Attribute to specify repeat count in ObjFIFO (Xilinx#1530)

Co-authored-by: AndraBisca <andrab@amd.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

include/aie/Dialect/AIE/IR/AIEOps.td

@@ -1668,7 +1668,9 @@ def AIE_ObjectFifoCreateOp: AIE_Op<"objectfifo", [HasParent<"DeviceOp">, Symbol]
         DefaultValuedAttr<BoolAttr, "false">:$plio,
         // via_shared_mem==0 means use producer tile's memory module
         // via_shared_mem==1 means use consumer tile's memory module
-        OptionalAttr<AIEI32Attr>:$via_shared_mem
+        OptionalAttr<AIEI32Attr>:$via_shared_mem,
+        // memtile_repeat==0 means "do it once" and don't repeat
+        OptionalAttr<AIEI32Attr>:$memtile_repeat
   );
 
   let assemblyFormat = [{
@@ -1763,16 +1765,10 @@ def AIE_ObjectFifoLinkOp: AIE_Op<"objectfifo.link", [HasParent<"DeviceOp">]> {
   let extraClassDeclaration = [{
     std::vector<ObjectFifoCreateOp> getInputObjectFifos();
     std::vector<ObjectFifoCreateOp> getOutputObjectFifos();
-
-    bool isJoin() {
-      return getFifoIns().size() > 1;
-    }
-
-    bool isDistribute() {
-      return getFifoOuts().size() > 1;
-    }
-
+    bool isJoin() { return getFifoIns().size() > 1; }
+    bool isDistribute() { return getFifoOuts().size() > 1; }
     std::optional<mlir::Value> getOptionalSharedTile();
+    std::optional<int> getRepeatCount();
     std::vector<int> getJoinTranferLengths();
     std::vector<int> getDistributeTranferLengths();
   }];

lib/Dialect/AIE/IR/AIEDialect.cpp

@@ -493,6 +493,12 @@ LogicalResult ObjectFifoCreateOp::verify() {
           "`via_shared_mem` can only be used in 1-to-1 object FIFOs");
   }
 
+  if (getMemtileRepeat().has_value()) {
+    if (!getProducerTileOp().isMemTile())
+      return emitError("`memtile_repeat` can only be used with a mem tile "
+                       "producer");
+  }
+
   return success();
 }
 
@@ -625,6 +631,26 @@ LogicalResult ObjectFifoLinkOp::verify() {
         return emitOpError("currently does not support objectFifos with "
                            "dimensionsFromStreamPerConsumer.");
     }
+
+    for (auto fifoOut : getOutputObjectFifos()) {
+      for (auto dims : fifoOut.getDimensionsFromStreamPerConsumer()) {
+        if (!dims.empty())
+          return emitOpError("currently does not support objectFifos with "
+                             "dimensionsFromStreamPerConsumer.");
+      }
+    }
+
+    std::vector<int> repeat_counts;
+    for (auto fifoOut : getOutputObjectFifos()) {
+      if (fifoOut.getMemtileRepeat().has_value())
+        repeat_counts.push_back(fifoOut.getMemtileRepeat().value());
+      else
+        repeat_counts.push_back(0);
+    }
+    for (auto repeat : repeat_counts)
+      if (repeat_counts[0] != repeat)
+        return emitError("repeat counts of output object FIFOs must be equal");
+
   } else {
     if (!getSrcOffsets().empty() && !getDstOffsets().empty())
       return emitOpError("all offsets should be empty if there is no "
@@ -732,6 +758,13 @@ std::vector<int> ObjectFifoLinkOp::getDistributeTranferLengths() {
   return lengths;
 }
 
+std::optional<int> ObjectFifoLinkOp::getRepeatCount() {
+  for (auto fifoOut : getOutputObjectFifos())
+    if (fifoOut.getMemtileRepeat().has_value())
+      return {fifoOut.getMemtileRepeat().value()};
+  return {};
+}
+
 //===----------------------------------------------------------------------===//
 // ObjectFifoRegisterExternalBuffersOp
 //===----------------------------------------------------------------------===//

lib/Dialect/AIE/Transforms/AIEObjectFifoStatefulTransform.cpp

@@ -427,6 +427,8 @@ struct AIEObjectFifoStatefulTransformPass
       of_elem_index++;
     }
     if (linked) {
+      if (linkOp->getRepeatCount().has_value())
+        numElem *= linkOp->getRepeatCount().value() + 1;
       if (linkOp->isDistribute())
         numElem *= linkOp->getFifoOuts().size();
       else if (linkOp->isJoin())
@@ -692,6 +694,19 @@ struct AIEObjectFifoStatefulTransformPass
     int acqNum = 1;
     int relNum = 1;
 
+    // check for repeat count
+    int repeatCount = 0;
+    if (!dims.getValue().empty()) {
+      auto highestStride = dims.getValue().begin()->getStride() - 1;
+      if (highestStride == 0) {
+        repeatCount = dims.getValue().begin()->getSize();
+        dims = AIE::BDDimLayoutArrayAttr::get(op->getContext(),
+                                              dims.getValue().drop_front(1));
+      }
+    }
+    if (op.getMemtileRepeat().has_value())
+      repeatCount = op.getMemtileRepeat().value();
+
     // search for the buffers/locks (based on if this objFifo has a link)
     // identify size difference between input and output memrefs
     ObjectFifoCreateOp target = op;
@@ -704,12 +719,20 @@ struct AIEObjectFifoStatefulTransformPass
         auto srcOffsets = linkOp->getSrcOffsets();
         auto dstOffsets = linkOp->getDstOffsets();
 
+        if (target == op) {
+          if (linkOp->getRepeatCount().has_value()) {
+            // +1 for original data movement
+            acqNum *= linkOp->getRepeatCount().value() + 1;
+            relNum *= linkOp->getRepeatCount().value() + 1;
+          }
+        }
+
         if (linkOp->isJoin()) {
           // compute offset and length
           isJoin = true;
           if (target == op) {
-            acqNum = linkOp->getFifoIns().size();
-            relNum = linkOp->getFifoIns().size();
+            acqNum *= linkOp->getFifoIns().size();
+            relNum *= linkOp->getFifoIns().size();
           } else {
             int i = 0;
             for (auto fifoIn : linkOp->getInputObjectFifos()) {
@@ -718,15 +741,14 @@ struct AIEObjectFifoStatefulTransformPass
               i++;
             }
             extraOffset = *getConstantIntValue(srcOffsets[i]);
-            if (dims.getValue().empty())
-              lenOut = linkOp->getJoinTranferLengths()[i];
+            lenOut = linkOp->getJoinTranferLengths()[i];
           }
         } else if (linkOp->isDistribute()) {
           // compute offset and length
           isDistribute = true;
           if (target == op) {
-            acqNum = linkOp->getFifoOuts().size();
-            relNum = linkOp->getFifoOuts().size();
+            acqNum *= linkOp->getFifoOuts().size();
+            relNum *= linkOp->getFifoOuts().size();
           } else {
             int i = 0;
             for (auto fifoOut : linkOp->getOutputObjectFifos()) {
@@ -735,8 +757,7 @@ struct AIEObjectFifoStatefulTransformPass
               i++;
             }
             extraOffset = *getConstantIntValue(dstOffsets[i]);
-            if (dims.getValue().empty())
-              lenOut = linkOp->getDistributeTranferLengths()[i];
+            lenOut = linkOp->getDistributeTranferLengths()[i];
           }
         } else {
           if (target != op) {
@@ -787,17 +808,6 @@ struct AIEObjectFifoStatefulTransformPass
     Block *dmaBlock = builder.createBlock(endBlock);
     Block *bdBlock = builder.createBlock(endBlock);
 
-    // check for repeat count in objfifo dims
-    int repeatCount = 0;
-    if (!dims.getValue().empty()) {
-      auto highestStride = dims.getValue().begin()->getStride();
-      if (highestStride == 0) {
-        repeatCount = dims.getValue().begin()->getSize();
-        dims = AIE::BDDimLayoutArrayAttr::get(op->getContext(),
-                                              dims.getValue().drop_front(1));
-      }
-    }
-
     // create DMA channel
     builder.setInsertionPointToStart(dmaBlock);
     builder.create<DMAStartOp>(builder.getUnknownLoc(), channelDir,

programming_examples/basic/memtile_repeat/distribute_repeat/CMakeLists.txt

@@ -0,0 +1,75 @@
+# This file is licensed under the Apache License v2.0 with LLVM Exceptions.
+# See https://llvm.org/LICENSE.txt for license information.
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+#
+# (c) Copyright 2023 Advanced Micro Devices, Inc.
+
+# parameters
+# -DBOOST_ROOT: Path to Boost install
+# -DXRT_INC_DIR: Full path to src/runtime_src/core/include in XRT cloned repo
+# -DXRT_LIB_DIR: Path to xrt_coreutil.lib
+# -DTARGET_NAME: Target name to be built
+
+# cmake needs this line
+cmake_minimum_required(VERSION 3.1)
+
+set(CMAKE_CXX_STANDARD 23) 
+set(CMAKE_CXX_STANDARD_REQUIRED YES)
+
+find_program(WSL NAMES powershell.exe)
+
+if (NOT WSL)
+    set(CMAKE_C_COMPILER gcc-13)
+    set(CMAKE_CXX_COMPILER g++-13)
+    set(BOOST_ROOT /usr/include/boost CACHE STRING "Path to Boost install")
+    set(XRT_INC_DIR /opt/xilinx/xrt/include CACHE STRING "Path to XRT cloned repo")
+    set(XRT_LIB_DIR /opt/xilinx/xrt/lib CACHE STRING "Path to xrt_coreutil.lib")
+else()
+    set(BOOST_ROOT C:/Technical/thirdParty/boost_1_83_0 CACHE STRING "Path to Boost install")
+    set(XRT_INC_DIR C:/Technical/XRT/src/runtime_src/core/include CACHE STRING "Path to XRT cloned repo")
+    set(XRT_LIB_DIR C:/Technical/xrtNPUfromDLL CACHE STRING "Path to xrt_coreutil.lib")
+endif()
+
+set(TARGET_NAME test CACHE STRING "Target to be built")
+
+SET (ProjectName proj_${TARGET_NAME})
+SET (currentTarget ${TARGET_NAME})
+
+if ( WSL )
+	set(CMAKE_RUNTIME_OUTPUT_DIRECTORY_RELEASE ${CMAKE_BINARY_DIR})
+endif ()
+
+project(${ProjectName})
+
+# Find packages
+find_package(Boost REQUIRED)
+
+add_executable(${currentTarget}
+    ${CMAKE_CURRENT_SOURCE_DIR}/../../../../runtime_lib/test_lib/test_utils.cpp
+    test.cpp
+)
+
+target_compile_definitions(${currentTarget} PUBLIC DISABLE_ABI_CHECK=1)
+
+target_include_directories (${currentTarget} PUBLIC 
+    ${XRT_INC_DIR}
+    ${Boost_INCLUDE_DIRS}
+    ${CMAKE_CURRENT_SOURCE_DIR}/../../../../runtime_lib/test_lib
+)
+
+target_link_directories(${currentTarget} PUBLIC
+    ${XRT_LIB_DIR}
+    ${Boost_LIBRARY_DIRS}
+)
+
+if (NOT WSL)
+    target_link_libraries(${currentTarget} PUBLIC
+        xrt_coreutil
+        boost_program_options
+        boost_filesystem
+    )
+else()
+    target_link_libraries(${currentTarget} PUBLIC
+        xrt_coreutil
+    )
+endif()

programming_examples/basic/memtile_repeat/distribute_repeat/Makefile

@@ -0,0 +1,69 @@
+##===- Makefile -----------------------------------------------------------===##
+# 
+# This file licensed under the Apache License v2.0 with LLVM Exceptions.
+# See https://llvm.org/LICENSE.txt for license information.
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+# 
+##===----------------------------------------------------------------------===##
+
+srcdir := $(shell dirname $(realpath $(firstword $(MAKEFILE_LIST))))
+
+include ${srcdir}/../../../makefile-common
+
+all: build/final.xclbin build/insts.txt
+
+devicename ?= npu
+targetname = distribute_repeat
+LENGTH ?= 36
+
+build/aie.mlir: ${srcdir}/aie2.py
+	mkdir -p ${@D}
+	python3 $< ${LENGTH} ${devicename} ${col} > $@
+
+.PHONY: inst/insts.txt
+inst/insts.txt: ${srcdir}/aie2.py
+	rm -rf inst
+	mkdir -p inst
+	python3 $< ${LENGTH} > inst/aie.mlir
+	pushd inst && aiecc.py --aie-only-generate-npu --npu-insts-name=insts.txt aie.mlir && popd
+	${powershell} ./build/${targetname}.exe -x build/final.xclbin -i inst/insts.txt -k MLIR_AIE -l ${LENGTH}
+
+build/final.xclbin: build/aie.mlir
+	mkdir -p ${@D}
+	cd ${@D} && aiecc.py --aie-generate-cdo --no-compile-host --xclbin-name=${@F} \
+				--aie-generate-npu --npu-insts-name=insts.txt $(<:%=../%)
+
+${targetname}.exe: ${srcdir}/test.cpp
+	rm -rf _build
+	mkdir -p _build
+	cd _build && ${powershell} cmake ${srcdir} -DTARGET_NAME=${targetname}
+	cd _build && ${powershell} cmake --build . --config Release
+ifeq "${powershell}" "powershell.exe"
+	cp _build/${targetname}.exe $@
+else
+	cp _build/${targetname} $@ 
+endif
+
+run: ${targetname}.exe build/final.xclbin build/insts.txt
+	${powershell} ./$< -x build/final.xclbin -i build/insts.txt -k MLIR_AIE -l ${LENGTH}
+
+# Changing variables when we target VCK5000
+vck5000: devicename=xcvc1902
+vck5000: col=6
+
+vck5000: build/aie.mlir
+	aiecc.py --link_against_hsa --host-target=x86_64-amd-linux-gnu build/aie.mlir \
+		-I/opt/xaiengine/include \
+		-I${srcdir}/../../../../install/runtime_lib/x86_64-hsa/test_lib/include \
+		-L/opt/xaiengine/lib \
+		-L/lib/x86_64-linux-gnu/ \
+		${srcdir}/test_vck5000.cpp \
+		${srcdir}/../../../../install/runtime_lib/x86_64-hsa/test_lib/src/test_library.cpp \
+		-Wl,-R/opt/xaiengine/lib \
+		-Wl,--whole-archive -Wl,--no-whole-archive -lstdc++ -ldl -lelf -o test.elf
+
+run_vck5000:
+	test.elf
+
+clean:
+	rm -rf build _build inst ${targetname}.exe

programming_examples/basic/memtile_repeat/distribute_repeat/README.md

@@ -0,0 +1,21 @@
+<!---//===- README.md --------------------------*- Markdown -*-===//
+//
+// This file is licensed under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+// Copyright (C) 2024, Advanced Micro Devices, Inc.
+// 
+//===----------------------------------------------------------------------===//-->
+
+# <ins>Distribute Repeat</ins>
+
+This reference design can be run on a Ryzen™ AI NPU.
+
+TODO
+
+To compile and run the design for NPU:
+```
+make
+make run
+```