Uplift LLVM and Stablehlo (#2211)

### Problem description Uplifting llvm and stablehlo versions ### What's changed Uplifted llvm to llvm/llvm-project@a854c26 Uplifted stablehlo to openxla/stablehlo@4598975 - `applyPatternsAndFoldGreedily` has been deprecated. - I've replaced all calls to this with `applyPatternsGreedily` as the deprecation instructs us to use `applyPatternsGreedily` instead. - After looking at the [source for `applyPatternsAndFoldGreedily`](https://github.com/llvm/llvm-project/blob/26a83994176fcdca6e77be4f221a15f561681621/mlir/include/mlir/Transforms/GreedyPatternRewriteDriver.h#L186) it simply calls `applyPatternsGreedily` with a `GreedyRewriteConfig` that has `fold = true`. However the [source of `GreedyRewriteConfig`](https://github.com/llvm/llvm-project/blob/26a83994176fcdca6e77be4f221a15f561681621/mlir/include/mlir/Transforms/GreedyPatternRewriteDriver.h#L96) has `fold = true` by default. So making this replacement should have no effect on how we expect the pattern driver to behave. - `FloatType::get<type>()` No longer exists, instead `FloatTypes` for each type of float have been defined - i.e `Float32Type` - `tosa::MatMulOp` had its quantization info attribute(s) removed in this commit: llvm/llvm-project@f0b8ff1 - `populateSCFToEmitCConversionPatterns` requires a type converter now.
tenstorrent · Feb 24, 2025 · 1d937da · 1d937da
1 parent 2c434ef
commit 1d937da
Show file tree

Hide file tree

Showing 13 changed files with 25 additions and 44 deletions.
diff --git a/env/CMakeLists.txt b/env/CMakeLists.txt
@@ -2,8 +2,8 @@ cmake_minimum_required(VERSION 3.20.0)
 project(ttmlir-toolchain LANGUAGES CXX C)
 
 set(FLATBUFFERS_VERSION "fb9afbafc7dfe226b9db54d4923bfb8839635274")
-set(LLVM_PROJECT_VERSION "e813750354bbc08551cf23ff559a54b4a9ea1f29")
-set(STABLEHLO_VERSION "d40285ef3db0687e3f1e2bb0d716d748485a9739")
+set(LLVM_PROJECT_VERSION "a854c266b98468ad4479a7d3c56a3fa76437e30d")
+set(STABLEHLO_VERSION "459897561d365ef97caba46984847f9184d472ec")
 set(SHARDY_VERSION "55f44c23b766be38bccb0b2394b0e8dfba45694e")
 set(LLVM_BUILD_TYPE MinSizeRel CACHE STRING "Build type for LLVM")
 

diff --git a/include/ttmlir/Dialect/TT/IR/TTOpsTypes.h b/include/ttmlir/Dialect/TT/IR/TTOpsTypes.h
@@ -115,23 +115,23 @@ inline Type dataTypeToElementType(::mlir::MLIRContext *context,
                                   DataType dtype) {
   switch (dtype) {
   case DataType::Float32:
-    return FloatType::getF32(context);
+    return Float32Type::get(context);
   case DataType::Float16:
-    return FloatType::getF16(context);
+    return Float16Type::get(context);
   case DataType::BFloat16:
-    return FloatType::getBF16(context);
+    return BFloat16Type::get(context);
   case DataType::BFP_Float8:
-    return FloatType::getF16(context);
+    return Float16Type::get(context);
   case DataType::BFP_BFloat8:
-    return FloatType::getBF16(context);
+    return BFloat16Type::get(context);
   case DataType::BFP_Float4:
-    return FloatType::getF16(context);
+    return Float16Type::get(context);
   case DataType::BFP_BFloat4:
-    return FloatType::getBF16(context);
+    return BFloat16Type::get(context);
   case DataType::BFP_Float2:
-    return FloatType::getF16(context);
+    return Float16Type::get(context);
   case DataType::BFP_BFloat2:
-    return FloatType::getBF16(context);
+    return BFloat16Type::get(context);
   case DataType::UInt32:
     return IntegerType::get(context, 32,
                             IntegerType::SignednessSemantics::Unsigned);

diff --git a/lib/Conversion/StableHLOToTTIR/StableHLOToTTIRPass.cpp b/lib/Conversion/StableHLOToTTIR/StableHLOToTTIRPass.cpp
@@ -89,7 +89,7 @@ class StablehloTypeConverter : public TypeConverter {
         }
         // Convert 64 bit float element type to 32 bit float.
         else if (isa<FloatType>(elementType)) {
-          elementType = FloatType::getF32(context);
+          elementType = Float32Type::get(context);
           changed = true;
         }
       }

diff --git a/lib/Conversion/TTIRToTTMetal/AttachMetalLayout.cpp b/lib/Conversion/TTIRToTTMetal/AttachMetalLayout.cpp
@@ -146,7 +146,7 @@ class TTIRAttachMetalLayout
     RewritePatternSet patterns(&getContext());
     patterns.add<TTIRLayoutTensorTypeRewriter>(typeConverter, &getContext());
     FrozenRewritePatternSet patternSet(std::move(patterns));
-    if (failed(applyPatternsAndFoldGreedily(getOperation(), patternSet))) {
+    if (failed(applyPatternsGreedily(getOperation(), patternSet))) {
       signalPassFailure();
       return;
     }

diff --git a/lib/Conversion/TTKernelToEmitC/TTKernelToEmitC.cpp b/lib/Conversion/TTKernelToEmitC/TTKernelToEmitC.cpp
@@ -530,7 +530,7 @@ class ConvertTTKernelToEmitCPass
       {
         populateArithToEmitCPatterns(typeConverter, patterns);
 
-        populateSCFToEmitCConversionPatterns(patterns);
+        populateSCFToEmitCConversionPatterns(patterns, typeConverter);
 
         populateMemRefToEmitCTypeConversion(typeConverter);
         populateMemRefToEmitCConversionPatterns(patterns, typeConverter);

diff --git a/lib/Conversion/TosaToTTIR/TosaToTTIRPatterns.cpp b/lib/Conversion/TosaToTTIR/TosaToTTIRPatterns.cpp
@@ -130,12 +130,6 @@ class TosaToTTIRMatmulOpConversionPattern
   LogicalResult
   matchAndRewrite(tosa::MatMulOp srcOp, Adaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
-    LogicalResult legalityResult =
-        checkConversionLegality(srcOp, adaptor, rewriter);
-    if (!legalityResult.succeeded()) {
-      return legalityResult;
-    }
-
     auto outputType = mlir::cast<RankedTensorType>(
         this->getTypeConverter()->convertType(srcOp.getResult().getType()));
 
@@ -144,17 +138,6 @@ class TosaToTTIRMatmulOpConversionPattern
 
     return success();
   }
-
-private:
-  LogicalResult
-  checkConversionLegality(tosa::MatMulOp srcOp, Adaptor adaptor,
-                          ConversionPatternRewriter &rewriter) const {
-    if (srcOp.getQuantizationInfo().has_value()) {
-      return rewriter.notifyMatchFailure(
-          srcOp, "TTIR MatmulOp currently doesn't support quantization.");
-    }
-    return success();
-  }
 };
 } // namespace
 

diff --git a/lib/Dialect/TTIR/Transforms/Broadcast.cpp b/lib/Dialect/TTIR/Transforms/Broadcast.cpp
@@ -94,7 +94,7 @@ class TTIRImplicitBroadcastFold
     patterns.add<TTIRImplicitBroadcastFoldRewriter>(&getContext());
     FrozenRewritePatternSet patternSet(std::move(patterns));
 
-    if (failed(applyPatternsAndFoldGreedily(getOperation(), patternSet))) {
+    if (failed(applyPatternsGreedily(getOperation(), patternSet))) {
       signalPassFailure();
       return;
     }

diff --git a/lib/Dialect/TTIR/Transforms/Constant.cpp b/lib/Dialect/TTIR/Transforms/Constant.cpp
@@ -40,7 +40,7 @@ class TTIRConstantAsFill
     RewritePatternSet patterns(&getContext());
     patterns.add<TTIRConstantAsFillRewriter>(&getContext());
     FrozenRewritePatternSet patternSet(std::move(patterns));
-    if (failed(applyPatternsAndFoldGreedily(getOperation(), patternSet))) {
+    if (failed(applyPatternsGreedily(getOperation(), patternSet))) {
       signalPassFailure();
       return;
     }

diff --git a/lib/Dialect/TTIR/Transforms/HoistCPUOps.cpp b/lib/Dialect/TTIR/Transforms/HoistCPUOps.cpp
@@ -70,7 +70,7 @@ static void hoistOperationToFunction(mlir::Operation *opToHoist,
   const llvm::SmallVector<int64_t, 4> ranks = getOperandTensorRanks(opToHoist);
   mlir::MLIRContext *context = sourceModule.getContext();
   mlir::OpBuilder typeBuilder(opToHoist);
-  auto f32Type = mlir::FloatType::getF32(context);
+  auto f32Type = mlir::Float32Type::get(context);
 
   // Convert operands and gather types for function signature
   llvm::SmallVector<mlir::Type> operandTypes;

diff --git a/lib/Dialect/TTIR/Transforms/Layout.cpp b/lib/Dialect/TTIR/Transforms/Layout.cpp
@@ -286,7 +286,7 @@ class TTIRLayout : public impl::TTIRLayoutBase<TTIRLayout> {
       RewritePatternSet patterns(&getContext());
       patterns.add<TTIRLayoutTensorTypeRewriter>(typeConverter, &getContext());
       FrozenRewritePatternSet patternSet(std::move(patterns));
-      if (failed(applyPatternsAndFoldGreedily(getOperation(), patternSet))) {
+      if (failed(applyPatternsGreedily(getOperation(), patternSet))) {
         signalPassFailure();
         return;
       }
@@ -300,8 +300,7 @@ class TTIRLayout : public impl::TTIRLayoutBase<TTIRLayout> {
       FrozenRewritePatternSet patternSet(std::move(patterns));
       GreedyRewriteConfig config = GreedyRewriteConfig();
       config.useTopDownTraversal = true;
-      if (failed(applyPatternsAndFoldGreedily(getOperation(), patternSet,
-                                              config))) {
+      if (failed(applyPatternsGreedily(getOperation(), patternSet, config))) {
         signalPassFailure();
         return;
       }
@@ -425,7 +424,7 @@ class TTIRSplitCompoundLayout
     RewritePatternSet patterns(&getContext());
     patterns.add<TTIRSplitCompoundLayoutRewriter>(&getContext());
     FrozenRewritePatternSet patternSet(std::move(patterns));
-    if (failed(applyPatternsAndFoldGreedily(getOperation(), patternSet))) {
+    if (failed(applyPatternsGreedily(getOperation(), patternSet))) {
       signalPassFailure();
       return;
     }

diff --git a/lib/Dialect/TTNN/Transforms/Passes.cpp b/lib/Dialect/TTNN/Transforms/Passes.cpp
@@ -254,7 +254,8 @@ class TTNNCreateInputGenerators
 
       // Create function type
       //
-      mlir::TypeRange returnTypeRange = mlir::TypeRange(rewriter.getI32Type());
+      mlir::Type i32Type = rewriter.getI32Type();
+      mlir::TypeRange returnTypeRange = mlir::TypeRange(i32Type);
       FunctionType functionType =
           mlir::FunctionType::get(&getContext(), {}, returnTypeRange);
 

diff --git a/lib/Dialect/TTNN/Transforms/TTNNLayout.cpp b/lib/Dialect/TTNN/Transforms/TTNNLayout.cpp
@@ -693,7 +693,7 @@ class TTNNLayout : public impl::TTNNLayoutBase<TTNNLayout> {
       patterns.add<TTNNLayoutFuncInputOutputTypeRewriter>(
           &getContext(), device.getWorkerGrid());
       FrozenRewritePatternSet patternSet(std::move(patterns));
-      if (failed(applyPatternsAndFoldGreedily(getOperation(), patternSet))) {
+      if (failed(applyPatternsGreedily(getOperation(), patternSet))) {
         signalPassFailure();
         return;
       }
@@ -711,8 +711,7 @@ class TTNNLayout : public impl::TTNNLayoutBase<TTNNLayout> {
       FrozenRewritePatternSet patternSet(std::move(patterns));
       GreedyRewriteConfig config = GreedyRewriteConfig();
       config.useTopDownTraversal = true;
-      if (failed(applyPatternsAndFoldGreedily(getOperation(), patternSet,
-                                              config))) {
+      if (failed(applyPatternsGreedily(getOperation(), patternSet, config))) {
         signalPassFailure();
         return;
       }

diff --git a/lib/Dialect/TTNN/Transforms/Workarounds/TTNNWorkarounds.cpp b/lib/Dialect/TTNN/Transforms/Workarounds/TTNNWorkarounds.cpp
@@ -475,8 +475,7 @@ class TTNNWorkarounds : public impl::TTNNWorkaroundsBase<TTNNWorkarounds> {
     // This configuration specifies that the rewriter should traverse the IR
     // in a top-down order.
     config.useTopDownTraversal = true;
-    if (failed(
-            applyPatternsAndFoldGreedily(getOperation(), patternSet, config))) {
+    if (failed(applyPatternsGreedily(getOperation(), patternSet, config))) {
       signalPassFailure();
       return;
     }