taco
taco copied to clipboard
tensor-compiler
The Tensor Algebra Compiler (taco) computes sparse tensor expressions on CPUs and GPUs
The command line parameter `-write-source=` writes several functions to the specified file. This includes utility functions and macros, I think it is intended to write everything needed to run the...
This is a proposal for adding a reduction node/nodes to the index expression IR. The rationale is that our current approach of inferring summations based on a convention is flawed,...
When contracting two sparse matrices, the following code does not work: A(i, k) = B(i, j) * C(j, k); Here A, B, C are matrices which are sparse in all...
This command is the offending one: ``` ./bin/taco "J(i,j) = A1(i,j) + A2(i,j,k)*x(k) + A2(i,l,j)*x(l)" -f=J:sd:0,1 -f=A1:sd:0,1 -f=A2:sss:0,1,2 -f=x:d:0 -write-source=taco_kernel.c -write-compute=taco_compute.c -write-assembly=taco_assembly.c ``` The problem lies with the `A2` tensor....
Is there a way to avoid zero-initializing a tensor before assigning to it? I want to do something like: ``` Tensor A({8}, Format{Dense}); Tensor B({3}, Format{Dense}); for (int i =...
# Summary Hello! I'm playing around with TACO and would like to pass some NumPy tensors to C++ and convert them to TACO's `Tensor` type before working with them in...
# Summary Hello! I'm working with a long expressions that require several, sequential sets of tensor contractions. In Python they look similar to: ```python Fki = 2*np.einsum('kcld,ilcd->ki', eris_ovov, t2) Fki...
Hi! I have opened this issue to ask some general questions regarding the GSoC project that aims to complete the Python bindings for TACO. From what I've discussed with Hameer...
I have 4 tensors: the output tensor Soperator with three indices, 2 larger Tensors S and Q with 6 indices each and an input tensor fnow with three indices. All...
Metadata
Owner
Metadata
The Tensor Algebra Compiler (taco) computes sparse tensor expressions on CPUs and GPUs