MCPcopy Create free account
hub / github.com/DeepRec-AI/DeepRec / BesselI1e

Function BesselI1e

tensorflow/compiler/xla/client/lib/math.cc:1580–1593  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

1578}
1579
1580XlaOp BesselI1e(XlaOp x) {
1581 auto& b = *x.builder();
1582 return b.ReportErrorOrReturn([&]() -> StatusOr<XlaOp> {
1583 TF_RETURN_IF_ERROR(EnsureOperandIsRealFp("BesselI1e", x));
1584 TF_ASSIGN_OR_RETURN(auto shape, b.GetShape(x));
1585 if (shape.element_type() == F64) {
1586 return I1eImpl64(x);
1587 }
1588 // I1eF32Impl don't have enough precision when run with bf16 intermediates
1589 // (not surprising!), so upcast to f32 in this case.
1590 return DoWithUpcastToF32(x, {BF16, F16},
1591 [](XlaOp x) { return I1eImpl32(x); });
1592 });
1593}
1594
1595// I J Thompson and A R Barnett. 1986. Coulomb and Bessel functions of complex
1596// arguments and order. J. Comput. Phys. 64, 2 (June 1986), 490-509.

Callers 1

XLA_TEST_FFunction · 0.70

Calls 7

EnsureOperandIsRealFpFunction · 0.85
I1eImpl64Function · 0.85
DoWithUpcastToF32Function · 0.85
I1eImpl32Function · 0.85
ReportErrorOrReturnMethod · 0.80
builderMethod · 0.45
element_typeMethod · 0.45

Tested by 1

XLA_TEST_FFunction · 0.56