diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2019-09-05 22:44:06 +0000 |
|---|---|---|
| committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2019-09-05 22:44:06 +0000 |
| commit | f54daffc2d73866312e6f50b75fe15035e62b4e8 (patch) | |
| tree | 828fbb8a0ade62ef05e8c3176ff5ba31bf2efd1e /llvm/utils/TableGen/CodeGenMapTable.cpp | |
| parent | 3e144ab7fd0fb516b4819504aa423a70cd6732e9 (diff) | |
| download | bcm5719-llvm-f54daffc2d73866312e6f50b75fe15035e62b4e8.tar.gz bcm5719-llvm-f54daffc2d73866312e6f50b75fe15035e62b4e8.zip | |
AMDGPU: Avoid constructing new std::vector in initCandidate
Approximately 30% of the time was spent in the std::vector
constructor. In one testcase this pushes the scheduler to being the
second slowest pass.
I'm not sure I understand why these vector are necessary. The default
scheduler initCandidate seems to use some pre-existing vectors for the
pressure.
llvm-svn: 371136
Diffstat (limited to 'llvm/utils/TableGen/CodeGenMapTable.cpp')
0 files changed, 0 insertions, 0 deletions
