00001 /* 00002 ----------------------------------------------------------------------------- 00003 This source file is part of OGRE 00004 (Object-oriented Graphics Rendering Engine) 00005 For the latest info, see http://www.ogre3d.org/ 00006 00007 Copyright (c) 2000-2006 Torus Knot Software Ltd 00008 Also see acknowledgements in Readme.html 00009 00010 This program is free software; you can redistribute it and/or modify it under 00011 the terms of the GNU Lesser General Public License as published by the Free Software 00012 Foundation; either version 2 of the License, or (at your option) any later 00013 version. 00014 00015 This program is distributed in the hope that it will be useful, but WITHOUT 00016 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS 00017 FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. 00018 00019 You should have received a copy of the GNU Lesser General Public License along with 00020 this program; if not, write to the Free Software Foundation, Inc., 59 Temple 00021 Place - Suite 330, Boston, MA 02111-1307, USA, or go to 00022 http://www.gnu.org/copyleft/lesser.txt. 00023 00024 You may alternatively use this source under the terms of a specific version of 00025 the OGRE Unrestricted License provided you have obtained such a license from 00026 Torus Knot Software Ltd. 00027 ----------------------------------------------------------------------------- 00028 */ 00029 #ifndef __SIMDHelper_H__ 00030 #define __SIMDHelper_H__ 00031 00032 #include "OgrePrerequisites.h" 00033 #include "OgrePlatformInformation.h" 00034 00035 // Stack-alignment hackery. 00036 // 00037 // If macro __OGRE_SIMD_ALIGN_STACK defined, means there requests 00038 // special code to ensure stack align to a 16-bytes boundary. 00039 // 00040 // Note: 00041 // This macro can only guarantee callee stack pointer (esp) align 00042 // to a 16-bytes boundary, but not that for frame pointer (ebp). 00043 // Because most compiler might use frame pointer to access to stack 00044 // variables, so you need to wrap those alignment required functions 00045 // with extra function call. 00046 // 00047 #if defined(__INTEL_COMPILER) 00048 // For intel's compiler, simply calling alloca seems to do the right 00049 // thing. The size of the allocated block seems to be irrelevant. 00050 #define __OGRE_SIMD_ALIGN_STACK() _alloca(16) 00051 00052 #elif OGRE_CPU == OGRE_CPU_X86 && OGRE_COMPILER == OGRE_COMPILER_GNUC 00053 // 00054 // Horrible hack to align the stack to a 16-bytes boundary for gcc. 00055 // 00056 // We assume a gcc version >= 2.95 so that 00057 // -mpreferred-stack-boundary works. Otherwise, all bets are 00058 // off. However, -mpreferred-stack-boundary does not create a 00059 // stack alignment, but it only preserves it. Unfortunately, 00060 // since Ogre are designed as a flexibility library, user might 00061 // compile their application with wrong stack alignment, even 00062 // if user taken care with stack alignment, but many versions 00063 // of libc on linux call main() with the wrong initial stack 00064 // alignment the result that the code is now pessimally aligned 00065 // instead of having a 50% chance of being correct. 00066 // 00067 #define __OGRE_SIMD_ALIGN_STACK() \ 00068 { \ 00069 /* Use alloca to allocate some memory on the stack. */ \ 00070 /* This alerts gcc that something funny is going on, */ \ 00071 /* so that it does not omit the frame pointer etc. */ \ 00072 (void)__builtin_alloca(16); \ 00073 /* Now align the stack pointer */ \ 00074 __asm__ __volatile__ ("andl $-16, %esp"); \ 00075 } 00076 00077 #elif defined(_MSC_VER) 00078 // Fortunately, MSVC will align the stack automatically 00079 00080 #endif 00081 00082 00083 // Additional platform-dependent header files and declares. 00084 // 00085 // NOTE: Should be sync with __OGRE_HAVE_SSE macro. 00086 // 00087 00088 #if OGRE_DOUBLE_PRECISION == 0 && OGRE_CPU == OGRE_CPU_X86 00089 00090 #if OGRE_COMPILER == OGRE_COMPILER_MSVC || defined(__INTEL_COMPILER) 00091 #include "OgreNoMemoryMacros.h" 00092 #include <xmmintrin.h> 00093 #include "OgreMemoryMacros.h" 00094 00095 #elif OGRE_COMPILER == OGRE_COMPILER_GNUC 00096 // Don't define ourself version SSE intrinsics if "xmmintrin.h" already included. 00097 // 00098 // Note: gcc in some platform already included "xmmintrin.h" for some reason. 00099 // I pick up macro _XMMINTRIN_H_INCLUDED here which based on the "xmmintrin.h" 00100 // comes with cygwin gcc 3.4.4, guess it should be solved duplicate definition 00101 // problem on gcc for x86. 00102 // 00103 #if !defined(_XMMINTRIN_H_INCLUDED) 00104 00105 // Simulate VC/ICC intrinsics. Only used intrinsics are declared here. 00106 00107 typedef float __m128 __attribute__ ((mode(V4SF),aligned(16))); 00108 typedef int __m64 __attribute__ ((mode(V2SI))); 00109 00110 // Macro for declare intrinsic routines always inline even if in debug build 00111 #define __ALWAYS_INLINE FORCEINLINE __attribute__ ((__always_inline__)) 00112 00113 // Shuffle instruction must be declare as macro 00114 00115 #define _MM_SHUFFLE(fp3,fp2,fp1,fp0) \ 00116 (((fp3) << 6) | ((fp2) << 4) | ((fp1) << 2) | ((fp0))) 00117 00118 #define _mm_shuffle_ps(a, b, imm8) __extension__ \ 00119 ({ \ 00120 __m128 result; \ 00121 __asm__("shufps %3, %2, %0" : "=x" (result) : "0" (a), "xm" (b), "N" (imm8)); \ 00122 result; \ 00123 }) 00124 00125 00126 // Load/store instructions 00127 00128 #define __MM_DECL_LD(name, instruction, type) \ 00129 static __ALWAYS_INLINE __m128 _mm_##name(const type *addr) \ 00130 { \ 00131 __m128 result; \ 00132 __asm__( #instruction " %1, %0" : "=x" (result) : "m" (*addr)); \ 00133 return result; \ 00134 } 00135 00136 #define __MM_DECL_LD2(name, instruction, type) \ 00137 static __ALWAYS_INLINE __m128 _mm_##name(__m128 val, const type *addr) \ 00138 { \ 00139 __m128 result; \ 00140 __asm__( #instruction " %2, %0" : "=x" (result) : "0"(val), "m" (*addr)); \ 00141 return result; \ 00142 } 00143 00144 #define __MM_DECL_ST(name, instruction, type) \ 00145 static __ALWAYS_INLINE void _mm_##name(type *addr, __m128 val) \ 00146 { \ 00147 __asm__( #instruction " %1, %0" : "=m" (*addr) : "x" (val)); \ 00148 } 00149 00150 __MM_DECL_LD(loadu_ps, movups, float) 00151 __MM_DECL_ST(storeu_ps, movups, float) 00152 00153 __MM_DECL_LD(load_ss, movss, float) 00154 __MM_DECL_ST(store_ss, movss, float) 00155 00156 __MM_DECL_ST(storel_pi, movlps, __m64) 00157 __MM_DECL_ST(storeh_pi, movhps, __m64) 00158 __MM_DECL_LD2(loadl_pi, movlps, __m64) 00159 __MM_DECL_LD2(loadh_pi, movhps, __m64) 00160 00161 #undef __MM_DECL_LD 00162 #undef __MM_DECL_LD2 00163 #undef __MM_DECL_ST 00164 00165 // Two operand instructions 00166 00167 #define __MM_DECL_OP2(name, instruction, constraint) \ 00168 static __ALWAYS_INLINE __m128 _mm_##name(__m128 a, __m128 b) \ 00169 { \ 00170 __m128 result; \ 00171 __asm__( #instruction " %2, %0" : "=x" (result) : "0" (a), #constraint (b)); \ 00172 return result; \ 00173 } 00174 00175 __MM_DECL_OP2(add_ps, addps, xm) 00176 __MM_DECL_OP2(add_ss, addss, xm) 00177 __MM_DECL_OP2(sub_ps, subps, xm) 00178 __MM_DECL_OP2(sub_ss, subss, xm) 00179 __MM_DECL_OP2(mul_ps, mulps, xm) 00180 __MM_DECL_OP2(mul_ss, mulss, xm) 00181 00182 __MM_DECL_OP2(xor_ps, xorps, xm) 00183 00184 __MM_DECL_OP2(unpacklo_ps, unpcklps, xm) 00185 __MM_DECL_OP2(unpackhi_ps, unpckhps, xm) 00186 00187 __MM_DECL_OP2(movehl_ps, movhlps, x) 00188 __MM_DECL_OP2(movelh_ps, movlhps, x) 00189 00190 __MM_DECL_OP2(cmpnle_ps, cmpnleps, xm) 00191 00192 #undef __MM_DECL_OP2 00193 00194 // Other used instructions 00195 00196 static __ALWAYS_INLINE __m128 _mm_load_ps1(const float *addr) 00197 { 00198 __m128 tmp = _mm_load_ss(addr); 00199 return _mm_shuffle_ps(tmp, tmp, 0); 00200 } 00201 00202 static __ALWAYS_INLINE __m128 _mm_setzero_ps(void) 00203 { 00204 __m128 result; 00205 __asm__("xorps %0, %0" : "=x" (result)); 00206 return result; 00207 } 00208 00209 static __ALWAYS_INLINE __m128 _mm_rsqrt_ps(__m128 val) 00210 { 00211 __m128 result; 00212 __asm__("rsqrtps %1, %0" : "=x" (result) : "xm" (val)); 00213 //__asm__("rsqrtps %0, %0" : "=x" (result) : "0" (val)); 00214 return result; 00215 } 00216 00217 static __ALWAYS_INLINE int _mm_movemask_ps(__m128 val) 00218 { 00219 int result; 00220 __asm__("movmskps %1, %0" : "=r" (result) : "x" (val)); 00221 return result; 00222 } 00223 00224 #endif // !defined(_XMMINTRIN_H_INCLUDED) 00225 00226 #endif // OGRE_COMPILER == OGRE_COMPILER_GNUC 00227 00228 #endif // OGRE_DOUBLE_PRECISION == 0 && OGRE_CPU == OGRE_CPU_X86 00229 00230 00231 00232 //--------------------------------------------------------------------- 00233 // SIMD macros and helpers 00234 //--------------------------------------------------------------------- 00235 00236 00237 namespace Ogre { 00238 00239 #if __OGRE_HAVE_SSE 00240 00251 #if 1 00252 #define __MM_RSQRT_PS(x) _mm_rsqrt_ps(x) 00253 #else 00254 #define __MM_RSQRT_PS(x) __mm_rsqrt_nr_ps(x) // Implemented below 00255 #endif 00256 00265 #define __MM_TRANSPOSE4x4_PS(r0, r1, r2, r3) \ 00266 { \ 00267 __m128 t3, t2, t1, t0; \ 00268 \ 00269 /* r00 r01 r02 r03 */ \ 00270 /* r10 r11 r12 r13 */ \ 00271 /* r20 r21 r22 r23 */ \ 00272 /* r30 r31 r32 r33 */ \ 00273 \ 00274 t0 = _mm_unpacklo_ps(r0, r1); /* r00 r10 r01 r11 */ \ 00275 t2 = _mm_unpackhi_ps(r0, r1); /* r02 r12 r03 r13 */ \ 00276 t1 = _mm_unpacklo_ps(r2, r3); /* r20 r30 r21 r31 */ \ 00277 t3 = _mm_unpackhi_ps(r2, r3); /* r22 r32 r23 r33 */ \ 00278 \ 00279 r0 = _mm_movelh_ps(t0, t1); /* r00 r10 r20 r30 */ \ 00280 r1 = _mm_movehl_ps(t1, t0); /* r01 r11 r21 r31 */ \ 00281 r2 = _mm_movelh_ps(t2, t3); /* r02 r12 r22 r32 */ \ 00282 r3 = _mm_movehl_ps(t3, t2); /* r03 r13 r23 r33 */ \ 00283 } 00284 00293 #define __MM_TRANSPOSE4x3_PS(v0, v1, v2) \ 00294 { \ 00295 __m128 t0, t1, t2; \ 00296 \ 00297 /* r00 r01 r02 r10 */ \ 00298 /* r11 r12 r20 r21 */ \ 00299 /* r22 r30 r31 r32 */ \ 00300 \ 00301 t0 = _mm_shuffle_ps(v0, v2, _MM_SHUFFLE(3,0,3,0)); /* r00 r10 r22 r32 */ \ 00302 t1 = _mm_shuffle_ps(v0, v1, _MM_SHUFFLE(1,0,2,1)); /* r01 r02 r11 r12 */ \ 00303 t2 = _mm_shuffle_ps(v1, v2, _MM_SHUFFLE(2,1,3,2)); /* r20 r21 r30 r31 */ \ 00304 \ 00305 v0 = _mm_shuffle_ps(t0, t2, _MM_SHUFFLE(2,0,1,0)); /* r00 r10 r20 r30 */ \ 00306 v1 = _mm_shuffle_ps(t1, t2, _MM_SHUFFLE(3,1,2,0)); /* r01 r11 r21 r31 */ \ 00307 v2 = _mm_shuffle_ps(t1, t0, _MM_SHUFFLE(3,2,3,1)); /* r02 r12 r22 r32 */ \ 00308 } 00309 00317 #define __MM_TRANSPOSE3x4_PS(v0, v1, v2) \ 00318 { \ 00319 __m128 t0, t1, t2; \ 00320 \ 00321 /* r00 r10 r20 r30 */ \ 00322 /* r01 r11 r21 r31 */ \ 00323 /* r02 r12 r22 r32 */ \ 00324 \ 00325 t0 = _mm_shuffle_ps(v0, v2, _MM_SHUFFLE(2,0,3,1)); /* r10 r30 r02 r22 */ \ 00326 t1 = _mm_shuffle_ps(v1, v2, _MM_SHUFFLE(3,1,3,1)); /* r11 r31 r12 r32 */ \ 00327 t2 = _mm_shuffle_ps(v0, v1, _MM_SHUFFLE(2,0,2,0)); /* r00 r20 r01 r21 */ \ 00328 \ 00329 v0 = _mm_shuffle_ps(t2, t0, _MM_SHUFFLE(0,2,2,0)); /* r00 r01 r02 r10 */ \ 00330 v1 = _mm_shuffle_ps(t1, t2, _MM_SHUFFLE(3,1,2,0)); /* r11 r12 r20 r21 */ \ 00331 v2 = _mm_shuffle_ps(t0, t1, _MM_SHUFFLE(3,1,1,3)); /* r22 r30 r31 r32 */ \ 00332 } 00333 00337 #define __MM_SELECT(v, fp) \ 00338 _mm_shuffle_ps((v), (v), _MM_SHUFFLE((fp),(fp),(fp),(fp))) 00339 00341 #define __MM_ACCUM4_PS(a, b, c, d) \ 00342 _mm_add_ps(_mm_add_ps(a, b), _mm_add_ps(c, d)) 00343 00347 #define __MM_DOT4x4_PS(a0, a1, a2, a3, b0, b1, b2, b3) \ 00348 __MM_ACCUM4_PS(_mm_mul_ps(a0, b0), _mm_mul_ps(a1, b1), _mm_mul_ps(a2, b2), _mm_mul_ps(a3, b3)) 00349 00353 #define __MM_DOT4x3_PS(r0, r1, r2, r3, v0, v1, v2) \ 00354 __MM_ACCUM4_PS(_mm_mul_ps(r0, v0), _mm_mul_ps(r1, v1), _mm_mul_ps(r2, v2), r3) 00355 00357 #define __MM_ACCUM3_PS(a, b, c) \ 00358 _mm_add_ps(_mm_add_ps(a, b), c) 00359 00363 #define __MM_DOT3x3_PS(r0, r1, r2, v0, v1, v2) \ 00364 __MM_ACCUM3_PS(_mm_mul_ps(r0, v0), _mm_mul_ps(r1, v1), _mm_mul_ps(r2, v2)) 00365 00367 #define __MM_MADD_PS(a, b, c) \ 00368 _mm_add_ps(_mm_mul_ps(a, b), c) 00369 00371 #define __MM_LERP_PS(t, a, b) \ 00372 __MM_MADD_PS(_mm_sub_ps(b, a), t, a) 00373 00375 #define __MM_MADD_SS(a, b, c) \ 00376 _mm_add_ss(_mm_mul_ss(a, b), c) 00377 00379 #define __MM_LERP_SS(t, a, b) \ 00380 __MM_MADD_SS(_mm_sub_ss(b, a), t, a) 00381 00383 #define __MM_LOAD_PS(p) \ 00384 (*(__m128*)(p)) 00385 00387 #define __MM_STORE_PS(p, v) \ 00388 (*(__m128*)(p) = (v)) 00389 00390 00393 template <bool aligned = false> 00394 struct SSEMemoryAccessor 00395 { 00396 static FORCEINLINE __m128 load(const float *p) 00397 { 00398 return _mm_loadu_ps(p); 00399 } 00400 static FORCEINLINE void store(float *p, const __m128& v) 00401 { 00402 _mm_storeu_ps(p, v); 00403 } 00404 }; 00405 // Special aligned accessor 00406 template <> 00407 struct SSEMemoryAccessor<true> 00408 { 00409 static FORCEINLINE const __m128& load(const float *p) 00410 { 00411 return __MM_LOAD_PS(p); 00412 } 00413 static FORCEINLINE void store(float *p, const __m128& v) 00414 { 00415 __MM_STORE_PS(p, v); 00416 } 00417 }; 00418 00421 static FORCEINLINE bool _isAlignedForSSE(const void *p) 00422 { 00423 return (((size_t)p) & 15) == 0; 00424 } 00425 00429 static FORCEINLINE __m128 __mm_rsqrt_nr_ps(const __m128& x) 00430 { 00431 static const __m128 v0pt5 = { 0.5f, 0.5f, 0.5f, 0.5f }; 00432 static const __m128 v3pt0 = { 3.0f, 3.0f, 3.0f, 3.0f }; 00433 __m128 t = _mm_rsqrt_ps(x); 00434 return _mm_mul_ps(_mm_mul_ps(v0pt5, t), 00435 _mm_sub_ps(v3pt0, _mm_mul_ps(_mm_mul_ps(x, t), t))); 00436 } 00437 00438 // Macro to check the stack aligned for SSE 00439 #if OGRE_DEBUG_MODE 00440 #define __OGRE_CHECK_STACK_ALIGNED_FOR_SSE() \ 00441 { \ 00442 __m128 test; \ 00443 assert(_isAlignedForSSE(&test)); \ 00444 } 00445 00446 #else // !OGRE_DEBUG_MODE 00447 #define __OGRE_CHECK_STACK_ALIGNED_FOR_SSE() 00448 00449 #endif // OGRE_DEBUG_MODE 00450 00451 00452 #endif // __OGRE_HAVE_SSE 00453 00454 } 00455 00456 #endif // __SIMDHelper_H__
Copyright © 2000-2005 by The OGRE Team
This work is licensed under a Creative Commons Attribution-ShareAlike 2.5 License.
Last modified Thu Mar 6 09:46:26 2008