diff options
author | Andy Polyakov <appro@openssl.org> | 2012-04-18 13:01:36 +0000 |
---|---|---|
committer | Andy Polyakov <appro@openssl.org> | 2012-04-18 13:01:36 +0000 |
commit | 3e181369dd635c8ce5d133e346928d6670cf83f6 (patch) | |
tree | 47abf2c0adeeffb0ae4d59c87c0075b87bc06f65 | |
parent | d3ddf0228e1977e5328277d751774bc9f9d385d8 (diff) |
C64x+ assembler pack. linux-c64xplus build is *not* tested nor can it be
tested, because kernel is not in shape to handle it *yet*. The code is
committed mostly to stimulate the kernel development.
-rwxr-xr-x | Configure | 3 | ||||
-rw-r--r-- | TABLE | 33 | ||||
-rw-r--r-- | crypto/aes/asm/aes-c64xplus.pl | 1361 | ||||
-rw-r--r-- | crypto/bn/asm/bn-c64xplus.asm | 333 | ||||
-rw-r--r-- | crypto/bn/asm/c64xplus-gf2m.pl | 146 | ||||
-rw-r--r-- | crypto/c64xpluscpuid.pl | 246 | ||||
-rw-r--r-- | crypto/modes/asm/ghash-c64xplus.pl | 231 | ||||
-rw-r--r-- | crypto/sha/asm/sha1-c64xplus.pl | 323 | ||||
-rw-r--r-- | crypto/sha/asm/sha256-c64xplus.pl | 302 | ||||
-rw-r--r-- | crypto/sha/asm/sha512-c64xplus.pl | 421 |
10 files changed, 3399 insertions, 0 deletions
@@ -399,6 +399,9 @@ my %table=( "linux-alpha+bwx-gcc","gcc:-O3 -DL_ENDIAN -DTERMIO::-D_REENTRANT::-ldl:SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_RISC1 DES_UNROLL:${alpha_asm}:dlfcn:linux-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)", "linux-alpha-ccc","ccc:-fast -readonly_strings -DL_ENDIAN -DTERMIO::-D_REENTRANT:::SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_PTR DES_RISC1 DES_UNROLL:${alpha_asm}", "linux-alpha+bwx-ccc","ccc:-fast -readonly_strings -DL_ENDIAN -DTERMIO::-D_REENTRANT:::SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_INT DES_PTR DES_RISC1 DES_UNROLL:${alpha_asm}", +# +# TI_CGT_C6000_7.3.x is a requirement +"linux-c64xplus","cl6x:--linux --strip_coff_underscore -ea=.s -eo=.o -mv6400+ -o2 -ox -ms -pden -DOPENSSL_SMALL_FOOTPRINT::-D_REENTRANT:::BN_LLONG:c64xpluscpuid.o:bn-c64xplus.o c64xplus-gf2m.o::aes-c64xplus.o aes_cbc.o aes_ctr.o:::sha1-c64xplus.o sha256-c64xplus.o sha512-c64xplus.o:::::::ghash-c64xplus.o::void:dlfcn:linux-shared:--pic:-z --sysv --shared:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR):true", # Android: linux-* but without -DTERMIO and pointers to headers and libs. "android","gcc:-mandroid -I\$(ANDROID_DEV)/include -B\$(ANDROID_DEV)/lib -O3 -fomit-frame-pointer -Wall::-D_REENTRANT::-ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_INT DES_UNROLL BF_PTR:${no_asm}:dlfcn:linux-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)", @@ -3927,6 +3927,39 @@ $ranlib = $arflags = $multilib = +*** linux-c64xplus +$cc = cl6x +$cflags = --linux --strip_coff_underscore -ea=.s -eo=.o -mv6400+ -o2 -ox -ms -pden -DOPENSSL_SMALL_FOOTPRINT +$unistd = +$thread_cflag = -D_REENTRANT +$sys_id = +$lflags = +$bn_ops = BN_LLONG +$cpuid_obj = c64xpluscpuid.o +$bn_obj = bn-c64xplus.o c64xplus-gf2m.o +$des_obj = +$aes_obj = aes-c64xplus.o aes_cbc.o aes_ctr.o +$bf_obj = +$md5_obj = +$sha1_obj = sha1-c64xplus.o sha256-c64xplus.o sha512-c64xplus.o +$cast_obj = +$rc4_obj = +$rmd160_obj = +$rc5_obj = +$wp_obj = +$cmll_obj = +$modes_obj = ghash-c64xplus.o +$engines_obj = +$perlasm_scheme = void +$dso_scheme = dlfcn +$shared_target= linux-shared +$shared_cflag = --pic +$shared_ldflag = -z --sysv --shared +$shared_extension = .so.$(SHLIB_MAJOR).$(SHLIB_MINOR) +$ranlib = true +$arflags = +$multilib = + *** linux-elf $cc = gcc $cflags = -DL_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -Wall diff --git a/crypto/aes/asm/aes-c64xplus.pl b/crypto/aes/asm/aes-c64xplus.pl new file mode 100644 index 0000000000..ad0c15a36f --- /dev/null +++ b/crypto/aes/asm/aes-c64xplus.pl @@ -0,0 +1,1361 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# [Endian-neutral] AES for C64x+. +# +# Even though SPLOOPs are scheduled for 13 cycles, and thus expected +# performance is ~8.5 cycles per byte processed with 128-bit key, +# measured performance turned to be ~10 cycles per byte. Discrepancy +# must be caused by limitations of L1D memory banking(*), see SPRU871 +# TI publication for further details. If any consolation it's still +# ~20% faster than TI's linear assembly module anyway... Compared to +# aes_core.c compiled with cl6x 6.0 with -mv6400+ -o2 options this +# code is 3.75x faster and almost 3x smaller (tables included). +# +# (*) This means that there might be subtle correlation between data +# and timing and one can wonder if it can be ... attacked:-( +# On the other hand this also means that *if* one chooses to +# implement *4* T-tables variant [instead of 1 T-table as in +# this implementation, or in addition to], then one ought to +# *interleave* them. Even though it complicates addressing, +# references to interleaved tables would be guaranteed not to +# clash. I reckon that it should be possible to break 8 cycles +# per byte "barrier," i.e. improve by ~20%, naturally at the +# cost of 8x increased pressure on L1D. 8x because you'd have +# to interleave both Te and Td tables... + +while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} +open STDOUT,">$output"; + +($TEA,$TEB)=("A5","B5"); +($KPA,$KPB)=("A3","B1"); +@K=("A6","B6","A7","B7"); +@s=("A8","B8","A9","B9"); +@Te0=@Td0=("A16","B16","A17","B17"); +@Te1=@Td1=("A18","B18","A19","B19"); +@Te2=@Td2=("A20","B20","A21","B21"); +@Te3=@Td3=("A22","B22","A23","B23"); + +$code=<<___; + .text + .if __TI_EABI__ + .nocmp + .endif + + .asg B3,RA + .asg A4,INP + .asg B4,OUT + .asg A6,KEY + .asg A4,RET + .asg B15,SP + + .eval 24,EXT0 + .eval 16,EXT1 + .eval 8,EXT2 + .eval 0,EXT3 + .eval 8,TBL1 + .eval 16,TBL2 + .eval 24,TBL3 + + .if .BIG_ENDIAN + .eval 24-EXT0,EXT0 + .eval 24-EXT1,EXT1 + .eval 24-EXT2,EXT2 + .eval 24-EXT3,EXT3 + .eval 32-TBL1,TBL1 + .eval 32-TBL2,TBL2 + .eval 32-TBL3,TBL3 + .endif + + .global _AES_encrypt +_AES_encrypt: + .asmfunc + MVK 1,B2 +__encrypt: + .if __TI_EABI__ + [B2] LDNDW *INP++,A9:A8 ; load input +|| MVKL \$PCR_OFFSET(AES_Te,_AES_encrypt),$TEA +|| ADDKPC _AES_encrypt,B0 + [B2] LDNDW *INP++,B9:B8 +|| MVKH \$PCR_OFFSET(AES_Te,_AES_encrypt),$TEA +|| ADD 0,KEY,$KPA +|| ADD 4,KEY,$KPB + .else + [B2] LDNDW *INP++,A9:A8 ; load input +|| MVKL (AES_Te-_AES_encrypt),$TEA +|| ADDKPC _AES_encrypt,B0 + [B2] LDNDW *INP++,B9:B8 +|| MVKH (AES_Te-_AES_encrypt),$TEA +|| ADD 0,KEY,$KPA +|| ADD 4,KEY,$KPB + .endif + LDW *$KPA++[2],$Te0[0] ; zero round key +|| LDW *$KPB++[2],$Te0[1] +|| MVK 60,A0 +|| ADD B0,$TEA,$TEA ; AES_Te + LDW *KEY[A0],B0 ; rounds +|| MVK 1024,A0 ; sizeof(AES_Te) + LDW *$KPA++[2],$Te0[2] +|| LDW *$KPB++[2],$Te0[3] +|| MV $TEA,$TEB + NOP + .if .BIG_ENDIAN + MV A9,$s[0] +|| MV A8,$s[1] +|| MV B9,$s[2] +|| MV B8,$s[3] + .else + MV A8,$s[0] +|| MV A9,$s[1] +|| MV B8,$s[2] +|| MV B9,$s[3] + .endif + XOR $Te0[0],$s[0],$s[0] +|| XOR $Te0[1],$s[1],$s[1] +|| LDW *$KPA++[2],$K[0] ; 1st round key +|| LDW *$KPB++[2],$K[1] + SUB B0,2,B0 + + SPLOOPD 13 +|| MVC B0,ILC +|| LDW *$KPA++[2],$K[2] +|| LDW *$KPB++[2],$K[3] +;;==================================================================== + EXTU $s[1],EXT1,24,$Te1[1] +|| EXTU $s[0],EXT3,24,$Te3[0] + LDW *${TEB}[$Te1[1]],$Te1[1] ; Te1[s1>>8], t0 +|| LDW *${TEA}[$Te3[0]],$Te3[0] ; Te3[s0>>24], t1 +|| XOR $s[2],$Te0[2],$s[2] ; modulo-scheduled +|| XOR $s[3],$Te0[3],$s[3] ; modulo-scheduled +|| EXTU $s[1],EXT3,24,$Te3[1] +|| EXTU $s[0],EXT1,24,$Te1[0] + LDW *${TEB}[$Te3[1]],$Te3[1] ; Te3[s1>>24], t2 +|| LDW *${TEA}[$Te1[0]],$Te1[0] ; Te1[s0>>8], t3 +|| EXTU $s[2],EXT2,24,$Te2[2] +|| EXTU $s[3],EXT2,24,$Te2[3] + LDW *${TEA}[$Te2[2]],$Te2[2] ; Te2[s2>>16], t0 +|| LDW *${TEB}[$Te2[3]],$Te2[3] ; Te2[s3>>16], t1 +|| EXTU $s[3],EXT3,24,$Te3[3] +|| EXTU $s[2],EXT1,24,$Te1[2] + LDW *${TEB}[$Te3[3]],$Te3[3] ; Te3[s3>>24], t0 +|| LDW *${TEA}[$Te1[2]],$Te1[2] ; Te1[s2>>8], t1 +|| EXTU $s[0],EXT2,24,$Te2[0] +|| EXTU $s[1],EXT2,24,$Te2[1] + LDW *${TEA}[$Te2[0]],$Te2[0] ; Te2[s0>>16], t2 +|| LDW *${TEB}[$Te2[1]],$Te2[1] ; Te2[s1>>16], t3 +|| EXTU $s[3],EXT1,24,$Te1[3] +|| EXTU $s[2],EXT3,24,$Te3[2] + LDW *${TEB}[$Te1[3]],$Te1[3] ; Te1[s3>>8], t2 +|| LDW *${TEA}[$Te3[2]],$Te3[2] ; Te3[s2>>24], t3 +|| ROTL $Te1[1],TBL1,$Te3[0] ; t0 +|| ROTL $Te3[0],TBL3,$Te1[1] ; t1 +|| EXTU $s[0],EXT0,24,$Te0[0] +|| EXTU $s[1],EXT0,24,$Te0[1] + LDW *${TEA}[$Te0[0]],$Te0[0] ; Te0[s0], t0 +|| LDW *${TEB}[$Te0[1]],$Te0[1] ; Te0[s1], t1 +|| ROTL $Te3[1],TBL3,$Te1[0] ; t2 +|| ROTL $Te1[0],TBL1,$Te3[1] ; t3 +|| EXTU $s[2],EXT0,24,$Te0[2] +|| EXTU $s[3],EXT0,24,$Te0[3] + LDW *${TEA}[$Te0[2]],$Te0[2] ; Te0[s2], t2 +|| LDW *${TEB}[$Te0[3]],$Te0[3] ; Te0[s3], t3 +|| ROTL $Te2[2],TBL2,$Te2[2] ; t0 +|| ROTL $Te2[3],TBL2,$Te2[3] ; t1 +|| XOR $K[0],$Te3[0],$s[0] +|| XOR $K[1],$Te1[1],$s[1] + ROTL $Te3[3],TBL3,$Te1[2] ; t0 +|| ROTL $Te1[2],TBL1,$Te3[3] ; t1 +|| XOR $K[2],$Te1[0],$s[2] +|| XOR $K[3],$Te3[1],$s[3] +|| LDW *$KPA++[2],$K[0] ; next round key +|| LDW *$KPB++[2],$K[1] + ROTL $Te2[0],TBL2,$Te2[0] ; t2 +|| ROTL $Te2[1],TBL2,$Te2[1] ; t3 +|| XOR $s[0],$Te2[2],$s[0] +|| XOR $s[1],$Te2[3],$s[1] +|| LDW *$KPA++[2],$K[2] +|| LDW *$KPB++[2],$K[3] + ROTL $Te1[3],TBL1,$Te3[2] ; t2 +|| ROTL $Te3[2],TBL3,$Te1[3] ; t3 +|| XOR $s[0],$Te1[2],$s[0] +|| XOR $s[1],$Te3[3],$s[1] + XOR $s[2],$Te2[0],$s[2] +|| XOR $s[3],$Te2[1],$s[3] +|| XOR $s[0],$Te0[0],$s[0] +|| XOR $s[1],$Te0[1],$s[1] + SPKERNEL +|| XOR.L $s[2],$Te3[2],$s[2] +|| XOR.L $s[3],$Te1[3],$s[3] +;;==================================================================== + ADD.D ${TEA},A0,${TEA} ; point to Te4 +|| ADD.D ${TEB},A0,${TEB} +|| EXTU $s[1],EXT1,24,$Te1[1] +|| EXTU $s[0],EXT3,24,$Te3[0] + LDBU *${TEB}[$Te1[1]],$Te1[1] ; Te1[s1>>8], t0 +|| LDBU *${TEA}[$Te3[0]],$Te3[0] ; Te3[s0>>24], t1 +|| XOR $s[2],$Te0[2],$s[2] ; modulo-scheduled +|| XOR $s[3],$Te0[3],$s[3] ; modulo-scheduled +|| EXTU $s[0],EXT0,24,$Te0[0] +|| EXTU $s[1],EXT0,24,$Te0[1] + LDBU *${TEA}[$Te0[0]],$Te0[0] ; Te0[s0], t0 +|| LDBU *${TEB}[$Te0[1]],$Te0[1] ; Te0[s1], t1 +|| EXTU $s[3],EXT3,24,$Te3[3] +|| EXTU $s[2],EXT1,24,$Te1[2] + LDBU *${TEB}[$Te3[3]],$Te3[3] ; Te3[s3>>24], t0 +|| LDBU *${TEA}[$Te1[2]],$Te1[2] ; Te1[s2>>8], t1 +|| EXTU $s[2],EXT2,24,$Te2[2] +|| EXTU $s[3],EXT2,24,$Te2[3] + LDBU *${TEA}[$Te2[2]],$Te2[2] ; Te2[s2>>16], t0 +|| LDBU *${TEB}[$Te2[3]],$Te2[3] ; Te2[s3>>16], t1 +|| EXTU $s[1],EXT3,24,$Te3[1] +|| EXTU $s[0],EXT1,24,$Te1[0] + LDBU *${TEB}[$Te3[1]],$Te3[1] ; Te3[s1>>24], t2 +|| LDBU *${TEA}[$Te1[0]],$Te1[0] ; Te1[s0>>8], t3 +|| EXTU $s[3],EXT1,24,$Te1[3] +|| EXTU $s[2],EXT3,24,$Te3[2] + LDBU *${TEB}[$Te1[3]],$Te1[3] ; Te1[s3>>8], t2 +|| LDBU *${TEA}[$Te3[2]],$Te3[2] ; Te3[s2>>24], t3 +|| EXTU $s[2],EXT0,24,$Te0[2] +|| EXTU $s[3],EXT0,24,$Te0[3] + LDBU *${TEA}[$Te0[2]],$Te0[2] ; Te0[s2], t2 +|| LDBU *${TEB}[$Te0[3]],$Te0[3] ; Te0[s3], t3 +|| EXTU $s[0],EXT2,24,$Te2[0] +|| EXTU $s[1],EXT2,24,$Te2[1] + LDBU *${TEA}[$Te2[0]],$Te2[0] ; Te2[s0>>16], t2 +|| LDBU *${TEB}[$Te2[1]],$Te2[1] ; Te2[s1>>16], t3 + + .if .BIG_ENDIAN + PACK2 $Te0[0],$Te1[1],$Te0[0] +|| PACK2 $Te0[1],$Te1[2],$Te0[1] + PACK2 $Te2[2],$Te3[3],$Te2[2] +|| PACK2 $Te2[3],$Te3[0],$Te2[3] + PACKL4 $Te0[0],$Te2[2],$Te0[0] +|| PACKL4 $Te0[1],$Te2[3],$Te0[1] + XOR $K[0],$Te0[0],$Te0[0] ; s[0] +|| XOR $K[1],$Te0[1],$Te0[1] ; s[1] + + PACK2 $Te0[2],$Te1[3],$Te0[2] +|| PACK2 $Te0[3],$Te1[0],$Te0[3] + PACK2 $Te2[0],$Te3[1],$Te2[0] +|| PACK2 $Te2[1],$Te3[2],$Te2[1] +|| BNOP RA + PACKL4 $Te0[2],$Te2[0],$Te0[2] +|| PACKL4 $Te0[3],$Te2[1],$Te0[3] + XOR $K[2],$Te0[2],$Te0[2] ; s[2] +|| XOR $K[3],$Te0[3],$Te0[3] ; s[3] + + MV $Te0[0],A9 +|| MV $Te0[1],A8 + MV $Te0[2],B9 +|| MV $Te0[3],B8 +|| [B2] STNDW A9:A8,*OUT++ + [B2] STNDW B9:B8,*OUT++ + .else + PACK2 $Te1[1],$Te0[0],$Te1[1] +|| PACK2 $Te1[2],$Te0[1],$Te1[2] + PACK2 $Te3[3],$Te2[2],$Te3[3] +|| PACK2 $Te3[0],$Te2[3],$Te3[0] + PACKL4 $Te3[3],$Te1[1],$Te1[1] +|| PACKL4 $Te3[0],$Te1[2],$Te1[2] + XOR $K[0],$Te1[1],$Te1[1] ; s[0] +|| XOR $K[1],$Te1[2],$Te1[2] ; s[1] + + PACK2 $Te1[3],$Te0[2],$Te1[3] +|| PACK2 $Te1[0],$Te0[3],$Te1[0] + PACK2 $Te3[1],$Te2[0],$Te3[1] +|| PACK2 $Te3[2],$Te2[1],$Te3[2] +|| BNOP RA + PACKL4 $Te3[1],$Te1[3],$Te1[3] +|| PACKL4 $Te3[2],$Te1[0],$Te1[0] + XOR $K[2],$Te1[3],$Te1[3] ; s[2] +|| XOR $K[3],$Te1[0],$Te1[0] ; s[3] + + MV $Te1[1],A8 +|| MV $Te1[2],A9 + MV $Te1[3],B8 +|| MV $Te1[0],B9 +|| [B2] STNDW A9:A8,*OUT++ + [B2] STNDW B9:B8,*OUT++ + .endif + .endasmfunc + + .global _AES_decrypt +_AES_decrypt: + .asmfunc + MVK 1,B2 +__decrypt: + .if __TI_EABI__ + [B2] LDNDW *INP++,A9:A8 ; load input +|| MVKL \$PCR_OFFSET(AES_Td,_AES_decrypt),$TEA +|| ADDKPC _AES_decrypt,B0 + [B2] LDNDW *INP++,B9:B8 +|| MVKH \$PCR_OFFSET(AES_Td,_AES_decrypt),$TEA +|| ADD 0,KEY,$KPA +|| ADD 4,KEY,$KPB + .else + [B2] LDNDW *INP++,A9:A8 ; load input +|| MVKL (AES_Td-_AES_decrypt),$TEA +|| ADDKPC _AES_decrypt,B0 + [B2] LDNDW *INP++,B9:B8 +|| MVKH (AES_Td-_AES_decrypt),$TEA +|| ADD 0,KEY,$KPA +|| ADD 4,KEY,$KPB + .endif + LDW *$KPA++[2],$Td0[0] ; zero round key +|| LDW *$KPB++[2],$Td0[1] +|| MVK 60,A0 +|| ADD B0,$TEA,$TEA ; AES_Td + LDW *KEY[A0],B0 ; rounds +|| MVK 1024,A0 ; sizeof(AES_Td) + LDW *$KPA++[2],$Td0[2] +|| LDW *$KPB++[2],$Td0[3] +|| MV $TEA,$TEB + NOP + .if .BIG_ENDIAN + MV A9,$s[0] +|| MV A8,$s[1] +|| MV B9,$s[2] +|| MV B8,$s[3] + .else + MV A8,$s[0] +|| MV A9,$s[1] +|| MV B8,$s[2] +|| MV B9,$s[3] + .endif + XOR $Td0[0],$s[0],$s[0] +|| XOR $Td0[1],$s[1],$s[1] +|| LDW *$KPA++[2],$K[0] ; 1st round key +|| LDW *$KPB++[2],$K[1] + SUB B0,2,B0 + + SPLOOPD 13 +|| MVC B0,ILC +|| LDW *$KPA++[2],$K[2] +|| LDW *$KPB++[2],$K[3] +;;==================================================================== + EXTU $s[1],EXT3,24,$Td3[1] +|| EXTU $s[0],EXT1,24,$Td1[0] + LDW *${TEB}[$Td3[1]],$Td3[1] ; Td3[s1>>24], t0 +|| LDW *${TEA}[$Td1[0]],$Td1[0] ; Td1[s0>>8], t1 +|| XOR $s[2],$Td0[2],$s[2] ; modulo-scheduled +|| XOR $s[3],$Td0[3],$s[3] ; modulo-scheduled +|| EXTU $s[1],EXT1,24,$Td1[1] +|| EXTU $s[0],EXT3,24,$Td3[0] + LDW *${TEB}[$Td1[1]],$Td1[1] ; Td1[s1>>8], t2 +|| LDW *${TEA}[$Td3[0]],$Td3[0] ; Td3[s0>>24], t3 +|| EXTU $s[2],EXT2,24,$Td2[2] +|| EXTU $s[3],EXT2,24,$Td2[3] + LDW *${TEA}[$Td2[2]],$Td2[2] ; Td2[s2>>16], t0 +|| LDW *${TEB}[$Td2[3]],$Td2[3] ; Td2[s3>>16], t1 +|| EXTU $s[3],EXT1,24,$Td1[3] +|| EXTU $s[2],EXT3,24,$Td3[2] + LDW *${TEB}[$Td1[3]],$Td1[3] ; Td1[s3>>8], t0 +|| LDW *${TEA}[$Td3[2]],$Td3[2] ; Td3[s2>>24], t1 +|| EXTU $s[0],EXT2,24,$Td2[0] +|| EXTU $s[1],EXT2,24,$Td2[1] + LDW *${TEA}[$Td2[0]],$Td2[0] ; Td2[s0>>16], t2 +|| LDW *${TEB}[$Td2[1]],$Td2[1] ; Td2[s1>>16], t3 +|| EXTU $s[3],EXT3,24,$Td3[3] +|| EXTU $s[2],EXT1,24,$Td1[2] + LDW *${TEB}[$Td3[3]],$Td3[3] ; Td3[s3>>24], t2 +|| LDW *${TEA}[$Td1[2]],$Td1[2] ; Td1[s2>>8], t3 +|| ROTL $Td3[1],TBL3,$Td1[0] ; t0 +|| ROTL $Td1[0],TBL1,$Td3[1] ; t1 +|| EXTU $s[0],EXT0,24,$Td0[0] +|| EXTU $s[1],EXT0,24,$Td0[1] + LDW *${TEA}[$Td0[0]],$Td0[0] ; Td0[s0], t0 +|| LDW *${TEB}[$Td0[1]],$Td0[1] ; Td0[s1], t1 +|| ROTL $Td1[1],TBL1,$Td3[0] ; t2 +|| ROTL $Td3[0],TBL3,$Td1[1] ; t3 +|| EXTU $s[2],EXT0,24,$Td0[2] +|| EXTU $s[3],EXT0,24,$Td0[3] + LDW *${TEA}[$Td0[2]],$Td0[2] ; Td0[s2], t2 +|| LDW *${TEB}[$Td0[3]],$Td0[3] ; Td0[s3], t3 +|| ROTL $Td2[2],TBL2,$Td2[2] ; t0 +|| ROTL $Td2[3],TBL2,$Td2[3] ; t1 +|| XOR $K[0],$Td1[0],$s[0] +|| XOR $K[1],$Td3[1],$s[1] + ROTL $Td1[3],TBL1,$Td3[2] ; t0 +|| ROTL $Td3[2],TBL3,$Td1[3] ; t1 +|| XOR $K[2],$Td3[0],$s[2] +|| XOR $K[3],$Td1[1],$s[3] +|| LDW *$KPA++[2],$K[0] ; next round key +|| LDW *$KPB++[2],$K[1] + ROTL $Td2[0],TBL2,$Td2[0] ; t2 +|| ROTL $Td2[1],TBL2,$Td2[1] ; t3 +|| XOR $s[0],$Td2[2],$s[0] +|| XOR $s[1],$Td2[3],$s[1] +|| LDW *$KPA++[2],$K[2] +|| LDW *$KPB++[2],$K[3] + ROTL $Td3[3],TBL3,$Td1[2] ; t2 +|| ROTL $Td1[2],TBL1,$Td3[3] ; t3 +|| XOR $s[0],$Td3[2],$s[0] +|| XOR $s[1],$Td1[3],$s[1] + XOR $s[2],$Td2[0],$s[2] +|| XOR $s[3],$Td2[1],$s[3] +|| XOR $s[0],$Td0[0],$s[0] +|| XOR $s[1],$Td0[1],$s[1] + SPKERNEL +|| XOR.L $s[2],$Td1[2],$s[2] +|| XOR.L $s[3],$Td3[3],$s[3] +;;==================================================================== + ADD.D ${TEA},A0,${TEA} ; point to Td4 +|| ADD.D ${TEB},A0,${TEB} +|| EXTU $s[1],EXT3,24,$Td3[1] +|| EXTU $s[0],EXT1,24,$Td1[0] + LDBU *${TEB}[$Td3[1]],$Td3[1] ; Td3[s1>>24], t0 +|| LDBU *${TEA}[$Td1[0]],$Td1[0] ; Td1[s0>>8], t1 +|| XOR $s[2],$Td0[2],$s[2] ; modulo-scheduled +|| XOR $s[3],$Td0[3],$s[3] ; modulo-scheduled +|| EXTU $s[0],EXT0,24,$Td0[0] +|| EXTU $s[1],EXT0,24,$Td0[1] + LDBU *${TEA}[$Td0[0]],$Td0[0] ; Td0[s0], t0 +|| LDBU *${TEB}[$Td0[1]],$Td0[1] ; Td0[s1], t1 +|| EXTU $s[2],EXT2,24,$Td2[2] +|| EXTU $s[3],EXT2,24,$Td2[3] + LDBU *${TEA}[$Td2[2]],$Td2[2] ; Td2[s2>>16], t0 +|| LDBU *${TEB}[$Td2[3]],$Td2[3] ; Td2[s3>>16], t1 +|| EXTU $s[3],EXT1,24,$Td1[3] +|| EXTU $s[2],EXT3,24,$Td3[2] + LDBU *${TEB}[$Td1[3]],$Td1[3] ; Td1[s3>>8], t0 +|| LDBU *${TEA}[$Td3[2]],$Td3[2] ; Td3[s2>>24], t1 +|| EXTU $s[1],EXT1,24,$Td1[1] +|| EXTU $s[0],EXT3,24,$Td3[0] + LDBU *${TEB}[$Td1[1]],$Td1[1] ; Td1[s1>>8], t2 +|| LDBU *${TEA}[$Td3[0]],$Td3[0] ; Td3[s0>>24], t3 +|| EXTU $s[0],EXT2,24,$Td2[0] +|| EXTU $s[1],EXT2,24,$Td2[1] + LDBU *${TEA}[$Td2[0]],$Td2[0] ; Td2[s0>>16], t2 +|| LDBU *${TEB}[$Td2[1]],$Td2[1] ; Td2[s1>>16], t3 +|| EXTU $s[3],EXT3,24,$Td3[3] +|| EXTU $s[2],EXT1,24,$Td1[2] + LDBU *${TEB}[$Td3[3]],$Td3[3] ; Td3[s3>>24], t2 +|| LDBU *${TEA}[$Td1[2]],$Td1[2] ; Td1[s2>>8], t3 +|| EXTU $s[2],EXT0,24,$Td0[2] +|| EXTU $s[3],EXT0,24,$Td0[3] + LDBU *${TEA}[$Td0[2]],$Td0[2] ; Td0[s2], t2 +|| LDBU *${TEB}[$Td0[3]],$Td0[3] ; Td0[s3], t3 + + .if .BIG_ENDIAN + PACK2 $Td0[0],$Td1[3],$Td0[0] +|| PACK2 $Td0[1],$Td1[0],$Td0[1] + PACK2 $Td2[2],$Td3[1],$Td2[2] +|| PACK2 $Td2[3],$Td3[2],$Td2[3] + PACKL4 $Td0[0],$Td2[2],$Td0[0] +|| PACKL4 $Td0[1],$Td2[3],$Td0[1] + XOR $K[0],$Td0[0],$Td0[0] ; s[0] +|| XOR $K[1],$Td0[1],$Td0[1] ; s[1] + + PACK2 $Td0[2],$Td1[1],$Td0[2] +|| PACK2 $Td0[3],$Td1[2],$Td0[3] + PACK2 $Td2[0],$Td3[3],$Td2[0] +|| PACK2 $Td2[1],$Td3[0],$Td2[1] +|| BNOP RA + PACKL4 $Td0[2],$Td2[0],$Td0[2] +|| PACKL4 $Td0[3],$Td2[1],$Td0[3] + XOR $K[2],$Td0[2],$Td0[2] ; s[2] +|| XOR $K[3],$Td0[3],$Td0[3] ; s[3] + + MV $Td0[0],A9 +|| MV $Td0[1],A8 + MV $Td0[2],B9 +|| MV $Td0[3],B8 +|| [B2] STNDW A9:A8,*OUT++ + [B2] STNDW B9:B8,*OUT++ + .else + PACK2 $Td1[3],$Td0[0],$Td1[3] +|| PACK2 $Td1[0],$Td0[1],$Td1[0] + PACK2 $Td3[1],$Td2[2],$Td3[1] +|| PACK2 $Td3[2],$Td2[3],$Td3[2] + PACKL4 $Td3[1],$Td1[3],$Td1[3] +|| PACKL4 $Td3[2],$Td1[0],$Td1[0] + XOR $K[0],$Td1[3],$Td1[3] ; s[0] +|| XOR $K[1],$Td1[0],$Td1[0] ; s[1] + + PACK2 $Td1[1],$Td0[2],$Td1[1] +|| PACK2 $Td1[2],$Td0[3],$Td1[2] + PACK2 $Td3[3],$Td2[0],$Td3[3] +|| PACK2 $Td3[0],$Td2[1],$Td3[0] +|| BNOP RA + PACKL4 $Td3[3],$Td1[1],$Td1[1] +|| PACKL4 $Td3[0],$Td1[2],$Td1[2] + XOR $K[2],$Td1[1],$Td1[1] ; s[2] +|| XOR $K[3],$Td1[2],$Td1[2] ; s[3] + + MV $Td1[3],A8 +|| MV $Td1[0],A9 + MV $Td1[1],B8 +|| MV $Td1[2],B9 +|| [B2] STNDW A9:A8,*OUT++ + [B2] STNDW B9:B8,*OUT++ + .endif + .endasmfunc +___ +{ +my @K=(@K,@s); # extended key +my @Te4=map("B$_",(16..19)); + +my @Kx9=@Te0; # used in AES_set_decrypt_key +my @KxB=@Te1; +my @KxD=@Te2; +my @KxE=@Te3; + +$code.=<<___; + .asg OUT,BITS + + .global _AES_set_encrypt_key +_AES_set_encrypt_key: +__set_encrypt_key: + .asmfunc + MV INP,A0 +|| SHRU BITS,5,BITS ; 128-192-256 -> 4-6-8 +|| MV KEY,A1 + [!A0] B RA +||[!A0] MVK -1,RET +||[!A0] MVK 1,A1 ; only one B RA + [!A1] B RA +||[!A1] MVK -1,RET +||[!A1] MVK 0,A0 +|| MVK 0,B0 +|| MVK 0,A1 + [A0] LDNDW *INP++,A9:A8 +|| [A0] CMPEQ 4,BITS,B0 +|| [A0] CMPLT 3,BITS,A1 + [B0] B key128? +|| [A1] LDNDW *INP++,B9:B8 +|| [A0] CMPEQ 6,BITS,B0 +|| [A0] CMPLT 5,BITS,A1 + [B0] B key192? +|| [A1] LDNDW *INP++,B17:B16 +|| [A0] CMPEQ 8,BITS,B0 +|| [A0] CMPLT 7,BITS,A1 + [B0] B key256? +|| [A1] LDNDW *INP++,B19:B18 + + .if __TI_EABI__ + [A0] ADD 0,KEY,$KPA +|| [A0] ADD 4,KEY,$KPB +|| [A0] MVKL \$PCR_OFFSET(AES_Te4,_AES_set_encrypt_key),$TEA +|| [A0] ADDKPC _AES_set_encrypt_key,B6 + [A0] MVKH \$PCR_OFFSET(AES_Te4,_AES_set_encrypt_key),$TEA + [A0] ADD B6,$TEA,$TEA ; AES_Te4 + .else + [A0] ADD 0,KEY,$KPA +|| [A0] ADD 4,KEY,$KPB +|| [A0] MVKL (AES_Te4-_AES_set_encrypt_key),$TEA +|| [A0] ADDKPC _AES_set_encrypt_key,B6 + [A0] MVKH (AES_Te4-_AES_set_encrypt_key),$TEA + [A0] ADD B6,$TEA,$TEA ; AES_Te4 + .endif + NOP + NOP + + BNOP RA,5 +|| MVK -2,RET ; unknown bit lenght +|| MVK 0,B0 ; redundant +;;==================================================================== +;;==================================================================== +key128?: + .if .BIG_ENDIAN + MV A9,$K[0] +|| MV A8,$K[1] +|| MV B9,$Te4[2] +|| MV B8,$K[3] + .else + MV A8,$K[0] +|| MV A9,$K[1] +|| MV B8,$Te4[2] +|| MV B9,$K[3] + .endif + + MVK 256,A0 +|| MVK 9,B0 + + SPLOOPD 14 +|| MVC B0,ILC +|| MV $TEA,$TEB +|| ADD $TEA,A0,A30 ; rcon +;;==================================================================== + LDW *A30++[1],A31 ; rcon[i] +|| MV $Te4[2],$K[2] +|| EXTU $K[3],EXT1,24,$Te4[0] + LDBU *${TEB}[$Te4[0]],$Te4[0] +|| MV $K[3],A0 +|| EXTU $K[3],EXT2,24,$Te4[1] + LDBU *${TEB}[$Te4[1]],$Te4[1] +|| EXTU A0,EXT3,24,A0 +|| EXTU $K[3],EXT0,24,$Te4[3] + .if .BIG_ENDIAN + LDBU *${TEA}[A0],$Te4[3] +|| LDBU *${TEB}[$Te4[3]],A0 + .else + LDBU *${TEA}[A0],A0 +|| LDBU *${TEB}[$Te4[3]],$Te4[3] + .endif + + STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + + XOR A31,$K[0],$K[0] ; ^=rcon[i] + .if .BIG_ENDIAN + PACK2 $Te4[0],$Te4[1],$Te4[1] + PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[1],$Te4[3],$Te4[3] + .else + PACK2 $Te4[1],$Te4[0],$Te4[1] + PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[3],$Te4[1],$Te4[3] + .endif + XOR $Te4[3],$K[0],$Te4[0] ; K[0] + XOR $Te4[0],$K[1],$K[1] ; K[1] + MV $Te4[0],$K[0] +|| XOR $K[1],$K[2],$Te4[2] ; K[2] + XOR $Te4[2],$K[3],$K[3] ; K[3] + SPKERNEL +;;==================================================================== + BNOP RA + MV $Te4[2],$K[2] +|| STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + MVK 10,B0 ; rounds + STW B0,*++${KPB}[15] + MVK 0,RET +;;==================================================================== +;;==================================================================== +key192?: + .if .BIG_ENDIAN + MV A9,$K[0] +|| MV A8,$K[1] +|| MV B9,$K[2] +|| MV B8,$K[3] + MV B17,$Te4[2] +|| MV B16,$K[5] + .else + MV A8,$K[0] +|| MV A9,$K[1] +|| MV B8,$K[2] +|| MV B9,$K[3] + MV B16,$Te4[2] +|| MV B17,$K[5] + .endif + + MVK 256,A0 +|| MVK 6,B0 + MV $TEA,$TEB +|| ADD $TEA,A0,A30 ; rcon +;;==================================================================== +loop192?: + LDW *A30++[1],A31 ; rcon[i] +|| MV $Te4[2],$K[4] +|| EXTU $K[5],EXT1,24,$Te4[0] + LDBU *${TEB}[$Te4[0]],$Te4[0] +|| MV $K[5],A0 +|| EXTU $K[5],EXT2,24,$Te4[1] + LDBU *${TEB}[$Te4[1]],$Te4[1] +|| EXTU A0,EXT3,24,A0 +|| EXTU $K[5],EXT0,24,$Te4[3] + .if .BIG_ENDIAN + LDBU *${TEA}[A0],$Te4[3] +|| LDBU *${TEB}[$Te4[3]],A0 + .else + LDBU *${TEA}[A0],A0 +|| LDBU *${TEB}[$Te4[3]],$Te4[3] + .endif + + STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + STW $K[4],*$KPA++[2] +|| STW $K[5],*$KPB++[2] + + XOR A31,$K[0],$K[0] ; ^=rcon[i] + .if .BIG_ENDIAN + PACK2 $Te4[0],$Te4[1],$Te4[1] +|| PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[1],$Te4[3],$Te4[3] + .else + PACK2 $Te4[1],$Te4[0],$Te4[1] +|| PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[3],$Te4[1],$Te4[3] + .endif + BDEC loop192?,B0 +|| XOR $Te4[3],$K[0],$Te4[0] ; K[0] + XOR $Te4[0],$K[1],$K[1] ; K[1] + MV $Te4[0],$K[0] +|| XOR $K[1],$K[2],$Te4[2] ; K[2] + XOR $Te4[2],$K[3],$K[3] ; K[3] + MV $Te4[2],$K[2] +|| XOR $K[3],$K[4],$Te4[2] ; K[4] + XOR $Te4[2],$K[5],$K[5] ; K[5] +;;==================================================================== + BNOP RA + STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + MVK 12,B0 ; rounds + STW B0,*++${KPB}[7] + MVK 0,RET +;;==================================================================== +;;==================================================================== +key256?: + .if .BIG_ENDIAN + MV A9,$K[0] +|| MV A8,$K[1] +|| MV B9,$K[2] +|| MV B8,$K[3] + MV B17,$K[4] +|| MV B16,$K[5] +|| MV B19,$Te4[2] +|| MV B18,$K[7] + .else + MV A8,$K[0] +|| MV A9,$K[1] +|| MV B8,$K[2] +|| MV B9,$K[3] + MV B16,$K[4] +|| MV B17,$K[5] +|| MV B18,$Te4[2] +|| MV B19,$K[7] + .endif + + MVK 256,A0 +|| MVK 6,B0 + MV $TEA,$TEB +|| ADD $TEA,A0,A30 ; rcon +;;==================================================================== +loop256?: + LDW *A30++[1],A31 ; rcon[i] +|| MV $Te4[2],$K[6] +|| EXTU $K[7],EXT1,24,$Te4[0] + LDBU *${TEB}[$Te4[0]],$Te4[0] +|| MV $K[7],A0 +|| EXTU $K[7],EXT2,24,$Te4[1] + LDBU *${TEB}[$Te4[1]],$Te4[1] +|| EXTU A0,EXT3,24,A0 +|| EXTU $K[7],EXT0,24,$Te4[3] + .if .BIG_ENDIAN + LDBU *${TEA}[A0],$Te4[3] +|| LDBU *${TEB}[$Te4[3]],A0 + .else + LDBU *${TEA}[A0],A0 +|| LDBU *${TEB}[$Te4[3]],$Te4[3] + .endif + + STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + STW $K[4],*$KPA++[2] +|| STW $K[5],*$KPB++[2] + STW $K[6],*$KPA++[2] +|| STW $K[7],*$KPB++[2] +|| XOR A31,$K[0],$K[0] ; ^=rcon[i] + .if .BIG_ENDIAN + PACK2 $Te4[0],$Te4[1],$Te4[1] +|| PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[1],$Te4[3],$Te4[3] +||[!B0] B done256? + .else + PACK2 $Te4[1],$Te4[0],$Te4[1] +|| PACK2 $Te4[3],A0,$Te4[3] + PACKL4 $Te4[3],$Te4[1],$Te4[3] +||[!B0] B done256? + .endif + XOR $Te4[3],$K[0],$Te4[0] ; K[0] + XOR $Te4[0],$K[1],$K[1] ; K[1] + MV $Te4[0],$K[0] +|| XOR $K[1],$K[2],$Te4[2] ; K[2] + XOR $Te4[2],$K[3],$K[3] ; K[3] + + MV $Te4[2],$K[2] +|| [B0] EXTU $K[3],EXT0,24,$Te4[0] +|| [B0] SUB B0,1,B0 + LDBU *${TEB}[$Te4[0]],$Te4[0] +|| MV $K[3],A0 +|| EXTU $K[3],EXT1,24,$Te4[1] + LDBU *${TEB}[$Te4[1]],$Te4[1] +|| EXTU A0,EXT2,24,A0 +|| EXTU $K[3],EXT3,24,$Te4[3] + + .if .BIG_ENDIAN + LDBU *${TEA}[A0],$Te4[3] +|| LDBU *${TEB}[$Te4[3]],A0 + NOP 3 + PACK2 $Te4[0],$Te4[1],$Te4[1] + PACK2 $Te4[3],A0,$Te4[3] +|| B loop256? + PACKL4 $Te4[1],$Te4[3],$Te4[3] + .else + LDBU *${TEA}[A0],A0 +|| LDBU *${TEB}[$Te4[3]],$Te4[3] + NOP 3 + PACK2 $Te4[1],$Te4[0],$Te4[1] + PACK2 $Te4[3],A0,$Te4[3] +|| B loop256? + PACKL4 $Te4[3],$Te4[1],$Te4[3] + .endif + + XOR $Te4[3],$K[4],$Te4[0] ; K[4] + XOR $Te4[0],$K[5],$K[5] ; K[5] + MV $Te4[0],$K[4] +|| XOR $K[5],$K[6],$Te4[2] ; K[6] + XOR $Te4[2],$K[7],$K[7] ; K[7] +;;==================================================================== +done256?: + BNOP RA + STW $K[0],*$KPA++[2] +|| STW $K[1],*$KPB++[2] + STW $K[2],*$KPA++[2] +|| STW $K[3],*$KPB++[2] + MVK 14,B0 ; rounds + STW B0,*--${KPB}[1] + MVK 0,RET + .endasmfunc + + .global _AES_set_decrypt_key +_AES_set_decrypt_key: + .asmfunc + B __set_encrypt_key ; guarantee local call + MV KEY,B30 ; B30 is not modified + MV RA, B31 ; B31 is not modified + ADDKPC ret?,RA,2 +ret?: ; B0 holds rounds or zero + [!B0] BNOP B31 ; return if zero + [B0] SHL B0,4,A0 ; offset to last round key + [B0] SHRU B0,1,B1 + [B0] SUB B1,1,B1 + [B0] MVK 0x0000001B,B3 ; AES polynomial + [B0] MVKH 0x07000000,B3 + + SPLOOPD 9 ; flip round keys +|| MVC B1,ILC +|| MV B30,$KPA +|| ADD B30,A0,$KPB +|| MVK 16,A0 ; sizeof(round key) +;;==================================================================== + LDW *${KPA}[0],A16 +|| LDW *${KPB}[0],B16 + LDW *${KPA}[1],A17 +|| LDW *${KPB}[1],B17 + LDW *${KPA}[2],A18 +|| LDW *${KPB}[2],B18 + LDW *${KPA}[3],A19 +|| ADD $KPA,A0,$KPA +|| LDW *${KPB}[3],B19 +|| SUB $KPB,A0,$KPB + NOP + STW B16,*${KPA}[-4] +|| STW A16,*${KPB}[4] + STW B17,*${KPA}[-3] +|| STW A17,*${KPB}[5] + STW B18,*${KPA}[-2] +|| STW A18,*${KPB}[6] + STW B19,*${KPA}[-1] +|| STW A19,*${KPB}[7] + SPKERNEL +;;==================================================================== + SUB B0,1,B0 ; skip last round +|| ADD B30,A0,$KPA ; skip first round +|| ADD B30,A0,$KPB +|| MVC GFPGFR,B30 ; save GFPGFR + LDW *${KPA}[0],$K[0] +|| LDW *${KPB}[1],$K[1] +|| MVC B3,GFPGFR + LDW *${KPA}[2],$K[2] +|| LDW *${KPB}[3],$K[3] + MVK 0x00000909,A24 +|| MVK 0x00000B0B,B24 + MVKH 0x09090000,A24 +|| MVKH 0x0B0B0000,B24 + MVC B0,ILC +|| SUB B0,1,B0 + + GMPY4 $K[0],A24,$Kx9[0] ; ·0x09 +|| GMPY4 $K[1],A24,$Kx9[1] +|| MVK 0x00000D0D,A25 +|| MVK 0x00000E0E,B25 + GMPY4 $K[2],A24,$Kx9[2] +|| GMPY4 $K[3],A24,$Kx9[3] +|| MVKH 0x0D0D0000,A25 +|| MVKH 0x0E0E0000,B25 + + GMPY4 $K[0],B24,$KxB[0] ; ·0x0B +|| GMPY4 $K[1],B24,$KxB[1] + GMPY4 $K[2],B24,$KxB[2] +|| GMPY4 $K[3],B24,$KxB[3] + + SPLOOP 11 ; InvMixColumns +;;==================================================================== + GMPY4 $K[0],A25,$KxD[0] ; ·0x0D +|| GMPY4 $K[1],A25,$KxD[1] +|| SWAP2 $Kx9[0],$Kx9[0] ; rotate by 16 +|| SWAP2 $Kx9[1],$Kx9[1] +|| MV $K[0],$s[0] ; this or DINT +|| MV $K[1],$s[1] +|| [B0] LDW *${KPA}[4],$K[0] +|| [B0] LDW *${KPB}[5],$K[1] + GMPY4 $K[2],A25,$KxD[2] +|| GMPY4 $K[3],A25,$KxD[3] +|| SWAP2 $Kx9[2],$Kx9[2] +|| SWAP2 $Kx9[3],$Kx9[3] +|| MV $K[2],$s[2] +|| MV $K[3],$s[3] +|| [B0] LDW *${KPA}[6],$K[2] +|| [B0] LDW *${KPB}[7],$K[3] + + GMPY4 $s[0],B25,$KxE[0] ; ·0x0E +|| GMPY4 $s[1],B25,$KxE[1] +|| XOR $Kx9[0],$KxB[0],$KxB[0] +|| XOR $Kx9[1],$KxB[1],$KxB[1] + GMPY4 $s[2],B25,$KxE[2] +|| GMPY4 $s[3],B25,$KxE[3] +|| XOR $Kx9[2],$KxB[2],$KxB[2] +|| XOR $Kx9[3],$KxB[3],$KxB[3] + + ROTL $KxB[0],TBL3,$KxB[0] +|| ROTL $KxB[1],TBL3,$KxB[1] +|| SWAP2 $KxD[0],$KxD[0] ; rotate by 16 +|| SWAP2 $KxD[1],$KxD[1] + ROTL $KxB[2],TBL3,$KxB[2] +|| ROTL $KxB[3],TBL3,$KxB[3] +|| SWAP2 $KxD[2],$KxD[2] +|| SWAP2 $KxD[3],$KxD[3] + + XOR $KxE[0],$KxD[0],$KxE[0] +|| XOR $KxE[1],$KxD[1],$KxE[1] +|| [B0] GMPY4 $K[0],A24,$Kx9[0] ; ·0x09 +|| [B0] GMPY4 $K[1],A24,$Kx9[1] +|| ADDAW $KPA,4,$KPA + XOR $KxE[2],$KxD[2],$KxE[2] +|| XOR $KxE[3],$KxD[3],$KxE[3] +|| [B0] GMPY4 $K[2],A24,$Kx9[2] +|| [B0] GMPY4 $K[3],A24,$Kx9[3] +|| ADDAW $KPB,4,$KPB + + XOR $KxB[0],$KxE[0],$KxE[0] +|| XOR $KxB[1],$KxE[1],$KxE[1] +|| [B0] GMPY4 $K[0],B24,$KxB[0] ; ·0x0B +|| [B0] GMPY4 $K[1],B24,$KxB[1] + XOR $KxB[2],$KxE[2],$KxE[2] +|| XOR $KxB[3],$KxE[3],$KxE[3] +|| [B0] GMPY4 $K[2],B24,$KxB[2] +|| [B0] GMPY4 $K[3],B24,$KxB[3] +|| STW $KxE[0],*${KPA}[-4] +|| STW $KxE[1],*${KPB}[-3] + STW $KxE[2],*${KPA}[-2] +|| STW $KxE[3],*${KPB}[-1] +|| [B0] SUB B0,1,B0 + SPKERNEL +;;==================================================================== + BNOP B31,3 + MVC B30,GFPGFR ; restore GFPGFR(*) + MVK 0,RET + .endasmfunc +___ +# (*) Even though ABI doesn't specify GFPGFR as non-volatile, there +# are code samples out there that *assume* its default value. +} +{ +my ($inp,$out,$blocks,$key,$ivp)=("A4","B4","A6","B6","A8"); +$code.=<<___; + .global _AES_ctr32_encrypt +_AES_ctr32_encrypt: + .asmfunc + LDNDW *${ivp}[0],A31:A30 ; load counter value +|| MV $blocks,A2 ; reassign $blocks +|| DMV RA,$key,B27:B26 ; reassign RA and $key + LDNDW *${ivp}[1],B31:B30 +|| MVK 0,B2 ; don't let __encrypt load input +|| MVK 0,A1 ; and postpone writing output + .if .BIG_ENDIAN + NOP + .else + NOP 4 + SWAP2 B31,B31 ; keep least significant 32 bits + SWAP4 B31,B31 ; in host byte order + .endif +ctr32_loop?: |