rename and clean up

programming_examples/basic/passthrough_hardware/Makefile → programming_examples/basic/passthrough_dmas/Makefile

@@ -10,7 +10,7 @@ include ../makefile-common
 
 SHELL := /bin/bash
 
-targetname = passThroughHardware
+targetname = passThroughDMAs
 LENGTH ?= 4096
 
 all: build/final.xclbin build/insts.txt
@@ -47,4 +47,4 @@ run: ${targetname}.exe build/final.xclbin build/insts.txt
 	${powershell} ./$< -x build/final.xclbin -i build/insts.txt -k MLIR_AIE -l ${LENGTH}
 
 clean:
-	rm -rf build _build inst
+	rm -rf build _build inst ${targetname}.exe

programming_examples/basic/passthrough_hardware/aie2.py → programming_examples/basic/passthrough_dmas/aie2.py

@@ -14,7 +14,6 @@
 from aie.extras.context import mlir_mod_ctx
 
 N = 4096
-N_in_bytes = N * 4
 
 if len(sys.argv) == 2:
     N = int(sys.argv[1])
@@ -41,9 +40,8 @@ def device_body():
             # Compute tile 2
             @core(ComputeTile2)
             def core_body():
-                tmp = memref.alloc(1, T.i32())
-                v0 = arith.constant(0, T.i32())
-                memref.store(v0, tmp, [0])
+                for _ in for_(sys.maxsize):
+                    yield_([]) 
 
             # To/from AIE-array data movement
             tensor_ty = T.memref(N, T.i32())

programming_examples/basic/passthrough_kernel/CMakeLists.txt

@@ -2,11 +2,10 @@
 # See https://llvm.org/LICENSE.txt for license information.
 # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 #
-# (c) Copyright 2023 Xilinx Inc.
+# (c) Copyright 2024 Xilinx Inc.
 
 # parameters
 # -DBOOST_ROOT: Path to Boost install
-# -DOpenCV_DIR: Path to OpenCV 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

programming_examples/basic/passthrough_kernel/Makefile

@@ -12,7 +12,7 @@ VPATH := ../../../aie_kernels/aie_generic
 
 PASSTHROUGH_SIZE = 4096
 
-targetname = passthrough_kernel
+targetname = passThroughKernel
 
 .PHONY: all template clean
 

programming_examples/basic/passthrough_kernel/aie2.py

@@ -3,7 +3,7 @@
 # See https://llvm.org/LICENSE.txt for license information.
 # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 #
-# (c) Copyright 2023 AMD Inc.
+# (c) Copyright 2024 AMD Inc.
 
 import sys
 
@@ -12,30 +12,30 @@
 from aie.dialects.scf import *
 from aie.extras.context import mlir_mod_ctx
 
-width = 1024  
+N = 1024  
 
 if len(sys.argv) == 2:
-    width = int(sys.argv[1])
+    N = int(sys.argv[1])
 
-lineWidthInBytes = width
+lineWidthInBytes = N
 lineWidthInInt32s = lineWidthInBytes // 4
 
 enableTrace = False
 traceSizeInBytes = 8192
 traceSizeInInt32s = traceSizeInBytes // 4
 
 
-def passThroughAIE2():
+def passthroughKernel():
     with mlir_mod_ctx() as ctx:
 
         @device(AIEDevice.ipu)
         def device_body():
             # define types
-            line_ty = T.memref(lineWidthInBytes, T.ui8())
+            memRef_ty = T.memref(lineWidthInBytes, T.ui8())
 
             # AIE Core Function declarations
             passThroughLine = external_func(
-                "passThroughLine", inputs=[line_ty, line_ty, T.i32()]
+                "passThroughLine", inputs=[memRef_ty, memRef_ty, T.i32()]
             )
 
             # Tile declarations
@@ -46,8 +46,8 @@ def device_body():
                 flow(ComputeTile2, "Trace", 0, ShimTile, "DMA", 1)
 
             # AIE-array data movement with object fifos
-            of_in = object_fifo("in", ShimTile, ComputeTile2, 2, line_ty)
-            of_out = object_fifo("out", ComputeTile2, ShimTile, 2, line_ty)
+            of_in = object_fifo("in", ShimTile, ComputeTile2, 2, memRef_ty)
+            of_out = object_fifo("out", ComputeTile2, ShimTile, 2, memRef_ty)
 
             # Set up compute tiles
 
@@ -57,16 +57,16 @@ def core_body():
                 for _ in for_(sys.maxsize):
                     elemOut = of_out.acquire(ObjectFifoPort.Produce, 1)
                     elemIn = of_in.acquire(ObjectFifoPort.Consume, 1)
-                    call(passThroughLine, [elemIn, elemOut, width])
+                    call(passThroughLine, [elemIn, elemOut, lineWidthInBytes])
                     of_in.release(ObjectFifoPort.Consume, 1)
                     of_out.release(ObjectFifoPort.Produce, 1)
                     yield_([])
 
             #    print(ctx.module.operation.verify())
 
-            tensorSize = width 
+            tensorSize = N 
             tensorSizeInInt32s = tensorSize // 4
-            tensor_ty = T.memref(tensorSizeInInt32s, T.i32())
+            tensor_ty = T.memref(lineWidthInInt32s, T.i32())
 
             @FuncOp.from_py_func(tensor_ty, tensor_ty, tensor_ty)
             def sequence(inTensor, outTensor, notUsed):
@@ -150,20 +150,10 @@ def sequence(inTensor, outTensor, notUsed):
                     )
                     IpuWrite32(0, 0, 0x1D20C, 0x3)
 
-                ipu_dma_memcpy_nd(
-                    metadata="in",
-                    bd_id=0,
-                    mem=inTensor,
-                    sizes=[1, 1, 1, tensorSizeInInt32s],
-                )
-                ipu_dma_memcpy_nd(
-                    metadata="out",
-                    bd_id=1,
-                    mem=outTensor,
-                    sizes=[1, 1, 1, tensorSizeInInt32s],
-                )
+                ipu_dma_memcpy_nd(metadata="in", bd_id=0, mem=inTensor, sizes=[1, 1, 1, tensorSizeInInt32s])
+                ipu_dma_memcpy_nd(metadata="out", bd_id=1, mem=outTensor, sizes=[1, 1, 1, tensorSizeInInt32s])
                 ipu_sync(column=0, row=0, direction=0, channel=0)
 
     print(ctx.module)
 
-passThroughAIE2()
+passthroughKernel()

programming_examples/basic/passthrough_kernel/run.lit

@@ -0,0 +1,12 @@
+// (c) Copyright 2023 Advanced Micro Devices, Inc.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+// REQUIRES: ryzen_ai, chess
+//
+// RUN: xchesscc_wrapper aie2 -I %aietools/include -DBIT_WIDTH=8 -c %S/../../../aie_kernels/aie_generic/passThrough.cc -o passThrough.cc.o
+// RUN: %python %S/aie2.py 4096 | aie-opt -cse -canonicalize -o ./aie.mlir
+// RUN: %python aiecc.py --xbridge --aie-generate-cdo --aie-generate-ipu --no-compile-host --xclbin-name=aie.xclbin --ipu-insts-name=insts.txt ./aie.mlir
+// RUN: clang %S/test.cpp -o test.exe -std=c++11 -Wall -DPASSTHROUGH_SIZE=4096 -I%S/../../utils %S/../../utils/xrtUtils.cpp %xrt_flags -lrt -lstdc++ -lboost_program_options -lboost_filesystem
+// RUN: %run_on_ipu ./test.exe -x aie.xclbin -k MLIR_AIE -i insts.txt | FileCheck %s
+// CHECK: PASS!
+

programming_examples/basic/passthrough_kernel/test.cpp

@@ -109,22 +109,21 @@ int main(int argc, const char *argv[]) {
   bo_out.sync(XCL_BO_SYNC_BO_FROM_DEVICE);
 
   // Compare out to in
-  int numberOfDifferences = 0;
+  int errors = 0;
   for(int i = 0; i < PASSTHROUGH_SIZE; i++) {
     if(bufOut[i] != bufInA[i])
-      numberOfDifferences++;
+      errors++;
   }
 
   // Print Pass/Fail result of our test
-  int res = 0;
-  if (numberOfDifferences == 0) {
-    printf("PASS!\n");
-    res = 0;
+  if (!errors) {
+    std::cout << std::endl << "PASS!" << std::endl << std::endl;
+    return 0;
   } else {
-    printf("Fail!\n");
-    res = -1;
+    std::cout << std::endl
+              << errors << " mismatches." << std::endl
+              << std::endl;
+    std::cout << std::endl << "fail." << std::endl << std::endl;
+    return 1;
   }
-
-  printf("Testing passThrough done!\n");
-  return res;
 }