Use Lit for ML example makefiles (#1408)

Co-authored-by: Gagandeep Singh <gagan.posted@gmail.com> Co-authored-by: singagan <53442471+singagan@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Kristof Denolf <kristof.denolf@amd.com>
Xilinx · Apr 25, 2024 · a74a480 · a74a480
1 parent 6c7255f
commit a74a480
Show file tree

Hide file tree

Showing 29 changed files with 195 additions and 278 deletions.
diff --git a/programming_examples/ml/bottleneck/Makefile b/programming_examples/ml/bottleneck/Makefile
@@ -5,36 +5,41 @@
 #
 # Copyright (C) 2024, Advanced Micro Devices, Inc.
 
-include ../../makefile-common
+srcdir := $(shell dirname $(realpath $(firstword $(MAKEFILE_LIST))))
+
+include ${srcdir}/../../makefile-common
 
 mlirFileName = aie
 
+VPATH := ${srcdir}/../../../aie_kernels/aie2
+
 all: build/conv2dk1.o build/conv2dk3.o build/conv2dk1_skip.o build/final.xclbin
 
-build/${mlirFileName}.mlir: aie2.py
+build/${mlirFileName}.mlir: ${srcdir}/aie2.py
 	mkdir -p ${@D}
 	python3 $< > $@
 
 insts.txt: build/${mlirFileName}.mlir
 	aiecc.py -v --aie-only-generate-npu --npu-insts-name=$@ $<
 
-build/conv2dk1.o: ../../../aie_kernels/aie2/conv2dk1.cc
+build/conv2dk1.o: conv2dk1.cc
 	xchesscc -d ${CHESSCC2_FLAGS} -DINT8_ACT -c $< -o $@
 
-build/conv2dk3.o: ../../../aie_kernels/aie2/conv2dk3.cc
+build/conv2dk3.o: conv2dk3.cc
 	xchesscc -d ${CHESSCC2_FLAGS} -DUINT8_ACT -c $< -o $@
 
-build/conv2dk1_skip.o: ../../../aie_kernels/aie2/conv2dk1_skip.cc
+build/conv2dk1_skip.o: conv2dk1_skip.cc
 	xchesscc -d ${CHESSCC2_FLAGS} -DINT8_ACT -c $< -o $@
 
-build/final.xclbin: build/${mlirFileName}.mlir 
+build/final.xclbin: build/${mlirFileName}.mlir build/conv2dk1.o build/conv2dk3.o build/conv2dk1_skip.o  
 	cd build && aiecc.py -v --aie-generate-cdo --aie-generate-npu --no-compile-host \
-		--xclbin-name=${@F} --npu-insts-name=insts.txt $(<:%=../%)
+		--basic-alloc-scheme \
+		--xclbin-name=${@F} --npu-insts-name=insts.txt ${<F}
 
 clean:
 	rm -rf build *.elf* *.lst *.bif ${mlirFileName}.mlir.prj log .xclbin sim \
 		chess* *.o insts.txt \
 		*.log aie_partition.json *.bin BOOT.BIN _x test.exe
 
 run_py: 
-	${powershell} python3 test.py -x build/final.xclbin -i build/insts.txt -k MLIR_AIE
+	${powershell} python3 ${srcdir}/test.py -x build/final.xclbin -i build/insts.txt -k MLIR_AIE
diff --git a/programming_examples/ml/bottleneck/run.lit b/programming_examples/ml/bottleneck/run.lit
diff --git a/programming_examples/ml/bottleneck/run_makefile.lit b/programming_examples/ml/bottleneck/run_makefile.lit
@@ -0,0 +1,9 @@
+// (c) Copyright 2024 Advanced Micro Devices, Inc.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+// REQUIRES: ryzen_ai, chess, torch
+//
+// RUN: make -f %S/Makefile clean
+// RUN: make -f %S/Makefile 
+// RUN: %run_on_npu make -f %S/Makefile run_py | FileCheck %s
+// CHECK: PASS!
diff --git a/programming_examples/ml/bottleneck/test.py b/programming_examples/ml/bottleneck/test.py
@@ -189,14 +189,17 @@ def forward(self, x):
     # ------------------------------------------------------
     print("\nAvg NPU time: {}us.".format(int((npu_time_total / num_iter) / 1000)))
 
-    assert np.allclose(
+    if np.allclose(
         ofm_mem_fmt_out.detach().numpy(),
         golden_output.detach().numpy(),
         rtol=0,
         atol=inp_scale4,
-    )
-
-    print("\nPASS!\n")
+    ):
+        print("\nPASS!\n")
+        exit(0)
+    else:
+        print("\nFailed.\n")
+        exit(-1)
 
 
 if __name__ == "__main__":

diff --git a/programming_examples/ml/conv2d/Makefile b/programming_examples/ml/conv2d/Makefile
@@ -5,33 +5,33 @@
 #
 # Copyright (C) 2024, Advanced Micro Devices, Inc.
 
-include ../../makefile-common
+srcdir := $(shell dirname $(realpath $(firstword $(MAKEFILE_LIST))))
 
-mlirFileName = aieWithTrace_1core
+include ${srcdir}/../../makefile-common
 
-VPATH := ../../../aie_kernels/aie2
+mlirFileName = aie
+
+VPATH := ${srcdir}/../../../aie_kernels/aie2
 
 all: build/conv2dk1_i8.o build/final.xclbin
 
+build/conv2dk1_i8.o: conv2dk1_i8.cc
+	mkdir -p ${@D}
+	cd ${@D} && xchesscc_wrapper ${CHESSCCWRAP2_FLAGS} -DINT8_ACT -c $< -o ${@F}
 
-build/${mlirFileName}.mlir: aie2.py
+build/${mlirFileName}.mlir: ${srcdir}/aie2.py
 	mkdir -p ${@D}
 	python3 $< > $@
 
-
-insts.txt: build/${mlirFileName}.mlir
-	aiecc.py -v --aie-only-generate-npu --npu-insts-name=$@ $<
-
-build/conv2dk1_i8.o: conv2dk1_i8.cc
-	xchesscc -d ${CHESSCC2_FLAGS} -DINT8_ACT -c $< -o $@
-
-build/final.xclbin: build/${mlirFileName}.mlir 
-	cd build && aiecc.py -v --aie-generate-cdo --aie-generate-npu --no-compile-host \
+build/final.xclbin: build/${mlirFileName}.mlir build/conv2dk1_i8.o 
+	mkdir -p ${@D} 
+	cd ${@D} && aiecc.py -v --aie-generate-cdo --aie-generate-npu --no-compile-host \
 		--xclbin-name=${@F} --npu-insts-name=insts.txt $(<:%=../%)
 
+run_py: build/final.xclbin build/insts.txt
+	${powershell} python3 ${srcdir}/test.py -x build/final.xclbin -i build/insts.txt -k MLIR_AIE
+
 clean:
 	rm -rf build *.elf* *.lst *.bif ${mlirFileName}.mlir.prj log* *.xclbin sim \
 		chess* *.o insts.txt \
-		*.log aie_partition.json *.bin BOOT.BIN _x test.exe
-run_py: 
-	${powershell} python3 test.py -x build/final.xclbin -i build/insts.txt -k MLIR_AIE
+		*.log aie_partition.json *.bin BOOT.BIN _x test.exe
diff --git a/programming_examples/ml/conv2d/aie2.py b/programming_examples/ml/conv2d/aie2.py
@@ -34,10 +34,6 @@
 bufOut = actOut * 2  # double buffer
 actOutInt32s = actOut // 4
 
-enableTrace = False
-trace_size = 16384
-traceSizeInInt32s = trace_size // 4
-
 
 def conv2dk1():
     with mlir_mod_ctx() as ctx:
@@ -83,9 +79,6 @@ def device_body():
             ComputeTile2 = tile(0, 2)
             compute_tile2_col, compute_tile2_row = 0, 2
 
-            if enableTrace:
-                flow(ComputeTile2, WireBundle.Trace, 0, ShimTile, WireBundle.DMA, 1)
-
             # AIE-array data movement with object fifos
             # Input
             of_inOF_act_L3L2 = object_fifo(
@@ -155,85 +148,6 @@ def core_body():
 
             @FuncOp.from_py_func(tensor_ty, memRef_wts_ty, tensor_ty)
             def sequence(I, W, O):
-                if enableTrace:
-                    # 0x340D0: Trace Control 0
-                    #          0xAABB---C
-                    #            AA        <- Event to stop trace capture
-                    #              BB      <- Event to start trace capture
-                    #                   C  <- Trace mode, 00=event=time, 01=event-PC, 10=execution
-                    # Configure so that "Event 1" (always true) causes tracing to start
-                    npu_write32(
-                        column=compute_tile2_col,
-                        row=compute_tile2_row,
-                        address=0x340D0,
-                        value=0x00010000,
-                    )
-                    # 0x340D4: Trace Control 1
-                    npu_write32(
-                        column=compute_tile2_col,
-                        row=compute_tile2_row,
-                        address=0x340D4,
-                        value=0x00000000,
-                    )
-                    # 0x340E0: Trace Event Group 1  (Which events to trace)
-                    #          0xAABBCCDD    AA, BB, CC, DD <- four event slots
-                    npu_write32(
-                        column=compute_tile2_col,
-                        row=compute_tile2_row,
-                        address=0x340E0,
-                        value=0x4B222125,
-                    )
-                    # 0x340E4: Trace Event Group 2  (Which events to trace)
-                    #          0xAABBCCDD    AA, BB, CC, DD <- four event slots
-                    npu_write32(
-                        column=compute_tile2_col,
-                        row=compute_tile2_row,
-                        address=0x340E4,
-                        value=0x2D2C1A4F,
-                    )
-
-                    npu_write32(
-                        column=compute_tile2_col,
-                        row=compute_tile2_row,
-                        address=0x3FF00,
-                        value=0x00000121,
-                    )
-
-                    # Configure a buffer descriptor to write tracing information that has been routed into this shim tile
-                    # out to host DDR memory
-                    trace_bd_id = 13  # use BD 13 for writing trace output from compute tile to DDR host memory
-                    output_size = bufOut
-                    npu_writebd_shimtile(
-                        bd_id=trace_bd_id,
-                        buffer_length=trace_size,
-                        buffer_offset=output_size,
-                        enable_packet=0,
-                        out_of_order_id=0,
-                        packet_id=0,
-                        packet_type=0,
-                        column=0,
-                        column_num=1,
-                        d0_size=0,
-                        d0_stride=0,
-                        d1_size=0,
-                        d1_stride=0,
-                        d2_stride=0,
-                        ddr_id=2,
-                        iteration_current=0,
-                        iteration_size=0,
-                        iteration_stride=0,
-                        lock_acq_enable=0,
-                        lock_acq_id=0,
-                        lock_acq_val=0,
-                        lock_rel_id=0,
-                        lock_rel_val=0,
-                        next_bd=0,
-                        use_next_bd=0,
-                        valid_bd=1,
-                    )
-                    # Set start BD to our shim bd_Id (3)
-                    npu_write32(column=0, row=0, address=0x1D20C, value=trace_bd_id)
-
                 NpuWriteRTPOp("rtp2", col=0, row=2, index=0, value=10)
 
                 npu_dma_memcpy_nd(

diff --git a/programming_examples/ml/conv2d/run.lit b/programming_examples/ml/conv2d/run.lit
diff --git a/programming_examples/ml/conv2d/run_makefile.lit b/programming_examples/ml/conv2d/run_makefile.lit
@@ -0,0 +1,9 @@
+// (c) Copyright 2024 Advanced Micro Devices, Inc.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+// REQUIRES: ryzen_ai, chess, torch
+//
+// RUN: make -f %S/Makefile clean
+// RUN: make -f %S/Makefile 
+// RUN: %run_on_npu make -f %S/Makefile run_py | FileCheck %s
+// CHECK: PASS!
diff --git a/programming_examples/ml/conv2d/test.py b/programming_examples/ml/conv2d/test.py
@@ -147,13 +147,17 @@ def forward(self, x):
 
     print("\nAvg NPU time: {}us.".format(int((npu_time_total / num_iter) / 1000)))
 
-    assert np.allclose(
+    if np.allclose(
         ofm_mem_fmt_out.detach().numpy(),
         golden_output.detach().numpy(),
         rtol=0,
         atol=2 * int8_scale,
-    )
-    print("\nPASS!\n")
+    ):
+        print("\nPASS!\n")
+        exit(0)
+    else:
+        print("\nFailed.\n")
+        exit(-1)
 
 
 if __name__ == "__main__":

diff --git a/programming_examples/ml/conv2d_fused_relu/Makefile b/programming_examples/ml/conv2d_fused_relu/Makefile
@@ -5,15 +5,17 @@
 #
 # Copyright (C) 2024, Advanced Micro Devices, Inc.
 
-include ../../makefile-common
+srcdir := $(shell dirname $(realpath $(firstword $(MAKEFILE_LIST))))
+
+include ${srcdir}/../../makefile-common
 
 mlirFileName = aieWithTrace_1core
 
-VPATH := ../../../aie_kernels/aie2
+VPATH := ${srcdir}/../../../aie_kernels/aie2
 
 all: build/conv2dk1.o build/final.xclbin
 
-build/${mlirFileName}.mlir: aie2.py
+build/${mlirFileName}.mlir: ${srcdir}/aie2.py
 	mkdir -p ${@D}
 	python3 $< > $@
 
@@ -24,14 +26,14 @@ insts.txt: build/${mlirFileName}.mlir
 build/conv2dk1.o: conv2dk1.cc
 	xchesscc -d ${CHESSCC2_FLAGS} -DINT8_ACT -c $< -o $@
 
-build/final.xclbin: build/${mlirFileName}.mlir 
+build/final.xclbin: build/${mlirFileName}.mlir build/conv2dk1.o  
 	cd build && aiecc.py -v --aie-generate-cdo --aie-generate-npu --no-compile-host \
-		--xclbin-name=${@F} --npu-insts-name=insts.txt $(<:%=../%)
+		--xclbin-name=${@F} --npu-insts-name=insts.txt ${<F}
 
 clean:
 	rm -rf build *.elf* *.lst *.bif ${mlirFileName}.mlir.prj log* *.xclbin sim \
 		chess* *.o insts.txt \
 		*.log aie_partition.json *.bin BOOT.BIN _x test.exe
 
 run_py: 
-	${powershell} python3 test.py -x build/final.xclbin -i build/insts.txt -k MLIR_AIE
+	${powershell} python3  ${srcdir}/test.py -x build/final.xclbin -i build/insts.txt -k MLIR_AIE
diff --git a/programming_examples/ml/conv2d_fused_relu/run.lit b/programming_examples/ml/conv2d_fused_relu/run.lit
diff --git a/programming_examples/ml/conv2d_fused_relu/run_makefile.lit b/programming_examples/ml/conv2d_fused_relu/run_makefile.lit
@@ -0,0 +1,9 @@
+// (c) Copyright 2024 Advanced Micro Devices, Inc.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+// REQUIRES: ryzen_ai, chess, torch
+//
+// RUN: make -f %S/Makefile clean
+// RUN: make -f %S/Makefile 
+// RUN: %run_on_npu make -f %S/Makefile run_py | FileCheck %s
+// CHECK: PASS!
diff --git a/programming_examples/ml/conv2d_fused_relu/test.py b/programming_examples/ml/conv2d_fused_relu/test.py
@@ -148,14 +148,17 @@ def forward(self, x):
     # ------------------------------------------------------
     print("\nAvg NPU time: {}us.".format(int((npu_time_total / num_iter) / 1000)))
 
-    assert np.allclose(
+    if np.allclose(
         ofm_mem_fmt_out.detach().numpy(),
         golden_output.detach().numpy(),
         rtol=0,
         atol=2 * relu_scale,
-    )
-
-    print("\nPASS!\n")
+    ):
+        print("\nPASS!\n")
+        exit(0)
+    else:
+        print("\nFailed.\n")
+        exit(-1)
 
 
 if __name__ == "__main__":