Add standalone cost model tests

Author: MacDue

llvm/include/llvm/Analysis/TargetTransformInfo.h

@@ -132,7 +132,7 @@ class IntrinsicCostAttributes {
   IntrinsicCostAttributes(
       Intrinsic::ID Id, const CallBase &CI,
       InstructionCost ScalarCost = InstructionCost::getInvalid(),
-      bool TypeBasedOnly = false);
+      bool TypeBasedOnly = false, TargetLibraryInfo const *LibInfo = nullptr);
 
   IntrinsicCostAttributes(
       Intrinsic::ID Id, Type *RTy, ArrayRef<Type *> Tys,

llvm/lib/Analysis/CostModel.cpp

@@ -17,13 +17,15 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Analysis/CostModel.h"
+#include "llvm/Analysis/TargetLibraryInfo.h"
 #include "llvm/Analysis/TargetTransformInfo.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/PassManager.h"
 #include "llvm/Pass.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/raw_ostream.h"
+
 using namespace llvm;
 
 static cl::opt<TargetTransformInfo::TargetCostKind> CostKind(
@@ -48,19 +50,20 @@ static cl::opt<bool> TypeBasedIntrinsicCost("type-based-intrinsic-cost",
 PreservedAnalyses CostModelPrinterPass::run(Function &F,
                                             FunctionAnalysisManager &AM) {
   auto &TTI = AM.getResult<TargetIRAnalysis>(F);
+  auto &TLI = AM.getResult<TargetLibraryAnalysis>(F);
   OS << "Printing analysis 'Cost Model Analysis' for function '" << F.getName() << "':\n";
   for (BasicBlock &B : F) {
     for (Instruction &Inst : B) {
       // TODO: Use a pass parameter instead of cl::opt CostKind to determine
       // which cost kind to print.
       InstructionCost Cost;
       auto *II = dyn_cast<IntrinsicInst>(&Inst);
-      if (II && TypeBasedIntrinsicCost) {
-        IntrinsicCostAttributes ICA(II->getIntrinsicID(), *II,
-                                    InstructionCost::getInvalid(), true);
+      if (II) {
+        IntrinsicCostAttributes ICA(
+            II->getIntrinsicID(), *II, InstructionCost::getInvalid(),
+            /*TypeBasedOnly=*/TypeBasedIntrinsicCost, &TLI);
         Cost = TTI.getIntrinsicInstrCost(ICA, CostKind);
-      }
-      else {
+      } else {
         Cost = TTI.getInstructionCost(&Inst, CostKind);
       }
 

llvm/lib/Analysis/TargetTransformInfo.cpp

@@ -69,9 +69,9 @@ bool HardwareLoopInfo::canAnalyze(LoopInfo &LI) {
 
 IntrinsicCostAttributes::IntrinsicCostAttributes(
     Intrinsic::ID Id, const CallBase &CI, InstructionCost ScalarizationCost,
-    bool TypeBasedOnly)
+    bool TypeBasedOnly, TargetLibraryInfo const *LibInfo)
     : II(dyn_cast<IntrinsicInst>(&CI)), RetTy(CI.getType()), IID(Id),
-      ScalarizationCost(ScalarizationCost) {
+      ScalarizationCost(ScalarizationCost), LibInfo(LibInfo) {
 
   if (const auto *FPMO = dyn_cast<FPMathOperator>(&CI))
     FMF = FPMO->getFastMathFlags();

llvm/test/Analysis/CostModel/AArch64/sincos.ll

@@ -0,0 +1,67 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --filter "sincos"
+; RUN: opt < %s -mtriple=aarch64-gnu-linux -mattr=+neon,+sve -passes="print<cost-model>" -cost-kind=throughput 2>&1 -disable-output | FileCheck %s
+; RUN: opt < %s -mtriple=aarch64-gnu-linux -mattr=+neon,+sve -vector-library=ArmPL -passes="print<cost-model>" -cost-kind=throughput 2>&1 -disable-output | FileCheck %s -check-prefix=CHECK-VECLIB
+
+define void @sincos(
+; CHECK-LABEL: 'sincos'
+; CHECK:  Cost Model: Found an estimated cost of 1 for instruction: %f16 = call { half, half } @llvm.sincos.f16(half %x_f16)
+; CHECK:  Cost Model: Found an estimated cost of 10 for instruction: %f32 = call { float, float } @llvm.sincos.f32(float %x_f32)
+; CHECK:  Cost Model: Found an estimated cost of 10 for instruction: %f64 = call { double, double } @llvm.sincos.f64(double %x_f64)
+; CHECK:  Cost Model: Found an estimated cost of 10 for instruction: %f128 = call { fp128, fp128 } @llvm.sincos.f128(fp128 %x_f128)
+;
+; CHECK:  Cost Model: Found an estimated cost of 50 for instruction: %v8f16 = call { <8 x half>, <8 x half> } @llvm.sincos.v8f16(<8 x half> %x_v8xf16)
+; CHECK:  Cost Model: Found an estimated cost of 58 for instruction: %v4f32 = call { <4 x float>, <4 x float> } @llvm.sincos.v4f32(<4 x float> %x_v4xf32)
+; CHECK:  Cost Model: Found an estimated cost of 26 for instruction: %v2f64 = call { <2 x double>, <2 x double> } @llvm.sincos.v2f64(<2 x double> %x_v2xf64)
+; CHECK:  Cost Model: Found an estimated cost of 10 for instruction: %v1f128 = call { <1 x fp128>, <1 x fp128> } @llvm.sincos.v1f128(<1 x fp128> %x_v1xf128)
+;
+; CHECK:  Cost Model: Invalid cost for instruction: %nxv8f16 = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.sincos.nxv8f16(<vscale x 8 x half> %x_nxv8xf16)
+; CHECK:  Cost Model: Invalid cost for instruction: %nxv4f32 = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.sincos.nxv4f32(<vscale x 4 x float> %x_nxv4xf32)
+; CHECK:  Cost Model: Invalid cost for instruction: %nxv2f64 = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.sincos.nxv2f64(<vscale x 2 x double> %x_nxv2xf64)
+; CHECK:  Cost Model: Invalid cost for instruction: %nxv1f128 = call { <vscale x 1 x fp128>, <vscale x 1 x fp128> } @llvm.sincos.nxv1f128(<vscale x 1 x fp128> %x_nxv1xf128)
+
+; CHECK-VECLIB-LABEL: 'sincos'
+; CHECK-VECLIB:  Cost Model: Found an estimated cost of 1 for instruction: %f16 = call { half, half } @llvm.sincos.f16(half %x_f16)
+; CHECK-VECLIB:  Cost Model: Found an estimated cost of 10 for instruction: %f32 = call { float, float } @llvm.sincos.f32(float %x_f32)
+; CHECK-VECLIB:  Cost Model: Found an estimated cost of 10 for instruction: %f64 = call { double, double } @llvm.sincos.f64(double %x_f64)
+; CHECK-VECLIB:  Cost Model: Found an estimated cost of 10 for instruction: %f128 = call { fp128, fp128 } @llvm.sincos.f128(fp128 %x_f128)
+;
+; CHECK-VECLIB:  Cost Model: Found an estimated cost of 50 for instruction: %v8f16 = call { <8 x half>, <8 x half> } @llvm.sincos.v8f16(<8 x half> %x_v8xf16)
+; CHECK-VECLIB:  Cost Model: Found an estimated cost of 12 for instruction: %v4f32 = call { <4 x float>, <4 x float> } @llvm.sincos.v4f32(<4 x float> %x_v4xf32)
+; CHECK-VECLIB:  Cost Model: Found an estimated cost of 12 for instruction: %v2f64 = call { <2 x double>, <2 x double> } @llvm.sincos.v2f64(<2 x double> %x_v2xf64)
+; CHECK-VECLIB:  Cost Model: Found an estimated cost of 10 for instruction: %v1f128 = call { <1 x fp128>, <1 x fp128> } @llvm.sincos.v1f128(<1 x fp128> %x_v1xf128)
+;
+; CHECK-VECLIB:  Cost Model: Invalid cost for instruction: %nxv8f16 = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.sincos.nxv8f16(<vscale x 8 x half> %x_nxv8xf16)
+; CHECK-VECLIB:  Cost Model: Found an estimated cost of 13 for instruction: %nxv4f32 = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.sincos.nxv4f32(<vscale x 4 x float> %x_nxv4xf32)
+; CHECK-VECLIB:  Cost Model: Found an estimated cost of 13 for instruction: %nxv2f64 = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.sincos.nxv2f64(<vscale x 2 x double> %x_nxv2xf64)
+; CHECK-VECLIB:  Cost Model: Invalid cost for instruction: %nxv1f128 = call { <vscale x 1 x fp128>, <vscale x 1 x fp128> } @llvm.sincos.nxv1f128(<vscale x 1 x fp128> %x_nxv1xf128)
+
+  half %x_f16,
+  float %x_f32,
+  double %x_f64,
+  fp128 %x_f128,
+  <8 x half> %x_v8xf16,
+  <4 x float> %x_v4xf32,
+  <2 x double> %x_v2xf64,
+  <1 x fp128> %x_v1xf128,
+  <vscale x 8 x half> %x_nxv8xf16,
+  <vscale x 4 x float> %x_nxv4xf32,
+  <vscale x 2 x double> %x_nxv2xf64,
+  <vscale x 1 x fp128> %x_nxv1xf128
+) {
+  %f16 = call { half, half } @llvm.sincos.f16(half %x_f16)
+  %f32 = call { float, float } @llvm.sincos.f32(float %x_f32)
+  %f64 = call { double, double } @llvm.sincos.f64(double %x_f64)
+  %f128 = call { fp128, fp128 } @llvm.sincos.f128(fp128 %x_f128)
+
+  %v8f16 = call { <8 x half>, <8 x half> } @llvm.sincos.v8f16(<8 x half> %x_v8xf16)
+  %v4f32 = call { <4 x float>, <4 x float> } @llvm.sincos.v4f32(<4 x float> %x_v4xf32)
+  %v2f64 = call { <2 x double>, <2 x double> } @llvm.sincos.v2f64(<2 x double> %x_v2xf64)
+  %v1f128 = call { <1 x fp128>, <1 x fp128> } @llvm.sincos.v1f128(<1 x fp128> %x_v1xf128)
+
+  %nxv8f16 = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.sincos.v8f16(<vscale x 8 x half> %x_nxv8xf16)
+  %nxv4f32 = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.sincos.v4f32(<vscale x 4 x float> %x_nxv4xf32)
+  %nxv2f64 = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.sincos.v2f64(<vscale x 2 x double> %x_nxv2xf64)
+  %nxv1f128 = call { <vscale x 1 x fp128>, <vscale x 1 x fp128> } @llvm.sincos.v1f128(<vscale x 1 x fp128> %x_nxv1xf128)
+
+  ret void
+}