CUDA Math Library- Possible Overflow Exceptions?

[shemalrathnasuriya]

Hi! Robert,
I have provided a test case here,

Possible Rounding/Precision Errors in CUDA Math APIs?

Hi! Robert, Thank you for your prompt reply. Below is the sample template we use to test all APIs. We focus on floating point precision at the moment. API Testing Template # cat acos.c #include <cmath> #include <fenv.h> extern "C" float cpp_kernel_1(float x0, int *flag) { float result; // Enable floating-point exceptions feclearexcept(FE_ALL_EXCEPT); result = acos(x0); int exceptions = fetestexcept(FE_ALL_EXCEPT); if (exceptions & FE_DIVBYZERO) { *flag = 1; …