2 # Copyright 2005-2016 The OpenSSL Project Authors. All Rights Reserved.
4 # Licensed under the OpenSSL license (the "License"). You may not use
5 # this file except in compliance with the License. You can obtain a copy
6 # in the file LICENSE in the source distribution or at
7 # https://www.openssl.org/source/license.html
12 if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
14 $win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
16 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
17 ( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
18 ( $xlate="${dir}perlasm/x86_64-xlate.pl" and -f $xlate) or
19 die "can't locate x86_64-xlate.pl";
21 open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
24 ($arg1,$arg2,$arg3,$arg4)=$win64?("%rcx","%rdx","%r8", "%r9") : # Win64 order
25 ("%rdi","%rsi","%rdx","%rcx"); # Unix order
28 .extern OPENSSL_cpuid_setup
29 .hidden OPENSSL_cpuid_setup
31 call OPENSSL_cpuid_setup
33 .hidden OPENSSL_ia32cap_P
34 .comm OPENSSL_ia32cap_P,16,4
38 .globl OPENSSL_atomic_add
39 .type OPENSSL_atomic_add,\@abi-omnipotent
43 .Lspin: leaq ($arg2,%rax),%r8
48 .byte 0x48,0x98 # cltq/cdqe
50 .size OPENSSL_atomic_add,.-OPENSSL_atomic_add
53 .type OPENSSL_rdtsc,\@abi-omnipotent
60 .size OPENSSL_rdtsc,.-OPENSSL_rdtsc
62 .globl OPENSSL_ia32_cpuid
63 .type OPENSSL_ia32_cpuid,\@function,1
67 mov %rbx,%r8 # save %rbx
68 .cfi_register %rbx,%r8
71 mov %rax,8(%rdi) # clear extended feature flags
73 mov %eax,%r11d # max value for standard query level
76 cmp \$0x756e6547,%ebx # "Genu"
79 cmp \$0x49656e69,%edx # "ineI"
82 cmp \$0x6c65746e,%ecx # "ntel"
84 or %eax,%r9d # 0 indicates Intel CPU
87 cmp \$0x68747541,%ebx # "Auth"
90 cmp \$0x69746E65,%edx # "enti"
93 cmp \$0x444D4163,%ecx # "cAMD"
95 or %eax,%r10d # 0 indicates AMD CPU
101 cmp \$0x80000001,%eax
104 mov \$0x80000001,%eax
107 and \$0x00000801,%r9d # isolate AMD XOP bit, 1<<11
109 cmp \$0x80000008,%r10d
112 mov \$0x80000008,%eax
114 movzb %cl,%r10 # number of cores - 1
115 inc %r10 # number of cores
119 bt \$28,%edx # test hyper-threading bit
121 shr \$16,%ebx # number of logical processors
124 and \$0xefffffff,%edx # ~(1<<28)
133 mov \$0,%ecx # query L1D
137 and \$0xfff,%r10d # number of cores -1 per L1D
142 and \$0xbfefffff,%edx # force reserved bits to 0
145 or \$0x40000000,%edx # set reserved bit#30 on Intel CPUs
147 cmp \$15,%ah # examine Family ID
149 or \$0x00100000,%edx # set reserved bit#20 to engage RC4_CHAR
153 and \$0x0fff0ff0,%eax
154 cmp \$0x00050670,%eax # Knights Landing
156 cmp \$0x00080650,%eax # Knights Mill (according to sde)
159 and \$0xfbffffff,%ecx # clear XSAVE flag to mimic Silvermont
162 bt \$28,%edx # test hyper-threading bit
164 and \$0xefffffff,%edx # ~(1<<28)
168 or \$0x10000000,%edx # 1<<28
170 cmp \$1,%bl # see if cache is shared
172 and \$0xefffffff,%edx # ~(1<<28)
174 and \$0x00000800,%r9d # isolate AMD XOP flag
175 and \$0xfffff7ff,%ecx
176 or %ecx,%r9d # merge AMD XOP flag
178 mov %edx,%r10d # %r9d:%r10d is copy of %ecx:%edx
181 jb .Lno_extended_info
185 bt \$26,%r9d # check XSAVE bit, cleared on Knights
187 and \$0xfff7ffff,%ebx # clear ADCX/ADOX flag
189 mov %ebx,8(%rdi) # save extended feature flags
193 bt \$27,%r9d # check OSXSAVE bit
196 .byte 0x0f,0x01,0xd0 # xgetbv
197 and \$0xe6,%eax # isolate XMM, YMM and ZMM state support
200 andl \$0xfffeffff,8(%rdi) # clear AVX512F, ~(1<<16)
201 # note that we don't touch other AVX512
202 # extensions, because they can be used
203 # with YMM (without opmasking though)
204 and \$6,%eax # isolate XMM and YMM state support
208 mov \$0xefffe7ff,%eax # ~(1<<28|1<<12|1<<11)
209 and %eax,%r9d # clear AVX, FMA and AMD XOP bits
210 mov \$0x3fdeffdf,%eax # ~(1<<31|1<<30|1<<21|1<<16|1<<5)
211 and %eax,8(%rdi) # clear AVX2 and AVX512* bits
215 mov %r8,%rbx # restore %rbx
220 .size OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
222 .globl OPENSSL_cleanse
223 .type OPENSSL_cleanse,\@abi-omnipotent
255 .size OPENSSL_cleanse,.-OPENSSL_cleanse
258 .type CRYPTO_memcmp,\@abi-omnipotent
277 .size CRYPTO_memcmp,.-CRYPTO_memcmp
280 print<<___ if (!$win64);
281 .globl OPENSSL_wipe_cpu
282 .type OPENSSL_wipe_cpu,\@abi-omnipotent
311 .size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
313 print<<___ if ($win64);
314 .globl OPENSSL_wipe_cpu
315 .type OPENSSL_wipe_cpu,\@abi-omnipotent
332 .size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
340 my $redzone=win64?8:-8;
343 .globl OPENSSL_instrument_bus
344 .type OPENSSL_instrument_bus,\@abi-omnipotent
346 OPENSSL_instrument_bus:
347 mov $arg1,$out # tribute to Win64
351 rdtsc # collect 1st tick
352 mov %eax,$lasttick # lasttick = tick
353 mov \$0,$lastdiff # lastdiff = 0
373 .size OPENSSL_instrument_bus,.-OPENSSL_instrument_bus
375 .globl OPENSSL_instrument_bus2
376 .type OPENSSL_instrument_bus2,\@abi-omnipotent
378 OPENSSL_instrument_bus2:
379 mov $arg1,$out # tribute to Win64
382 mov $cnt,$redzone(%rsp)
384 rdtsc # collect 1st tick
385 mov %eax,$lasttick # lasttick = tick
386 mov \$0,$lastdiff # lastdiff = 0
392 rdtsc # collect 1st diff
394 sub $lasttick,%eax # diff
395 mov %edx,$lasttick # lasttick = tick
396 mov %eax,$lastdiff # lastdiff = diff
400 add %eax,($out) # accumulate diff
407 sub $lasttick,%eax # diff
408 mov %edx,$lasttick # lasttick = tick
410 mov %eax,$lastdiff # lastdiff = diff
413 sub %rdx,$cnt # conditional --$cnt
414 lea ($out,%rdx,4),$out # conditional ++$out
418 mov $redzone(%rsp),%rax
421 .size OPENSSL_instrument_bus2,.-OPENSSL_instrument_bus2
428 .globl OPENSSL_ia32_${rdop}
429 .type OPENSSL_ia32_${rdop},\@abi-omnipotent
431 OPENSSL_ia32_${rdop}:
441 .size OPENSSL_ia32_${rdop},.-OPENSSL_ia32_${rdop}
443 .globl OPENSSL_ia32_${rdop}_bytes
444 .type OPENSSL_ia32_${rdop}_bytes,\@abi-omnipotent
446 OPENSSL_ia32_${rdop}_bytes:
447 xor %rax, %rax # return value
449 je .Ldone_${rdop}_bytes
454 jc .Lbreak_${rdop}_bytes
456 jnz .Loop_${rdop}_bytes
457 jmp .Ldone_${rdop}_bytes
460 .Lbreak_${rdop}_bytes:
462 jb .Ltail_${rdop}_bytes
467 jz .Ldone_${rdop}_bytes
469 jmp .Loop_${rdop}_bytes
472 .Ltail_${rdop}_bytes:
478 jnz .Ltail_${rdop}_bytes
480 .Ldone_${rdop}_bytes:
482 .size OPENSSL_ia32_${rdop}_bytes,.-OPENSSL_ia32_${rdop}_bytes
485 gen_random("rdrand");
486 gen_random("rdseed");
488 close STDOUT; # flush