3 # ====================================================================
4 # Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
5 # project. The module is, however, dual licensed under OpenSSL and
6 # CRYPTOGAMS licenses depending on where you obtain it. For further
7 # details see http://www.openssl.org/~appro/cryptogams/.
8 # ====================================================================
10 # SHA256 block procedure for ARMv4. May 2007.
12 # Performance is ~2x better than gcc 3.4 generated code and in "abso-
13 # lute" terms is ~2250 cycles per 64-byte block or ~35 cycles per
14 # byte [on single-issue Xscale PXA250 core].
18 # Rescheduling for dual-issue pipeline resulted in 22% improvement on
19 # Cortex A8 core and ~20 cycles per processed byte.
21 while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
22 open STDOUT,">$output";
36 @V=($A,$B,$C,$D,$E,$F,$G,$H);
46 my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_;
48 $code.=<<___ if ($i<16);
49 ldrb $T1,[$inp,#3] @ $i
54 orr $T1,$T1,$t1,lsl#16
55 orr $T1,$T1,$t0,lsl#24
56 `"str $inp,[sp,#17*4]" if ($i==15)`
59 ldr $t2,[$Ktbl],#4 @ *K256++
60 mov $t0,$e,ror#$Sigma1[0]
61 str $T1,[sp,#`$i%16`*4]
62 eor $t0,$t0,$e,ror#$Sigma1[1]
64 eor $t0,$t0,$e,ror#$Sigma1[2] @ Sigma1(e)
67 eor $t1,$t1,$g @ Ch(e,f,g)
69 mov $h,$a,ror#$Sigma0[0]
71 eor $h,$h,$a,ror#$Sigma0[1]
73 eor $h,$h,$a,ror#$Sigma0[2] @ Sigma0(a)
78 orr $t0,$t0,$t1 @ Maj(a,b,c)
85 my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_;
88 ldr $t1,[sp,#`($i+1)%16`*4] @ $i
89 ldr $t2,[sp,#`($i+14)%16`*4]
90 ldr $T1,[sp,#`($i+0)%16`*4]
91 mov $t0,$t1,ror#$sigma0[0]
92 ldr $inp,[sp,#`($i+9)%16`*4]
93 eor $t0,$t0,$t1,ror#$sigma0[1]
94 eor $t0,$t0,$t1,lsr#$sigma0[2] @ sigma0(X[i+1])
95 mov $t1,$t2,ror#$sigma1[0]
97 eor $t1,$t1,$t2,ror#$sigma1[1]
99 eor $t1,$t1,$t2,lsr#$sigma1[2] @ sigma1(X[i+14])
112 .word 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
113 .word 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5
114 .word 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3
115 .word 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174
116 .word 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc
117 .word 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da
118 .word 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7
119 .word 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967
120 .word 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13
121 .word 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85
122 .word 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3
123 .word 0xd192e819,0xd6990624,0xf40e3585,0x106aa070
124 .word 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5
125 .word 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3
126 .word 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
127 .word 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
130 .global sha256_block_data_order
131 .type sha256_block_data_order,%function
132 sha256_block_data_order:
133 sub r3,pc,#8 @ sha256_block_data_order
134 add $len,$inp,$len,lsl#6 @ len to point at the end of inp
135 stmdb sp!,{$ctx,$inp,$len,r4-r12,lr}
136 ldmia $ctx,{$A,$B,$C,$D,$E,$F,$G,$H}
137 sub $Ktbl,r3,#256 @ K256
138 sub sp,sp,#16*4 @ alloca(X[16])
141 for($i=0;$i<16;$i++) { &BODY_00_15($i,@V); unshift(@V,pop(@V)); }
142 $code.=".Lrounds_16_xx:\n";
143 for (;$i<32;$i++) { &BODY_16_XX($i,@V); unshift(@V,pop(@V)); }
149 ldr $T1,[sp,#16*4] @ pull ctx
164 ldr $inp,[sp,#17*4] @ pull inp
165 ldr $t2,[sp,#18*4] @ pull inp+len
168 stmia $T1,{$A,$B,$C,$D,$E,$F,$G,$H}
170 sub $Ktbl,$Ktbl,#256 @ rewind Ktbl
173 add sp,sp,#`16+3`*4 @ destroy frame
174 ldmia sp!,{r4-r12,lr}
176 moveq pc,lr @ be binary compatible with V4, yet
177 bx lr @ interoperable with Thumb ISA:-)
178 .size sha256_block_data_order,.-sha256_block_data_order
179 .asciz "SHA256 block transform for ARMv4, CRYPTOGAMS by <appro\@openssl.org>"
183 $code =~ s/\`([^\`]*)\`/eval $1/gem;
184 $code =~ s/\bbx\s+lr\b/.word\t0xe12fff1e/gm; # make it possible to compile with -march=armv4
186 close STDOUT; # enforce flush