Engine directory for ticket #1

Engine/source/math/mMathAltivec.cpp

@@ -0,0 +1,136 @@
+//-----------------------------------------------------------------------------
+// Copyright (c) 2012 GarageGames, LLC
+//
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to
+// deal in the Software without restriction, including without limitation the
+// rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+// sell copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in
+// all copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+// FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+// IN THE SOFTWARE.
+//-----------------------------------------------------------------------------
+
+// This file is Mac specific.
+#if defined( __APPLE__ )
+
+#include <vecLib/vecLib.h>
+#include "math/mMathFn.h"
+#include "console/console.h"
+#include "platform/profiler.h"
+
+#if defined( __VEC__ )
+
+// tests show BLAS to be about 4x slower than aligned altivec, 3x slower than unaligned altivec code below.
+
+/// Altivec 4x4 Matrix multiplication.
+/// Most of our time is spent moving data in & out of the vector registers.
+/// Alignment of the matrix data to 16-byte boundaries is very important,
+/// because we get a much better speed gain if we can assume the data is aligned.
+void vec_MatrixF_x_MatrixF(const F32 *matA, const F32 *matB, F32 *result)
+{
+   vector float A[4][1];
+   vector float B[4][1];
+   vector float C[4][1];
+
+   /// If the incoming pointers are not 16-byte aligned, we have to load & store the slow way.
+   if((int)matA & 0xF || (int)matB & 0xF || (int)result & 0xF)
+   {
+      F32 *loader;
+      loader = (F32*) &A;
+      loader[0] = matA[0];
+      loader[1] = matA[1];
+      loader[2] = matA[2];
+      loader[3] = matA[3];
+      loader[4] = matA[4];
+      loader[5] = matA[5];
+      loader[6] = matA[6];
+      loader[7] = matA[7];
+      loader[8] = matA[8];
+      loader[9] = matA[9];
+      loader[10] = matA[10];
+      loader[11] = matA[11];
+      loader[12] = matA[12];
+      loader[13] = matA[13];
+      loader[14] = matA[14];
+      loader[15] = matA[15];
+      loader = (F32*) &B;
+      loader[0] = matB[0];
+      loader[1] = matB[1];
+      loader[2] = matB[2];
+      loader[3] = matB[3];
+      loader[4] = matB[4];
+      loader[5] = matB[5];
+      loader[6] = matB[6];
+      loader[7] = matB[7];
+      loader[8] = matB[8];
+      loader[9] = matB[9];
+      loader[10] = matB[10];
+      loader[11] = matB[11];
+      loader[12] = matB[12];
+      loader[13] = matB[13];
+      loader[14] = matB[14];
+      loader[15] = matB[15];
+      
+      vMultMatMat_4x4( A, B, C);
+      
+      loader = (F32*) &C;
+      
+      result[0] = loader[0];
+      result[1] = loader[1];
+      result[2] = loader[2];
+      result[3] = loader[3];
+      result[4] = loader[4];
+      result[5] = loader[5];
+      result[6] = loader[6];
+      result[7] = loader[7];
+      result[8] = loader[8];
+      result[9] = loader[9];
+      result[10] = loader[10];
+      result[11] = loader[11];
+      result[12] = loader[12];
+      result[13] = loader[13];
+      result[14] = loader[14];
+      result[15] = loader[15];
+   }
+   else
+   {
+      A[0][0] = vec_ld(0,  matA);
+      A[1][0] = vec_ld(16, matA);
+      A[2][0] = vec_ld(32, matA);
+      A[3][0] = vec_ld(48, matA);
+      B[0][0] = vec_ld(0,  matB);
+      B[1][0] = vec_ld(16, matB);
+      B[2][0] = vec_ld(32, matB);
+      B[3][0] = vec_ld(48, matB);
+      
+      vMultMatMat_4x4( A, B, C);
+      
+      vec_st(C[0][0], 0,  result);
+      vec_st(C[1][0], 16, result);
+      vec_st(C[2][0], 32, result);
+      vec_st(C[3][0], 48, result);
+   }
+}
+
+void mInstallLibrary_Vec()
+{
+   m_matF_x_matF           = vec_MatrixF_x_MatrixF;
+}
+#else // defined(__VEC__)
+void mInstallLibrary_Vec()
+{
+   Con::warnf("Cannot use altivec math, this build does not support altivec.");
+}
+#endif// defined(__VEC__)
+
+#endif// defined(__APPLE__)