76 lines
6.9 KiB
MLIR
76 lines
6.9 KiB
MLIR
// RUN: scalehls-opt -scalehls-tosa-node-fusion %s | FileCheck %s
|
|
|
|
module {
|
|
// CHECK: func @forward(%arg0: tensor<1x3x32x32xi8>) -> tensor<1x10xi8> {
|
|
func @forward(%arg0: tensor<1x3x32x32xi8>) -> tensor<1x10xi8> {
|
|
%0 = "tosa.const"() {value = dense<0> : tensor<1x10xi8>} : () -> tensor<1x10xi8>
|
|
%1 = "tosa.const"() {value = dense<1> : tensor<1x64x10xi8>} : () -> tensor<1x64x10xi8>
|
|
%2 = "tosa.const"() {value = dense<2> : tensor<64x3x3x64xi8>} : () -> tensor<64x3x3x64xi8>
|
|
%3 = "tosa.const"() {value = dense<3> : tensor<64x3x3x64xi8>} : () -> tensor<64x3x3x64xi8>
|
|
%4 = "tosa.const"() {value = dense<4> : tensor<64x3x3x3xi8>} : () -> tensor<64x3x3x3xi8>
|
|
%5 = "tosa.const"() {value = dense<[0, 3, 1, 2]> : tensor<4xi32>} : () -> tensor<4xi32>
|
|
%6 = "tosa.const"() {value = dense<[0, 2, 3, 1]> : tensor<4xi32>} : () -> tensor<4xi32>
|
|
%7 = "tosa.const"() {value = dense<5> : tensor<64xi8>} : () -> tensor<64xi8>
|
|
|
|
// CHECK: %8 = "tosa.transpose"(%arg0, %6) : (tensor<1x3x32x32xi8>, tensor<4xi32>) -> tensor<1x32x32x3xi8>
|
|
%8 = "tosa.transpose"(%arg0, %6) : (tensor<1x3x32x32xi8>, tensor<4xi32>) -> tensor<1x32x32x3xi8>
|
|
|
|
// CHECK: %9:2 = hls.dataflow.node() -> tensor<1x32x32x64xi8>, tensor<1x32x32x64xi8> {
|
|
// CHECK: %16 = "tosa.conv2d"(%8, %4, %7) {dilation = [1, 1], pad = [1, 1, 1, 1], quantization_info = {input_zp = 0 : i32, weight_zp = 0 : i32}, stride = [1, 1]} : (tensor<1x32x32x3xi8>, tensor<64x3x3x3xi8>, tensor<64xi8>) -> tensor<1x32x32x64xi8>
|
|
// CHECK: %17 = "tosa.clamp"(%16) {max_fp = 3.40282347E+38 : f32, max_int = 2147483647 : i64, min_fp = 0.000000e+00 : f32, min_int = 0 : i64} : (tensor<1x32x32x64xi8>) -> tensor<1x32x32x64xi8>
|
|
// CHECK: "hls.dataflow.output"(%16, %17) : (tensor<1x32x32x64xi8>, tensor<1x32x32x64xi8>) -> ()
|
|
// CHECK: }
|
|
%9 = "tosa.conv2d"(%8, %4, %7) {dilation = [1, 1], pad = [1, 1, 1, 1], quantization_info = {input_zp = 0 : i32, weight_zp = 0 : i32}, stride = [1, 1]} : (tensor<1x32x32x3xi8>, tensor<64x3x3x3xi8>, tensor<64xi8>) -> tensor<1x32x32x64xi8>
|
|
%10 = "tosa.clamp"(%9) {max_fp = 3.40282347E+38 : f32, max_int = 2147483647 : i64, min_fp = 0.000000e+00 : f32, min_int = 0 : i64} : (tensor<1x32x32x64xi8>) -> tensor<1x32x32x64xi8>
|
|
|
|
// CHECK: %10:2 = hls.dataflow.node() -> tensor<1x32x32x64xi8>, tensor<1x32x32x64xi8> {
|
|
// CHECK: %16 = "tosa.conv2d"(%9#1, %3, %7) {dilation = [1, 1], pad = [1, 1, 1, 1], quantization_info = {input_zp = 0 : i32, weight_zp = 0 : i32}, stride = [1, 1]} : (tensor<1x32x32x64xi8>, tensor<64x3x3x64xi8>, tensor<64xi8>) -> tensor<1x32x32x64xi8>
|
|
// CHECK: %17 = "tosa.clamp"(%16) {max_fp = 3.40282347E+38 : f32, max_int = 2147483647 : i64, min_fp = 0.000000e+00 : f32, min_int = 0 : i64} : (tensor<1x32x32x64xi8>) -> tensor<1x32x32x64xi8>
|
|
// CHECK: "hls.dataflow.output"(%16, %17) : (tensor<1x32x32x64xi8>, tensor<1x32x32x64xi8>) -> ()
|
|
// CHECK: }
|
|
%11 = "tosa.conv2d"(%10, %3, %7) {dilation = [1, 1], pad = [1, 1, 1, 1], quantization_info = {input_zp = 0 : i32, weight_zp = 0 : i32}, stride = [1, 1]} : (tensor<1x32x32x64xi8>, tensor<64x3x3x64xi8>, tensor<64xi8>) -> tensor<1x32x32x64xi8>
|
|
%12 = "tosa.clamp"(%11) {max_fp = 3.40282347E+38 : f32, max_int = 2147483647 : i64, min_fp = 0.000000e+00 : f32, min_int = 0 : i64} : (tensor<1x32x32x64xi8>) -> tensor<1x32x32x64xi8>
|
|
|
|
// CHECK: %11 = hls.dataflow.node() -> tensor<1x32x32x64xi8> {
|
|
// CHECK: %16 = "tosa.conv2d"(%10#1, %2, %7) {dilation = [1, 1], pad = [1, 1, 1, 1], quantization_info = {input_zp = 0 : i32, weight_zp = 0 : i32}, stride = [1, 1]} : (tensor<1x32x32x64xi8>, tensor<64x3x3x64xi8>, tensor<64xi8>) -> tensor<1x32x32x64xi8>
|
|
// CHECK: "hls.dataflow.output"(%16) : (tensor<1x32x32x64xi8>) -> ()
|
|
// CHECK: }
|
|
%13 = "tosa.conv2d"(%12, %2, %7) {dilation = [1, 1], pad = [1, 1, 1, 1], quantization_info = {input_zp = 0 : i32, weight_zp = 0 : i32}, stride = [1, 1]} : (tensor<1x32x32x64xi8>, tensor<64x3x3x64xi8>, tensor<64xi8>) -> tensor<1x32x32x64xi8>
|
|
|
|
// CHECK: %12:2 = hls.dataflow.node() -> tensor<1x32x32x64xi8>, tensor<1x32x32x64xi8> {
|
|
// CHECK: %16 = "tosa.add"(%11, %9#1) : (tensor<1x32x32x64xi8>, tensor<1x32x32x64xi8>) -> tensor<1x32x32x64xi8>
|
|
// CHECK: %17 = "tosa.clamp"(%16) {max_fp = 3.40282347E+38 : f32, max_int = 2147483647 : i64, min_fp = 0.000000e+00 : f32, min_int = 0 : i64} : (tensor<1x32x32x64xi8>) -> tensor<1x32x32x64xi8>
|
|
// CHECK: "hls.dataflow.output"(%16, %17) : (tensor<1x32x32x64xi8>, tensor<1x32x32x64xi8>) -> ()
|
|
// CHECK: }
|
|
%14 = "tosa.add"(%13, %10) : (tensor<1x32x32x64xi8>, tensor<1x32x32x64xi8>) -> tensor<1x32x32x64xi8>
|
|
%15 = "tosa.clamp"(%14) {max_fp = 3.40282347E+38 : f32, max_int = 2147483647 : i64, min_fp = 0.000000e+00 : f32, min_int = 0 : i64} : (tensor<1x32x32x64xi8>) -> tensor<1x32x32x64xi8>
|
|
|
|
// CHECK: %13:2 = hls.dataflow.node() -> tensor<1x1x1x64xi8>, tensor<1x64x1x1xi8> {
|
|
// CHECK: %16 = "tosa.avg_pool2d"(%12#1) {kernel = [32, 32], pad = [0, 0, 0, 0], quantization_info = {input_zp = 0 : i32, output_zp = 0 : i32}, stride = [32, 32]} : (tensor<1x32x32x64xi8>) -> tensor<1x1x1x64xi8>
|
|
// CHECK: %17 = "tosa.transpose"(%16, %5) : (tensor<1x1x1x64xi8>, tensor<4xi32>) -> tensor<1x64x1x1xi8>
|
|
// CHECK: "hls.dataflow.output"(%16, %17) : (tensor<1x1x1x64xi8>, tensor<1x64x1x1xi8>) -> ()
|
|
// CHECK: }
|
|
%16 = "tosa.avg_pool2d"(%15) {kernel = [32, 32], pad = [0, 0, 0, 0], quantization_info = {input_zp = 0 : i32, output_zp = 0 : i32}, stride = [32, 32]} : (tensor<1x32x32x64xi8>) -> tensor<1x1x1x64xi8>
|
|
%17 = "tosa.transpose"(%16, %5) : (tensor<1x1x1x64xi8>, tensor<4xi32>) -> tensor<1x64x1x1xi8>
|
|
|
|
// CHECK: %14:2 = hls.dataflow.node() -> tensor<1x1x64xi8>, tensor<1x1x10xi8> {
|
|
// CHECK: %16 = "tosa.reshape"(%13#1) {new_shape = [1, 1, 64]} : (tensor<1x64x1x1xi8>) -> tensor<1x1x64xi8>
|
|
// CHECK: %17 = "tosa.matmul"(%16, %1) {quantization_info = {a_zp = 0 : i32, b_zp = 0 : i32}} : (tensor<1x1x64xi8>, tensor<1x64x10xi8>) -> tensor<1x1x10xi8>
|
|
// CHECK: "hls.dataflow.output"(%16, %17) : (tensor<1x1x64xi8>, tensor<1x1x10xi8>) -> ()
|
|
// CHECK: }
|
|
%18 = "tosa.reshape"(%17) {new_shape = [1, 1, 64]} : (tensor<1x64x1x1xi8>) -> tensor<1x1x64xi8>
|
|
%19 = "tosa.matmul"(%18, %1) {quantization_info = {a_zp = 0 : i32, b_zp = 0 : i32}} : (tensor<1x1x64xi8>, tensor<1x64x10xi8>) -> tensor<1x1x10xi8>
|
|
|
|
// CHECK: %15:2 = hls.dataflow.node() -> tensor<1x10xi8>, tensor<1x10xi8> {
|
|
// CHECK: %16 = "tosa.reshape"(%14#1) {new_shape = [1, 10]} : (tensor<1x1x10xi8>) -> tensor<1x10xi8>
|
|
// CHECK: %17 = "tosa.add"(%16, %0) : (tensor<1x10xi8>, tensor<1x10xi8>) -> tensor<1x10xi8>
|
|
// CHECK: "hls.dataflow.output"(%16, %17) : (tensor<1x10xi8>, tensor<1x10xi8>) -> ()
|
|
// CHECK: }
|
|
%20 = "tosa.reshape"(%19) {new_shape = [1, 10]} : (tensor<1x1x10xi8>) -> tensor<1x10xi8>
|
|
%21 = "tosa.add"(%20, %0) : (tensor<1x10xi8>, tensor<1x10xi8>) -> tensor<1x10xi8>
|
|
|
|
// CHECK: return %15#1 : tensor<1x10xi8>
|
|
return %21 : tensor<1x10xi8>
|
|
}
|
|
}
|