18: 0000000000403dc0     0 OBJECT  LOCAL  DEFAULT   20 __frame_dummy_in[...]
    19: 0000000000000000     0 FILE    LOCAL  DEFAULT  ABS sum.cpp
    20: 0000000000000000     0 FILE    LOCAL  DEFAULT  ABS multiply.cpp
    21: 0000000000000000     0 FILE    LOCAL  DEFAULT  ABS crtstuff.c
    22: 00000000004021e0     0 OBJECT  LOCAL  DEFAULT   19 __FRAME_END__
    23: 0000000000000000     0 FILE    LOCAL  DEFAULT  ABS

        when:
        succeeds("unexport")
        then:
        file("build/relocated").assertHasDescendants(objectFile("greeter").name, objectFile("sum").name, objectFile("multiply").name)
    }

Status AddV2(AbstractContext* ctx, AbstractTensorHandle* const x,
             AbstractTensorHandle* const y, AbstractTensorHandle** z,
             const char* name = nullptr, const char* raw_device_name = nullptr);

// Multiply the matrix "a" by the matrix "b".
Status MatMul(AbstractContext* ctx, AbstractTensorHandle* const a,
              AbstractTensorHandle* const b, AbstractTensorHandle** product,

        app.test.greeterTest.withTestableImport("Hello")
        app.test.greeterTest.withTestableImport("Log")

        app.writeToProject(file("app"))
        app.sum.writeToProject(file("app"))
        app.multiply.writeToProject(file("app"))
        app.greeter.writeToProject(file("hello"))
        app.logger.writeToProject(file("log"))

        settingsFile.text =  """
            include 'app', 'log', 'hello'

			x1.Mul(x1, x4)

			if (i+j)&1 != 0 {
				x1.Neg(x1)
			}

			a.set(i, j, x1)
		}
	}
	return a
}

func (a *matrix) mul(b *matrix) *matrix {
	if a.m != b.n {
		panic("illegal matrix multiply")
	}
	c := newMatrix(a.n, b.m)
	for i := 0; i < c.n; i++ {
		for j := 0; j < c.m; j++ {
			x := NewRat(0, 1)
			for k := 0; k < a.m; k++ {
				x.Add(x, new(Rat).Mul(a.at(i, k), b.at(k, j)))
			}

// Checks if the value has only one user.
def HasOneUse : Constraint<CPred<"$0.hasOneUse()">>;

// If we see a Conv2D op followed by Mul, then multiply the filter
// with the value in Mul.
def FuseMulAndConv2D :
  Pat<(TF_MulOp:$mul (TF_Conv2DOp:$conv $input,
                          (Arith_ConstantOp:$filter F32ElementsAttr:$filter_value),

				sz = listSize.Multiply(elemSize)
			}

			if len(args) > 0 {
				sepSize := l.sizeEstimate(args[0])
				minSeparators := uint64(0)
				maxSeparators := uint64(0)
				if listSize.Min > 0 {
					minSeparators = listSize.Min - 1
				}
				if listSize.Max > 0 {
					maxSeparators = listSize.Max - 1
				}

		t.Fatal(err)
	}
	for i, v := range vectors {
		expectedToVerify := true
		for _, f := range v.Flags {
			switch f {
			// We use the simplified verification formula that doesn't multiply
			// by the cofactor, so any low order residue will cause the
			// signature not to verify.
			//
			// This is allowed, but not required, by RFC 8032.
			case "LowOrderResidue":
				expectedToVerify = false

        expect:
        succeeds "assemble"
        result.assertTasksExecuted(assembleAppTasks)
        result.assertTasksNotSkipped(assembleAppTasks)

        outputs.deletedClasses("multiply", "sum")

        // See https://github.com/gradle/gradle-native/issues/1004
        if (toolchainUnderTest.version.major == 5) {
            outputs.recompiledClasses('renamed-sum')
        } else {

  ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());

  // Loop to make synchronization failures more deterministic
  for (int i = 0; i < 100; ++i) {
    TensorHandlePtr multiply_result(
        Multiply(context.get(), combined_value.get(), combined_value.get(),
                 status.get()));
    ASSERT_TRUE(TF_GetCode(status.get()) == TF_OK) << TF_Message(status.get());
    std::array<TensorHandlePtr, 2> out_components;