diff options
| author | Andy Polyakov <appro@openssl.org> | 2019-02-15 22:16:41 +0100 |
|---|---|---|
| committer | Richard Levitte <levitte@openssl.org> | 2019-02-16 17:01:15 +0100 |
| commit | db42bb440e76399b89fc8ae04644441a2a5f6821 (patch) | |
| tree | 0ddc0a16632b60834e44805d2e329b774381f323 /crypto/poly1305/asm | |
| parent | 3405db97e5448c784729b56837f3f8c776a01067 (diff) | |
| download | openssl-new-db42bb440e76399b89fc8ae04644441a2a5f6821.tar.gz | |
ARM64 assembly pack: make it Windows-friendly.
"Windows friendliness" means a) unified PIC-ification, unified across
all platforms; b) unified commantary delimiter; c) explicit ldur/stur,
as Visual Studio assembler can't automatically encode ldr/str as
ldur/stur when needed.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/8256)
Diffstat (limited to 'crypto/poly1305/asm')
| -rwxr-xr-x | crypto/poly1305/asm/poly1305-armv8.pl | 31 |
1 files changed, 12 insertions, 19 deletions
diff --git a/crypto/poly1305/asm/poly1305-armv8.pl b/crypto/poly1305/asm/poly1305-armv8.pl index 1aded5a275..b7aa7dc90b 100755 --- a/crypto/poly1305/asm/poly1305-armv8.pl +++ b/crypto/poly1305/asm/poly1305-armv8.pl @@ -71,17 +71,12 @@ poly1305_init: csel x0,xzr,x0,eq b.eq .Lno_key -#ifdef __ILP32__ - ldrsw $t1,.LOPENSSL_armcap_P -#else - ldr $t1,.LOPENSSL_armcap_P -#endif - adr $t0,.LOPENSSL_armcap_P + adrp x17,OPENSSL_armcap_P + ldr w17,[x17,#:lo12:OPENSSL_armcap_P] ldp $r0,$r1,[$inp] // load key mov $s1,#0xfffffffc0fffffff movk $s1,#0x0fff,lsl#48 - ldr w17,[$t0,$t1] #ifdef __ARMEB__ rev $r0,$r0 // flip bytes rev $r1,$r1 @@ -93,10 +88,10 @@ poly1305_init: tst w17,#ARMV7_NEON - adr $d0,poly1305_blocks - adr $r0,poly1305_blocks_neon - adr $d1,poly1305_emit - adr $r1,poly1305_emit_neon + adr $d0,.Lpoly1305_blocks + adr $r0,.Lpoly1305_blocks_neon + adr $d1,.Lpoly1305_emit + adr $r1,.Lpoly1305_emit_neon csel $d0,$d0,$r0,eq csel $d1,$d1,$r1,eq @@ -115,6 +110,7 @@ poly1305_init: .type poly1305_blocks,%function .align 5 poly1305_blocks: +.Lpoly1305_blocks: ands $len,$len,#-16 b.eq .Lno_data @@ -179,6 +175,7 @@ poly1305_blocks: .type poly1305_emit,%function .align 5 poly1305_emit: +.Lpoly1305_emit: ldp $h0,$h1,[$ctx] // load hash base 2^64 ldr $h2,[$ctx,#16] ldp $t0,$t1,[$nonce] // load nonce @@ -285,10 +282,11 @@ poly1305_splat: .type poly1305_blocks_neon,%function .align 5 poly1305_blocks_neon: +.Lpoly1305_blocks_neon: ldr $is_base2_26,[$ctx,#24] cmp $len,#128 b.hs .Lblocks_neon - cbz $is_base2_26,poly1305_blocks + cbz $is_base2_26,.Lpoly1305_blocks .Lblocks_neon: .inst 0xd503233f // paciasp @@ -431,7 +429,7 @@ poly1305_blocks_neon: csel $in2,$zeros,$in2,lo mov x4,#1 - str x4,[$ctx,#-24] // set is_base2_26 + stur x4,[$ctx,#-24] // set is_base2_26 sub $ctx,$ctx,#48 // restore original $ctx b .Ldo_neon @@ -868,6 +866,7 @@ poly1305_blocks_neon: .type poly1305_emit_neon,%function .align 5 poly1305_emit_neon: +.Lpoly1305_emit_neon: ldr $is_base2_26,[$ctx,#24] cbz $is_base2_26,poly1305_emit @@ -920,12 +919,6 @@ poly1305_emit_neon: .align 5 .Lzeros: .long 0,0,0,0,0,0,0,0 -.LOPENSSL_armcap_P: -#ifdef __ILP32__ -.long OPENSSL_armcap_P-. -#else -.quad OPENSSL_armcap_P-. -#endif .asciz "Poly1305 for ARMv8, CRYPTOGAMS by <appro\@openssl.org>" .align 2 ___ |