rpm/openssl-1_1
SHA256
1
0
Fork 0
forked from pool/openssl-1_1
Code Pull Requests Activity

Accepting request 1119558 from home:ohollmann:branches:security:tls

Browse Source 
- Performance enhancements for cryptography from OpenSSL 3.x
  [jsc#PED-5086, jsc#PED-3514]
  * Add patches:
    - openssl-ec-Use-static-linkage-on-nistp521-felem_-square-mul-.patch
    - openssl-ec-56-bit-Limb-Solinas-Strategy-for-secp384r1.patch
    - openssl-ec-powerpc64le-Add-asm-implementation-of-felem_-squa.patch
    - openssl-ecc-Remove-extraneous-parentheses-in-secp384r1.patch
    - openssl-powerpc-ecc-Fix-stack-allocation-secp384r1-asm.patch
    - openssl-Improve-performance-for-6x-unrolling-with-vpermxor-i.patch

OBS-URL: https://build.opensuse.org/request/show/1119558
OBS-URL: https://build.opensuse.org/package/show/security:tls/openssl-1_1?expand=0&rev=148
This commit is contained in:
Otto Hollmann 2023-10-25 07:52:22 +00:00 committed by Git OBS Bridge
parent 861b7f632f
commit b51c004cd8
8 changed files with 3360 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
openssl-1_1.changes
View File

@ -1,3 +1,16 @@
-------------------------------------------------------------------
Thu Oct 19 15:03:14 UTC 2023 - Otto Hollmann <otto.hollmann@suse.com>
- Performance enhancements for cryptography from OpenSSL 3.x
[jsc#PED-5086, jsc#PED-3514]
* Add patches:
- openssl-ec-Use-static-linkage-on-nistp521-felem_-square-mul-.patch
- openssl-ec-56-bit-Limb-Solinas-Strategy-for-secp384r1.patch
- openssl-ec-powerpc64le-Add-asm-implementation-of-felem_-squa.patch
- openssl-ecc-Remove-extraneous-parentheses-in-secp384r1.patch
- openssl-powerpc-ecc-Fix-stack-allocation-secp384r1-asm.patch
- openssl-Improve-performance-for-6x-unrolling-with-vpermxor-i.patch
-------------------------------------------------------------------
Wed Oct 4 07:15:29 UTC 2023 - Otto Hollmann <otto.hollmann@suse.com>

8
openssl-1_1.spec
View File

@ -177,6 +177,14 @@ Patch106: openssl-s_client-check-ocsp-status.patch
Patch107: openssl-dont-pass-zero-length-input-to-EVP_Cipher.patch
#PATCH-FIX-SUSE bsc#1215215 FIPS: Add "fips" to version string
Patch108: openssl-1_1-fips-bsc1215215_fips_in_version_string.patch
# PATCH-FIX-UPSTREAM jsc#PED-5086, jsc#PED-3514
# POWER10 performance enhancements for cryptography
Patch109: openssl-ec-Use-static-linkage-on-nistp521-felem_-square-mul-.patch
Patch110: openssl-ec-56-bit-Limb-Solinas-Strategy-for-secp384r1.patch
Patch111: openssl-ec-powerpc64le-Add-asm-implementation-of-felem_-squa.patch
Patch112: openssl-ecc-Remove-extraneous-parentheses-in-secp384r1.patch
Patch113: openssl-powerpc-ecc-Fix-stack-allocation-secp384r1-asm.patch
Patch114: openssl-Improve-performance-for-6x-unrolling-with-vpermxor-i.patch
BuildRequires: jitterentropy-devel >= 3.4.0
BuildRequires: pkgconfig
BuildRequires: pkgconfig(zlib)

495
openssl-Improve-performance-for-6x-unrolling-with-vpermxor-i.patch Normal file
View File

@ -0,0 +1,495 @@
From 3d3a7ecd1ae5ab08d22041f7b3b035c34f12fa02 Mon Sep 17 00:00:00 2001
From: Danny Tsen <dtsen@linux.ibm.com>
Date: Tue, 22 Aug 2023 15:58:53 -0400
Subject: [PATCH] Improve performance for 6x unrolling with vpermxor
instruction
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/21812)
---
crypto/aes/asm/aesp8-ppc.pl | 145 +++++++++++++++++++++++-------------
1 file changed, 95 insertions(+), 50 deletions(-)
diff --git a/crypto/aes/asm/aesp8-ppc.pl b/crypto/aes/asm/aesp8-ppc.pl
index 60cf86f52aed2..38b9405a283b7 100755
--- a/crypto/aes/asm/aesp8-ppc.pl
+++ b/crypto/aes/asm/aesp8-ppc.pl
@@ -99,11 +99,12 @@
.long 0x1b000000, 0x1b000000, 0x1b000000, 0x1b000000 ?rev
.long 0x0d0e0f0c, 0x0d0e0f0c, 0x0d0e0f0c, 0x0d0e0f0c ?rev
.long 0,0,0,0 ?asis
+.long 0x0f102132, 0x43546576, 0x8798a9ba, 0xcbdcedfe
Lconsts:
mflr r0
bcl 20,31,\$+4
mflr $ptr #vvvvv "distance between . and rcon
- addi $ptr,$ptr,-0x48
+ addi $ptr,$ptr,-0x58
mtlr r0
blr
.long 0
@@ -2405,7 +2406,7 @@ ()
my $key_=$key2;
my ($x00,$x10,$x20,$x30,$x40,$x50,$x60,$x70)=map("r$_",(0,3,26..31));
$x00=0 if ($flavour =~ /osx/);
-my ($in0, $in1, $in2, $in3, $in4, $in5 )=map("v$_",(0..5));
+my ($in0, $in1, $in2, $in3, $in4, $in5)=map("v$_",(0..5));
my ($out0, $out1, $out2, $out3, $out4, $out5)=map("v$_",(7,12..16));
my ($twk0, $twk1, $twk2, $twk3, $twk4, $twk5)=map("v$_",(17..22));
my $rndkey0="v23"; # v24-v25 rotating buffer for first found keys
@@ -2460,6 +2461,18 @@ ()
li $x70,0x70
mtspr 256,r0
+ # Reverse eighty7 to 0x010101..87
+ xxlor 2, 32+$eighty7, 32+$eighty7
+ vsldoi $eighty7,$tmp,$eighty7,1 # 0x010101..87
+ xxlor 1, 32+$eighty7, 32+$eighty7
+
+ # Load XOR contents. 0xf102132435465768798a9bacbdcedfe
+ mr $x70, r6
+ bl Lconsts
+ lxvw4x 0, $x40, r6 # load XOR contents
+ mr r6, $x70
+ li $x70,0x70
+
subi $rounds,$rounds,3 # -4 in total
lvx $rndkey0,$x00,$key1 # load key schedule
@@ -2502,69 +2515,77 @@ ()
?vperm v31,v31,$twk5,$keyperm
lvx v25,$x10,$key_ # pre-load round[2]
+ # Switch to use the following codes with 0x010101..87 to generate tweak.
+ # eighty7 = 0x010101..87
+ # vsrab tmp, tweak, seven # next tweak value, right shift 7 bits
+ # vand tmp, tmp, eighty7 # last byte with carry
+ # vaddubm tweak, tweak, tweak # left shift 1 bit (x2)
+ # xxlor vsx, 0, 0
+ # vpermxor tweak, tweak, tmp, vsx
+
vperm $in0,$inout,$inptail,$inpperm
subi $inp,$inp,31 # undo "caller"
vxor $twk0,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
vand $tmp,$tmp,$eighty7
vxor $out0,$in0,$twk0
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in1, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in1
lvx_u $in1,$x10,$inp
vxor $twk1,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
le?vperm $in1,$in1,$in1,$leperm
vand $tmp,$tmp,$eighty7
vxor $out1,$in1,$twk1
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in2, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in2
lvx_u $in2,$x20,$inp
andi. $taillen,$len,15
vxor $twk2,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
le?vperm $in2,$in2,$in2,$leperm
vand $tmp,$tmp,$eighty7
vxor $out2,$in2,$twk2
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in3, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in3
lvx_u $in3,$x30,$inp
sub $len,$len,$taillen
vxor $twk3,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
le?vperm $in3,$in3,$in3,$leperm
vand $tmp,$tmp,$eighty7
vxor $out3,$in3,$twk3
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in4, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in4
lvx_u $in4,$x40,$inp
subi $len,$len,0x60
vxor $twk4,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
le?vperm $in4,$in4,$in4,$leperm
vand $tmp,$tmp,$eighty7
vxor $out4,$in4,$twk4
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in5, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in5
lvx_u $in5,$x50,$inp
addi $inp,$inp,0x60
vxor $twk5,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
le?vperm $in5,$in5,$in5,$leperm
vand $tmp,$tmp,$eighty7
vxor $out5,$in5,$twk5
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in0, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in0
vxor v31,v31,$rndkey0
mtctr $rounds
@@ -2590,6 +2611,8 @@ ()
lvx v25,$x10,$key_ # round[4]
bdnz Loop_xts_enc6x
+ xxlor 32+$eighty7, 1, 1 # 0x010101..87
+
subic $len,$len,96 # $len-=96
vxor $in0,$twk0,v31 # xor with last round key
vcipher $out0,$out0,v24
@@ -2599,7 +2622,6 @@ ()
vaddubm $tweak,$tweak,$tweak
vcipher $out2,$out2,v24
vcipher $out3,$out3,v24
- vsldoi $tmp,$tmp,$tmp,15
vcipher $out4,$out4,v24
vcipher $out5,$out5,v24
@@ -2607,7 +2629,8 @@ ()
vand $tmp,$tmp,$eighty7
vcipher $out0,$out0,v25
vcipher $out1,$out1,v25
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in1, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in1
vcipher $out2,$out2,v25
vcipher $out3,$out3,v25
vxor $in1,$twk1,v31
@@ -2618,13 +2641,13 @@ ()
and r0,r0,$len
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
vcipher $out0,$out0,v26
vcipher $out1,$out1,v26
vand $tmp,$tmp,$eighty7
vcipher $out2,$out2,v26
vcipher $out3,$out3,v26
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in2, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in2
vcipher $out4,$out4,v26
vcipher $out5,$out5,v26
@@ -2638,7 +2661,6 @@ ()
vaddubm $tweak,$tweak,$tweak
vcipher $out0,$out0,v27
vcipher $out1,$out1,v27
- vsldoi $tmp,$tmp,$tmp,15
vcipher $out2,$out2,v27
vcipher $out3,$out3,v27
vand $tmp,$tmp,$eighty7
@@ -2646,7 +2668,8 @@ ()
vcipher $out5,$out5,v27
addi $key_,$sp,$FRAME+15 # rewind $key_
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in3, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in3
vcipher $out0,$out0,v28
vcipher $out1,$out1,v28
vxor $in3,$twk3,v31
@@ -2655,7 +2678,6 @@ ()
vcipher $out2,$out2,v28
vcipher $out3,$out3,v28
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
vcipher $out4,$out4,v28
vcipher $out5,$out5,v28
lvx v24,$x00,$key_ # re-pre-load round[1]
@@ -2663,7 +2685,8 @@ ()
vcipher $out0,$out0,v29
vcipher $out1,$out1,v29
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in4, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in4
vcipher $out2,$out2,v29
vcipher $out3,$out3,v29
vxor $in4,$twk4,v31
@@ -2673,14 +2696,14 @@ ()
vcipher $out5,$out5,v29
lvx v25,$x10,$key_ # re-pre-load round[2]
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
vcipher $out0,$out0,v30
vcipher $out1,$out1,v30
vand $tmp,$tmp,$eighty7
vcipher $out2,$out2,v30
vcipher $out3,$out3,v30
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in5, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in5
vcipher $out4,$out4,v30
vcipher $out5,$out5,v30
vxor $in5,$twk5,v31
@@ -2690,7 +2713,6 @@ ()
vcipherlast $out0,$out0,$in0
lvx_u $in0,$x00,$inp # load next input block
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
vcipherlast $out1,$out1,$in1
lvx_u $in1,$x10,$inp
vcipherlast $out2,$out2,$in2
@@ -2703,7 +2725,10 @@ ()
vcipherlast $out4,$out4,$in4
le?vperm $in2,$in2,$in2,$leperm
lvx_u $in4,$x40,$inp
- vxor $tweak,$tweak,$tmp
+ xxlor 10, 32+$in0, 32+$in0
+ xxlor 32+$in0, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in0
+ xxlor 32+$in0, 10, 10
vcipherlast $tmp,$out5,$in5 # last block might be needed
# in stealing mode
le?vperm $in3,$in3,$in3,$leperm
@@ -2736,6 +2761,8 @@ ()
mtctr $rounds
beq Loop_xts_enc6x # did $len-=96 borrow?
+ xxlor 32+$eighty7, 2, 2 # 0x870101..01
+
addic. $len,$len,0x60
beq Lxts_enc6x_zero
cmpwi $len,0x20
@@ -3112,6 +3139,18 @@ ()
li $x70,0x70
mtspr 256,r0
+ # Reverse eighty7 to 0x010101..87
+ xxlor 2, 32+$eighty7, 32+$eighty7
+ vsldoi $eighty7,$tmp,$eighty7,1 # 0x010101..87
+ xxlor 1, 32+$eighty7, 32+$eighty7
+
+ # Load XOR contents. 0xf102132435465768798a9bacbdcedfe
+ mr $x70, r6
+ bl Lconsts
+ lxvw4x 0, $x40, r6 # load XOR contents
+ mr r6, $x70
+ li $x70,0x70
+
subi $rounds,$rounds,3 # -4 in total
lvx $rndkey0,$x00,$key1 # load key schedule
@@ -3159,64 +3198,64 @@ ()
vxor $twk0,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
vand $tmp,$tmp,$eighty7
vxor $out0,$in0,$twk0
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in1, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in1
lvx_u $in1,$x10,$inp
vxor $twk1,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
le?vperm $in1,$in1,$in1,$leperm
vand $tmp,$tmp,$eighty7
vxor $out1,$in1,$twk1
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in2, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in2
lvx_u $in2,$x20,$inp
andi. $taillen,$len,15
vxor $twk2,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
le?vperm $in2,$in2,$in2,$leperm
vand $tmp,$tmp,$eighty7
vxor $out2,$in2,$twk2
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in3, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in3
lvx_u $in3,$x30,$inp
sub $len,$len,$taillen
vxor $twk3,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
le?vperm $in3,$in3,$in3,$leperm
vand $tmp,$tmp,$eighty7
vxor $out3,$in3,$twk3
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in4, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in4
lvx_u $in4,$x40,$inp
subi $len,$len,0x60
vxor $twk4,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
le?vperm $in4,$in4,$in4,$leperm
vand $tmp,$tmp,$eighty7
vxor $out4,$in4,$twk4
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in5, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in5
lvx_u $in5,$x50,$inp
addi $inp,$inp,0x60
vxor $twk5,$tweak,$rndkey0
vsrab $tmp,$tweak,$seven # next tweak value
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
le?vperm $in5,$in5,$in5,$leperm
vand $tmp,$tmp,$eighty7
vxor $out5,$in5,$twk5
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in0, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in0
vxor v31,v31,$rndkey0
mtctr $rounds
@@ -3242,6 +3281,8 @@ ()
lvx v25,$x10,$key_ # round[4]
bdnz Loop_xts_dec6x
+ xxlor 32+$eighty7, 1, 1
+
subic $len,$len,96 # $len-=96
vxor $in0,$twk0,v31 # xor with last round key
vncipher $out0,$out0,v24
@@ -3251,7 +3292,6 @@ ()
vaddubm $tweak,$tweak,$tweak
vncipher $out2,$out2,v24
vncipher $out3,$out3,v24
- vsldoi $tmp,$tmp,$tmp,15
vncipher $out4,$out4,v24
vncipher $out5,$out5,v24
@@ -3259,7 +3299,8 @@ ()
vand $tmp,$tmp,$eighty7
vncipher $out0,$out0,v25
vncipher $out1,$out1,v25
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in1, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in1
vncipher $out2,$out2,v25
vncipher $out3,$out3,v25
vxor $in1,$twk1,v31
@@ -3270,13 +3311,13 @@ ()
and r0,r0,$len
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
vncipher $out0,$out0,v26
vncipher $out1,$out1,v26
vand $tmp,$tmp,$eighty7
vncipher $out2,$out2,v26
vncipher $out3,$out3,v26
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in2, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in2
vncipher $out4,$out4,v26
vncipher $out5,$out5,v26
@@ -3290,7 +3331,6 @@ ()
vaddubm $tweak,$tweak,$tweak
vncipher $out0,$out0,v27
vncipher $out1,$out1,v27
- vsldoi $tmp,$tmp,$tmp,15
vncipher $out2,$out2,v27
vncipher $out3,$out3,v27
vand $tmp,$tmp,$eighty7
@@ -3298,7 +3338,8 @@ ()
vncipher $out5,$out5,v27
addi $key_,$sp,$FRAME+15 # rewind $key_
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in3, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in3
vncipher $out0,$out0,v28
vncipher $out1,$out1,v28
vxor $in3,$twk3,v31
@@ -3307,7 +3348,6 @@ ()
vncipher $out2,$out2,v28
vncipher $out3,$out3,v28
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
vncipher $out4,$out4,v28
vncipher $out5,$out5,v28
lvx v24,$x00,$key_ # re-pre-load round[1]
@@ -3315,7 +3355,8 @@ ()
vncipher $out0,$out0,v29
vncipher $out1,$out1,v29
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in4, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in4
vncipher $out2,$out2,v29
vncipher $out3,$out3,v29
vxor $in4,$twk4,v31
@@ -3325,14 +3366,14 @@ ()
vncipher $out5,$out5,v29
lvx v25,$x10,$key_ # re-pre-load round[2]
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
vncipher $out0,$out0,v30
vncipher $out1,$out1,v30
vand $tmp,$tmp,$eighty7
vncipher $out2,$out2,v30
vncipher $out3,$out3,v30
- vxor $tweak,$tweak,$tmp
+ xxlor 32+$in5, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in5
vncipher $out4,$out4,v30
vncipher $out5,$out5,v30
vxor $in5,$twk5,v31
@@ -3342,7 +3383,6 @@ ()
vncipherlast $out0,$out0,$in0
lvx_u $in0,$x00,$inp # load next input block
vaddubm $tweak,$tweak,$tweak
- vsldoi $tmp,$tmp,$tmp,15
vncipherlast $out1,$out1,$in1
lvx_u $in1,$x10,$inp
vncipherlast $out2,$out2,$in2
@@ -3355,7 +3395,10 @@ ()
vncipherlast $out4,$out4,$in4
le?vperm $in2,$in2,$in2,$leperm
lvx_u $in4,$x40,$inp
- vxor $tweak,$tweak,$tmp
+ xxlor 10, 32+$in0, 32+$in0
+ xxlor 32+$in0, 0, 0
+ vpermxor $tweak, $tweak, $tmp, $in0
+ xxlor 32+$in0, 10, 10
vncipherlast $out5,$out5,$in5
le?vperm $in3,$in3,$in3,$leperm
lvx_u $in5,$x50,$inp
@@ -3386,6 +3429,8 @@ ()
mtctr $rounds
beq Loop_xts_dec6x # did $len-=96 borrow?
+ xxlor 32+$eighty7, 2, 2
+
addic. $len,$len,0x60
beq Lxts_dec6x_zero
cmpwi $len,0x20

2197
openssl-ec-56-bit-Limb-Solinas-Strategy-for-secp384r1.patch Normal file
View File

File diff suppressed because it is too large Load Diff

65
openssl-ec-Use-static-linkage-on-nistp521-felem_-square-mul-.patch Normal file
View File

@ -0,0 +1,65 @@
From 3e47a286dc3274bda72a196c3a4030a1fc8302f1 Mon Sep 17 00:00:00 2001
From: Rohan McLure <rohanmclure@linux.ibm.com>
Date: Fri, 23 Jun 2023 16:41:48 +1000
Subject: [PATCH] ec: Use static linkage on nistp521 felem_{square,mul}
wrappers
Runtime selection of implementations for felem_{square,mul} depends on
felem_{square,mul}_wrapper functions, which overwrite function points in
a similar design to that of .plt.got sections used by program loaders
during dynamic linking.
There's no reason why these functions need to have external linkage.
Mark static.
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)
---
crypto/ec/ecp_nistp521.c | 10 +++++-----
1 file changed, 5 insertions(+), 5 deletions(-)
diff --git a/crypto/ec/ecp_nistp521.c b/crypto/ec/ecp_nistp521.c
index 97815cac1f13..32a9268ecf17 100644
--- a/crypto/ec/ecp_nistp521.c
+++ b/crypto/ec/ecp_nistp521.c
@@ -676,8 +676,8 @@ static void felem_reduce(felem out, const largefelem in)
}
#if defined(ECP_NISTP521_ASM)
-void felem_square_wrapper(largefelem out, const felem in);
-void felem_mul_wrapper(largefelem out, const felem in1, const felem in2);
+static void felem_square_wrapper(largefelem out, const felem in);
+static void felem_mul_wrapper(largefelem out, const felem in1, const felem in2);
static void (*felem_square_p)(largefelem out, const felem in) =
felem_square_wrapper;
@@ -691,7 +691,7 @@ void p521_felem_mul(largefelem out, const felem in1, const felem in2);
# include "../ppc_arch.h"
# endif
-void felem_select(void)
+static void felem_select(void)
{
# if defined(_ARCH_PPC64)
if ((OPENSSL_ppccap_P & PPC_MADD300) && (OPENSSL_ppccap_P & PPC_ALTIVEC)) {
@@ -707,13 +707,13 @@ void felem_select(void)
felem_mul_p = felem_mul_ref;
}
-void felem_square_wrapper(largefelem out, const felem in)
+static void felem_square_wrapper(largefelem out, const felem in)
{
felem_select();
felem_square_p(out, in);
}
-void felem_mul_wrapper(largefelem out, const felem in1, const felem in2)
+static void felem_mul_wrapper(largefelem out, const felem in1, const felem in2)
{
felem_select();
felem_mul_p(out, in1, in2);

410
openssl-ec-powerpc64le-Add-asm-implementation-of-felem_-squa.patch Normal file
View File

@ -0,0 +1,410 @@
From 966047ee13188e8634af25af348940acceb9316d Mon Sep 17 00:00:00 2001
From: Rohan McLure <rohanmclure@linux.ibm.com>
Date: Wed, 31 May 2023 14:32:26 +1000
Subject: [PATCH] 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)
---
crypto/ec/asm/ecp_nistp384-ppc64.pl | 355 ++++++++++++++++++++++++++++++++++++
crypto/ec/build.info | 2
crypto/ec/ecp_nistp384.c | 9
3 files changed, 366 insertions(+)
create mode 100755 crypto/ec/asm/ecp_nistp384-ppc64.pl
--- /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: $!";
--- a/crypto/ec/build.info
+++ b/crypto/ec/build.info
@@ -31,6 +31,8 @@ GENERATE[ecp_nistz256-armv8.S]=asm/ecp_n
INCLUDE[ecp_nistz256-armv8.o]=..
GENERATE[ecp_nistz256-ppc64.s]=asm/ecp_nistz256-ppc64.pl $(PERLASM_SCHEME)
+GENERATE[ecp_nistp384-ppc64.s]=asm/ecp_nistp384-ppc64.pl $(PERLASM_SCHEME)
+INCLUDE[ecp_nistp384.o]=..
GENERATE[ecp_nistp521-ppc64.s]=asm/ecp_nistp521-ppc64.pl $(PERLASM_SCHEME)
GENERATE[x25519-x86_64.s]=asm/x25519-x86_64.pl $(PERLASM_SCHEME)
--- a/crypto/ec/ecp_nistp384.c
+++ b/crypto/ec/ecp_nistp384.c
@@ -691,6 +691,15 @@ void p384_felem_mul(widefelem out, const
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;

76
openssl-ecc-Remove-extraneous-parentheses-in-secp384r1.patch Normal file
View File

@ -0,0 +1,76 @@
From 670e73d9084465384b11ef24802ca4a313e1d2f4 Mon Sep 17 00:00:00 2001
From: Rohan McLure <rohanmclure@linux.ibm.com>
Date: Tue, 15 Aug 2023 15:20:20 +1000
Subject: [PATCH] ecc: Remove extraneous parentheses in secp384r1
Substitutions in the felem_reduce() method feature unecessary
parentheses, remove them.
Signed-off-by: Rohan McLure <rohan.mclure@linux.ibm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/21749)
---
crypto/ec/ecp_nistp384.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/crypto/ec/ecp_nistp384.c b/crypto/ec/ecp_nistp384.c
index 14f9530d07c6..ff68f9cc7ad0 100644
--- a/crypto/ec/ecp_nistp384.c
+++ b/crypto/ec/ecp_nistp384.c
@@ -540,7 +540,7 @@ static void felem_reduce(felem out, const widefelem in)
acc[7] += in[12] >> 8;
acc[6] += (in[12] & 0xff) << 48;
acc[6] -= in[12] >> 16;
- acc[5] -= ((in[12] & 0xffff) << 40);
+ acc[5] -= (in[12] & 0xffff) << 40;
acc[6] += in[12] >> 48;
acc[5] += (in[12] & 0xffffffffffff) << 8;
@@ -549,7 +549,7 @@ static void felem_reduce(felem out, const widefelem in)
acc[6] += in[11] >> 8;
acc[5] += (in[11] & 0xff) << 48;
acc[5] -= in[11] >> 16;
- acc[4] -= ((in[11] & 0xffff) << 40);
+ acc[4] -= (in[11] & 0xffff) << 40;
acc[5] += in[11] >> 48;
acc[4] += (in[11] & 0xffffffffffff) << 8;
@@ -558,7 +558,7 @@ static void felem_reduce(felem out, const widefelem in)
acc[5] += in[10] >> 8;
acc[4] += (in[10] & 0xff) << 48;
acc[4] -= in[10] >> 16;
- acc[3] -= ((in[10] & 0xffff) << 40);
+ acc[3] -= (in[10] & 0xffff) << 40;
acc[4] += in[10] >> 48;
acc[3] += (in[10] & 0xffffffffffff) << 8;
@@ -567,7 +567,7 @@ static void felem_reduce(felem out, const widefelem in)
acc[4] += in[9] >> 8;
acc[3] += (in[9] & 0xff) << 48;
acc[3] -= in[9] >> 16;
- acc[2] -= ((in[9] & 0xffff) << 40);
+ acc[2] -= (in[9] & 0xffff) << 40;
acc[3] += in[9] >> 48;
acc[2] += (in[9] & 0xffffffffffff) << 8;
@@ -582,7 +582,7 @@ static void felem_reduce(felem out, const widefelem in)
acc[3] += acc[8] >> 8;
acc[2] += (acc[8] & 0xff) << 48;
acc[2] -= acc[8] >> 16;
- acc[1] -= ((acc[8] & 0xffff) << 40);
+ acc[1] -= (acc[8] & 0xffff) << 40;
acc[2] += acc[8] >> 48;
acc[1] += (acc[8] & 0xffffffffffff) << 8;
@@ -591,7 +591,7 @@ static void felem_reduce(felem out, const widefelem in)
acc[2] += acc[7] >> 8;
acc[1] += (acc[7] & 0xff) << 48;
acc[1] -= acc[7] >> 16;
- acc[0] -= ((acc[7] & 0xffff) << 40);
+ acc[0] -= (acc[7] & 0xffff) << 40;
acc[1] += acc[7] >> 48;
acc[0] += (acc[7] & 0xffffffffffff) << 8;

96
openssl-powerpc-ecc-Fix-stack-allocation-secp384r1-asm.patch Normal file
View File

@ -0,0 +1,96 @@
From 50f8b936b00dc18ce1f622a7a6aa46daf03da48b Mon Sep 17 00:00:00 2001
From: Rohan McLure <rohanmclure@linux.ibm.com>
Date: Wed, 16 Aug 2023 16:52:47 +1000
Subject: [PATCH] powerpc: ecc: Fix stack allocation secp384r1 asm
Assembly acceleration secp384r1 opts to not use any callee-save VSRs, as
VSX enabled systems make extensive use of renaming, and so writebacks in
felem_{mul,square}() can be reordered for best cache effects.
Remove stack allocations. This in turn fixes unmatched push/pops in
felem_{mul,square}().
Signed-off-by: Rohan McLure <rohan.mclure@linux.ibm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/21749)
---
crypto/ec/asm/ecp_nistp384-ppc64.pl | 49 -----------------------------
1 file changed, 49 deletions(-)
diff --git a/crypto/ec/asm/ecp_nistp384-ppc64.pl b/crypto/ec/asm/ecp_nistp384-ppc64.pl
index 3f86b391af69..28f4168e5218 100755
--- a/crypto/ec/asm/ecp_nistp384-ppc64.pl
+++ b/crypto/ec/asm/ecp_nistp384-ppc64.pl
@@ -62,51 +62,6 @@ ($)
___
}
-
-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($$)
{
my ($pointer, $reg_list) = @_;
@@ -162,8 +117,6 @@ ($$)
startproc("p384_felem_mul");
- push_vrs(52, 63);
-
$code.=<<___;
vspltisw $vzero,0
@@ -268,8 +221,6 @@ ($$)
startproc("p384_felem_square");
- push_vrs(52, 63);
-
$code.=<<___;
vspltisw $vzero,0