for i := 0; i < b.N; i++ {
		x.Add(x, y)
	}
}

func BenchmarkMultiply(b *testing.B) {
	x := new(Element).One()
	y := new(Element).Add(x, x)
	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		x.Multiply(x, y)
	}
}

func BenchmarkSquare(b *testing.B) {
	x := new(Element).Add(feOne, feOne)
	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		x.Square(x)
	}
}

func BenchmarkInvert(b *testing.B) {

        this("greeter")
    }

    SwiftLibWithXCTest(String projectName) {
        super(projectName)
        main = new SwiftLib(projectName)
        test = new SwiftLibTest(main, main.greeter, main.sum, main.multiply)
    }

                        Multiplier multiplier = new Multiplier();
                        Assert.assertEquals(1, multiplier.multiply(1, 1));
                        Assert.assertEquals(4, multiplier.multiply(2, 2));
                        Assert.assertEquals(2, multiplier.multiply(1, 2));
                    }
                }
            """
            file("transitive/build.gradle") << """

  // CHECK: %[[X_CENTER:.+]] = mhlo.subtract %[[X]], %[[MEAN_BCAST]] : tensor<4x256xf32>
  // CHECK: %[[X_SCALED:.+]] = mhlo.multiply %[[X_CENTER]], %[[SCALE_BCAST]] : tensor<4x256xf32>
  // CHECK: %[[X_NORMED:.+]] = mhlo.divide %[[X_SCALED]], %[[STDDEV_BCAST]] : tensor<4x256xf32>

        List<XCTestSourceFileElement> getTestSuites() {
            return [
                new XCTestSourceFileElement("CombinedTests") {
                    final delegate = new SwiftLibTest(main, main.greeter, main.sum, main.multiply)

                    @Override
                    List<XCTestCaseElement> getTestCases() {
                        return delegate.sumTest.testCases + delegate.greeterTest.testCases + delegate.multiplyTest.testCases

    auto subtraction_result = Subtract(scope, grad_grad, squeeze_result);
    auto multiply_result = Multiply(scope, subtraction_result, logits_softmax);
    grad = Add(scope, grad, multiply_result);
  }
  auto minus_log_softmax = Multiply(scope, LogSoftmax(scope, logits), -1.0f);
  grad_outputs->push_back(grad);
  grad_outputs->push_back(BroadcastMul(scope, grad_loss, minus_log_softmax));

        }
        return new Amount<>(units.scaleTo(value, other.units).subtract(other.value), other.units);
    }

    public Amount<Q> multiply(BigDecimal other) {
        return new Amount<>(value.multiply(other), units);
    }

    public Amount<Q> div(long other) {
        return div(BigDecimal.valueOf(other));
    }
    public Amount<Q> div(BigDecimal other) {

  %10 = mhlo.add %arg0, %5 : tensor<2xf32>
  %11 = mhlo.clamp %7, %10, %3 : tensor<2xf32>
  %12 = mhlo.clamp %9, %11, %1 : tensor<2xf32>
  %13 = mhlo.multiply %arg0, %12 : tensor<2xf32>
  %14 = mhlo.divide %13, %1 : tensor<2xf32>
  return %14 : tensor<2xf32>
}

// CHECK-LABEL:   func.func @hardswish(

// It also runs ssa/check and gccheck to be sure that those are checked at least a
// little in each run.bash.  It does not check or run the generated code.
// The test file is however a useful example of fused-vs-cascaded multiply-add.
func TestFmaHash(t *testing.T) {
	switch runtime.GOOS {
	case "linux", "darwin":
	default:
		t.Skipf("Slow test, usually avoid it, os=%s not linux or darwin", runtime.GOOS)
	}
	switch runtime.GOARCH {

	}

	if y32+v32_n < 0 {
		t.Errorf("'%#x + %#x < 0' failed", y32, v32_n)
	}

	if y32+v32_n <= 0 {
		t.Errorf("'%#x + %#x <= 0' failed", y32, v32_n)
	}
}

// multiply-add
func testMAddVar64(t *testing.T) {
	if x64+v64*one64 < 0 {
	} else {
		t.Errorf("'%#x + %#x*1 < 0' failed", x64, v64)
	}

	if x64+v64*one64 <= 0 {
	} else {