diff options
| author | Koakuma <koachan@protonmail.com> | 2024-01-15 04:28:51 +0700 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-01-14 16:28:51 -0500 |
| commit | 63f98290d09e1da05fb1217d1b760cbe24b76db9 (patch) | |
| tree | ddb7d53229c8c157070315db9882dd746128aefe /polly | |
| parent | 72990df072a56996612169f07c5752a6924288bb (diff) | |
| download | llvm-users/koachan/main.sparc-prefer-rdpc-over-call-to-implement-getpcx-for-64-bit-target.zip llvm-users/koachan/main.sparc-prefer-rdpc-over-call-to-implement-getpcx-for-64-bit-target.tar.gz llvm-users/koachan/main.sparc-prefer-rdpc-over-call-to-implement-getpcx-for-64-bit-target.tar.bz2 | |
[SPARC] Prefer RDPC over CALL to implement GETPCX for 64-bit target (#77196)users/koachan/main.sparc-prefer-rdpc-over-call-to-implement-getpcx-for-64-bit-target
On 64-bit target, prefer usng RDPC over CALL to get the value of %pc.
This is faster on modern processors (Niagara T1 and newer) and avoids polluting
the processor's predictor state.
The old behavior of using a fake CALL is still done when tuning for classic
UltraSPARC processors, since RDPC is much slower there.
A quick pgbench test on a SPARC T4 shows about 2% speedup on SELECT loads,
and about 7% speedup on INSERT/UPDATE loads.
Diffstat (limited to 'polly')
0 files changed, 0 insertions, 0 deletions