Avoid excess range in results from i386 exp, hypot, pow functions (bug 18980).

i386 exp, hypot and pow functions can return overflowing and
underflowing values with excess range and precision; ; Wilco
Dijkstra's patches to make isfinite etc. expand inline cause this
pre-existing issue to result in test failures.

This patch fixes those functions to avoid excess range and precision
in their return values.  Appropriate macros are added for the repeated
code sequences; in future I'll add more such macros and refactor
existing code forcing underflow (with or without also eliminating
excess range and precision from the return value) to use such macros.

Tested for x86.  If, after this patch, you still see x86 libm test
failures with excess range or precision, please file bugs in Bugzilla.

	[BZ #18980]
	* sysdeps/i386/fpu/i386-math-asm.h (DEFINE_FLT_MIN): New macro.
	(DEFINE_DBL_MIN): Likewise.
	(FLT_NARROW_EVAL_UFLOW_NONNEG_NAN): Likewise.
	(DBL_NARROW_EVAL_UFLOW_NONNEG_NAN): Likewise.
	(FLT_NARROW_EVAL_UFLOW_NONNEG): Likewise.
	(DBL_NARROW_EVAL_UFLOW_NONNEG): Likewise.
	* sysdeps/i386/fpu/e_exp.S: Include <i386-math-asm.h>.
	(dbl_min): Replace with use of DEFINE_DBL_MIN.
	(__ieee754_exp): Use DBL_NARROW_EVAL_UFLOW_NONNEG_NAN.
	(__exp_finite): Use DBL_NARROW_EVAL_UFLOW_NONNEG.
	* sysdeps/i386/fpu/e_exp10.S: Include <i386-math-asm.h>.
	(dbl_min): Replace with use of DEFINE_DBL_MIN.
	(__ieee754_exp10): Use DBL_NARROW_EVAL_UFLOW_NONNEG_NAN.
	* sysdeps/i386/fpu/e_exp10f.S: Include <i386-math-asm.h>.
	(flt_min): Replace with use of DEFINE_FLT_MIN.
	(__ieee754_exp10f): Use FLT_NARROW_EVAL_UFLOW_NONNEG_NAN.
	* sysdeps/i386/fpu/e_exp2.S: Include <i386-math-asm.h>.
	(dbl_min): Replace with use of DEFINE_DBL_MIN.
	(__ieee754_exp2): Use DBL_NARROW_EVAL_UFLOW_NONNEG_NAN.
	* sysdeps/i386/fpu/e_exp2f.S: Include <i386-math-asm.h>.
	(flt_min): Replace with use of DEFINE_FLT_MIN.
	(__ieee754_exp2f): Use FLT_NARROW_EVAL_UFLOW_NONNEG_NAN.
	* sysdeps/i386/fpu/e_expf.S: Include <i386-math-asm.h>.
	(flt_min): Replace with use of DEFINE_FLT_MIN.
	(__ieee754_expf): Use FLT_NARROW_EVAL_UFLOW_NONNEG_NAN.
	(__expf_finite): Use FLT_NARROW_EVAL_UFLOW_NONNEG.
	* sysdeps/i386/fpu/e_hypot.S: Include <i386-math-asm.h>.
	(__ieee754_hypot): Use DBL_NARROW_EVAL.
	* sysdeps/i386/fpu/e_hypotf.S: Include <i386-math-asm.h>.
	(__ieee754_hypotf): Use FLT_NARROW_EVAL.
	* sysdeps/i386/fpu/e_pow.S: Include <i386-math-asm.h>.
	(__ieee754_pow): Use DBL_NARROW_EVAL.
	* sysdeps/i386/fpu/e_powf.S: Include <i386-math-asm.h>.
	(__ieee754_powf): Use FLT_NARROW_EVAL.
	* sysdeps/i386/i686/fpu/multiarch/e_expf-sse2.S
	(__ieee754_expf_sse2): Convert double-precision result to single
	precision.
	* sysdeps/i386/fpu/libm-test-ulps: Update.

13 files changed, 249 insertions, 282 deletions

diff --git a/sysdeps/i386/fpu/e_exp.S b/sysdeps/i386/fpu/e_exp.S
index c00beed..a7e7f13 100644
--- a/sysdeps/i386/fpu/e_exp.S
+++ b/sysdeps/i386/fpu/e_exp.S
@@ -4,13 +4,9 @@
  */
 
 #include <machine/asm.h>
+#include <i386-math-asm.h>
 
-	.section .rodata.cst8,"aM",@progbits,8
-
-	.p2align 3
-	.type dbl_min,@object
-dbl_min:	.byte 0, 0, 0, 0, 0, 0, 0x10, 0
-	ASM_SIZE_DIRECTIVE(dbl_min)
+DEFINE_DBL_MIN
 
 #ifdef PIC
 # define MO(op) op##@GOTOFF(%ecx)
@@ -46,22 +42,8 @@ ENTRY(__ieee754_exp)
 	faddp				/* 2^(fract(x * log2(e))) */
 	fscale				/* e^x */
 	fstp	%st(1)
-	fldl	MO(dbl_min)
-	fld	%st(1)
-	fucompp
-	fnstsw
-	sahf
-	jnc 3f
-	subl	$8, %esp
-	cfi_adjust_cfa_offset (8)
-	fld	%st(0)
-	fmul	%st(0)
-	fstpl	(%esp)
-	fstpl	(%esp)
-	fldl	(%esp)
-	addl	$8, %esp
-	cfi_adjust_cfa_offset (-8)
-3:	ret
+	DBL_NARROW_EVAL_UFLOW_NONNEG_NAN
+	ret
 
 1:	testl	$0x200, %eax		/* Test sign.  */
 	jz	2f			/* If positive, jump.  */
@@ -86,20 +68,6 @@ ENTRY(__exp_finite)
 	faddp				/* 2^(fract(x * log2(e))) */
 	fscale				/* e^x */
 	fstp	%st(1)
-	fldl	MO(dbl_min)
-	fld	%st(1)
-	fucompp
-	fnstsw
-	sahf
-	jnc 4f
-	subl	$8, %esp
-	cfi_adjust_cfa_offset (8)
-	fld	%st(0)
-	fmul	%st(0)
-	fstpl	(%esp)
-	fstpl	(%esp)
-	fldl	(%esp)
-	addl	$8, %esp
-	cfi_adjust_cfa_offset (-8)
-4:	ret
+	DBL_NARROW_EVAL_UFLOW_NONNEG
+	ret
 END(__exp_finite)
diff --git a/sysdeps/i386/fpu/e_exp10.S b/sysdeps/i386/fpu/e_exp10.S
index fa54732..acb5160 100644
--- a/sysdeps/i386/fpu/e_exp10.S
+++ b/sysdeps/i386/fpu/e_exp10.S
@@ -3,13 +3,9 @@
  */
 
 #include <machine/asm.h>
+#include <i386-math-asm.h>
 
-	.section .rodata.cst8,"aM",@progbits,8
-
-	.p2align 3
-	.type dbl_min,@object
-dbl_min:	.byte 0, 0, 0, 0, 0, 0, 0x10, 0
-	ASM_SIZE_DIRECTIVE(dbl_min)
+DEFINE_DBL_MIN
 
 #ifdef PIC
 # define MO(op) op##@GOTOFF(%ecx)
@@ -45,22 +41,8 @@ ENTRY(__ieee754_exp10)
 	faddp				/* 2^(fract(x * log2(10))) */
 	fscale				/* e^x */
 	fstp	%st(1)
-	fldl	MO(dbl_min)
-	fld	%st(1)
-	fucompp
-	fnstsw
-	sahf
-	jnc 3f
-	subl	$8, %esp
-	cfi_adjust_cfa_offset (8)
-	fld	%st(0)
-	fmul	%st(0)
-	fstpl	(%esp)
-	fstpl	(%esp)
-	fldl	(%esp)
-	addl	$8, %esp
-	cfi_adjust_cfa_offset (-8)
-3:	ret
+	DBL_NARROW_EVAL_UFLOW_NONNEG_NAN
+	ret
 
 1:	testl	$0x200, %eax		/* Test sign.  */
 	jz	2f			/* If positive, jump.  */
diff --git a/sysdeps/i386/fpu/e_exp10f.S b/sysdeps/i386/fpu/e_exp10f.S
index a84b2ae..1812b34 100644
--- a/sysdeps/i386/fpu/e_exp10f.S
+++ b/sysdeps/i386/fpu/e_exp10f.S
@@ -3,13 +3,9 @@
  */
 
 #include <machine/asm.h>
+#include <i386-math-asm.h>
 
-	.section .rodata.cst4,"aM",@progbits,4
-
-	.p2align 2
-	.type flt_min,@object
-flt_min:	.byte 0, 0, 0x80, 0
-	ASM_SIZE_DIRECTIVE(flt_min)
+DEFINE_FLT_MIN
 
 #ifdef PIC
 # define MO(op) op##@GOTOFF(%ecx)
@@ -45,22 +41,8 @@ ENTRY(__ieee754_exp10f)
 	faddp				/* 2^(fract(x * log2(10))) */
 	fscale				/* e^x */
 	fstp	%st(1)
-	flds	MO(flt_min)
-	fld	%st(1)
-	fucompp
-	fnstsw
-	sahf
-	jnc 3f
-	subl	$4, %esp
-	cfi_adjust_cfa_offset (4)
-	fld	%st(0)
-	fmul	%st(0)
-	fstps	(%esp)
-	fstps	(%esp)
-	flds	(%esp)
-	addl	$4, %esp
-	cfi_adjust_cfa_offset (-4)
-3:	ret
+	FLT_NARROW_EVAL_UFLOW_NONNEG_NAN
+	ret
 
 1:	testl	$0x200, %eax		/* Test sign.  */
 	jz	2f			/* If positive, jump.  */
diff --git a/sysdeps/i386/fpu/e_exp2.S b/sysdeps/i386/fpu/e_exp2.S
index b75a63a..fc16a96 100644
--- a/sysdeps/i386/fpu/e_exp2.S
+++ b/sysdeps/i386/fpu/e_exp2.S
@@ -5,13 +5,9 @@
  */
 
 #include <machine/asm.h>
+#include <i386-math-asm.h>
 
-	.section .rodata.cst8,"aM",@progbits,8
-
-	.p2align 3
-	.type dbl_min,@object
-dbl_min:	.byte 0, 0, 0, 0, 0, 0, 0x10, 0
-	ASM_SIZE_DIRECTIVE(dbl_min)
+DEFINE_DBL_MIN
 
 #ifdef PIC
 # define MO(op) op##@GOTOFF(%ecx)
@@ -44,22 +40,8 @@ ENTRY(__ieee754_exp2)
 	faddp				/* 2^(fract(x)) */
 	fscale				/* e^x */
 	fstp	%st(1)
-	fldl	MO(dbl_min)
-	fld	%st(1)
-	fucompp
-	fnstsw
-	sahf
-	jnc 3f
-	subl	$8, %esp
-	cfi_adjust_cfa_offset (8)
-	fld	%st(0)
-	fmul	%st(0)
-	fstpl	(%esp)
-	fstpl	(%esp)
-	fldl	(%esp)
-	addl	$8, %esp
-	cfi_adjust_cfa_offset (-8)
-3:	ret
+	DBL_NARROW_EVAL_UFLOW_NONNEG_NAN
+	ret
 
 1:	testl	$0x200, %eax		/* Test sign.  */
 	jz	2f			/* If positive, jump.  */
diff --git a/sysdeps/i386/fpu/e_exp2f.S b/sysdeps/i386/fpu/e_exp2f.S
index 042c83b..30623cd 100644
--- a/sysdeps/i386/fpu/e_exp2f.S
+++ b/sysdeps/i386/fpu/e_exp2f.S
@@ -5,13 +5,9 @@
  */
 
 #include <machine/asm.h>
+#include <i386-math-asm.h>
 
-	.section .rodata.cst4,"aM",@progbits,4
-
-	.p2align 2
-	.type flt_min,@object
-flt_min:	.byte 0, 0, 0x80, 0
-	ASM_SIZE_DIRECTIVE(flt_min)
+DEFINE_FLT_MIN
 
 #ifdef PIC
 # define MO(op) op##@GOTOFF(%ecx)
@@ -44,22 +40,8 @@ ENTRY(__ieee754_exp2f)
 	faddp				/* 2^(fract(x)) */
 	fscale				/* e^x */
 	fstp	%st(1)
-	flds	MO(flt_min)
-	fld	%st(1)
-	fucompp
-	fnstsw
-	sahf
-	jnc 3f
-	subl	$4, %esp
-	cfi_adjust_cfa_offset (4)
-	fld	%st(0)
-	fmul	%st(0)
-	fstps	(%esp)
-	fstps	(%esp)
-	flds	(%esp)
-	addl	$4, %esp
-	cfi_adjust_cfa_offset (-4)
-3:	ret
+	FLT_NARROW_EVAL_UFLOW_NONNEG_NAN
+	ret
 
 1:	testl	$0x200, %eax		/* Test sign.  */
 	jz	2f			/* If positive, jump.  */
diff --git a/sysdeps/i386/fpu/e_expf.S b/sysdeps/i386/fpu/e_expf.S
index 306afd1..65cb4ec 100644
--- a/sysdeps/i386/fpu/e_expf.S
+++ b/sysdeps/i386/fpu/e_expf.S
@@ -5,13 +5,9 @@
  */
 
 #include <machine/asm.h>
+#include <i386-math-asm.h>
 
-	.section .rodata.cst4,"aM",@progbits,4
-
-	.p2align 2
-	.type flt_min,@object
-flt_min:	.byte 0, 0, 0x80, 0
-	ASM_SIZE_DIRECTIVE(flt_min)
+DEFINE_FLT_MIN
 
 #ifdef PIC
 # define MO(op) op##@GOTOFF(%ecx)
@@ -47,22 +43,8 @@ ENTRY(__ieee754_expf)
 	faddp				/* 2^(fract(x * log2(e))) */
 	fscale				/* e^x */
 	fstp	%st(1)
-	flds	MO(flt_min)
-	fld	%st(1)
-	fucompp
-	fnstsw
-	sahf
-	jnc 3f
-	subl	$4, %esp
-	cfi_adjust_cfa_offset (4)
-	fld	%st(0)
-	fmul	%st(0)
-	fstps	(%esp)
-	fstps	(%esp)
-	flds	(%esp)
-	addl	$4, %esp
-	cfi_adjust_cfa_offset (-4)
-3:	ret
+	FLT_NARROW_EVAL_UFLOW_NONNEG_NAN
+	ret
 
 1:	testl	$0x200, %eax		/* Test sign.  */
 	jz	2f			/* If positive, jump.  */
@@ -87,20 +69,6 @@ ENTRY(__expf_finite)
 	faddp				/* 2^(fract(x * log2(e))) */
 	fscale				/* e^x */
 	fstp	%st(1)
-	flds	MO(flt_min)
-	fld	%st(1)
-	fucompp
-	fnstsw
-	sahf
-	jnc 4f
-	subl	$4, %esp
-	cfi_adjust_cfa_offset (4)
-	fld	%st(0)
-	fmul	%st(0)
-	fstps	(%esp)
-	fstps	(%esp)
-	flds	(%esp)
-	addl	$4, %esp
-	cfi_adjust_cfa_offset (-4)
-4:	ret
+	FLT_NARROW_EVAL_UFLOW_NONNEG
+	ret
 END(__expf_finite)
diff --git a/sysdeps/i386/fpu/e_hypot.S b/sysdeps/i386/fpu/e_hypot.S
index 63083ad..5323fde 100644
--- a/sysdeps/i386/fpu/e_hypot.S
+++ b/sysdeps/i386/fpu/e_hypot.S
@@ -18,6 +18,7 @@
    <http://www.gnu.org/licenses/>.  */
 
 #include <sysdep.h>
+#include <i386-math-asm.h>
 
 	.text
 ENTRY(__ieee754_hypot)
@@ -37,6 +38,7 @@ ENTRY(__ieee754_hypot)
 	fmul	%st(0)		// x * x : y * y
 	faddp			// x * x + y * y
 	fsqrt
+	DBL_NARROW_EVAL
 2:	ret
 
 	// We have to test whether any of the parameters is Inf.
diff --git a/sysdeps/i386/fpu/e_hypotf.S b/sysdeps/i386/fpu/e_hypotf.S
index 4e22d33..fd11ea7 100644
--- a/sysdeps/i386/fpu/e_hypotf.S
+++ b/sysdeps/i386/fpu/e_hypotf.S
@@ -18,6 +18,7 @@
    <http://www.gnu.org/licenses/>.  */
 
 #include <sysdep.h>
+#include <i386-math-asm.h>
 
 	.text
 ENTRY(__ieee754_hypotf)
@@ -37,6 +38,7 @@ ENTRY(__ieee754_hypotf)
 	fmul	%st(0)		// x * x : y * y
 	faddp			// x * x + y * y
 	fsqrt
+	FLT_NARROW_EVAL
 2:	ret
 
 	// We have to test whether any of the parameters is Inf.
diff --git a/sysdeps/i386/fpu/e_pow.S b/sysdeps/i386/fpu/e_pow.S
index 40f8227..2903e13 100644
--- a/sysdeps/i386/fpu/e_pow.S
+++ b/sysdeps/i386/fpu/e_pow.S
@@ -18,6 +18,7 @@
    <http://www.gnu.org/licenses/>.  */
 
 #include <machine/asm.h>
+#include <i386-math-asm.h>
 
 	.section .rodata.cst8,"aM",@progbits,8
 
@@ -165,6 +166,7 @@ ENTRY(__ieee754_pow)
 	orl	%edx, %ecx
 	jnz	6b
 	fstp	%st(0)		// ST*x
+	DBL_NARROW_EVAL
 	ret
 
 	/* y is ±NAN */
@@ -257,6 +259,7 @@ ENTRY(__ieee754_pow)
 	cfi_adjust_cfa_offset (-8)
 292:	fscale			// +/- 2^fract(y*log2(x))*2^int(y*log2(x)) : int(y*log2(x))
 	fstp	%st(1)		// +/- 2^fract(y*log2(x))*2^int(y*log2(x))
+	DBL_NARROW_EVAL
 	ret
 
 
diff --git a/sysdeps/i386/fpu/e_powf.S b/sysdeps/i386/fpu/e_powf.S
index 0233828..d929bdc 100644
--- a/sysdeps/i386/fpu/e_powf.S
+++ b/sysdeps/i386/fpu/e_powf.S
@@ -18,6 +18,7 @@
    <http://www.gnu.org/licenses/>.  */
 
 #include <machine/asm.h>
+#include <i386-math-asm.h>
 
 	.section .rodata.cst8,"aM",@progbits,8
 
@@ -148,6 +149,7 @@ ENTRY(__ieee754_powf)
 	testl	%edx, %edx
 	jnz	6b
 	fstp	%st(0)		// ST*x
+	FLT_NARROW_EVAL
 	ret
 
 	/* y is ±NAN */
@@ -196,6 +198,7 @@ ENTRY(__ieee754_powf)
 32:	addl	$4, %esp
 	cfi_adjust_cfa_offset (-4)
 	fstp	%st(1)		// 2^fract(y*log2(x))*2^int(y*log2(x))
+	FLT_NARROW_EVAL
 	ret
 
 
diff --git a/sysdeps/i386/fpu/i386-math-asm.h b/sysdeps/i386/fpu/i386-math-asm.h
index fd43130..c15029d 100644
--- a/sysdeps/i386/fpu/i386-math-asm.h
+++ b/sysdeps/i386/fpu/i386-math-asm.h
@@ -36,4 +36,91 @@
 	addl	$8, %esp;			\
 	cfi_adjust_cfa_offset (-8);
 
+/* Define constants for the minimum value of a floating-point
+   type.  */
+#define DEFINE_FLT_MIN				\
+	.section .rodata.cst4,"aM",@progbits,4;	\
+	.p2align 2;				\
+	.type flt_min,@object;			\
+flt_min:					\
+	.byte 0, 0, 0x80, 0;			\
+	.size flt_min, .-flt_min;
+#define DEFINE_DBL_MIN				\
+	.section .rodata.cst8,"aM",@progbits,8;	\
+	.p2align 3;				\
+	.type dbl_min,@object;			\
+dbl_min:					\
+	.byte 0, 0, 0, 0, 0, 0, 0x10, 0;	\
+	.size dbl_min, .-dbl_min;
+
+/* Remove excess range and precision by storing a value on the stack
+   and loading it back.  The value is given to be nonnegative or NaN;
+   if it is subnormal, also force an underflow exception.  The
+   relevant constant for the minimum of the type must have been
+   defined, the MO macro must have been defined for access to memory
+   operands, and, if PIC, the PIC register must have been loaded.  */
+#define FLT_NARROW_EVAL_UFLOW_NONNEG_NAN	\
+	subl	$4, %esp;			\
+	cfi_adjust_cfa_offset (4);		\
+	flds	MO(flt_min);			\
+	fld	%st(1);				\
+	fucompp;				\
+	fnstsw;					\
+	sahf;					\
+	jnc 6424f;				\
+	fld	%st(0);				\
+	fmul	%st(0);				\
+	fstps	(%esp);				\
+6424:	fstps	(%esp);				\
+	flds	(%esp);				\
+	addl	$4, %esp;			\
+	cfi_adjust_cfa_offset (-4);
+#define DBL_NARROW_EVAL_UFLOW_NONNEG_NAN	\
+	subl	$8, %esp;			\
+	cfi_adjust_cfa_offset (8);		\
+	fldl	MO(dbl_min);			\
+	fld	%st(1);				\
+	fucompp;				\
+	fnstsw;					\
+	sahf;					\
+	jnc 6453f;				\
+	fld	%st(0);				\
+	fmul	%st(0);				\
+	fstpl	(%esp);				\
+6453:	fstpl	(%esp);				\
+	fldl	(%esp);				\
+	addl	$8, %esp;			\
+	cfi_adjust_cfa_offset (-8);
+
+/* Likewise, but the argument is not a NaN (so fcom instructions,
+   which support memory operands, can be used).  */
+#define FLT_NARROW_EVAL_UFLOW_NONNEG		\
+	subl	$4, %esp;			\
+	cfi_adjust_cfa_offset (4);		\
+	fcoms	MO(flt_min);			\
+	fnstsw;					\
+	sahf;					\
+	jnc 6424f;				\
+	fld	%st(0);				\
+	fmul	%st(0);				\
+	fstps	(%esp);				\
+6424:	fstps	(%esp);				\
+	flds	(%esp);				\
+	addl	$4, %esp;			\
+	cfi_adjust_cfa_offset (-4);
+#define DBL_NARROW_EVAL_UFLOW_NONNEG		\
+	subl	$8, %esp;			\
+	cfi_adjust_cfa_offset (8);		\
+	fcoml	MO(dbl_min);			\
+	fnstsw;					\
+	sahf;					\
+	jnc 6453f;				\
+	fld	%st(0);				\
+	fmul	%st(0);				\
+	fstpl	(%esp);				\
+6453:	fstpl	(%esp);				\
+	fldl	(%esp);				\
+	addl	$8, %esp;			\
+	cfi_adjust_cfa_offset (-8);
+
 #endif /* i386-math-asm.h.  */
diff --git a/sysdeps/i386/fpu/libm-test-ulps b/sysdeps/i386/fpu/libm-test-ulps
index ddaa8fc..32f24d0 100644
--- a/sysdeps/i386/fpu/libm-test-ulps
+++ b/sysdeps/i386/fpu/libm-test-ulps
@@ -238,9 +238,9 @@ ildouble: 2
 ldouble: 2
 
 Function: Imaginary part of "cacos_downward":
-double: 3
+double: 4
 float: 3
-idouble: 3
+idouble: 4
 ifloat: 3
 ildouble: 5
 ldouble: 5
@@ -254,9 +254,9 @@ ildouble: 2
 ldouble: 2
 
 Function: Imaginary part of "cacos_towardzero":
-double: 3
+double: 4
 float: 3
-idouble: 3
+idouble: 4
 ifloat: 3
 ildouble: 5
 ldouble: 5
@@ -294,9 +294,9 @@ ildouble: 1
 ldouble: 1
 
 Function: Real part of "cacosh_downward":
-double: 3
+double: 4
 float: 3
-idouble: 3
+idouble: 4
 ifloat: 3
 ildouble: 5
 ldouble: 5
@@ -310,9 +310,9 @@ ildouble: 2
 ldouble: 2
 
 Function: Real part of "cacosh_towardzero":
-double: 3
+double: 4
 float: 3
-idouble: 3
+idouble: 4
 ifloat: 3
 ildouble: 5
 ldouble: 5
@@ -396,9 +396,9 @@ ildouble: 2
 ldouble: 2
 
 Function: Imaginary part of "casin_downward":
-double: 3
+double: 4
 float: 3
-idouble: 3
+idouble: 4
 ifloat: 3
 ildouble: 5
 ldouble: 5
@@ -412,9 +412,9 @@ ildouble: 2
 ldouble: 2
 
 Function: Imaginary part of "casin_towardzero":
-double: 3
+double: 4
 float: 3
-idouble: 3
+idouble: 4
 ifloat: 3
 ildouble: 5
 ldouble: 5
@@ -452,9 +452,9 @@ ildouble: 1
 ldouble: 1
 
 Function: Real part of "casinh_downward":
-double: 3
+double: 4
 float: 3
-idouble: 3
+idouble: 4
 ifloat: 3
 ildouble: 5
 ldouble: 5
@@ -468,9 +468,9 @@ ildouble: 2
 ldouble: 2
 
 Function: Real part of "casinh_towardzero":
-double: 3
+double: 4
 float: 3
-idouble: 3
+idouble: 4
 ifloat: 3
 ildouble: 5
 ldouble: 5
@@ -668,10 +668,10 @@ ildouble: 3
 ldouble: 3
 
 Function: Imaginary part of "ccos_downward":
-double: 2
-float: 2
-idouble: 2
-ifloat: 2
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
 ildouble: 3
 ldouble: 3
 
@@ -684,25 +684,25 @@ ildouble: 3
 ldouble: 3
 
 Function: Imaginary part of "ccos_towardzero":
-double: 2
-float: 2
-idouble: 2
-ifloat: 2
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
 ildouble: 3
 ldouble: 3
 
 Function: Real part of "ccos_upward":
-double: 1
-float: 1
-idouble: 1
-ifloat: 1
+double: 2
+float: 2
+idouble: 2
+ifloat: 2
 ildouble: 2
 ldouble: 2
 
 Function: Imaginary part of "ccos_upward":
-double: 1
+double: 2
 float: 2
-idouble: 1
+idouble: 2
 ifloat: 2
 ildouble: 2
 ldouble: 2
@@ -725,77 +725,69 @@ ldouble: 1
 
 Function: Real part of "ccosh_downward":
 double: 1
-float: 1
+float: 2
 idouble: 1
-ifloat: 1
+ifloat: 2
 ildouble: 3
 ldouble: 3
 
 Function: Imaginary part of "ccosh_downward":
-double: 2
-float: 2
-idouble: 2
-ifloat: 2
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
 ildouble: 3
 ldouble: 3
 
 Function: Real part of "ccosh_towardzero":
 double: 1
-float: 1
+float: 2
 idouble: 1
-ifloat: 1
+ifloat: 2
 ildouble: 3
 ldouble: 3
 
 Function: Imaginary part of "ccosh_towardzero":
-double: 2
-float: 2
-idouble: 2
-ifloat: 2
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
 ildouble: 3
 ldouble: 3
 
 Function: Real part of "ccosh_upward":
-double: 1
-float: 1
-idouble: 1
-ifloat: 1
+double: 2
+float: 2
+idouble: 2
+ifloat: 2
 ildouble: 2
 ldouble: 2
 
 Function: Imaginary part of "ccosh_upward":
-double: 1
+double: 3
 float: 2
-idouble: 1
+idouble: 3
 ifloat: 2
 ildouble: 2
 ldouble: 2
 
 Function: Real part of "cexp":
-double: 1
+double: 2
 float: 1
-idouble: 1
+idouble: 2
 ifloat: 1
 ildouble: 1
 ldouble: 1
 
 Function: Imaginary part of "cexp":
 double: 1
-float: 1
+float: 2
 idouble: 1
-ifloat: 1
+ifloat: 2
 ildouble: 1
 ldouble: 1
 
 Function: Real part of "cexp_downward":
-double: 1
-float: 1
-idouble: 1
-ifloat: 1
-ildouble: 3
-ldouble: 3
-
-Function: Imaginary part of "cexp_downward":
 double: 2
 float: 2
 idouble: 2
@@ -803,15 +795,15 @@ ifloat: 2
 ildouble: 3
 ldouble: 3
 
-Function: Real part of "cexp_towardzero":
-double: 1
-float: 1
-idouble: 1
-ifloat: 1
+Function: Imaginary part of "cexp_downward":
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
 ildouble: 3
 ldouble: 3
 
-Function: Imaginary part of "cexp_towardzero":
+Function: Real part of "cexp_towardzero":
 double: 2
 float: 2
 idouble: 2
@@ -819,19 +811,27 @@ ifloat: 2
 ildouble: 3
 ldouble: 3
 
+Function: Imaginary part of "cexp_towardzero":
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
+ildouble: 3
+ldouble: 3
+
 Function: Real part of "cexp_upward":
 double: 1
-float: 1
+float: 2
 idouble: 1
-ifloat: 1
+ifloat: 2
 ildouble: 2
 ldouble: 2
 
 Function: Imaginary part of "cexp_upward":
-double: 1
-float: 1
-idouble: 1
-ifloat: 1
+double: 3
+float: 2
+idouble: 3
+ifloat: 2
 ildouble: 3
 ldouble: 3
 
@@ -983,27 +983,30 @@ ldouble: 2
 
 Function: "cosh":
 double: 1
+float: 1
 idouble: 1
 ildouble: 2
 ldouble: 2
 
 Function: "cosh_downward":
-double: 1
+double: 2
+float: 1
 idouble: 1
 ifloat: 1
 ildouble: 2
 ldouble: 3
 
 Function: "cosh_towardzero":
-double: 1
+double: 2
+float: 1
 idouble: 1
 ifloat: 1
 ildouble: 2
 ldouble: 2
 
 Function: "cosh_upward":
-double: 1
-float: 1
+double: 4
+float: 2
 idouble: 1
 ifloat: 1
 ildouble: 2
@@ -1019,9 +1022,9 @@ ldouble: 3
 
 Function: Imaginary part of "cpow":
 double: 1
-float: 1
+float: 2
 idouble: 1
-ifloat: 1
+ifloat: 2
 ildouble: 4
 ldouble: 4
 
@@ -1034,10 +1037,10 @@ ildouble: 7
 ldouble: 7
 
 Function: Imaginary part of "cpow_downward":
-double: 1
-float: 1
-idouble: 1
-ifloat: 1
+double: 2
+float: 2
+idouble: 2
+ifloat: 2
 ildouble: 2
 ldouble: 2
 
@@ -1050,10 +1053,10 @@ ildouble: 7
 ldouble: 7
 
 Function: Imaginary part of "cpow_towardzero":
-double: 1
-float: 1
-idouble: 1
-ifloat: 1
+double: 2
+float: 2
+idouble: 2
+ifloat: 2
 ildouble: 1
 ldouble: 1
 
@@ -1067,9 +1070,9 @@ ldouble: 2
 
 Function: Imaginary part of "cpow_upward":
 double: 1
-float: 1
+float: 2
 idouble: 1
-ifloat: 1
+ifloat: 2
 ildouble: 2
 ldouble: 2
 
@@ -1088,10 +1091,10 @@ idouble: 1
 ifloat: 1
 
 Function: Real part of "csin_downward":
-double: 2
-float: 2
-idouble: 2
-ifloat: 2
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
 ildouble: 3
 ldouble: 3
 
@@ -1104,10 +1107,10 @@ ildouble: 3
 ldouble: 3
 
 Function: Real part of "csin_towardzero":
-double: 2
-float: 2
-idouble: 2
-ifloat: 2
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
 ildouble: 3
 ldouble: 3
 
@@ -1120,9 +1123,9 @@ ildouble: 3
 ldouble: 3
 
 Function: Real part of "csin_upward":
-double: 1
+double: 3
 float: 2
-idouble: 1
+idouble: 3
 ifloat: 2
 ildouble: 3
 ldouble: 3
@@ -1160,10 +1163,10 @@ ildouble: 3
 ldouble: 3
 
 Function: Imaginary part of "csinh_downward":
-double: 2
-float: 2
-idouble: 2
-ifloat: 2
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
 ildouble: 3
 ldouble: 3
 
@@ -1176,10 +1179,10 @@ ildouble: 3
 ldouble: 3
 
 Function: Imaginary part of "csinh_towardzero":
-double: 2
-float: 2
-idouble: 2
-ifloat: 2
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
 ildouble: 3
 ldouble: 3
 
@@ -1192,9 +1195,9 @@ ildouble: 3
 ldouble: 3
 
 Function: Imaginary part of "csinh_upward":
-double: 1
+double: 3
 float: 2
-idouble: 1
+idouble: 3
 ifloat: 2
 ildouble: 3
 ldouble: 3
@@ -1207,7 +1210,9 @@ ldouble: 2
 
 Function: Imaginary part of "csqrt":
 double: 1
+float: 1
 idouble: 1
+ifloat: 1
 ildouble: 2
 ldouble: 2
 
@@ -1253,9 +1258,9 @@ ldouble: 5
 
 Function: Imaginary part of "csqrt_upward":
 double: 1
-float: 1
+float: 2
 idouble: 1
-ifloat: 1
+ifloat: 2
 ildouble: 4
 ldouble: 4
 
@@ -1429,9 +1434,9 @@ ldouble: 3
 
 Function: "erfc_downward":
 double: 2
-float: 2
+float: 3
 idouble: 2
-ifloat: 2
+ifloat: 3
 ildouble: 4
 ldouble: 4
 
@@ -1445,9 +1450,9 @@ ldouble: 4
 
 Function: "erfc_upward":
 double: 2
-float: 2
+float: 3
 idouble: 2
-ifloat: 2
+ifloat: 3
 ildouble: 5
 ldouble: 5
 
diff --git a/sysdeps/i386/i686/fpu/multiarch/e_expf-sse2.S b/sysdeps/i386/i686/fpu/multiarch/e_expf-sse2.S
index be3b145..2e23595 100644
--- a/sysdeps/i386/i686/fpu/multiarch/e_expf-sse2.S
+++ b/sysdeps/i386/i686/fpu/multiarch/e_expf-sse2.S
@@ -113,11 +113,12 @@ ENTRY(__ieee754_expf_sse2)
 	mulsd	MO2(DP_T,%eax,8), %xmm0	/* DP P(y)*T[j] */
 	addsd	MO2(DP_T,%eax,8), %xmm0	/* DP T[j]*(P(y)+1) */
 	mulsd	%xmm1, %xmm0		/* DP result=2^n*(T[j]*(P(y)+1)) */
+	cvtsd2ss	%xmm0, %xmm1
 
-	lea	-8(%esp), %esp		/* Borrow 8 bytes of stack frame */
-	movsd	%xmm0, 0(%esp)		/* Move result from sse... */
-	fldl	0(%esp)			/* ...to FPU. */
-	lea	8(%esp), %esp		/* Return back 8 bytes of stack frame */
+	lea	-4(%esp), %esp		/* Borrow 4 bytes of stack frame */
+	movss	%xmm1, 0(%esp)		/* Move result from sse... */
+	flds	0(%esp)			/* ...to FPU. */
+	lea	4(%esp), %esp		/* Return back 4 bytes of stack frame */
 	ret
 
 	.p2align	4