[llvm] a0df86d - [RISCV] Add TuneNoDefaultUnroll to spacemit-x100 (#209069)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Jul 13 03:28:26 PDT 2026
Author: Mark Zhuang
Date: 2026-07-13T18:28:21+08:00
New Revision: a0df86deb4295ff9ebebe54eb515bc49a7fb65f8
URL: https://github.com/llvm/llvm-project/commit/a0df86deb4295ff9ebebe54eb515bc49a7fb65f8
DIFF: https://github.com/llvm/llvm-project/commit/a0df86deb4295ff9ebebe54eb515bc49a7fb65f8.diff
LOG: [RISCV] Add TuneNoDefaultUnroll to spacemit-x100 (#209069)
In the spec2k6int tests, some subtests improved while others
regressed. The overall score shows a 0.7% gain. Given that
most OoO CPUs enable this feature by default, also enable
it by default for X100.
Added:
Modified:
llvm/lib/Target/RISCV/RISCVProcessors.td
Removed:
################################################################################
diff --git a/llvm/lib/Target/RISCV/RISCVProcessors.td b/llvm/lib/Target/RISCV/RISCVProcessors.td
index daf7bcd0aad82..a006cea4c3157 100644
--- a/llvm/lib/Target/RISCV/RISCVProcessors.td
+++ b/llvm/lib/Target/RISCV/RISCVProcessors.td
@@ -938,6 +938,7 @@ def SPACEMIT_X100 : RISCVProcessorModel<"spacemit-x100",
TuneOptimizedNF3SegmentLoadStore,
TuneOptimizedNF4SegmentLoadStore,
TuneVXRMPipelineFlush,
+ TuneNoDefaultUnroll,
TuneFusionLogicRegReg,
TuneFusionLogicRegImm,
TuneFusionLogicImmReg,
More information about the llvm-commits
mailing list