diff --git a/ChangeLog b/ChangeLog
index 2ec26e19d..052afa672 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,19 @@
+2012-12-29 Paulo Andrade <pcpa@gnu.org>
+
+	* check/float.ok, check/float.tst: New test cases implementing
+	extensive validation of float comparison and branch code
+	generation as well as integer conversion, involving NaN and
+	[+-]Inf.
+
+	* lib/jit_arm-swf.c, lib/jit_x86-sse.c, lib/jit_x86-x87.c:
+	Correct bugs found by new float test case.
+
+	* lib/jit_x86.c: Correct cut&paste error added in commit to
+	convert jit_arg* return value to a jit_node_t*,	that would
+	cause it to not properly handle double arguments in ix86.
+
+	* check/Makefile.am: Update for the new test case.
+
 2012-12-28 Paulo Andrade <pcpa@gnu.org>
 
 	* check/lightning.c, include/lightning.h, lib/jit_arm.c,
diff --git a/TODO b/TODO
index 8cb265714..22d33b42d 100644
--- a/TODO
+++ b/TODO
@@ -9,3 +9,7 @@
 
 	* Test and correct the ppc and mips ports, after the import and
 	adaptation of the code to lightning.
+
+	* Add new float/double comparison and branch codes:
+	unne:	proper inverse of eq	(nan/nan == 0)
+	eqeq:	proper inverse of ltgt	(nan/nan == 1)
diff --git a/check/Makefile.am b/check/Makefile.am
index c1eaf6a69..70bb439e1 100644
--- a/check/Makefile.am
+++ b/check/Makefile.am
@@ -64,6 +64,7 @@ EXTRA_DIST =				\
 	clobber.tst	clobber.ok	\
 	carry.tst	carry.ok	\
 	call.tst	call.ok		\
+	float.tst	float.ok	\
 	check.sh			\
 	check.x87.sh			\
 	check.arm.sh	check.swf.sh	\
@@ -84,7 +85,8 @@ base_TESTS =				\
 	alu_com alu_neg			\
 	fop_abs fop_sqrt		\
 	varargs stack			\
-	clobber carry call
+	clobber carry call		\
+	float
 
 $(base_TESTS):	check.sh
 	$(LN_S) $(srcdir)/check.sh $@
diff --git a/check/float.ok b/check/float.ok
new file mode 100644
index 000000000..9766475a4
--- /dev/null
+++ b/check/float.ok
@@ -0,0 +1 @@
+ok
diff --git a/check/float.tst b/check/float.tst
new file mode 100644
index 000000000..de5886880
--- /dev/null
+++ b/check/float.tst
@@ -0,0 +1,338 @@
+
+.data	4
+ok:
+.c	"ok"
+
+.	$($NaN  =		 0.0 / 0.0)
+.	$($pInf =		 1.0 / 0.0)
+.	$($nInf =		-1.0 / 0.0)
+#if __WORDSIZE == 32
+#  define x7f			0x7fffffff
+#  define x80			0x80000000
+#else
+#  define x7f			0x7fffffffffffffff
+#  define x80			0x8000000000000000
+#endif
+
+#if __mips__
+#  define wnan			x7f
+#elif __arm__
+#  define wnan			0
+#else
+#  define wnan			x80
+#endif
+#if __mips__ || __arm__
+#  define wpinf			x7f
+#else
+#  define wpinf			x80
+#endif
+#define wninf			x80
+
+/* ensure result is correct and 0 or 1 in the result register */
+#define xtcmp(l, t, op, r0, f0, f1, li, ri)		\
+	movi##t %f0 li					\
+	movi##t %f1 ri					\
+	op##r##t %r0 %f0 %f1				\
+	bnei T##op##r##t##r0##f0##f1##l %r0 0		\
+	calli @abort					\
+T##op##r##t##r0##f0##f1##l:				\
+	movi##t %f0 li					\
+	movi##t %f1 ri					\
+	b##op##r##t bT##op##r##t##r0##f0##f1##l %f0 %f1	\
+	calli @abort					\
+bT##op##r##t##r0##f0##f1##l:				\
+	movi##t %f1 li					\
+	op##i##t %r0 %f0 ri				\
+	bnei T##op##i##t##r0##f0##f1##l %r0 0		\
+	calli @abort					\
+T##op##i##t##r0##f0##f1##l:				\
+	movi##t %f1 li					\
+	b##op##i##t bT##op##i##t##r0##f0##f1##l %f0 ri	\
+	calli @abort					\
+bT##op##i##t##r0##f0##f1##l:				\
+	movi##t %f0 li					\
+	movi##t %f1 ri					\
+	op##r##t %r0 %f0 %f1				\
+	beqi F##op##r##t##r0##f0##f1##l %r0 1		\
+	calli @abort					\
+F##op##r##t##r0##f0##f1##l:				\
+	movi##t %f1 li					\
+	op##i##t %r0 %f0 ri				\
+	beqi F##op##i##t##r0##f0##f1##l %r0 1		\
+	calli @abort					\
+F##op##i##t##r0##f0##f1##l:
+#define tcmp1(l, t, op, r0, li, ri)			\
+	xtcmp(l, t, op, r0, f0, f1, li, ri)		\
+	xtcmp(l, t, op, r0, f1, f2, li, ri)		\
+	xtcmp(l, t, op, r0, f2, f3, li, ri)		\
+	xtcmp(l, t, op, r0, f3, f4, li, ri)		\
+	xtcmp(l, t, op, r0, f4, f5, li, ri)
+#define tcmp0(l, t, op, li, ri)				\
+	tcmp1(l, t, op, v0, li, ri)			\
+	tcmp1(l, t, op, v1, li, ri)			\
+	tcmp1(l, t, op, v2, li, ri)			\
+	tcmp1(l, t, op, r0, li, ri)			\
+	tcmp1(l, t, op, r1, li, ri)			\
+	tcmp1(l, t, op, r2, li, ri)
+#define tcmp(l, op, li, ri)				\
+	tcmp0(l, _f, op, li, ri)			\
+	tcmp0(l, _d, op, li, ri)
+
+#define xfcmp(l, t, op, r0, f0, f1, li, ri)		\
+	movi##t %f0 li					\
+	movi##t %f1 ri					\
+	op##r##t %r0 %f0 %f1				\
+	beqi T##op##r##t##r0##f0##f1##l %r0 0		\
+	calli @abort					\
+T##op##r##t##r0##f0##f1##l:				\
+	movi##t %f1 li					\
+	op##i##t %r0 %f0 ri				\
+	beqi T##op##i##t##r0##f0##f1##l %r0 0		\
+	calli @abort					\
+T##op##i##t##r0##f0##f1##l:				\
+	movi##t %f0 li					\
+	movi##t %f1 ri					\
+	op##r##t %r0 %f0 %f1				\
+	bnei F##op##r##t##r0##f0##f1##l %r0 1		\
+	calli @abort					\
+F##op##r##t##r0##f0##f1##l:				\
+	movi##t %f1 li					\
+	op##i##t %r0 %f0 ri				\
+	bnei F##op##i##t##r0##f0##f1##l %r0 1		\
+	calli @abort					\
+F##op##i##t##r0##f0##f1##l:
+#define fcmp1(l, t, op, r0, li, ri)			\
+	xfcmp(l, t, op, r0, f0, f1, li, ri)		\
+	xfcmp(l, t, op, r0, f1, f2, li, ri)		\
+	xfcmp(l, t, op, r0, f2, f3, li, ri)		\
+	xfcmp(l, t, op, r0, f3, f4, li, ri)		\
+	xfcmp(l, t, op, r0, f4, f5, li, ri)
+#define fcmp0(l, t, op, li, ri)				\
+	fcmp1(l, t, op, v0, li, ri)			\
+	fcmp1(l, t, op, v1, li, ri)			\
+	fcmp1(l, t, op, v2, li, ri)			\
+	fcmp1(l, t, op, r0, li, ri)			\
+	fcmp1(l, t, op, r1, li, ri)			\
+	fcmp1(l, t, op, r2, li, ri)
+#define fcmp(l, op, li, ri)				\
+	fcmp0(l, _f, op, li, ri)			\
+	fcmp0(l, _d, op, li, ri)
+
+#define xf2w(l, f, r0, f0, iv, fv)			\
+	movi##f %f0 fv					\
+	truncr##f %r0 %f0				\
+	beqi W##f##r0##f0##l %r0 iv			\
+	calli @abort					\
+W##f##r0##f0##l:
+#define f2w1(l, t, r0, iv, fv)				\
+	xf2w(l, t, r0, f0, iv, fv)			\
+	xf2w(l, t, r0, f1, iv, fv)			\
+	xf2w(l, t, r0, f2, iv, fv)			\
+	xf2w(l, t, r0, f3, iv, fv)			\
+	xf2w(l, t, r0, f4, iv, fv)			\
+	xf2w(l, t, r0, f5, iv, fv)
+#define f2w0(l, t, iv, fv)				\
+	f2w1(l, t, v0, iv, fv)				\
+	f2w1(l, t, v1, iv, fv)				\
+	f2w1(l, t, v2, iv, fv)				\
+	f2w1(l, t, r0, iv, fv)				\
+	f2w1(l, t, r1, iv, fv)				\
+	f2w1(l, t, r2, iv, fv)
+#define f2w(l, iv, fv)					\
+	f2w0(l, _f, iv, fv)				\
+	f2w0(l, _d, iv, fv)
+
+.code
+	prolog
+
+	tcmp(__LINE__, lt, 0, 1)
+	tcmp(__LINE__, lt, $nInf, $pInf)
+	tcmp(__LINE__, lt, $nInf, 0)
+	tcmp(__LINE__, lt, 0, $pInf)
+	fcmp(__LINE__, lt, $NaN, 0)
+	fcmp(__LINE__, lt, $NaN, $NaN)
+	fcmp(__LINE__, lt, $nInf, $NaN)
+	fcmp(__LINE__, lt, 1, 0)
+	fcmp(__LINE__, lt, 0, 0)
+	fcmp(__LINE__, lt, $pInf, $nInf)
+	fcmp(__LINE__, lt, 0, $nInf)
+	fcmp(__LINE__, lt, 0, $NaN)
+
+	tcmp(__LINE__, le, 0, 1)
+	tcmp(__LINE__, le, 0, 0)
+	tcmp(__LINE__, le, 1, 1)
+	tcmp(__LINE__, le, $nInf, $pInf)
+	tcmp(__LINE__, le, $nInf, 0)
+	tcmp(__LINE__, le, 0, $pInf)
+	fcmp(__LINE__, le, $NaN, 0)
+	fcmp(__LINE__, le, $NaN, $NaN)
+	fcmp(__LINE__, le, $nInf, $NaN)
+	fcmp(__LINE__, le, 1, 0)
+	fcmp(__LINE__, le, $pInf, $nInf)
+	fcmp(__LINE__, le, 0, $nInf)
+	fcmp(__LINE__, le, 0, $NaN)
+
+	tcmp(__LINE__, eq, 0, 0)
+	tcmp(__LINE__, eq, 1, 1)
+	fcmp(__LINE__, eq, $NaN, 0)
+	fcmp(__LINE__, eq, $NaN, $NaN)
+	fcmp(__LINE__, eq, $nInf, $NaN)
+	fcmp(__LINE__, eq, 0, 1)
+	fcmp(__LINE__, eq, 1, 0)
+	fcmp(__LINE__, eq, $pInf, $nInf)
+	fcmp(__LINE__, eq, 0, $nInf)
+	fcmp(__LINE__, eq, 0, $NaN)
+
+	tcmp(__LINE__, ge, 1, 0)
+	tcmp(__LINE__, ge, 0, 0)
+	tcmp(__LINE__, ge, 1, 1)
+	tcmp(__LINE__, ge, $pInf, $nInf)
+	tcmp(__LINE__, ge, 0, $nInf)
+	tcmp(__LINE__, ge, $pInf, 0)
+	fcmp(__LINE__, ge, $NaN, 0)
+	fcmp(__LINE__, ge, $NaN, $NaN)
+	fcmp(__LINE__, ge, $nInf, $NaN)
+	fcmp(__LINE__, ge, 0, 1)
+	fcmp(__LINE__, ge, $nInf, $pInf)
+	fcmp(__LINE__, ge, $nInf, 0)
+	fcmp(__LINE__, ge, 0, $NaN)
+
+	tcmp(__LINE__, gt, 1, 0)
+	tcmp(__LINE__, gt, $pInf, $nInf)
+	tcmp(__LINE__, gt, 0, $nInf)
+	tcmp(__LINE__, gt, $pInf, 0)
+	fcmp(__LINE__, gt, $NaN, 0)
+	fcmp(__LINE__, gt, $NaN, $NaN)
+	fcmp(__LINE__, gt, $nInf, $NaN)
+	fcmp(__LINE__, gt, 0, 1)
+	fcmp(__LINE__, gt, 0, 0)
+	fcmp(__LINE__, gt, $nInf, $pInf)
+	fcmp(__LINE__, gt, $nInf, 0)
+	fcmp(__LINE__, gt, 0, $NaN)
+
+	tcmp(__LINE__, ne, 0, 1)
+	tcmp(__LINE__, ne, 1, 0)
+	tcmp(__LINE__, ne, $NaN, $NaN)
+	tcmp(__LINE__, ne, $nInf, $pInf)
+	tcmp(__LINE__, ne, $NaN, 0)
+	tcmp(__LINE__, ne, $nInf, $NaN)
+	tcmp(__LINE__, ne, $pInf, $nInf)
+	tcmp(__LINE__, ne, 0, $nInf)
+	tcmp(__LINE__, ne, 0, $NaN)
+	fcmp(__LINE__, ne, 0, 0)
+	fcmp(__LINE__, ne, 1, 1)
+
+	tcmp(__LINE__, unlt, 0, 1)
+	tcmp(__LINE__, unlt, $nInf, $pInf)
+	tcmp(__LINE__, unlt, $nInf, 0)
+	tcmp(__LINE__, unlt, 0, $pInf)
+	tcmp(__LINE__, unlt, $NaN, 0)
+	tcmp(__LINE__, unlt, $NaN, $NaN)
+	tcmp(__LINE__, unlt, $nInf, $NaN)
+	tcmp(__LINE__, unlt, 0, $NaN)
+	fcmp(__LINE__, unlt, 1, 0)
+	fcmp(__LINE__, unlt, 0, 0)
+	fcmp(__LINE__, unlt, $pInf, $nInf)
+	fcmp(__LINE__, unlt, 0, $nInf)
+
+	tcmp(__LINE__, unle, 0, 1)
+	tcmp(__LINE__, unle, 0, 0)
+	tcmp(__LINE__, unle, 1, 1)
+	tcmp(__LINE__, unle, $nInf, $pInf)
+	tcmp(__LINE__, unle, $nInf, 0)
+	tcmp(__LINE__, unle, 0, $pInf)
+	tcmp(__LINE__, unle, $NaN, 0)
+	tcmp(__LINE__, unle, $NaN, $NaN)
+	tcmp(__LINE__, unle, $nInf, $NaN)
+	tcmp(__LINE__, unle, 0, $NaN)
+	fcmp(__LINE__, unle, 1, 0)
+	fcmp(__LINE__, unle, $pInf, $nInf)
+	fcmp(__LINE__, unle, 0, $nInf)
+
+	tcmp(__LINE__, uneq, 0, 0)
+	tcmp(__LINE__, uneq, 1, 1)
+	tcmp(__LINE__, uneq, $NaN, 0)
+	tcmp(__LINE__, uneq, $NaN, $NaN)
+	tcmp(__LINE__, uneq, $nInf, $NaN)
+	tcmp(__LINE__, uneq, 0, $NaN)
+	fcmp(__LINE__, uneq, 0, 1)
+	fcmp(__LINE__, uneq, 1, 0)
+	fcmp(__LINE__, uneq, $pInf, $nInf)
+	fcmp(__LINE__, uneq, 0, $nInf)
+
+	tcmp(__LINE__, unge, 1, 0)
+	tcmp(__LINE__, unge, 0, 0)
+	tcmp(__LINE__, unge, 1, 1)
+	tcmp(__LINE__, unge, $pInf, $nInf)
+	tcmp(__LINE__, unge, 0, $nInf)
+	tcmp(__LINE__, unge, $pInf, 0)
+	tcmp(__LINE__, unge, $NaN, 0)
+	tcmp(__LINE__, unge, $NaN, $NaN)
+	tcmp(__LINE__, unge, $nInf, $NaN)
+	tcmp(__LINE__, unge, 0, $NaN)
+	fcmp(__LINE__, unge, 0, 1)
+	fcmp(__LINE__, unge, $nInf, $pInf)
+	fcmp(__LINE__, unge, $nInf, 0)
+
+	tcmp(__LINE__, ungt, 1, 0)
+	tcmp(__LINE__, ungt, $pInf, $nInf)
+	tcmp(__LINE__, ungt, 0, $nInf)
+	tcmp(__LINE__, ungt, $pInf, 0)
+	tcmp(__LINE__, ungt, $NaN, 0)
+	tcmp(__LINE__, ungt, $NaN, $NaN)
+	tcmp(__LINE__, ungt, $nInf, $NaN)
+	tcmp(__LINE__, ungt, 0, $NaN)
+	fcmp(__LINE__, ungt, 0, 1)
+	fcmp(__LINE__, ungt, 0, 0)
+	fcmp(__LINE__, ungt, $nInf, $pInf)
+	fcmp(__LINE__, ungt, $nInf, 0)
+
+	tcmp(__LINE__, ltgt, 0, 1)
+	tcmp(__LINE__, ltgt, 1, 0)
+	tcmp(__LINE__, ltgt, $nInf, $pInf)
+	tcmp(__LINE__, ltgt, $pInf, $nInf)
+	tcmp(__LINE__, ltgt, 0, $nInf)
+	fcmp(__LINE__, ltgt, $NaN, $NaN)
+	fcmp(__LINE__, ltgt, $NaN, 0)
+	fcmp(__LINE__, ltgt, $nInf, $NaN)
+	fcmp(__LINE__, ltgt, 0, $NaN)
+	fcmp(__LINE__, ltgt, 0, 0)
+	fcmp(__LINE__, ltgt, 1, 1)
+
+	tcmp(__LINE__, ord, 0, 1)
+	tcmp(__LINE__, ord, 1, 0)
+	tcmp(__LINE__, ord, $nInf, $pInf)
+	tcmp(__LINE__, ord, $pInf, $nInf)
+	tcmp(__LINE__, ord, 0, $nInf)
+	tcmp(__LINE__, ord, 0, 0)
+	tcmp(__LINE__, ord, 1, 1)
+	fcmp(__LINE__, ord, $NaN, $NaN)
+	fcmp(__LINE__, ord, $NaN, 0)
+	fcmp(__LINE__, ord, $nInf, $NaN)
+	fcmp(__LINE__, ord, 0, $NaN)
+
+	tcmp(__LINE__, unord, $NaN, $NaN)
+	tcmp(__LINE__, unord, $NaN, 0)
+	tcmp(__LINE__, unord, $nInf, $NaN)
+	tcmp(__LINE__, unord, 0, $NaN)
+	fcmp(__LINE__, unord, 0, 1)
+	fcmp(__LINE__, unord, 1, 0)
+	fcmp(__LINE__, unord, $nInf, $pInf)
+	fcmp(__LINE__, unord, $pInf, $nInf)
+	fcmp(__LINE__, unord, 0, $nInf)
+	fcmp(__LINE__, unord, 0, 0)
+	fcmp(__LINE__, unord, 1, 1)
+
+	f2w(__LINE__, 0, 0)
+	f2w(__LINE__, 1, 1)
+	f2w(__LINE__, wninf, $nInf)
+	f2w(__LINE__, wpinf, $pInf)
+	f2w(__LINE__, wnan, $NaN)
+
+	prepare
+		pushargi ok
+		ellipsis
+	finishi @puts
+
+	ret
+	epilog
diff --git a/lib/jit_arm-swf.c b/lib/jit_arm-swf.c
index 1315693f7..89aefed35 100644
--- a/lib/jit_arm-swf.c
+++ b/lib/jit_arm-swf.c
@@ -175,7 +175,7 @@ static void _swf_negr_d(jit_state_t*,jit_int32_t,jit_int32_t);
 #  define swf_ger_d(r0,r1,r2)		swf_idd(__aeabi_dcmpge,r0,r1,r2)
 #  define swf_gei_d(r0,r1,i0)		swf_idd_(__aeabi_dcmpge,r0,r1,i0)
 #  define swf_gtr_f(r0,r1,r2)		swf_iff(__aeabi_fcmpgt,r0,r1,r2)
-#  define swf_gti_f(r0,r1,i0)		swf_iff(__aeabi_fcmpgt,r0,r1,i0)
+#  define swf_gti_f(r0,r1,i0)		swf_iff_(__aeabi_fcmpgt,r0,r1,i0)
 #  define swf_gtr_d(r0,r1,r2)		swf_idd(__aeabi_dcmpgt,r0,r1,r2)
 #  define swf_gti_d(r0,r1,i0)		swf_idd_(__aeabi_dcmpgt,r0,r1,i0)
 #  define swf_ner_f(r0,r1,r2)		_swf_ner_f(_jit,r0,r1,r2)
@@ -688,7 +688,7 @@ _swf_iunff(jit_state_t *_jit, int (*i0)(float, float),
 	movr(_R1_REGNO, r2);
     swf_call(__aeabi_fcmpun, fcmpun, _R2_REGNO);
     if (jit_thumb_p()) {
-	T1_CMPI(_R0, 0);
+	T1_CMPI(_R0_REGNO, 0);
 	IT(ARM_CC_NE);
 	if (r0 < 8)
 	    T1_MOVI(r0, 1);
@@ -698,7 +698,7 @@ _swf_iunff(jit_state_t *_jit, int (*i0)(float, float),
 	T2_CC_B(ARM_CC_NE, 0);
     }
     else {
-	CMPI(_R0, 0);
+	CMPI(_R0_REGNO, 0);
 	CC_MOVI(ARM_CC_NE, r0, 1);
 	instr = _jit->pc.w;
 	CC_B(ARM_CC_NE, 0);
@@ -749,7 +749,7 @@ _swf_iundd(jit_state_t *_jit, int (*i0)(double, double),
     }
     swf_call_with_get_reg(__aeabi_dcmpun, dcmpun);
     if (jit_thumb_p()) {
-	T1_CMPI(_R0, 0);
+	T1_CMPI(_R0_REGNO, 0);
 	IT(ARM_CC_NE);
 	if (r0 < 8)
 	    T1_MOVI(r0, 1);
@@ -759,7 +759,7 @@ _swf_iundd(jit_state_t *_jit, int (*i0)(double, double),
 	T2_CC_B(ARM_CC_NE, 0);
     }
     else {
-	CMPI(_R0, 0);
+	CMPI(_R0_REGNO, 0);
 	CC_MOVI(ARM_CC_NE, r0, 1);
 	instr = _jit->pc.w;
 	CC_B(ARM_CC_NE, 0);
@@ -812,7 +812,7 @@ _swf_iunff_(jit_state_t *_jit, int (*i0)(float, float),
     movi(_R1_REGNO, data.i);
     swf_call(__aeabi_fcmpun, fcmpun, _R2_REGNO);
     if (jit_thumb_p()) {
-	T1_CMPI(_R0, 0);
+	T1_CMPI(__R0_REGNO, 0);
 	IT(ARM_CC_NE);
 	if (r0 < 8)
 	    T1_MOVI(r0, 1);
@@ -822,7 +822,7 @@ _swf_iunff_(jit_state_t *_jit, int (*i0)(float, float),
 	T2_CC_B(ARM_CC_NE, 0);
     }
     else {
-	CMPI(_R0, 0);
+	CMPI(__R0_REGNO, 0);
 	CC_MOVI(ARM_CC_NE, r0, 1);
 	instr = _jit->pc.w;
 	CC_B(ARM_CC_NE, 0);
@@ -865,7 +865,7 @@ _swf_iundd_(jit_state_t *_jit, int (*i0)(double, double),
     movi(_R3_REGNO, data.i[1]);
     swf_call_with_get_reg(__aeabi_dcmpun, dcmpun);
     if (jit_thumb_p()) {
-	T1_CMPI(_R0, 0);
+	T1_CMPI(__R0_REGNO, 0);
 	IT(ARM_CC_NE);
 	if (r0 < 8)
 	    T1_MOVI(r0, 1);
@@ -875,7 +875,7 @@ _swf_iundd_(jit_state_t *_jit, int (*i0)(double, double),
 	T2_CC_B(ARM_CC_NE, 0);
     }
     else {
-	CMPI(_R0, 0);
+	CMPI(__R0_REGNO, 0);
 	CC_MOVI(ARM_CC_NE, r0, 1);
 	instr = _jit->pc.w;
 	CC_B(ARM_CC_NE, 0);
diff --git a/lib/jit_x86-sse.c b/lib/jit_x86-sse.c
index d67ab9dc8..afc7c9a7a 100644
--- a/lib/jit_x86-sse.c
+++ b/lib/jit_x86-sse.c
@@ -191,7 +191,7 @@ static void _sse_ungei_f(jit_state_t*,jit_int32_t,jit_int32_t,jit_float32_t*);
 static void _sse_unger_f(jit_state_t*, jit_int32_t, jit_int32_t, jit_int32_t);
 #  define sse_ungti_f(r0, r1, i0)	_sse_ungti_f(_jit, r0, r1, i0)
 static void _sse_ungti_f(jit_state_t*,jit_int32_t,jit_int32_t,jit_float32_t*);
-#  define sse_ungtr_f(r0, r1, r2)	ssecmpf(X86_CC_NAE, r0, r2, r1)
+#  define sse_ungtr_f(r0, r1, r2)	ssecmpf(X86_CC_NAE, r0, r1, r2)
 #  define sse_ltgti_f(r0, r1, i0)	_sse_ltgti_f(_jit, r0, r1, i0)
 static void _sse_ltgti_f(jit_state_t*,jit_int32_t,jit_int32_t,jit_float32_t*);
 #  define sse_ltgtr_f(r0, r1, r2)	_sse_ltgtr_f(_jit, r0, r1, r2)
@@ -323,7 +323,7 @@ static void _sse_uneqi_d(jit_state_t*,jit_int32_t,jit_int32_t,jit_float64_t*);
 static void _sse_unger_d(jit_state_t*, jit_int32_t, jit_int32_t, jit_int32_t);
 #  define sse_ungei_d(r0, r1, i0)	_sse_ungei_d(_jit, r0, r1, i0)
 static void _sse_ungei_d(jit_state_t*,jit_int32_t,jit_int32_t,jit_float64_t*);
-#  define sse_ungtr_d(r0, r1, r2)	ssecmpd(X86_CC_NAE, r0, r2, r1)
+#  define sse_ungtr_d(r0, r1, r2)	ssecmpd(X86_CC_NAE, r0, r1, r2)
 #  define sse_ungti_d(r0, r1, i0)	_sse_ungti_d(_jit, r0, r1, i0)
 static void _sse_ungti_d(jit_state_t*,jit_int32_t,jit_int32_t,jit_float64_t*);
 #  define sse_ltgtr_d(r0, r1, r2)	_sse_ltgtr_d(_jit, r0, r1, r2)
diff --git a/lib/jit_x86-x87.c b/lib/jit_x86-x87.c
index 579c66136..7df115496 100644
--- a/lib/jit_x86-x87.c
+++ b/lib/jit_x86-x87.c
@@ -166,7 +166,7 @@ static void _x87_unlti_f(jit_state_t*,jit_int32_t,jit_int32_t,jit_float32_t*);
 #  define x87_unler_f(r0, r1, r2)	x87cmp(X86_CC_NA, r0, r1, r2)
 #  define x87_unlei_f(r0, r1, i0)	_x87_unlei_f(_jit, r0, r1, i0)
 static void _x87_unlei_f(jit_state_t*,jit_int32_t,jit_int32_t,jit_float32_t*);
-#  define x87_uneqr_f(r0, r1, r2)	x87cmp2(X86_CC_E, r0, r1, r1)
+#  define x87_uneqr_f(r0, r1, r2)	x87cmp2(X86_CC_E, r0, r1, r2)
 #  define x87_uneqi_f(r0, r1, i0)	_x87_uneqi_f(_jit, r0, r1, i0)
 static void _x87_uneqi_f(jit_state_t*,jit_int32_t,jit_int32_t,jit_float32_t*);
 #  define x87_unger_f(r0, r1, r2)	x87cmp(X86_CC_NA, r0, r2, r1)
@@ -210,7 +210,7 @@ static void _x87_unlti_d(jit_state_t*,jit_int32_t,jit_int32_t,jit_float64_t*);
 #  define x87_unler_d(r0, r1, r2)	x87cmp(X86_CC_NA, r0, r1, r2)
 #  define x87_unlei_d(r0, r1, i0)	_x87_unlei_d(_jit, r0, r1, i0)
 static void _x87_unlei_d(jit_state_t*,jit_int32_t,jit_int32_t,jit_float64_t*);
-#  define x87_uneqr_d(r0, r1, r2)	x87cmp2(X86_CC_E, r0, r1, r1)
+#  define x87_uneqr_d(r0, r1, r2)	x87cmp2(X86_CC_E, r0, r1, r2)
 #  define x87_uneqi_d(r0, r1, i0)	_x87_uneqi_d(_jit, r0, r1, i0)
 static void _x87_uneqi_d(jit_state_t*,jit_int32_t,jit_int32_t,jit_float64_t*);
 #  define x87_unger_d(r0, r1, r2)	x87cmp(X86_CC_NA, r0, r2, r1)
diff --git a/lib/jit_x86.c b/lib/jit_x86.c
index 7104edbfa..6b1480824 100644
--- a/lib/jit_x86.c
+++ b/lib/jit_x86.c
@@ -480,11 +480,7 @@ _jit_arg_d(jit_state_t *_jit)
 #endif
     {
 	offset = _jit->function->self.size;
-#if __WORDSIZE == 32
-	_jit->function->self.size += sizeof(jit_float32_t);
-#else
 	_jit->function->self.size += sizeof(jit_float64_t);
-#endif
     }
     return (jit_new_node_w(jit_code_arg_d, offset));
 }