} : (tensor<3x2x!tf_type.qint32>, tensor<3x2x!tf_type.qint32>, tensor<3x2x!tf_type.qint32>, tensor<2xf32>, tensor<2xi32>) -> tensor<3x2x!tf_type.qint32>
  func.return %1 : tensor<3x2x!tf_type.qint32>
}

// -----

// CHECK-LABEL: func @uniform_quantized_clip_by_value_min_not_const

}

// Ilogb returns the binary exponent of x as an integer.
//
// Special cases are:
//
//	Ilogb(±Inf) = MaxInt32
//	Ilogb(0) = MinInt32
//	Ilogb(NaN) = MaxInt32
func Ilogb(x float64) int {
	// special cases
	switch {
	case x == 0:
		return MinInt32
	case IsNaN(x):
		return MaxInt32
	case IsInf(x, 0):
		return MaxInt32
	}
	return ilogb(x)
}

)

const (
	_ int32 = minInt32 /* ERROR "overflows" */ - 1
	_ int32 = minInt32
	_ int32 = maxInt32
	_ int32 = maxInt32 /* ERROR "overflows" */ + 1
	_ int32 = smallestFloat64 /* ERROR "truncated" */

	_ = int32(minInt32 /* ERROR "overflows" */ - 1)
	_ = int32(minInt32)
	_ = int32(maxInt32)
	_ = int32(maxInt32 /* ERROR "overflows" */ + 1)
	_ = int32(smallestFloat64 /* ERROR "cannot convert" */)
)

		t.Errorf("MaxInt8 should wrap around to MinInt8: %d", v+1)
	}
	if v := int16(MaxInt16); v+1 != MinInt16 {
		t.Errorf("MaxInt16 should wrap around to MinInt16: %d", v+1)
	}
	if v := int32(MaxInt32); v+1 != MinInt32 {
		t.Errorf("MaxInt32 should wrap around to MinInt32: %d", v+1)
	}
	if v := int64(MaxInt64); v+1 != MinInt64 {
		t.Errorf("MaxInt64 should wrap around to MinInt64: %d", v+1)
	}

package runtime

import (
	"internal/cpu"
)

var useAVXmemmove bool

func init() {
	// Let's remove stepping and reserved fields
	processor := processorVersionInfo & 0x0FFF3FF0

	isIntelBridgeFamily := isIntel &&
		processor == 0x206A0 ||
		processor == 0x206D0 ||
		processor == 0x306A0 ||
		processor == 0x306E0

	useAVXmemmove = cpu.X86.HasAVX && !isIntelBridgeFamily

	MinInt    = -1 << (intSize - 1) // MinInt32 or MinInt64 depending on intSize.
	MaxInt8   = 1<<7 - 1            // 127
	MinInt8   = -1 << 7             // -128
	MaxInt16  = 1<<15 - 1           // 32767
	MinInt16  = -1 << 15            // -32768
	MaxInt32  = 1<<31 - 1           // 2147483647
	MinInt32  = -1 << 31            // -2147483648
	MaxInt64  = 1<<63 - 1           // 9223372036854775807

	"cmd/compile/internal/base"
	"cmd/internal/src"
)

var basicTypes = [...]struct {
	name  string
	etype Kind
}{
	{"int8", TINT8},
	{"int16", TINT16},
	{"int32", TINT32},
	{"int64", TINT64},
	{"uint8", TUINT8},
	{"uint16", TUINT16},
	{"uint32", TUINT32},
	{"uint64", TUINT64},
	{"float32", TFLOAT32},
	{"float64", TFLOAT64},
	{"complex64", TCOMPLEX64},
	{"complex128", TCOMPLEX128},
	{"bool", TBOOL},

type MyComplex64 complex64
type MyFloat32 float32
type MyFloat64 float64
type MyInt int
type MyInt8 int8
type MyInt16 int16
type MyInt32 int32
type MyInt64 int64
type MyString string
type MyUint uint
type MyUint8 uint8
type MyUint16 uint16
type MyUint32 uint32
type MyUint64 uint64
type MyUintptr uintptr

func panicCustomComplex64() {
	panic(MyComplex64(0.11 + 3i))
}

func panicCustomComplex128() {

    rhs_quantization_min_val = -2147483648 : i64
  } : (
    tensor<10x10x!tf_type.qint32>, tensor<10x10x!tf_type.qint32>, tensor<f32>,
    tensor<i32>, tensor<f32>, tensor<i32>, tensor<f32>, tensor<i32>
  ) -> tensor<10x10x!tf_type.qint32>
  %1 = "tf.Cast"(%0) {} : (tensor<10x10x!tf_type.qint32>) ->  tensor<10x10xi32>
  return %1 : tensor<10x10xi32>
})mlir";

 * (Marsaglia & Tsang, 2000)
 * http://www.jstatsoft.org/v05/i08/paper [pdf]
 */

const (
	rn = 3.442619855899
)

func absInt32(i int32) uint32 {
	if i < 0 {
		return uint32(-i)
	}
	return uint32(i)
}

// NormFloat64 returns a normally distributed float64 in
// the range -[math.MaxFloat64] through +[math.MaxFloat64] inclusive,