
+ vmovdqu XMMWORD PTR[8*16 + 0*16 + esp], TMP0
+
+ cmp len, 16*7
+ jb LEncDataSingles
+; Prepare the "top" counters
+ vmovdqu XMMWORD PTR[8*16 + 1*16 + esp], TMP0
+ vmovdqu XMMWORD PTR[8*16 + 2*16 + esp], TMP0
+ vmovdqu XMMWORD PTR[8*16 + 3*16 + esp], TMP0
+ vmovdqu XMMWORD PTR[8*16 + 4*16 + esp], TMP0
+ vmovdqu XMMWORD PTR[8*16 + 5*16 + esp], TMP0
+ vmovdqu XMMWORD PTR[8*16 + 6*16 + esp], TMP0
+
+ vmovdqu CTR0, XMMWORD PTR[16*16 + 2*16 + Gctx]
+ vpshufb CTR0, CTR0, XMMWORD PTR[Lbswap_mask]
+; Encrypt the initial 7 blocks
+ sub len, 16*7
+ vpaddd CTR1, CTR0, XMMWORD PTR[Lone]
+ vpaddd CTR2, CTR0, XMMWORD PTR[Ltwo]
+ vpaddd CTR3, CTR2, XMMWORD PTR[Lone]
+ vpaddd CTR4, CTR2, XMMWORD PTR[Ltwo]
+ vpaddd CTR5, CTR4, XMMWORD PTR[Lone]
+ vpaddd CTR6, CTR4, XMMWORD PTR[Ltwo]
+
+ vpshufb CTR0, CTR0, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR1, CTR1, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR2, CTR2, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR3, CTR3, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR4, CTR4, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR5, CTR5, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR6, CTR6, XMMWORD PTR[Lbswap_mask]
+
+ vmovdqu xmm7, XMMWORD PTR[0*16 + KS]
+ vpxor CTR0, CTR0, xmm7
+ vpxor CTR1, CTR1, xmm7
+ vpxor CTR2, CTR2, xmm7
+ vpxor CTR3, CTR3, xmm7
+ vpxor CTR4, CTR4, xmm7
+ vpxor CTR5, CTR5, xmm7
+ vpxor CTR6, CTR6, xmm7
+
+ ROUND 1
+
+ add aluCTR, 7
+ mov aluTMP, aluCTR
+ bswap aluTMP
+ xor aluTMP, [KS + 3*4]
+ mov [8*16 + 0*16 + 3*4 + esp], aluTMP
+
+ ROUND 2
+ NEXTCTR 1
+ ROUND 3
+ NEXTCTR 2
+ ROUND 4
+ NEXTCTR 3
+ ROUND 5
+ NEXTCTR 4
+ ROUND 6
+ NEXTCTR 5
+ ROUND 7
+ NEXTCTR 6
+ ROUND 8
+ ROUND 9
+ vmovdqu xmm7, XMMWORD PTR[10*16 + KS]
+ cmp NR, 10
+ je @f
+
+ ROUND 10
+ ROUND 11
+ vmovdqu xmm7, XMMWORD PTR[12*16 + KS]
+ cmp NR, 12
+ je @f
+
+ ROUND 12
+ ROUND 13
+ vmovdqu xmm7, XMMWORD PTR[14*16 + KS]
+@@:
+ vaesenclast CTR0, CTR0, xmm7
+ vaesenclast CTR1, CTR1, xmm7
+ vaesenclast CTR2, CTR2, xmm7
+ vaesenclast CTR3, CTR3, xmm7
+ vaesenclast CTR4, CTR4, xmm7
+ vaesenclast CTR5, CTR5, xmm7
+ vaesenclast CTR6, CTR6, xmm7
+
+ vpxor CTR0, CTR0, XMMWORD PTR[0*16 + PT]
+ vpxor CTR1, CTR1, XMMWORD PTR[1*16 + PT]
+ vpxor CTR2, CTR2, XMMWORD PTR[2*16 + PT]
+ vpxor CTR3, CTR3, XMMWORD PTR[3*16 + PT]
+ vpxor CTR4, CTR4, XMMWORD PTR[4*16 + PT]
+ vpxor CTR5, CTR5, XMMWORD PTR[5*16 + PT]
+ vpxor CTR6, CTR6, XMMWORD PTR[6*16 + PT]
+
+ vmovdqu XMMWORD PTR[0*16 + CT], CTR0
+ vmovdqu XMMWORD PTR[1*16 + CT], CTR1
+ vmovdqu XMMWORD PTR[2*16 + CT], CTR2
+ vmovdqu XMMWORD PTR[3*16 + CT], CTR3
+ vmovdqu XMMWORD PTR[4*16 + CT], CTR4
+ vmovdqu XMMWORD PTR[5*16 + CT], CTR5
+ vmovdqu XMMWORD PTR[6*16 + CT], CTR6
+
+ vpshufb CTR0, CTR0, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR1, CTR1, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR2, CTR2, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR3, CTR3, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR4, CTR4, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR5, CTR5, XMMWORD PTR[Lbswap_mask]
+ vpshufb TMP5, CTR6, XMMWORD PTR[Lbswap_mask]
+
+ vmovdqa XMMWORD PTR[1*16 + esp], CTR5
+ vmovdqa XMMWORD PTR[2*16 + esp], CTR4
+ vmovdqa XMMWORD PTR[3*16 + esp], CTR3
+ vmovdqa XMMWORD PTR[4*16 + esp], CTR2
+ vmovdqa XMMWORD PTR[5*16 + esp], CTR1
+ vmovdqa XMMWORD PTR[6*16 + esp], CTR0
+
+ lea CT, [7*16 + CT]
+ lea PT, [7*16 + PT]
+ jmp LEncData7
+
+LEncData7:
+ cmp len, 16*7
+ jb LEndEnc7
+ sub len, 16*7
+
+ vpshufd TMP4, TMP5, 78
+ vpxor TMP4, TMP4, TMP5
+ vpclmulqdq TMP0, TMP4, XMMWORD PTR[0*16 + 8*16 + Htbl], 000h
+ vmovdqu TMP4, XMMWORD PTR[0*16 + Htbl]
+ vpclmulqdq TMP1, TMP5, TMP4, 011h
+ vpclmulqdq TMP2, TMP5, TMP4, 000h
+
+ vmovdqu TMP5, XMMWORD PTR[1*16 + esp]
+ KARATSUBA 1
+ vmovdqu TMP5, XMMWORD PTR[2*16 + esp]
+ KARATSUBA 2
+ vmovdqu TMP5, XMMWORD PTR[3*16 + esp]
+ KARATSUBA 3
+ vmovdqu TMP5, XMMWORD PTR[4*16 + esp]
+ KARATSUBA 4
+ vmovdqu TMP5, XMMWORD PTR[5*16 + esp]
+ KARATSUBA 5
+ vmovdqu TMP5, XMMWORD PTR[6*16 + esp]
+ vpxor TMP5, TMP5, T
+ KARATSUBA 6
+
+ vpxor TMP0, TMP0, TMP1
+ vpxor TMP0, TMP0, TMP2
+ vpsrldq TMP3, TMP0, 8
+ vpxor TMP4, TMP1, TMP3
+ vpslldq TMP3, TMP0, 8
+ vpxor TMP5, TMP2, TMP3
+
+ vpclmulqdq TMP1, TMP5, XMMWORD PTR[Lpoly], 010h
+ vpalignr TMP5,TMP5,TMP5,8
+ vpxor TMP5, TMP5, TMP1
+
+ vpclmulqdq TMP1, TMP5, XMMWORD PTR[Lpoly], 010h
+ vpalignr TMP5,TMP5,TMP5,8
+ vpxor TMP5, TMP5, TMP1
+
+ vpxor TMP5, TMP5, TMP4
+ vmovdqu T, TMP5
+
+ vmovdqa CTR0, XMMWORD PTR[8*16 + 0*16 + esp]
+ vmovdqa CTR1, XMMWORD PTR[8*16 + 1*16 + esp]
+ vmovdqa CTR2, XMMWORD PTR[8*16 + 2*16 + esp]
+ vmovdqa CTR3, XMMWORD PTR[8*16 + 3*16 + esp]
+ vmovdqa CTR4, XMMWORD PTR[8*16 + 4*16 + esp]
+ vmovdqa CTR5, XMMWORD PTR[8*16 + 5*16 + esp]
+ vmovdqa CTR6, XMMWORD PTR[8*16 + 6*16 + esp]
+
+ ROUND 1
+ NEXTCTR 0
+ ROUND 2
+ NEXTCTR 1
+ ROUND 3
+ NEXTCTR 2
+ ROUND 4
+ NEXTCTR 3
+ ROUND 5
+ NEXTCTR 4
+ ROUND 6
+ NEXTCTR 5
+ ROUND 7
+ NEXTCTR 6
+
+ ROUND 8
+ ROUND 9
+
+ vmovdqu xmm7, XMMWORD PTR[10*16 + KS]
+ cmp NR, 10
+ je @f
+
+ ROUND 10
+ ROUND 11
+ vmovdqu xmm7, XMMWORD PTR[12*16 + KS]
+ cmp NR, 12
+ je @f
+
+ ROUND 12
+ ROUND 13
+ vmovdqu xmm7, XMMWORD PTR[14*16 + KS]
+@@:
+ vaesenclast CTR0, CTR0, xmm7
+ vaesenclast CTR1, CTR1, xmm7
+ vaesenclast CTR2, CTR2, xmm7
+ vaesenclast CTR3, CTR3, xmm7
+ vaesenclast CTR4, CTR4, xmm7
+ vaesenclast CTR5, CTR5, xmm7
+ vaesenclast CTR6, CTR6, xmm7
+
+ vpxor CTR0, CTR0, XMMWORD PTR[0*16 + PT]
+ vpxor CTR1, CTR1, XMMWORD PTR[1*16 + PT]
+ vpxor CTR2, CTR2, XMMWORD PTR[2*16 + PT]
+ vpxor CTR3, CTR3, XMMWORD PTR[3*16 + PT]
+ vpxor CTR4, CTR4, XMMWORD PTR[4*16 + PT]
+ vpxor CTR5, CTR5, XMMWORD PTR[5*16 + PT]
+ vpxor CTR6, CTR6, XMMWORD PTR[6*16 + PT]
+
+ vmovdqu XMMWORD PTR[0*16 + CT], CTR0
+ vmovdqu XMMWORD PTR[1*16 + CT], CTR1
+ vmovdqu XMMWORD PTR[2*16 + CT], CTR2
+ vmovdqu XMMWORD PTR[3*16 + CT], CTR3
+ vmovdqu XMMWORD PTR[4*16 + CT], CTR4
+ vmovdqu XMMWORD PTR[5*16 + CT], CTR5
+ vmovdqu XMMWORD PTR[6*16 + CT], CTR6
+
+ vpshufb CTR0, CTR0, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR1, CTR1, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR2, CTR2, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR3, CTR3, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR4, CTR4, XMMWORD PTR[Lbswap_mask]
+ vpshufb CTR5, CTR5, XMMWORD PTR[Lbswap_mask]
+ vpshufb TMP5, CTR6, XMMWORD PTR[Lbswap_mask]
+
+ vmovdqa XMMWORD PTR[1*16 + esp], CTR5
+ vmovdqa XMMWORD PTR[2*16 + esp], CTR4
+ vmovdqa XMMWORD PTR[3*16 + esp], CTR3
+ vmovdqa XMMWORD PTR[4*16 + esp], CTR2
+ vmovdqa XMMWORD PTR[5*16 + esp], CTR1
+ vmovdqa XMMWORD PTR[6*16 + esp], CTR0
+
+ lea CT, [7*16 + CT]
+ lea PT, [7*16 + PT]
+ jmp LEncData7
+
+LEndEnc7:
+
+ vpshufd TMP4, TMP5, 78
+ vpxor TMP4, TMP4, TMP5
+ vpclmulqdq TMP0, TMP4, XMMWORD PTR[0*16 + 8*16 + Htbl], 000h
+ vmovdqu TMP4, XMMWORD PTR[0*16 + Htbl]
+ vpclmulqdq TMP1, TMP5, TMP4, 011h
+ vpclmulqdq TMP2, TMP5, TMP4, 000h
+
+ vmovdqu TMP5, XMMWORD PTR[1*16 + esp]
+ KARATSUBA 1
+ vmovdqu TMP5, XMMWORD PTR[2*16 + esp]
+ KARATSUBA 2
+ vmovdqu TMP5, XMMWORD PTR[3*16 + esp]
+ KARATSUBA 3
+ vmovdqu TMP5, XMMWORD PTR[4*16 + esp]
+ KARATSUBA 4
+ vmovdqu TMP5, XMMWORD PTR[5*16 + esp]
+ KARATSUBA 5
+ vmovdqu TMP5, XMMWORD PTR[6*16 + esp]
+ vpxor TMP5, TMP5, T
+ KARATSUBA 6
+
+ vpxor TMP0, TMP0, TMP1
+ vpxor TMP0, TMP0, TMP2
+ vpsrldq TMP3, TMP0, 8
+ vpxor TMP4, TMP1, TMP3
+ vpslldq TMP3, TMP0, 8
+ vpxor TMP5, TMP2, TMP3
+
+ vpclmulqdq TMP1, TMP5, XMMWORD PTR[Lpoly], 010h
+ vpalignr TMP5,TMP5,TMP5,8
+ vpxor TMP5, TMP5, TMP1
+
+ vpclmulqdq TMP1, TMP5, XMMWORD PTR[Lpoly], 010h
+ vpalignr TMP5,TMP5,TMP5,8
+ vpxor TMP5, TMP5, TMP1
+
+ vpxor TMP5, TMP5, TMP4
+ vmovdqu T, TMP5
+
+ sub aluCTR, 6
+
+LEncDataSingles:
+
+ cmp len, 16
+ jb LEncDataTail
+ sub len, 16
+
+ vmovdqa TMP1, XMMWORD PTR[8*16 + 0*16 + esp]
+ NEXTCTR 0
+
+ vaesenc TMP1, TMP1, XMMWORD PTR[1*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[2*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[3*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[4*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[5*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[6*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[7*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[8*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[9*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[10*16 + KS]
+ cmp NR, 10
+ je @f
+ vaesenc TMP1, TMP1, XMMWORD PTR[10*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[11*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[12*16 + KS]
+ cmp NR, 12
+ je @f
+ vaesenc TMP1, TMP1, XMMWORD PTR[12*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[13*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[14*16 + KS]
+@@:
+ vaesenclast TMP1, TMP1, TMP2
+ vpxor TMP1, TMP1, XMMWORD PTR[PT]
+ vmovdqu XMMWORD PTR[CT], TMP1
+
+ lea PT, [16+PT]
+ lea CT, [16+CT]
+
+ vpshufb TMP1, TMP1, XMMWORD PTR[Lbswap_mask]
+ vpxor TMP1, TMP1, T
+
+ vmovdqu TMP0, XMMWORD PTR[Htbl]
+ GFMUL TMP1, TMP1, TMP0, TMP5, TMP2, TMP3, TMP4
+ vmovdqu T, TMP1
+
+ jmp LEncDataSingles
+
+LEncDataTail:
+
+ cmp len, 0
+ je LEncDataEnd
+
+ vmovdqa TMP1, XMMWORD PTR[8*16 + 0*16 + esp]
+
+ vaesenc TMP1, TMP1, XMMWORD PTR[1*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[2*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[3*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[4*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[5*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[6*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[7*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[8*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[9*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[10*16 + KS]
+ cmp NR, 10
+ je @f
+ vaesenc TMP1, TMP1, XMMWORD PTR[10*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[11*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[12*16 + KS]
+ cmp NR, 12
+ je @f
+ vaesenc TMP1, TMP1, XMMWORD PTR[12*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[13*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[14*16 + KS]
+@@:
+ vaesenclast TMP1, TMP1, TMP2
+; zero a temp location
+ vpxor TMP2, TMP2, TMP2
+ vmovdqa XMMWORD PTR[esp], TMP2
+; copy as many bytes as needed
+ xor KS, KS
+ mov aluTMP, edx
+@@:
+ cmp len, KS
+ je @f
+ mov dl, BYTE PTR[PT + KS]
+ mov BYTE PTR[esp + KS], dl
+ inc KS
+ jmp @b
+@@:
+ vpxor TMP1, TMP1, XMMWORD PTR[esp]
+ vmovdqa XMMWORD PTR[esp], TMP1
+ xor KS, KS
+@@:
+ cmp len, KS
+ je @f
+ mov dl, BYTE PTR[esp + KS]
+ mov BYTE PTR[CT + KS], dl
+ inc KS
+ jmp @b
+@@:
+ cmp KS, 16
+ je @f
+ mov BYTE PTR[esp + KS], 0
+ inc KS
+ jmp @b
+@@:
+ mov edx, aluTMP
+ vmovdqa TMP1, XMMWORD PTR[esp]
+ vpshufb TMP1, TMP1, XMMWORD PTR[Lbswap_mask]
+ vpxor TMP1, TMP1, T
+
+ vmovdqu TMP0, XMMWORD PTR[Htbl]
+ GFMUL TMP1, TMP1, TMP0, TMP5, TMP2, TMP3, TMP4
+ vmovdqu T, TMP1
+
+LEncDataEnd:
+ inc aluCTR
+ bswap aluCTR
+ mov [16*16 + 2*16 + 3*4 + Gctx], aluCTR
+
+ mov esp, ebp
+ pop edi
+ pop esi
+ pop ebx
+ pop ebp
+
+
+ vzeroupper
+
+ ret
+intel_aes_gcmENC ENDP
+
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+;
+; Decrypt and Authenticate
+; void intel_aes_gcmDEC(uint8_t* PT, uint8_t* CT, void *Gctx, unsigned int len);
+;
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+
+
+NEXTCTR MACRO i
+ add aluCTR, 1
+ mov aluTMP, aluCTR
+ bswap aluTMP
+ xor aluTMP, [3*4 + KS]
+ mov [3*4 + i*16 + esp], aluTMP
+ENDM
+
+intel_aes_gcmDEC PROC
+
+ cmp DWORD PTR[1*4 + 3*4 + esp], 0
+ jne LbeginDEC
+ ret
+
+LbeginDEC:
+
+ vzeroupper
+ push ebp
+ push ebx
+ push esi
+ push edi
+
+ mov ebp, esp
+ sub esp, 8*16
+ and esp, -16
+
+ mov CT, [ebp + 5*4 + 0*4]
+ mov PT, [ebp + 5*4 + 1*4]
+ mov Gctx, [ebp + 5*4 + 2*4]
+
+ mov KS, [16*16 + 3*16 + Gctx]
+
+ mov aluCTR, [16*16 + 2*16 + 3*4 + Gctx]
+ bswap aluCTR
+
+
+ vmovdqu TMP0, XMMWORD PTR[0*16 + KS]
+ vpxor TMP0, TMP0, XMMWORD PTR[16*16 + 2*16 + Gctx]
+ vmovdqu XMMWORD PTR[0*16 + esp], TMP0
+
+ cmp len, 16*7
+ jb LDecDataSingles
+ vmovdqu XMMWORD PTR[1*16 + esp], TMP0
+ vmovdqu XMMWORD PTR[2*16 + esp], TMP0
+ vmovdqu XMMWORD PTR[3*16 + esp], TMP0
+ vmovdqu XMMWORD PTR[4*16 + esp], TMP0
+ vmovdqu XMMWORD PTR[5*16 + esp], TMP0
+ vmovdqu XMMWORD PTR[6*16 + esp], TMP0
+ dec aluCTR
+
+LDecData7:
+ cmp len, 16*7
+ jb LDecData7End
+ sub len, 16*7
+
+ vmovdqu TMP5, XMMWORD PTR[0*16 + CT]
+ vpshufb TMP5, TMP5, XMMWORD PTR[Lbswap_mask]
+ vpxor TMP5, TMP5, T
+ vpshufd TMP4, TMP5, 78
+ vpxor TMP4, TMP4, TMP5
+ vpclmulqdq TMP0, TMP4, XMMWORD PTR[6*16 + 8*16 + Htbl], 000h
+ vmovdqu TMP4, XMMWORD PTR[6*16 + Htbl]
+ vpclmulqdq TMP1, TMP5, TMP4, 011h
+ vpclmulqdq TMP2, TMP5, TMP4, 000h
+
+ NEXTCTR 0
+ vmovdqu TMP5, XMMWORD PTR[1*16 + CT]
+ vpshufb TMP5, TMP5, XMMWORD PTR[Lbswap_mask]
+ KARATSUBA 5
+ NEXTCTR 1
+ vmovdqu TMP5, XMMWORD PTR[2*16 + CT]
+ vpshufb TMP5, TMP5, XMMWORD PTR[Lbswap_mask]
+ KARATSUBA 4
+ NEXTCTR 2
+ vmovdqu TMP5, XMMWORD PTR[3*16 + CT]
+ vpshufb TMP5, TMP5, XMMWORD PTR[Lbswap_mask]
+ KARATSUBA 3
+ NEXTCTR 3
+ vmovdqu TMP5, XMMWORD PTR[4*16 + CT]
+ vpshufb TMP5, TMP5, XMMWORD PTR[Lbswap_mask]
+ KARATSUBA 2
+ NEXTCTR 4
+ vmovdqu TMP5, XMMWORD PTR[5*16 + CT]
+ vpshufb TMP5, TMP5, XMMWORD PTR[Lbswap_mask]
+ KARATSUBA 1
+ NEXTCTR 5
+ vmovdqu TMP5, XMMWORD PTR[6*16 + CT]
+ vpshufb TMP5, TMP5, XMMWORD PTR[Lbswap_mask]
+ KARATSUBA 0
+ NEXTCTR 6
+
+ vpxor TMP0, TMP0, TMP1
+ vpxor TMP0, TMP0, TMP2
+ vpsrldq TMP3, TMP0, 8
+ vpxor TMP4, TMP1, TMP3
+ vpslldq TMP3, TMP0, 8
+ vpxor TMP5, TMP2, TMP3
+
+ vpclmulqdq TMP1, TMP5, XMMWORD PTR[Lpoly], 010h
+ vpalignr TMP5,TMP5,TMP5,8
+ vpxor TMP5, TMP5, TMP1
+
+ vpclmulqdq TMP1, TMP5, XMMWORD PTR[Lpoly], 010h
+ vpalignr TMP5,TMP5,TMP5,8
+ vpxor TMP5, TMP5, TMP1
+
+ vpxor TMP5, TMP5, TMP4
+ vmovdqu T, TMP5
+
+ vmovdqa CTR0, XMMWORD PTR[0*16 + esp]
+ vmovdqa CTR1, XMMWORD PTR[1*16 + esp]
+ vmovdqa CTR2, XMMWORD PTR[2*16 + esp]
+ vmovdqa CTR3, XMMWORD PTR[3*16 + esp]
+ vmovdqa CTR4, XMMWORD PTR[4*16 + esp]
+ vmovdqa CTR5, XMMWORD PTR[5*16 + esp]
+ vmovdqa CTR6, XMMWORD PTR[6*16 + esp]
+
+ ROUND 1
+ ROUND 2
+ ROUND 3
+ ROUND 4
+ ROUND 5
+ ROUND 6
+ ROUND 7
+ ROUND 8
+ ROUND 9
+ vmovdqu xmm7, XMMWORD PTR[10*16 + KS]
+ cmp NR, 10
+ je @f
+
+ ROUND 10
+ ROUND 11
+ vmovdqu xmm7, XMMWORD PTR[12*16 + KS]
+ cmp NR, 12
+ je @f
+
+ ROUND 12
+ ROUND 13
+ vmovdqu xmm7, XMMWORD PTR[14*16 + KS]
+@@:
+ vaesenclast CTR0, CTR0, xmm7
+ vaesenclast CTR1, CTR1, xmm7
+ vaesenclast CTR2, CTR2, xmm7
+ vaesenclast CTR3, CTR3, xmm7
+ vaesenclast CTR4, CTR4, xmm7
+ vaesenclast CTR5, CTR5, xmm7
+ vaesenclast CTR6, CTR6, xmm7
+
+ vpxor CTR0, CTR0, XMMWORD PTR[0*16 + CT]
+ vpxor CTR1, CTR1, XMMWORD PTR[1*16 + CT]
+ vpxor CTR2, CTR2, XMMWORD PTR[2*16 + CT]
+ vpxor CTR3, CTR3, XMMWORD PTR[3*16 + CT]
+ vpxor CTR4, CTR4, XMMWORD PTR[4*16 + CT]
+ vpxor CTR5, CTR5, XMMWORD PTR[5*16 + CT]
+ vpxor CTR6, CTR6, XMMWORD PTR[6*16 + CT]
+
+ vmovdqu XMMWORD PTR[0*16 + PT], CTR0
+ vmovdqu XMMWORD PTR[1*16 + PT], CTR1
+ vmovdqu XMMWORD PTR[2*16 + PT], CTR2
+ vmovdqu XMMWORD PTR[3*16 + PT], CTR3
+ vmovdqu XMMWORD PTR[4*16 + PT], CTR4
+ vmovdqu XMMWORD PTR[5*16 + PT], CTR5
+ vmovdqu XMMWORD PTR[6*16 + PT], CTR6
+
+ lea CT, [7*16 + CT]
+ lea PT, [7*16 + PT]
+ jmp LDecData7
+
+LDecData7End:
+
+ NEXTCTR 0
+
+LDecDataSingles:
+
+ cmp len, 16
+ jb LDecDataTail
+ sub len, 16
+
+ vmovdqu TMP1, XMMWORD PTR[CT]
+ vpshufb TMP1, TMP1, XMMWORD PTR[Lbswap_mask]
+ vpxor TMP1, TMP1, T
+
+ vmovdqu TMP0, XMMWORD PTR[Htbl]
+ GFMUL TMP1, TMP1, TMP0, TMP5, TMP2, TMP3, TMP4
+ vmovdqu T, TMP1
+
+ vmovdqa TMP1, XMMWORD PTR[0*16 + esp]
+ NEXTCTR 0
+
+ vaesenc TMP1, TMP1, XMMWORD PTR[1*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[2*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[3*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[4*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[5*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[6*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[7*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[8*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[9*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[10*16 + KS]
+ cmp NR, 10
+ je @f
+ vaesenc TMP1, TMP1, XMMWORD PTR[10*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[11*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[12*16 + KS]
+ cmp NR, 12
+ je @f
+ vaesenc TMP1, TMP1, XMMWORD PTR[12*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[13*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[14*16 + KS]
+@@:
+ vaesenclast TMP1, TMP1, TMP2
+ vpxor TMP1, TMP1, XMMWORD PTR[CT]
+ vmovdqu XMMWORD PTR[PT], TMP1
+
+ lea PT, [16+PT]
+ lea CT, [16+CT]
+ jmp LDecDataSingles
+
+LDecDataTail:
+
+ cmp len, 0
+ je LDecDataEnd
+
+ vmovdqa TMP1, XMMWORD PTR[0*16 + esp]
+ inc aluCTR
+ vaesenc TMP1, TMP1, XMMWORD PTR[1*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[2*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[3*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[4*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[5*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[6*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[7*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[8*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[9*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[10*16 + KS]
+ cmp NR, 10
+ je @f
+ vaesenc TMP1, TMP1, XMMWORD PTR[10*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[11*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[12*16 + KS]
+ cmp NR, 12
+ je @f
+ vaesenc TMP1, TMP1, XMMWORD PTR[12*16 + KS]
+ vaesenc TMP1, TMP1, XMMWORD PTR[13*16 + KS]
+ vmovdqu TMP2, XMMWORD PTR[14*16 + KS]
+@@:
+ vaesenclast xmm7, TMP1, TMP2
+
+; copy as many bytes as needed
+ xor KS, KS
+ mov aluTMP, edx
+@@:
+ cmp len, KS
+ je @f
+ mov dl, BYTE PTR[CT + KS]
+ mov BYTE PTR[esp + KS], dl
+ inc KS
+ jmp @b
+@@:
+ cmp KS, 16
+ je @f
+ mov BYTE PTR[esp + KS], 0
+ inc KS
+ jmp @b
+@@:
+ mov edx, aluTMP
+ vmovdqa TMP1, XMMWORD PTR[esp]
+ vpshufb TMP1, TMP1, XMMWORD PTR[Lbswap_mask]
+ vpxor TMP1, TMP1, T
+
+ vmovdqu TMP0, XMMWORD PTR[Htbl]
+ GFMUL TMP1, TMP1, TMP0, TMP5, TMP2, TMP3, TMP4
+ vmovdqu T, TMP1
+
+ vpxor xmm7, xmm7, XMMWORD PTR[esp]
+ vmovdqa XMMWORD PTR[esp], xmm7
+ xor KS, KS
+ mov aluTMP, edx
+@@:
+ cmp len, KS
+ je @f
+ mov dl, BYTE PTR[esp + KS]
+ mov BYTE PTR[PT + KS], dl
+ inc KS
+ jmp @b
+@@:
+ mov edx, aluTMP
+
+LDecDataEnd:
+
+ bswap aluCTR
+ mov [16*16 + 2*16 + 3*4 + Gctx], aluCTR
+
+ mov esp, ebp
+ pop edi
+ pop esi
+ pop ebx
+ pop ebp
+
+ vzeroupper
+
+ ret
+intel_aes_gcmDEC ENDP
+
+
+END