ec: powerpc64le: Add asm implementation of felem_{square,mul}

Add an assembly implementation of felem_{square,mul}, which will be
implemented whenever Altivec support is present and the core implements
ISA 3.0 (Power 9) or greater.

Signed-off-by: Rohan McLure <rohanmclure@linux.ibm.com>

Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Todd Short <todd.short@me.com>
(Merged from https://github.com/openssl/openssl/pull/21471)

3 files changed, 368 insertions, 2 deletions

diff --git a/crypto/ec/asm/ecp_nistp384-ppc64.pl b/crypto/ec/asm/ecp_nistp384-ppc64.pl
new file mode 100755
index 0000000000..3f86b391af
--- /dev/null
+++ b/crypto/ec/asm/ecp_nistp384-ppc64.pl
@@ -0,0 +1,355 @@
+#! /usr/bin/env perl
+# Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the Apache License 2.0 (the "License").  You may not use
+# this file except in compliance with the License.  You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+#
+# ====================================================================
+# Written by Rohan McLure <rmclure@linux.ibm.com> for the OpenSSL
+# project.
+# ====================================================================
+#
+# p384 lower-level primitives for PPC64 using vector instructions.
+#
+
+use strict;
+use warnings;
+
+my $flavour = shift;
+my $output = "";
+while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {}
+if (!$output) {
+    $output = "-";
+}
+
+my ($xlate, $dir);
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
+die "can't locate ppc-xlate.pl";
+
+open OUT,"| \"$^X\" $xlate $flavour $output";
+*STDOUT=*OUT;
+
+my $code = "";
+
+my ($sp, $outp, $savelr, $savesp) = ("r1", "r3", "r10", "r12");
+
+my $vzero = "v32";
+
+sub startproc($)
+{
+    my ($name) = @_;
+
+    $code.=<<___;
+    .globl ${name}
+    .align 5
+${name}:
+
+___
+}
+
+sub endproc($)
+{
+    my ($name) = @_;
+
+    $code.=<<___;
+    blr
+        .size ${name},.-${name}
+
+___
+}
+
+
+sub push_vrs($$)
+{
+    my ($min, $max) = @_;
+
+    my $count = $max - $min + 1;
+
+    $code.=<<___;
+    mr      $savesp,$sp
+    stdu        $sp,-16*`$count+1`($sp)
+
+___
+        for (my $i = $min; $i <= $max; $i++) {
+            my $mult = $max - $i + 1;
+            $code.=<<___;
+    stxv        $i,-16*$mult($savesp)
+___
+
+    }
+
+    $code.=<<___;
+
+___
+}
+
+sub pop_vrs($$)
+{
+    my ($min, $max) = @_;
+
+    $code.=<<___;
+    ld      $savesp,0($sp)
+___
+    for (my $i = $min; $i <= $max; $i++) {
+        my $mult = $max - $i + 1;
+        $code.=<<___;
+    lxv     $i,-16*$mult($savesp)
+___
+    }
+
+    $code.=<<___;
+    mr      $sp,$savesp
+
+___
+}
+
+sub load_vrs($$)
+{
+    my ($pointer, $reg_list) = @_;
+
+    for (my $i = 0; $i <= 6; $i++) {
+        my $offset = $i * 8;
+        $code.=<<___;
+    lxsd        $reg_list->[$i],$offset($pointer)
+___
+    }
+
+    $code.=<<___;
+
+___
+}
+
+sub store_vrs($$)
+{
+    my ($pointer, $reg_list) = @_;
+
+    for (my $i = 0; $i <= 12; $i++) {
+        my $offset = $i * 16;
+        $code.=<<___;
+    stxv        $reg_list->[$i],$offset($pointer)
+___
+    }
+
+    $code.=<<___;
+
+___
+}
+
+$code.=<<___;
+.machine    "any"
+.text
+
+___
+
+{
+    # mul/square common
+    my ($t1, $t2, $t3, $t4) = ("v33", "v34", "v42", "v43");
+    my ($zero, $one) = ("r8", "r9");
+    my $out = "v51";
+
+    {
+        #
+        # p384_felem_mul
+        #
+
+        my ($in1p, $in2p) = ("r4", "r5");
+        my @in1 = map("v$_",(44..50));
+        my @in2 = map("v$_",(35..41));
+
+        startproc("p384_felem_mul");
+
+        push_vrs(52, 63);
+
+        $code.=<<___;
+    vspltisw    $vzero,0
+
+___
+
+        load_vrs($in1p, \@in1);
+        load_vrs($in2p, \@in2);
+
+        $code.=<<___;
+    vmsumudm    $out,$in1[0],$in2[0],$vzero
+    stxv        $out,0($outp)
+
+    xxpermdi    $t1,$in1[0],$in1[1],0b00
+    xxpermdi    $t2,$in2[1],$in2[0],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    stxv        $out,16($outp)
+
+    xxpermdi    $t2,$in2[2],$in2[1],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    vmsumudm    $out,$in1[2],$in2[0],$out
+    stxv        $out,32($outp)
+
+    xxpermdi    $t2,$in2[1],$in2[0],0b00
+    xxpermdi    $t3,$in1[2],$in1[3],0b00
+    xxpermdi    $t4,$in2[3],$in2[2],0b00
+    vmsumudm    $out,$t1,$t4,$vzero
+    vmsumudm    $out,$t3,$t2,$out
+    stxv        $out,48($outp)
+
+    xxpermdi    $t2,$in2[4],$in2[3],0b00
+    xxpermdi    $t4,$in2[2],$in2[1],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    vmsumudm    $out,$t3,$t4,$out
+    vmsumudm    $out,$in1[4],$in2[0],$out
+    stxv        $out,64($outp)
+
+    xxpermdi    $t2,$in2[5],$in2[4],0b00
+    xxpermdi    $t4,$in2[3],$in2[2],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    vmsumudm    $out,$t3,$t4,$out
+    xxpermdi    $t4,$in2[1],$in2[0],0b00
+    xxpermdi    $t1,$in1[4],$in1[5],0b00
+    vmsumudm    $out,$t1,$t4,$out
+    stxv        $out,80($outp)
+
+    xxpermdi    $t1,$in1[0],$in1[1],0b00
+    xxpermdi    $t2,$in2[6],$in2[5],0b00
+    xxpermdi    $t4,$in2[4],$in2[3],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    vmsumudm    $out,$t3,$t4,$out
+    xxpermdi    $t2,$in2[2],$in2[1],0b00
+    xxpermdi    $t1,$in1[4],$in1[5],0b00
+    vmsumudm    $out,$t1,$t2,$out
+    vmsumudm    $out,$in1[6],$in2[0],$out
+    stxv        $out,96($outp)
+
+    xxpermdi    $t1,$in1[1],$in1[2],0b00
+    xxpermdi    $t2,$in2[6],$in2[5],0b00
+    xxpermdi    $t3,$in1[3],$in1[4],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    vmsumudm    $out,$t3,$t4,$out
+    xxpermdi    $t3,$in2[2],$in2[1],0b00
+    xxpermdi    $t1,$in1[5],$in1[6],0b00
+    vmsumudm    $out,$t1,$t3,$out
+    stxv        $out,112($outp)
+
+    xxpermdi    $t1,$in1[2],$in1[3],0b00
+    xxpermdi    $t3,$in1[4],$in1[5],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    vmsumudm    $out,$t3,$t4,$out
+    vmsumudm    $out,$in1[6],$in2[2],$out
+    stxv        $out,128($outp)
+
+    xxpermdi    $t1,$in1[3],$in1[4],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    xxpermdi    $t1,$in1[5],$in1[6],0b00
+    vmsumudm    $out,$t1,$t4,$out
+    stxv        $out,144($outp)
+
+    vmsumudm    $out,$t3,$t2,$vzero
+    vmsumudm    $out,$in1[6],$in2[4],$out
+    stxv        $out,160($outp)
+
+    vmsumudm    $out,$t1,$t2,$vzero
+    stxv        $out,176($outp)
+
+    vmsumudm    $out,$in1[6],$in2[6],$vzero
+    stxv        $out,192($outp)
+___
+
+        endproc("p384_felem_mul");
+    }
+
+    {
+        #
+        # p384_felem_square
+        #
+
+        my ($inp) = ("r4");
+        my @in = map("v$_",(44..50));
+        my @inx2 = map("v$_",(35..41));
+
+        startproc("p384_felem_square");
+
+        push_vrs(52, 63);
+
+        $code.=<<___;
+    vspltisw    $vzero,0
+
+___
+
+        load_vrs($inp, \@in);
+
+        $code.=<<___;
+    li        $zero,0
+    li        $one,1
+    mtvsrdd        $t1,$one,$zero
+___
+
+        for (my $i = 0; $i <= 6; $i++) {
+            $code.=<<___;
+    vsld        $inx2[$i],$in[$i],$t1
+___
+        }
+
+        $code.=<<___;
+    vmsumudm    $out,$in[0],$in[0],$vzero
+    stxv        $out,0($outp)
+
+    vmsumudm    $out,$in[0],$inx2[1],$vzero
+    stxv        $out,16($outp)
+
+    vmsumudm    $out,$in[0],$inx2[2],$vzero
+    vmsumudm    $out,$in[1],$in[1],$out
+    stxv        $out,32($outp)
+
+    xxpermdi    $t1,$in[0],$in[1],0b00
+    xxpermdi    $t2,$inx2[3],$inx2[2],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    stxv        $out,48($outp)
+
+    xxpermdi    $t4,$inx2[4],$inx2[3],0b00
+    vmsumudm    $out,$t1,$t4,$vzero
+    vmsumudm    $out,$in[2],$in[2],$out
+    stxv        $out,64($outp)
+
+    xxpermdi    $t2,$inx2[5],$inx2[4],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    vmsumudm    $out,$in[2],$inx2[3],$out
+    stxv        $out,80($outp)
+
+    xxpermdi    $t2,$inx2[6],$inx2[5],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    vmsumudm    $out,$in[2],$inx2[4],$out
+    vmsumudm    $out,$in[3],$in[3],$out
+    stxv        $out,96($outp)
+
+    xxpermdi    $t3,$in[1],$in[2],0b00
+    vmsumudm    $out,$t3,$t2,$vzero
+    vmsumudm    $out,$in[3],$inx2[4],$out
+    stxv        $out,112($outp)
+
+    xxpermdi    $t1,$in[2],$in[3],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    vmsumudm    $out,$in[4],$in[4],$out
+    stxv        $out,128($outp)
+
+    xxpermdi    $t1,$in[3],$in[4],0b00
+    vmsumudm    $out,$t1,$t2,$vzero
+    stxv        $out,144($outp)
+
+    vmsumudm    $out,$in[4],$inx2[6],$vzero
+    vmsumudm    $out,$in[5],$in[5],$out
+    stxv        $out,160($outp)
+
+    vmsumudm    $out,$in[5],$inx2[6],$vzero
+    stxv        $out,176($outp)
+
+    vmsumudm    $out,$in[6],$in[6],$vzero
+    stxv        $out,192($outp)
+___
+
+        endproc("p384_felem_square");
+    }
+}
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/crypto/ec/build.info b/crypto/ec/build.info
index 1fa60a1ded..4077bead7b 100644
--- a/crypto/ec/build.info
+++ b/crypto/ec/build.info
@@ -39,8 +39,9 @@ IF[{- !$disabled{asm} -}]
   $ECASM_ppc64=ecp_nistz256.c ecp_ppc.c ecp_nistz256-ppc64.s
   $ECDEF_ppc64=ECP_NISTZ256_ASM
   IF[{- !$disabled{'ec_nistp_64_gcc_128'} -}]
-    $ECASM_ppc64=$ECASM_ppc64 ecp_nistp521-ppc64.s
-    $ECDEF_ppc64=$ECDEF_ppc64 ECP_NISTP521_ASM
+    $ECASM_ppc64=$ECASM_ppc64 ecp_nistp384-ppc64.s ecp_nistp521-ppc64.s
+    $ECDEF_ppc64=$ECDEF_ppc64 ECP_NISTP384_ASM ECP_NISTP521_ASM
+    INCLUDE[ecp_nistp384.o]=..
     INCLUDE[ecp_nistp521.o]=..
   ENDIF
   IF[{- !$disabled{'ecx'} -}]
@@ -119,6 +120,7 @@ GENERATE[ecp_nistz256-armv8.S]=asm/ecp_nistz256-armv8.pl
 INCLUDE[ecp_nistz256-armv8.o]=..
 GENERATE[ecp_nistz256-ppc64.s]=asm/ecp_nistz256-ppc64.pl
 
+GENERATE[ecp_nistp384-ppc64.s]=asm/ecp_nistp384-ppc64.pl
 GENERATE[ecp_nistp521-ppc64.s]=asm/ecp_nistp521-ppc64.pl
 
 IF[{- !$disabled{'ecx'} -}]
diff --git a/crypto/ec/ecp_nistp384.c b/crypto/ec/ecp_nistp384.c
index a0559487ed..14f9530d07 100644
--- a/crypto/ec/ecp_nistp384.c
+++ b/crypto/ec/ecp_nistp384.c
@@ -691,6 +691,15 @@ void p384_felem_mul(widefelem out, const felem in1, const felem in2);
 
 static void felem_select(void)
 {
+# if defined(_ARCH_PPC64)
+    if ((OPENSSL_ppccap_P & PPC_MADD300) && (OPENSSL_ppccap_P & PPC_ALTIVEC)) {
+        felem_square_p = p384_felem_square;
+        felem_mul_p = p384_felem_mul;
+
+        return;
+    }
+# endif
+
     /* Default */
     felem_square_p = felem_square_ref;
     felem_mul_p = felem_mul_ref;