Enable Some Programming Examples for Strix (#1952)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: [email protected] <Joe Melber>

aie_kernels/aie2/mul.cc

@@ -8,10 +8,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define __AIENGINE__ 2
-#define NOCPP
-#define __AIEARCH__ 20
-
 #include <stdint.h>
 #include <stdio.h>
 #include <stdlib.h>

programming_examples/basic/passthrough_kernel/Makefile

@@ -12,6 +12,7 @@ srcdir := $(shell dirname $(realpath $(firstword $(MAKEFILE_LIST))))
 
 include ${srcdir}/../../makefile-common
 
+device = npu
 targetname = passThroughKernel
 VPATH := ${srcdir}/../../../aie_kernels/generic
 data_size = 4096
@@ -31,27 +32,44 @@ all: build/final_${data_size}.xclbin
 
 build/aie2_lineBased_8b_${data_size}.mlir: ${srcdir}/${aie_py_src}
 	mkdir -p ${@D}
-	python3 $< ${data_size} 0 > $@
+	python3 $< ${device} ${data_size} 0 > $@
 
 build/aie_trace__lineBased_8b_${data_size}.mlir: ${srcdir}/${aie_py_src}
 	mkdir -p ${@D}
-	python3 $< ${data_size} ${trace_size} > $@
+	python3 $< ${device} ${data_size} ${trace_size} > $@
 
 build/passThrough.cc.o: passThrough.cc
 	mkdir -p ${@D}
+ifeq ($(device),npu)
 	cd ${@D} && ${PEANO_INSTALL_DIR}/bin/clang++ ${PEANOWRAP2_FLAGS} -DBIT_WIDTH=8 -c $< -o ${@F}
+else ifeq ($(device),npu2)
+	cd ${@D} && xchesscc_wrapper ${CHESSCCWRAP2P_FLAGS} -DBIT_WIDTH=8 -c $< -o ${@F}
+else
+	echo "Device type not supported"
+endif
 
 build/final_${data_size}.xclbin: build/aie2_lineBased_8b_${data_size}.mlir build/passThrough.cc.o
 	mkdir -p ${@D}
+ifeq ($(device),npu)
 	cd ${@D} && aiecc.py --aie-generate-cdo --aie-generate-npu --no-compile-host \
 		--no-xchesscc --no-xbridge \
 		--xclbin-name=${@F} --npu-insts-name=insts_${data_size}.txt $(<:%=../%)
+else
+	cd ${@D} && aiecc.py --aie-generate-cdo --aie-generate-npu --no-compile-host \
+		--xclbin-name=${@F} --npu-insts-name=insts_${data_size}.txt $(<:%=../%)
+endif
 
 build/final_trace_${data_size}.xclbin: build/aie2_lineBased_8b_${data_size}.mlir build/passThrough.cc.o
 	mkdir -p ${@D}
+ifeq ($(device),npu)
 	cd ${@D} && aiecc.py --aie-generate-cdo --aie-generate-npu --no-compile-host \
 		--no-xchesscc --no-xbridge \
 		--xclbin-name=${@F} --npu-insts-name=insts_${data_size}.txt $(<:%=../%)
+else
+	cd ${@D} && aiecc.py --aie-generate-cdo --aie-generate-npu --no-compile-host \
+		--xclbin-name=${@F} --npu-insts-name=insts_${data_size}.txt $(<:%=../%)
+endif
+
 
 ${targetname}_${data_size}.exe: ${srcdir}/test.cpp
 	rm -rf _build

programming_examples/basic/passthrough_kernel/aie2.py

@@ -16,11 +16,11 @@
 import aie.utils.trace as trace_utils
 
 
-def passthroughKernel(vector_size, trace_size):
+def passthroughKernel(dev, vector_size, trace_size):
     N = vector_size
     lineWidthInBytes = N // 4  # chop input in 4 sub-tensors
 
-    @device(AIEDevice.npu1_1col)
+    @device(dev)
     def device_body():
         # define types
         vector_ty = np.ndarray[(N,), np.dtype[np.uint8]]
@@ -55,8 +55,6 @@ def core_body():
                 of_in.release(ObjectFifoPort.Consume, 1)
                 of_out.release(ObjectFifoPort.Produce, 1)
 
-        #    print(ctx.module.operation.verify())
-
         @runtime_sequence(vector_ty, vector_ty, vector_ty)
         def sequence(inTensor, outTensor, notUsed):
             if trace_size > 0:
@@ -85,13 +83,20 @@ def sequence(inTensor, outTensor, notUsed):
 
 
 try:
-    vector_size = int(sys.argv[1])
+    device_name = str(sys.argv[1])
+    if device_name == "npu":
+        dev = AIEDevice.npu1_1col
+    elif device_name == "npu2":
+        dev = AIEDevice.npu2
+    else:
+        raise ValueError("[ERROR] Device name {} is unknown".format(sys.argv[1]))
+    vector_size = int(sys.argv[2])
     if vector_size % 64 != 0 or vector_size < 512:
         print("Vector size must be a multiple of 64 and greater than or equal to 512")
         raise ValueError
-    trace_size = 0 if (len(sys.argv) != 3) else int(sys.argv[2])
+    trace_size = 0 if (len(sys.argv) != 4) else int(sys.argv[3])
 except ValueError:
     print("Argument has inappropriate value")
 with mlir_mod_ctx() as ctx:
-    passthroughKernel(vector_size, trace_size)
+    passthroughKernel(dev, vector_size, trace_size)
     print(ctx.module)

programming_examples/basic/passthrough_kernel/aie2_alt.py

@@ -16,11 +16,11 @@
 import aie.utils.trace as trace_utils
 
 
-def passthroughKernel(vector_size, trace_size):
+def passthroughKernel(dev, vector_size, trace_size):
     N = vector_size
     lineWidthInBytes = N // 4  # chop input in 4 sub-tensors
 
-    @device(AIEDevice.npu1_1col)
+    @device(dev)
     def device_body():
         # define types
         vector_ty = np.ndarray[(N,), np.dtype[np.uint8]]
@@ -79,13 +79,20 @@ def sequence(inTensor, outTensor, notUsed):
 
 
 try:
-    vector_size = int(sys.argv[1])
+    device_name = str(sys.argv[1])
+    if device_name == "npu":
+        dev = AIEDevice.npu1_1col
+    elif device_name == "npu2":
+        dev = AIEDevice.npu2
+    else:
+        raise ValueError("[ERROR] Device name {} is unknown".format(sys.argv[1]))
+    vector_size = int(sys.argv[2])
     if vector_size % 64 != 0 or vector_size < 512:
         print("Vector size must be a multiple of 64 and greater than or equal to 512")
         raise ValueError
-    trace_size = 0 if (len(sys.argv) != 3) else int(sys.argv[2])
+    trace_size = 0 if (len(sys.argv) != 4) else int(sys.argv[3])
 except ValueError:
     print("Argument has inappropriate value")
 with mlir_mod_ctx() as ctx:
-    passthroughKernel(vector_size, trace_size)
+    passthroughKernel(dev, vector_size, trace_size)
     print(ctx.module)

programming_examples/basic/passthrough_kernel/run_makefile.lit

@@ -3,9 +3,9 @@
 //
 // REQUIRES: ryzen_ai, peano 
 //
+// RUN: mkdir -p test
+// RUN: cd test
 // RUN: make -f %S/Makefile clean
 // RUN: make -f %S/Makefile 
-// RUN: %run_on_npu make -f %S/Makefile run | FileCheck %s
-// RUN: %run_on_npu make -f %S/Makefile run_py | FileCheck %s
-// CHECK: Running...
-// CHECK: PASS!
+// RUN: %run_on_npu make -f %S/Makefile run
+// RUN: %run_on_npu make -f %S/Makefile run_py

programming_examples/basic/passthrough_kernel/run_makefile_alt.lit

@@ -7,6 +7,5 @@
 // RUN: cd test_alt
 // RUN: make -f %S/Makefile clean
 // RUN: env use_alt=1 make -f %S/Makefile 
-// RUN: %run_on_npu make -f %S/Makefile run_py | FileCheck %s
-// CHECK: PASS!
+// RUN: %run_on_npu make -f %S/Makefile run_py
 

programming_examples/basic/passthrough_kernel/run_strix_makefile.lit

@@ -0,0 +1,10 @@
+// (c) Copyright 2024 Advanced Micro Devices, Inc.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+// REQUIRES: ryzen_ai, chess
+//
+// RUN: mkdir -p test_stx
+// RUN: cd test_stx
+// RUN: make -f %S/Makefile clean
+// RUN: make -f %S/Makefile device=npu2 
+// RUN: %run_on_2npu make -f %S/Makefile run device=npu2

programming_examples/basic/passthrough_kernel/test.py

@@ -40,11 +40,11 @@ def main(opts):
 
     if not errors:
         print("\nPASS!\n")
-        exit(0)
+        sys.exit(0)
     else:
         print("\nError count: ", errors)
         print("\nFailed.\n")
-        exit(-1)
+        sys.exit(1)
 
 
 if __name__ == "__main__":

programming_examples/basic/vector_scalar_mul/Makefile

@@ -14,6 +14,7 @@ include ${srcdir}/../../makefile-common
 
 VPATH := ${srcdir}/../../../aie_kernels/aie2
 
+device = npu
 targetname = vectorScalar
 data_size = 4096
 trace_size = 8192
@@ -32,19 +33,25 @@ kristof: build/insts_${data_size}.txt
 
 build/%.o: %.cc
 	mkdir -p ${@D}
+ifeq ($(device),npu)
 ifeq ($(CHESS), true)
 	cd ${@D} && xchesscc_wrapper ${CHESSCCWRAP2_FLAGS} -c $< -o ${@F}; 
 else 
 	cd ${@D} && ${PEANO_INSTALL_DIR}/bin/clang++ ${PEANOWRAP2_FLAGS} -c $< -o ${@F}; 
 endif
+else ifeq ($(device),npu2)
+	cd ${@D} && xchesscc_wrapper ${CHESSCCWRAP2P_FLAGS} -DBIT_WIDTH=8 -c $< -o ${@F}
+else
+	echo "Device type not supported"
+endif
 
 build/aie_${data_size}.mlir: ${srcdir}/${aie_py_src}
 	mkdir -p ${@D}
-	python3 $< ${data_size} 0 > $@
+	python3 $< ${device} ${data_size} 0 > $@
 
 build/aie_trace_${data_size}.mlir: ${srcdir}/${aie_py_src}
 	mkdir -p ${@D}
-	python3 $< ${data_size} ${trace_size} > $@
+	python3 $< ${device} ${data_size} ${trace_size} > $@
 
 #build/insts_${data_size}.txt: build/final_${data_size}.xclbin
 build/final_${data_size}.xclbin: build/aie_${data_size}.mlir build/scale.o

programming_examples/basic/vector_scalar_mul/aie2.py

@@ -16,7 +16,7 @@
 import aie.utils.trace as trace_utils
 
 
-def my_vector_scalar(vector_size, trace_size):
+def my_vector_scalar(dev, vector_size, trace_size):
     N = vector_size
     N_in_bytes = N * 2
     N_div_n = 4  # chop input vector into 4 sub-vectors
@@ -26,7 +26,7 @@ def my_vector_scalar(vector_size, trace_size):
 
     vectorized = True
 
-    @device(AIEDevice.npu1_1col)
+    @device(dev)
     def device_body():
         tensor_ty = np.ndarray[(N,), np.dtype[np.int16]]
         tile_ty = np.ndarray[(n,), np.dtype[np.int16]]
@@ -93,13 +93,20 @@ def sequence(A, F, C):
 
 
 try:
-    vector_size = int(sys.argv[1])
+    device_name = str(sys.argv[1])
+    if device_name == "npu":
+        dev = AIEDevice.npu1_1col
+    elif device_name == "npu2":
+        dev = AIEDevice.npu2
+    else:
+        raise ValueError("[ERROR] Device name {} is unknown".format(sys.argv[1]))
+    vector_size = int(sys.argv[2])
     if vector_size % 64 != 0 or vector_size < 512:
         print("Vector size must be a multiple of 64 and greater than or equal to 512")
         raise ValueError
-    trace_size = 0 if (len(sys.argv) != 3) else int(sys.argv[2])
+    trace_size = 0 if (len(sys.argv) != 4) else int(sys.argv[3])
 except ValueError:
     print("Argument has inappropriate value")
 with mlir_mod_ctx() as ctx:
-    my_vector_scalar(vector_size, trace_size)
-    print(ctx.module)
+    my_vector_scalar(dev, vector_size, trace_size)
+print(ctx.module)

programming_examples/basic/vector_scalar_mul/aie2_alt.py

@@ -16,7 +16,7 @@
 import aie.utils.trace as trace_utils
 
 
-def my_vector_scalar(vector_size, trace_size):
+def my_vector_scalar(dev, vector_size, trace_size):
     N = vector_size
     N_in_bytes = N * 2
     N_div_n = 4  # chop input vector into 4 sub-vectors
@@ -26,7 +26,7 @@ def my_vector_scalar(vector_size, trace_size):
 
     vectorized = True
 
-    @device(AIEDevice.npu1_1col)
+    @device(dev)
     def device_body():
         tensor_ty = np.ndarray[(N,), np.dtype[np.int16]]
         tile_ty = np.ndarray[(n,), np.dtype[np.int16]]
@@ -97,13 +97,20 @@ def sequence(A, F, C):
 
 
 try:
-    vector_size = int(sys.argv[1])
+    device_name = str(sys.argv[1])
+    if device_name == "npu":
+        dev = AIEDevice.npu1_1col
+    elif device_name == "npu2":
+        dev = AIEDevice.npu2
+    else:
+        raise ValueError("[ERROR] Device name {} is unknown".format(sys.argv[1]))
+    vector_size = int(sys.argv[2])
     if vector_size % 64 != 0 or vector_size < 512:
         print("Vector size must be a multiple of 64 and greater than or equal to 512")
         raise ValueError
-    trace_size = 0 if (len(sys.argv) != 3) else int(sys.argv[2])
+    trace_size = 0 if (len(sys.argv) != 4) else int(sys.argv[3])
 except ValueError:
     print("Argument has inappropriate value")
 with mlir_mod_ctx() as ctx:
-    my_vector_scalar(vector_size, trace_size)
-    print(ctx.module)
+    my_vector_scalar(dev, vector_size, trace_size)
+print(ctx.module)

programming_examples/basic/vector_scalar_mul/run_makefile.lit

@@ -7,9 +7,8 @@
 // RUN: cd test_peano
 // RUN: make -f %S/Makefile clean
 // RUN: env CHESS=false make -f %S/Makefile 
-// RUN: %run_on_npu make -f %S/Makefile run | FileCheck %s
-// RUN: %run_on_npu make -f %S/Makefile run_py | FileCheck %s
+// RUN: %run_on_npu make -f %S/Makefile run 
+// RUN: %run_on_npu make -f %S/Makefile run_py 
 // RUN: make -f %S/Makefile clean
-// RUN: env CHESS=false %run_on_npu make -f %S/Makefile trace | FileCheck %s
-// RUN: env CHESS=false %run_on_npu make -f %S/Makefile trace_py | FileCheck %s
-// CHECK: PASS!
+// RUN: env CHESS=false %run_on_npu make -f %S/Makefile trace 
+// RUN: env CHESS=false %run_on_npu make -f %S/Makefile trace_py 

programming_examples/basic/vector_scalar_mul/run_makefile_alt.lit

@@ -7,6 +7,5 @@
 // RUN: cd test_alt
 // RUN: make -f %S/Makefile clean
 // RUN: env CHESS=true use_alt=1 make -f %S/Makefile 
-// RUN: %run_on_npu make -f %S/Makefile run | FileCheck %s
-// CHECK: PASS!
-
+// RUN: %run_on_npu make -f %S/Makefile run 
+

programming_examples/basic/vector_scalar_mul/run_makefile_chess.lit

@@ -7,9 +7,8 @@
 // RUN: cd test_chess
 // RUN: make -f %S/Makefile clean
 // RUN: env CHESS=true make -f %S/Makefile 
-// RUN: %run_on_npu make -f %S/Makefile run | FileCheck %s
-// RUN: %run_on_npu make -f %S/Makefile run_py | FileCheck %s
+// RUN: %run_on_npu make -f %S/Makefile run
+// RUN: %run_on_npu make -f %S/Makefile run_py
 // RUN: make -f %S/Makefile clean
-// RUN: env CHESS=true %run_on_npu make -f %S/Makefile trace | FileCheck %s
-// RUN: env CHESS=true %run_on_npu make -f %S/Makefile trace_py | FileCheck %s
-// CHECK: PASS!
+// RUN: env CHESS=true %run_on_npu make -f %S/Makefile trace
+// RUN: env CHESS=true %run_on_npu make -f %S/Makefile trace_py

programming_examples/basic/vector_scalar_mul/run_strix_makefile.lit

@@ -0,0 +1,10 @@
+// (c) Copyright 2024 Advanced Micro Devices, Inc.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+// REQUIRES: ryzen_ai, chess
+//
+// RUN: mkdir -p test_stx
+// RUN: cd test_stx
+// RUN: make -f %S/Makefile clean
+// RUN: make -f %S/Makefile device=npu2 
+// RUN: %run_on_2npu make -f %S/Makefile run device=npu2

programming_examples/basic/vector_scalar_mul/test.py

@@ -71,11 +71,11 @@ def main(opts):
 
     if not errors:
         print("\nPASS!\n")
-        exit(0)
+        sys.exit(0)
     else:
         print("\nError count: ", errors)
         print("\nFailed.\n")
-        exit(-1)
+        sys.exit(1)
 
 
 if __name__ == "__main__":