Add @llvm.clear_cache builtin
Implementing the LLVM part of the call to __builtin___clear_cache
which translates into an intrinsic @llvm.clear_cache and is lowered
by each target, either to a call to __clear_cache or nothing at all
incase the caches are unified.
Updating LangRef and adding some tests for the implemented architectures.
Other archs will have to implement the method in case this builtin
has to be compiled for it, since the default behaviour is to bail
unimplemented.
A Clang patch is required for the builtin to be lowered into the
llvm intrinsic. This will be done next.
llvm-svn: 204802
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index c959f56..d18d4eb 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -6939,6 +6939,36 @@
Note that runtime support may be conditional on the privilege-level code is
running at and the host platform.
+'``llvm.clear_cache``' Intrinsic
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Syntax:
+"""""""
+
+::
+
+ declare void @llvm.clear_cache(i8*, i8*)
+
+Overview:
+"""""""""
+
+The '``llvm.clear_cache``' intrinsic provides access to the systemcall
+that flushes the cache at the range specified. Some targets need this
+to specifically flush the instruction cache when executable data changes
+in memory (self-modifying code). Other targets have unified intruction
+and data cache, so they don't need any calls.
+
+Semantics:
+""""""""""
+
+When directly supported, this intrinsic will either return a call to
+the appropriate cache clearing system call (usually ``__clear_cache``)
+when the caches are not unified (ARM, Mips) or just remove the call
+altogether when they are (ex. x86_64).
+
+Targets must implement it directly to have either behaviour, as the
+default is to bail with "Not Implemented" message.
+
Standard C Library Intrinsics
-----------------------------
diff --git a/llvm/include/llvm/IR/Intrinsics.td b/llvm/include/llvm/IR/Intrinsics.td
index 94c7b50..0c9cc3b 100644
--- a/llvm/include/llvm/IR/Intrinsics.td
+++ b/llvm/include/llvm/IR/Intrinsics.td
@@ -513,6 +513,11 @@
def int_convertuu : Intrinsic<[llvm_anyint_ty],
[llvm_anyint_ty, llvm_i32_ty, llvm_i32_ty]>;
+// Clear cache intrinsic, default to ignore (ie. emit nothing)
+// maps to void __clear_cache() on supporting platforms
+def int_clear_cache : Intrinsic<[], [llvm_ptr_ty, llvm_ptr_ty],
+ [], "llvm.clear_cache">;
+
//===----------------------------------------------------------------------===//
// Target-specific intrinsics
//===----------------------------------------------------------------------===//
diff --git a/llvm/include/llvm/Target/TargetLowering.h b/llvm/include/llvm/Target/TargetLowering.h
index 8f00ac8f..fbc8b74 100644
--- a/llvm/include/llvm/Target/TargetLowering.h
+++ b/llvm/include/llvm/Target/TargetLowering.h
@@ -2108,6 +2108,11 @@
return false;
}
+ /// Return the builtin name for the __builtin___clear_cache intrinsic
+ virtual const char * getClearCacheBuiltinName() const {
+ llvm_unreachable("Not Implemented");
+ }
+
/// Return the type that should be used to zero or sign extend a
/// zeroext/signext integer argument or return value. FIXME: Most C calling
/// convention requires the return type to be promoted, but this is not true
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index bd80a50..639ff82 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -5386,6 +5386,8 @@
(void)getControlRoot();
return 0;
}
+ case Intrinsic::clear_cache:
+ return TLI->getClearCacheBuiltinName();
case Intrinsic::donothing:
// ignore
return 0;
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.h b/llvm/lib/Target/ARM/ARMISelLowering.h
index 022945f..5f6ea75 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.h
+++ b/llvm/lib/Target/ARM/ARMISelLowering.h
@@ -384,6 +384,11 @@
bool shouldConvertConstantLoadToIntImm(const APInt &Imm,
Type *Ty) const override;
+ /// Clear cache library call
+ const char * getClearCacheBuiltinName() const {
+ return "__clear_cache";
+ }
+
protected:
std::pair<const TargetRegisterClass*, uint8_t>
findRepresentativeClass(MVT VT) const override;
diff --git a/llvm/lib/Target/Mips/MipsISelLowering.h b/llvm/lib/Target/Mips/MipsISelLowering.h
index 27492a8..b03cccf 100644
--- a/llvm/lib/Target/Mips/MipsISelLowering.h
+++ b/llvm/lib/Target/Mips/MipsISelLowering.h
@@ -583,6 +583,11 @@
bool MemcpyStrSrc,
MachineFunction &MF) const;
+ /// Clear cache library call
+ const char * getClearCacheBuiltinName() const {
+ return "__clear_cache";
+ }
+
/// isFPImmLegal - Returns true if the target can instruction select the
/// specified FP immediate natively. If false, the legalizer will
/// materialize the FP immediate as a load from a constant pool.
diff --git a/llvm/lib/Target/X86/X86ISelLowering.h b/llvm/lib/Target/X86/X86ISelLowering.h
index 894cfc6..ae13ca4 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.h
+++ b/llvm/lib/Target/X86/X86ISelLowering.h
@@ -770,6 +770,11 @@
bool shouldConvertConstantLoadToIntImm(const APInt &Imm,
Type *Ty) const override;
+ /// Intel processors have a unified instruction and data cache
+ const char * getClearCacheBuiltinName() const {
+ return 0; // nothing to do, move along.
+ }
+
/// createFastISel - This method returns a target specific FastISel object,
/// or null if the target does not support "fast" ISel.
FastISel *createFastISel(FunctionLoweringInfo &funcInfo,
diff --git a/llvm/test/CodeGen/ARM/cache-intrinsic.ll b/llvm/test/CodeGen/ARM/cache-intrinsic.ll
new file mode 100644
index 0000000..6048917
--- /dev/null
+++ b/llvm/test/CodeGen/ARM/cache-intrinsic.ll
@@ -0,0 +1,26 @@
+; RUN: llc %s -o - | FileCheck %s
+target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-n32-S64"
+target triple = "armv7--linux-gnueabihf"
+
+@buffer = global [32 x i8] c"This is a largely unused buffer\00", align 1
+@.str = private unnamed_addr constant [4 x i8] c"%s\0A\00", align 1
+@.str1 = private unnamed_addr constant [25 x i8] c"Still, largely unused...\00", align 1
+
+define i32 @main() {
+entry:
+ %retval = alloca i32, align 4
+ store i32 0, i32* %retval
+ %call = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([4 x i8]* @.str, i32 0, i32 0), i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0))
+ %call1 = call i8* @strcpy(i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0), i8* getelementptr inbounds ([25 x i8]* @.str1, i32 0, i32 0)) #3
+ call void @llvm.clear_cache(i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0), i8* getelementptr inbounds (i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0), i32 32)) #3
+ %call2 = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([4 x i8]* @.str, i32 0, i32 0), i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0))
+ ret i32 0
+}
+
+; CHECK: __clear_cache
+
+declare i32 @printf(i8*, ...)
+
+declare i8* @strcpy(i8*, i8*)
+
+declare void @llvm.clear_cache(i8*, i8*)
diff --git a/llvm/test/CodeGen/Mips/cache-intrinsic.ll b/llvm/test/CodeGen/Mips/cache-intrinsic.ll
new file mode 100644
index 0000000..2fa4115
--- /dev/null
+++ b/llvm/test/CodeGen/Mips/cache-intrinsic.ll
@@ -0,0 +1,26 @@
+; RUN: llc %s -o - | FileCheck %s
+target datalayout = "E-p:32:32:32-i1:8:8-i8:8:32-i16:16:32-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-n32-S64"
+target triple = "mips--linux-gnu"
+
+@buffer = global [32 x i8] c"This is a largely unused buffer\00", align 1
+@.str = private unnamed_addr constant [4 x i8] c"%s\0A\00", align 1
+@.str1 = private unnamed_addr constant [25 x i8] c"Still, largely unused...\00", align 1
+
+define i32 @main() {
+entry:
+ %retval = alloca i32, align 4
+ store i32 0, i32* %retval
+ %call = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([4 x i8]* @.str, i32 0, i32 0), i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0))
+ %call1 = call i8* @strcpy(i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0), i8* getelementptr inbounds ([25 x i8]* @.str1, i32 0, i32 0)) #3
+ call void @llvm.clear_cache(i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0), i8* getelementptr inbounds (i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0), i32 32)) #3
+ %call2 = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([4 x i8]* @.str, i32 0, i32 0), i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0))
+ ret i32 0
+}
+
+; CHECK: __clear_cache
+
+declare i32 @printf(i8*, ...)
+
+declare i8* @strcpy(i8*, i8*)
+
+declare void @llvm.clear_cache(i8*, i8*)
diff --git a/llvm/test/CodeGen/X86/cache-intrinsic.ll b/llvm/test/CodeGen/X86/cache-intrinsic.ll
new file mode 100644
index 0000000..3091b5f
--- /dev/null
+++ b/llvm/test/CodeGen/X86/cache-intrinsic.ll
@@ -0,0 +1,26 @@
+; RUN: llc %s -o - | FileCheck %s
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+@buffer = global [32 x i8] c"This is a largely unused buffer\00", align 16
+@.str = private unnamed_addr constant [4 x i8] c"%s\0A\00", align 1
+@.str1 = private unnamed_addr constant [25 x i8] c"Still, largely unused...\00", align 1
+
+define i32 @main() {
+entry:
+ %retval = alloca i32, align 4
+ store i32 0, i32* %retval
+ %call = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([4 x i8]* @.str, i32 0, i32 0), i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0))
+ %call1 = call i8* @strcpy(i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0), i8* getelementptr inbounds ([25 x i8]* @.str1, i32 0, i32 0)) #3
+ call void @llvm.clear_cache(i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0), i8* getelementptr inbounds (i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0), i32 32)) #3
+ %call3 = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([4 x i8]* @.str, i32 0, i32 0), i8* getelementptr inbounds ([32 x i8]* @buffer, i32 0, i32 0))
+ ret i32 0
+}
+
+; CHECK-NOT: __clear_cache
+
+declare i32 @printf(i8*, ...)
+
+declare i8* @strcpy(i8*, i8*)
+
+declare void @llvm.clear_cache(i8*, i8*)