CUDA.jl write to global memory in PTX

Specific Domains GPU
,
5

Great thanks, now it worked!

Also other great news: With this I was able to fix my problem.

Here is the updated PTX-Code:

.version 8.5 // changed from 7.1 to 8.5
.target sm_61
.address_size 64

.visible .entry ExpressionProcessing(
.param .u64 param_1) // changed from .u32 to u.64
{
	.reg .b64   %parameter<1>; // changed from .u32 to .b64
	.reg .b64   %i<1>; // changed from .u32 to .b64

	ld.param.u64   %i0, [param_1]; // changed from .u32 to .b64
	cvta.to.global.u64   %parameter0, %i0; // changed from .u32 to .b64

	st.global.f32  [%parameter0], 10.0;
	ret;
}

With these changes I was also able to get my more complicated PTX code running again!

Thank you guys for helping me out!

1 Like