ARM64 assembly pack: make it Windows-friendly.

[openssl.git] / crypto / chacha / asm / chacha-armv8.pl

diff --git a/crypto/chacha/asm/chacha-armv8.pl b/crypto/chacha/asm/chacha-armv8.pl
index 03e8616a035f81c772851cbcb0dc70003e1d25cf..56ba1c36ba6762efb197f51ea9da4db5a6c9452f 100755 (executable)
--- a/crypto/chacha/asm/chacha-armv8.pl
+++ b/crypto/chacha/asm/chacha-armv8.pl
@@ -1,7 +1,7 @@
 #! /usr/bin/env perl
 # Copyright 2016 The OpenSSL Project Authors. All Rights Reserved.
 #
-# Licensed under the OpenSSL license (the "License").  You may not use
+# 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
@@ -15,7 +15,7 @@
 # ====================================================================
 #
 # June 2015
-# 
+#
 # ChaCha20 for ARMv8.
 #
 # Performance in cycles per byte out of large buffer.
@@ -27,6 +27,8 @@
 # Cortex-A57           8.06/+43%       4.90            4.43(**)
 # Denver               4.50/+82%       2.63            2.67(*)
 # X-Gene               9.50/+46%       8.82            8.89(*)
+# Mongoose             8.00/+44%       3.64            3.25
+# Kryo                 8.17/+50%       4.83            4.65
 #
 # (*)  it's expected that doubling interleave factor doesn't help
 #      all processors, only those with higher NEON latency and
@@ -129,12 +131,6 @@ $code.=<<___;
 .quad  0x3320646e61707865,0x6b20657479622d32           // endian-neutral
 .Lone:
 .long  1,0,0,0
-.LOPENSSL_armcap_P:
-#ifdef __ILP32__
-.long  OPENSSL_armcap_P-.
-#else
-.quad  OPENSSL_armcap_P-.
-#endif
 .asciz "ChaCha20 for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
 
 .globl ChaCha20_ctr32
@@ -142,19 +138,16 @@ $code.=<<___;
 .align 5
 ChaCha20_ctr32:
        cbz     $len,.Labort
-       adr     @x[0],.LOPENSSL_armcap_P
        cmp     $len,#192
        b.lo    .Lshort
-#ifdef __ILP32__
-       ldrsw   @x[1],[@x[0]]
-#else
-       ldr     @x[1],[@x[0]]
-#endif
-       ldr     w17,[@x[1],@x[0]]
+
+       adrp    x17,OPENSSL_armcap_P
+       ldr     w17,[x17,#:lo12:OPENSSL_armcap_P]
        tst     w17,#ARMV7_NEON
-       b.ne    ChaCha20_neon
+       b.ne    .LChaCha20_neon
 
 .Lshort:
+       .inst   0xd503233f                      // paciasp
        stp     x29,x30,[sp,#-96]!
        add     x29,sp,#0
 
@@ -200,7 +193,7 @@ ChaCha20_ctr32:
        mov     $ctr,#10
        subs    $len,$len,#64
 .Loop:
-       sub     $ctr,$ctr,#1    
+       sub     $ctr,$ctr,#1
 ___
        foreach (&ROUND(0, 4, 8,12)) { eval; }
        foreach (&ROUND(0, 5,10,15)) { eval; }
@@ -274,6 +267,7 @@ $code.=<<___;
        ldp     x25,x26,[x29,#64]
        ldp     x27,x28,[x29,#80]
        ldp     x29,x30,[sp],#96
+       .inst   0xd50323bf                      // autiasp
 .Labort:
        ret
 
@@ -330,6 +324,7 @@ $code.=<<___;
        ldp     x25,x26,[x29,#64]
        ldp     x27,x28,[x29,#80]
        ldp     x29,x30,[sp],#96
+       .inst   0xd50323bf                      // autiasp
        ret
 .size  ChaCha20_ctr32,.-ChaCha20_ctr32
 ___
@@ -375,6 +370,8 @@ $code.=<<___;
 .type  ChaCha20_neon,%function
 .align 5
 ChaCha20_neon:
+.LChaCha20_neon:
+       .inst   0xd503233f                      // paciasp
        stp     x29,x30,[sp,#-96]!
        add     x29,sp,#0
 
@@ -573,6 +570,7 @@ $code.=<<___;
        ldp     x25,x26,[x29,#64]
        ldp     x27,x28,[x29,#80]
        ldp     x29,x30,[sp],#96
+       .inst   0xd50323bf                      // autiasp
        ret
 
 .Ltail_neon:
@@ -682,6 +680,7 @@ $code.=<<___;
        ldp     x25,x26,[x29,#64]
        ldp     x27,x28,[x29,#80]
        ldp     x29,x30,[sp],#96
+       .inst   0xd50323bf                      // autiasp
        ret
 .size  ChaCha20_neon,.-ChaCha20_neon
 ___
@@ -694,6 +693,7 @@ $code.=<<___;
 .type  ChaCha20_512_neon,%function
 .align 5
 ChaCha20_512_neon:
+       .inst   0xd503233f                      // paciasp
        stp     x29,x30,[sp,#-96]!
        add     x29,sp,#0
 
@@ -1112,6 +1112,7 @@ $code.=<<___;
        ldp     x25,x26,[x29,#64]
        ldp     x27,x28,[x29,#80]
        ldp     x29,x30,[sp],#96
+       .inst   0xd50323bf                      // autiasp
        ret
 .size  ChaCha20_512_neon,.-ChaCha20_512_neon
 ___