diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2019-02-05 19:23:57 +0000 |
|---|---|---|
| committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2019-02-05 19:23:57 +0000 |
| commit | a3f9b71c09613fa51df44a6253afc52e5fff2947 (patch) | |
| tree | 58a2b50ef896d214c294d0d81e66fbfb17147e06 /llvm/include | |
| parent | 51358e45e23ea0889d11a91c4077bb3af406be55 (diff) | |
| download | bcm5719-llvm-a3f9b71c09613fa51df44a6253afc52e5fff2947.tar.gz bcm5719-llvm-a3f9b71c09613fa51df44a6253afc52e5fff2947.zip | |
AMDGPU: Fix assert on trunc from bitcast of build_vector
The v2i64 argument is lowered to a bitcast of v4i32 build_vector.
This would then attempt to use the i32-element as the source of the
vector truncate. This really would need to collect 2 elements from the
build_vector to produce the intended truncate.
llvm-svn: 353202
Diffstat (limited to 'llvm/include')
0 files changed, 0 insertions, 0 deletions
