I am working on a project , it would sometimes need to store over 64 bits, so I can’t use long long type.
I am wondering if there is 128 bits data type supported ?
If yes. Can somebody tell me how to use it ?
If not, I think I would have to use array to simulate 128 bit data type.
It depends on what you mean by “supported.” You mean direct one-instruction operations on 128 bit integer types? Then no.
For storage of 128 bits? Sure, it’s just like the CPU. You can make your own struct, or use the standard 128 bit float4 type. If you need to do operations on it, it’s all classic C++, you can define your own operators for whatever abilities you need.
It depends on what you mean by “supported.” You mean direct one-instruction operations on 128 bit integer types? Then no.
For storage of 128 bits? Sure, it’s just like the CPU. You can make your own struct, or use the standard 128 bit float4 type. If you need to do operations on it, it’s all classic C++, you can define your own operators for whatever abilities you need.
Therez a “load.128-bit” PTX instrn… no?
Therez a “load.128-bit” PTX instrn… no?
Thanks for your reply.
Yes, I want 128 bit operation in one instruction. I know float4, int4. But I think they are just packed struct type.
I think what Sarnath says loading 128 bit instruction do exists, because I see it in the programming guide, it says
a thread can load 32 bit, 64 bit, 128 bit at a time.
What I do want is not really just load 128 bit at a time. I want add, sub… operation in 128 bits.
Is it supported in the hardware ?
Thanks for your reply.
Yes, I want 128 bit operation in one instruction. I know float4, int4. But I think they are just packed struct type.
I think what Sarnath says loading 128 bit instruction do exists, because I see it in the programming guide, it says
a thread can load 32 bit, 64 bit, 128 bit at a time.
What I do want is not really just load 128 bit at a time. I want add, sub… operation in 128 bits.
Is it supported in the hardware ?