Add builtins for bitreverse intrinsic

Follow the naming convention that bswap uses since it's a
similar sort of operation.

llvm-svn: 259671
This commit is contained in:
Matt Arsenault 2016-02-03 17:49:38 +00:00
parent 45860fac37
commit 105e892c2c
3 changed files with 57 additions and 42 deletions

View File

@ -404,6 +404,10 @@ BUILTIN(__builtin_bswap16, "UsUs", "nc")
BUILTIN(__builtin_bswap32, "UiUi", "nc")
BUILTIN(__builtin_bswap64, "ULLiULLi", "nc")
BUILTIN(__builtin_bitreverse16, "UsUs", "nc")
BUILTIN(__builtin_bitreverse32, "UiUi", "nc")
BUILTIN(__builtin_bitreverse64, "ULLiULLi", "nc")
// Random GCC builtins
BUILTIN(__builtin_constant_p, "i.", "nctu")
BUILTIN(__builtin_classify_type, "i.", "nctu")

View File

@ -288,6 +288,40 @@ static llvm::Value *EmitOverflowIntrinsic(CodeGenFunction &CGF,
return CGF.Builder.CreateExtractValue(Tmp, 0);
}
// Emit a simple mangled intrinsic that has 1 argument and a return type
// matching the argument type.
static Value *emitUnaryBuiltin(CodeGenFunction &CGF,
const CallExpr *E,
unsigned IntrinsicID) {
llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
return CGF.Builder.CreateCall(F, Src0);
}
// Emit an intrinsic that has 3 float or double operands.
static Value *emitTernaryFPBuiltin(CodeGenFunction &CGF,
const CallExpr *E,
unsigned IntrinsicID) {
llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
llvm::Value *Src2 = CGF.EmitScalarExpr(E->getArg(2));
Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
return CGF.Builder.CreateCall(F, {Src0, Src1, Src2});
}
// Emit an intrinsic that has 1 float or double operand, and 1 integer.
static Value *emitFPIntBuiltin(CodeGenFunction &CGF,
const CallExpr *E,
unsigned IntrinsicID) {
llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
return CGF.Builder.CreateCall(F, {Src0, Src1});
}
namespace {
struct WidthAndSignedness {
unsigned Width;
@ -645,10 +679,12 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
case Builtin::BI__builtin_bswap16:
case Builtin::BI__builtin_bswap32:
case Builtin::BI__builtin_bswap64: {
Value *ArgValue = EmitScalarExpr(E->getArg(0));
llvm::Type *ArgType = ArgValue->getType();
Value *F = CGM.getIntrinsic(Intrinsic::bswap, ArgType);
return RValue::get(Builder.CreateCall(F, ArgValue));
return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::bswap));
}
case Builtin::BI__builtin_bitreverse16:
case Builtin::BI__builtin_bitreverse32:
case Builtin::BI__builtin_bitreverse64: {
return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::bitreverse));
}
case Builtin::BI__builtin_object_size: {
unsigned Type =
@ -6990,39 +7026,6 @@ Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
}
}
// Emit an intrinsic that has 1 float or double.
static Value *emitUnaryFPBuiltin(CodeGenFunction &CGF,
const CallExpr *E,
unsigned IntrinsicID) {
llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
return CGF.Builder.CreateCall(F, Src0);
}
// Emit an intrinsic that has 3 float or double operands.
static Value *emitTernaryFPBuiltin(CodeGenFunction &CGF,
const CallExpr *E,
unsigned IntrinsicID) {
llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
llvm::Value *Src2 = CGF.EmitScalarExpr(E->getArg(2));
Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
return CGF.Builder.CreateCall(F, {Src0, Src1, Src2});
}
// Emit an intrinsic that has 1 float or double operand, and 1 integer.
static Value *emitFPIntBuiltin(CodeGenFunction &CGF,
const CallExpr *E,
unsigned IntrinsicID) {
llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
return CGF.Builder.CreateCall(F, {Src0, Src1});
}
Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
const CallExpr *E) {
switch (BuiltinID) {
@ -7072,13 +7075,13 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
return emitFPIntBuiltin(*this, E, Intrinsic::amdgcn_trig_preop);
case AMDGPU::BI__builtin_amdgcn_rcp:
case AMDGPU::BI__builtin_amdgcn_rcpf:
return emitUnaryFPBuiltin(*this, E, Intrinsic::amdgcn_rcp);
return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_rcp);
case AMDGPU::BI__builtin_amdgcn_rsq:
case AMDGPU::BI__builtin_amdgcn_rsqf:
return emitUnaryFPBuiltin(*this, E, Intrinsic::amdgcn_rsq);
return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_rsq);
case AMDGPU::BI__builtin_amdgcn_rsq_clamped:
case AMDGPU::BI__builtin_amdgcn_rsq_clampedf:
return emitUnaryFPBuiltin(*this, E, Intrinsic::amdgcn_rsq_clamped);
return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_rsq_clamped);
case AMDGPU::BI__builtin_amdgcn_ldexp:
case AMDGPU::BI__builtin_amdgcn_ldexpf:
return emitFPIntBuiltin(*this, E, Intrinsic::amdgcn_ldexp);
@ -7090,8 +7093,8 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
case AMDGPU::BI__builtin_amdgpu_rsq:
case AMDGPU::BI__builtin_amdgpu_rsqf: {
if (getTarget().getTriple().getArch() == Triple::amdgcn)
return emitUnaryFPBuiltin(*this, E, Intrinsic::amdgcn_rsq);
return emitUnaryFPBuiltin(*this, E, Intrinsic::r600_rsq);
return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_rsq);
return emitUnaryBuiltin(*this, E, Intrinsic::r600_rsq);
}
case AMDGPU::BI__builtin_amdgpu_ldexp:
case AMDGPU::BI__builtin_amdgpu_ldexpf: {

View File

@ -116,6 +116,14 @@ int main() {
P(bswap16, (N));
P(bswap32, (N));
P(bswap64, (N));
// CHECK: @llvm.bitreverse.i16
// CHECK: @llvm.bitreverse.i32
// CHECK: @llvm.bitreverse.i64
P(bitreverse16, (N));
P(bitreverse32, (N));
P(bitreverse64, (N));
// FIXME
// V(clear_cache, (&N, &N+1));
V(trap, ());