Dogcows Code - chaz/yoink/blob - doc/cml_parameters.txt

   1 CML_ENABLE_MATRIX_BRACES
   2 - Allow matrices to be accessed as A[row][col] in addition to A(row,col).
   3   This is primarily for testing, and is not used anywhere in the library
   4   proper since the proxy object mechanism that's used is not guaranteed to
   5   produce optimal code.
   6
   7 CML_NO_DEFAULT_EXPR_TRAITS
   8 - Don't use default expression traits for built-in types.  This is handy
   9   for debugging, since no classes or built-in types will have an expression
  10   traits class by default (i.e. the expression traits class must be created
  11   explicitly).
  12
  13 CML_DEFAULT_BASIS_ORIENTATION
  14 - Either cml::row_basis or cml::col_basis.  This determines the basis
  15   vector orientation for the matrix result of a vector outer product.  The
  16   default is cml::col_basis.
  17
  18 CML_DEFAULT_ARRAY_LAYOUT
  19 - Either cml::row_major or cml::col_major.  This determines the layout of the
  20   resulting matrix when promoting matrices with different layouts.  If the
  21   matrices have the same layout, the promoted type will also have that layout,
  22   unless CML_ALWAYS_PROMOTE_TO_DEFAULT_LAYOUT is defined.  This parameter also
  23   determines the default 2D array memory layout.
  24
  25 CML_DEFAULT_ARRAY_ALLOC
  26 - Defines the default allocator for arrays generated at run-time by the type
  27   deduction system.  The default value is std::allocator<void>.
  28
  29 CML_ALWAYS_PROMOTE_TO_DEFAULT_LAYOUT
  30 - If this is defined, 2D arrays will always be promoted to the default layout.
  31   This is useful if matrix expressions should always be handled in
  32   a particular orientation for efficiency.
  33
  34 CML_VECTOR_UNROLL_LIMIT=<N>
  35 - <N> is the maximum vector length that should be automatically unrolled by
  36   the expression template engine.  If the vector (expression) is longer,
  37   the normal for loop is used.  Note also that unrolling is available only
  38   for fixed-size vectors.
  39
  40 CML_2D_UNROLLER
  41 - Use a 2D unroller for matrix expressions.  This causes very poor code for
  42   most of the supported compilers, but with GCC4/G5, it's required for
  43   optimal code.  Note that unrolling is available only for fixed-size
  44   matrices.
  45
  46 CML_MATRIX_UNROLL_LIMIT=<N>
  47 - <N> is the maximum number of matrix elements (#rows x #cols) that
  48   a matrix (expression) can have before being unrolled.  A 2D loop is used
  49   for matrices with more than <N> elements.  When either CML_2D_UNROLLER or
  50   CML_2D_TO_1D_UNROLLER is defined, this must also be defined.
  51
  52 CML_NO_2D_UNROLLER
  53 - Don't unroll at all, just use a loop.  This seems to generate the best
  54   code on at least GCC4/x86 and Intel 9/Linux/x86.
  55
  56 CML_RECIPROCAL_OPTIMIZATION
  57 - Use "*= 1./x" instead of "/= x" for per-element division.  This may generate
  58   better code for certain systems, but isn't yet tested for any configuration.
  59
  60 CML_AUTOMATIC_VECTOR_RESIZE_ON_ASSIGNMENT
  61 - Enable automatic resizing of dynamic vectors on assignment and copy from
  62   other vectors and expressions.  This guarantees that vectors have the right
  63   size on assignment, but imposes a significant performance penalty when
  64   vectors already have the correct size.  Without this option, however,
  65   assigned-to vectors must be constructed with the correct size, or must be
  66   explicitly resized with the resize() method.  Note that this has no
  67   effect when assigning using a scalar operator, since the vector must already
  68   have the proper size.  Also note that temporaries will be automatically
  69   resized as necessary regardless of the setting of this option.
  70
  71 CML_CHECK_VECTOR_EXPR_SIZES
  72 - Verify that all arguments in a vector expression have the proper size.  This
  73   is particularly handy for debugging when CML_AUTOMATIC_VECTOR_RESIZE is
  74   undefined.  Note that the performance penalty for defining this option is
  75   dependent upon the number of binary expressions and the compiler.  Also,
  76   this has no effect for expressions composed of entirely fixed-size vectors,
  77   since the sizes are automatically checked at compile time.
  78
  79 CML_AUTOMATIC_MATRIX_RESIZE_ON_ASSIGNMENT
  80 CML_CHECK_MATRIX_EXPR_SIZES
  81 CML_CHECK_MATVEC_EXPR_SIZES
  82 - Same as CML_AUTOMATIC_VECTOR_RESIZE and CML_CHECK_VECTOR_EXPR_SIZES, but for
  83   matrices.
  84
  85 CML_USE_GENERATED_MATRIX_ASSIGN_OP
  86 - Allow the compiler to generate its own matrix assignment code.  This yields
  87   better performance at least on Intel C++ 9/x86 Linux.