pcsx2/common/include/x86emitter/implement/simd_arithmetic.h

322 lines
13 KiB
C++

/* PCSX2 - PS2 Emulator for PCs
* Copyright (C) 2002-2010 PCSX2 Dev Team
*
* PCSX2 is free software: you can redistribute it and/or modify it under the terms
* of the GNU Lesser General Public License as published by the Free Software Found-
* ation, either version 3 of the License, or (at your option) any later version.
*
* PCSX2 is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY;
* without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
* PURPOSE. See the GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License along with PCSX2.
* If not, see <http://www.gnu.org/licenses/>.
*/
#pragma once
namespace x86Emitter {
// --------------------------------------------------------------------------------------
// _SimdShiftHelper
// --------------------------------------------------------------------------------------
struct _SimdShiftHelper
{
u8 Prefix;
u16 Opcode;
u16 OpcodeImm;
u8 Modcode;
void operator()( const xRegisterSSE& to, const xRegisterSSE& from ) const;
void operator()( const xRegisterSSE& to, const ModSibBase& from ) const;
void operator()( const xRegisterMMX& to, const xRegisterMMX& from ) const;
void operator()( const xRegisterMMX& to, const ModSibBase& from ) const;
void operator()( const xRegisterSSE& to, u8 imm8 ) const;
void operator()( const xRegisterMMX& to, u8 imm8 ) const;
};
// --------------------------------------------------------------------------------------
// xImplSimd_Shift / xImplSimd_ShiftWithoutQ
// --------------------------------------------------------------------------------------
// Used for PSRA, which lacks the Q form.
//
struct xImplSimd_ShiftWithoutQ
{
const _SimdShiftHelper W;
const _SimdShiftHelper D;
};
// Implements PSRL and PSLL
//
struct xImplSimd_Shift
{
const _SimdShiftHelper W;
const _SimdShiftHelper D;
const _SimdShiftHelper Q;
void DQ( const xRegisterSSE& to, u8 imm8 ) const;
};
//////////////////////////////////////////////////////////////////////////////////////////
//
struct xImplSimd_AddSub
{
const xImplSimd_DestRegEither B;
const xImplSimd_DestRegEither W;
const xImplSimd_DestRegEither D;
const xImplSimd_DestRegEither Q;
// Add/Sub packed signed byte [8bit] integers from src into dest, and saturate the results.
const xImplSimd_DestRegEither SB;
// Add/Sub packed signed word [16bit] integers from src into dest, and saturate the results.
const xImplSimd_DestRegEither SW;
// Add/Sub packed unsigned byte [8bit] integers from src into dest, and saturate the results.
const xImplSimd_DestRegEither USB;
// Add/Sub packed unsigned word [16bit] integers from src into dest, and saturate the results.
const xImplSimd_DestRegEither USW;
};
//////////////////////////////////////////////////////////////////////////////////////////
//
struct xImplSimd_PMul
{
const xImplSimd_DestRegEither LW;
const xImplSimd_DestRegEither HW;
const xImplSimd_DestRegEither HUW;
const xImplSimd_DestRegEither UDQ;
// [SSE-3] PMULHRSW multiplies vertically each signed 16-bit integer from dest with the
// corresponding signed 16-bit integer of source, producing intermediate signed 32-bit
// integers. Each intermediate 32-bit integer is truncated to the 18 most significant
// bits. Rounding is always performed by adding 1 to the least significant bit of the
// 18-bit intermediate result. The final result is obtained by selecting the 16 bits
// immediately to the right of the most significant bit of each 18-bit intermediate
// result and packed to the destination operand.
//
// Both operands can be MMX or XMM registers. Source can be register or memory.
//
const xImplSimd_DestRegEither HRSW;
// [SSE-4.1] Multiply the packed dword signed integers in dest with src, and store
// the low 32 bits of each product in xmm1.
const xImplSimd_DestRegSSE LD;
// [SSE-4.1] Multiply the packed signed dword integers in dest with src.
const xImplSimd_DestRegSSE DQ;
};
//////////////////////////////////////////////////////////////////////////////////////////
// For instructions that have PS/SS form only (most commonly reciprocal Sqrt functions)
//
struct xImplSimd_rSqrt
{
const xImplSimd_DestRegSSE PS;
const xImplSimd_DestRegSSE SS;
};
//////////////////////////////////////////////////////////////////////////////////////////
// SQRT has PS/SS/SD forms, but not the PD form.
//
struct xImplSimd_Sqrt
{
const xImplSimd_DestRegSSE PS;
const xImplSimd_DestRegSSE SS;
const xImplSimd_DestRegSSE SD;
};
//////////////////////////////////////////////////////////////////////////////////////////
//
struct xImplSimd_AndNot
{
const xImplSimd_DestRegSSE PS;
const xImplSimd_DestRegSSE PD;
};
//////////////////////////////////////////////////////////////////////////////////////////
// Packed absolute value. [sSSE3 only]
//
struct xImplSimd_PAbsolute
{
// [sSSE-3] Computes the absolute value of bytes in the src, and stores the result
// in dest, as UNSIGNED.
const xImplSimd_DestRegEither B;
// [sSSE-3] Computes the absolute value of word in the src, and stores the result
// in dest, as UNSIGNED.
const xImplSimd_DestRegEither W;
// [sSSE-3] Computes the absolute value of doublewords in the src, and stores the
// result in dest, as UNSIGNED.
const xImplSimd_DestRegEither D;
};
//////////////////////////////////////////////////////////////////////////////////////////
// Packed Sign [sSSE3 only] - Negate/zero/preserve packed integers in dest depending on the
// corresponding sign in src.
//
struct xImplSimd_PSign
{
// [sSSE-3] negates each byte element of dest if the signed integer value of the
// corresponding data element in src is less than zero. If the signed integer value
// of a data element in src is positive, the corresponding data element in dest is
// unchanged. If a data element in src is zero, the corresponding data element in
// dest is set to zero.
const xImplSimd_DestRegEither B;
// [sSSE-3] negates each word element of dest if the signed integer value of the
// corresponding data element in src is less than zero. If the signed integer value
// of a data element in src is positive, the corresponding data element in dest is
// unchanged. If a data element in src is zero, the corresponding data element in
// dest is set to zero.
const xImplSimd_DestRegEither W;
// [sSSE-3] negates each doubleword element of dest if the signed integer value
// of the corresponding data element in src is less than zero. If the signed integer
// value of a data element in src is positive, the corresponding data element in dest
// is unchanged. If a data element in src is zero, the corresponding data element in
// dest is set to zero.
const xImplSimd_DestRegEither D;
};
//////////////////////////////////////////////////////////////////////////////////////////
// Packed Multiply and Add!!
//
struct xImplSimd_PMultAdd
{
// Multiplies the individual signed words of dest by the corresponding signed words
// of src, producing temporary signed, doubleword results. The adjacent doubleword
// results are then summed and stored in the destination operand.
//
// DEST[31:0] = ( DEST[15:0] * SRC[15:0]) + (DEST[31:16] * SRC[31:16] );
// DEST[63:32] = ( DEST[47:32] * SRC[47:32]) + (DEST[63:48] * SRC[63:48] );
// [.. repeat in the case of XMM src/dest operands ..]
//
const xImplSimd_DestRegEither WD;
// [sSSE-3] multiplies vertically each unsigned byte of dest with the corresponding
// signed byte of src, producing intermediate signed 16-bit integers. Each adjacent
// pair of signed words is added and the saturated result is packed to dest.
// For example, the lowest-order bytes (bits 7-0) in src and dest are multiplied
// and the intermediate signed word result is added with the corresponding
// intermediate result from the 2nd lowest-order bytes (bits 15-8) of the operands;
// the sign-saturated result is stored in the lowest word of dest (bits 15-0).
// The same operation is performed on the other pairs of adjacent bytes.
//
// In Coder Speak:
// DEST[15-0] = SaturateToSignedWord( SRC[15-8] * DEST[15-8] + SRC[7-0] * DEST[7-0] );
// DEST[31-16] = SaturateToSignedWord( SRC[31-24] * DEST[31-24] + SRC[23-16] * DEST[23-16] );
// [.. repeat for each 16 bits up to 64 (mmx) or 128 (xmm) ..]
//
const xImplSimd_DestRegEither UBSW;
};
//////////////////////////////////////////////////////////////////////////////////////////
// Packed Horizontal Add [SSE3 only]
//
struct xImplSimd_HorizAdd
{
// [SSE-3] Horizontal Add of Packed Data. A three step process:
// * Adds the single-precision floating-point values in the first and second dwords of
// dest and stores the result in the first dword of dest.
// * Adds single-precision floating-point values in the third and fourth dword of dest
// stores the result in the second dword of dest.
// * Adds single-precision floating-point values in the first and second dword of *src*
// and stores the result in the third dword of dest.
const xImplSimd_DestRegSSE PS;
// [SSE-3] Horizontal Add of Packed Data. A two step process:
// * Adds the double-precision floating-point values in the high and low quadwords of
// dest and stores the result in the low quadword of dest.
// * Adds the double-precision floating-point values in the high and low quadwords of
// *src* stores the result in the high quadword of dest.
const xImplSimd_DestRegSSE PD;
};
//////////////////////////////////////////////////////////////////////////////////////////
// DotProduct calculation (SSE4.1 only!)
//
struct xImplSimd_DotProduct
{
// [SSE-4.1] Conditionally multiplies the packed single precision floating-point
// values in dest with the packed single-precision floats in src depending on a
// mask extracted from the high 4 bits of the immediate byte. If a condition mask
// bit in Imm8[7:4] is zero, the corresponding multiplication is replaced by a value
// of 0.0. The four resulting single-precision values are summed into an inter-
// mediate result.
//
// The intermediate result is conditionally broadcasted to the destination using a
// broadcast mask specified by bits [3:0] of the immediate byte. If a broadcast
// mask bit is 1, the intermediate result is copied to the corresponding dword
// element in dest. If a broadcast mask bit is zero, the corresponding element in
// the destination is set to zero.
//
xImplSimd_DestRegImmSSE PS;
// [SSE-4.1]
xImplSimd_DestRegImmSSE PD;
};
//////////////////////////////////////////////////////////////////////////////////////////
// Rounds floating point values (packed or single scalar) by an arbitrary rounding mode.
// (SSE4.1 only!)
struct xImplSimd_Round
{
// [SSE-4.1] Rounds the 4 packed single-precision src values and stores them in dest.
//
// Imm8 specifies control fields for the rounding operation:
// Bit 3 - processor behavior for a precision exception (0: normal, 1: inexact)
// Bit 2 - If enabled, use MXCSR.RC, else use RC specified in bits 1:0 of this Imm8.
// Bits 1:0 - Specifies a rounding mode for this instruction only.
//
// Rounding Mode Reference:
// 0 - Nearest, 1 - Negative Infinity, 2 - Positive infinity, 3 - Truncate.
//
const xImplSimd_DestRegImmSSE PS;
// [SSE-4.1] Rounds the 2 packed double-precision src values and stores them in dest.
//
// Imm8 specifies control fields for the rounding operation:
// Bit 3 - processor behavior for a precision exception (0: normal, 1: inexact)
// Bit 2 - If enabled, use MXCSR.RC, else use RC specified in bits 1:0 of this Imm8.
// Bits 1:0 - Specifies a rounding mode for this instruction only.
//
// Rounding Mode Reference:
// 0 - Nearest, 1 - Negative Infinity, 2 - Positive infinity, 3 - Truncate.
//
const xImplSimd_DestRegImmSSE PD;
// [SSE-4.1] Rounds the single-precision src value and stores in dest.
//
// Imm8 specifies control fields for the rounding operation:
// Bit 3 - processor behavior for a precision exception (0: normal, 1: inexact)
// Bit 2 - If enabled, use MXCSR.RC, else use RC specified in bits 1:0 of this Imm8.
// Bits 1:0 - Specifies a rounding mode for this instruction only.
//
// Rounding Mode Reference:
// 0 - Nearest, 1 - Negative Infinity, 2 - Positive infinity, 3 - Truncate.
//
const xImplSimd_DestRegImmSSE SS;
// [SSE-4.1] Rounds the double-precision src value and stores in dest.
//
// Imm8 specifies control fields for the rounding operation:
// Bit 3 - processor behavior for a precision exception (0: normal, 1: inexact)
// Bit 2 - If enabled, use MXCSR.RC, else use RC specified in bits 1:0 of this Imm8.
// Bits 1:0 - Specifies a rounding mode for this instruction only.
//
// Rounding Mode Reference:
// 0 - Nearest, 1 - Negative Infinity, 2 - Positive infinity, 3 - Truncate.
//
const xImplSimd_DestRegImmSSE SD;
};
} // End namespace x86Emitter