x86_64 assembler pack to comply with updated styling x86_64-xlate.pl rules.
[openssl.git] / crypto / x86_64cpuid.pl
index 2f657ca9d8d7de7b1e8303673b5013b9572820f5..c54b9e3681c6cebba14a9509a43c13865079dccc 100644 (file)
 #!/usr/bin/env perl
 
-$output=shift;
-$masm=1 if ($output =~ /\.asm/);
-open STDOUT,">$output" || die "can't open $output: $!";
+$flavour = shift;
+$output  = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
 
-print<<___ if(defined($masm));
-_TEXT  SEGMENT
-PUBLIC OPENSSL_rdtsc
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
 
-PUBLIC OPENSSL_atomic_add
-ALIGN  16
-OPENSSL_atomic_add     PROC
-       mov     eax,DWORD PTR[rcx]
-\$Lspin:       lea     r8,DWORD PTR[rdx+rax]
-lock   cmpxchg DWORD PTR[rcx],r8d
-       jne     \$Lspin
-       mov     eax,r8d
-       cdqe    
-       ret
-OPENSSL_atomic_add     ENDP
-
-PUBLIC OPENSSL_wipe_cpu
-ALIGN  16
-OPENSSL_wipe_cpu       PROC
-       pxor    xmm0,xmm0
-       pxor    xmm1,xmm1
-       pxor    xmm2,xmm2
-       pxor    xmm3,xmm3
-       pxor    xmm4,xmm4
-       pxor    xmm5,xmm5
-       xor     rcx,rcx
-       xor     rdx,rdx
-       xor     r8,r8
-       xor     r9,r9
-       xor     r10,r10
-       xor     r11,r11
-       lea     rax,QWORD PTR[rsp+8]
-       ret
-OPENSSL_wipe_cpu       ENDP
-_TEXT  ENDS
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+open STDOUT,"| $^X ${dir}perlasm/x86_64-xlate.pl $flavour $output";
 
-CRT\$XIU       SEGMENT
-EXTRN  OPENSSL_cpuid_setup:PROC
-DQ     OPENSSL_cpuid_setup
-CRT\$XIU       ENDS
+if ($win64)    { $arg1="%rcx"; $arg2="%rdx"; }
+else           { $arg1="%rdi"; $arg2="%rsi"; }
+print<<___;
+.extern                OPENSSL_cpuid_setup
+.section       .init
+       call    OPENSSL_cpuid_setup
 
-___
-print<<___ if(!defined($masm));
 .text
 
 .globl OPENSSL_atomic_add
-.type  OPENSSL_atomic_add,\@function
+.type  OPENSSL_atomic_add,\@abi-omnipotent
 .align 16
 OPENSSL_atomic_add:
-       movl    (%rdi),%eax
-.Lspin:        leaq    (%rsi,%rax),%r8
-lock;  cmpxchgl        %r8d,(%rdi)
+       movl    ($arg1),%eax
+.Lspin:        leaq    ($arg2,%rax),%r8
+       .byte   0xf0            # lock
+       cmpxchgl        %r8d,($arg1)
        jne     .Lspin
        movl    %r8d,%eax
-       .byte   0x48,0x98
+       .byte   0x48,0x98       # cltq/cdqe
        ret
 .size  OPENSSL_atomic_add,.-OPENSSL_atomic_add
 
-.globl OPENSSL_wipe_cpu
-.type  OPENSSL_wipe_cpu,\@function
-.align 16
-OPENSSL_wipe_cpu:
-       pxor    %xmm0,%xmm0
-       pxor    %xmm1,%xmm1
-       pxor    %xmm2,%xmm2
-       pxor    %xmm3,%xmm3
-       pxor    %xmm4,%xmm4
-       pxor    %xmm5,%xmm5
-       pxor    %xmm6,%xmm6
-       pxor    %xmm7,%xmm7
-       pxor    %xmm8,%xmm8
-       pxor    %xmm9,%xmm9
-       pxor    %xmm10,%xmm10
-       pxor    %xmm11,%xmm11
-       pxor    %xmm12,%xmm12
-       pxor    %xmm13,%xmm13
-       pxor    %xmm14,%xmm14
-       pxor    %xmm15,%xmm15
-       xorq    %rcx,%rcx
-       xorq    %rdx,%rdx
-       xorq    %rsi,%rsi
-       xorq    %rdi,%rdi
-       xorq    %r8,%r8
-       xorq    %r9,%r9
-       xorq    %r10,%r10
-       xorq    %r11,%r11
-       leaq    8(%rsp),%rax
-       ret
-.size  OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
-
-.section       .init
-       call    OPENSSL_cpuid_setup
-
-___
-
-open STDOUT,"| $^X perlasm/x86_64-xlate.pl $output";
-print<<___;
-.text
-
 .globl OPENSSL_rdtsc
 .type  OPENSSL_rdtsc,\@abi-omnipotent
 .align 16
@@ -134,16 +63,16 @@ OPENSSL_ia32_cpuid:
 
        mov     \$1,%eax
        cpuid
-       bt      \$28,%edx               # test hyper-threading bit
-       jnc     .Ldone
        cmp     \$0,%r9d
        jne     .Lnotintel
-       or      \$1<<20,%edx            # use reserved bit to engage RC4_CHAR
+       or      \$0x00100000,%edx       # use reserved 20th bit to engage RC4_CHAR
        and     \$15,%ah
        cmp     \$15,%ah                # examine Family ID
        je      .Lnotintel
-       or      \$1<<30,%edx            # use reserved bit to skip unrolled loop
+       or      \$0x40000000,%edx       # use reserved bit to skip unrolled loop
 .Lnotintel:
+       bt      \$28,%edx               # test hyper-threading bit
+       jnc     .Ldone
        shr     \$16,%ebx
        cmp     \$1,%bl                 # see if cache is shared
        ja      .Ldone
@@ -157,34 +86,91 @@ OPENSSL_ia32_cpuid:
 .size  OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
 
 .globl  OPENSSL_cleanse
-.type   OPENSSL_cleanse,\@function,2
+.type   OPENSSL_cleanse,\@abi-omnipotent
 .align  16
 OPENSSL_cleanse:
        xor     %rax,%rax
-       cmp     \$15,%rsi
+       cmp     \$15,$arg2
        jae     .Lot
 .Little:
-       mov     %al,(%rdi)
-       sub     \$1,%rsi
-       lea     1(%rdi),%rdi
+       mov     %al,($arg1)
+       sub     \$1,$arg2
+       lea     1($arg1),$arg1
        jnz     .Little
        ret
 .align 16
 .Lot:
-       test    \$7,%rdi
+       test    \$7,$arg1
        jz      .Laligned
-       mov     %al,(%rdi)
-       lea     -1(%rsi),%rsi
-       lea     1(%rdi),%rdi
+       mov     %al,($arg1)
+       lea     -1($arg2),$arg2
+       lea     1($arg1),$arg1
        jmp     .Lot
 .Laligned:
-       mov     %rax,(%rdi)
-       lea     -8(%rsi),%rsi
-       test    \$-8,%rsi
-       lea     8(%rdi),%rdi
+       mov     %rax,($arg1)
+       lea     -8($arg2),$arg2
+       test    \$-8,$arg2
+       lea     8($arg1),$arg1
        jnz     .Laligned
-       cmp     \$0,%rsi
+       cmp     \$0,$arg2
        jne     .Little
        ret
 .size  OPENSSL_cleanse,.-OPENSSL_cleanse
 ___
+
+print<<___ if (!$win64);
+.globl OPENSSL_wipe_cpu
+.type  OPENSSL_wipe_cpu,\@abi-omnipotent
+.align 16
+OPENSSL_wipe_cpu:
+       pxor    %xmm0,%xmm0
+       pxor    %xmm1,%xmm1
+       pxor    %xmm2,%xmm2
+       pxor    %xmm3,%xmm3
+       pxor    %xmm4,%xmm4
+       pxor    %xmm5,%xmm5
+       pxor    %xmm6,%xmm6
+       pxor    %xmm7,%xmm7
+       pxor    %xmm8,%xmm8
+       pxor    %xmm9,%xmm9
+       pxor    %xmm10,%xmm10
+       pxor    %xmm11,%xmm11
+       pxor    %xmm12,%xmm12
+       pxor    %xmm13,%xmm13
+       pxor    %xmm14,%xmm14
+       pxor    %xmm15,%xmm15
+       xorq    %rcx,%rcx
+       xorq    %rdx,%rdx
+       xorq    %rsi,%rsi
+       xorq    %rdi,%rdi
+       xorq    %r8,%r8
+       xorq    %r9,%r9
+       xorq    %r10,%r10
+       xorq    %r11,%r11
+       leaq    8(%rsp),%rax
+       ret
+.size  OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
+___
+print<<___ if ($win64);
+.globl OPENSSL_wipe_cpu
+.type  OPENSSL_wipe_cpu,\@abi-omnipotent
+.align 16
+OPENSSL_wipe_cpu:
+       pxor    %xmm0,%xmm0
+       pxor    %xmm1,%xmm1
+       pxor    %xmm2,%xmm2
+       pxor    %xmm3,%xmm3
+       pxor    %xmm4,%xmm4
+       pxor    %xmm5,%xmm5
+       xorq    %rcx,%rcx
+       xorq    %rdx,%rdx
+       xorq    %r8,%r8
+       xorq    %r9,%r9
+       xorq    %r10,%r10
+       xorq    %r11,%r11
+       leaq    8(%rsp),%rax
+       ret
+.size  OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
+___
+
+close STDOUT;  # flush