pow/pown ignore the sign of their first argument (e.g. pow(-2,3) gives 8 instead of -8)
This patch is from: https://bugs.debian.org/cgi-bin/bugreport.cgi?msg=5;filename=Fix-pow-erf-tgamma.patch;att=3;bug=768090 Signed-off-by: Zhigang Gong <zhigang.g...@intel.com> --- backend/src/builtin_vector_proto.def | 3 +-- backend/src/ocl_stdlib.tmpl.h | 19 ++++++++++++++++--- 2 files changed, 17 insertions(+), 5 deletions(-) diff --git a/backend/src/builtin_vector_proto.def b/backend/src/builtin_vector_proto.def index 18d23ca..811ab6c 100644 --- a/backend/src/builtin_vector_proto.def +++ b/backend/src/builtin_vector_proto.def @@ -94,8 +94,7 @@ floatn pown (floatn x, intn y) float pown (float x, int y) doublen pown (doublen x, intn y) double pown (double x, int y) -#XXX we define powr as pow -#gentype powr (gentype x, gentype y) +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) diff --git a/backend/src/ocl_stdlib.tmpl.h b/backend/src/ocl_stdlib.tmpl.h index f648a8c..d6900ee 100755 --- a/backend/src/ocl_stdlib.tmpl.h +++ b/backend/src/ocl_stdlib.tmpl.h @@ -3190,7 +3190,6 @@ INLINE_OVERLOADABLE float __gen_ocl_internal_exp10(float x){ #define atan2pi __gen_ocl_internal_atan2pi #define atanpi __gen_ocl_internal_atanpi #define atanh __gen_ocl_internal_atanh -#define pow powr #define cbrt __gen_ocl_internal_cbrt #define rint __gen_ocl_internal_rint #define copysign __gen_ocl_internal_copysign @@ -3729,11 +3728,25 @@ INLINE_OVERLOADABLE float remquo(float x, float y, private int *quo) { BODY; } #undef BODY INLINE_OVERLOADABLE float native_divide(float x, float y) { return x/y; } INLINE_OVERLOADABLE float pown(float x, int n) { - if (x == 0 && n == 0) - return 1; + if (x == 0.f && n == 0) + return 1.f; + if (x < 0.f && (n&1) ) + return -powr(-x, n); return powr(x, n); } +INLINE_OVERLOADABLE float pow(float x, float y) { + int n; + if (x == 0.f && y == 0.f) + return 1.f; + if (x >= 0.f) + return powr(x, y); + n = y; + if ((float)n == y)//is exact integer + return pown(x, n); + return NAN; +} + INLINE_OVERLOADABLE float internal_rootn(float x, int n, const bool isFastpath) { float ax,re; -- 1.8.3.2 _______________________________________________ Beignet mailing list Beignet@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/beignet
