From b56cb7c6eac9613dbbaed3b32cf121d16583810a Mon Sep 17 00:00:00 2001
From: Andy Polyakov <appro@openssl.org>
Date: Mon, 23 Jul 2007 16:18:36 +0000
Subject: [PATCH] x86*cpuid update [from HEAD].

---
 crypto/x86_64cpuid.pl | 99 ++++++++++++++++++++++++++-----------------
 crypto/x86cpuid.pl    | 31 +++++++++++++-
 2 files changed, 89 insertions(+), 41 deletions(-)

diff --git a/crypto/x86_64cpuid.pl b/crypto/x86_64cpuid.pl
index 4d88ad191b..0c5d3397d5 100644
--- a/crypto/x86_64cpuid.pl
+++ b/crypto/x86_64cpuid.pl
@@ -1,19 +1,12 @@
 #!/usr/bin/env perl
 
 $output=shift;
-$win64a=1 if ($output =~ /win64a\.[s|asm]/);
+$masm=1 if ($output =~ /\.asm/);
 open STDOUT,">$output" || die "can't open $output: $!";
 
-print<<___ if(defined($win64a));
+print<<___ if(defined($masm));
 _TEXT	SEGMENT
 PUBLIC	OPENSSL_rdtsc
-ALIGN	16
-OPENSSL_rdtsc	PROC
-	rdtsc
-	shl	rdx,32
-	or	rax,rdx
-	ret
-OPENSSL_rdtsc	ENDP
 
 PUBLIC	OPENSSL_atomic_add
 ALIGN	16
@@ -45,35 +38,16 @@ OPENSSL_wipe_cpu	PROC
 	lea	rax,QWORD PTR[rsp+8]
 	ret
 OPENSSL_wipe_cpu	ENDP
-
-OPENSSL_ia32_cpuid	PROC
-	mov	r8,rbx
-	mov	eax,1
-	cpuid
-	shl	rcx,32
-	mov	eax,edx
-	mov	rbx,r8
-	or	rax,rcx
-	ret
-OPENSSL_ia32_cpuid	ENDP
 _TEXT	ENDS
 
 CRT\$XIU	SEGMENT
 EXTRN	OPENSSL_cpuid_setup:PROC
 DQ	OPENSSL_cpuid_setup
 CRT\$XIU	ENDS
-END
+
 ___
-print<<___ if(!defined($win64a));
+print<<___ if(!defined($masm));
 .text
-.globl	OPENSSL_rdtsc
-.align	16
-OPENSSL_rdtsc:
-	rdtsc
-	shlq	\$32,%rdx
-	orq	%rdx,%rax
-	ret
-.size	OPENSSL_rdtsc,.-OPENSSL_rdtsc
 
 .globl	OPENSSL_atomic_add
 .type	OPENSSL_atomic_add,\@function
@@ -120,19 +94,66 @@ OPENSSL_wipe_cpu:
 	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
+OPENSSL_rdtsc:
+	rdtsc
+	shl	\$32,%rdx
+	or	%rdx,%rax
+	ret
+.size	OPENSSL_rdtsc,.-OPENSSL_rdtsc
+
 .globl	OPENSSL_ia32_cpuid
+.type	OPENSSL_ia32_cpuid,\@abi-omnipotent
 .align	16
 OPENSSL_ia32_cpuid:
-	movq	%rbx,%r8
-	movl	\$1,%eax
+	mov	%rbx,%r8
+
+	xor	%eax,%eax
+	cpuid
+	xor	%eax,%eax
+	cmp	\$0x756e6547,%ebx	# "Genu"
+	setne	%al
+	mov	%eax,%r9d
+	cmp	\$0x49656e69,%edx	# "ineI"
+	setne	%al
+	or	%eax,%r9d
+	cmp	\$0x6c65746e,%ecx	# "ntel"
+	setne	%al
+	or	%eax,%r9d
+
+	mov	\$1,%eax
 	cpuid
-	shlq	\$32,%rcx
-	movl	%edx,%eax
-	movq	%r8,%rbx
-	orq	%rcx,%rax
+	cmp	\$0,%r9d
+	jne	.Lnotintel
+	or	\$1<<20,%edx		# use reserved 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
+.Lnotintel:
+	bt	\$28,%edx		# test hyper-threading bit
+	jnc	.Ldone
+	shr	\$16,%ebx
+	cmp	\$1,%bl			# see if cache is shared
+	ja	.Ldone
+	and	\$0xefffffff,%edx	# ~(1<<28)
+.Ldone:
+	shl	\$32,%rcx
+	mov	%edx,%eax
+	mov	%r8,%rbx
+	or	%rcx,%rax
 	ret
 .size	OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
-
-.section	.init
-	call	OPENSSL_cpuid_setup
 ___
+close STDOUT;	# flush
diff --git a/crypto/x86cpuid.pl b/crypto/x86cpuid.pl
index c53c9bc998..6bf6a7b57f 100644
--- a/crypto/x86cpuid.pl
+++ b/crypto/x86cpuid.pl
@@ -19,13 +19,40 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
 	&pop	("eax");
 	&xor	("ecx","eax");
 	&bt	("ecx",21);
-	&jnc	(&label("nocpuid"));
+	&jnc	(&label("done"));
+	&xor	("eax","eax");
+	&cpuid	();
+	&xor	("eax","eax");
+	&cmp	("ebx",0x756e6547);	# "Genu"
+	&setne	(&LB("eax"));
+	&mov	("ebp","eax");
+	&cmp	("edx",0x49656e69);	# "ineI"
+	&setne	(&LB("eax"));
+	&or	("ebp","eax");
+	&cmp	("ecx",0x6c65746e);	# "ntel"
+	&setne	(&LB("eax"));
+	&or	("ebp","eax");
 	&mov	("eax",1);
 	&cpuid	();
-&set_label("nocpuid");
+	&cmp	("ebp",0);
+	&jne	(&label("notP4"));
+	&and	(&HB("eax"),15);	# familiy ID
+	&cmp	(&HB("eax"),15);	# P4?
+	&jne	(&label("notP4"));
+	&or	("edx",1<<20);		# use reserved bit to engage RC4_CHAR
+&set_label("notP4");
+	&bt	("edx",28);		# test hyper-threading bit
+	&jnc	(&label("done"));
+	&shr	("ebx",16);
+	&cmp	(&LB("ebx"),1);		# see if cache is shared(*)
+	&ja	(&label("done"));
+	&and	("edx",0xefffffff);	# clear hyper-threading bit if not
+&set_label("done");
 	&mov	("eax","edx");
 	&mov	("edx","ecx");
 &function_end("OPENSSL_ia32_cpuid");
+# (*)	on Core2 this value is set to 2 denoting the fact that L2
+#	cache is shared between cores.
 
 &external_label("OPENSSL_ia32cap_P");
 
-- 
2.25.1