[Beignet] [PATCH] Added memory space parameters support at the autogeneration script.
Zhigang Gong
zhigang.gong at linux.intel.com
Mon Jul 29 20:57:30 PDT 2013
Ping for comments. Thanks.
On Thu, Jul 25, 2013 at 03:07:02PM +0800, Zhigang Gong wrote:
> Enhance the python script to support pointer with memory space
> type, such as :
>
> gentype fract (gentype x, __global gentype *iptr)
> gentype fract (gentype x, __local gentype *iptr)
> gentype fract (gentype x, __private gentype *iptr)
>
> So enable the following function at the builtin function spec file
> fract/frexp/modf/nextafter/remquo/sincos.
>
> Remove the duplicate at the ocl_stdlib.tmp.h.
>
> Signed-off-by: Zhigang Gong <zhigang.gong at linux.intel.com>
> ---
> backend/src/builtin_vector_proto.def | 68 +++++++--------
> backend/src/gen_builtin_vector.py | 69 ++++++++++++---
> backend/src/ocl_stdlib.tmpl.h | 157 ----------------------------------
> 3 files changed, 89 insertions(+), 205 deletions(-)
>
> diff --git a/backend/src/builtin_vector_proto.def b/backend/src/builtin_vector_proto.def
> index 13d7a34..d55272d 100644
> --- a/backend/src/builtin_vector_proto.def
> +++ b/backend/src/builtin_vector_proto.def
> @@ -35,21 +35,21 @@ gentype fmin (gentype x, gentype y)
> gentypef fmin (gentypef x, float y)
> gentyped fmin (gentyped x, double y)
> gentype fmod (gentype x, gentype y)
> -#gentype fract (gentype x, __global gentype *iptr)
> -#gentype fract (gentype x, __local gentype *iptr)
> -#gentype fract (gentype x, __private gentype *iptr)
> -#floatn frexp (floatn x, __global intn *exp)
> -#floatn frexp (floatn x, __local intn *exp)
> -#floatn frexp (floatn x, __private intn *exp)
> -#float frexp (float x, __global int *exp)
> -#float frexp (float x, __local int *exp)
> -#float frexp (float x, __private int *exp)
> -#doublen frexp (doublen x, __global intn *exp)
> -#doublen frexp (doublen x, __local intn *exp)
> -#doublen frexp (doublen x, __private intn *exp)
> -#double frexp (double x, __global int *exp)
> -#double frexp (double x, __local int *exp)
> -#double frexp (double x, __private int *exp)
> +gentype fract (gentype x, __global gentype *iptr)
> +gentype fract (gentype x, __local gentype *iptr)
> +gentype fract (gentype x, __private gentype *iptr)
> +floatn frexp (floatn x, __global intn *exp)
> +floatn frexp (floatn x, __local intn *exp)
> +floatn frexp (floatn x, __private intn *exp)
> +float frexp (float x, __global int *exp)
> +float frexp (float x, __local int *exp)
> +float frexp (float x, __private int *exp)
> +doublen frexp (doublen x, __global intn *exp)
> +doublen frexp (doublen x, __local intn *exp)
> +doublen frexp (doublen x, __private intn *exp)
> +double frexp (double x, __global int *exp)
> +double frexp (double x, __local int *exp)
> +double frexp (double x, __private int *exp)
> gentype hypot (gentype x, gentype y)
> intn ilogb (floatn x)
> int ilogb (float x)
> @@ -82,14 +82,14 @@ gentype logb (gentype x)
> gentype mad (gentype a, gentype b, gentype c)
> gentype maxmag (gentype x, gentype y)
> gentype minmag (gentype x, gentype y)
> -#gentype modf (gentype x, __global gentype *iptr)
> -#gentype modf (gentype x, __local gentype *iptr)
> -#gentype modf (gentype x, __private gentype *iptr)
> +gentype modf (gentype x, __global gentype *iptr)
> +gentype modf (gentype x, __local gentype *iptr)
> +gentype modf (gentype x, __private gentype *iptr)
> floatn nan (uintn nancode)
> float nan (uint nancode)
> doublen nan (ulongn nancode)
> double nan (ulong nancode)
> -#gentype nextafter (gentype x, gentype y)
> +gentype nextafter (gentype x, gentype y)
> gentype pow (gentype x, gentype y)
> floatn pown (floatn x, intn y)
> float pown (float x, int y)
> @@ -98,18 +98,18 @@ double pown (double x, int y)
> #XXX we define powr as pow
> #gentype powr (gentype x, gentype y)
> gentype remainder (gentype x, gentype y)
> -#floatn remquo (floatn x, floatn y, __global intn *quo)
> -#floatn remquo (floatn x, floatn y, __local intn *quo)
> -#floatn remquo (floatn x, floatn y, __private intn *quo)
> -#float remquo (float x, float y, __global int *quo)
> -#float remquo (float x, float y, __local int *quo)
> -#float remquo (float x, float y, __private int *quo)
> -#doublen remquo (doublen x, doublen y, __global intn *quo)
> -#doublen remquo (doublen x, doublen y, __local intn *quo)
> -#doublen remquo (doublen x, doublen y, __private intn *quo)
> -#double remquo (double x, double y, __global int *quo)
> -#double remquo (double x, double y, __local int *quo)
> -#double remquo (double x, double y, __private int *quo)
> +floatn remquo (floatn x, floatn y, __global intn *quo)
> +floatn remquo (floatn x, floatn y, __local intn *quo)
> +floatn remquo (floatn x, floatn y, __private intn *quo)
> +float remquo (float x, float y, __global int *quo)
> +float remquo (float x, float y, __local int *quo)
> +float remquo (float x, float y, __private int *quo)
> +doublen remquo (doublen x, doublen y, __global intn *quo)
> +doublen remquo (doublen x, doublen y, __local intn *quo)
> +doublen remquo (doublen x, doublen y, __private intn *quo)
> +double remquo (double x, double y, __global int *quo)
> +double remquo (double x, double y, __local int *quo)
> +double remquo (double x, double y, __private int *quo)
> gentype rint (gentype)
> floatn rootn (floatn x, intn y)
>
> @@ -118,9 +118,9 @@ doublen rootn (double x, int y)
> gentype round (gentype x)
> gentype rsqrt (gentype)
> gentype sin (gentype)
> -#gentype sincos (gentype x, __global gentype *cosval)
> -#gentype sincos (gentype x, __local gentype *cosval)
> -#gentype sincos (gentype x, __private gentype *cosval)
> +gentype sincos (gentype x, __global gentype *cosval)
> +gentype sincos (gentype x, __local gentype *cosval)
> +gentype sincos (gentype x, __private gentype *cosval)
> gentype sinh (gentype)
> gentype sinpi (gentype x)
> gentype sqrt (gentype)
> diff --git a/backend/src/gen_builtin_vector.py b/backend/src/gen_builtin_vector.py
> index f5b0f3c..b073682 100755
> --- a/backend/src/gen_builtin_vector.py
> +++ b/backend/src/gen_builtin_vector.py
> @@ -39,6 +39,19 @@ def gen_vector_type(type_set, vector_set = all_vector):
> ret.append((t, i))
> return ret
>
> +def set_vector_memspace(vector_type_set, memspace):
> + ret = []
> + if memspace == '':
> + return vector_type_set
> + for t in vector_type_set:
> + ret.append((t[0], t[1], memspace))
> + return ret
> +
> +# if we have 3 elements in the type tuple, we are a pointer with a memory space type
> +# at the third element.
> +def isPointer(t):
> + return len(t) == 3
> +
> all_itype = "char","short","int","long"
> all_utype = "uchar","ushort","uint","ulong"
> all_int_type = all_itype + all_utype
> @@ -98,8 +111,18 @@ def _prefix(prefix, dtype):
> return prefix + '_' + dtype
> return dtype
>
> +memspaces = ["__local ", "__private ", "__global "]
> +
> +def stripMemSpace(t):
> + if t[0:2] == '__':
> + for memspace in memspaces :
> + if t[0:len(memspace)] == memspace:
> + return memspace, t[len(memspace):]
> + return '', t
> +
> def check_type(types):
> for t in types:
> + memspace, t = stripMemSpace(t)
> if not t in type_dict:
> print t
> raise "found invalid type."
> @@ -167,7 +190,6 @@ def fixup_type(dstType, srcType, n):
> print dstType, srcType
> raise "type mispatch"
>
> -
> class builtinProto():
> valueTypeStr = ""
> functionName = ""
> @@ -191,7 +213,6 @@ class builtinProto():
>
> def append(self, line, nextInit = ""):
> self.outputStr.append(line);
> - #print line
> return nextInit;
>
> def indentSpace(self):
> @@ -203,25 +224,34 @@ class builtinProto():
>
> def init_from_line(self, t):
> self.append('//{}'.format(t))
> - line = filter(None, re.split(',| |\(', t.rstrip('\n)')))
> + line = filter(None, re.split(',| |\(', t.rstrip(')\n')))
> self.paramCount = 0
> + stripped = 0
> + memSpace = ''
> for i, text in enumerate(line):
> - if i == 0:
> + idx = i - stripped
> + if idx == 0:
> self.valueTypeStr = _prefix(self.prefix, line[i])
> continue
>
> - if i == 1:
> + if idx == 1:
> self.functionName = line[i];
> continue
>
> - if i % 2 == 0:
> + if idx % 2 == 0:
> if line[i][0] == '(':
> tmpType = line[i][1:]
> else:
> tmpType = line[i]
> -
> - self.paramTypeStrs.append(_prefix(self.prefix, tmpType))
> - self.paramCount += 1;
> + if tmpType == '__local' or \
> + tmpType == '__private' or \
> + tmpType == '__global':
> + memSpace = tmpType + ' '
> + stripped += 1
> + continue
> + self.paramTypeStrs.append(memSpace + _prefix(self.prefix, tmpType))
> + memSpace = ''
> + self.paramCount += 1
>
> def gen_proto_str_1(self, vtypeSeq, ptypeSeqs, i):
> for n in range(0, self.paramCount):
> @@ -243,10 +273,16 @@ class builtinProto():
> if vtype[1] == 1:
> return
>
> + if isPointer(ptype):
> + formatStr += ptype[2]
> + pointerStr = '*'
> + else:
> + pointerStr = ''
> +
> if ptype[1] != 1:
> - formatStr += '{}{} param{}'.format(ptype[0], ptype[1], n)
> + formatStr += '{}{} {}param{}'.format(ptype[0], ptype[1], pointerStr, n)
> else:
> - formatStr += '{} param{}'.format(ptype[0], n)
> + formatStr += '{} {}param{}'.format(ptype[0], pointerStr, n)
>
> formatStr += ')'
> formatStr = self.append(formatStr, '{{return ({}{})('.format(vtype[0], vtype[1]))
> @@ -269,11 +305,15 @@ class builtinProto():
> if vtype[1] != ptype[1]:
> if ptype[1] != 1:
> raise "parameter is not a scalar but has different width with result value."
> -
> + if isPointer(ptype):
> + formatStr += '&'
> formatStr += 'param{}'.format(n)
> continue
>
> - formatStr += 'param{}.s{:x}'.format(n, j)
> + if (isPointer(ptype)):
> + formatStr += '({} {} *)param{} + {:2d}'.format(ptype[2], ptype[0], n, j)
> + else:
> + formatStr += 'param{}.s{:x}'.format(n, j)
>
> formatStr += ')'
>
> @@ -296,7 +336,8 @@ class builtinProto():
> ptypeSeqs = []
> count = len(vtypeSeq);
> for t in self.paramTypeStrs:
> - ptypeSeqs.append(type_dict[t])
> + memspace,t = stripMemSpace(t)
> + ptypeSeqs.append(set_vector_memspace(type_dict[t], memspace))
> count = max(count, len(type_dict[t]))
>
> for i in range(count):
> diff --git a/backend/src/ocl_stdlib.tmpl.h b/backend/src/ocl_stdlib.tmpl.h
> index 5badd56..cc626a1 100644
> --- a/backend/src/ocl_stdlib.tmpl.h
> +++ b/backend/src/ocl_stdlib.tmpl.h
> @@ -548,44 +548,6 @@ INLINE_OVERLOADABLE float sincos(float x, global float *cosval) { return __gen_o
> INLINE_OVERLOADABLE float sincos(float x, local float *cosval) { return __gen_ocl_internal_sincos(x, (float*)cosval); }
> INLINE_OVERLOADABLE float sincos(float x, private float *cosval) { return __gen_ocl_internal_sincos(x, (float*)cosval); }
>
> -INLINE_OVERLOADABLE float2 __gen_ocl_internal_sincos(float2 x, float2 *cosval) {
> - return (float2)(__gen_ocl_internal_sincos(x.s0, (float *)cosval),
> - __gen_ocl_internal_sincos(x.s1, 1 + (float *)cosval));
> -}
> -INLINE_OVERLOADABLE float4 __gen_ocl_internal_sincos(float4 x, float4 *cosval) {
> - return (float4)(__gen_ocl_internal_sincos(x.s0, (float *)cosval),
> - __gen_ocl_internal_sincos(x.s1, 1 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s2, 2 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s3, 3 + (float *)cosval));
> -}
> -INLINE_OVERLOADABLE float8 __gen_ocl_internal_sincos(float8 x, float8 *cosval) {
> - return (float8)(__gen_ocl_internal_sincos(x.s0, (float *)cosval),
> - __gen_ocl_internal_sincos(x.s1, 1 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s2, 2 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s3, 3 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s4, 4 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s5, 5 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s6, 6 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s7, 7 + (float *)cosval));
> -}
> -INLINE_OVERLOADABLE float16 __gen_ocl_internal_sincos(float16 x, float16 *cosval) {
> - return (float16)(__gen_ocl_internal_sincos(x.s0, (float *)cosval),
> - __gen_ocl_internal_sincos(x.s1, 1 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s2, 2 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s3, 3 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s4, 4 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s5, 5 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s6, 6 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s7, 7 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s8, 8 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.s9, 9 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.sa, 10 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.sb, 11 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.sc, 12 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.sd, 13 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.se, 14 + (float *)cosval),
> - __gen_ocl_internal_sincos(x.sf, 15 + (float *)cosval));
> -}
> INLINE_OVERLOADABLE float __gen_ocl_internal_sinh(float x) {
> return (1 - native_exp(-2 * x)) / (2 * native_exp(-x));
> }
> @@ -779,26 +741,6 @@ INLINE_OVERLOADABLE float frexp(float x, global int *exp) { return __gen_ocl_fre
> INLINE_OVERLOADABLE float frexp(float x, local int *exp) { return __gen_ocl_frexp(x, (int *)exp); }
> INLINE_OVERLOADABLE float frexp(float x, private int *exp) { return __gen_ocl_frexp(x, (int *)exp); }
>
> -INLINE_OVERLOADABLE float2 frexp(float2 x, int2 *exp) {
> - return (float2)(frexp(x.s0, (int *)exp), frexp(x.s1, 1 + (int *)exp));
> -}
> -
> -INLINE_OVERLOADABLE float3 frexp(float3 x, int3 *exp) {
> - return (float3)(frexp(x.s0, (int *)exp), frexp(x.s1, 1 + (int *)exp), frexp(x.s2, 2 + (int *)exp));
> -}
> -
> -INLINE_OVERLOADABLE float4 frexp(float4 x, int4 *exp) {
> - return (float4)(frexp(x.s0, (int *)exp), frexp(x.s1, 1 + (int *)exp), frexp(x.s2, 2 + (int *)exp), frexp(x.s3, 3 + (int *)exp));
> -}
> -
> -INLINE_OVERLOADABLE float8 frexp(float8 x, int8 *exp) {
> - return (float8)(frexp(x.s0, (int *)exp), frexp(x.s1, 1 + (int *)exp), frexp(x.s2, 2 + (int *)exp), frexp(x.s3, 3 + (int *)exp), frexp(x.s4, 4 + (int *)exp), frexp(x.s5, 5 + (int *)exp), frexp(x.s6, 6 + (int *)exp), frexp(x.s7, 7 + (int *)exp));
> -}
> -
> -INLINE_OVERLOADABLE float16 frexp(float16 x, int16 *exp) {
> - return (float16)(frexp(x.s0, (int *)exp), frexp(x.s1, 1 + (int *)exp), frexp(x.s2, 2 + (int *)exp), frexp(x.s3, 3 + (int *)exp), frexp(x.s4, 4 + (int *)exp), frexp(x.s5, 5 + (int *)exp), frexp(x.s6, 6 + (int *)exp), frexp(x.s7, 7 + (int *)exp), frexp(x.s8, 8 + (int *)exp), frexp(x.s9, 9 + (int *)exp), frexp(x.sa, 10 + (int *)exp), frexp(x.sb, 11 + (int *)exp), frexp(x.sc, 12 + (int *)exp), frexp(x.sd, 13 + (int *)exp), frexp(x.se, 14 + (int *)exp), frexp(x.sf, 15 + (int *)exp));
> -}
> -
> INLINE_OVERLOADABLE float nextafter(float x, float y) {
> uint hx = as_uint(x), ix = hx & 0x7FFFFFFF;
> uint hy = as_uint(y), iy = hy & 0x7FFFFFFF;
> @@ -829,39 +771,9 @@ INLINE_OVERLOADABLE float __gen_ocl_modf(float x, float *i) {
> return x - *i;
> }
>
> -INLINE_OVERLOADABLE float2 __gen_ocl_modf(float2 x, float2 *i) {
> - return (float2)(__gen_ocl_modf(x.s0, (float *)i), __gen_ocl_modf(x.s1, 1 + (float *)i));
> -}
> -
> -INLINE_OVERLOADABLE float3 __gen_ocl_modf(float3 x, float3 *i) {
> - return (float3)(__gen_ocl_modf(x.s0, (float *)i), __gen_ocl_modf(x.s1, 1 + (float *)i), __gen_ocl_modf(x.s2, 2 + (float *)i));
> -}
> -
> -INLINE_OVERLOADABLE float4 __gen_ocl_modf(float4 x, float4 *i) {
> - return (float4)(__gen_ocl_modf(x.s0, (float *)i), __gen_ocl_modf(x.s1, 1 + (float *)i), __gen_ocl_modf(x.s2, 2 + (float *)i), __gen_ocl_modf(x.s3, 3 + (float *)i));
> -}
> -
> -INLINE_OVERLOADABLE float8 __gen_ocl_modf(float8 x, float8 *i) {
> - return (float8)(__gen_ocl_modf(x.s0, (float *)i), __gen_ocl_modf(x.s1, 1 + (float *)i), __gen_ocl_modf(x.s2, 2 + (float *)i), __gen_ocl_modf(x.s3, 3 + (float *)i), __gen_ocl_modf(x.s4, 4 + (float *)i), __gen_ocl_modf(x.s5, 5 + (float *)i), __gen_ocl_modf(x.s6, 6 + (float *)i), __gen_ocl_modf(x.s7, 7 + (float *)i));
> -}
> -
> -INLINE_OVERLOADABLE float16 __gen_ocl_modf(float16 x, float16 *i) {
> - return (float16)(__gen_ocl_modf(x.s0, (float *)i), __gen_ocl_modf(x.s1, 1 + (float *)i), __gen_ocl_modf(x.s2, 2 + (float *)i), __gen_ocl_modf(x.s3, 3 + (float *)i), __gen_ocl_modf(x.s4, 4 + (float *)i), __gen_ocl_modf(x.s5, 5 + (float *)i), __gen_ocl_modf(x.s6, 6 + (float *)i), __gen_ocl_modf(x.s7, 7 + (float *)i), __gen_ocl_modf(x.s8, 8 + (float *)i), __gen_ocl_modf(x.s9, 9 + (float *)i), __gen_ocl_modf(x.sa, 10 + (float *)i), __gen_ocl_modf(x.sb, 11 + (float *)i), __gen_ocl_modf(x.sc, 12 + (float *)i), __gen_ocl_modf(x.sd, 13 + (float *)i), __gen_ocl_modf(x.se, 14 + (float *)i), __gen_ocl_modf(x.sf, 15 + (float *)i));
> -}
> -
> INLINE_OVERLOADABLE float modf(float x, global float *i) { return __gen_ocl_modf(x, (float *)i); }
> INLINE_OVERLOADABLE float modf(float x, local float *i) { return __gen_ocl_modf(x, (float *)i); }
> INLINE_OVERLOADABLE float modf(float x, private float *i) { return __gen_ocl_modf(x, (float *)i); }
> -#define DEF(n) \
> - INLINE_OVERLOADABLE float##n modf(float##n x, global float##n *i) { return __gen_ocl_modf(x, (float##n *)i); } \
> - INLINE_OVERLOADABLE float##n modf(float##n x, local float##n *i) { return __gen_ocl_modf(x, (float##n *)i); } \
> - INLINE_OVERLOADABLE float##n modf(float##n x, private float##n *i) { return __gen_ocl_modf(x, (float##n *)i); }
> -DEF(2)
> -DEF(3)
> -DEF(4)
> -DEF(8)
> -DEF(16)
> -#undef DEF
>
> INLINE_OVERLOADABLE float degrees(float radians) { return (180 / M_PI_F) * radians; }
> INLINE_OVERLOADABLE float radians(float degrees) { return (M_PI_F / 180) * degrees; }
> @@ -903,45 +815,6 @@ INLINE_OVERLOADABLE float fract(float x, global float *p) { return __gen_ocl_fra
> INLINE_OVERLOADABLE float fract(float x, local float *p) { return __gen_ocl_fract(x, (float *)p); }
> INLINE_OVERLOADABLE float fract(float x, private float *p) { return __gen_ocl_fract(x, (float *)p); }
>
> -INLINE_OVERLOADABLE float2 fract(float2 x, float2 *p) {
> - return (float2)(fract(x.s0, (float *)p),
> - fract(x.s1, 1 + (float *)p));
> -}
> -INLINE_OVERLOADABLE float4 fract(float4 x, float4 *p) {
> - return (float4)(fract(x.s0, (float *)p),
> - fract(x.s1, 1 + (float *)p),
> - fract(x.s2, 2 + (float *)p),
> - fract(x.s3, 3 + (float *)p));
> -}
> -INLINE_OVERLOADABLE float8 fract(float8 x, float8 *p) {
> - return (float8)(fract(x.s0, (float *)p),
> - fract(x.s1, 1 + (float *)p),
> - fract(x.s2, 2 + (float *)p),
> - fract(x.s3, 3 + (float *)p),
> - fract(x.s4, 4 + (float *)p),
> - fract(x.s5, 5 + (float *)p),
> - fract(x.s6, 6 + (float *)p),
> - fract(x.s7, 7 + (float *)p));
> -}
> -INLINE_OVERLOADABLE float16 fract(float16 x, float16 *p) {
> - return (float16)(fract(x.s0, (float *)p),
> - fract(x.s1, 1 + (float *)p),
> - fract(x.s2, 2 + (float *)p),
> - fract(x.s3, 3 + (float *)p),
> - fract(x.s4, 4 + (float *)p),
> - fract(x.s5, 5 + (float *)p),
> - fract(x.s6, 6 + (float *)p),
> - fract(x.s7, 7 + (float *)p),
> - fract(x.s8, 8 + (float *)p),
> - fract(x.s9, 9 + (float *)p),
> - fract(x.sa, 10 + (float *)p),
> - fract(x.sb, 11 + (float *)p),
> - fract(x.sc, 12 + (float *)p),
> - fract(x.sd, 13 + (float *)p),
> - fract(x.se, 14 + (float *)p),
> - fract(x.sf, 15 + (float *)p));
> -}
> -
> INLINE_OVERLOADABLE float __gen_ocl_remquo(float x, float y, int *quo) {
> uint hx = as_uint(x), ix = hx & 0x7FFFFFFF, hy = as_uint(y), iy = hy & 0x7FFFFFFF;
> if (ix > 0x7F800000 || iy > 0x7F800000 || ix == 0x7F800000 || iy == 0)
> @@ -956,39 +829,9 @@ INLINE_OVERLOADABLE float __gen_ocl_remquo(float x, float y, int *quo) {
> return as_float(hr);
> }
>
> -INLINE_OVERLOADABLE float2 __gen_ocl_remquo(float2 x, float2 y, int2 *i) {
> - return (float2)(__gen_ocl_remquo(x.s0, y.s0, (int *)i), __gen_ocl_remquo(x.s1, y.s1, 1 + (int *)i));
> -}
> -
> -INLINE_OVERLOADABLE float3 __gen_ocl_remquo(float3 x, float3 y, int3 *i) {
> - return (float3)(__gen_ocl_remquo(x.s0, y.s0, (int *)i), __gen_ocl_remquo(x.s1, y.s1, 1 + (int *)i), __gen_ocl_remquo(x.s2, y.s2, 2 + (int *)i));
> -}
> -
> -INLINE_OVERLOADABLE float4 __gen_ocl_remquo(float4 x, float4 y, int4 *i) {
> - return (float4)(__gen_ocl_remquo(x.s0, y.s0, (int *)i), __gen_ocl_remquo(x.s1, y.s1, 1 + (int *)i), __gen_ocl_remquo(x.s2, y.s2, 2 + (int *)i), __gen_ocl_remquo(x.s3, y.s3, 3 + (int *)i));
> -}
> -
> -INLINE_OVERLOADABLE float8 __gen_ocl_remquo(float8 x, float8 y, int8 *i) {
> - return (float8)(__gen_ocl_remquo(x.s0, y.s0, (int *)i), __gen_ocl_remquo(x.s1, y.s1, 1 + (int *)i), __gen_ocl_remquo(x.s2, y.s2, 2 + (int *)i), __gen_ocl_remquo(x.s3, y.s3, 3 + (int *)i), __gen_ocl_remquo(x.s4, y.s4, 4 + (int *)i), __gen_ocl_remquo(x.s5, y.s5, 5 + (int *)i), __gen_ocl_remquo(x.s6, y.s6, 6 + (int *)i), __gen_ocl_remquo(x.s7, y.s7, 7 + (int *)i));
> -}
> -
> -INLINE_OVERLOADABLE float16 __gen_ocl_remquo(float16 x, float16 y, int16 *i) {
> - return (float16)(__gen_ocl_remquo(x.s0, y.s0, (int *)i), __gen_ocl_remquo(x.s1, y.s1, 1 + (int *)i), __gen_ocl_remquo(x.s2, y.s2, 2 + (int *)i), __gen_ocl_remquo(x.s3, y.s3, 3 + (int *)i), __gen_ocl_remquo(x.s4, y.s4, 4 + (int *)i), __gen_ocl_remquo(x.s5, y.s5, 5 + (int *)i), __gen_ocl_remquo(x.s6, y.s6, 6 + (int *)i), __gen_ocl_remquo(x.s7, y.s7, 7 + (int *)i), __gen_ocl_remquo(x.s8, y.s8, 8 + (int *)i), __gen_ocl_remquo(x.s9, y.s9, 9 + (int *)i), __gen_ocl_remquo(x.sa, y.sa, 10 + (int *)i), __gen_ocl_remquo(x.sb, y.sb, 11 + (int *)i), __gen_ocl_remquo(x.sc, y.sc, 12 + (int *)i), __gen_ocl_remquo(x.sd, y.sd, 13 + (int *)i), __gen_ocl_remquo(x.se, y.se, 14 + (int *)i), __gen_ocl_remquo(x.sf, y.sf, 15 + (int *)i));
> -}
> -
> INLINE_OVERLOADABLE float remquo(float x, float y, global int *quo) { return __gen_ocl_remquo(x, y, (int *)quo); }
> INLINE_OVERLOADABLE float remquo(float x, float y, local int *quo) { return __gen_ocl_remquo(x, y, (int *)quo); }
> INLINE_OVERLOADABLE float remquo(float x, float y, private int *quo) { return __gen_ocl_remquo(x, y, (int *)quo); }
> -#define DEF(n) \
> - INLINE_OVERLOADABLE float##n remquo(float##n x, float##n y, global int##n *quo) { return __gen_ocl_remquo(x, y, (int##n *)quo); } \
> - INLINE_OVERLOADABLE float##n remquo(float##n x, float##n y, local int##n *quo) { return __gen_ocl_remquo(x, y, (int##n *)quo); } \
> - INLINE_OVERLOADABLE float##n remquo(float##n x, float##n y, private int##n *quo) { return __gen_ocl_remquo(x, y, (int##n *)quo); }
> -DEF(2)
> -DEF(3)
> -DEF(4)
> -DEF(8)
> -DEF(16)
> -#undef DEF
>
> INLINE_OVERLOADABLE float native_divide(float x, float y) { return x/y; }
> INLINE_OVERLOADABLE float ldexp(float x, int n) {
> --
> 1.7.9.5
>
> _______________________________________________
> Beignet mailing list
> Beignet at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/beignet
More information about the Beignet
mailing list