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 # SHA1 block procedure for MIPS.
12 # Performance improvement is 30% on unaligned input. The "secret" is
13 # to deploy lwl/lwr pair to load unaligned input. One could have
14 # vectorized Xupdate on MIPSIII/IV, but the goal was to code MIPS32-
15 # compatible subroutine. There is room for minor optimization on
16 # little-endian platforms...
20 # Add MIPS32r2 code (>25% less instructions).
22 ######################################################################
23 # There is a number of MIPS ABI in use, O32 and N32/64 are most
24 # widely used. Then there is a new contender: NUBI. It appears that if
25 # one picks the latter, it's possible to arrange code in ABI neutral
26 # manner. Therefore let's stick to NUBI register layout:
28 ($zero,$at,$t0,$t1,$t2)=map("\$$_",(0..2,24,25));
29 ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
30 ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7,$s8,$s9,$s10,$s11)=map("\$$_",(12..23));
31 ($gp,$tp,$sp,$fp,$ra)=map("\$$_",(3,28..31));
33 # The return value is placed in $a0. Following coding rules facilitate
36 # - never ever touch $tp, "thread pointer", former $gp;
37 # - copy return value to $t0, former $v0 [or to $a0 if you're adapting
39 # - on O32 populate $a4-$a7 with 'lw $aN,4*N($sp)' if necessary;
41 # For reference here is register layout for N32/64 MIPS ABIs:
43 # ($zero,$at,$v0,$v1)=map("\$$_",(0..3));
44 # ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
45 # ($t0,$t1,$t2,$t3,$t8,$t9)=map("\$$_",(12..15,24,25));
46 # ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23));
47 # ($gp,$sp,$fp,$ra)=map("\$$_",(28..31));
49 $flavour = shift || "o32"; # supported flavours are o32,n32,64,nubi32,nubi64
51 if ($flavour =~ /64|n32/i) {
52 $PTR_ADD="dadd"; # incidentally works even on n32
53 $PTR_SUB="dsub"; # incidentally works even on n32
56 $PTR_SLL="dsll"; # incidentally works even on n32
69 ######################################################################
71 $big_endian=(`echo MIPSEL | $ENV{CC} -E -P -`=~/MIPSEL/)?1:0;
73 for (@ARGV) { $output=$_ if (/^\w[\w\-]*\.\w+$/); }
74 open STDOUT,">$output";
76 if (!defined($big_endian))
77 { $big_endian=(unpack('L',pack('N',1))==1); }
79 # offsets of the Most and Least Significant Bytes
83 @X=map("\$$_",(8..23)); # a4-a7,s0-s11
92 $E="\$24"; @V=($A,$B,$C,$D,$E);
94 $t1=$num; # $num is offloaded to stack
99 my ($i,$a,$b,$c,$d,$e)=@_;
101 $code.=<<___ if (!$big_endian);
102 #if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2)
103 wsbh @X[$i],@X[$i] # byte swap($i)
104 rotr @X[$i],@X[$i],16
106 srl $t0,@X[$i],24 # byte swap($i)
108 andi $t2,@X[$i],0xFF00
118 #if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2)
122 lwl @X[$j],$j*4+$MSB($inp)
125 lwr @X[$j],$j*4+$LSB($inp)
131 lwl @X[$j],$j*4+$MSB($inp)
134 lwr @X[$j],$j*4+$LSB($inp)
151 my ($i,$a,$b,$c,$d,$e)=@_;
154 $code.=<<___ if (!$big_endian && $i==15);
155 #if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2)
156 wsbh @X[$i],@X[$i] # byte swap($i)
157 rotr @X[$i],@X[$i],16
159 srl $t0,@X[$i],24 # byte swap($i)
161 andi $t2,@X[$i],0xFF00
171 #if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2)
173 xor @X[$j%16],@X[($j+2)%16]
176 xor @X[$j%16],@X[($j+8)%16]
179 xor @X[$j%16],@X[($j+13)%16]
182 rotr @X[$j%16],@X[$j%16],31
186 xor @X[$j%16],@X[($j+2)%16]
191 xor @X[$j%16],@X[($j+8)%16]
194 xor @X[$j%16],@X[($j+13)%16]
198 addu @X[$j%16],@X[$j%16]
210 my ($i,$a,$b,$c,$d,$e)=@_;
212 $code.=<<___ if ($i<79);
213 #if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2)
214 xor @X[$j%16],@X[($j+2)%16]
217 xor @X[$j%16],@X[($j+8)%16]
220 xor @X[$j%16],@X[($j+13)%16]
223 rotr @X[$j%16],@X[$j%16],31
227 xor @X[$j%16],@X[($j+2)%16]
232 xor @X[$j%16],@X[($j+8)%16]
235 xor @X[$j%16],@X[($j+13)%16]
239 addu @X[$j%16],@X[$j%16]
247 $code.=<<___ if ($i==79);
248 #if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2)
285 my ($i,$a,$b,$c,$d,$e)=@_;
287 $code.=<<___ if ($i<79);
288 #if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2)
291 xor @X[$j%16],@X[($j+2)%16]
294 xor @X[$j%16],@X[($j+8)%16]
297 xor @X[$j%16],@X[($j+13)%16]
300 rotr @X[$j%16],@X[$j%16],31
304 xor @X[$j%16],@X[($j+2)%16]
309 xor @X[$j%16],@X[($j+8)%16]
312 xor @X[$j%16],@X[($j+13)%16]
317 addu @X[$j%16],@X[$j%16]
328 $FRAMESIZE=16; # large enough to accomodate NUBI saved registers
329 $SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? 0xc0fff008 : 0xc0ff0000;
332 #ifdef OPENSSL_FIPSCANISTER
333 # include <openssl/fipssyms.h>
336 #if defined(__mips_smartmips) && !defined(_MIPS_ARCH_MIPS32R2)
337 #define _MIPS_ARCH_MIPS32R2
345 .globl sha1_block_data_order
346 .ent sha1_block_data_order
347 sha1_block_data_order:
348 .frame $sp,$FRAMESIZE*$SZREG,$ra
349 .mask $SAVED_REGS_MASK,-$SZREG
351 $PTR_SUB $sp,$FRAMESIZE*$SZREG
352 $REG_S $ra,($FRAMESIZE-1)*$SZREG($sp)
353 $REG_S $fp,($FRAMESIZE-2)*$SZREG($sp)
354 $REG_S $s11,($FRAMESIZE-3)*$SZREG($sp)
355 $REG_S $s10,($FRAMESIZE-4)*$SZREG($sp)
356 $REG_S $s9,($FRAMESIZE-5)*$SZREG($sp)
357 $REG_S $s8,($FRAMESIZE-6)*$SZREG($sp)
358 $REG_S $s7,($FRAMESIZE-7)*$SZREG($sp)
359 $REG_S $s6,($FRAMESIZE-8)*$SZREG($sp)
360 $REG_S $s5,($FRAMESIZE-9)*$SZREG($sp)
361 $REG_S $s4,($FRAMESIZE-10)*$SZREG($sp)
363 $code.=<<___ if ($flavour =~ /nubi/i); # optimize non-nubi prologue
364 $REG_S $s3,($FRAMESIZE-11)*$SZREG($sp)
365 $REG_S $s2,($FRAMESIZE-12)*$SZREG($sp)
366 $REG_S $s1,($FRAMESIZE-13)*$SZREG($sp)
367 $REG_S $s0,($FRAMESIZE-14)*$SZREG($sp)
368 $REG_S $gp,($FRAMESIZE-15)*$SZREG($sp)
386 ori $K,0x7999 # K_00_19
388 for ($i=0;$i<15;$i++) { &BODY_00_14($i,@V); unshift(@V,pop(@V)); }
389 for (;$i<20;$i++) { &BODY_15_19($i,@V); unshift(@V,pop(@V)); }
392 ori $K,0xeba1 # K_20_39
394 for (;$i<40;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
397 ori $K,0xbcdc # K_40_59
399 for (;$i<60;$i++) { &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
402 ori $K,0xc1d6 # K_60_79
404 for (;$i<80;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
424 $REG_L $ra,($FRAMESIZE-1)*$SZREG($sp)
425 $REG_L $fp,($FRAMESIZE-2)*$SZREG($sp)
426 $REG_L $s11,($FRAMESIZE-3)*$SZREG($sp)
427 $REG_L $s10,($FRAMESIZE-4)*$SZREG($sp)
428 $REG_L $s9,($FRAMESIZE-5)*$SZREG($sp)
429 $REG_L $s8,($FRAMESIZE-6)*$SZREG($sp)
430 $REG_L $s7,($FRAMESIZE-7)*$SZREG($sp)
431 $REG_L $s6,($FRAMESIZE-8)*$SZREG($sp)
432 $REG_L $s5,($FRAMESIZE-9)*$SZREG($sp)
433 $REG_L $s4,($FRAMESIZE-10)*$SZREG($sp)
435 $code.=<<___ if ($flavour =~ /nubi/i);
436 $REG_L $s3,($FRAMESIZE-11)*$SZREG($sp)
437 $REG_L $s2,($FRAMESIZE-12)*$SZREG($sp)
438 $REG_L $s1,($FRAMESIZE-13)*$SZREG($sp)
439 $REG_L $s0,($FRAMESIZE-14)*$SZREG($sp)
440 $REG_L $gp,($FRAMESIZE-15)*$SZREG($sp)
444 $PTR_ADD $sp,$FRAMESIZE*$SZREG
445 .end sha1_block_data_order
447 .asciiz "SHA1 for MIPS, CRYPTOGAMS by <appro\@openssl.org>"