ART: Add entrypoint and intrinsic for Math.pow().

MathBenchmarks.java#timePow results on taimen's little cores
fixed at frequency 1401600 with forced JIT compilation:
  - before:
    - X32: 356.33 (@FastNative), 315.39 (@CriticalNative)
    - X64: 357.31 (@FastNative), 315.37 (@CriticalNative)
  - after (LICM defeats the benchmark):
    - X32: 2.88
    - X64: 2.87
  - after but with kAllSideEffects to prevent LICM:
    - X32: 275.42
    - X64: 275.67

Test: Rely on TreeHugger.
Bug: 70727450
Change-Id: Iaa31f70acabbd57c163cfeafe02eed67c1348861

21 files changed, 68 insertions, 7 deletions

diff --git a/compiler/optimizing/intrinsics_arm64.cc b/compiler/optimizing/intrinsics_arm64.cc
index ca1b451e6b..2f8e33f941 100644
--- a/compiler/optimizing/intrinsics_arm64.cc
+++ b/compiler/optimizing/intrinsics_arm64.cc
@@ -2011,6 +2011,14 @@ void IntrinsicCodeGeneratorARM64::VisitMathAtan2(HInvoke* invoke) {
   GenFPToFPCall(invoke, codegen_, kQuickAtan2);
 }
 
+void IntrinsicLocationsBuilderARM64::VisitMathPow(HInvoke* invoke) {
+  CreateFPFPToFPCallLocations(allocator_, invoke);
+}
+
+void IntrinsicCodeGeneratorARM64::VisitMathPow(HInvoke* invoke) {
+  GenFPToFPCall(invoke, codegen_, kQuickPow);
+}
+
 void IntrinsicLocationsBuilderARM64::VisitMathHypot(HInvoke* invoke) {
   CreateFPFPToFPCallLocations(allocator_, invoke);
 }
diff --git a/compiler/optimizing/intrinsics_arm_vixl.cc b/compiler/optimizing/intrinsics_arm_vixl.cc
index 99b8b5df74..830d0403e4 100644
--- a/compiler/optimizing/intrinsics_arm_vixl.cc
+++ b/compiler/optimizing/intrinsics_arm_vixl.cc
@@ -2811,6 +2811,14 @@ void IntrinsicCodeGeneratorARMVIXL::VisitMathAtan2(HInvoke* invoke) {
   GenFPFPToFPCall(invoke, GetAssembler(), codegen_, kQuickAtan2);
 }
 
+void IntrinsicLocationsBuilderARMVIXL::VisitMathPow(HInvoke* invoke) {
+  CreateFPFPToFPCallLocations(allocator_, invoke);
+}
+
+void IntrinsicCodeGeneratorARMVIXL::VisitMathPow(HInvoke* invoke) {
+  GenFPFPToFPCall(invoke, GetAssembler(), codegen_, kQuickPow);
+}
+
 void IntrinsicLocationsBuilderARMVIXL::VisitMathHypot(HInvoke* invoke) {
   CreateFPFPToFPCallLocations(allocator_, invoke);
 }
diff --git a/compiler/optimizing/intrinsics_mips.cc b/compiler/optimizing/intrinsics_mips.cc
index 113c9de5a2..cafa5228d9 100644
--- a/compiler/optimizing/intrinsics_mips.cc
+++ b/compiler/optimizing/intrinsics_mips.cc
@@ -2835,6 +2835,15 @@ void IntrinsicCodeGeneratorMIPS::VisitMathAtan2(HInvoke* invoke) {
   GenFPFPToFPCall(invoke, codegen_, kQuickAtan2);
 }
 
+// static double java.lang.Math.pow(double y, double x)
+void IntrinsicLocationsBuilderMIPS::VisitMathPow(HInvoke* invoke) {
+  CreateFPFPToFPCallLocations(allocator_, invoke);
+}
+
+void IntrinsicCodeGeneratorMIPS::VisitMathPow(HInvoke* invoke) {
+  GenFPFPToFPCall(invoke, codegen_, kQuickPow);
+}
+
 // static double java.lang.Math.cbrt(double a)
 void IntrinsicLocationsBuilderMIPS::VisitMathCbrt(HInvoke* invoke) {
   CreateFPToFPCallLocations(allocator_, invoke);
diff --git a/compiler/optimizing/intrinsics_mips64.cc b/compiler/optimizing/intrinsics_mips64.cc
index 521bad27e2..89f1818be2 100644
--- a/compiler/optimizing/intrinsics_mips64.cc
+++ b/compiler/optimizing/intrinsics_mips64.cc
@@ -2416,6 +2416,15 @@ void IntrinsicCodeGeneratorMIPS64::VisitMathAtan2(HInvoke* invoke) {
   GenFPFPToFPCall(invoke, codegen_, kQuickAtan2);
 }
 
+// static double java.lang.Math.pow(double y, double x)
+void IntrinsicLocationsBuilderMIPS64::VisitMathPow(HInvoke* invoke) {
+  CreateFPFPToFPCallLocations(allocator_, invoke);
+}
+
+void IntrinsicCodeGeneratorMIPS64::VisitMathPow(HInvoke* invoke) {
+  GenFPFPToFPCall(invoke, codegen_, kQuickPow);
+}
+
 // static double java.lang.Math.cbrt(double a)
 void IntrinsicLocationsBuilderMIPS64::VisitMathCbrt(HInvoke* invoke) {
   CreateFPToFPCallLocations(allocator_, invoke);
diff --git a/compiler/optimizing/intrinsics_x86.cc b/compiler/optimizing/intrinsics_x86.cc
index baa410b884..46b7f3f1ce 100644
--- a/compiler/optimizing/intrinsics_x86.cc
+++ b/compiler/optimizing/intrinsics_x86.cc
@@ -1105,6 +1105,14 @@ void IntrinsicCodeGeneratorX86::VisitMathAtan2(HInvoke* invoke) {
   GenFPToFPCall(invoke, codegen_, kQuickAtan2);
 }
 
+void IntrinsicLocationsBuilderX86::VisitMathPow(HInvoke* invoke) {
+  CreateFPFPToFPCallLocations(allocator_, invoke);
+}
+
+void IntrinsicCodeGeneratorX86::VisitMathPow(HInvoke* invoke) {
+  GenFPToFPCall(invoke, codegen_, kQuickPow);
+}
+
 void IntrinsicLocationsBuilderX86::VisitMathHypot(HInvoke* invoke) {
   CreateFPFPToFPCallLocations(allocator_, invoke);
 }
diff --git a/compiler/optimizing/intrinsics_x86_64.cc b/compiler/optimizing/intrinsics_x86_64.cc
index 6dd8b8e1f5..6483b7cb2a 100644
--- a/compiler/optimizing/intrinsics_x86_64.cc
+++ b/compiler/optimizing/intrinsics_x86_64.cc
@@ -897,6 +897,14 @@ void IntrinsicCodeGeneratorX86_64::VisitMathAtan2(HInvoke* invoke) {
   GenFPToFPCall(invoke, codegen_, kQuickAtan2);
 }
 
+void IntrinsicLocationsBuilderX86_64::VisitMathPow(HInvoke* invoke) {
+  CreateFPFPToFPCallLocations(allocator_, invoke);
+}
+
+void IntrinsicCodeGeneratorX86_64::VisitMathPow(HInvoke* invoke) {
+  GenFPToFPCall(invoke, codegen_, kQuickPow);
+}
+
 void IntrinsicLocationsBuilderX86_64::VisitMathHypot(HInvoke* invoke) {
   CreateFPFPToFPCallLocations(allocator_, invoke);
 }
diff --git a/compiler/utils/assembler_thumb_test_expected.cc.inc b/compiler/utils/assembler_thumb_test_expected.cc.inc
index 0a094352e4..674dc9a78b 100644
--- a/compiler/utils/assembler_thumb_test_expected.cc.inc
+++ b/compiler/utils/assembler_thumb_test_expected.cc.inc
@@ -153,7 +153,7 @@ const char* const VixlJniHelpersResults[] = {
   " 21c:	f8d9 8034 	ldr.w	r8, [r9, #52]	; 0x34\n",
   " 220:	4770      	bx	lr\n",
   " 222:	4660      	mov	r0, ip\n",
-  " 224:	f8d9 c2c0 	ldr.w	ip, [r9, #704]	; 0x2c0\n",
+  " 224:	f8d9 c2c4 	ldr.w	ip, [r9, #708]	; 0x2c4\n",
   " 228:	47e0      	blx	ip\n",
   nullptr
 };
diff --git a/dex2oat/linker/oat_writer_test.cc b/dex2oat/linker/oat_writer_test.cc
index c45166b17e..d5a87837f6 100644
--- a/dex2oat/linker/oat_writer_test.cc
+++ b/dex2oat/linker/oat_writer_test.cc
@@ -485,7 +485,7 @@ TEST_F(OatTest, OatHeaderSizeCheck) {
   EXPECT_EQ(76U, sizeof(OatHeader));
   EXPECT_EQ(4U, sizeof(OatMethodOffsets));
   EXPECT_EQ(24U, sizeof(OatQuickMethodHeader));
-  EXPECT_EQ(161 * static_cast<size_t>(GetInstructionSetPointerSize(kRuntimeISA)),
+  EXPECT_EQ(162 * static_cast<size_t>(GetInstructionSetPointerSize(kRuntimeISA)),
             sizeof(QuickEntryPoints));
 }
 
diff --git a/runtime/arch/arm/entrypoints_init_arm.cc b/runtime/arch/arm/entrypoints_init_arm.cc
index 78b9e46d77..80080e9832 100644
--- a/runtime/arch/arm/entrypoints_init_arm.cc
+++ b/runtime/arch/arm/entrypoints_init_arm.cc
@@ -144,6 +144,7 @@ void InitEntryPoints(JniEntryPoints* jpoints, QuickEntryPoints* qpoints) {
   qpoints->pAsin = asin;
   qpoints->pAtan = atan;
   qpoints->pAtan2 = atan2;
+  qpoints->pPow = pow;
   qpoints->pCbrt = cbrt;
   qpoints->pCosh = cosh;
   qpoints->pExp = exp;
diff --git a/runtime/arch/arm64/entrypoints_init_arm64.cc b/runtime/arch/arm64/entrypoints_init_arm64.cc
index 80bf3abc6f..4c43b7ed3d 100644
--- a/runtime/arch/arm64/entrypoints_init_arm64.cc
+++ b/runtime/arch/arm64/entrypoints_init_arm64.cc
@@ -168,6 +168,7 @@ void InitEntryPoints(JniEntryPoints* jpoints, QuickEntryPoints* qpoints) {
   qpoints->pAsin = asin;
   qpoints->pAtan = atan;
   qpoints->pAtan2 = atan2;
+  qpoints->pPow = pow;
   qpoints->pCbrt = cbrt;
   qpoints->pCosh = cosh;
   qpoints->pExp = exp;
diff --git a/runtime/arch/mips/entrypoints_init_mips.cc b/runtime/arch/mips/entrypoints_init_mips.cc
index 209f36705a..badee59568 100644
--- a/runtime/arch/mips/entrypoints_init_mips.cc
+++ b/runtime/arch/mips/entrypoints_init_mips.cc
@@ -348,6 +348,8 @@ void InitEntryPoints(JniEntryPoints* jpoints, QuickEntryPoints* qpoints) {
   static_assert(IsDirectEntrypoint(kQuickAtan), "Direct C stub marked non-direct.");
   qpoints->pAtan2 = atan2;
   static_assert(IsDirectEntrypoint(kQuickAtan2), "Direct C stub marked non-direct.");
+  qpoints->pPow = pow;
+  static_assert(IsDirectEntrypoint(kQuickPow), "Direct C stub marked non-direct.");
   qpoints->pCbrt = cbrt;
   static_assert(IsDirectEntrypoint(kQuickCbrt), "Direct C stub marked non-direct.");
   qpoints->pCosh = cosh;
diff --git a/runtime/arch/mips64/entrypoints_init_mips64.cc b/runtime/arch/mips64/entrypoints_init_mips64.cc
index 35cbd1dcc0..bdfb9421df 100644
--- a/runtime/arch/mips64/entrypoints_init_mips64.cc
+++ b/runtime/arch/mips64/entrypoints_init_mips64.cc
@@ -165,6 +165,7 @@ void InitEntryPoints(JniEntryPoints* jpoints, QuickEntryPoints* qpoints) {
   qpoints->pAsin = asin;
   qpoints->pAtan = atan;
   qpoints->pAtan2 = atan2;
+  qpoints->pPow = pow;
   qpoints->pCbrt = cbrt;
   qpoints->pCosh = cosh;
   qpoints->pExp = exp;
diff --git a/runtime/arch/x86/entrypoints_init_x86.cc b/runtime/arch/x86/entrypoints_init_x86.cc
index 24bf9cc07c..ffb0c94cc7 100644
--- a/runtime/arch/x86/entrypoints_init_x86.cc
+++ b/runtime/arch/x86/entrypoints_init_x86.cc
@@ -68,6 +68,7 @@ void InitEntryPoints(JniEntryPoints* jpoints, QuickEntryPoints* qpoints) {
   qpoints->pAsin = asin;
   qpoints->pAtan = atan;
   qpoints->pAtan2 = atan2;
+  qpoints->pPow = pow;
   qpoints->pCbrt = cbrt;
   qpoints->pCosh = cosh;
   qpoints->pExp = exp;
diff --git a/runtime/arch/x86_64/entrypoints_init_x86_64.cc b/runtime/arch/x86_64/entrypoints_init_x86_64.cc
index 3656f83b58..6bae69c495 100644
--- a/runtime/arch/x86_64/entrypoints_init_x86_64.cc
+++ b/runtime/arch/x86_64/entrypoints_init_x86_64.cc
@@ -91,6 +91,7 @@ void InitEntryPoints(JniEntryPoints* jpoints, QuickEntryPoints* qpoints) {
   qpoints->pAsin = asin;
   qpoints->pAtan = atan;
   qpoints->pAtan2 = atan2;
+  qpoints->pPow = pow;
   qpoints->pCbrt = cbrt;
   qpoints->pCosh = cosh;
   qpoints->pExp = exp;
diff --git a/runtime/asm_support.h b/runtime/asm_support.h
index 3cf2b93690..2f7d6ab98f 100644
--- a/runtime/asm_support.h
+++ b/runtime/asm_support.h
@@ -73,7 +73,7 @@ ADD_TEST_EQ(THREAD_LOCAL_OBJECTS_OFFSET,
 
 // Offset of field Thread::tlsPtr_.mterp_current_ibase.
 #define THREAD_CURRENT_IBASE_OFFSET \
-    (THREAD_LOCAL_OBJECTS_OFFSET + __SIZEOF_SIZE_T__ + (1 + 161) * __SIZEOF_POINTER__)
+    (THREAD_LOCAL_OBJECTS_OFFSET + __SIZEOF_SIZE_T__ + (1 + 162) * __SIZEOF_POINTER__)
 ADD_TEST_EQ(THREAD_CURRENT_IBASE_OFFSET,
             art::Thread::MterpCurrentIBaseOffset<POINTER_SIZE>().Int32Value())
 // Offset of field Thread::tlsPtr_.mterp_default_ibase.
diff --git a/runtime/entrypoints/quick/quick_entrypoints_list.h b/runtime/entrypoints/quick/quick_entrypoints_list.h
index 74e7c180b8..48a56f2fbf 100644
--- a/runtime/entrypoints/quick/quick_entrypoints_list.h
+++ b/runtime/entrypoints/quick/quick_entrypoints_list.h
@@ -91,6 +91,7 @@
   V(Asin, double, double) \
   V(Atan, double, double) \
   V(Atan2, double, double, double) \
+  V(Pow, double, double, double) \
   V(Cbrt, double, double) \
   V(Cosh, double, double) \
   V(Exp, double, double) \
diff --git a/runtime/entrypoints_order_test.cc b/runtime/entrypoints_order_test.cc
index 7c912d0a4a..1fdf439d3f 100644
--- a/runtime/entrypoints_order_test.cc
+++ b/runtime/entrypoints_order_test.cc
@@ -238,7 +238,8 @@ class EntrypointsOrderTest : public CommonRuntimeTest {
     EXPECT_OFFSET_DIFFNP(QuickEntryPoints, pAcos, pAsin, sizeof(void*));
     EXPECT_OFFSET_DIFFNP(QuickEntryPoints, pAsin, pAtan, sizeof(void*));
     EXPECT_OFFSET_DIFFNP(QuickEntryPoints, pAtan, pAtan2, sizeof(void*));
-    EXPECT_OFFSET_DIFFNP(QuickEntryPoints, pAtan2, pCbrt, sizeof(void*));
+    EXPECT_OFFSET_DIFFNP(QuickEntryPoints, pAtan2, pPow, sizeof(void*));
+    EXPECT_OFFSET_DIFFNP(QuickEntryPoints, pPow, pCbrt, sizeof(void*));
     EXPECT_OFFSET_DIFFNP(QuickEntryPoints, pCbrt, pCosh, sizeof(void*));
     EXPECT_OFFSET_DIFFNP(QuickEntryPoints, pCosh, pExp, sizeof(void*));
     EXPECT_OFFSET_DIFFNP(QuickEntryPoints, pExp, pExpm1, sizeof(void*));
diff --git a/runtime/image.cc b/runtime/image.cc
index dd0c1487c0..8e3615ffcf 100644
--- a/runtime/image.cc
+++ b/runtime/image.cc
@@ -26,7 +26,7 @@
 namespace art {
 
 const uint8_t ImageHeader::kImageMagic[] = { 'a', 'r', 't', '\n' };
-const uint8_t ImageHeader::kImageVersion[] = { '0', '5', '3', '\0' };  // ClassStatus in high bits.
+const uint8_t ImageHeader::kImageVersion[] = { '0', '5', '4', '\0' };  // Math.pow() intrinsic.
 
 ImageHeader::ImageHeader(uint32_t image_begin,
                          uint32_t image_size,
diff --git a/runtime/interpreter/interpreter_intrinsics.cc b/runtime/interpreter/interpreter_intrinsics.cc
index 99a4f763c9..681a582b5d 100644
--- a/runtime/interpreter/interpreter_intrinsics.cc
+++ b/runtime/interpreter/interpreter_intrinsics.cc
@@ -478,6 +478,7 @@ bool MterpHandleIntrinsic(ShadowFrame* shadow_frame,
     UNIMPLEMENTED_CASE(MathLog /* (D)D */)
     UNIMPLEMENTED_CASE(MathLog10 /* (D)D */)
     UNIMPLEMENTED_CASE(MathNextAfter /* (DD)D */)
+    UNIMPLEMENTED_CASE(MathPow /* (DD)D */)
     UNIMPLEMENTED_CASE(MathSinh /* (D)D */)
     INTRINSIC_CASE(MathTan)
     UNIMPLEMENTED_CASE(MathTanh /* (D)D */)
diff --git a/runtime/intrinsics_list.h b/runtime/intrinsics_list.h
index d007728750..da08793f59 100644
--- a/runtime/intrinsics_list.h
+++ b/runtime/intrinsics_list.h
@@ -136,6 +136,7 @@
   V(MathAsin, kStatic, kNeedsEnvironmentOrCache, kNoSideEffects, kNoThrow, "Ljava/lang/Math;", "asin", "(D)D") \
   V(MathAtan, kStatic, kNeedsEnvironmentOrCache, kNoSideEffects, kNoThrow, "Ljava/lang/Math;", "atan", "(D)D") \
   V(MathAtan2, kStatic, kNeedsEnvironmentOrCache, kNoSideEffects, kNoThrow, "Ljava/lang/Math;", "atan2", "(DD)D") \
+  V(MathPow, kStatic, kNeedsEnvironmentOrCache, kNoSideEffects, kNoThrow, "Ljava/lang/Math;", "pow", "(DD)D") \
   V(MathCbrt, kStatic, kNeedsEnvironmentOrCache, kNoSideEffects, kNoThrow, "Ljava/lang/Math;", "cbrt", "(D)D") \
   V(MathCosh, kStatic, kNeedsEnvironmentOrCache, kNoSideEffects, kNoThrow, "Ljava/lang/Math;", "cosh", "(D)D") \
   V(MathExp, kStatic, kNeedsEnvironmentOrCache, kNoSideEffects, kNoThrow, "Ljava/lang/Math;", "exp", "(D)D") \
diff --git a/runtime/oat.h b/runtime/oat.h
index 36099b93dc..8f81010a06 100644
--- a/runtime/oat.h
+++ b/runtime/oat.h
@@ -32,8 +32,8 @@ class InstructionSetFeatures;
 class PACKED(4) OatHeader {
  public:
   static constexpr uint8_t kOatMagic[] = { 'o', 'a', 't', '\n' };
-  // Last oat version changed reason: ClassStatus in high bits.
-  static constexpr uint8_t kOatVersion[] = { '1', '3', '7', '\0' };
+  // Last oat version changed reason: Math.pow() intrinsic.
+  static constexpr uint8_t kOatVersion[] = { '1', '3', '8', '\0' };
 
   static constexpr const char* kImageLocationKey = "image-location";
   static constexpr const char* kDex2OatCmdLineKey = "dex2oat-cmdline";