sha/asm/keccak1600-armv8.pl: add hardware-assisted ARMv8.2 subroutines.
authorAndy Polyakov <appro@openssl.org>
Tue, 13 Feb 2018 20:59:15 +0000 (21:59 +0100)
committerAndy Polyakov <appro@openssl.org>
Mon, 19 Feb 2018 13:15:31 +0000 (14:15 +0100)
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5358)

crypto/sha/asm/keccak1600-armv8.pl

index 5e4cf77003ddd3b6c2c91988eb6e2a28541d582e..a32b8bae84c401a673b88fda908498a7fd7d7b6e 100755 (executable)
@@ -1,5 +1,5 @@
 #!/usr/bin/env perl
-# Copyright 2017 The OpenSSL Project Authors. All Rights Reserved.
+# Copyright 2017-2018 The OpenSSL Project Authors. All Rights Reserved.
 #
 # Licensed under the OpenSSL license (the "License").  You may not use
 # this file except in compliance with the License.  You can obtain a copy
 # it gets compensated by availability of rotate. Not to forget that
 # most processors achieve higher issue rate with scalar instructions.
 #
+# February 2018.
+#
+# Add hardware-assisted ARMv8.2 implementation. It's KECCAK_1X_ALT
+# variant with register permutation/rotation twist that allows to
+# eliminate copies to temporary registers. If you look closely you'll
+# notice that it uses only one lane of vector registers. The new
+# instructions effectively facilitate parallel hashing, which we don't
+# support [yet?]. But lowest-level core procedure is prepared for it.
+# The inner round is 67 [vector] instructions, so it's not actually
+# obvious that it will provide performance improvement [in serial
+# hash] as long as vector instructions issue rate is limited to 1 per
+# cycle...
+#
 ######################################################################
 # Numbers are cycles per processed byte.
 #
@@ -56,12 +69,6 @@ die "can't locate arm-xlate.pl";
 open OUT,"| \"$^X\" $xlate $flavour $output";
 *STDOUT=*OUT;
 
-my @A = map([ "x$_", "x".($_+1), "x".($_+2), "x".($_+3), "x".($_+4) ],
-            (0, 5, 10, 15, 20));
-   $A[3][3] = "x25"; # x18 is reserved
-
-my @C = map("x$_", (26,27,28,30));
-
 my @rhotates = ([  0,  1, 62, 28, 27 ],
                 [ 36, 44,  6, 55, 20 ],
                 [  3, 10, 43, 25, 39 ],
@@ -101,7 +108,15 @@ iotas:
        .quad   0x0000000080000001
        .quad   0x8000000080008008
 .size  iotas,.-iotas
+___
+                                                               {{{
+my @A = map([ "x$_", "x".($_+1), "x".($_+2), "x".($_+3), "x".($_+4) ],
+            (0, 5, 10, 15, 20));
+   $A[3][3] = "x25"; # x18 is reserved
+
+my @C = map("x$_", (26,27,28,30));
 
+$code.=<<___;
 .type  KeccakF1600_int,%function
 .align 5
 KeccakF1600_int:
@@ -510,9 +525,342 @@ SHA3_squeeze:
        ldp     x29,x30,[sp],#48
        ret
 .size  SHA3_squeeze,.-SHA3_squeeze
+___
+}                                                              }}}
+                                                               {{{
+my @A = map([ "v".$_.".16b", "v".($_+1).".16b", "v".($_+2).".16b",
+                             "v".($_+3).".16b", "v".($_+4).".16b" ],
+            (0, 5, 10, 15, 20));
+
+my @C = map("v$_.16b", (25..31));
+
+$code.=<<___;
+.type  KeccakF1600_ce,%function
+.align 5
+KeccakF1600_ce:
+       mov     x9,#6
+       adr     x10,iotas
+       b       .Loop_ce
+.align 4
+.Loop_ce:
+___
+for($i=0; $i<4; $i++) {
+$code.=<<___;
+       ////////////////////////////////////////////////// Theta
+       eor3    $C[0],$A[0][0],$A[1][0],$A[2][0]
+       eor3    $C[1],$A[0][1],$A[1][1],$A[2][1]
+       eor3    $C[2],$A[0][2],$A[1][2],$A[2][2]
+       eor3    $C[3],$A[0][3],$A[1][3],$A[2][3]
+       eor3    $C[4],$A[0][4],$A[1][4],$A[2][4]
+       eor3    $C[0],$C[0],   $A[3][0],$A[4][0]
+       eor3    $C[1],$C[1],   $A[3][1],$A[4][1]
+       eor3    $C[2],$C[2],   $A[3][2],$A[4][2]
+       eor3    $C[3],$C[3],   $A[3][3],$A[4][3]
+       eor3    $C[4],$C[4],   $A[3][4],$A[4][4]
+
+       rax1    $C[5],$C[0],$C[2]                       // D[1]
+       rax1    $C[6],$C[1],$C[3]                       // D[2]
+       rax1    $C[2],$C[2],$C[4]                       // D[3]
+       rax1    $C[3],$C[3],$C[0]                       // D[4]
+       rax1    $C[4],$C[4],$C[1]                       // D[0]
+
+       ////////////////////////////////////////////////// Theta+Rho+Pi
+       xar     $C[0],   $A[1][1],$C[5],#64-$rhotates[1][1]     // C[0]=A[0][1]
+       xar     $A[1][1],$A[1][4],$C[3],#64-$rhotates[1][4]
+       xar     $A[1][4],$A[4][2],$C[6],#64-$rhotates[4][2]
+       xar     $A[4][2],$A[2][4],$C[3],#64-$rhotates[2][4]
+       xar     $A[2][4],$A[4][0],$C[4],#64-$rhotates[4][0]
+
+       xar     $A[4][0],$A[0][2],$C[6],#64-$rhotates[0][2]
+
+       xar     $A[0][2],$A[2][2],$C[6],#64-$rhotates[2][2]
+       xar     $A[2][2],$A[2][3],$C[2],#64-$rhotates[2][3]
+       xar     $A[2][3],$A[3][4],$C[3],#64-$rhotates[3][4]
+       xar     $A[3][4],$A[4][3],$C[2],#64-$rhotates[4][3]
+       xar     $A[4][3],$A[3][0],$C[4],#64-$rhotates[3][0]
+
+       xar     $A[3][0],$A[0][4],$C[3],#64-$rhotates[0][4]
+
+       eor     $A[0][0],$A[0][0],$C[4]
+       ldr     x11,[x10],#8
+
+       xar     $C[1]   ,$A[3][3],$C[2],#64-$rhotates[3][3]     // C[1]=A[0][3]
+       xar     $A[3][3],$A[3][2],$C[6],#64-$rhotates[3][2]
+       xar     $A[3][2],$A[2][1],$C[5],#64-$rhotates[2][1]
+       xar     $A[2][1],$A[1][2],$C[6],#64-$rhotates[1][2]
+       xar     $A[1][2],$A[2][0],$C[4],#64-$rhotates[2][0]
+
+       xar     $A[2][0],$A[0][1],$C[5],#64-$rhotates[0][1]     // *
+
+       xar     $A[0][4],$A[4][4],$C[3],#64-$rhotates[4][4]
+       xar     $A[4][4],$A[4][1],$C[5],#64-$rhotates[4][1]
+       xar     $A[4][1],$A[1][3],$C[2],#64-$rhotates[1][3]
+       xar     $A[1][3],$A[3][1],$C[5],#64-$rhotates[3][1]
+       xar     $A[3][1],$A[1][0],$C[4],#64-$rhotates[1][0]
+
+       xar     $A[1][0],$A[0][3],$C[2],#64-$rhotates[0][3]     // *
+
+       ////////////////////////////////////////////////// Chi+Iota
+       dup     $C[6],x11                               // borrow C[6]
+       bcax    $C[3],   $A[0][0],$A[0][2],$C[0]        // *
+       bcax    $A[0][1],$C[0],   $C[1],   $A[0][2]     // *
+       bcax    $A[0][2],$A[0][2],$A[0][4],$C[1]
+       bcax    $A[0][3],$C[1],   $A[0][0],$A[0][4]
+       bcax    $A[0][4],$A[0][4],$C[0],   $A[0][0]
+
+       bcax    $C[0],   $A[1][0],$A[1][2],$A[1][1]     // *
+       bcax    $C[1],   $A[1][1],$A[1][3],$A[1][2]     // *
+       bcax    $A[1][2],$A[1][2],$A[1][4],$A[1][3]
+       bcax    $A[1][3],$A[1][3],$A[1][0],$A[1][4]
+       bcax    $A[1][4],$A[1][4],$A[1][1],$A[1][0]
+
+       eor     $A[0][0],$C[3],$C[6]                    // Iota
+
+       bcax    $C[2],   $A[2][0],$A[2][2],$A[2][1]     // *
+       bcax    $C[3],   $A[2][1],$A[2][3],$A[2][2]     // *
+       bcax    $A[2][2],$A[2][2],$A[2][4],$A[2][3]
+       bcax    $A[2][3],$A[2][3],$A[2][0],$A[2][4]
+       bcax    $A[2][4],$A[2][4],$A[2][1],$A[2][0]
+
+       bcax    $A[2][0],$A[3][0],$A[3][2],$A[3][1]     // *
+       bcax    $A[2][1],$A[3][1],$A[3][3],$A[3][2]     // *
+       bcax    $A[3][2],$A[3][2],$A[3][4],$A[3][3]
+       bcax    $A[3][3],$A[3][3],$A[3][0],$A[3][4]
+       bcax    $A[3][4],$A[3][4],$A[3][1],$A[3][0]
+
+       bcax    $A[3][0],$A[4][0],$A[4][2],$A[4][1]     // *
+       bcax    $A[3][1],$A[4][1],$A[4][3],$A[4][2]     // *
+       bcax    $A[4][2],$A[4][2],$A[4][4],$A[4][3]
+       bcax    $A[4][3],$A[4][3],$A[4][0],$A[4][4]
+       bcax    $A[4][4],$A[4][4],$A[4][1],$A[4][0]
+___
+       ($A[1][0],$A[1][1], $C[0],$C[1])
+       =  ($C[0],$C[1],    $A[1][0],$A[1][1]);
+       ($A[2][0],$A[2][1], $A[3][0],$A[3][1], $A[4][0],$A[4][1], $C[2],$C[3])
+       =  ($C[2],$C[3],    $A[2][0],$A[2][1], $A[3][0],$A[3][1], $A[4][0],$A[4][1]);
+}
+$code.=<<___;
+       subs    x9,x9,#1
+       bne     .Loop_ce
+
+       ret
+.size  KeccakF1600_ce,.-KeccakF1600_ce
+
+.type  KeccakF1600_cext,%function
+.align 5
+KeccakF1600_cext:
+       stp     x29,x30,[sp,#-80]!
+       add     x29,sp,#0
+       stp     d8,d9,[sp,#16]          // per ABI requirement
+       stp     d10,d11,[sp,#32]
+       stp     d12,d13,[sp,#48]
+       stp     d14,d15,[sp,#64]
+___
+for($i=0; $i<24; $i+=2) {              # load A[5][5]
+my $j=$i+1;
+$code.=<<___;
+       ldp     d$i,d$j,[x0,#8*$i]
+___
+}
+$code.=<<___;
+       ldr     d24,[x0,#8*$i]
+       bl      KeccakF1600_ce
+       ldr     x30,[sp,#8]
+___
+for($i=0; $i<24; $i+=2) {              # store A[5][5]
+my $j=$i+1;
+$code.=<<___;
+       stp     d$i,d$j,[x0,#8*$i]
+___
+}
+$code.=<<___;
+       str     d24,[x0,#8*$i]
+
+       ldp     d8,d9,[sp,#16]
+       ldp     d10,d11,[sp,#32]
+       ldp     d12,d13,[sp,#48]
+       ldp     d14,d15,[sp,#64]
+       ldr     x29,[sp],#80
+       ret
+.size  KeccakF1600_cext,.-KeccakF1600_cext
+___
+
+{
+my ($ctx,$inp,$len,$bsz) = map("x$_",(0..3));
+
+$code.=<<___;
+.globl SHA3_absorb_cext
+.type  SHA3_absorb_cext,%function
+.align 5
+SHA3_absorb_cext:
+       stp     x29,x30,[sp,#-80]!
+       add     x29,sp,#0
+       stp     d8,d9,[sp,#16]          // per ABI requirement
+       stp     d10,d11,[sp,#32]
+       stp     d12,d13,[sp,#48]
+       stp     d14,d15,[sp,#64]
+___
+for($i=0; $i<24; $i+=2) {              # load A[5][5]
+my $j=$i+1;
+$code.=<<___;
+       ldp     d$i,d$j,[x0,#8*$i]
+___
+}
+$code.=<<___;
+       ldr     d24,[x0,#8*$i]
+       b       .Loop_absorb_ce
+
+.align 4
+.Loop_absorb_ce:
+       subs    $len,$len,$bsz          // len - bsz
+       blo     .Labsorbed_ce
+___
+for (my $i=0; $i<24; $i+=2) {
+my $j = $i+1;
+$code.=<<___;
+       ldr     d31,[$inp],#8           // *inp++
+#ifdef __AARCH64EB__
+       rev64   v31.16b,v31.16b
+#endif
+       eor     $A[$i/5][$i%5],$A[$i/5][$i%5],v31.16b
+       cmp     $bsz,#8*($i+2)
+       blo     .Lprocess_block_ce
+       ldr     d31,[$inp],#8           // *inp++
+#ifdef __AARCH64EB__
+       rev     v31.16b,v31.16b
+#endif
+       eor     $A[$j/5][$j%5],$A[$j/5][$j%5],v31.16b
+       beq     .Lprocess_block_ce
+___
+}
+$code.=<<___;
+       ldr     d31,[$inp],#8           // *inp++
+#ifdef __AARCH64EB__
+       rev     v31.16b,v31.16b
+#endif
+       eor     $A[4][4],$A[4][4],v31.16b
+
+.Lprocess_block_ce:
+
+       bl      KeccakF1600_ce
+
+       b       .Loop_absorb_ce
+
+.align 4
+.Labsorbed_ce:
+___
+for($i=0; $i<24; $i+=2) {              # store A[5][5]
+my $j=$i+1;
+$code.=<<___;
+       stp     d$i,d$j,[x0,#8*$i]
+___
+}
+$code.=<<___;
+       str     d24,[x0,#8*$i]
+       add     x0,$len,$bsz            // return value
+
+       ldp     d8,d9,[sp,#16]
+       ldp     d10,d11,[sp,#32]
+       ldp     d12,d13,[sp,#48]
+       ldp     d14,d15,[sp,#64]
+       ldp     x29,x30,[sp],#80
+       ret
+.size  SHA3_absorb_cext,.-SHA3_absorb_cext
+___
+}
+{
+my ($ctx,$out,$len,$bsz) = map("x$_",(0..3));
+$code.=<<___;
+.globl SHA3_squeeze_cext
+.type  SHA3_squeeze_cext,%function
+.align 5
+SHA3_squeeze_cext:
+       stp     x29,x30,[sp,#-16]!
+       add     x29,sp,#0
+       mov     x9,$ctx
+       mov     x10,$bsz
+
+.Loop_squeeze_ce:
+       ldr     x4,[x9],#8
+       cmp     $len,#8
+       blo     .Lsqueeze_tail_ce
+#ifdef __AARCH64EB__
+       rev     x4,x4
+#endif
+       str     x4,[$out],#8
+       beq     .Lsqueeze_done_ce
+
+       sub     $len,$len,#8
+       subs    x10,x10,#8
+       bhi     .Loop_squeeze_ce
+
+       bl      KeccakF1600_cext
+       ldr     x30,[sp,#8]
+       mov     x9,$ctx
+       mov     x10,$bsz
+       b       .Loop_squeeze_ce
+
+.align 4
+.Lsqueeze_tail_ce:
+       strb    w4,[$out],#1
+       lsr     x4,x4,#8
+       subs    $len,$len,#1
+       beq     .Lsqueeze_done_ce
+       strb    w4,[$out],#1
+       lsr     x4,x4,#8
+       subs    $len,$len,#1
+       beq     .Lsqueeze_done_ce
+       strb    w4,[$out],#1
+       lsr     x4,x4,#8
+       subs    $len,$len,#1
+       beq     .Lsqueeze_done_ce
+       strb    w4,[$out],#1
+       lsr     x4,x4,#8
+       subs    $len,$len,#1
+       beq     .Lsqueeze_done_ce
+       strb    w4,[$out],#1
+       lsr     x4,x4,#8
+       subs    $len,$len,#1
+       beq     .Lsqueeze_done_ce
+       strb    w4,[$out],#1
+       lsr     x4,x4,#8
+       subs    $len,$len,#1
+       beq     .Lsqueeze_done_ce
+       strb    w4,[$out],#1
+
+.Lsqueeze_done_ce:
+       ldr     x29,[sp],#16
+       ret
+.size  SHA3_squeeze_cext,.-SHA3_squeeze_cext
+___
+}                                                              }}}
+$code.=<<___;
 .asciz "Keccak-1600 absorb and squeeze for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
 ___
+
+{   my  %opcode = (
+       "rax1"  => 0xce608c00,  "eor3"  => 0xce000000,
+       "bcax"  => 0xce200000,  "xar"   => 0xce800000   );
+
+    sub unsha3 {
+       my ($mnemonic,$arg)=@_;
+
+       $arg =~ m/[qv]([0-9]+)[^,]*,\s*[qv]([0-9]+)[^,]*(?:,\s*[qv]([0-9]+)[^,]*(?:,\s*[qv#]([0-9\-]+))?)?/
+       &&
+       sprintf ".inst\t0x%08x\t//%s %s",
+                       $opcode{$mnemonic}|$1|($2<<5)|($3<<16)|(eval($4)<<10),
+                       $mnemonic,$arg;
+    }
+}
+
+foreach(split("\n",$code)) {
+
+       s/\`([^\`]*)\`/eval($1)/ge;
+
+       m/\bdup\b/ and s/\.16b/.2d/g    or
+       s/\b(eor3|rax1|xar|bcax)\s+(v.*)/unsha3($1,$2)/ge;
+
+       print $_,"\n";
 }
 
-print $code;
 close STDOUT;