Fix reduce min (#1263)

aie_kernels/aie2/reduce_max.cc

@@ -8,12 +8,12 @@
 void _reduce_max_vector(int32_t *restrict in, int32_t *restrict out,
                         const int32_t input_size) {
 
-  v16int32 tiny = broadcast_to_v16int32((int32_t)-2147483648);
+  v16int32 tiny = broadcast_to_v16int32((int32_t)INT32_MIN);
   const int32_t vector_size = 16;
   v16int32 after_vector;
   v16int32 running_max = tiny;
   for (int32_t i = 0; i < input_size; i += vector_size)
-    chess_prepare_for_pipelining chess_loop_range(64, 64) {
+    chess_prepare_for_pipelining chess_loop_range(8, ) {
       v16int32 next = *(v16int32 *)(in + i);
       v16int32 test = max(running_max, next);
       running_max = test;
@@ -34,7 +34,7 @@ void _reduce_max_vector(int32_t *restrict in, int32_t *restrict out,
 
 void _reduce_max_scalar(int32_t *restrict in, int32_t *restrict out,
                         const int32_t input_size) {
-  int32_t running_max = (int32_t)-2147483648;
+  int32_t running_max = (int32_t)INT32_MIN;
   for (int32_t i = 0; i < input_size; i++) {
     if (in[i] > running_max)
       running_max = in[i];

aie_kernels/generic/vector_min.cc → aie_kernels/aie2/reduce_min.cc

@@ -5,15 +5,15 @@
 
 #include <aie_api/aie.hpp>
 
-void vector(int32_t *restrict in, int32_t *restrict out) {
+void _reduce_min_vector(int32_t *restrict in, int32_t *restrict out,
+                        const int32_t input_size) {
 
-  v16int32 massive = broadcast_to_v16int32((int32_t)2147483647);
-  int32_t input_size = 1024;
-  int32_t vector_size = 16;
+  v16int32 massive = broadcast_to_v16int32((int32_t)INT32_MAX);
+  const int32_t vector_size = 16;
   v16int32 after_vector;
   v16int32 running_min = massive;
   for (int32_t i = 0; i < input_size; i += vector_size)
-    chess_prepare_for_pipelining chess_loop_range(64, 64) {
+    chess_prepare_for_pipelining chess_loop_range(8, ) {
       v16int32 next = *(v16int32 *)(in + i);
       v16int32 test = min(running_min, next);
       running_min = test;
@@ -32,9 +32,9 @@ void vector(int32_t *restrict in, int32_t *restrict out) {
   return;
 }
 
-void scalar(int32_t *restrict in, int32_t *restrict out) {
-  size_t input_size = 1024;
-  int32_t running_min = (int32_t)2147483647;
+void _reduce_min_scalar(int32_t *restrict in, int32_t *restrict out,
+                        const int32_t input_size) {
+  int32_t running_min = (int32_t)INT32_MAX;
   for (int32_t i = 0; i < input_size; i++) {
     if (in[i] < running_min)
       running_min = in[i];
@@ -46,8 +46,12 @@ void scalar(int32_t *restrict in, int32_t *restrict out) {
 
 extern "C" {
 
-void vector_min(int32_t *a_in, int32_t *c_out) { vector(a_in, c_out); }
+void reduce_min_vector(int32_t *a_in, int32_t *c_out, int32_t input_size) {
+  _reduce_min_vector(a_in, c_out, input_size);
+}
 
-void scalar_min(int32_t *a_in, int32_t *c_out) { scalar(a_in, c_out); }
+void reduce_min_scalar(int32_t *a_in, int32_t *c_out, int32_t input_size) {
+  _reduce_min_scalar(a_in, c_out, input_size);
+}
 
 } // extern "C"

programming_examples/basic/vector_min/CMakeLists.txt → programming_examples/basic/reduce_min/CMakeLists.txt

@@ -40,15 +40,17 @@ project(${ProjectName})
 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 
+    ../../utils
+    ${CMAKE_CURRENT_SOURCE_DIR}/../../../runtime_lib/test_lib
     ${XRT_INC_DIR}
     ${Boost_INCLUDE_DIRS}
-    ../../../programming_examples/utils
 )
 
 target_link_directories(${currentTarget} PUBLIC
@@ -66,4 +68,4 @@ else()
     target_link_libraries(${currentTarget} PUBLIC
         xrt_coreutil
     )
-endif()
+endif()

programming_examples/basic/vector_min_reduce/Makefile → programming_examples/basic/reduce_min/Makefile

@@ -10,23 +10,23 @@ include ../../makefile-common
 
 ACDC_AIE = $(dir $(shell which aie-opt))/..
 
-targetname = vector_min
+targetname = reduce_min
 devicename = ipu
 col = 0
 CHESS_FLAGS=${CHESSCCWRAP2_FLAGS}
-KERNEL_LIB=${ACDC_AIE}/../../aie_kernels/generic/
+KERNEL_LIB=../../../aie_kernels/aie2
 
 all: build/final.xclbin build/insts.txt
 
-build/vector_min.o: ${KERNEL_LIB}/vector_min.cc
+build/reduce_min.cc.o: ${KERNEL_LIB}/reduce_min.cc
 	mkdir -p ${@D}
-	cd ${@D} && xchesscc_wrapper ${CHESS_FLAGS} -DBIT_WIDTH=8 -c $< -o ${@F}
+	cd ${@D} && xchesscc_wrapper ${CHESS_FLAGS} -c $(<:%=../%) -o ${@F}
 
 build/aie.mlir: aie2.py
 	mkdir -p ${@D}
 	python3 $< ${devicename} ${col} > $@
 
-build/final.xclbin: build/aie.mlir build/vector_min.o
+build/final.xclbin: build/aie.mlir build/reduce_min.cc.o
 	mkdir -p ${@D}
 	cd ${@D} && aiecc.py --aie-generate-cdo --no-compile-host --xclbin-name=${@F} \
 				--aie-generate-ipu --ipu-insts-name=insts.txt $(<:%=../%)
@@ -73,4 +73,4 @@ vck5000: build/aie.mlir build/scale.o
 		-Wl,--whole-archive -Wl,--no-whole-archive -lstdc++ -ldl -lelf -o test.elf
 
 clean: clean_trace
-	rm -rf build _build inst aie.mlir.prj core_* test.elf ${targetname}.exe vector_min.o* vector_min.cc
+	rm -rf build _build inst aie.mlir.prj core_* test.elf ${targetname}.exe

programming_examples/basic/vector_min/aie2.py → programming_examples/basic/reduce_min/aie2.py

@@ -16,7 +16,7 @@
 import sys
 
 
-def my_vector_max():
+def my_reduce_min():
     N = 1024
 
     buffer_depth = 2
@@ -39,10 +39,9 @@ def device_body():
             memRef_O_ty = T.memref(1, T.i32())
 
             # AIE Core Function declarations
-
-            vector_min = external_func("vector_min", inputs=[memRef_I_ty, memRef_O_ty])
-
-            scalar_min = external_func("scalar_min", inputs=[memRef_I_ty, memRef_O_ty])
+            reduce_min_vector = external_func(
+                "reduce_min_vector", inputs=[memRef_I_ty, memRef_O_ty, T.i32()]
+            )
 
             # Tile declarations
             ShimTile = tile(int(sys.argv[2]), 0)
@@ -57,16 +56,12 @@ def device_body():
             # Set up compute tiles
 
             # Compute tile 2
-            @core(ComputeTile2, "vector_min.o")
+            @core(ComputeTile2, "reduce_min.cc.o")
             def core_body():
                 for _ in for_(0xFFFFFFFF):
                     elem_out = of_out.acquire(ObjectFifoPort.Produce, 1)
                     elem_in = of_in.acquire(ObjectFifoPort.Consume, 1)
-
-                    call(
-                        vector_min,
-                        [elem_in, elem_out],
-                    )
+                    call(reduce_min_vector, [elem_in, elem_out, N])
                     of_in.release(ObjectFifoPort.Consume, 1)
                     of_out.release(ObjectFifoPort.Produce, 1)
                     yield_([])
@@ -83,4 +78,4 @@ def sequence(A, C):
     print(ctx.module)
 
 
-my_vector_max()
+my_reduce_min()

programming_examples/basic/vector_min/run.lit → programming_examples/basic/reduce_min/run.lit

@@ -1,12 +1,11 @@
 // (c) Copyright 2023 Advanced Micro Devices, Inc.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
-// REQUIRES: ryzen_ai
+// REQUIRES: ryzen_ai, chess
 //
-// RUN: %python %S/aie2.py ipu 0 > ./aie.mlir
-// RUN: xchesscc_wrapper aie2 -I %aietools/include -c %S/../../../aie_kernels/generic/vector_min.cc -o vector_min.o
+// RUN: xchesscc_wrapper aie2 -I %aietools/include -c %S/../../../aie_kernels/aie2/reduce_min.cc -o reduce_min.cc.o
+// RUN: %python %S/aie2.py ipu 0 | aie-opt -cse -canonicalize -o ./aie.mlir
 // RUN: %python aiecc.py --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 -I%S/../../../runtime_lib/test_lib %S/../../../runtime_lib/test_lib/test_utils.cpp %xrt_flags -lrt -lstdc++ -lboost_program_options -lboost_filesystem
+// RUN: g++ %S/test.cpp -o test.exe -std=c++23 -Wall -I%S/../../../runtime_lib/test_lib %S/../../../runtime_lib/test_lib/test_utils.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/vector_min/test.cpp → programming_examples/basic/reduce_min/test.cpp

@@ -52,11 +52,8 @@ int main(int argc, const char *argv[]) {
   int n_warmup_iterations = vm["warmup"].as<int>();
   int trace_size = vm["trace_sz"].as<int>();
 
-  // ------------------------------------------------------
-  // Configure this to match your design's buffer size
-  // ------------------------------------------------------
-  int INOUT0_VOLUME = 1024; // Input only, 64x uint32_t in this example
-  int INOUT1_VOLUME = 1;    // Not used in this example
+  int INOUT0_VOLUME = 1024;
+  int INOUT1_VOLUME = 1;
 
   size_t INOUT0_SIZE = INOUT0_VOLUME * sizeof(INOUT0_DATATYPE);
   size_t INOUT1_SIZE = INOUT1_VOLUME * sizeof(INOUT1_DATATYPE);
@@ -76,46 +73,12 @@ int main(int argc, const char *argv[]) {
   // Get device, load the xclbin & kernel and register them
   // ------------------------------------------------------
   // Get a device handle
-  unsigned int device_index = 0;
-  auto device = xrt::device(device_index);
-
-  // Load the xclbin
-  if (verbosity >= 1)
-    std::cout << "Loading xclbin: " << vm["xclbin"].as<std::string>() << "\n";
-  auto xclbin = xrt::xclbin(vm["xclbin"].as<std::string>());
-
-  // Load the kernel
-  if (verbosity >= 1)
-    std::cout << "Kernel opcode: " << vm["kernel"].as<std::string>() << "\n";
-  std::string Node = vm["kernel"].as<std::string>();
-
-  // Get the kernel from the xclbin
-  auto xkernels = xclbin.get_kernels();
-  auto xkernel = *std::find_if(xkernels.begin(), xkernels.end(),
-                               [Node, verbosity](xrt::xclbin::kernel &k) {
-                                 auto name = k.get_name();
-                                 if (verbosity >= 1) {
-                                   std::cout << "Name: " << name << std::endl;
-                                 }
-                                 return name.rfind(Node, 0) == 0;
-                               });
-  auto kernelName = xkernel.get_name();
-
-  // Register xclbin
-  if (verbosity >= 1)
-    std::cout << "Registering xclbin: " << vm["xclbin"].as<std::string>()
-              << "\n";
-  device.register_xclbin(xclbin);
+  xrt::device device;
+  xrt::kernel kernel;
 
-  // Get a hardware context
-  if (verbosity >= 1)
-    std::cout << "Getting hardware context.\n";
-  xrt::hw_context context(device, xclbin.get_uuid());
-
-  // Get a kernel handle
-  if (verbosity >= 1)
-    std::cout << "Getting handle to kernel:" << kernelName << "\n";
-  auto kernel = xrt::kernel(context, kernelName);
+  test_utils::init_xrt_load_kernel(device, kernel, verbosity,
+                                   vm["xclbin"].as<std::string>(),
+                                   vm["kernel"].as<std::string>());
 
   // ------------------------------------------------------
   // Initialize input/ output buffer sizes and sync them
@@ -136,7 +99,7 @@ int main(int argc, const char *argv[]) {
 
   // Initialize Inout buffer 0
   INOUT0_DATATYPE *bufInOut0 = bo_inout0.map<INOUT0_DATATYPE *>();
-  std::int32_t min = (std::int32_t)2147483647;
+  std::int32_t min = (std::int32_t)INT32_MAX;
   for (int i = 0; i < INOUT0_VOLUME; i++) {
     std::int32_t next = test_utils::random_int32_t(100000);
     if (next < min)