[MLIR][LLVMIR] Import unregistered intrinsics via llvm.intrinsic_call #128626
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Currently, the llvm importer can only cover intrinsics that have a first class representation in an MLIR dialect (arm-neon, etc). This PR introduces a fallback mechanism that allow "unregistered" intrinsics to be imported by using the generic `llvm.intrinsic_call` operation. This is useful in several ways: 1. Allows round-trip the LLVM dialect output lowered from other dialects (example: ClangIR) 2. Enables MLIR-linking tools to operate on imported LLVM IR without requiring adding new operations to dozen of different targets (cc @xlauko @smeenai). Implemented: - Cover intrinsics across all current supported LLVM targets and generic ones. - Use existing LLVM tablegen files to populate intrinsic matching. - Matches with lower priority against dialect supported intrinsic conversion. Open question: how to add the proper include path for tablegen to find Intrinsics.td?
|
@llvm/pr-subscribers-mlir-core @llvm/pr-subscribers-mlir Author: Bruno Cardoso Lopes (bcardosolopes) ChangesCurrently, the llvm importer can only cover intrinsics that have a first class representation in an MLIR dialect (arm-neon, etc). This PR introduces a fallback mechanism that allow "unregistered" intrinsics to be imported by using the generic
Implemented:
Open question: how to add the proper include path for tablegen to find Intrinsics.td? Patch is 20.23 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/128626.diff 7 Files Affected:
diff --git a/mlir/include/mlir/Dialect/LLVMIR/CMakeLists.txt b/mlir/include/mlir/Dialect/LLVMIR/CMakeLists.txt
index 759de745440c2..cefd1bbe4b027 100644
--- a/mlir/include/mlir/Dialect/LLVMIR/CMakeLists.txt
+++ b/mlir/include/mlir/Dialect/LLVMIR/CMakeLists.txt
@@ -81,3 +81,29 @@ mlir_tablegen(VCIXConversions.inc -gen-llvmir-conversions)
mlir_tablegen(VCIXOpsAttributes.h.inc -gen-attrdef-decls -attrdefs-dialect=vcix)
mlir_tablegen(VCIXOpsAttributes.cpp.inc -gen-attrdef-defs -attrdefs-dialect=vcix)
add_public_tablegen_target(MLIRVCIXConversionsIncGen)
+
+# FIXME: Should emit extra file with intrinsics_gen if MLIR is enabled? Or maybe find
+# a better way to get the path for Intrinsics.td?
+set(LLVM_TARGET_DEFINITIONS ../../../../../llvm/include/llvm/IR/Intrinsics.td)
+
+mlir_tablegen(LLVMUnregisteredLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics)
+mlir_tablegen(LLVMUnregisteredAArch64LLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=aarch64)
+mlir_tablegen(LLVMUnregisteredAMDGPULLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=amdgcn)
+mlir_tablegen(LLVMUnregisteredARMLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=arm)
+mlir_tablegen(LLVMUnregisteredBPFLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=bpf)
+mlir_tablegen(LLVMUnregisteredDirectXLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=dx)
+mlir_tablegen(LLVMUnregisteredHexagonLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=hexagon)
+mlir_tablegen(LLVMUnregisteredLoongArchLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=loongarch)
+mlir_tablegen(LLVMUnregisteredMipsLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=mips)
+mlir_tablegen(LLVMUnregisteredNVPTXLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=nvvm)
+mlir_tablegen(LLVMUnregisteredPowerPCLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=ppc)
+mlir_tablegen(LLVMUnregisteredR600LLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=r600)
+mlir_tablegen(LLVMUnregisteredRISCVLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=riscv)
+mlir_tablegen(LLVMUnregisteredSPIRVLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=spv)
+mlir_tablegen(LLVMUnregisteredS390LLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=s390)
+mlir_tablegen(LLVMUnregisteredWebAssemblyLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=wasm)
+mlir_tablegen(LLVMUnregisteredX86LLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=x86)
+mlir_tablegen(LLVMUnregisteredXCoreLLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=xcore)
+mlir_tablegen(LLVMUnregisteredVELLVMIRIntrinsics.inc -gen-unregistered-llvmir-intrinsics -intrinsic-prefix=ve)
+
+add_public_tablegen_target(MLIRLLVMIntrinsicUnregisteredIncGen)
\ No newline at end of file
diff --git a/mlir/include/mlir/Target/LLVMIR/LLVMImportInterface.h b/mlir/include/mlir/Target/LLVMIR/LLVMImportInterface.h
index cc5a77ed35d2b..d122021e8586d 100644
--- a/mlir/include/mlir/Target/LLVMIR/LLVMImportInterface.h
+++ b/mlir/include/mlir/Target/LLVMIR/LLVMImportInterface.h
@@ -74,6 +74,10 @@ class LLVMImportDialectInterface
/// returns the list of supported intrinsic identifiers.
virtual ArrayRef<unsigned> getSupportedIntrinsics() const { return {}; }
+ /// Returns the list of LLVM IR intrinsic identifiers that are unsupported
+ /// but dialects might have a generic way to represent them.
+ virtual ArrayRef<unsigned> getUnregisteredIntrinsics() const { return {}; }
+
/// Hook for derived dialect interfaces to publish the supported instructions.
/// As every LLVM IR instruction has a unique integer identifier, the function
/// returns the list of supported instruction identifiers. These identifiers
@@ -139,6 +143,9 @@ class LLVMImportInterface
// Add a mapping for all supported intrinsic identifiers.
for (unsigned id : iface.getSupportedIntrinsics())
intrinsicToDialect[id] = iface.getDialect();
+ // Add a mapping for all unregistered intrinsic identifiers.
+ for (unsigned id : iface.getUnregisteredIntrinsics())
+ unregisteredIntrinscToDialect[id] = iface.getDialect();
// Add a mapping for all supported instruction identifiers.
for (unsigned id : iface.getSupportedInstructions())
instructionToDialect[id] = &iface;
@@ -155,7 +162,19 @@ class LLVMImportInterface
LogicalResult convertIntrinsic(OpBuilder &builder, llvm::CallInst *inst,
LLVM::ModuleImport &moduleImport) const {
// Lookup the dialect interface for the given intrinsic.
- Dialect *dialect = intrinsicToDialect.lookup(inst->getIntrinsicID());
+ llvm::Intrinsic::ID intrinId = inst->getIntrinsicID();
+ if (intrinId == llvm::Intrinsic::not_intrinsic)
+ return failure();
+
+ // First lookup intrinsic across different dialects for known
+ // supported converstions, examples include arm-neon, nvm-sve, etc
+ Dialect *dialect = intrinsicToDialect.lookup(intrinId);
+
+ // No specialized (supported) intrinsics, attempt to generate a generic
+ // version via llvm.call_intrinsic (if available).
+ if (!dialect)
+ dialect = unregisteredIntrinscToDialect.lookup(intrinId);
+
if (!dialect)
return failure();
@@ -227,6 +246,9 @@ class LLVMImportInterface
DenseMap<unsigned, Dialect *> intrinsicToDialect;
DenseMap<unsigned, const LLVMImportDialectInterface *> instructionToDialect;
DenseMap<unsigned, SmallVector<Dialect *, 1>> metadataToDialect;
+
+ /// Unregistered generic and target independent intrinsics.
+ DenseMap<unsigned, Dialect *> unregisteredIntrinscToDialect;
};
} // namespace mlir
diff --git a/mlir/lib/Dialect/LLVMIR/CMakeLists.txt b/mlir/lib/Dialect/LLVMIR/CMakeLists.txt
index c9a3b97294562..030aed9cdb06a 100644
--- a/mlir/lib/Dialect/LLVMIR/CMakeLists.txt
+++ b/mlir/lib/Dialect/LLVMIR/CMakeLists.txt
@@ -17,6 +17,9 @@ add_mlir_dialect_library(MLIRLLVMDialect
MLIRLLVMTypesIncGen
MLIRLLVMIntrinsicOpsIncGen
MLIRLLVMInterfacesIncGen
+ MLIRLLVMConversionsIncGen
+ MLIRLLVMIntrinsicConversionsIncGen
+ MLIRLLVMIntrinsicUnregisteredIncGen
MLIROpenMPOpsIncGen
intrinsics_gen
diff --git a/mlir/lib/Target/LLVMIR/Dialect/LLVMIR/LLVMIRToLLVMTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/LLVMIR/LLVMIRToLLVMTranslation.cpp
index 4fd043c7c93e6..88ea2ff58434d 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/LLVMIR/LLVMIRToLLVMTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/LLVMIR/LLVMIRToLLVMTranslation.cpp
@@ -24,6 +24,24 @@
#include "llvm/IR/InlineAsm.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
+#include "llvm/IR/IntrinsicsAArch64.h"
+#include "llvm/IR/IntrinsicsAMDGPU.h"
+#include "llvm/IR/IntrinsicsARM.h"
+#include "llvm/IR/IntrinsicsBPF.h"
+#include "llvm/IR/IntrinsicsDirectX.h"
+#include "llvm/IR/IntrinsicsHexagon.h"
+#include "llvm/IR/IntrinsicsLoongArch.h"
+#include "llvm/IR/IntrinsicsMips.h"
+#include "llvm/IR/IntrinsicsNVPTX.h"
+#include "llvm/IR/IntrinsicsPowerPC.h"
+#include "llvm/IR/IntrinsicsR600.h"
+#include "llvm/IR/IntrinsicsRISCV.h"
+#include "llvm/IR/IntrinsicsS390.h"
+#include "llvm/IR/IntrinsicsSPIRV.h"
+#include "llvm/IR/IntrinsicsVE.h"
+#include "llvm/IR/IntrinsicsWebAssembly.h"
+#include "llvm/IR/IntrinsicsX86.h"
+#include "llvm/IR/IntrinsicsXCore.h"
#include "llvm/Support/ModRef.h"
using namespace mlir;
@@ -56,6 +74,105 @@ static ArrayRef<unsigned> getSupportedIntrinsicsImpl() {
return convertibleIntrinsics;
}
+/// Returns true if the LLVM IR intrinsic is convertible to llvm.intrinsic_call
+/// Returns false otherwise.
+static bool isConvertibleUnregisteredIntrinsic(llvm::Intrinsic::ID id) {
+ static const DenseSet<unsigned> convertibleTargetIntrinsics = {
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredAArch64LLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredAMDGPULLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredARMLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredBPFLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredDirectXLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredHexagonLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredLoongArchLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredMipsLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredNVPTXLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredPowerPCLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredR600LLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredRISCVLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredS390LLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredSPIRVLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredVELLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredWebAssemblyLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredX86LLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredXCoreLLVMIRIntrinsics.inc"
+ };
+ return convertibleTargetIntrinsics.contains(id);
+}
+
+/// Returns the list of LLVM IR intrinsic identifiers that are not registered
+/// by any dialect but can be convertible to llvm.intrinsic_call operation.
+static ArrayRef<unsigned> getUnregisteredIntrinsicsImpl() {
+ static const SmallVector<unsigned> convertibleTargetIntrinsics = {
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredAArch64LLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredAMDGPULLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredARMLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredBPFLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredDirectXLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredHexagonLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredLoongArchLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredMipsLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredNVPTXLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredPowerPCLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredR600LLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredRISCVLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredS390LLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredSPIRVLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredVELLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredWebAssemblyLLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredX86LLVMIRIntrinsics.inc"
+#include "mlir/Dialect/LLVMIR/LLVMUnregisteredXCoreLLVMIRIntrinsics.inc"
+ };
+ return convertibleTargetIntrinsics;
+}
+
+/// Converts the LLVM intrinsic to a generic LLVM intrinsic call using
+/// llvm.intrinsic_call. Returns failure otherwise.
+static LogicalResult
+convertUnregisteredIntrinsicImpl(OpBuilder &odsBuilder, llvm::CallInst *inst,
+ LLVM::ModuleImport &moduleImport) {
+ llvm::Intrinsic::ID intrinsicID = inst->getIntrinsicID();
+ StringRef intrinName = inst->getCalledFunction()->getName();
+
+ // Sanity check the intrinsic ID.
+ assert(isConvertibleUnregisteredIntrinsic(intrinsicID));
+ SmallVector<llvm::Value *> args(inst->args());
+ ArrayRef<llvm::Value *> llvmOperands(args);
+
+ SmallVector<llvm::OperandBundleUse> llvmOpBundles;
+ llvmOpBundles.reserve(inst->getNumOperandBundles());
+ for (unsigned i = 0; i < inst->getNumOperandBundles(); ++i)
+ llvmOpBundles.push_back(inst->getOperandBundleAt(i));
+
+ SmallVector<Value> mlirOperands;
+ SmallVector<NamedAttribute> mlirAttrs;
+ if (failed(moduleImport.convertIntrinsicArguments(
+ llvmOperands, llvmOpBundles, false, {}, {}, mlirOperands, mlirAttrs)))
+ return failure();
+
+ mlir::Type results = moduleImport.convertType(inst->getType());
+ auto op = odsBuilder.create<::mlir::LLVM::CallIntrinsicOp>(
+ moduleImport.translateLoc(inst->getDebugLoc()), results,
+ StringAttr::get(odsBuilder.getContext(), intrinName),
+ ValueRange{mlirOperands}, FastmathFlagsAttr{});
+
+ moduleImport.setFastmathFlagsAttr(inst, op);
+
+ // Update importer tracking of results.
+ unsigned numRes = op.getNumResults();
+ if (numRes == 1)
+ moduleImport.mapValue(inst) = op.getResult(0);
+ else if (numRes == 0)
+ moduleImport.mapNoResultOp(inst);
+ else
+ return op.emitError(
+ "expected at most one result from target intrinsic call");
+
+ return success();
+}
+
/// Converts the LLVM intrinsic to an MLIR LLVM dialect operation if a
/// conversion exits. Returns failure otherwise.
static LogicalResult convertIntrinsicImpl(OpBuilder &odsBuilder,
@@ -75,6 +192,8 @@ static LogicalResult convertIntrinsicImpl(OpBuilder &odsBuilder,
llvmOpBundles.push_back(inst->getOperandBundleAt(i));
#include "mlir/Dialect/LLVMIR/LLVMIntrinsicFromLLVMIRConversions.inc"
+ } else if (isConvertibleUnregisteredIntrinsic(intrinsicID)) {
+ return convertUnregisteredIntrinsicImpl(odsBuilder, inst, moduleImport);
}
return failure();
@@ -422,6 +541,12 @@ class LLVMDialectLLVMIRImportInterface : public LLVMImportDialectInterface {
return getSupportedIntrinsicsImpl();
}
+ /// Returns the list of LLVM IR intrinsic identifiers that are unsupported
+ /// by existing dialects by are convertible to generic llvm.call_intrinsic.
+ ArrayRef<unsigned> getUnregisteredIntrinsics() const final {
+ return getUnregisteredIntrinsicsImpl();
+ }
+
/// Returns the list of LLVM IR metadata kinds that are convertible to MLIR
/// LLVM dialect attributes.
ArrayRef<unsigned>
diff --git a/mlir/test/Target/LLVMIR/Import/import-failure.ll b/mlir/test/Target/LLVMIR/Import/import-failure.ll
index d929a59284762..fc4ccddb756d5 100644
--- a/mlir/test/Target/LLVMIR/Import/import-failure.ll
+++ b/mlir/test/Target/LLVMIR/Import/import-failure.ll
@@ -38,18 +38,6 @@ bb1:
; // -----
-declare void @llvm.gcroot(ptr %arg1, ptr %arg2)
-
-; CHECK: <unknown>
-; CHECK-SAME: error: unhandled intrinsic: call void @llvm.gcroot(ptr %arg1, ptr null)
-define void @unhandled_intrinsic() gc "example" {
- %arg1 = alloca ptr
- call void @llvm.gcroot(ptr %arg1, ptr null)
- ret void
-}
-
-; // -----
-
; Check that debug intrinsics with an unsupported argument are dropped.
declare void @llvm.dbg.value(metadata, metadata, metadata)
diff --git a/mlir/test/Target/LLVMIR/Import/intrinsic-unregistered.ll b/mlir/test/Target/LLVMIR/Import/intrinsic-unregistered.ll
new file mode 100644
index 0000000000000..0f30ce438dbaa
--- /dev/null
+++ b/mlir/test/Target/LLVMIR/Import/intrinsic-unregistered.ll
@@ -0,0 +1,68 @@
+; RUN: mlir-translate -import-llvm %s -split-input-file | FileCheck %s
+
+declare i64 @llvm.aarch64.ldxr.p0(ptr)
+
+define dso_local void @t0(ptr %a) {
+ %x = call i64 @llvm.aarch64.ldxr.p0(ptr elementtype(i8) %a)
+ ret void
+}
+
+; CHECK-LABEL: llvm.func @llvm.aarch64.ldxr.p0(!llvm.ptr)
+; CHECK-LABEL: llvm.func @t0
+; CHECK: llvm.call_intrinsic "llvm.aarch64.ldxr.p0"({{.*}}) : (!llvm.ptr) -> i64
+; CHECK: llvm.return
+; CHECK: }
+
+; -----
+
+declare <8 x i8> @llvm.aarch64.neon.uabd.v8i8(<8 x i8>, <8 x i8>)
+
+define dso_local <8 x i8> @t1(<8 x i8> %lhs, <8 x i8> %rhs) {
+ %r = call <8 x i8> @llvm.aarch64.neon.uabd.v8i8(<8 x i8> %lhs, <8 x i8> %rhs)
+ ret <8 x i8> %r
+}
+
+; CHECK: llvm.func @t1(%[[A0:.*]]: vector<8xi8>, %[[A1:.*]]: vector<8xi8>) -> vector<8xi8> {{.*}} {
+; CHECK: %[[R:.*]] = llvm.call_intrinsic "llvm.aarch64.neon.uabd.v8i8"(%[[A0]], %[[A1]]) : (vector<8xi8>, vector<8xi8>) -> vector<8xi8>
+; CHECK: llvm.return %[[R]] : vector<8xi8>
+; CHECK: }
+
+; -----
+
+declare void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8>, <8 x i8>, ptr)
+
+define dso_local void @t2(<8 x i8> %lhs, <8 x i8> %rhs, ptr %a) {
+ call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> %lhs, <8 x i8> %rhs, ptr %a)
+ ret void
+}
+
+; CHECK: llvm.func @t2(%[[A0:.*]]: vector<8xi8>, %[[A1:.*]]: vector<8xi8>, %[[A2:.*]]: !llvm.ptr) {{.*}} {
+; CHECK: llvm.call_intrinsic "llvm.aarch64.neon.st2.v8i8.p0"(%[[A0]], %[[A1]], %[[A2]]) : (vector<8xi8>, vector<8xi8>, !llvm.ptr) -> !llvm.void
+; CHECK: llvm.return
+; CHECK: }
+
+; -----
+
+declare void @llvm.gcroot(ptr %arg1, ptr %arg2)
+define void @gctest() gc "example" {
+ %arg1 = alloca ptr
+ call void @llvm.gcroot(ptr %arg1, ptr null)
+ ret void
+}
+
+; CHECK-LABEL: @gctest
+; CHECK: llvm.call_intrinsic "llvm.gcroot"({{.*}}, {{.*}}) : (!llvm.ptr, !llvm.ptr) -> !llvm.void
+
+; -----
+
+; Test we get the supported version, not the unregistered one.
+
+declare i32 @llvm.lround.i32.f32(float)
+
+; CHECK-LABEL: llvm.func @lround_test
+define void @lround_test(float %0, double %1) {
+ ; CHECK-NOT: llvm.call_intrinsic "llvm.lround
+ ; CHECK: llvm.intr.lround(%{{.*}}) : (f32) -> i32
+ %3 = call i32 @llvm.lround.i32.f32(float %0)
+ ret void
+}
\ No newline at end of file
diff --git a/mlir/tools/mlir-tblgen/LLVMIRConversionGen.cpp b/mlir/tools/mlir-tblgen/LLVMIRConversionGen.cpp
index 9e19f479d673a..c78872fa2de39 100644
--- a/mlir/tools/mlir-tblgen/LLVMIRConversionGen.cpp
+++ b/mlir/tools/mlir-tblgen/LLVMIRConversionGen.cpp
@@ -19,6 +19,7 @@
#include "llvm/ADT/Sequence.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/ADT/Twine.h"
+#include "llvm/Support/CommandLine.h"
#include "llvm/Support/FormatVariadic.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/TableGen/Error.h"
@@ -564,6 +565,46 @@ static bool emitConvertibleIntrinsics(const RecordKeeper &records,
return false;
}
+static void emitOneUnregisteredIntrinsic(const Record &record, raw_ostream &os,
+ StringRef targetName) {
+ StringRef targetPrefix = record.getValueAsString("TargetPrefix");
+
+ // Not interested in target specific intrinsics in the generic namespace.
+ if (targetName.empty() && !targetPrefix.empty())
+ return;
+
+ // Skip unknown targets.
+ if (!targetName.empty() && targetPrefix != targetName)
+ return;
+
+ StringRef defName = record.getName();
+ ArrayRef<SMLoc> defLoc = record.getLoc();
+
+ // Sanity check the input.
+ if (!defName.starts_with("int_"))
+ PrintFatalError(defLoc,
+ "Intrinsic '" + defName + "' does not start with 'int_'!");
+
+ StringRef enumName = defName.substr(4);
+
+ os << "llvm::Intrinsic::";
+ if (!targetName.empty())
+ os << StringRef(targetName).upper() << "Intrinsics::";
+ os << enumName << ",\n";
+}
+
+// Emit the list of LLVM IR intrinsics enums, both target and generic. Those
+// are used for LLVMImporter's convenience when looking at intrinsics while
+// being up-to-date with new additions to LLVM.
+static bool emitUnregisteredIntrinsics(const RecordKeeper &records,
+ raw_ostream &os, StringRef targetName) {
+
+ for (const Record *def : records.getAllDerivedDefinitions("Intrinsic"))
+ emitOneUnregisteredIntrinsic(*def, os, targetName);
+
+ return false;
+}
+
static mlir::GenRegistration
genLLVMIRConversions("gen-llvmir-conversions",
"Generate LLVM IR conversions", emitBuilders);
@@ -590,3 +631,19 @@ static mlir::GenRegistration genConvertibleLLVMIRIntrinsics(
"gen-convertible-llvmir-intrinsics",
"Generate list of convertible LLVM IR intrinsics",
emitConvertibleIntrinsics);
+
+static llvm::cl::OptionCategory
+ genUnregIntrinsicCat("Options for -gen-unregistered-llvmir-intrinsics");
+static llvm::cl::opt<std::string> unregIntrinsicPrefix(
+ "intrinsic-prefix",
+ cl::desc("Specify target to generate intrinsic information"),
+ cl::value_desc("target prefix"), cl::cat(genUnregIntrinsicCat));
+
+static mlir::GenRegistration genUnregisteredLLVMIRIntrinsics(
+ "gen-unregistered-llvmir-intrinsics",
+ "Generate enum list of target specific or generic intrinsics according to "
+ "enums defined in LLVM by other tablege...
[truncated]
|
gysit
reviewed
Feb 25, 2025
I have been extra cautious without necessity - just simplified by applying your suggested change |
gysit
reviewed
Feb 25, 2025
mlir/lib/Target/LLVMIR/Dialect/LLVMIR/LLVMIRToLLVMTranslation.cpp
Outdated
Show resolved
Hide resolved
mlir/lib/Target/LLVMIR/Dialect/LLVMIR/LLVMIRToLLVMTranslation.cpp
Outdated
Show resolved
Hide resolved
Co-authored-by: Tobias Gysi <tobias.gysi@nextsilicon.com>
Co-authored-by: Tobias Gysi <tobias.gysi@nextsilicon.com>
Also done! |
gysit
approved these changes
Feb 26, 2025
mlir/lib/Target/LLVMIR/Dialect/LLVMIR/LLVMIRToLLVMTranslation.cpp
Outdated
Show resolved
Hide resolved
mlir/lib/Target/LLVMIR/Dialect/LLVMIR/LLVMIRToLLVMTranslation.cpp
Outdated
Show resolved
Hide resolved
|
Besides @gysit nits lgtm. Will try to look later today at some sources that were failing on our side, whether this fixes all intrinsic issues. |
|
I've checked the fails we've seen with @xlauko in coreutils. Seems to be working fine based on quick testing |
Co-authored-by: Tobias Gysi <tobias.gysi@nextsilicon.com>
Co-authored-by: Tobias Gysi <tobias.gysi@nextsilicon.com>
Co-authored-by: Tobias Gysi <tobias.gysi@nextsilicon.com>
|
Thanks for the extra checking folks! Applied last round of changes, will land as soon as it passes the tests |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/138/builds/10857 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/205/builds/1727 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/203/builds/2936 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/204/builds/1746 Here is the relevant piece of the build log for the reference |
llvm-sync bot
pushed a commit
to arm/arm-toolchain
that referenced
this pull request
Feb 27, 2025
…llvm.intrinsic_call" (#128973) Reverts llvm/llvm-project#128626 Looks like the static definition broke some bots! Co-authored-by: Bruno Cardoso Lopes <bcardosolopes@users.noreply.github.com>
bcardosolopes
added a commit
to bcardosolopes/llvm-project
that referenced
this pull request
Feb 28, 2025
…sic_call Original introduced in llvm#128626, reverted in llvm#128973 Reproduced the issue on a shared lib build locally on Linux, moved content arround to statisfy both static and shared lib builds. Currently, the llvm importer can only cover intrinsics that have a first class representation in an MLIR dialect (arm-neon, etc). This PR introduces a fallback mechanism that allow "unregistered" intrinsics to be imported by using the generic `llvm.intrinsic_call` operation. This is useful in several ways: 1. Allows round-trip the LLVM dialect output lowered from other dialects (example: ClangIR) 2. Enables MLIR-linking tools to operate on imported LLVM IR without requiring to add new operations to dozen of different targets (cc @xlauko @smeenai). If multiple dialects implement this interface hook, the last one to register is the one converting all unregistered intrinsics.
bcardosolopes
added a commit
that referenced
this pull request
Mar 3, 2025
#129174) …sic_call Original introduced in #128626, reverted in #128973 Reproduced the issue on a shared lib build locally on Linux, moved content around to satisfy both static and shared lib builds. ### Original commit message Currently, the llvm importer can only cover intrinsics that have a first class representation in an MLIR dialect (arm-neon, etc). This PR introduces a fallback mechanism that allow "unregistered" intrinsics to be imported by using the generic `llvm.intrinsic_call` operation. This is useful in several ways: 1. Allows round-trip the LLVM dialect output lowered from other dialects (example: ClangIR) 2. Enables MLIR-linking tools to operate on imported LLVM IR without requiring to add new operations to dozen of different targets. If multiple dialects implement this interface hook, the last one to register is the one converting all unregistered intrinsics. --------- Co-authored-by: Bruno Cardoso Lopes <bcardosolopes@users.noreply.github.com>
llvm-sync bot
pushed a commit
to arm/arm-toolchain
that referenced
this pull request
Mar 3, 2025
…llvm.intrin… (#129174) …sic_call Original introduced in llvm/llvm-project#128626, reverted in llvm/llvm-project#128973 Reproduced the issue on a shared lib build locally on Linux, moved content around to satisfy both static and shared lib builds. ### Original commit message Currently, the llvm importer can only cover intrinsics that have a first class representation in an MLIR dialect (arm-neon, etc). This PR introduces a fallback mechanism that allow "unregistered" intrinsics to be imported by using the generic `llvm.intrinsic_call` operation. This is useful in several ways: 1. Allows round-trip the LLVM dialect output lowered from other dialects (example: ClangIR) 2. Enables MLIR-linking tools to operate on imported LLVM IR without requiring to add new operations to dozen of different targets. If multiple dialects implement this interface hook, the last one to register is the one converting all unregistered intrinsics. --------- Co-authored-by: Bruno Cardoso Lopes <bcardosolopes@users.noreply.github.com>
joaosaffran
pushed a commit
to joaosaffran/llvm-project
that referenced
this pull request
Mar 3, 2025
…sic_call" (llvm#128973) Reverts llvm#128626 Looks like the static definition broke some bots! Co-authored-by: Bruno Cardoso Lopes <bcardosolopes@users.noreply.github.com>
jph-13
pushed a commit
to jph-13/llvm-project
that referenced
this pull request
Mar 21, 2025
llvm#129174) …sic_call Original introduced in llvm#128626, reverted in llvm#128973 Reproduced the issue on a shared lib build locally on Linux, moved content around to satisfy both static and shared lib builds. ### Original commit message Currently, the llvm importer can only cover intrinsics that have a first class representation in an MLIR dialect (arm-neon, etc). This PR introduces a fallback mechanism that allow "unregistered" intrinsics to be imported by using the generic `llvm.intrinsic_call` operation. This is useful in several ways: 1. Allows round-trip the LLVM dialect output lowered from other dialects (example: ClangIR) 2. Enables MLIR-linking tools to operate on imported LLVM IR without requiring to add new operations to dozen of different targets. If multiple dialects implement this interface hook, the last one to register is the one converting all unregistered intrinsics. --------- Co-authored-by: Bruno Cardoso Lopes <bcardosolopes@users.noreply.github.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Currently, the llvm importer can only cover intrinsics that have a first class representation in an MLIR dialect (arm-neon, etc). This PR introduces a fallback mechanism that allow "unregistered" intrinsics to be imported by using the generic
llvm.intrinsic_calloperation. This is useful in several ways:If multiple dialects implement this interface hook, the last one to register is the one converting all unregistered intrinsics.