การเรียกที่กำหนดเองของ XLA

// Handler that always returns an error.
auto always_error = Ffi::Bind().To(
    []() { return Error(ErrorCode::kInternal, "Oops!"); });

// Handler that always returns a success.
auto always_success = Ffi::Bind().To(
    []() { return Error::Success(); });

%0 = "stablehlo.custom_call"(%arg0) {
  call_target_name = "foo",
  api_version = 4 : i32
} : (tensor<2x2xf32>) -> tensor<2x2xf32>

// Buffers of any rank and data type.
auto handler = Ffi::Bind().Arg<AnyBuffer>().Ret<AnyBuffer>().To(
    [](AnyBuffer arg, Result<AnyBuffer> res) -> Error {
      void* arg_data = arg.untyped_data();
      void* res_data = res->untyped_data();
      return Error::Success();
    });

// Buffers of any rank and F32 data type.
auto handler = Ffi::Bind().Arg<Buffer<F32>>().Ret<Buffer<F32>>().To(
    [](Buffer<F32> arg, Result<Buffer<F32>> res) -> Error {
      float* arg_data = arg.typed_data();
      float* res_data = res->typed_data();
      return Error::Success();
    });

// Buffers of rank 2 and F32 data type.
auto handler = Ffi::Bind().Arg<BufferR2<F32>>().Ret<BufferR2<F32>>().To(
    [](BufferR2<F32> arg, Result<BufferR2<F32>> res) -> Error {
      float* arg_data = arg.typed_data();
      float* res_data = res->typed_data();
      return Error::Success();
    });

auto handler = Ffi::Bind().RemainingArgs().RemainingRets().To(
    [](RemainingArgs args, RemainingRets results) -> Error {
      ErrorOr<AnyBuffer> arg = args.get<AnyBuffer>(0);
      ErrorOr<Result<AnyBuffer>> res = results.get<AnyBuffer>(0);

      if (!arg.has_value()) {
        return Error(ErrorCode::kInternal, arg.error());
      }

      if (!res.has_value()) {
        return Error(ErrorCode::kInternal, res.error());
      }

      return Error::Success();
    });

auto handler =
    Ffi::Bind()
        .Arg<AnyBuffer>()
        .RemainingArgs()
        .Ret<AnyBuffer>()
        .RemainingRets()
        .To([](AnyBuffer arg, RemainingArgs args, AnyBuffer ret,
               RemainingRets results) -> Error { return Error::Success(); });

%0 = "stablehlo.custom_call"(%arg0) {
  call_target_name = "foo",
  backend_config= {
    i32 = 42 : i32,
    str = "string"
  },
  api_version = 4 : i32
} : (tensor<f32>) -> tensor<f32>

auto handler = Ffi::Bind()
  .Arg<BufferR0<F32>>()
  .Attr<int32_t>("i32")
  .Attr<std::string_view>("str")
  .To([](BufferR0<F32> buffer, int32_t i32, std::string_view str) {
    return Error::Success();
  });

%0 = "stablehlo.custom_call"(%arg0) {
  call_target_name = "foo",
  backend_config= {
    command = 0 : i32
  },
  api_version = 4 : i32
} : (tensor<f32>) -> tensor<f32>

enum class Command : int32_t {
  kAdd = 0,
  kMul = 1,
};

XLA_FFI_REGISTER_ENUM_ATTR_DECODING(Command);

auto handler = Ffi::Bind().Attr<Command>("command").To(
    [](Command command) -> Error { return Error::Success(); });

auto handler = Ffi::Bind().Attrs().To([](Dictionary attrs) -> Error {
  ErrorOr<int32_t> i32 = attrs.get<int32_t>("i32");
  return Error::Success();
});

%0 = "stablehlo.custom_call"(%arg0) {
  call_target_name = "foo",
  backend_config= {
    range = { lo = 0 : i64, hi = 42 : i64 }
  },
  api_version = 4 : i32
} : (tensor<f32>) -> tensor<f32>

struct Range {
  int64_t lo;
  int64_t hi;
};

XLA_FFI_REGISTER_STRUCT_ATTR_DECODING(Range, StructMember<int64_t>("i64"),
                                             StructMember<int64_t>("i64"));

auto handler = Ffi::Bind().Attr<Range>("range").To([](Range range) -> Error{
  return Error::Success();
});

auto handler = Ffi::Bind().Attrs().To([](Dictionary attrs) -> Error {
  ErrorOr<Range> range = attrs.get<Range>("range");
  return Error::Success();
});

#include "xla/client/xla_builder.h"
#include "xla/service/custom_call_target_registry.h"

void do_it() {
  xla::XlaBuilder b("do_it");
  xla::XlaOp param0 =
      xla::Parameter(&b, 0, xla::ShapeUtil::MakeShape(xla::F32, {128}), "p0");
  xla::XlaOp param1 =
      xla::Parameter(&b, 1, xla::ShapeUtil::MakeShape(xla::F32, {2048}), "p1");
  xla::XlaOp custom_call =
      xla::CustomCall(&b, "do_custom_call", /*operands=*/{param0, param1},
        /*shape=*/xla::ShapeUtil::MakeShape(xla::F32, {2048}),
        /*opaque=*/"", /*has_side_effect=*/false,
        /*output_operand_aliasing=*/{}, /*literal=*/nullptr,
        /*schedule=*/CustomCallSchedule::SCHEDULE_NONE,
        /*api_version=*/CustomCallApiVersion::API_VERSION_TYPED_FFI);
}

// Constrain custom call arguments to rank-1 buffers of F32 data type.
using BufferF32 = xla::ffi::BufferR1<xla::ffi::DataType::F32>;

// Implement a custom call as a C+ function. Note that we can use `Buffer` type
// defined by XLA FFI that gives us access to buffer data type and shape.
xla::ffi::Error do_custom_call(BufferF32 in0, BufferF32 in1,
                               xla::ffi::Result<BufferF32> out) {
  size_t d0 = in0.dimensions[0];
  size_t d1 = in1.dimensions[0];

  // Check that dimensions are compatible.
  assert(out->dimensions[0] == d1 && "unexpected dimensions");

  for (size_t i = 0; i < d1; ++i) {
    out->data[i] = in0.data[i % d0] + in1.data[i];
  }
}

// Explicitly define an XLA FFI handler signature and bind it to the
// `do_custom_call` implementation. XLA FFI handler can automatically infer
// type signature from the custom call function, but it relies on magical
// template metaprogramming an explicit binding provides and extra level of
// type checking and clearly states custom call author intentions.
XLA_FFI_DEFINE_HANDLER(handler, do_custom_call,
                       ffi::Ffi::Bind()
                           .Arg<Buffer>()
                           .Arg<Buffer>()
                           .Ret<Buffer>());

// Registers `handler` with and XLA FFI on a "Host" platform.
XLA_FFI_REGISTER_HANDLER(xla::ffi::GetXlaFfiApi(), "do_custom_call",
                         "Host", handler);

void do_it() { /* same implementation as above */ }

__global__ custom_call_kernel(const float* in0, const float* in1, float* out) {
  size_t idx = blockIdx.x * blockDim.x + threadIdx.x;
  out[idx] = in0[idx % 128] + in1[idx];
}

void do_custom_call(CUstream stream, BufferF32 in0, BufferF32 in1,
                    xla::ffi::Result<BufferF32> out) {
  size_t d0 = in0.dimensions[0];
  size_t d1 = in1.dimensions[0];
  size_t d2 = out->dimensions[0];

  assert(d0 == 128 && d1 == 2048 && d2 == 2048 && "unexpected dimensions");

  const int64_t block_dim = 64;
  const int64_t grid_dim = 2048 / block_dim;
  custom_call_kernel<<<grid_dim, block_dim, 0, stream>>>(
    in0.data, in1.data, out->data);
}

XLA_FFI_DEFINE_HANDLER(handler, do_custom_call,
                       ffi::Ffi::Bind()
                           .Ctx<xla::ffi::PlatformStream<CUstream>>()
                           .Arg<BufferF32>()
                           .Arg<BufferF32>()
                           .Ret<BufferF32>());

XLA_FFI_REGISTER_HANDLER(xla::ffi::GetXlaFfiApi(), "do_custom_call",
                         "CUDA", handler);

using xla::ShapeUtil;
using xla::F32;
Shape p0_shape = ShapeUtil::MakeTuple({
    ShapeUtil::MakeShape(F32, {32}),
    ShapeUtil::MakeTuple({
        ShapeUtil::MakeShape(F32, {64}),
        ShapeUtil::MakeShape(F32, {128}),
    }),
    ShapeUtil::MakeShape(F32, {256}),
});
xla::XlaOp p0 = xla::Parameter(0, p0_shape, "p0");

Shape out_shape = ShapeUtil::MakeTuple({
  ShapeUtil::MakeShape(F32, {512}),
  ShapeUtil::MakeShape(F32, {1024}),
});
xla::CustomCall(&b, "do_custom_call", /*operands=*/{p0}, out_shape, ...);