stdml/stdtensor

`cudaMalloc<T>` allocates invalid memory

Opened this issue · 3 comments

`cudaMalloc` allocates invalid memory

happened again when used with thrust.

sjdrc commented

Is this still an issue as of #44?

Since #44, the template version of cudaMalloc is replaced by the non-template version, and I don't see cuda error anymore. But I'd like to re-introduce cudaMalloc<T> in the future with careful test, and a guard macro.