[PATCH] D141499: [AArch64] Set MaxInterleaveFactor for Apple A14, A15, A16.

[Florian Hahn via Phabricator via llvm-commits]

fhahn created this revision.
fhahn added reviewers: t.p.northover, ab, jroelofs.
Herald added subscribers: StephenFan, hiraditya, kristof.beyls.
Herald added a project: All.
fhahn requested review of this revision.
Herald added a subscriber: pcwang-thead.
Herald added a project: LLVM.

Those CPUs can benefit from additional interleaving.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D141499

Files:
  llvm/lib/Target/AArch64/AArch64Subtarget.cpp
  llvm/test/Transforms/LoopVectorize/AArch64/interleaving-load-store.ll
  llvm/test/Transforms/LoopVectorize/AArch64/interleaving-reduction.ll


Index: llvm/test/Transforms/LoopVectorize/AArch64/interleaving-reduction.ll
===================================================================
--- llvm/test/Transforms/LoopVectorize/AArch64/interleaving-reduction.ll
+++ llvm/test/Transforms/LoopVectorize/AArch64/interleaving-reduction.ll
@@ -1,9 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=cortex-a57 -S %s | FileCheck --check-prefix=INTERLEAVE-4 %s
 ; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=cortex-a75 -S %s | FileCheck --check-prefix=INTERLEAVE-2 %s
-; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-m1 -S %s | FileCheck --check-prefix=INTERLEAVE-2 %s
-; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-a14 -S %s | FileCheck --check-prefix=INTERLEAVE-2 %s
-; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-a15 -S %s | FileCheck --check-prefix=INTERLEAVE-2 %s
+; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-m1 -S %s | FileCheck --check-prefix=INTERLEAVE-4 %s
+; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-a14 -S %s | FileCheck --check-prefix=INTERLEAVE-4 %s
+; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-a15 -S %s | FileCheck --check-prefix=INTERLEAVE-4 %s
+; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-a16 -S %s | FileCheck --check-prefix=INTERLEAVE-4 %s
 
 ; Tests for selecting the interleave count for loops with reductions.
 
Index: llvm/test/Transforms/LoopVectorize/AArch64/interleaving-load-store.ll
===================================================================
--- llvm/test/Transforms/LoopVectorize/AArch64/interleaving-load-store.ll
+++ llvm/test/Transforms/LoopVectorize/AArch64/interleaving-load-store.ll
@@ -1,9 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=cortex-a57 -S %s | FileCheck --check-prefix=INTERLEAVE-4 %s
 ; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=cortex-a75 -S %s | FileCheck --check-prefix=INTERLEAVE-2 %s
-; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-m1 -S %s | FileCheck --check-prefix=INTERLEAVE-2 %s
-; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-a14 -S %s | FileCheck --check-prefix=INTERLEAVE-2 %s
-; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-a15 -S %s | FileCheck --check-prefix=INTERLEAVE-2 %s
+; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-m1 -S %s | FileCheck --check-prefix=INTERLEAVE-4 %s
+; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-a14 -S %s | FileCheck --check-prefix=INTERLEAVE-4 %s
+; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-a15 -S %s | FileCheck --check-prefix=INTERLEAVE-4 %s
+; RUN: opt -passes=loop-vectorize -mtriple=arm64-apple-macos -mcpu=apple-a16 -S %s | FileCheck --check-prefix=INTERLEAVE-4 %s
 
 ; Tests for selecting interleave counts for loops with loads and stores.
 
Index: llvm/lib/Target/AArch64/AArch64Subtarget.cpp
===================================================================
--- llvm/lib/Target/AArch64/AArch64Subtarget.cpp
+++ llvm/lib/Target/AArch64/AArch64Subtarget.cpp
@@ -172,6 +172,15 @@
     PrefetchDistance = 280;
     MinPrefetchStride = 2048;
     MaxPrefetchIterationsAhead = 3;
+    switch (ARMProcFamily) {
+    case AppleA14:
+    case AppleA15:
+    case AppleA16:
+      MaxInterleaveFactor = 4;
+      break;
+    default:
+      break;
+    }
     break;
   case ExynosM3:
     MaxInterleaveFactor = 4;


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D141499.488209.patch
Type: text/x-patch
Size: 3773 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230111/0c04d4bf/attachment.bin>