OSDN Git Service

Merging r339190:
authorHans Wennborg <hans@hanshq.net>
Wed, 8 Aug 2018 11:35:18 +0000 (11:35 +0000)
committerHans Wennborg <hans@hanshq.net>
Wed, 8 Aug 2018 11:35:18 +0000 (11:35 +0000)
commit62ca9ef3b9aff92f53de29cb74429f58d4201c82
tree85f98aa8a8aa7d07ef7637d68550dbc194e3bbda
parentb87385fdc0bb895bd93cbf00640e5495becfc22e
Merging r339190:
------------------------------------------------------------------------
r339190 | jvesely | 2018-08-07 23:54:37 +0200 (Tue, 07 Aug 2018) | 12 lines

AMDGPU: Remove broken i16 ternary patterns

Fixup test to check for GCN prefix
These patterns always zero extend the result even though it might need sign extension.
This has been broken since the addition of i16 support.
It has popped up in mad_sat(char) test since min(max()) combination is turned into v_med3, resulting in the following (incorrect) sequence:
        v_mad_i16 v2, v10, v9, v11
        v_med3_i32 v2, v2, v8, v7

Fixes mad_sat(char) piglit on VI.

Differential Revision: https://reviews.llvm.org/D49836
------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_70@339235 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/AMDGPU/VOP3Instructions.td
test/CodeGen/AMDGPU/mad_uint24.ll