diff options
| author | Joseph Huber <huberjn@outlook.com> | 2024-02-01 07:19:57 -0600 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-02-01 07:19:57 -0600 |
| commit | f956e7fbf161447b9236f7c4448a9d02d3564261 (patch) | |
| tree | 75efdcc8e95d0be467d7976dea21b448c0a461f1 /clang/lib/CodeGen/CodeGenModule.cpp | |
| parent | d2565bb11308f6cf98d838e828d9bcbe2d51e0e4 (diff) | |
| download | llvm-f956e7fbf161447b9236f7c4448a9d02d3564261.zip llvm-f956e7fbf161447b9236f7c4448a9d02d3564261.tar.gz llvm-f956e7fbf161447b9236f7c4448a9d02d3564261.tar.bz2 | |
[AMDGPU] Prefer `s_memtime` for `readcyclecounter` on GFX10 (#80211)
Summary:
The old `s_memtime` instruction was supported until the GFX10
architecture. Although this instruction has a higher latency than the
new shader counter, it's much more usable as a processor clock as it is
a full 64-bit counter. The new shader counter is only a 20-bit counter,
which makes it difficult to use as a standard cycle counter as it will
overflow in a few milliseconds. This patch suggests preferring
`s_memtime` for this instrinsic if it is still available.
Diffstat (limited to 'clang/lib/CodeGen/CodeGenModule.cpp')
0 files changed, 0 insertions, 0 deletions