Merge lp:~schnetter/pocl/main into lp:~pocl/pocl/trunk

[Erik Schnetter]

[lp:~schnetter/pocl/main]

77. By Erik Schnetter on 2011-11-05

Add missing lib/kernel/tce that got lost in the merge

[Carlos Sánchez de La Lama (csanchezdll)]

review: Approve

=== modified file 'include/_kernel.h'
--- include/_kernel.h	2011-11-03 18:58:00 +0000
+++ include/_kernel.h	2011-11-05 00:27:24 +0000
@@ -48,9 +48,6 @@
 #define cl_khr_fp64
 #define cl_khr_int64
 
-/* Shouldn't the compiler define this? */
-//#define __SSE4_1__
-
 #endif
 
 /* Enable double precision. This should really only be done when
@@ -928,6 +925,7 @@
 #define M_SQRT2_F    1.41421356237309504880168872421f
 #define M_SQRT1_2_F  0.707106781186547524400844362105f
 
+#ifdef cl_khr_fp64
 #define HUGE_VAL __builtin_huge_val()
 
 #define DBL_DIG        15
@@ -953,6 +951,7 @@
 #define M_2_SQRTPI 1.12837916709551257389615890312
 #define M_SQRT2    1.41421356237309504880168872421
 #define M_SQRT1_2  0.707106781186547524400844362105
+#endif
 
 
 
 /* Math Functions */
@@ -1344,6 +1343,29 @@
   int16 _cl_overloadable NAME(float16 );
 #endif
 #ifdef cl_khr_fp64
+#define _CL_DECLARE_FUNC_K_V(NAME)              \
+  int    _cl_overloadable NAME(float   );       \
+  int2   _cl_overloadable NAME(float2  );       \
+  int3   _cl_overloadable NAME(float3  );       \
+  int4   _cl_overloadable NAME(float4  );       \
+  int8   _cl_overloadable NAME(float8  );       \
+  int16  _cl_overloadable NAME(float16 );       \
+  int    _cl_overloadable NAME(double  );       \
+  long2  _cl_overloadable NAME(double2 );       \
+  long3  _cl_overloadable NAME(double3 );       \
+  long4  _cl_overloadable NAME(double4 );       \
+  long8  _cl_overloadable NAME(double8 );       \
+  long16 _cl_overloadable NAME(double16);
+#else
+#define _CL_DECLARE_FUNC_K_V(NAME)              \
+  int   _cl_overloadable NAME(float   );        \
+  int2  _cl_overloadable NAME(float2  );        \
+  int3  _cl_overloadable NAME(float3  );        \
+  int4  _cl_overloadable NAME(float4  );        \
+  int8  _cl_overloadable NAME(float8  );        \
+  int16 _cl_overloadable NAME(float16 );
+#endif
+#ifdef cl_khr_fp64
 #define _CL_DECLARE_FUNC_S_V(NAME)              \
   float  _cl_overloadable NAME(float   );       \
   float  _cl_overloadable NAME(float2  );       \
@@ -2242,7 +2264,7 @@
 _CL_DECLARE_FUNC_J_VV(isnormal)
 _CL_DECLARE_FUNC_J_VV(isordered)
 _CL_DECLARE_FUNC_J_VV(isunordered)
-_CL_DECLARE_FUNC_J_V(signbit)
+_CL_DECLARE_FUNC_K_V(signbit)
 _CL_DECLARE_FUNC_I_IG(any)
 _CL_DECLARE_FUNC_I_IG(all)
 _CL_DECLARE_FUNC_G_GGG(bitselect)
 
=== modified file 'lib/kernel/acospi.cl'
--- lib/kernel/acospi.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/acospi.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,8 @@
 
 #include "templates.h"
 
+#ifdef cl_khr_fp64
 DEFINE_EXPR_V_V(acospi, acos(a)/(stype)M_PI)
+#else
+DEFINE_EXPR_V_V(acospi, acos(a)/M_PI_F)
+#endif
 
=== modified file 'lib/kernel/asinpi.cl'
--- lib/kernel/asinpi.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/asinpi.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,8 @@
 
 #include "templates.h"
 
+#ifdef cl_khr_fp64
 DEFINE_EXPR_V_V(asinpi, asin(a)/(stype)M_PI)
+#else
+DEFINE_EXPR_V_V(asinpi, asin(a)/M_PI_F)
+#endif
 
=== modified file 'lib/kernel/atan2pi.cl'
--- lib/kernel/atan2pi.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/atan2pi.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,8 @@
 
 #include "templates.h"
 
+#ifdef cl_khr_fp64
 DEFINE_EXPR_V_VV(atan2pi, atan2(a, b)/(stype)M_PI)
+#else
+DEFINE_EXPR_V_VV(atan2pi, atan2(a, b)/M_PI_F)
+#endif
 
=== modified file 'lib/kernel/atanpi.cl'
--- lib/kernel/atanpi.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/atanpi.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,8 @@
 
 #include "templates.h"
 
+#ifdef cl_khr_fp64
 DEFINE_EXPR_V_V(atanpi, atan(a)/(stype)M_PI)
+#else
+DEFINE_EXPR_V_V(atanpi, atan(a)/M_PI_F)
+#endif
 
=== modified file 'lib/kernel/cospi.cl'
--- lib/kernel/cospi.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/cospi.cl	2011-11-05 00:27:24 +0000
@@ -23,5 +23,8 @@
 
 #include "templates.h"
 
-#undef cospi
+#ifdef cl_khr_fp64
 DEFINE_EXPR_V_V(cospi, cos((stype)M_PI*a))
+#else
+DEFINE_EXPR_V_V(cospi, cos(M_PI_F*a))
+#endif
 
=== modified file 'lib/kernel/degrees.cl'
--- lib/kernel/degrees.cl	2011-10-26 21:01:40 +0000
+++ lib/kernel/degrees.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,8 @@
 
 #include "templates.h"
 
-DEFINE_EXPR_V_V(degrees, (stype)(180.0 / M_PI) * a)
+#ifdef cl_khr_fp64
+DEFINE_EXPR_V_V(degrees, (stype)(180 / M_PI) * a)
+#else
+DEFINE_EXPR_V_V(degrees, (180 / M_PI_F) * a)
+#endif
 
=== modified file 'lib/kernel/exp10.cl'
--- lib/kernel/exp10.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/exp10.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,8 @@
 
 #include "templates.h"
 
+#ifdef cl_khr_fp64
 DEFINE_EXPR_V_V(exp10, exp((stype)M_LN10*a))
+#else
+DEFINE_EXPR_V_V(exp10, exp(M_LN10_F*a))
+#endif
 
=== modified file 'lib/kernel/fdim.cl'
--- lib/kernel/fdim.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/fdim.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,4 @@
 
 #include "templates.h"
 
-DEFINE_EXPR_V_VV(fdim, fmax(a-b, (stype)0.0))
+DEFINE_EXPR_V_VV(fdim, fmax(a-b, (stype)0))
 
=== modified file 'lib/kernel/fract.cl'
--- lib/kernel/fract.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/fract.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,9 @@
 
 #include "templates.h"
 
+
+#ifdef cl_khr_fp64
 DEFINE_EXPR_V_VPV(fract, fmin(a - floor(a), (vtype)(stype)(sizeof(stype)==4 ? 0x1.fffffep-1f : 0x1.fffffffffffffp-1)))
+#else
+DEFINE_EXPR_V_VPV(fract, fmin(a - floor(a), (vtype)(stype)0x1.fffffep-1f))
+#endif
 
=== modified file 'lib/kernel/radians.cl'
--- lib/kernel/radians.cl	2011-10-26 21:01:40 +0000
+++ lib/kernel/radians.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,8 @@
 
 #include "templates.h"
 
-DEFINE_EXPR_V_V(radians, (stype)(M_PI / 180.0) * a)
+#ifdef cl_khr_fp64
+DEFINE_EXPR_V_V(radians, (stype)(M_PI / 180) * a)
+#else
+DEFINE_EXPR_V_V(radians, (M_PI_F / 180) * a)
+#endif
 
=== modified file 'lib/kernel/rsqrt.cl'
--- lib/kernel/rsqrt.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/rsqrt.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,4 @@
 
 #include "templates.h"
 
-DEFINE_EXPR_V_V(rsqrt, (stype)1.0/sqrt(a))
+DEFINE_EXPR_V_V(rsqrt, (stype)1/sqrt(a))
 
=== modified file 'lib/kernel/signbit.cl'
--- lib/kernel/signbit.cl	2011-10-27 01:35:56 +0000
+++ lib/kernel/signbit.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,4 @@
 
 #include "templates.h"
 
-DEFINE_BUILTIN_J_V(signbit)
+DEFINE_BUILTIN_K_V(signbit)
 
=== modified file 'lib/kernel/sinpi.cl'
--- lib/kernel/sinpi.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/sinpi.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,8 @@
 
 #include "templates.h"
 
+#ifdef cl_khr_fp64
 DEFINE_EXPR_V_V(sinpi, sin((stype)M_PI*a))
+#else
+DEFINE_EXPR_V_V(sinpi, sin(M_PI_F*a))
+#endif
 
=== modified file 'lib/kernel/smoothstep.cl'
--- lib/kernel/smoothstep.cl	2011-10-27 00:18:42 +0000
+++ lib/kernel/smoothstep.cl	2011-11-05 00:27:24 +0000
@@ -25,7 +25,7 @@
 
 DEFINE_EXPR_V_VVV(smoothstep,
                   ({
-                    vtype t = clamp((c - a) / (b - a), (stype)0.0, (stype)1.0);
-                    t * t * mad((vtype)-2.0, t, (vtype)3.0);
+                    vtype t = clamp((c - a) / (b - a), (stype)0, (stype)1);
+                    t * t * mad((vtype)-2, t, (vtype)3);
                   }))
 DEFINE_EXPR_V_SSV(smoothstep, smoothstep((vtype)a, (vtype)b, c))
 
=== modified file 'lib/kernel/step.cl'
--- lib/kernel/step.cl	2011-10-27 00:18:42 +0000
+++ lib/kernel/step.cl	2011-11-05 00:27:24 +0000
@@ -27,8 +27,8 @@
 // DEFINE_EXPR_V_VV(step, b < a ? (vtype)0.0 : (vtype)1.0)
 DEFINE_EXPR_V_VV(step,
                  ({
-                   vtype zero = 0.0;
-                   vtype one  = 1.0;
+                   vtype zero = 0;
+                   vtype one  = 1;
                    jtype result = b < a ? *(jtype*)&zero : *(jtype*)&one;
                    *(vtype*)&result;
                  }))
 
=== modified file 'lib/kernel/tanpi.cl'
--- lib/kernel/tanpi.cl	2011-10-26 03:01:29 +0000
+++ lib/kernel/tanpi.cl	2011-11-05 00:27:24 +0000
@@ -23,4 +23,8 @@
 
 #include "templates.h"
 
+#ifdef cl_khr_fp64
 DEFINE_EXPR_V_V(tanpi, tan((stype)M_PI*a))
+#else
+DEFINE_EXPR_V_V(tanpi, tan(M_PI_F*a))
+#endif
 
=== added directory 'lib/kernel/tce'
=== removed directory 'lib/kernel/tce'
=== added file 'lib/kernel/tce/Makefile.am'
--- lib/kernel/tce/Makefile.am	1970-01-01 00:00:00 +0000
+++ lib/kernel/tce/Makefile.am	2011-11-05 00:27:24 +0000
@@ -0,0 +1,52 @@
+# Process this file with automake to produce Makefile.in (in this,
+# and all subdirectories).
+# Makefile.am for pocl/lib/kernel/tce. 
+#
+# The TCE (http://tce.cs.tut.fi) target.
+# 
+# Copyright (c) 2011 Universidad Rey Juan Carlos
+# 
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+# 
+# The above copyright notice and this permission notice shall be included in
+# all copies or substantial portions of the Software.
+# 
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+# THE SOFTWARE.
+
+targetpkglibdir = $(pkglibdir)/tce
+targetpkglib_LIBRARIES = libkernel.a
+
+vpath %.cl @srcdir@/..
+vpath %.c @srcdir@/..
+vpath %.ll @srcdir@/..
+
+include ../sources.mk
+
+libkernel_a_LIBADD = barrier.o
+EXTRA_DIST = barrier.ll
+
+RANLIB = `@LLVM_CONFIG@ --bindir`/llvm-ranlib
+AR = `@LLVM_CONFIG@ --bindir`/llvm-ar
+
+.cl.o:
+	$(CLANG) $(AM_CPPFLAGS) $(CLANGFLAGS) -ccc-host-triple tce-tut-llvm -c -emit-llvm -include $(top_srcdir)/include/_kernel.h -o $@ $<
+
+.c.o:
+	$(CLANG) $(AM_CPPFLAGS) $(CLANGFLAGS) -ccc-host-triple tce-tut-llvm -c -emit-llvm -include $(top_srcdir)/include/_kernel.h -o $@ $<
+
+barrier.o: barrier.ll
+	$(LLVM_AS) -o $@ $<
+
+$(libkernel_a_SOURCES:.c=.o):  $(top_srcdir)/include/_kernel.h ../templates.h
+$(libkernel_a_SOURCES:.cl=.o):  $(top_srcdir)/include/_kernel.h ../templates.h
 
=== removed file 'lib/kernel/tce/Makefile.am'
--- lib/kernel/tce/Makefile.am	2011-11-03 18:58:00 +0000
+++ lib/kernel/tce/Makefile.am	1970-01-01 00:00:00 +0000
@@ -1,52 +0,0 @@
-# Process this file with automake to produce Makefile.in (in this,
-# and all subdirectories).
-# Makefile.am for pocl/lib/kernel/tce. 
-#
-# The TCE (http://tce.cs.tut.fi) target.
-# 
-# Copyright (c) 2011 Universidad Rey Juan Carlos
-# 
-# Permission is hereby granted, free of charge, to any person obtaining a copy
-# of this software and associated documentation files (the "Software"), to deal
-# in the Software without restriction, including without limitation the rights
-# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-# copies of the Software, and to permit persons to whom the Software is
-# furnished to do so, subject to the following conditions:
-# 
-# The above copyright notice and this permission notice shall be included in
-# all copies or substantial portions of the Software.
-# 
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
-# THE SOFTWARE.
-
-targetpkglibdir = $(pkglibdir)/tce
-targetpkglib_LIBRARIES = libkernel.a
-
-vpath %.cl @srcdir@/..
-vpath %.c @srcdir@/..
-vpath %.ll @srcdir@/..
-
-include ../sources.mk
-
-libkernel_a_LIBADD = barrier.o
-EXTRA_DIST = barrier.ll
-
-RANLIB = `@LLVM_CONFIG@ --bindir`/llvm-ranlib
-AR = `@LLVM_CONFIG@ --bindir`/llvm-ar
-
-.cl.o:
-	$(CLANG) $(AM_CPPFLAGS) $(CLANGFLAGS) -ccc-host-triple tce-tut-llvm -c -emit-llvm -include $(top_srcdir)/include/_kernel.h -o $@ $<
-
-.c.o:
-	$(CLANG) $(AM_CPPFLAGS) $(CLANGFLAGS) -ccc-host-triple tce-tut-llvm -c -emit-llvm -include $(top_srcdir)/include/_kernel.h -o $@ $<
-
-barrier.o: barrier.ll
-	$(LLVM_AS) -o $@ $<
-
-$(libkernel_a_SOURCES:.c=.o):  $(top_srcdir)/include/_kernel.h ../templates.h
-$(libkernel_a_SOURCES:.cl=.o):  $(top_srcdir)/include/_kernel.h ../templates.h
 
=== modified file 'lib/kernel/templates.h'
--- lib/kernel/templates.h	2011-11-01 16:33:17 +0000
+++ lib/kernel/templates.h	2011-11-05 00:27:24 +0000
@@ -359,6 +359,48 @@
   IMPLEMENT_BUILTIN_J_V(NAME, int16, float16 , lo, hi)
 #endif
 
+#define IMPLEMENT_BUILTIN_K_V(NAME, JTYPE, VTYPE, LO, HI)       \
+  JTYPE __attribute__ ((overloadable))                          \
+  NAME(VTYPE a)                                                 \
+  {                                                             \
+    return (JTYPE)(NAME(a.LO), NAME(a.HI));                     \
+  }
+#ifdef cl_khr_fp64
+#define DEFINE_BUILTIN_K_V(NAME)                        \
+  int __attribute__ ((overloadable))                    \
+  NAME(float a)                                         \
+  {                                                     \
+    return __builtin_##NAME##f(a);                      \
+  }                                                     \
+  int __attribute__ ((overloadable))                    \
+  NAME(double a)                                        \
+  {                                                     \
+    return __builtin_##NAME(a);                         \
+  }                                                     \
+  IMPLEMENT_BUILTIN_K_V(NAME, int2  , float2  , lo, hi) \
+  IMPLEMENT_BUILTIN_K_V(NAME, int3  , float3  , lo, s2) \
+  IMPLEMENT_BUILTIN_K_V(NAME, int4  , float4  , lo, hi) \
+  IMPLEMENT_BUILTIN_K_V(NAME, int8  , float8  , lo, hi) \
+  IMPLEMENT_BUILTIN_K_V(NAME, int16 , float16 , lo, hi) \
+  IMPLEMENT_BUILTIN_K_V(NAME, long2 , double2 , lo, hi) \
+  IMPLEMENT_BUILTIN_K_V(NAME, long3 , double3 , lo, s2) \
+  IMPLEMENT_BUILTIN_K_V(NAME, long4 , double4 , lo, hi) \
+  IMPLEMENT_BUILTIN_K_V(NAME, long8 , double8 , lo, hi) \
+  IMPLEMENT_BUILTIN_K_V(NAME, long16, double16, lo, hi)
+#else
+#define DEFINE_BUILTIN_K_V(NAME)                        \
+  int __attribute__ ((overloadable))                    \
+  NAME(float a)                                         \
+  {                                                     \
+    return __builtin_##NAME##f(a);                      \
+  }                                                     \
+  IMPLEMENT_BUILTIN_K_V(NAME, int2 , float2  , lo, hi)  \
+  IMPLEMENT_BUILTIN_K_V(NAME, int3 , float3  , lo, s2)  \
+  IMPLEMENT_BUILTIN_K_V(NAME, int4 , float4  , lo, hi)  \
+  IMPLEMENT_BUILTIN_K_V(NAME, int8 , float8  , lo, hi)  \
+  IMPLEMENT_BUILTIN_K_V(NAME, int16, float16 , lo, hi)
+#endif
+
 #define IMPLEMENT_EXPR_V_V(NAME, EXPR, VTYPE, STYPE)    \
   VTYPE __attribute__ ((overloadable))                  \
   NAME(VTYPE a, VTYPE b)                                \
 
=== added file 'lib/kernel/x86_64/signbit.cl'
--- lib/kernel/x86_64/signbit.cl	1970-01-01 00:00:00 +0000
+++ lib/kernel/x86_64/signbit.cl	2011-11-05 00:27:24 +0000
@@ -0,0 +1,55 @@
+/* OpenCL built-in library: signbit()
+
+   Copyright (c) 2011 Universidad Rey Juan Carlos
+   
+   Permission is hereby granted, free of charge, to any person obtaining a copy
+   of this software and associated documentation files (the "Software"), to deal
+   in the Software without restriction, including without limitation the rights
+   to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+   copies of the Software, and to permit persons to whom the Software is
+   furnished to do so, subject to the following conditions:
+   
+   The above copyright notice and this permission notice shall be included in
+   all copies or substantial portions of the Software.
+   
+   THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+   IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+   FITNESS FOR A PARTICULAR PURPOSE AND NONORDEREDRINGEMENT. IN NO EVENT SHALL THE
+   AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+   LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+   OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+   THE SOFTWARE.
+*/
+
+#define IMPLEMENT_SIGNBIT_BUILTIN_FLOAT  __builtin_signbitf(a)
+#define IMPLEMENT_SIGNBIT_BUILTIN_DOUBLE __builtin_signbit(a)
+#define IMPLEMENT_SIGNBIT_DIRECT                \
+  ({                                            \
+    int bits = CHAR_BIT * sizeof(stype);        \
+    *(jtype*)&a >> (jtype)(bits-1);             \
+  })
+
+#define IMPLEMENT_DIRECT(NAME, VTYPE, STYPE, JTYPE, EXPR)       \
+  JTYPE _cl_overloadable NAME(VTYPE a)                          \
+  {                                                             \
+    typedef VTYPE vtype;                                        \
+    typedef STYPE stype;                                        \
+    typedef JTYPE jtype;                                        \
+    return EXPR;                                                \
+  }
+
+
+
+IMPLEMENT_DIRECT(signbit, float  , float, int  , IMPLEMENT_SIGNBIT_BUILTIN_FLOAT)
+IMPLEMENT_DIRECT(signbit, float2 , float, int2 , IMPLEMENT_SIGNBIT_DIRECT)
+IMPLEMENT_DIRECT(signbit, float3 , float, int3 , IMPLEMENT_SIGNBIT_DIRECT)
+IMPLEMENT_DIRECT(signbit, float4 , float, int4 , IMPLEMENT_SIGNBIT_DIRECT)
+IMPLEMENT_DIRECT(signbit, float8 , float, int8 , IMPLEMENT_SIGNBIT_DIRECT)
+IMPLEMENT_DIRECT(signbit, float16, float, int16, IMPLEMENT_SIGNBIT_DIRECT)
+
+IMPLEMENT_DIRECT(signbit, double  , double, int   , IMPLEMENT_SIGNBIT_BUILTIN_DOUBLE)
+IMPLEMENT_DIRECT(signbit, double2 , double, long2 , IMPLEMENT_SIGNBIT_DIRECT)
+IMPLEMENT_DIRECT(signbit, double3 , double, long3 , IMPLEMENT_SIGNBIT_DIRECT)
+IMPLEMENT_DIRECT(signbit, double4 , double, long4 , IMPLEMENT_SIGNBIT_DIRECT)
+IMPLEMENT_DIRECT(signbit, double8 , double, long8 , IMPLEMENT_SIGNBIT_DIRECT)
+IMPLEMENT_DIRECT(signbit, double16, double, long16, IMPLEMENT_SIGNBIT_DIRECT)