diff options
| author | Andy Polyakov <appro@openssl.org> | 2019-02-15 09:44:39 +0100 |
|---|---|---|
| committer | Richard Levitte <levitte@openssl.org> | 2019-02-16 16:59:23 +0100 |
| commit | 3405db97e5448c784729b56837f3f8c776a01067 (patch) | |
| tree | 3038d1d73fb0964915fdfd6377bb717722c02ea7 /crypto/poly1305 | |
| parent | b2b580fe445e064da50c13d3e00f71022da16ece (diff) | |
ARM assembly pack: make it Windows-friendly.
"Windows friendliness" means a) flipping .thumb and .text directives,
b) always generate Thumb-2 code when asked(*); c) Windows-specific
references to external OPENSSL_armcap_P.
(*) so far *some* modules were compiled as .code 32 even if Thumb-2
was targeted. It works at hardware level because processor can alternate
between the modes with no overhead. But clang --target=arm-windows's
builtin assembler just refuses to compile .code 32...
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/8252)
Diffstat (limited to 'crypto/poly1305')
| -rwxr-xr-x | crypto/poly1305/asm/poly1305-armv4.pl | 48 |
1 files changed, 24 insertions, 24 deletions
diff --git a/crypto/poly1305/asm/poly1305-armv4.pl b/crypto/poly1305/asm/poly1305-armv4.pl index 7003cc7770..38622af1ab 100755 --- a/crypto/poly1305/asm/poly1305-armv4.pl +++ b/crypto/poly1305/asm/poly1305-armv4.pl @@ -48,7 +48,6 @@ if ($flavour && $flavour ne "void") { $code.=<<___; #include "arm_arch.h" -.text #if defined(__thumb2__) .syntax unified .thumb @@ -56,6 +55,8 @@ $code.=<<___; .code 32 #endif +.text + .globl poly1305_emit .globl poly1305_blocks .globl poly1305_init @@ -100,8 +101,10 @@ poly1305_init: and r4,r4,r10 #if __ARM_MAX_ARCH__>=7 +# if !defined(_WIN32) ldr r12,[r11,r12] @ OPENSSL_armcap_P -# ifdef __APPLE__ +# endif +# if defined(__APPLE__) || defined(_WIN32) ldr r12,[r12] # endif #endif @@ -116,31 +119,21 @@ poly1305_init: #if __ARM_MAX_ARCH__>=7 tst r12,#ARMV7_NEON @ check for NEON -# ifdef __APPLE__ - adr r9,poly1305_blocks_neon - adr r11,poly1305_blocks -# ifdef __thumb2__ - it ne -# endif +# ifdef __thumb2__ + adr r9,.Lpoly1305_blocks_neon + adr r11,.Lpoly1305_blocks + adr r12,.Lpoly1305_emit + adr r10,.Lpoly1305_emit_neon + itt ne movne r11,r9 - adr r12,poly1305_emit - adr r10,poly1305_emit_neon -# ifdef __thumb2__ - it ne -# endif movne r12,r10 + orr r11,r11,#1 @ thumb-ify address + orr r12,r12,#1 # else -# ifdef __thumb2__ - itete eq -# endif - addeq r12,r11,#(poly1305_emit-.Lpoly1305_init) - addne r12,r11,#(poly1305_emit_neon-.Lpoly1305_init) - addeq r11,r11,#(poly1305_blocks-.Lpoly1305_init) - addne r11,r11,#(poly1305_blocks_neon-.Lpoly1305_init) -# endif -# ifdef __thumb2__ - orr r12,r12,#1 @ thumb-ify address - orr r11,r11,#1 + addeq r12,r11,#(.Lpoly1305_emit-.Lpoly1305_init) + addne r12,r11,#(.Lpoly1305_emit_neon-.Lpoly1305_init) + addeq r11,r11,#(.Lpoly1305_blocks-.Lpoly1305_init) + addne r11,r11,#(.Lpoly1305_blocks_neon-.Lpoly1305_init) # endif #endif ldrb r9,[$inp,#11] @@ -352,6 +345,7 @@ $code.=<<___; .type poly1305_emit,%function .align 5 poly1305_emit: +.Lpoly1305_emit: stmdb sp!,{r4-r11} .Lpoly1305_emit_enter: @@ -671,6 +665,7 @@ poly1305_init_neon: .type poly1305_blocks_neon,%function .align 5 poly1305_blocks_neon: +.Lpoly1305_blocks_neon: ldr ip,[$ctx,#36] @ is_base2_26 ands $len,$len,#-16 beq .Lno_data_neon @@ -1157,6 +1152,7 @@ poly1305_blocks_neon: .type poly1305_emit_neon,%function .align 5 poly1305_emit_neon: +.Lpoly1305_emit_neon: ldr ip,[$ctx,#36] @ is_base2_26 stmdb sp!,{r4-r11} @@ -1229,7 +1225,11 @@ poly1305_emit_neon: .Lzeros: .long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 .LOPENSSL_armcap: +# ifdef _WIN32 +.word OPENSSL_armcap_P +# else .word OPENSSL_armcap_P-.Lpoly1305_init +# endif #endif ___ } } |