fix regression in x86_64 math asm with old binutils

the implicit-operand form of fucomip is rejected by binutils 2.19 and
perhaps other versions still in use. writing both operands explicitly
fixes the issue. there is no change to the resulting output.

commit a732e80d33b4fd6f510f7cec4f5573ef5d89bc4e was the source of this
regression.

src/math/x32/exp2l.s

@@ -6,7 +6,7 @@ expm1l:
 	fmulp
 	movl $0xc2820000,-4(%esp)
 	flds -4(%esp)
-	fucomip %st(1)
+	fucomip %st(1),%st
 	fld1
 	jb 1f
 		# x*log2e <= -65, return -1 without underflow
@@ -15,7 +15,7 @@ expm1l:
 	ret
 1:	fld %st(1)
 	fabs
-	fucomip %st(1)
+	fucomip %st(1),%st
 	fstp %st(0)
 	ja 1f
 	f2xm1
@@ -48,7 +48,7 @@ exp2l:
 	fld %st(1)
 	fsub %st(1)
 	faddp
-	fucomip %st(1)
+	fucomip %st(1),%st
 	je 2f             # x - 0x1p63 + 0x1p63 == x
 	movl $1,(%esp)
 	flds (%esp)       # 0x1p-149

src/math/x86_64/exp2l.s

@@ -6,7 +6,7 @@ expm1l:
 	fmulp
 	movl $0xc2820000,-4(%rsp)
 	flds -4(%rsp)
-	fucomip %st(1)
+	fucomip %st(1),%st
 	fld1
 	jb 1f
 		# x*log2e <= -65, return -1 without underflow
@@ -15,7 +15,7 @@ expm1l:
 	ret
 1:	fld %st(1)
 	fabs
-	fucomip %st(1)
+	fucomip %st(1),%st
 	fstp %st(0)
 	ja 1f
 	f2xm1
@@ -48,7 +48,7 @@ exp2l:
 	fld %st(1)
 	fsub %st(1)
 	faddp
-	fucomip %st(1)
+	fucomip %st(1),%st
 	je 2f             # x - 0x1p63 + 0x1p63 == x
 	movl $1,(%rsp)
 	flds (%rsp)       # 0x1p-149