Comments (4)
Hi @sksg, when you're passing a tensor to a device/global function the tensor must first be converted to tensor_impl_t
. The reason is that tensor_t
maintains some state, like a reference count, that is not usable or wanted in device code. tensor_impl_t
strips a tensor to the bare minimum needed to be used on device code. To convert, you use the base_type
trait:
MatX/include/matx/operators/set.h
Line 66 in d9a32d5
So your code may be something like
int main() {
auto a = matx::make_tensor<float>({10});
auto a_base = base_type<decltype(a)>(a);
kernel_launch<<<0, 10>>>([=] __device__() { a_base(threadIdx.x) = threadIdx.x; }); // <-- error line
}
I have not tried this yet to see if it works, but I will test and get back to you today.
from matx.
Hi @sksg, the issue is that there are two definitions of operator()
, a const and a non-const version. Since lambdas capture by const, it's choosing the const version that cannot modify an lvalue. Instead you can pass it as a function parameter:
using namespace matx;
template<typename T, typename FUNC>
__global__ void kernel_launch(T t, FUNC kernel) {
kernel(t);
}
kernel_launch<<<0, 10>>>(a_base, [=] __device__(auto ab) {
ab(threadIdx.x) = threadIdx.x; });
And this works as well. Let me know if that suffices.
from matx.
Thanks for the speedy feedback.
I did not know that lambdas were const by default. It then makes perfect sense. I did try it with a mutable lambda instead:
#include <matx.h>
template<typename FUNC>
__global__ void kernel_launch(FUNC kernel) {
kernel();
}
int main() {
auto a = matx::make_tensor<float>({10});
kernel_launch<<<0, 10>>>([=] __device__() mutable { a(threadIdx.x) = threadIdx.x; }); // <-- No errors!
}
Technically, I can verify that the device array a
has the correct values. But I see your point wrt. matx::base_type
. I do not know if any of the host state will be corrupted by using a
directly rather than a_base
. The above syntax is very concise, though.
Thanks for helping out. This is perfectly usable for me.
from matx.
Thanks for the report!
from matx.
Related Issues (20)
- [QST] Cannot call device executor using host compiler HOT 7
- print on diag operator appears broken [BUG] HOT 3
- [FEA] Add vector and matrix norm operators
- [BUG] Renamed `rmax`/`rmin`
- [QST] After running some computations on a tensor, how to fetch the entire tensor back to host memory? HOT 3
- [DOC] Document enumerated types
- [DOC] Add host/cuda support to docs
- [FEA] Consider custom reduction if using non-contiguous stride
- [FEA] Add support for mdspan
- [FEA] Add support to extract the exponents of tensor with float / complex values HOT 1
- [BUG] `cmake_sample_project` cannot build when setting `MATX_EN_CUTLASS=ON` HOT 6
- [BUG] `cmake_sample_project` cannot build when setting `MATX_EN_PYBIND11=ON` on Python >= 3.11 HOT 6
- [FEA] cuTENSOR 2.0 support HOT 9
- [BUG] `clone(slice(diag(A)))` failed HOT 1
- [BUG] Simple einsum program failed to compile HOT 4
- [FEA] Add a `find_peaks` function
- [BUG] Compilation error with MATX_ENABLE_CUTLASS in transforms/matmul.h HOT 3
- [BUG]: #define print replaces print functions in other libraries HOT 3
- [BUG] Clang cannot compile as a host compiler
- [BUG] Conversion issue with matxBf16 and logical > with float literal HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from matx.