[Mlir-commits] [mlir] [MLIR][XeGPU] Add support for Convert Layout from Wg to Sg (PR #178922)

Nishant Patel llvmlistbot at llvm.org
Thu Feb 12 09:25:27 PST 2026


https://github.com/nbpatel updated https://github.com/llvm/llvm-project/pull/178922

>From de5c4b67dde500560422435b0e222291da148aae Mon Sep 17 00:00:00 2001
From: nbpatel <nishant.b.patel at intel.com>
Date: Fri, 23 Jan 2026 15:22:34 +0000
Subject: [PATCH 1/4] Add convert layout via SLM

---
 .../Transforms/XeGPUWgToSgDistribute.cpp      | 152 ++++++++++++++----
 .../XeGPU/xegpu-wg-to-sg-unify-ops.mlir       |  37 +++++
 2 files changed, 162 insertions(+), 27 deletions(-)

diff --git a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
index 8328c2797be4f..450dc34a4263d 100644
--- a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
+++ b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
@@ -604,44 +604,142 @@ struct WgToSgElementwiseOp : public ConversionPattern {
 struct WgToSgConvertLayoutOp
     : public OpConversionPattern<xegpu::ConvertLayoutOp> {
   using OpConversionPattern<xegpu::ConvertLayoutOp>::OpConversionPattern;
+
   LogicalResult
   matchAndRewrite(xegpu::ConvertLayoutOp op, OneToNOpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
-    // TODO: currently, we only support LayoutAttr
-    auto input = dyn_cast<xegpu::LayoutAttr>(op.getInputLayout());
-    auto target = dyn_cast<xegpu::LayoutAttr>(op.getTargetLayout());
+    Location loc = op.getLoc();
+
+    VectorType resultType = op.getResult().getType();
+    ArrayRef<int64_t> wgShape = resultType.getShape();
+    auto inputLayout = dyn_cast<xegpu::LayoutAttr>(op.getInputLayout());
+    auto targetLayout = dyn_cast<xegpu::LayoutAttr>(op.getTargetLayout());
 
-    if (!input || !target || !input.isForWorkgroup() ||
-        !target.isForWorkgroup())
+    if (!inputLayout || !targetLayout || !inputLayout.isForWorkgroup() ||
+        !targetLayout.isForWorkgroup())
       return rewriter.notifyMatchFailure(
           op, "Input and target layouts must have subgroup layout");
 
-    DenseI32ArrayAttr inputSgLayout = input.getSgLayout();
-    DenseI32ArrayAttr inputSgData = input.getSgData();
-    DenseI32ArrayAttr inputOrder = input.getOrder();
-    DenseI32ArrayAttr targetSgLayout = target.getSgLayout();
-    DenseI32ArrayAttr targetSgData = target.getSgData();
-    DenseI32ArrayAttr targetOrder = target.getOrder();
-
-    // TODO: currently we only support for optimal case, where input and
-    // output has the same sg_layout and sg_data, so SLM is not involved.
-    if (inputSgLayout != targetSgLayout || inputSgData != targetSgData ||
-        inputOrder != targetOrder)
+    SmallVector<int64_t> inputSgLayout =
+        inputLayout.getEffectiveSgLayoutAsInt();
+    SmallVector<int64_t> inputSgData = inputLayout.getEffectiveSgDataAsInt();
+    SmallVector<int64_t> targetSgLayout =
+        targetLayout.getEffectiveSgLayoutAsInt();
+    SmallVector<int64_t> targetSgData = targetLayout.getEffectiveSgDataAsInt();
+
+    // if sg_layout and sg_data are identical, no SLM needed
+    if (inputSgLayout == targetSgLayout && inputSgData == targetSgData) {
+      inputLayout = inputLayout.dropSgLayoutAndData();
+      targetLayout = targetLayout.dropSgLayoutAndData();
+
+      SmallVector<Value> newOps(adaptor.getSource());
+      if (inputLayout && targetLayout) {
+        for (auto [i, src] : llvm::enumerate(adaptor.getSource())) {
+          auto newOp = xegpu::ConvertLayoutOp::create(
+              rewriter, loc, src.getType(), src, inputLayout, targetLayout);
+          newOps[i] = newOp;
+        }
+      }
+      rewriter.replaceOpWithMultiple(op, {newOps});
+      return success();
+    }
+
+    // SLM path: layouts differ, need cross-subgroup data redistribution
+    auto srcVectorType = cast<VectorType>(op.getSource().getType());
+    Type elemTy = srcVectorType.getElementType();
+
+    // Calculate SLM size requirements
+    auto slmShape = wgShape;
+    auto bitWidth = elemTy.getIntOrFloatBitWidth();
+    auto bytesPerElement = bitWidth / 8;
+    auto slmSize = computeProduct(slmShape) * bytesPerElement;
+
+    // Allocate SLM
+    auto slmTy = MemRefType::get({slmSize}, rewriter.getI8Type(), {}, 3);
+    auto slm = memref::AllocaOp::create(rewriter, loc, slmTy);
+
+    auto memDescType = xegpu::MemDescType::get(rewriter.getContext(), slmShape,
+                                               elemTy, nullptr);
+    auto memDesc =
+        xegpu::CreateMemDescOp::create(rewriter, loc, memDescType, slm);
+
+    auto sgId = gpu::SubgroupIdOp::create(rewriter, loc,
+                                          rewriter.getIndexType(), nullptr);
+
+    // STORE PHASE: Store input data to SLM using input layout
+    // Convert input sg_layout to Values for delinearizeIndex
+    SmallVector<Value> inputSgLayoutValues;
+    for (int64_t dim : inputSgLayout) {
+      inputSgLayoutValues.push_back(
+          arith::ConstantIndexOp::create(rewriter, loc, dim));
+    }
+
+    auto inputSgIdsResult = affine::delinearizeIndex(
+        rewriter, loc, sgId.getResult(), inputSgLayoutValues);
+    if (failed(inputSgIdsResult))
       return failure();
+    SmallVector<Value> inputSgIds = *inputSgIdsResult;
+
+    // Calculate store offsets based on input subgroup position and sg_data
+    SmallVector<Value> storeOffsets;
+    for (size_t i = 0; i < inputSgIds.size(); ++i) {
+      Value sgDataVal =
+          arith::ConstantIndexOp::create(rewriter, loc, inputSgData[i]);
+      Value offset =
+          arith::MulIOp::create(rewriter, loc, inputSgIds[i], sgDataVal);
+      storeOffsets.push_back(offset);
+    }
 
-    input = input.dropSgLayoutAndData();
-    target = target.dropSgLayoutAndData();
+    SmallVector<OpFoldResult> storeMatrixOffsets(storeOffsets.begin(),
+                                                 storeOffsets.end());
 
-    SmallVector<Value> newOps(adaptor.getSource());
-    if (input && target) {
-      // keep the ConvertLayoutOp for rest fields, e.g., inst_data.
-      for (auto [i, src] : llvm::enumerate(adaptor.getSource())) {
-        auto newOp = xegpu::ConvertLayoutOp::create(
-            rewriter, op.getLoc(), src.getType(), src, input, target);
-        newOps[i] = newOp;
-      }
+    for (auto src : adaptor.getSource()) {
+      xegpu::StoreMatrixOp::create(rewriter, loc, src, memDesc.getResult(),
+                                   storeMatrixOffsets,
+                                   targetLayout.dropSgLayoutAndData());
     }
-    rewriter.replaceOpWithMultiple(op, {newOps});
+
+    gpu::BarrierOp::create(rewriter, loc);
+
+    // LOAD PHASE: Load data from SLM using target layout
+    // Convert target sg_layout to Values for delinearizeIndex
+    SmallVector<Value> targetSgLayoutValues;
+    for (int64_t dim : targetSgLayout) {
+      targetSgLayoutValues.push_back(
+          arith::ConstantIndexOp::create(rewriter, loc, dim));
+    }
+
+    auto targetSgIdsResult = affine::delinearizeIndex(
+        rewriter, loc, sgId.getResult(), targetSgLayoutValues);
+    if (failed(targetSgIdsResult))
+      return failure();
+    SmallVector<Value> targetSgIds = *targetSgIdsResult;
+
+    // Calculate load offsets based on target subgroup position and sg_data
+    SmallVector<Value> loadOffsets;
+    for (size_t i = 0; i < targetSgIds.size(); ++i) {
+      Value sgDataVal =
+          arith::ConstantIndexOp::create(rewriter, loc, targetSgData[i]);
+      Value offset =
+          arith::MulIOp::create(rewriter, loc, targetSgIds[i], sgDataVal);
+      loadOffsets.push_back(offset);
+    }
+
+    SmallVector<OpFoldResult> loadMatrixOffsets(loadOffsets.begin(),
+                                                loadOffsets.end());
+
+    VectorType targetVectorType = VectorType::get(targetSgData, elemTy);
+
+    SmallVector<Value> loadedVectors;
+    for (size_t i = 0; i < adaptor.getSource().size(); ++i) {
+      auto loadOp =
+          xegpu::LoadMatrixOp::create(rewriter, loc, targetVectorType,
+                                      memDesc.getResult(), loadMatrixOffsets,
+                                      /*layout=*/nullptr);
+      loadedVectors.push_back(loadOp.getResult());
+    }
+
+    rewriter.replaceOpWithMultiple(op, {loadedVectors});
     return success();
   }
 };
diff --git a/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir b/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir
index ff0946d100a63..06faf167803fc 100644
--- a/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir
+++ b/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir
@@ -8,6 +8,9 @@
 // CHECK-DAG: #map5 = affine_map<()[s0] -> ((s0 mod 32) floordiv 16)>
 // CHECK-DAG: #map6 = affine_map<()[s0] -> (s0 mod 16)>
 // CHECK-DAG: #map7 = affine_map<()[s0] -> ((s0 mod 16) floordiv 4)>
+// CHECK-DAG: #map8 = affine_map<()[s0] -> (s0 floordiv 16)>
+// CHECK-DAG: #map9 = affine_map<()[s0] -> (s0 floordiv 8)>
+// CHECK-DAG: #map10 = affine_map<()[s0] -> (s0 mod 8)>
 gpu.module @test_distribution {
   // CHECK-LABEL: create_nd_tdesc_no_offset
   // CHECK-SAME: %[[ARG_0:.*]]: memref<256x128xf32>
@@ -838,4 +841,38 @@ gpu.module @test_distribution {
       -> vector<256x128xf32>
     gpu.return
   }
+
+  // CHECK-LABEL: convert_layout_slm
+  // CHECK-SAME: %[[ARG0:.*]]: memref<128x256xf32>
+  gpu.func @convert_layout_slm(%arg0: memref<128x256xf32>) {
+    // CHECK-DAG: %[[SGID:.*]] = gpu.subgroup_id : index
+    // CHECK-DAG: %[[SGIDX:.*]] = arith.remui %[[SGID]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[SGIDY_TMP:.*]] = arith.divui %[[SGID]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[SGIDY:.*]] = arith.remui %[[SGIDY_TMP]], %[[C4:.*]] : index
+    // CHECK-DAG: %[[MUL_Y:.*]] = arith.muli %[[SGIDY]], %[[C32:.*]] : index
+    // CHECK-DAG: %[[MUL_X:.*]] = arith.muli %[[SGIDX]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[OFF_Y:.*]] = arith.remui %[[MUL_Y]], %[[C128:.*]] : index
+    // CHECK-DAG: %[[OFF_X:.*]] = arith.remui %[[MUL_X]], %[[C256:.*]] : index
+    // CHECK-DAG: %[[TDESC:.*]] = xegpu.create_nd_tdesc %[[ARG0]][%[[OFF_Y]], %[[OFF_X]]] : memref<128x256xf32> -> !xegpu.tensor_desc<32x16xf32, #xegpu.layout<inst_data = [16, 16]>>
+    // CHECK-DAG: %[[LOAD:.*]] = xegpu.load_nd %[[TDESC]] <{layout = #xegpu.layout<inst_data = [16, 16]>}> : !xegpu.tensor_desc<32x16xf32, #xegpu.layout<inst_data = [16, 16]>> -> vector<32x16xf32>
+    // CHECK-DAG: %[[ALLOCA:.*]] = memref.alloca() : memref<131072xi8, 3>
+    // CHECK-DAG: %[[MDESC:.*]] = xegpu.create_mem_desc %[[ALLOCA]] : memref<131072xi8, 3> -> !xegpu.mem_desc<128x256xf32>
+    // CHECK-DAG: %[[SGID_:.*]] = gpu.subgroup_id : index
+    // CHECK-DAG: %[[AFFINE1:.*]] = affine.apply #map8()[%[[SGID_]]]
+    // CHECK-DAG: %[[AFFINE2:.*]] = affine.apply #map6()[%[[SGID_]]]
+    // CHECK-DAG: %[[ROW_OFF:.*]] = arith.muli %[[AFFINE1]], %[[C32:.*]] : index
+    // CHECK-DAG: %[[COL_OFF:.*]] = arith.muli %[[AFFINE2]], %[[C16:.*]] : index
+    // CHECK-DAG: xegpu.store_matrix %[[LOAD]], %[[MDESC]][%[[ROW_OFF]], %[[COL_OFF]]] <{layout = #xegpu.layout<inst_data = [16, 16]>}>: vector<32x16xf32>, !xegpu.mem_desc<128x256xf32>, index, index
+    // CHECK-DAG: gpu.barrier
+    // CHECK-DAG: %[[AFFINE3:.*]] = affine.apply #map9()[%[[SGID_]]]
+    // CHECK-DAG: %[[AFFINE4:.*]] = affine.apply #map10()[%[[SGID_]]]
+    // CHECK-DAG: %[[ROW_OFF2:.*]] = arith.muli %[[AFFINE3]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[COL_OFF2:.*]] = arith.muli %[[AFFINE4]], %[[C32:.*]] : index
+    // CHECK-DAG: %[[LOAD_SLM:.*]] = xegpu.load_matrix %[[MDESC]][%[[ROW_OFF2]], %[[COL_OFF2]]] : !xegpu.mem_desc<128x256xf32>, index, index -> vector<16x32xf32>
+    %0 = xegpu.create_nd_tdesc %arg0[0, 0] : memref<128x256xf32> -> !xegpu.tensor_desc<128x256xf32, #xegpu.layout<sg_layout = [4, 16], sg_data = [32, 16], inst_data = [16, 16]>>
+    %1 = xegpu.load_nd %0 {layout = #xegpu.layout<sg_layout = [4, 16], sg_data = [32, 16], inst_data = [16, 16]>} : !xegpu.tensor_desc<128x256xf32, #xegpu.layout<sg_layout = [4, 16], sg_data = [32, 16], inst_data = [16, 16]>> -> vector<128x256xf32>
+    %2 = xegpu.convert_layout %1 <{input_layout = #xegpu.layout<sg_layout = [4, 16], sg_data = [32, 16], inst_data = [16, 16]>,
+                                   target_layout = #xegpu.layout<sg_layout = [8, 8], sg_data = [16, 32], inst_data = [16, 16]>}> : vector<128x256xf32>
+    gpu.return
+  }
 }

>From 3fa629332a679a27aab27270b33f322e6983593d Mon Sep 17 00:00:00 2001
From: nbpatel <nishant.b.patel at intel.com>
Date: Mon, 26 Jan 2026 22:58:01 +0000
Subject: [PATCH 2/4] Add nD support

---
 .../include/mlir/Dialect/XeGPU/IR/XeGPUOps.td |  10 +-
 mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp        |   4 +-
 .../Transforms/XeGPUWgToSgDistribute.cpp      | 112 ++++++++----------
 mlir/test/Dialect/XeGPU/invalid.mlir          |   4 +-
 .../XeGPU/xegpu-wg-to-sg-unify-ops.mlir       |  72 ++++++++---
 5 files changed, 112 insertions(+), 90 deletions(-)

diff --git a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td
index 2cbec50772b98..fefc8c9903497 100644
--- a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td
+++ b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td
@@ -1593,9 +1593,8 @@ def XeGPU_LoadMatrixOp: XeGPU_Op<"load_matrix", [MemoryEffects<[MemRead]>,
   }];
 
   let description = [{
-    This operation loads a 2D block of data from shared local memory (SLM) as specified
-    by the provided 2D `mem_desc`. Only 2D memory descriptors are supported; use the
-    subview operation to obtain a compatible 2D `mem_desc` from a higher-rank descriptor if needed.
+    This operation loads an nD block of data from shared local memory (SLM) as specified
+    by the provided nD `mem_desc`. Memory descriptors of any rank are supported.
 
     This operation serves as an anchor through which users assign a layout attribute
     to govern computation distribution.
@@ -1665,9 +1664,8 @@ def XeGPU_StoreMatrixOp: XeGPU_Op<"store_matrix", [MemoryEffects<[MemWrite]>,
   let assemblyFormat = [{ $data `,` $mem_desc `` custom<DynamicIndexList>($offsets, $const_offsets)
                           prop-dict attr-dict `` `:` type(operands)}];
   let description = [{
-    This operation stores a 2D `data` fragment into the shared local memory region
-    specified by a 2D `mem_desc`. Only 2D memory descriptors are supported; use the
-    subview operation to obtain a 2D `mem_desc` from a higher-rank descriptor if needed.
+    This operation stores an nD `data` fragment into the shared local memory region
+    specified by an nD `mem_desc`. Memory descriptors of any rank are supported.
 
     This operation serves as an anchor through which users assign a layout attribute
     to govern computation distribution.
diff --git a/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp b/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp
index 91ba07a8e0256..c7226c7ebbd5d 100644
--- a/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp
+++ b/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp
@@ -186,8 +186,8 @@ IsValidMatrixOpParams(VectorType dataTy, MemDescType mdescTy,
       return success();
   }
 
-  if (mdescTy.getRank() != 2)
-    return emitError() << "mem_desc must be 2D.";
+  if (mdescTy.getRank() < 2)
+    return emitError() << "mem_desc must be 2D or greater.";
 
   ArrayRef<int64_t> dataShape = dataTy.getShape();
   ArrayRef<int64_t> mdescShape = mdescTy.getShape();
diff --git a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
index 44659bf1dec31..8dbca952cc8c2 100644
--- a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
+++ b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
@@ -627,7 +627,22 @@ struct WgToSgConvertLayoutOp
         targetLayout.getEffectiveSgLayoutAsInt();
     SmallVector<int64_t> targetSgData = targetLayout.getEffectiveSgDataAsInt();
 
-    // if sg_layout and sg_data are identical, no SLM needed
+    auto hasUnitLeadingDims = [](ArrayRef<int64_t> shape) {
+      if (shape.size() <= 2)
+        return true;
+      for (size_t i = 0; i + 2 < shape.size(); ++i)
+        if (shape[i] != 1)
+          return false;
+      return true;
+    };
+
+    if (wgShape.size() > 2) {
+      if (!hasUnitLeadingDims(inputSgData) || !hasUnitLeadingDims(targetSgData))
+        return rewriter.notifyMatchFailure(
+            op, "rank > 2 requires unit leading dims for sg_data");
+    }
+
+    // Fast path: if sg_layout and sg_data are identical, no SLM needed
     if (inputSgLayout == targetSgLayout && inputSgData == targetSgData) {
       inputLayout = inputLayout.dropSgLayoutAndData();
       targetLayout = targetLayout.dropSgLayoutAndData();
@@ -645,11 +660,11 @@ struct WgToSgConvertLayoutOp
     }
 
     // SLM path: layouts differ, need cross-subgroup data redistribution
-    auto srcVectorType = cast<VectorType>(op.getSource().getType());
-    Type elemTy = srcVectorType.getElementType();
+    Type elemTy = cast<VectorType>(op.getSource().getType()).getElementType();
+
+    SmallVector<int64_t> slmShape = llvm::to_vector(wgShape);
 
     // Calculate SLM size requirements
-    auto slmShape = wgShape;
     auto bitWidth = elemTy.getIntOrFloatBitWidth();
     auto bytesPerElement = bitWidth / 8;
     auto slmSize = computeProduct(slmShape) * bytesPerElement;
@@ -666,80 +681,47 @@ struct WgToSgConvertLayoutOp
     auto sgId = gpu::SubgroupIdOp::create(rewriter, loc,
                                           rewriter.getIndexType(), nullptr);
 
-    // STORE PHASE: Store input data to SLM using input layout
-    // Convert input sg_layout to Values for delinearizeIndex
-    SmallVector<Value> inputSgLayoutValues;
-    for (int64_t dim : inputSgLayout) {
-      inputSgLayoutValues.push_back(
-          arith::ConstantIndexOp::create(rewriter, loc, dim));
-    }
-
-    auto inputSgIdsResult = affine::delinearizeIndex(
-        rewriter, loc, sgId.getResult(), inputSgLayoutValues);
-    if (failed(inputSgIdsResult))
+    // STORE PHASE: Each subgroup stores in SLM using input layout
+    auto storeCoords = inputLayout.computeDistributedCoords(
+        rewriter, loc, sgId.getResult(), wgShape);
+    if (failed(storeCoords))
       return failure();
-    SmallVector<Value> inputSgIds = *inputSgIdsResult;
-
-    // Calculate store offsets based on input subgroup position and sg_data
-    SmallVector<Value> storeOffsets;
-    for (size_t i = 0; i < inputSgIds.size(); ++i) {
-      Value sgDataVal =
-          arith::ConstantIndexOp::create(rewriter, loc, inputSgData[i]);
-      Value offset =
-          arith::MulIOp::create(rewriter, loc, inputSgIds[i], sgDataVal);
-      storeOffsets.push_back(offset);
-    }
-
-    SmallVector<OpFoldResult> storeMatrixOffsets(storeOffsets.begin(),
-                                                 storeOffsets.end());
 
-    for (auto src : adaptor.getSource()) {
+    // Store to SLM
+    for (auto [src, coords] : llvm::zip(adaptor.getSource(), *storeCoords)) {
+      SmallVector<OpFoldResult> storeMatrixOffsets;
+      for (Value coord : coords) {
+        storeMatrixOffsets.push_back(coord);
+      }
       xegpu::StoreMatrixOp::create(rewriter, loc, src, memDesc.getResult(),
-                                   storeMatrixOffsets,
-                                   targetLayout.dropSgLayoutAndData());
+                                   storeMatrixOffsets, nullptr /*layout*/);
     }
 
     gpu::BarrierOp::create(rewriter, loc);
 
-    // LOAD PHASE: Load data from SLM using target layout
-    // Convert target sg_layout to Values for delinearizeIndex
-    SmallVector<Value> targetSgLayoutValues;
-    for (int64_t dim : targetSgLayout) {
-      targetSgLayoutValues.push_back(
-          arith::ConstantIndexOp::create(rewriter, loc, dim));
-    }
-
-    auto targetSgIdsResult = affine::delinearizeIndex(
-        rewriter, loc, sgId.getResult(), targetSgLayoutValues);
-    if (failed(targetSgIdsResult))
+    // LOAD PHASE: Each target subgroup loads from SLM using target layout
+    auto loadCoords = targetLayout.computeDistributedCoords(
+        rewriter, loc, sgId.getResult(), wgShape);
+    if (failed(loadCoords))
       return failure();
-    SmallVector<Value> targetSgIds = *targetSgIdsResult;
-
-    // Calculate load offsets based on target subgroup position and sg_data
-    SmallVector<Value> loadOffsets;
-    for (size_t i = 0; i < targetSgIds.size(); ++i) {
-      Value sgDataVal =
-          arith::ConstantIndexOp::create(rewriter, loc, targetSgData[i]);
-      Value offset =
-          arith::MulIOp::create(rewriter, loc, targetSgIds[i], sgDataVal);
-      loadOffsets.push_back(offset);
-    }
 
-    SmallVector<OpFoldResult> loadMatrixOffsets(loadOffsets.begin(),
-                                                loadOffsets.end());
+    VectorType loadType = VectorType::get(targetSgData, elemTy);
 
-    VectorType targetVectorType = VectorType::get(targetSgData, elemTy);
+    // Load vectors from SLM
+    SmallVector<Value> finalResults;
+    for (auto coords : *loadCoords) {
+      SmallVector<OpFoldResult> loadMatrixOffsets;
+      for (Value coord : coords) {
+        loadMatrixOffsets.push_back(coord);
+      }
+      auto loadOp = xegpu::LoadMatrixOp::create(
+          rewriter, loc, loadType, memDesc.getResult(), loadMatrixOffsets,
+          targetLayout.dropSgLayoutAndData());
 
-    SmallVector<Value> loadedVectors;
-    for (size_t i = 0; i < adaptor.getSource().size(); ++i) {
-      auto loadOp =
-          xegpu::LoadMatrixOp::create(rewriter, loc, targetVectorType,
-                                      memDesc.getResult(), loadMatrixOffsets,
-                                      /*layout=*/nullptr);
-      loadedVectors.push_back(loadOp.getResult());
+      finalResults.push_back(loadOp.getResult());
     }
 
-    rewriter.replaceOpWithMultiple(op, {loadedVectors});
+    rewriter.replaceOpWithMultiple(op, {finalResults});
     return success();
   }
 };
diff --git a/mlir/test/Dialect/XeGPU/invalid.mlir b/mlir/test/Dialect/XeGPU/invalid.mlir
index f2011ab86e9e9..e6376e3ecb4cd 100644
--- a/mlir/test/Dialect/XeGPU/invalid.mlir
+++ b/mlir/test/Dialect/XeGPU/invalid.mlir
@@ -852,7 +852,7 @@ func.func @load_mem_desc_invalid_result_size(%arg0: !xegpu.mem_desc<16x64xf16>)
 
 // -----
 func.func @load_mem_desc_invalid_rank(%arg0: !xegpu.mem_desc<64xf16>) {
-  // expected-error at +1 {{mem_desc must be 2D}}
+  // expected-error at +1 {{mem_desc must be 2D or greater}}
   %data = xegpu.load_matrix %arg0[16]: !xegpu.mem_desc<64xf16> -> vector<16xf16>
   return
 }
@@ -873,7 +873,7 @@ func.func @store_mem_desc_invalid_data_size(%arg0: !xegpu.mem_desc<16x64xf16>, %
 
 // -----
 func.func @store_mem_desc_invalid_rank(%arg0: !xegpu.mem_desc<64xf16>, %arg1: vector<32xf16>) {
-  // expected-error at +1 {{mem_desc must be 2D.}}
+  // expected-error at +1 {{mem_desc must be 2D or greater}}
   xegpu.store_matrix %arg1, %arg0[32] : vector<32xf16>, !xegpu.mem_desc<64xf16>
   return
 }
diff --git a/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir b/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir
index d0419d1a11a8d..d4b611c713674 100644
--- a/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir
+++ b/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir
@@ -8,9 +8,6 @@
 // CHECK-DAG: #map5 = affine_map<()[s0] -> ((s0 mod 32) floordiv 16)>
 // CHECK-DAG: #map6 = affine_map<()[s0] -> (s0 mod 16)>
 // CHECK-DAG: #map7 = affine_map<()[s0] -> ((s0 mod 16) floordiv 4)>
-// CHECK-DAG: #map8 = affine_map<()[s0] -> (s0 floordiv 16)>
-// CHECK-DAG: #map9 = affine_map<()[s0] -> (s0 floordiv 8)>
-// CHECK-DAG: #map10 = affine_map<()[s0] -> (s0 mod 8)>
 gpu.module @test_distribution {
   // CHECK-LABEL: create_nd_tdesc_no_offset
   // CHECK-SAME: %[[ARG_0:.*]]: memref<256x128xf32>
@@ -846,7 +843,7 @@ gpu.module @test_distribution {
   // CHECK-SAME: %[[ARG0:.*]]: memref<128x256xf32>
   gpu.func @convert_layout_slm(%arg0: memref<128x256xf32>) {
     // CHECK-DAG: %[[SGID:.*]] = gpu.subgroup_id : index
-    // CHECK-DAG: %[[SGIDX:.*]] = arith.remui %[[SGID]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[SGIDX:.*]] = arith.remui %[[SGID]],  %[[C16:.*]] : index
     // CHECK-DAG: %[[SGIDY_TMP:.*]] = arith.divui %[[SGID]], %[[C16:.*]] : index
     // CHECK-DAG: %[[SGIDY:.*]] = arith.remui %[[SGIDY_TMP]], %[[C4:.*]] : index
     // CHECK-DAG: %[[MUL_Y:.*]] = arith.muli %[[SGIDY]], %[[C32:.*]] : index
@@ -857,24 +854,69 @@ gpu.module @test_distribution {
     // CHECK-DAG: %[[LOAD:.*]] = xegpu.load_nd %[[TDESC]] <{layout = #xegpu.layout<inst_data = [16, 16]>}> : !xegpu.tensor_desc<32x16xf32, #xegpu.layout<inst_data = [16, 16]>> -> vector<32x16xf32>
     // CHECK-DAG: %[[ALLOCA:.*]] = memref.alloca() : memref<131072xi8, 3>
     // CHECK-DAG: %[[MDESC:.*]] = xegpu.create_mem_desc %[[ALLOCA]] : memref<131072xi8, 3> -> !xegpu.mem_desc<128x256xf32>
-    // CHECK-DAG: %[[SGID_:.*]] = gpu.subgroup_id : index
-    // CHECK-DAG: %[[AFFINE1:.*]] = affine.apply #map8()[%[[SGID_]]]
-    // CHECK-DAG: %[[AFFINE2:.*]] = affine.apply #map6()[%[[SGID_]]]
-    // CHECK-DAG: %[[ROW_OFF:.*]] = arith.muli %[[AFFINE1]], %[[C32:.*]] : index
-    // CHECK-DAG: %[[COL_OFF:.*]] = arith.muli %[[AFFINE2]], %[[C16:.*]] : index
-    // CHECK-DAG: xegpu.store_matrix %[[LOAD]], %[[MDESC]][%[[ROW_OFF]], %[[COL_OFF]]] <{layout = #xegpu.layout<inst_data = [16, 16]>}>: vector<32x16xf32>, !xegpu.mem_desc<128x256xf32>, index, index
+    // CHECK-DAG: %[[SGID_STORE:.*]] = gpu.subgroup_id : index
+    // CHECK-DAG: %[[STORE_X:.*]] = arith.remui %[[SGID_STORE]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[STORE_Y_TMP:.*]] = arith.divui %[[SGID_STORE]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[STORE_Y:.*]] = arith.remui %[[STORE_Y_TMP]], %[[C4:.*]] : index
+    // CHECK-DAG: %[[STORE_MUL_Y:.*]] = arith.muli %[[STORE_Y]], %[[C32:.*]] : index
+    // CHECK-DAG: %[[STORE_MUL_X:.*]] = arith.muli %[[STORE_X]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[STORE_OFF_Y:.*]] = arith.remui %[[STORE_MUL_Y]], %[[C128:.*]] : index
+    // CHECK-DAG: %[[STORE_OFF_X:.*]] = arith.remui %[[STORE_MUL_X]], %[[C256:.*]] : index
+    // CHECK-DAG: xegpu.store_matrix %[[LOAD]], %[[MDESC]][%[[STORE_OFF_Y]], %[[STORE_OFF_X]]] : vector<32x16xf32>, !xegpu.mem_desc<128x256xf32>, index, index
     // CHECK-DAG: gpu.barrier
-    // CHECK-DAG: %[[AFFINE3:.*]] = affine.apply #map9()[%[[SGID_]]]
-    // CHECK-DAG: %[[AFFINE4:.*]] = affine.apply #map10()[%[[SGID_]]]
-    // CHECK-DAG: %[[ROW_OFF2:.*]] = arith.muli %[[AFFINE3]], %[[C16:.*]] : index
-    // CHECK-DAG: %[[COL_OFF2:.*]] = arith.muli %[[AFFINE4]], %[[C32:.*]] : index
-    // CHECK-DAG: %[[LOAD_SLM:.*]] = xegpu.load_matrix %[[MDESC]][%[[ROW_OFF2]], %[[COL_OFF2]]] : !xegpu.mem_desc<128x256xf32>, index, index -> vector<16x32xf32>
+    // CHECK-DAG: %[[LOAD_X:.*]] = arith.remui %[[SGID_STORE]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[LOAD_Y_TMP:.*]] = arith.divui %[[SGID_STORE]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[LOAD_Y:.*]] = arith.remui %[[LOAD_Y_TMP]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[LOAD_MUL_Y:.*]] = arith.muli %[[LOAD_Y]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[LOAD_MUL_X:.*]] = arith.muli %[[LOAD_X]], %[[C32:.*]] : index
+    // CHECK-DAG: %[[LOAD_OFF_Y:.*]] = arith.remui %[[LOAD_MUL_Y]], %[[C128:.*]] : index
+    // CHECK-DAG: %[[LOAD_OFF_X:.*]] = arith.remui %[[LOAD_MUL_X]], %[[C256:.*]] : index
+    // CHECK-DAG: %[[LOAD_SLM:.*]] = xegpu.load_matrix %[[MDESC]][%[[LOAD_OFF_Y]], %[[LOAD_OFF_X]]] <{layout = #xegpu.layout<inst_data = [16, 16]>}>: !xegpu.mem_desc<128x256xf32>, index, index -> vector<16x32xf32>
     %0 = xegpu.create_nd_tdesc %arg0[0, 0] : memref<128x256xf32> -> !xegpu.tensor_desc<128x256xf32, #xegpu.layout<sg_layout = [4, 16], sg_data = [32, 16], inst_data = [16, 16]>>
     %1 = xegpu.load_nd %0 {layout = #xegpu.layout<sg_layout = [4, 16], sg_data = [32, 16], inst_data = [16, 16]>} : !xegpu.tensor_desc<128x256xf32, #xegpu.layout<sg_layout = [4, 16], sg_data = [32, 16], inst_data = [16, 16]>> -> vector<128x256xf32>
     %2 = xegpu.convert_layout %1 <{input_layout = #xegpu.layout<sg_layout = [4, 16], sg_data = [32, 16], inst_data = [16, 16]>,
                                    target_layout = #xegpu.layout<sg_layout = [8, 8], sg_data = [16, 32], inst_data = [16, 16]>}> : vector<128x256xf32>
     gpu.return
   }
+
+  gpu.func @convert_layout_3D(%arg0: memref<?xf32>) {
+    // CHECK-DAG: %[[CST:.*]] = arith.constant {layout_result_0 = #xegpu.layout<inst_data = [1, 16, 16]>} dense<0> : vector<1x32x16xindex>
+    // CHECK-DAG: %[[CST_0:.*]] = arith.constant {layout_result_0 = #xegpu.layout<inst_data = [1, 16, 16]>} dense<true> : vector<1x32x16xi1>
+    // CHECK-DAG: %[[LOAD:.*]] = xegpu.load %{{.*}}[%[[CST]]], %[[CST_0]] <{chunk_size = 1 : i64, layout = #xegpu.layout<inst_data = [1, 16, 16]>}> : memref<?xf32>, vector<1x32x16xindex>, vector<1x32x16xi1> -> vector<1x32x16xf32>
+    // CHECK-DAG: %[[ALLOCA:.*]] = memref.alloca() : memref<1048576xi8, 3>
+    // CHECK-DAG: %[[MDESC:.*]] = xegpu.create_mem_desc %[[ALLOCA]] : memref<1048576xi8, 3> -> !xegpu.mem_desc<8x128x256xf32>
+    // CHECK-DAG: %[[SGID:.*]] = gpu.subgroup_id : index
+    // CHECK-DAG: %[[STORE_X:.*]] = arith.remui %[[SGID]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[STORE_YZ_TMP:.*]] = arith.divui %[[SGID]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[STORE_Y:.*]] = arith.remui %[[STORE_YZ_TMP]], %[[C4:.*]] : index
+    // CHECK-DAG: %[[STORE_Z_TMP:.*]] = arith.divui %[[STORE_YZ_TMP]], %[[C4:.*]] : index
+    // CHECK-DAG: %[[STORE_Z:.*]] = arith.remui %[[STORE_Z_TMP]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[STORE_MUL_Y:.*]] = arith.muli %[[STORE_Y]], %[[C32:.*]] : index
+    // CHECK-DAG: %[[STORE_MUL_X:.*]] = arith.muli %[[STORE_X]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[STORE_OFF_Z:.*]] = arith.remui %[[STORE_Z]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[STORE_OFF_Y:.*]] = arith.remui %[[STORE_MUL_Y]], %[[C128:.*]] : index
+    // CHECK-DAG: %[[STORE_OFF_X:.*]] = arith.remui %[[STORE_MUL_X]], %[[C256:.*]] : index
+    // CHECK-DAG: xegpu.store_matrix %[[LOAD]], %[[MDESC]][%[[STORE_OFF_Z]], %[[STORE_OFF_Y]], %[[STORE_OFF_X]]] : vector<1x32x16xf32>, !xegpu.mem_desc<8x128x256xf32>, index, index, index
+    // CHECK-DAG: gpu.barrier
+    // CHECK-DAG: %[[LOAD_X:.*]] = arith.remui %[[SGID]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[LOAD_YZ_TMP:.*]] = arith.divui %[[SGID]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[LOAD_Y:.*]] = arith.remui %[[LOAD_YZ_TMP]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[LOAD_Z_TMP:.*]] = arith.divui %[[LOAD_YZ_TMP]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[LOAD_Z:.*]] = arith.remui %[[LOAD_Z_TMP]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[LOAD_MUL_Y:.*]] = arith.muli %[[LOAD_Y]], %[[C16:.*]] : index
+    // CHECK-DAG: %[[LOAD_MUL_X:.*]] = arith.muli %[[LOAD_X]], %[[C32:.*]] : index
+    // CHECK-DAG: %[[LOAD_OFF_Z:.*]] = arith.remui %[[LOAD_Z]], %[[C8:.*]] : index
+    // CHECK-DAG: %[[LOAD_OFF_Y:.*]] = arith.remui %[[LOAD_MUL_Y]], %[[C128:.*]] : index
+    // CHECK-DAG: %[[LOAD_OFF_X:.*]] = arith.remui %[[LOAD_MUL_X]], %[[C256:.*]] : index
+    // CHECK-DAG: %[[LOAD_SLM:.*]] = xegpu.load_matrix %[[MDESC]][%[[LOAD_OFF_Z]], %[[LOAD_OFF_Y]], %[[LOAD_OFF_X]]] <{layout = #xegpu.layout<inst_data = [1, 16, 16]>}>: !xegpu.mem_desc<8x128x256xf32>, index, index, index -> vector<1x16x32xf32>
+    %offset = arith.constant {layout_result_0 = #xegpu.layout<sg_layout = [8, 4, 16], sg_data = [1, 32, 16], inst_data = [1, 16, 16]>} dense<0> : vector<8x128x256xindex>
+    %mask = arith.constant {layout_result_0 = #xegpu.layout<sg_layout = [8, 4, 16], sg_data = [1, 32, 16], inst_data = [1, 16, 16]>} dense<true> : vector<8x128x256xi1>
+    %1 = xegpu.load %arg0[%offset], %mask {chunk_size = 1, layout = #xegpu.layout<sg_layout = [8, 4, 16], sg_data = [1, 32, 16], inst_data = [1, 16, 16]>} : memref<?xf32>, vector<8x128x256xindex>, vector<8x128x256xi1> -> vector<8x128x256xf32>
+    %2 = xegpu.convert_layout %1 <{input_layout = #xegpu.layout<sg_layout = [8, 4, 16], sg_data = [1, 32, 16], inst_data = [1, 16, 16]>,
+                                   target_layout = #xegpu.layout<sg_layout = [8, 8, 8], sg_data = [1, 16, 32], inst_data = [1, 16, 16]>}> : vector<8x128x256xf32>
+    gpu.return
+  }
+
   // CHECK-LABEL: distribute_nested_slice
   // CHECK: %[[V0:.*]] = vector.shape_cast %{{.*}} : vector<32x32xf32> to vector<32x1x32x1xf32>
   // CHECK: %[[V1:.*]] = vector.broadcast %[[V0]] : vector<32x1x32x1xf32> to vector<32x16x32x16xf32>

>From 1bf0df3c93110565642e3db583699179cc77de73 Mon Sep 17 00:00:00 2001
From: nbpatel <nishant.b.patel at intel.com>
Date: Thu, 12 Feb 2026 15:57:03 +0000
Subject: [PATCH 3/4] Clean up

---
 .../XeGPU/Transforms/XeGPUWgToSgDistribute.cpp    | 15 ---------------
 1 file changed, 15 deletions(-)

diff --git a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
index 8dbca952cc8c2..2f240b974892e 100644
--- a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
+++ b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
@@ -627,21 +627,6 @@ struct WgToSgConvertLayoutOp
         targetLayout.getEffectiveSgLayoutAsInt();
     SmallVector<int64_t> targetSgData = targetLayout.getEffectiveSgDataAsInt();
 
-    auto hasUnitLeadingDims = [](ArrayRef<int64_t> shape) {
-      if (shape.size() <= 2)
-        return true;
-      for (size_t i = 0; i + 2 < shape.size(); ++i)
-        if (shape[i] != 1)
-          return false;
-      return true;
-    };
-
-    if (wgShape.size() > 2) {
-      if (!hasUnitLeadingDims(inputSgData) || !hasUnitLeadingDims(targetSgData))
-        return rewriter.notifyMatchFailure(
-            op, "rank > 2 requires unit leading dims for sg_data");
-    }
-
     // Fast path: if sg_layout and sg_data are identical, no SLM needed
     if (inputSgLayout == targetSgLayout && inputSgData == targetSgData) {
       inputLayout = inputLayout.dropSgLayoutAndData();

>From 9c436b36478139ad12a66e32765bb3e83a3b1160 Mon Sep 17 00:00:00 2001
From: nbpatel <nishant.b.patel at intel.com>
Date: Thu, 12 Feb 2026 16:12:24 +0000
Subject: [PATCH 4/4] address feedback

---
 .../Transforms/XeGPUWgToSgDistribute.cpp      |  3 +-
 .../XeGPU/xegpu-wg-to-sg-unify-ops.mlir       | 32 +++++++++++++++++--
 2 files changed, 32 insertions(+), 3 deletions(-)

diff --git a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
index d3c666d00e99e..88290d2e5bb9c 100644
--- a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
+++ b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
@@ -626,7 +626,8 @@ struct WgToSgConvertLayoutOp
     SmallVector<int64_t> targetSgData = targetLayout.getEffectiveSgDataAsInt();
 
     // Fast path: if sg_layout and sg_data are identical, no SLM needed
-    if (inputSgLayout == targetSgLayout && inputSgData == targetSgData) {
+    if (llvm::equal(inputSgLayout, targetSgLayout) &&
+        llvm::equal(inputSgData, targetSgData)) {
       inputLayout = inputLayout.dropSgLayoutAndData();
       targetLayout = targetLayout.dropSgLayoutAndData();
 
diff --git a/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir b/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir
index 8606e89616c91..e2e94c5f0300f 100644
--- a/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir
+++ b/mlir/test/Dialect/XeGPU/xegpu-wg-to-sg-unify-ops.mlir
@@ -828,6 +828,34 @@ gpu.module @test_distribution {
     gpu.return
   }
 
+  // CHECK-LABEL: convert_layout_no_slm
+  gpu.func @convert_layout_no_slm(%arg0: memref<4096x4096xf32>, %arg1: memref<4096x4096xf16>, %arg2: memref<4096x4096xf16>) {
+    %c32 = arith.constant 32 : index
+    %c4096 = arith.constant 4096 : index
+    %c0 = arith.constant 0 : index
+    %c256 = arith.constant 256 : index
+    %block_id_x = gpu.block_id  x
+    %block_id_y = gpu.block_id  y
+    %0 = arith.muli %block_id_x, %c256 overflow<nsw> : index
+    %1 = arith.muli %block_id_y, %c256 overflow<nsw> : index
+    %2 = xegpu.create_nd_tdesc %arg0 : memref<4096x4096xf32> -> !xegpu.tensor_desc<256x256xf32, #xegpu.block_tdesc_attr<boundary_check = false>, #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [8, 16]>>
+    %3 = xegpu.load_nd %2[%0, %1] <{layout = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [8, 16]>}> : !xegpu.tensor_desc<256x256xf32, #xegpu.block_tdesc_attr<boundary_check = false>, #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [8, 16]>> -> vector<256x256xf32>
+    %4 = xegpu.create_nd_tdesc %arg1 : memref<4096x4096xf16> -> !xegpu.tensor_desc<256x32xf16, #xegpu.block_tdesc_attr<boundary_check = false>, #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [32, 16]>>
+    %5 = xegpu.create_nd_tdesc %arg2 : memref<4096x4096xf16> -> !xegpu.tensor_desc<32x256xf16, #xegpu.block_tdesc_attr<boundary_check = false>, #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [32, 16]>>
+    %6 = scf.for %arg3 = %c0 to %c4096 step %c32 iter_args(%arg4 = %3) -> (vector<256x256xf32>) {
+      %7 = xegpu.load_nd %4[%0, %arg3] <{layout = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [32, 16]>}> : !xegpu.tensor_desc<256x32xf16, #xegpu.block_tdesc_attr<boundary_check = false>, #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [32, 16]>> -> vector<256x32xf16>
+      %8 = xegpu.load_nd %5[%arg3, %1] <{layout = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [32, 16]>}> : !xegpu.tensor_desc<32x256xf16, #xegpu.block_tdesc_attr<boundary_check = false>, #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [32, 16]>> -> vector<32x256xf16>
+      // CHECK: %[[CONVERT_A:.*]] = xegpu.convert_layout %{{.*}} <{input_layout = #xegpu.layout<inst_data = [32, 16]>, target_layout = #xegpu.layout<inst_data = [8, 16]>}> : vector<32x32xf16>
+      // CHECK: %[[CONVERT_B:.*]] = xegpu.convert_layout %{{.*}} <{input_layout = #xegpu.layout<inst_data = [32, 16]>, target_layout = #xegpu.layout<inst_data = [16, 16]>}> : vector<32x32xf16>
+      %9 = xegpu.convert_layout %7 <{input_layout = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [32, 16]>, target_layout = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [8, 16]>}> : vector<256x32xf16>
+      %10 = xegpu.convert_layout %8 <{input_layout = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [32, 16]>, target_layout = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [16, 16]>}> : vector<32x256xf16>
+      %11 = xegpu.dpas %9, %10, %arg4 {layout_a = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [8, 16]>, layout_b = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [16, 16]>, layout_cd = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [8, 16]>} : vector<256x32xf16>, vector<32x256xf16>, vector<256x256xf32> -> vector<256x256xf32>
+      scf.yield %11 : vector<256x256xf32>
+    } {layout_result_0 = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [8, 16]>}
+    xegpu.store_nd %6, %2[%0, %1] <{layout = #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [8, 16]>}> : vector<256x256xf32>, !xegpu.tensor_desc<256x256xf32, #xegpu.block_tdesc_attr<boundary_check = false>, #xegpu.layout<sg_layout = [8, 8], sg_data = [32, 32], inst_data = [8, 16]>>
+    gpu.return
+  }
+  
   // CHECK-LABEL: convert_layout_slm
   // CHECK-SAME: %[[ARG0:.*]]: memref<128x256xf32>
   gpu.func @convert_layout_slm(%arg0: memref<128x256xf32>) {
@@ -869,8 +897,8 @@ gpu.module @test_distribution {
   }
 
   gpu.func @convert_layout_3D(%arg0: memref<?xf32>) {
-    // CHECK-DAG: %[[CST:.*]] = arith.constant {layout_result_0 = #xegpu.layout<inst_data = [1, 16, 16]>} dense<0> : vector<1x32x16xindex>
-    // CHECK-DAG: %[[CST_0:.*]] = arith.constant {layout_result_0 = #xegpu.layout<inst_data = [1, 16, 16]>} dense<true> : vector<1x32x16xi1>
+    // CHECK-DAG: %[[CST:.*]] = arith.constant dense<0> : vector<1x32x16xindex>
+    // CHECK-DAG: %[[CST_0:.*]] = arith.constant dense<true> : vector<1x32x16xi1>
     // CHECK-DAG: %[[LOAD:.*]] = xegpu.load %{{.*}}[%[[CST]]], %[[CST_0]] <{chunk_size = 1 : i64, layout = #xegpu.layout<inst_data = [1, 16, 16]>}> : memref<?xf32>, vector<1x32x16xindex>, vector<1x32x16xi1> -> vector<1x32x16xf32>
     // CHECK-DAG: %[[ALLOCA:.*]] = memref.alloca() : memref<1048576xi8, 3>
     // CHECK-DAG: %[[MDESC:.*]] = xegpu.create_mem_desc %[[ALLOCA]] : memref<1048576xi8, 3> -> !xegpu.mem_desc<8x128x256xf32>



More information about the Mlir-commits mailing list