sleeping_kangaroo12 0.0.3 → 0.0.4

data/ext/{xkcp/lib/low/KeccakP-1600/AVX2 → k12/lib/Optimized64}/KeccakP-1600-AVX2.s

@@ -1,6 +1,3 @@
-# The eXtended Keccak Code Package (XKCP)
-# https://github.com/XKCP/XKCP
-#
 # Copyright (c) 2006-2017, CRYPTOGAMS by <appro@openssl.org>
 # Copyright (c) 2017 Ronny Van Keer
 # All rights reserved.
@@ -14,22 +11,22 @@
 # (https://github.com/dot-asm/cryptogams/blob/master/x86_64/keccak1600-avx2.pl).
 # The rest of the code was written by Ronny Van Keer.
 # Adaptations for macOS by Stéphane Léon.
-# Adaptations for mingw-w64 (changes macOS too) by Jorrit Jongma.
 
 .text
 
 # -----------------------------------------------------------------------------
 #
-# void KeccakP1600_Initialize(void *state);
+# void KeccakP1600_AVX2_Initialize(void *state);
 #
-.globl  KeccakP1600_Initialize
-.globl _KeccakP1600_Initialize
-.ifndef old_gas_syntax
-.type   KeccakP1600_Initialize,@function
+.ifdef macOS
+.globl  _KeccakP1600_AVX2_Initialize
+_KeccakP1600_AVX2_Initialize:
+.else
+.globl  KeccakP1600_AVX2_Initialize
+.type   KeccakP1600_AVX2_Initialize,@function
+KeccakP1600_AVX2_Initialize:
 .endif
-KeccakP1600_Initialize:
-_KeccakP1600_Initialize:
-.balign 32
+.balign  32
     vpxor       %ymm0,%ymm0,%ymm0
     vmovdqu     %ymm0,0*32(%rdi)
     vmovdqu     %ymm0,1*32(%rdi)
@@ -39,22 +36,24 @@ _KeccakP1600_Initialize:
     vmovdqu     %ymm0,5*32(%rdi)
     movq        $0,6*32(%rdi)
     ret
-.ifndef old_gas_syntax
-.size   KeccakP1600_Initialize,.-KeccakP1600_Initialize
+.ifdef macOS
+.else
+.size   KeccakP1600_AVX2_Initialize,.-KeccakP1600_AVX2_Initialize
 .endif
 
 # -----------------------------------------------------------------------------
 #
-# void KeccakP1600_AddByte(void *state, unsigned char data, unsigned int offset);
+# void KeccakP1600_AVX2_AddByte(void *state, unsigned char data, unsigned int offset);
 #                                %rdi                 %rsi               %rdx
 #
-.globl  KeccakP1600_AddByte
-.globl _KeccakP1600_AddByte
-.ifndef old_gas_syntax
-.type   KeccakP1600_AddByte,@function
+.ifdef macOS
+.globl  _KeccakP1600_AVX2_AddByte
+_KeccakP1600_AVX2_AddByte:
+.else
+.globl  KeccakP1600_AVX2_AddByte
+.type   KeccakP1600_AVX2_AddByte,@function
+KeccakP1600_AVX2_AddByte:
 .endif
-KeccakP1600_AddByte:
-_KeccakP1600_AddByte:
 .balign 32
     mov         %rdx, %rax
     and         $7, %rax
@@ -65,31 +64,33 @@ _KeccakP1600_AddByte:
     add         %rax, %rdi
     xorb        %sil, (%rdi)
     ret
-.ifndef old_gas_syntax
-.size   KeccakP1600_AddByte,.-KeccakP1600_AddByte
+.ifdef macOS
+.else
+.size   KeccakP1600_AVX2_AddByte,.-KeccakP1600_AVX2_AddByte
 .endif
 
 # -----------------------------------------------------------------------------
 #
-# void KeccakP1600_AddBytes(void *state, const unsigned char *data, unsigned int offset, unsigned int length);
+# void KeccakP1600_AVX2_AddBytes(void *state, const unsigned char *data, unsigned int offset, unsigned int length);
 #                                %rdi                         %rsi               %rdx                 %rcx
 #
-.globl  KeccakP1600_AddBytes
-.globl _KeccakP1600_AddBytes
-.ifndef old_gas_syntax
-.type   KeccakP1600_AddBytes,@function
+.ifdef macOS
+.globl  _KeccakP1600_AVX2_AddBytes
+_KeccakP1600_AVX2_AddBytes:
+.else
+.globl  KeccakP1600_AVX2_AddBytes
+.type   KeccakP1600_AVX2_AddBytes,@function
+KeccakP1600_AVX2_AddBytes:
 .endif
-KeccakP1600_AddBytes:
-_KeccakP1600_AddBytes:
 .balign 32
     cmp         $0, %rcx
-    jz          KeccakP1600_AddBytes_Exit
+    jz          KeccakP1600_AVX2_AddBytes_Exit
     mov         %rdx, %rax                              # rax offset in lane
     and         $0xFFFFFFF8, %edx                       # rdx pointer into state index mapper
     lea         mapState(%rip), %r9
     add         %r9, %rdx
     and         $7, %rax
-    jz          KeccakP1600_AddBytes_LaneAlignedCheck
+    jz          KeccakP1600_AVX2_AddBytes_LaneAlignedCheck
     mov         $8, %r9                                 # r9 is (max) length of incomplete lane
     sub         %rax, %r9
     cmp         %rcx, %r9
@@ -98,170 +99,66 @@ _KeccakP1600_AddBytes:
     add         (%rdx), %rax                            # rax = pointer to state lane
     add         $8, %rdx
     add         %rdi, %rax
-KeccakP1600_AddBytes_NotAlignedLoop:
+KeccakP1600_AVX2_AddBytes_NotAlignedLoop:
     mov         (%rsi), %r8b
     inc         %rsi
     xorb        %r8b, (%rax)
     inc         %rax
     dec         %r9
-    jnz         KeccakP1600_AddBytes_NotAlignedLoop
-    jmp         KeccakP1600_AddBytes_LaneAlignedCheck
-KeccakP1600_AddBytes_LaneAlignedLoop:
+    jnz         KeccakP1600_AVX2_AddBytes_NotAlignedLoop
+    jmp         KeccakP1600_AVX2_AddBytes_LaneAlignedCheck
+KeccakP1600_AVX2_AddBytes_LaneAlignedLoop:
     mov         (%rsi), %r8
     add         $8, %rsi
     mov         (%rdx), %rax
     add         $8, %rdx
     add         %rdi, %rax
     xor         %r8, (%rax)
-KeccakP1600_AddBytes_LaneAlignedCheck:
+KeccakP1600_AVX2_AddBytes_LaneAlignedCheck:
     sub         $8, %rcx
-    jnc         KeccakP1600_AddBytes_LaneAlignedLoop
-KeccakP1600_AddBytes_LastIncompleteLane:
+    jnc         KeccakP1600_AVX2_AddBytes_LaneAlignedLoop
+KeccakP1600_AVX2_AddBytes_LastIncompleteLane:
     add         $8, %rcx
-    jz          KeccakP1600_AddBytes_Exit
+    jz          KeccakP1600_AVX2_AddBytes_Exit
     mov         (%rdx), %rax
     add         %rdi, %rax
-KeccakP1600_AddBytes_LastIncompleteLaneLoop:
+KeccakP1600_AVX2_AddBytes_LastIncompleteLaneLoop:
     mov         (%rsi), %r8b
     inc         %rsi
     xor         %r8b, (%rax)
     inc         %rax
     dec         %rcx
-    jnz         KeccakP1600_AddBytes_LastIncompleteLaneLoop
-KeccakP1600_AddBytes_Exit:
+    jnz         KeccakP1600_AVX2_AddBytes_LastIncompleteLaneLoop
+KeccakP1600_AVX2_AddBytes_Exit:
     ret
-.ifndef old_gas_syntax
-.size   KeccakP1600_AddBytes,.-KeccakP1600_AddBytes
-.endif
-
-# -----------------------------------------------------------------------------
-#
-# void KeccakP1600_OverwriteBytes(void *state, const unsigned char *data, unsigned int offset, unsigned int length);
-#                                       %rdi                        %rsi               %rdx                 %rcx
-#
-.globl  KeccakP1600_OverwriteBytes
-.globl _KeccakP1600_OverwriteBytes
-.ifndef old_gas_syntax
-.type   KeccakP1600_OverwriteBytes,@function
-.endif
-KeccakP1600_OverwriteBytes:
-_KeccakP1600_OverwriteBytes:
-.balign 32
-    cmp         $0, %rcx
-    jz          KeccakP1600_OverwriteBytes_Exit
-    mov         %rdx, %rax                              # rax offset in lane
-    and         $0xFFFFFFF8, %edx                       # rdx pointer into state index mapper
-    lea         mapState(%rip), %r9
-    add         %r9, %rdx
-    and         $7, %rax
-    jz          KeccakP1600_OverwriteBytes_LaneAlignedCheck
-    mov         $8, %r9                                 # r9 is (max) length of incomplete lane
-    sub         %rax, %r9
-    cmp         %rcx, %r9
-    cmovae      %rcx, %r9
-    sub         %r9, %rcx                               # length -= length of incomplete lane
-    add         (%rdx), %rax                            # rax = pointer to state lane
-    add         $8, %rdx
-    add         %rdi, %rax
-KeccakP1600_OverwriteBytes_NotAlignedLoop:
-    mov         (%rsi), %r8b
-    inc         %rsi
-    mov         %r8b, (%rax)
-    inc         %rax
-    dec         %r9
-    jnz         KeccakP1600_OverwriteBytes_NotAlignedLoop
-    jmp         KeccakP1600_OverwriteBytes_LaneAlignedCheck
-KeccakP1600_OverwriteBytes_LaneAlignedLoop:
-    mov         (%rsi), %r8
-    add         $8, %rsi
-    mov         (%rdx), %rax
-    add         $8, %rdx
-    add         %rdi, %rax
-    mov         %r8, (%rax)
-KeccakP1600_OverwriteBytes_LaneAlignedCheck:
-    sub         $8, %rcx
-    jnc         KeccakP1600_OverwriteBytes_LaneAlignedLoop
-KeccakP1600_OverwriteBytes_LastIncompleteLane:
-    add         $8, %rcx
-    jz          KeccakP1600_OverwriteBytes_Exit
-    mov         (%rdx), %rax
-    add         %rdi, %rax
-KeccakP1600_OverwriteBytes_LastIncompleteLaneLoop:
-    mov         (%rsi), %r8b
-    inc         %rsi
-    mov         %r8b, (%rax)
-    inc         %rax
-    dec         %rcx
-    jnz         KeccakP1600_OverwriteBytes_LastIncompleteLaneLoop
-KeccakP1600_OverwriteBytes_Exit:
-    ret
-.ifndef old_gas_syntax
-.size   KeccakP1600_OverwriteBytes,.-KeccakP1600_OverwriteBytes
-.endif
-
-# -----------------------------------------------------------------------------
-#
-# void KeccakP1600_OverwriteWithZeroes(void *state, unsigned int byteCount);
-#                                            %rdi                %rsi
-#
-.globl  KeccakP1600_OverwriteWithZeroes
-.globl _KeccakP1600_OverwriteWithZeroes
-.ifndef old_gas_syntax
-.type   KeccakP1600_OverwriteWithZeroes,@function
-.endif
-KeccakP1600_OverwriteWithZeroes:
-_KeccakP1600_OverwriteWithZeroes:
-.balign 32
-    cmp         $0, %rsi
-    jz          KeccakP1600_OverwriteWithZeroes_Exit
-    lea         mapState(%rip), %rdx                          # rdx pointer into state index mapper
-    jmp         KeccakP1600_OverwriteWithZeroes_LaneAlignedCheck
-KeccakP1600_OverwriteWithZeroes_LaneAlignedLoop:
-    mov         (%rdx), %rax
-    add         $8, %rdx
-    add         %rdi, %rax
-    movq        $0, (%rax)
-KeccakP1600_OverwriteWithZeroes_LaneAlignedCheck:
-    sub         $8, %rsi
-    jnc         KeccakP1600_OverwriteWithZeroes_LaneAlignedLoop
-KeccakP1600_OverwriteWithZeroes_LastIncompleteLane:
-    add         $8, %rsi
-    jz          KeccakP1600_OverwriteWithZeroes_Exit
-    mov         (%rdx), %rax
-    add         %rdi, %rax
-KeccakP1600_OverwriteWithZeroes_LastIncompleteLaneLoop:
-    movb        $0, (%rax)
-    inc         %rax
-    dec         %rsi
-    jnz         KeccakP1600_OverwriteWithZeroes_LastIncompleteLaneLoop
-KeccakP1600_OverwriteWithZeroes_Exit:
-    ret
-.ifndef old_gas_syntax
-.size   KeccakP1600_OverwriteWithZeroes,.-KeccakP1600_OverwriteWithZeroes
+.ifdef macOS
+.else
+.size   KeccakP1600_AVX2_AddBytes,.-KeccakP1600_AVX2_AddBytes
 .endif
 
 # -----------------------------------------------------------------------------
 #
-# void KeccakP1600_ExtractBytes(const void *state, unsigned char *data, unsigned int offset, unsigned int length);
+# void KeccakP1600_AVX2_ExtractBytes(const void *state, unsigned char *data, unsigned int offset, unsigned int length);
 #                                           %rdi                  %rsi               %rdx                 %rcx
 #
-.globl  KeccakP1600_ExtractBytes
-.globl _KeccakP1600_ExtractBytes
-.ifndef old_gas_syntax
-.type   KeccakP1600_ExtractBytes,@function
+.ifdef macOS
+.globl  _KeccakP1600_AVX2_ExtractBytes
+_KeccakP1600_AVX2_ExtractBytes:
+.else
+.globl  KeccakP1600_AVX2_ExtractBytes
+.type   KeccakP1600_AVX2_ExtractBytes,@function
+KeccakP1600_AVX2_ExtractBytes:
 .endif
-KeccakP1600_ExtractBytes:
-_KeccakP1600_ExtractBytes:
-.balign 32
+.balign  32
     push        %rbx
     cmp         $0, %rcx
-    jz          KeccakP1600_ExtractBytes_Exit
+    jz          KeccakP1600_AVX2_ExtractBytes_Exit
     mov         %rdx, %rax                              # rax offset in lane
     and         $0xFFFFFFF8, %edx                       # rdx pointer into state index mapper
     lea         mapState(%rip), %r9
     add         %r9, %rdx
     and         $7, %rax
-    jz          KeccakP1600_ExtractBytes_LaneAlignedCheck
+    jz          KeccakP1600_AVX2_ExtractBytes_LaneAlignedCheck
     mov         $8, %rbx                                # rbx is (max) length of incomplete lane
     sub         %rax, %rbx
     cmp         %rcx, %rbx
@@ -271,127 +168,53 @@ _KeccakP1600_ExtractBytes:
     add         $8, %rdx
     add         %rdi, %r9
     add         %rax, %r9
-KeccakP1600_ExtractBytes_NotAlignedLoop:
+KeccakP1600_AVX2_ExtractBytes_NotAlignedLoop:
     mov         (%r9), %r8b
     inc         %r9
     mov         %r8b, (%rsi)
     inc         %rsi
     dec         %rbx
-    jnz         KeccakP1600_ExtractBytes_NotAlignedLoop
-    jmp         KeccakP1600_ExtractBytes_LaneAlignedCheck
-KeccakP1600_ExtractBytes_LaneAlignedLoop:
+    jnz         KeccakP1600_AVX2_ExtractBytes_NotAlignedLoop
+    jmp         KeccakP1600_AVX2_ExtractBytes_LaneAlignedCheck
+KeccakP1600_AVX2_ExtractBytes_LaneAlignedLoop:
     mov         (%rdx), %rax
     add         $8, %rdx
     add         %rdi, %rax
     mov         (%rax), %r8
     mov         %r8, (%rsi)
     add         $8, %rsi
-KeccakP1600_ExtractBytes_LaneAlignedCheck:
+KeccakP1600_AVX2_ExtractBytes_LaneAlignedCheck:
     sub         $8, %rcx
-    jnc         KeccakP1600_ExtractBytes_LaneAlignedLoop
-KeccakP1600_ExtractBytes_LastIncompleteLane:
+    jnc         KeccakP1600_AVX2_ExtractBytes_LaneAlignedLoop
+KeccakP1600_AVX2_ExtractBytes_LastIncompleteLane:
     add         $8, %rcx
-    jz          KeccakP1600_ExtractBytes_Exit
+    jz          KeccakP1600_AVX2_ExtractBytes_Exit
     mov         (%rdx), %rax
     add         %rdi, %rax
     mov         (%rax), %r8
-KeccakP1600_ExtractBytes_LastIncompleteLaneLoop:
+KeccakP1600_AVX2_ExtractBytes_LastIncompleteLaneLoop:
     mov         %r8b, (%rsi)
     shr         $8, %r8
     inc         %rsi
     dec         %rcx
-    jnz         KeccakP1600_ExtractBytes_LastIncompleteLaneLoop
-KeccakP1600_ExtractBytes_Exit:
+    jnz         KeccakP1600_AVX2_ExtractBytes_LastIncompleteLaneLoop
+KeccakP1600_AVX2_ExtractBytes_Exit:
     pop         %rbx
     ret
-.ifndef old_gas_syntax
-.size   KeccakP1600_ExtractBytes,.-KeccakP1600_ExtractBytes
-.endif
-
-# -----------------------------------------------------------------------------
-#
-# void KeccakP1600_ExtractAndAddBytes(const void *state, const unsigned char *input, unsigned char *output, unsigned int offset, unsigned int length);
-#                                                 %rdi                        %rsi                  %rdx                 %rcx                  %r8
-#
-.globl  KeccakP1600_ExtractAndAddBytes
-.globl _KeccakP1600_ExtractAndAddBytes
-.ifndef old_gas_syntax
-.type   KeccakP1600_ExtractAndAddBytes,@function
-.endif
-KeccakP1600_ExtractAndAddBytes:
-_KeccakP1600_ExtractAndAddBytes:
-.balign 32
-    push        %rbx
-    push        %r10
-    cmp         $0, %r8
-    jz          KeccakP1600_ExtractAndAddBytes_Exit
-    mov         %rcx, %rax                              # rax offset in lane
-    and         $0xFFFFFFF8, %ecx                       # rcx pointer into state index mapper
-    lea         mapState(%rip), %r9
-    add         %r9, %rcx
-    and         $7, %rax
-    jz          KeccakP1600_ExtractAndAddBytes_LaneAlignedCheck
-    mov         $8, %rbx                                # rbx is (max) length of incomplete lane
-    sub         %rax, %rbx
-    cmp         %r8, %rbx
-    cmovae      %r8, %rbx
-    sub         %rbx, %r8                               # length -= length of incomplete lane
-    mov         (%rcx), %r9
-    add         $8, %rcx
-    add         %rdi, %r9
-    add         %rax, %r9
-KeccakP1600_ExtractAndAddBytes_NotAlignedLoop:
-    mov         (%r9), %r10b
-    inc         %r9
-    xor         (%rsi), %r10b
-    inc         %rsi
-    mov         %r10b, (%rdx)
-    inc         %rdx
-    dec         %rbx
-    jnz         KeccakP1600_ExtractAndAddBytes_NotAlignedLoop
-    jmp         KeccakP1600_ExtractAndAddBytes_LaneAlignedCheck
-KeccakP1600_ExtractAndAddBytes_LaneAlignedLoop:
-    mov         (%rcx), %rax
-    add         $8, %rcx
-    add         %rdi, %rax
-    mov         (%rax), %r10
-    xor         (%rsi), %r10
-    add         $8, %rsi
-    mov         %r10, (%rdx)
-    add         $8, %rdx
-KeccakP1600_ExtractAndAddBytes_LaneAlignedCheck:
-    sub         $8, %r8
-    jnc         KeccakP1600_ExtractAndAddBytes_LaneAlignedLoop
-KeccakP1600_ExtractAndAddBytes_LastIncompleteLane:
-    add         $8, %r8
-    jz          KeccakP1600_ExtractAndAddBytes_Exit
-    mov         (%rcx), %rax
-    add         %rdi, %rax
-    mov         (%rax), %r10
-KeccakP1600_ExtractAndAddBytes_LastIncompleteLaneLoop:
-    xor         (%rsi), %r10b
-    inc         %rsi
-    mov         %r10b, (%rdx)
-    inc         %rdx
-    shr         $8, %r10
-    dec         %r8
-    jnz         KeccakP1600_ExtractAndAddBytes_LastIncompleteLaneLoop
-KeccakP1600_ExtractAndAddBytes_Exit:
-    pop         %r10
-    pop         %rbx
-    ret
-.ifndef old_gas_syntax
-.size   KeccakP1600_ExtractAndAddBytes,.-KeccakP1600_ExtractAndAddBytes
+.ifdef macOS
+.else
+.size   KeccakP1600_AVX2_ExtractBytes,.-KeccakP1600_AVX2_ExtractBytes
 .endif
 
 # -----------------------------------------------------------------------------
 #
 # internal    
 #
-.ifndef old_gas_syntax
+.ifdef macOS
+.else
 .type    __KeccakF1600,@function
 .endif
-.balign 32
+.balign  32
 __KeccakF1600:
 .Loop_avx2:
     ######################################### Theta
@@ -530,63 +353,22 @@ __KeccakF1600:
     dec         %eax
     jnz         .Loop_avx2
     ret
-.ifndef old_gas_syntax
+.ifdef macOS
+.else
 .size   __KeccakF1600,.-__KeccakF1600
 .endif
 
-# -----------------------------------------------------------------------------
-#
-# void KeccakP1600_Permute_24rounds(void *state);
-#                                        %rdi
-#
-.globl  KeccakP1600_Permute_24rounds
-.globl _KeccakP1600_Permute_24rounds
-.ifndef old_gas_syntax
-.type   KeccakP1600_Permute_24rounds,@function
-.endif
-KeccakP1600_Permute_24rounds:
-_KeccakP1600_Permute_24rounds:
-.balign 32
-    lea             rhotates_left+96(%rip),%r8
-    lea             rhotates_right+96(%rip),%r9
-    lea             iotas(%rip),%r10
-    mov             $24,%eax
-    lea             96(%rdi),%rdi
-    vzeroupper
-    vpbroadcastq    -96(%rdi),%ymm0         # load A[5][5]
-    vmovdqu         8+32*0-96(%rdi),%ymm1
-    vmovdqu         8+32*1-96(%rdi),%ymm2
-    vmovdqu         8+32*2-96(%rdi),%ymm3
-    vmovdqu         8+32*3-96(%rdi),%ymm4
-    vmovdqu         8+32*4-96(%rdi),%ymm5
-    vmovdqu         8+32*5-96(%rdi),%ymm6
-    call            __KeccakF1600
-    vmovq           %xmm0,-96(%rdi)
-    vmovdqu         %ymm1,8+32*0-96(%rdi)
-    vmovdqu         %ymm2,8+32*1-96(%rdi)
-    vmovdqu         %ymm3,8+32*2-96(%rdi)
-    vmovdqu         %ymm4,8+32*3-96(%rdi)
-    vmovdqu         %ymm5,8+32*4-96(%rdi)
-    vmovdqu         %ymm6,8+32*5-96(%rdi)
-    vzeroupper
-    ret
-.ifndef old_gas_syntax
-.size   KeccakP1600_Permute_24rounds,.-KeccakP1600_Permute_24rounds
-.endif
 
-# -----------------------------------------------------------------------------
-#
-# void KeccakP1600_Permute_12rounds(void *state);
-#                                        %rdi
-#
-.globl  KeccakP1600_Permute_12rounds
-.globl _KeccakP1600_Permute_12rounds
-.ifndef old_gas_syntax
-.type   KeccakP1600_Permute_12rounds,@function
+
+.ifdef macOS
+.globl  _KeccakP1600_AVX2_Permute_12rounds
+_KeccakP1600_AVX2_Permute_12rounds:
+.else
+.globl  KeccakP1600_AVX2_Permute_12rounds
+.type   KeccakP1600_AVX2_Permute_12rounds,@function
+KeccakP1600_AVX2_Permute_12rounds:
 .endif
-KeccakP1600_Permute_12rounds:
-_KeccakP1600_Permute_12rounds:
-.balign 32
+.balign  32
     lea             rhotates_left+96(%rip),%r8
     lea             rhotates_right+96(%rip),%r9
     lea             iotas+12*4*8(%rip),%r10
@@ -610,253 +392,34 @@ _KeccakP1600_Permute_12rounds:
     vmovdqu         %ymm6,8+32*5-96(%rdi)
     vzeroupper
     ret
-.ifndef old_gas_syntax
-.size   KeccakP1600_Permute_12rounds,.-KeccakP1600_Permute_12rounds
-.endif
-
-# -----------------------------------------------------------------------------
-#
-# void KeccakP1600_Permute_Nrounds(void *state, unsigned int nrounds);
-#                                        %rdi                %rsi
-#
-.globl  KeccakP1600_Permute_Nrounds
-.globl _KeccakP1600_Permute_Nrounds
-.ifndef old_gas_syntax
-.type   KeccakP1600_Permute_Nrounds,@function
-.endif
-KeccakP1600_Permute_Nrounds:
-_KeccakP1600_Permute_Nrounds:
-.balign 32
-    lea             rhotates_left+96(%rip),%r8
-    lea             rhotates_right+96(%rip),%r9
-    lea             iotas+24*4*8(%rip),%r10
-    mov             %rsi,%rax
-    shl             $2+3,%rsi
-    sub             %rsi, %r10
-    lea             96(%rdi),%rdi
-    vzeroupper
-    vpbroadcastq    -96(%rdi),%ymm0         # load A[5][5]
-    vmovdqu         8+32*0-96(%rdi),%ymm1
-    vmovdqu         8+32*1-96(%rdi),%ymm2
-    vmovdqu         8+32*2-96(%rdi),%ymm3
-    vmovdqu         8+32*3-96(%rdi),%ymm4
-    vmovdqu         8+32*4-96(%rdi),%ymm5
-    vmovdqu         8+32*5-96(%rdi),%ymm6
-    call            __KeccakF1600
-    vmovq           %xmm0,-96(%rdi)
-    vmovdqu         %ymm1,8+32*0-96(%rdi)
-    vmovdqu         %ymm2,8+32*1-96(%rdi)
-    vmovdqu         %ymm3,8+32*2-96(%rdi)
-    vmovdqu         %ymm4,8+32*3-96(%rdi)
-    vmovdqu         %ymm5,8+32*4-96(%rdi)
-    vmovdqu         %ymm6,8+32*5-96(%rdi)
-    vzeroupper
-    ret
-.ifndef old_gas_syntax
-.size   KeccakP1600_Permute_Nrounds,.-KeccakP1600_Permute_Nrounds
-.endif
-
-# -----------------------------------------------------------------------------
-#
-# size_t KeccakF1600_FastLoop_Absorb(void *state, unsigned int laneCount, const unsigned char *data, size_t dataByteLen);
-#                                          %rdi                %rsi                            %rdx         %rcx
-#
-.globl  KeccakF1600_FastLoop_Absorb
-.globl _KeccakF1600_FastLoop_Absorb
-.ifndef old_gas_syntax
-.type   KeccakF1600_FastLoop_Absorb,@function
-.endif
-KeccakF1600_FastLoop_Absorb:
-_KeccakF1600_FastLoop_Absorb:
-.balign 32
-    push            %rbx
-    push            %r10
-    shr             $3, %rcx                # rcx = data length in lanes
-    mov             %rdx, %rbx              # rbx = initial data pointer
-    cmp             %rsi, %rcx
-    jb              KeccakF1600_FastLoop_Absorb_Exit
-    vzeroupper
-    cmp             $21, %rsi    
-    jnz             KeccakF1600_FastLoop_Absorb_Not21Lanes
-    sub             $21, %rcx
-    lea             rhotates_left+96(%rip),%r8
-    lea             rhotates_right+96(%rip),%r9
-    lea             96(%rdi),%rdi
-    vpbroadcastq    -96(%rdi),%ymm0         # load A[5][5]
-    vmovdqu         8+32*0-96(%rdi),%ymm1
-    vmovdqu         8+32*1-96(%rdi),%ymm2
-    vmovdqu         8+32*2-96(%rdi),%ymm3
-    vmovdqu         8+32*3-96(%rdi),%ymm4
-    vmovdqu         8+32*4-96(%rdi),%ymm5
-    vmovdqu         8+32*5-96(%rdi),%ymm6
-KeccakF1600_FastLoop_Absorb_Loop21Lanes:    
-    vpbroadcastq    (%rdx),%ymm7
-    vmovdqu         8(%rdx),%ymm8
-
-    vmovdqa         map2(%rip), %xmm15
-    vpcmpeqd        %ymm14, %ymm14, %ymm14
-    vpgatherdq      %ymm14, (%rdx, %xmm15, 1), %ymm9
-
-    vmovdqa         mask3_21(%rip), %ymm14
-    vpxor           %ymm10, %ymm10, %ymm10
-    vmovdqa         map3(%rip), %xmm15
-    vpgatherdq      %ymm14, (%rdx, %xmm15, 1), %ymm10
-
-    vmovdqa         mask4_21(%rip), %ymm14
-    vpxor           %ymm11, %ymm11, %ymm11
-    vmovdqa         map4(%rip), %xmm15
-    vpgatherdq      %ymm14, (%rdx, %xmm15, 1), %ymm11
-
-    vmovdqa         mask5_21(%rip), %ymm14
-    vpxor           %ymm12, %ymm12, %ymm12
-    vmovdqa         map5(%rip), %xmm15
-    vpgatherdq      %ymm14, (%rdx, %xmm15, 1), %ymm12
-
-    vmovdqa         mask6_21(%rip), %ymm14
-    vpxor           %ymm13, %ymm13, %ymm13
-    vmovdqa         map6(%rip), %xmm15
-    vpgatherdq      %ymm14, (%rdx, %xmm15, 1), %ymm13
-
-    vpxor           %ymm7,%ymm0,%ymm0
-    vpxor           %ymm8,%ymm1,%ymm1
-    vpxor           %ymm9,%ymm2,%ymm2
-    vpxor           %ymm10,%ymm3,%ymm3
-    vpxor           %ymm11,%ymm4,%ymm4
-    vpxor           %ymm12,%ymm5,%ymm5
-    vpxor           %ymm13,%ymm6,%ymm6
-    add             $21*8, %rdx
-    lea             iotas(%rip),%r10
-    mov             $24,%eax
-    call            __KeccakF1600
-    sub             $21, %rcx
-    jnc             KeccakF1600_FastLoop_Absorb_Loop21Lanes
-KeccakF1600_FastLoop_Absorb_SaveAndExit:
-    vmovq           %xmm0,-96(%rdi)
-    vmovdqu         %ymm1,8+32*0-96(%rdi)
-    vmovdqu         %ymm2,8+32*1-96(%rdi)
-    vmovdqu         %ymm3,8+32*2-96(%rdi)
-    vmovdqu         %ymm4,8+32*3-96(%rdi)
-    vmovdqu         %ymm5,8+32*4-96(%rdi)
-    vmovdqu         %ymm6,8+32*5-96(%rdi)
-KeccakF1600_FastLoop_Absorb_Exit:
-    vzeroupper
-    mov             %rdx, %rax               # return number of bytes processed
-    sub             %rbx, %rax
-    pop             %r10
-    pop             %rbx
-    ret
-KeccakF1600_FastLoop_Absorb_Not21Lanes:
-    cmp             $17, %rsi    
-    jnz             KeccakF1600_FastLoop_Absorb_Not17Lanes
-    sub             $17, %rcx
-    lea             rhotates_left+96(%rip),%r8
-    lea             rhotates_right+96(%rip),%r9
-    lea             96(%rdi),%rdi
-    vpbroadcastq    -96(%rdi),%ymm0         # load A[5][5]
-    vmovdqu         8+32*0-96(%rdi),%ymm1
-    vmovdqu         8+32*1-96(%rdi),%ymm2
-    vmovdqu         8+32*2-96(%rdi),%ymm3
-    vmovdqu         8+32*3-96(%rdi),%ymm4
-    vmovdqu         8+32*4-96(%rdi),%ymm5
-    vmovdqu         8+32*5-96(%rdi),%ymm6
-KeccakF1600_FastLoop_Absorb_Loop17Lanes:    
-    vpbroadcastq    (%rdx),%ymm7
-    vmovdqu         8(%rdx),%ymm8
-
-    vmovdqa         mask2_17(%rip), %ymm14
-    vpxor           %ymm9, %ymm9, %ymm9
-    vmovdqa         map2(%rip), %xmm15
-    vpgatherdq      %ymm14, (%rdx, %xmm15, 1), %ymm9
-
-    vmovdqa         mask3_17(%rip), %ymm14
-    vpxor           %ymm10, %ymm10, %ymm10
-    vmovdqa         map3(%rip), %xmm15
-    vpgatherdq      %ymm14, (%rdx, %xmm15, 1), %ymm10
-
-    vmovdqa         mask4_17(%rip), %ymm14
-    vpxor           %ymm11, %ymm11, %ymm11
-    vmovdqa         map4(%rip), %xmm15
-    vpgatherdq      %ymm14, (%rdx, %xmm15, 1), %ymm11
-
-    vmovdqa         mask5_17(%rip), %ymm14
-    vpxor           %ymm12, %ymm12, %ymm12
-    vmovdqa         map5(%rip), %xmm15
-    vpgatherdq      %ymm14, (%rdx, %xmm15, 1), %ymm12
-
-    vmovdqa         mask6_17(%rip), %ymm14
-    vpxor           %ymm13, %ymm13, %ymm13
-    vmovdqa         map6(%rip), %xmm15
-    vpgatherdq      %ymm14, (%rdx, %xmm15, 1), %ymm13
-
-    vpxor           %ymm7,%ymm0,%ymm0
-    vpxor           %ymm8,%ymm1,%ymm1
-    vpxor           %ymm9,%ymm2,%ymm2
-    vpxor           %ymm10,%ymm3,%ymm3
-    vpxor           %ymm11,%ymm4,%ymm4
-    vpxor           %ymm12,%ymm5,%ymm5
-    vpxor           %ymm13,%ymm6,%ymm6
-    add             $17*8, %rdx
-    lea             iotas(%rip),%r10
-    mov             $24,%eax
-    call            __KeccakF1600
-    sub             $17, %rcx
-    jnc             KeccakF1600_FastLoop_Absorb_Loop17Lanes
-    jmp             KeccakF1600_FastLoop_Absorb_SaveAndExit
-KeccakF1600_FastLoop_Absorb_Not17Lanes:
-    lea             mapState(%rip), %r9
-    mov             %rsi, %rax
-KeccakF1600_FastLoop_Absorb_LanesAddLoop:
-    mov             (%rdx), %r8
-    add             $8, %rdx
-    mov             (%r9), %r10
-    add             $8, %r9
-    add             %rdi, %r10
-    xor             %r8, (%r10)
-    sub             $1, %rax
-    jnz             KeccakF1600_FastLoop_Absorb_LanesAddLoop
-    sub             %rsi, %rcx
-    push            %rdi
-    push            %rsi
-    push            %rdx
-    push            %rcx
-.ifdef no_plt
-    call            KeccakP1600_Permute_24rounds
+.ifdef macOS
 .else
-    call            KeccakP1600_Permute_24rounds@PLT
-.endif
-    pop             %rcx
-    pop             %rdx
-    pop             %rsi
-    pop             %rdi
-    cmp             %rsi, %rcx
-    jae             KeccakF1600_FastLoop_Absorb_Not17Lanes
-    jmp             KeccakF1600_FastLoop_Absorb_Exit
-.ifndef old_gas_syntax
-.size   KeccakF1600_FastLoop_Absorb,.-KeccakF1600_FastLoop_Absorb
+.size   KeccakP1600_AVX2_Permute_12rounds,.-KeccakP1600_AVX2_Permute_12rounds
 .endif
 
 # -----------------------------------------------------------------------------
 #
-# size_t KeccakP1600_12rounds_FastLoop_Absorb(void *state, unsigned int laneCount, const unsigned char *data, size_t dataByteLen);
+# size_t KeccakP1600_AVX2_12rounds_FastLoop_Absorb(void *state, unsigned int laneCount, const unsigned char *data, size_t dataByteLen);
 #                                          %rdi                %rsi                            %rdx         %rcx
 #
-.globl  KeccakP1600_12rounds_FastLoop_Absorb
-.globl _KeccakP1600_12rounds_FastLoop_Absorb
-.ifndef old_gas_syntax
-.type   KeccakP1600_12rounds_FastLoop_Absorb,@function
+.ifdef macOS
+.globl  _KeccakP1600_AVX2_12rounds_FastLoop_Absorb
+_KeccakP1600_AVX2_12rounds_FastLoop_Absorb:
+.else
+.globl  KeccakP1600_AVX2_12rounds_FastLoop_Absorb
+.type   KeccakP1600_AVX2_12rounds_FastLoop_Absorb,@function
+KeccakP1600_AVX2_12rounds_FastLoop_Absorb:
 .endif
-KeccakP1600_12rounds_FastLoop_Absorb:
-_KeccakP1600_12rounds_FastLoop_Absorb:
-.balign 32
+.balign  32
     push            %rbx
     push            %r10
     shr             $3, %rcx                # rcx = data length in lanes
     mov             %rdx, %rbx              # rbx = initial data pointer
     cmp             %rsi, %rcx
-    jb              KeccakP1600_12rounds_FastLoop_Absorb_Exit
+    jb              KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Exit
     vzeroupper
     cmp             $21, %rsi    
-    jnz             KeccakP1600_12rounds_FastLoop_Absorb_Not21Lanes
+    jnz             KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Not21Lanes
     sub             $21, %rcx
     lea             rhotates_left+96(%rip),%r8
     lea             rhotates_right+96(%rip),%r9
@@ -868,7 +431,7 @@ _KeccakP1600_12rounds_FastLoop_Absorb:
     vmovdqu         8+32*3-96(%rdi),%ymm4
     vmovdqu         8+32*4-96(%rdi),%ymm5
     vmovdqu         8+32*5-96(%rdi),%ymm6
-KeccakP1600_12rounds_FastLoop_Absorb_Loop21Lanes:    
+KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Loop21Lanes:
     vpbroadcastq    (%rdx),%ymm7
     vmovdqu         8(%rdx),%ymm8
 
@@ -908,8 +471,8 @@ KeccakP1600_12rounds_FastLoop_Absorb_Loop21Lanes:
     mov             $12,%eax
     call            __KeccakF1600
     sub             $21, %rcx
-    jnc             KeccakP1600_12rounds_FastLoop_Absorb_Loop21Lanes
-KeccakP1600_12rounds_FastLoop_Absorb_SaveAndExit:
+    jnc             KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Loop21Lanes
+KeccakP1600_AVX2_12rounds_FastLoop_Absorb_SaveAndExit:
     vmovq           %xmm0,-96(%rdi)
     vmovdqu         %ymm1,8+32*0-96(%rdi)
     vmovdqu         %ymm2,8+32*1-96(%rdi)
@@ -917,16 +480,16 @@ KeccakP1600_12rounds_FastLoop_Absorb_SaveAndExit:
     vmovdqu         %ymm4,8+32*3-96(%rdi)
     vmovdqu         %ymm5,8+32*4-96(%rdi)
     vmovdqu         %ymm6,8+32*5-96(%rdi)
-KeccakP1600_12rounds_FastLoop_Absorb_Exit:
+KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Exit:
     vzeroupper
     mov             %rdx, %rax               # return number of bytes processed
     sub             %rbx, %rax
     pop             %r10
     pop             %rbx
     ret
-KeccakP1600_12rounds_FastLoop_Absorb_Not21Lanes:
+KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Not21Lanes:
     cmp             $17, %rsi    
-    jnz             KeccakP1600_12rounds_FastLoop_Absorb_Not17Lanes
+    jnz             KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Not17Lanes
     sub             $17, %rcx
     lea             rhotates_left+96(%rip),%r8
     lea             rhotates_right+96(%rip),%r9
@@ -938,7 +501,7 @@ KeccakP1600_12rounds_FastLoop_Absorb_Not21Lanes:
     vmovdqu         8+32*3-96(%rdi),%ymm4
     vmovdqu         8+32*4-96(%rdi),%ymm5
     vmovdqu         8+32*5-96(%rdi),%ymm6
-KeccakP1600_12rounds_FastLoop_Absorb_Loop17Lanes:    
+KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Loop17Lanes:
     vpbroadcastq    (%rdx),%ymm7
     vmovdqu         8(%rdx),%ymm8
 
@@ -979,12 +542,12 @@ KeccakP1600_12rounds_FastLoop_Absorb_Loop17Lanes:
     mov             $12,%eax
     call            __KeccakF1600
     sub             $17, %rcx
-    jnc             KeccakP1600_12rounds_FastLoop_Absorb_Loop17Lanes
-    jmp             KeccakP1600_12rounds_FastLoop_Absorb_SaveAndExit
-KeccakP1600_12rounds_FastLoop_Absorb_Not17Lanes:
+    jnc             KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Loop17Lanes
+    jmp             KeccakP1600_AVX2_12rounds_FastLoop_Absorb_SaveAndExit
+KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Not17Lanes:
     lea             mapState(%rip), %r9
     mov             %rsi, %rax
-KeccakP1600_12rounds_FastLoop_Absorb_LanesAddLoop:
+KeccakP1600_AVX2_12rounds_FastLoop_Absorb_LanesAddLoop:
     mov             (%rdx), %r8
     add             $8, %rdx
     mov             (%r9), %r10
@@ -992,31 +555,32 @@ KeccakP1600_12rounds_FastLoop_Absorb_LanesAddLoop:
     add             %rdi, %r10
     xor             %r8, (%r10)
     sub             $1, %rax
-    jnz             KeccakP1600_12rounds_FastLoop_Absorb_LanesAddLoop
+    jnz             KeccakP1600_AVX2_12rounds_FastLoop_Absorb_LanesAddLoop
     sub             %rsi, %rcx
     push            %rdi
     push            %rsi
     push            %rdx
     push            %rcx
-.ifdef no_plt
-    call            KeccakP1600_Permute_12rounds
+.ifdef macOS
+    call            _KeccakP1600_AVX2_Permute_12rounds
 .else
-    call            KeccakP1600_Permute_12rounds@PLT
+    call            KeccakP1600_AVX2_Permute_12rounds@PLT
 .endif
     pop             %rcx
     pop             %rdx
     pop             %rsi
     pop             %rdi
     cmp             %rsi, %rcx
-    jae             KeccakP1600_12rounds_FastLoop_Absorb_Not17Lanes
-    jmp             KeccakP1600_12rounds_FastLoop_Absorb_Exit
-.ifndef old_gas_syntax
-.size   KeccakP1600_12rounds_FastLoop_Absorb,.-KeccakP1600_12rounds_FastLoop_Absorb
+    jae             KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Not17Lanes
+    jmp             KeccakP1600_AVX2_12rounds_FastLoop_Absorb_Exit
+.ifdef macOS
+.else
+.size   KeccakP1600_AVX2_12rounds_FastLoop_Absorb,.-KeccakP1600_AVX2_12rounds_FastLoop_Absorb
 .endif
 
 .equ    ALLON,        0xFFFFFFFFFFFFFFFF
 
-.balign 64
+.balign  64
 rhotates_left:
     .quad     3,   18,    36,    41         # [2][0] [4][0] [1][0] [3][0]
     .quad     1,   62,    28,    27         # [0][1] [0][2] [0][3] [0][4]
@@ -1064,7 +628,7 @@ mapState:
     .quad     8*8,  9*8, 18*8, 23*8, 16*8
     .quad     6*8, 17*8, 14*8, 11*8, 24*8
 
-.balign 16
+    .balign   16
 map2:
     .long    10*8, 20*8,  5*8, 15*8
 map3:
@@ -1076,7 +640,7 @@ map5:
 map6:
     .long     6*8, 12*8, 18*8, 24*8
 
-.balign 32
+    .balign   32
 mask3_21:
     .quad    ALLON, ALLON,     0, ALLON
 mask4_21: