/* ES40 emulator. * Copyright (C) 2007-2008 by the ES40 Emulator Project * * Website: http://sourceforge.net/projects/es40 * E-mail : camiel@camicom.com * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public License * as published by the Free Software Foundation; either version 2 * of the License, or (at your option) any later version. * * This program 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 this program; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. * * Although this is not required, the author would appreciate being notified of, * and receiving any modifications you may make to the source code that might serve * the general public. */ /** * \file * Contains code macros for the processor MVI (multimedia) instructions. * * $Id$ * * X-1.4 Camiel Vanderhoeven 14-MAR-2008 * 1. More meaningful exceptions replace throwing (int) 1. * 2. U64 macro replaces X64 macro. * * X-1.3 Camiel Vanderhoeven 11-APR-2007 * Moved all data that should be saved to a state file to a structure * "state". * * X-1.2 Camiel Vanderhoeven 30-MAR-2007 * Added old changelog comments. * * X-1.1 Camiel Vanderhoeven 18-FEB-2007 * File created. Contains code previously found in AlphaCPU.h * * \author Camiel Vanderhoeven (camiel@camicom.com / http://www.camicom.com) **/ #if defined(_MSC_VER) && defined(_M_X64) #include #define ES40_HAVE_SSE2 1 // x64 on MSVC guarantees SSE2 /* Optional: define these if you compile with /arch:AVX (/arch:AVX2) or explicitly via project Preprocessor Definitions. Otherwise the SSE2 fallbacks are used automatically. */ #if defined(__AVX2__) || defined(__AVX__) || defined(__SSE4_1__) #define ES40_HAVE_SSE41 1 #endif #if defined(__SSSE3__) || defined(__SSE4_1__) || defined(__AVX__) || defined(__AVX2__) #define ES40_HAVE_SSSE3 1 #endif #endif #if defined(ES40_HAVE_SSE2) /* tiny helpers for 64b <-> XMM */ static __forceinline __m128i es40_loadq(u64 x) { return _mm_cvtsi64_si128((__int64)x); } static __forceinline u64 es40_storeq(__m128i v) { return (u64)_mm_cvtsi128_si64(v); } #endif #if defined(ES40_HAVE_SSE2) #define DO_MINUB8 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ __m128i m = _mm_min_epu8(a, b); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #else #define DO_MINUB8 do { \ temp_64 = 0; temp_64_1 = state.r[REG_1]; temp_64_2 = V_2; \ for (i = 0; i < 64; i += 8) { \ u8 a = (u8)((temp_64_1 >> i) & X64_BYTE), b = (u8)((temp_64_2 >> i) & X64_BYTE); \ temp_64 |= ((u64)(a < b ? a : b) << i); \ } \ state.r[REG_3] = temp_64; \ } while (0) #endif #if defined(ES40_HAVE_SSE2) && defined(ES40_HAVE_SSE41) #define DO_MINSB8 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ __m128i m = _mm_min_epi8(a, b); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #elif defined(ES40_HAVE_SSE2) #define DO_MINSB8 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ __m128i mask = _mm_cmpgt_epi8(b, a); /* mask=(b>a) */ \ __m128i m = _mm_or_si128(_mm_and_si128(a, mask), _mm_andnot_si128(mask, b)); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #else #define DO_MINSB8 do { \ temp_64 = 0; temp_64_1 = state.r[REG_1]; temp_64_2 = V_2; \ for (i = 0; i < 64; i += 8) { \ s8 a = (s8)((temp_64_1 >> i) & X64_BYTE), b = (s8)((temp_64_2 >> i) & X64_BYTE); \ temp_64 |= (u64)((a < b ? (u8)a : (u8)b)) << i; \ } \ state.r[REG_3] = temp_64; \ } while (0) #endif #if defined(ES40_HAVE_SSE2) && defined(ES40_HAVE_SSE41) #define DO_MINUW4 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ __m128i m = _mm_min_epu16(a, b); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #elif defined(ES40_HAVE_SSE2) #define DO_MINUW4 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ const __m128i bias = _mm_set1_epi16((short)0x8000); \ __m128i ax = _mm_xor_si128(a, bias); \ __m128i bx = _mm_xor_si128(b, bias); \ __m128i m = _mm_xor_si128(_mm_min_epi16(ax, bx), bias); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #else #define DO_MINUW4 do { \ temp_64 = 0; temp_64_1 = state.r[REG_1]; temp_64_2 = V_2; \ for (i = 0; i < 64; i += 16) { \ u16 a = (u16)((temp_64_1 >> i) & X64_WORD), b = (u16)((temp_64_2 >> i) & X64_WORD); \ temp_64 |= ((u64)(a < b ? a : b) << i); \ } \ state.r[REG_3] = temp_64; \ } while (0) #endif #if defined(ES40_HAVE_SSE2) #define DO_MINSW4 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ __m128i m = _mm_min_epi16(a, b); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #else #define DO_MINSW4 do { \ temp_64 = 0; temp_64_1 = state.r[REG_1]; temp_64_2 = V_2; \ for (i = 0; i < 64; i += 16) { \ s16 a = (s16)((temp_64_1 >> i) & X64_WORD), b = (s16)((temp_64_2 >> i) & X64_WORD); \ temp_64 |= (u64)((a < b ? (u16)a : (u16)b)) << i; \ } \ state.r[REG_3] = temp_64; \ } while (0) #endif #if defined(ES40_HAVE_SSE2) #define DO_MAXUB8 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ __m128i m = _mm_max_epu8(a, b); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #else #define DO_MAXUB8 do { \ temp_64 = 0; temp_64_1 = state.r[REG_1]; temp_64_2 = V_2; \ for (i = 0; i < 64; i += 8) { \ u8 a = (u8)((temp_64_1 >> i) & X64_BYTE), b = (u8)((temp_64_2 >> i) & X64_BYTE); \ temp_64 |= ((u64)(a > b ? a : b) << i); \ } \ state.r[REG_3] = temp_64; \ } while (0) #endif #if defined(ES40_HAVE_SSE2) && defined(ES40_HAVE_SSE41) #define DO_MAXSB8 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ __m128i m = _mm_max_epi8(a, b); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #elif defined(ES40_HAVE_SSE2) #define DO_MAXSB8 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ __m128i mask = _mm_cmpgt_epi8(a, b); /* mask=(a>b) */ \ __m128i m = _mm_or_si128(_mm_and_si128(a, mask), _mm_andnot_si128(mask, b)); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #else #define DO_MAXSB8 do { \ temp_64 = 0; temp_64_1 = state.r[REG_1]; temp_64_2 = V_2; \ for (i = 0; i < 64; i += 8) { \ s8 a = (s8)((temp_64_1 >> i) & X64_BYTE), b = (s8)((temp_64_2 >> i) & X64_BYTE); \ temp_64 |= (u64)((a > b ? (u8)a : (u8)b)) << i; \ } \ state.r[REG_3] = temp_64; \ } while (0) #endif #if defined(ES40_HAVE_SSE2) && defined(ES40_HAVE_SSE41) #define DO_MAXUW4 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ __m128i m = _mm_max_epu16(a, b); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #elif defined(ES40_HAVE_SSE2) #define DO_MAXUW4 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ const __m128i bias = _mm_set1_epi16((short)0x8000); \ __m128i ax = _mm_xor_si128(a, bias); \ __m128i bx = _mm_xor_si128(b, bias); \ __m128i m = _mm_xor_si128(_mm_max_epi16(ax, bx), bias); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #else #define DO_MAXUW4 do { \ temp_64 = 0; temp_64_1 = state.r[REG_1]; temp_64_2 = V_2; \ for (i = 0; i < 64; i += 16) { \ u16 a = (u16)((temp_64_1 >> i) & X64_WORD), b = (u16)((temp_64_2 >> i) & X64_WORD); \ temp_64 |= (u64)(a > b ? a : b) << i; \ } \ state.r[REG_3] = temp_64; \ } while (0) #endif #if defined(ES40_HAVE_SSE2) #define DO_MAXSW4 do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ __m128i m = _mm_max_epi16(a, b); \ state.r[REG_3] = es40_storeq(m); \ } while (0) #else #define DO_MAXSW4 do { \ temp_64 = 0; temp_64_1 = state.r[REG_1]; temp_64_2 = V_2; \ for (i = 0; i < 64; i += 16) { \ s16 a = (s16)((temp_64_1 >> i) & X64_WORD), b = (s16)((temp_64_2 >> i) & X64_WORD); \ temp_64 |= (u64)((a > b ? (u16)a : (u16)b)) << i; \ } \ state.r[REG_3] = temp_64; \ } while (0) #endif #if defined(ES40_HAVE_SSE2) /* PERR: scalar sum of 8 *unsigned* byte absolute differences (HRM 4.13.3). */ #define DO_PERR do { \ __m128i a = es40_loadq(state.r[REG_1]); \ __m128i b = es40_loadq(V_2); \ state.r[REG_3] = es40_storeq(_mm_sad_epu8(a, b)); \ } while (0) #else #define DO_PERR do { \ temp_64 = 0; temp_64_1 = state.r[REG_1]; temp_64_2 = V_2; \ for (i = 0; i < 64; i += 8) { \ u8 a = (u8)((temp_64_1 >> i) & X64_BYTE); \ u8 b = (u8)((temp_64_2 >> i) & X64_BYTE); \ temp_64 += (a > b) ? (a - b) : (b - a); \ } \ state.r[REG_3] = temp_64; \ } while (0) #endif #if defined(ES40_HAVE_SSSE3) #define DO_PKLB do { \ __m128i x = es40_loadq(V_2); \ const __m128i m = _mm_setr_epi8(0,4,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80, \ (char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80); \ state.r[REG_3] = es40_storeq(_mm_shuffle_epi8(x, m)); \ } while (0) #else #define DO_PKLB do { u64 t = V_2; state.r[REG_3] = (t & U64(0x00000000000000ff)) | ((t & U64(0x000000ff00000000)) >> 24); } while (0) #endif #if defined(ES40_HAVE_SSSE3) #define DO_PKWB do { \ __m128i x = es40_loadq(V_2); \ const __m128i m = _mm_setr_epi8(0,2,4,6,(char)0x80,(char)0x80,(char)0x80,(char)0x80, \ (char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80); \ state.r[REG_3] = es40_storeq(_mm_shuffle_epi8(x, m)); \ } while (0) #else #define DO_PKWB do { u64 t = V_2; state.r[REG_3] = (t & U64(0x00000000000000ff)) | \ ((t & U64(0x0000000000ff0000)) >> 8) | ((t & U64(0x000000ff00000000)) >> 16) | ((t & U64(0x00ff000000000000)) >> 24); } while (0) #endif #if defined(ES40_HAVE_SSSE3) #define DO_UNPKBL do { \ __m128i x = es40_loadq(V_2); \ const __m128i m = _mm_setr_epi8(0,(char)0x80,(char)0x80,(char)0x80, 1,(char)0x80,(char)0x80,(char)0x80, \ (char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80); \ state.r[REG_3] = es40_storeq(_mm_shuffle_epi8(x, m)); \ } while (0) #else #define DO_UNPKBL do { u64 t = V_2; state.r[REG_3] = (t & U64(0x000000ff)) | ((t & U64(0x0000ff00)) << 24); } while (0) #endif #if defined(ES40_HAVE_SSSE3) #define DO_UNPKBW do { \ __m128i x = es40_loadq(V_2); \ const __m128i m = _mm_setr_epi8(0,(char)0x80,1,(char)0x80,2,(char)0x80,3,(char)0x80, \ (char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80,(char)0x80); \ state.r[REG_3] = es40_storeq(_mm_shuffle_epi8(x, m)); \ } while (0) #else #define DO_UNPKBW do { u64 t = V_2; state.r[REG_3] = (t & U64(0x000000ff)) | ((t & U64(0x0000ff00)) << 8) | \ ((t & U64(0x00ff0000)) << 16) | ((t & U64(0xff000000)) << 24); } while (0) #endif