Fix i386 exp10 missing underflows (bug 18966) [committed]

[Joseph Myers <joseph at codesourcery dot com>]

On i386, the double version of exp10 can miss underflow exceptions if
the result is in the subnormal range for double but the last 11 bits
of the 64-bit extended-precision mantissa happen to be zero.  This
patch forces the exception in a similar way to previous fixes.

As with the exp2 and exp fixes, the exp10f changes may in fact not be
needed to ensure underflow exceptions, but are included for
consistency and to fix the exp10 part of bug 18875 by ensuring that
excess range and precision is removed from underflowing return values.

Tested for x86_64 and x86.  Committed.

(auto-libm-test-out diffs omitted below.)

2015-09-15  Joseph Myers  <joseph@codesourcery.com>

	[BZ #18875]
	[BZ #18966]
	* sysdeps/i386/fpu/e_exp10.S (dbl_min): New object.
	(MO): New macro.
	(__ieee754_exp10): For small results, force underflow exception
	and remove excess range and precision from return value.
	* sysdeps/i386/fpu/e_exp10f.S (flt_min): New object.
	(MO): New macro.
	(__ieee754_exp10f): For small results, force underflow exception
	and remove excess range and precision from return value.
	* math/auto-libm-test-in: Add more tests of exp10.
	* math/auto-libm-test-out: Regenerated.

diff --git a/math/auto-libm-test-in b/math/auto-libm-test-in
index 9da01fb..4a8017c 100644
--- a/math/auto-libm-test-in
+++ b/math/auto-libm-test-in
@@ -1458,6 +1458,9 @@ exp10 -max
 exp10 0.75
 # GCC bug 59666: results on directed rounding may be incorrect.
 exp10 0x1.348e45573a1dd72cp+8 xfail-rounding:ldbl-128ibm
+exp10 -0x1.33aa03p+8
+exp10 -0x1.33ad17p+8
+exp10 -0x1.33afcap+8
 exp10 0x1p-10
 exp10 -0x1p-10
 exp10 0x1p-20
diff --git a/sysdeps/i386/fpu/e_exp10.S b/sysdeps/i386/fpu/e_exp10.S
index 1e32b07..fa54732 100644
--- a/sysdeps/i386/fpu/e_exp10.S
+++ b/sysdeps/i386/fpu/e_exp10.S
@@ -4,8 +4,25 @@
 
 #include <machine/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)
+
+#ifdef PIC
+# define MO(op) op##@GOTOFF(%ecx)
+#else
+# define MO(op) op
+#endif
+
+	.text
 /* 10^x = 2^(x * log2(10)) */
 ENTRY(__ieee754_exp10)
+#ifdef  PIC
+	LOAD_PIC_REG (cx)
+#endif
 	fldl	4(%esp)
 /* I added the following ugly construct because exp(+-Inf) resulted
    in NaN.  The ugliness results from the bright minds at Intel.
@@ -28,7 +45,22 @@ ENTRY(__ieee754_exp10)
 	faddp				/* 2^(fract(x * log2(10))) */
 	fscale				/* e^x */
 	fstp	%st(1)
-	ret
+	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
 
 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 6144964..a84b2ae 100644
--- a/sysdeps/i386/fpu/e_exp10f.S
+++ b/sysdeps/i386/fpu/e_exp10f.S
@@ -4,8 +4,25 @@
 
 #include <machine/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)
+
+#ifdef PIC
+# define MO(op) op##@GOTOFF(%ecx)
+#else
+# define MO(op) op
+#endif
+
+	.text
 /* 10^x = 2^(x * log2(10)) */
 ENTRY(__ieee754_exp10f)
+#ifdef  PIC
+	LOAD_PIC_REG (cx)
+#endif
 	flds	4(%esp)
 /* I added the following ugly construct because exp(+-Inf) resulted
    in NaN.  The ugliness results from the bright minds at Intel.
@@ -28,7 +45,22 @@ ENTRY(__ieee754_exp10f)
 	faddp				/* 2^(fract(x * log2(10))) */
 	fscale				/* e^x */
 	fstp	%st(1)
-	ret
+	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
 
 1:	testl	$0x200, %eax		/* Test sign.  */
 	jz	2f			/* If positive, jump.  */

-- 
Joseph S. Myers
joseph@codesourcery.com