[llvm] [LoongArch][TTI] Add CostModel for @llvm.speculative.load (PR #227551)

[via llvm-commits]

https://github.com/lrzlin created https://github.com/llvm/llvm-project/pull/227551

Add LoongArch TTI support for @llvm.speculative.load, using 4096-byte page boundary as an constraint.

Follows: https://github.com/llvm/llvm-project/pull/180036

>From 1fd73af641df7f2fb5fb619be4e44748e9e1a267 Mon Sep 17 00:00:00 2001
From: Lin Runze <linrunze at loongson.cn>
Date: Wed, 30 Sep 2026 11:58:02 +0800
Subject: [PATCH 1/2] [LoongArch][NFC] Add TTI test for speculative loads

---
 .../CostModel/LoongArch/speculative-load.ll   | 27 +++++++++++++++++++
 1 file changed, 27 insertions(+)
 create mode 100644 llvm/test/Analysis/CostModel/LoongArch/speculative-load.ll

diff --git a/llvm/test/Analysis/CostModel/LoongArch/speculative-load.ll b/llvm/test/Analysis/CostModel/LoongArch/speculative-load.ll
new file mode 100644
index 0000000000000..56ea0e7c66353
--- /dev/null
+++ b/llvm/test/Analysis/CostModel/LoongArch/speculative-load.ll
@@ -0,0 +1,27 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -mtriple=loongarch64 < %s | FileCheck %s
+
+define void @speculative_load_cost(ptr %p) {
+; CHECK-LABEL: 'speculative_load_cost'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = call b8 (ptr, i1, ...) @llvm.speculative.load.b8.p0(ptr %p, i1 false, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = call b16 (ptr, i1, ...) @llvm.speculative.load.b16.p0(ptr %p, i1 false, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = call b32 (ptr, i1, ...) @llvm.speculative.load.b32.p0(ptr %p, i1 false, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = call b64 (ptr, i1, ...) @llvm.speculative.load.b64.p0(ptr %p, i1 false, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %5 = call <4 x i32> (ptr, i1, ...) @llvm.speculative.load.v4i32.p0(ptr %p, i1 false, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %6 = call <8 x i32> (ptr, i1, ...) @llvm.speculative.load.v8i32.p0(ptr %p, i1 false, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = call <2 x i64> (ptr, i1, ...) @llvm.speculative.load.v2i64.p0(ptr %p, i1 false, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = call <4 x float> (ptr, i1, ...) @llvm.speculative.load.v4f32.p0(ptr %p, i1 false, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %9 = call <2 x double> (ptr, i1, ...) @llvm.speculative.load.v2f64.p0(ptr %p, i1 false, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  call b8 (ptr, i1, ...) @llvm.speculative.load.b8.p0(ptr %p, i1 false, i64 0)
+  call b16 (ptr, i1, ...) @llvm.speculative.load.b16.p0(ptr %p, i1 false, i64 0)
+  call b32 (ptr, i1, ...) @llvm.speculative.load.b32.p0(ptr %p, i1 false, i64 0)
+  call b64 (ptr, i1, ...) @llvm.speculative.load.b64.p0(ptr %p, i1 false, i64 0)
+  call <4 x i32> (ptr, i1, ...) @llvm.speculative.load.v4i32.p0(ptr %p, i1 false, i64 0)
+  call <8 x i32> (ptr, i1, ...) @llvm.speculative.load.v8i32.p0(ptr %p, i1 false, i64 0)
+  call <2 x i64> (ptr, i1, ...) @llvm.speculative.load.v2i64.p0(ptr %p, i1 false, i64 0)
+  call <4 x float> (ptr, i1, ...) @llvm.speculative.load.v4f32.p0(ptr %p, i1 false, i64 0)
+  call <2 x double> (ptr, i1, ...) @llvm.speculative.load.v2f64.p0(ptr %p, i1 false, i64 0)
+  ret void
+}

>From c3d3dd60f00f33bb1519fa3c3d9cd9f03a581883 Mon Sep 17 00:00:00 2001
From: Lin Runze <linrunze at loongson.cn>
Date: Wed, 30 Sep 2026 12:04:03 +0800
Subject: [PATCH 2/2] [LoongArch][TTI] Add CostModel for @llvm.speculative.load

---
 .../Target/LoongArch/LoongArchTargetTransformInfo.cpp | 11 +++++++++++
 .../Target/LoongArch/LoongArchTargetTransformInfo.h   |  3 +++
 2 files changed, 14 insertions(+)

diff --git a/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.cpp b/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.cpp
index 6646841b93215..49d89c3addcc2 100644
--- a/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.cpp
@@ -113,6 +113,17 @@ bool LoongArchTTIImpl::shouldExpandReduction(const IntrinsicInst *II) const {
   }
 }
 
+bool LoongArchTTIImpl::isLegalSpeculativeLoad(Type *DataType,
+                                            unsigned AddressSpace) const {
+  // Matches LoongArchTargetLowering::emitCanLoadSpeculatively: only address
+  // space 0 and power-of-2 sizes up to the 4096-byte size page boundary.
+  if (AddressSpace != 0)
+    return false;
+  TypeSize Size = DL.getTypeStoreSize(DataType);
+  return !Size.isScalable() && isPowerOf2_64(Size.getFixedValue()) &&
+         Size.getFixedValue() <= 4096;
+}
+
 LoongArchTTIImpl::TTI::MemCmpExpansionOptions
 LoongArchTTIImpl::enableMemCmpExpansion(bool OptSize, bool IsZeroCmp) const {
   TTI::MemCmpExpansionOptions Options;
diff --git a/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.h b/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.h
index 2392b709767ee..28cbc8b113e02 100644
--- a/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.h
+++ b/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.h
@@ -56,6 +56,9 @@ class LoongArchTTIImpl : public BasicTTIImplBase<LoongArchTTIImpl> {
 
   bool shouldExpandReduction(const IntrinsicInst *II) const override;
 
+  bool isLegalSpeculativeLoad(Type *DataType,
+                              unsigned AddressSpace) const override;
+
   TTI::MemCmpExpansionOptions
   enableMemCmpExpansion(bool OptSize, bool IsZeroCmp) const override;