audk/EdkModulePkg/Universal/Ebc/Dxe/x64/x64Math.c

452 lines
7.5 KiB
C

/*++
Copyright (c) 2006 , Intel Corporation
All rights reserved. This program and the accompanying materials
are licensed and made available under the terms and conditions of the BSD License
which accompanies this distribution. The full text of the license may be found at
http://opensource.org/licenses/bsd-license.php
THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,
WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.
Module Name:
x64math.c
Abstract:
Math routines for x64.
--*/
UINT64
LeftShiftU64 (
IN UINT64 Operand,
IN UINT64 Count
)
/*++
Routine Description:
Left-shift a 64 bit value.
Arguments:
Operand - 64-bit value to shift
Count - shift count
Returns:
Operand << Count
--*/
{
if (Count > 63) {
return 0;
}
return Operand << Count;
}
UINT64
RightShiftU64 (
IN UINT64 Operand,
IN UINT64 Count
)
/*++
Routine Description:
Right-shift a 64 bit value.
Arguments:
Operand - 64-bit value to shift
Count - shift count
Returns:
Operand >> Count
--*/
{
if (Count > 63) {
return 0;
}
return Operand >> Count;
}
INT64
ARightShift64 (
IN INT64 Operand,
IN UINT64 Count
)
/*++
Routine Description:
Right-shift a 64 bit signed value.
Arguments:
Operand - 64-bit value to shift
Count - shift count
Returns:
Operand >> Count
--*/
{
if (Count > 63) {
if (Operand & 0x8000000000000000ULL) {
return (INT64)~0;
}
return 0;
}
return Operand >> Count;
}
#if 0
//
// The compiler generates true assembly for these, so we don't need them.
//
INT32
ARightShift32 (
IN INT32 Operand,
IN UINTN Count
)
/*++
Routine Description:
Right shift a 32-bit value
Arguments:
Operand - value to shift
Count - shift count
Returns:
Operand >> Count
--*/
{
return Operand >> (Count & 0x1f);
}
INT32
MulS32x32 (
INT32 Value1,
INT32 Value2,
INT32 *ResultHigh
)
/*++
Routine Description:
Multiply two signed 32-bit numbers.
Arguments:
Value1 - first value to multiply
Value2 - value to multiply Value1 by
ResultHigh - overflow
Returns:
Value1 * Value2
Notes:
The 64-bit result is the concatenation of *ResultHigh and the return value
The product fits in 32 bits if
(*ResultHigh == 0x00000000 AND *ResultLow_bit31 == 0)
OR
(*ResultHigh == 0xffffffff AND *ResultLow_bit31 == 1)
--*/
{
INT64 Rres64;
INT32 Result;
Res64 = (INT64) Value1 * (INT64) Value2;
*ResultHigh = (Res64 >> 32) & 0xffffffff;
Result = Res64 & 0xffffffff;
return Result;
}
UINT32
MulU32x32 (
UINT32 Value1,
UINT32 Value2,
UINT32 *ResultHigh
)
/*++
Routine Description:
Multiply two unsigned 32-bit values.
Arguments:
Value1 - first number
Value2 - number to multiply by Value1
ResultHigh - overflow
Returns:
Value1 * Value2
Notes:
The 64-bit result is the concatenation of *ResultHigh and the return value.
The product fits in 32 bits if *ResultHigh == 0x00000000
--*/
{
UINT64 Res64;
UINT32 Result;
Res64 = (INT64) Value1 * (INT64) Value2;
*ResultHigh = (Res64 >> 32) & 0xffffffff;
Result = Res64 & 0xffffffff;
return Result;
}
INT32
DivS32x32 (
INT32 Value1,
INT32 Value2,
INT32 *Remainder,
UINTN *error
)
//
// signed 32-bit by signed 32-bit divide; the 32-bit remainder is
// in *Remainder and the quotient is the return value; *error = 1 if the
// divisor is 0, and it is 1 otherwise
//
{
INT32 Result;
*error = 0;
if (Value2 == 0x0) {
*error = 1;
Result = 0x80000000;
*Remainder = 0x80000000;
} else {
Result = Value1 / Value2;
*Remainder = Value1 - Result * Value2;
}
return Result;
}
UINT32
DivU32x32 (
UINT32 Value1,
UINT32 Value2,
UINT32 *Remainder,
UINTN *Error
)
//
// unsigned 32-bit by unsigned 32-bit divide; the 32-bit remainder is
// in *Remainder and the quotient is the return value; *error = 1 if the
// divisor is 0, and it is 1 otherwise
//
{
UINT32 Result;
*Error = 0;
if (Value2 == 0x0) {
*Error = 1;
Result = 0x80000000;
*Remainder = 0x80000000;
} else {
Result = Value1 / Value2;
*Remainder = Value1 - Result * Value2;
}
return Result;
}
#endif
INT64
MulS64x64 (
INT64 Value1,
INT64 Value2,
INT64 *ResultHigh
)
/*++
Routine Description:
Multiply two signed 32-bit numbers.
Arguments:
Value1 - first value to multiply
Value2 - value to multiply Value1 by
ResultHigh - overflow
Returns:
Value1 * Value2
Notes:
The 64-bit result is the concatenation of *ResultHigh and the return value
The product fits in 32 bits if
(*ResultHigh == 0x00000000 AND *ResultLow_bit31 == 0)
OR
(*ResultHigh == 0xffffffff AND *ResultLow_bit31 == 1)
--*/
{
INT64 Result;
Result = Value1 * Value2;
return Result;
}
UINT64
MulU64x64 (
UINT64 Value1,
UINT64 Value2,
UINT64 *ResultHigh
)
/*++
Routine Description:
Multiply two unsigned 32-bit values.
Arguments:
Value1 - first number
Value2 - number to multiply by Value1
ResultHigh - overflow
Returns:
Value1 * Value2
Notes:
The 64-bit result is the concatenation of *ResultHigh and the return value.
The product fits in 32 bits if *ResultHigh == 0x00000000
--*/
{
UINT64 Result;
Result = Value1 * Value2;
return Result;
}
INT64
DivS64x64 (
INT64 Value1,
INT64 Value2,
INT64 *Remainder,
UINTN *Error
)
/*++
Routine Description:
Divide two 64-bit signed values.
Arguments:
Value1 - dividend
Value2 - divisor
Remainder - remainder of Value1/Value2
Error - to flag errors (divide-by-0)
Returns:
Value1 / Valu2
Note:
The 64-bit remainder is in *Remainder and the quotient is the return value.
*Error = 1 if the divisor is 0, and it is 1 otherwise
--*/
{
INT64 Result;
*Error = 0;
if (Value2 == 0x0) {
*Error = 1;
Result = 0x8000000000000000;
*Remainder = 0x8000000000000000;
} else {
Result = Value1 / Value2;
*Remainder = Value1 - Result * Value2;
}
return Result;
}
UINT64
DivU64x64 (
UINT64 Value1,
UINT64 Value2,
UINT64 *Remainder,
UINTN *Error
)
/*++
Routine Description:
Divide two 64-bit unsigned values.
Arguments:
Value1 - dividend
Value2 - divisor
Remainder - remainder of Value1/Value2
Error - to flag errors (divide-by-0)
Returns:
Value1 / Valu2
Note:
The 64-bit remainder is in *Remainder and the quotient is the return value.
*Error = 1 if the divisor is 0, and it is 1 otherwise
--*/
{
UINT64 Result;
*Error = 0;
if (Value2 == 0x0) {
*Error = 1;
Result = 0x8000000000000000;
*Remainder = 0x8000000000000000;
} else {
Result = Value1 / Value2;
*Remainder = Value1 - Result * Value2;
}
return Result;
}