4 Commits

Author SHA1 Message Date
khaki3
7d166fa384
[flang][cuda] Correct the number of blocks when setting the grid to * (#121000)
We set the `gridX` argument of `_FortranACUFLaunchKernel` to `-1` when
`*` is passed to the grid parameter. We store it in one of `dim3`
members. However, `dim3` members are unsigned, so positive-value checks
we use later, such as `gridDim.x > 0`, are invalid. This PR utilizes the
original gird-size arguments to compute the number of blocks.
2024-12-23 17:14:38 -08:00
Valentin Clement (バレンタイン クレメン)
6b21cf8cca
[flang][cuda] Compute grid x when calling a kernel with <<<*, block>>> (#115538)
`-1, 1, 1` is passed when calling a kernel with the `<<<*, block>>>`
syntax. Query the device to compute the grid.x value.
2024-11-08 14:34:26 -08:00
Valentin Clement (バレンタイン クレメン)
0b700f2333
[flang][cuda] Add entry point to launch global function with cluster_dims (#113958) 2024-10-29 10:01:49 -07:00
Valentin Clement (バレンタイン クレメン)
e2766b2bce
[flang][cuda] Add entry point to launch cuda fortran kernel (#113490) 2024-10-23 13:44:02 -07:00