tree f7f8597308f408c643164302fc5443239f1e76fd
parent da22c84fa2261cf5df566029725402d565d1e7b0
author Han-Chung Wang <hanchung@google.com> 1677786356 -0800
committer GitHub <noreply@github.com> 1677786356 +0000
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJkAPz0CRBK7hj4Ov3rIwAA904IAF9/xyUP2j0fgNAEJV9FsaSx
 rp+U5PDJli2VYel3iY11btPjS4fQW45D1VTEQI5ipxz+94Mszr0R/puh5yDcSHi6
 MEN+e1fmbnCnQv7r4g3h2rn8SSRCf2FJIaNdT65kkJqvYbqzUyGLvhy5gFCgs7Qf
 lRCYlpD9kplXjvo5XrvQ4rRB/KxolW2wVJtlXytvQH0NtgmJackkpExFVyu1+umD
 E+N2t0Ufr7c3L96GQFMdQLgoi1tu7X/7CMZEHs3WltW5K2EqP/O8x1kjBMGb8Pq/
 qyrVRZbBX/GSolVbi43dv2ifeCkaKqsCaZWdL4JYT8IO+F7Ky5WwGatzkOxk1WY=
 =iIve
 -----END PGP SIGNATURE-----
 

[LLVMGPU] Enhance TensorPad pass to handle tensor.unpack ops. (#12458)

The unpack op can have extract_slice semantics. The pass pads the
tensor.unpack ops that gets rid of extract_slice. This helps vector
transfer ops be in bounds.