From b9d89f8aaa82845d09463f36829e189b0e28c15d Mon Sep 17 00:00:00 2001
From: Xianyi Zhang <xianyi@iscas.ac.cn>
Date: Wed, 31 Aug 2011 18:21:37 +0800
Subject: [PATCH 1/2] Fixed the bug about installation. f77blas.h works OK now.

---
 Makefile                   |  7 +++++++
 common_interface.h         | 11 +++++++++++
 openblas_config_template.h | 20 ++++++++++++++++++++
 3 files changed, 38 insertions(+)

diff --git a/Makefile b/Makefile
index 6789272a3..d86fbadf3 100644
--- a/Makefile
+++ b/Makefile
@@ -118,6 +118,13 @@ endif
 #Save the config files for installation
 	cp Makefile.conf Makefile.conf_last
 	cp config.h config_last.h
+ifdef QUAD_PRECISION
+	echo "#define QUAD_PRECISION">> config_last.h
+endif
+ifeq ($(EXPRECISION), 1)
+	echo "#define EXPRECISION">> config_last.h
+endif
+## 
 ifdef DYNAMIC_ARCH
 	  $(MAKE) -C kernel commonlibs || exit 1
 	for d in $(DYNAMIC_CORE) ; \
diff --git a/common_interface.h b/common_interface.h
index 36bf5aa48..898d91001 100644
--- a/common_interface.h
+++ b/common_interface.h
@@ -38,6 +38,11 @@
 
 #ifndef ASSEMBLER
 
+#ifdef __cplusplus
+extern "C" {
+	/* Assume C declarations for C++ */
+#endif  /* __cplusplus */
+
 int    BLASFUNC(xerbla)(char *, blasint *info, blasint);
 
 FLOATRET  BLASFUNC(sdot)  (blasint *, float  *, blasint *, float  *, blasint *);
@@ -733,4 +738,10 @@ xdouble   BLASFUNC(qlamch)(char *);
 FLOATRET  BLASFUNC(slamc3)(float *, float *);
 double    BLASFUNC(dlamc3)(double *, double *);
 xdouble   BLASFUNC(qlamc3)(xdouble *, xdouble *);
+
+#ifdef __cplusplus
+}
+     
+#endif  /* __cplusplus */
+
 #endif
diff --git a/openblas_config_template.h b/openblas_config_template.h
index 9fb80aa4f..8bf972593 100644
--- a/openblas_config_template.h
+++ b/openblas_config_template.h
@@ -6,6 +6,16 @@
 #define BLASFUNC(FUNC) FUNC
 #endif
 
+#ifdef QUAD_PRECISION
+typedef struct {
+  unsigned long x[2];
+}  xdouble;
+#elif defined EXPRECISION
+#define xdouble long double
+#else
+#define xdouble double
+#endif
+
 #if defined(OS_WINDOWS) && defined(__64BIT__)
 typedef long long BLASLONG;
 typedef unsigned long long BLASULONG;
@@ -19,3 +29,13 @@ typedef BLASLONG blasint;
 #else
 typedef int blasint;
 #endif
+
+#if defined(XDOUBLE) || defined(DOUBLE)
+#define FLOATRET	FLOAT
+#else
+#ifdef NEED_F2CCONV
+#define FLOATRET	double
+#else
+#define FLOATRET	float
+#endif
+#endif

From 7b410b7f0e94edde2a606593086694ae6bb17be8 Mon Sep 17 00:00:00 2001
From: Zhang Xiianyi <traits@debian32bits.rdcps.ac.cn>
Date: Wed, 14 Sep 2011 23:52:51 +0800
Subject: [PATCH 2/2] Fixed #58 zdot SEGFAULT bug with GCC-4.6. Thank Mr. John
 for this patch.

In i386 calling convention, the caller put the address of return value of zdot into the first hidden parameter.
Thus, the callee should delete this address before return.
Actually, I have fixed the same bug on x86/zdot_sse2.S (issue #32). However, that is not a good implementation which uses 3 instructions. Mr. John told me used "ret $0x4" to skip the first hidden address (4 bytes).
---
 kernel/x86/xdot.S      | 10 +++++++++-
 kernel/x86/zdot.S      | 10 +++++++++-
 kernel/x86/zdot_sse2.S |  4 +---
 3 files changed, 19 insertions(+), 5 deletions(-)

diff --git a/kernel/x86/xdot.S b/kernel/x86/xdot.S
index 4a5af4642..929763271 100644
--- a/kernel/x86/xdot.S
+++ b/kernel/x86/xdot.S
@@ -307,7 +307,11 @@
 	popl	%ebx
 	popl	%esi
 	popl	%edi
+#if defined(F_INTERFACE) && defined(RETURN_BY_STACK)
+	ret	$0x4
+#else
 	ret
+#endif
 	ALIGN_3
 
 .L88:
@@ -326,6 +330,10 @@
 	popl	%ebx
 	popl	%esi
 	popl	%edi
-	ret
 
+#if defined(F_INTERFACE) && defined(RETURN_BY_STACK)
+	ret	$0x4
+#else
+	ret
+#endif
 	EPILOGUE
diff --git a/kernel/x86/zdot.S b/kernel/x86/zdot.S
index aa4481f97..9d8866ad0 100644
--- a/kernel/x86/zdot.S
+++ b/kernel/x86/zdot.S
@@ -283,7 +283,11 @@
 	popl	%ebx
 	popl	%esi
 	popl	%edi
+#if defined(DOUBLE) || defined(XDOUBLE)
+	ret	$0x4
+#else
 	ret
+#endif
 	ALIGN_3
 
 .L88:
@@ -305,6 +309,10 @@
 	popl	%ebx
 	popl	%esi
 	popl	%edi
-	ret
 
+#if defined(DOUBLE) || defined(XDOUBLE)
+	ret	$0x4
+#else
+	ret
+#endif
 	EPILOGUE
diff --git a/kernel/x86/zdot_sse2.S b/kernel/x86/zdot_sse2.S
index 2a174fb5d..efebe637b 100644
--- a/kernel/x86/zdot_sse2.S
+++ b/kernel/x86/zdot_sse2.S
@@ -1542,7 +1542,5 @@
 	popl	%esi
 	popl	%edi
 /*remove the hidden return value address from the stack.*/
-	popl	%ecx
-	xchgl	%ecx, 0(%esp)
-	ret
+	ret	$0x4
 	EPILOGUE