.file "fdim.s"

// Copyright (c) 2001 - 2003, Intel Corporation

// All rights reserved.

//

// Contributed 2001 by the Intel Numerics Group, Intel Corporation

//

// Redistribution and use in source and binary forms, with or without

// modification, are permitted provided that the following conditions are

// met:

//

// * Redistributions of source code must retain the above copyright

// notice, this list of conditions and the following disclaimer.

//

// * Redistributions in binary form must reproduce the above copyright

// notice, this list of conditions and the following disclaimer in the

// documentation and/or other materials provided with the distribution.

//

// * The name of Intel Corporation may not be used to endorse or promote

// products derived from this software without specific prior written

// permission.

// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS

// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT

// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR

// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS

// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,

// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,

// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR

// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY

// OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING

// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS

// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

//

// Intel Corporation is the author of this code, and requests that all

// problem reports or change requests be submitted to it directly at

// http://www.intel.com/software/products/opensource/libraries/num.htm.

//

// History

//==============================================================

// 06/08/01 Initial version

// 08/23/01 Corrected error tag number

// 05/20/02 Cleaned up namespace and sf0 syntax

// 01/28/03 Improved performance

//

// API

//==============================================================

// double fdim( double x, double y );

// input  floating point f8, f9

// output floating point f8

//

//

// Overview of operation

//==============================================================

// fdim determines the positive difference between the arguments

//   Result = x - y   if x > y

//          = +0      if x <= y

//

// Error support is called if x-y overflows for x > y

//

// Registers used

//==============================================================

// General purpose registers: r14, r32 - r39

rExpBig               = r14

// r36-39 parameters for libm_error_support

GR_SAVE_B0            = r33

GR_SAVE_GP            = r34

GR_SAVE_PFS           = r35

GR_Parameter_X        = r36

GR_Parameter_Y        = r37

GR_Parameter_RESULT   = r38

GR_Parameter_TAG      = r39

// Floating-point registers: f8 - f12

f_tmp_result          = f10

fBig                  = f11

fNormX                = f12

// Predicate registers: p6 - p10

.section .text

GLOBAL_LIBM_ENTRY(fdim)

{ .mfi

      mov             rExpBig = 0x103ff  // Exponent to indicate overflow

      fcmp.le.s1      p6,p7 = f8, f9     // Is x <= y?

      nop.i           0

}

{ .mfi

      nop.m           0

      fnorm.s1        fNormX = f8        // Save x

      nop.i           0

}

;;

{ .mfi

      setf.exp        fBig = rExpBig     // Constant to test for overflow

      fcmp.eq.s0      p8,p0 = f8, f9     // Dummy op to set Denormal or Invalid

      nop.i           0

}

;;

{ .mfi

      nop.m           0

      fclass.m        p9,p10 = f8, 0x1e3 // Test for x natval, nan, inf

      nop.i           0

}

;;

{ .mfi

      nop.m           0

(p6)  fmerge.s        f8 = f0, f0        // Result is +0 if x <= y

      nop.i           0

}

{ .mfi

      nop.m           0

(p7)  fms.d.s0        f8 = f8, f1, f9    // Result is x - y if x > y

      nop.i           0

}

;;

{ .mfi

      nop.m           0

(p10) fclass.m        p9,p10 = f9, 0x1e3 // Test for y natval, nan, inf

      nop.i           0

}

;;

{ .mfi

      nop.m           0

(p10) fcmp.ge.s1      p8,p0 = f8, fBig   // Test result for overflow

      nop.i           0

}

;;

{ .mbb

(p9)  cmp.ne          p8,p0 = r0,r0      // Clear p8 if x or y natval,nan,inf

(p8)  br.cond.spnt    FDIM_OVERFLOW      // Branch if result overflows

      br.ret.sptk     b0                 // Normal return

}

;;

// Here if result will overflow

FDIM_OVERFLOW:

{ .mfi

      alloc           r32=ar.pfs,2,2,4,0

      fms.d.s0        f_tmp_result = f8,f1,f9 // Normalize result force overflow

      nop.i           0

}

{ .mfb

      mov             GR_Parameter_TAG = 196  // Error code

      nop.f           0

      br.cond.sptk    __libm_error_region     // Branch to error code

}

;;

GLOBAL_LIBM_END(fdim)

libm_alias_double_other (fdim, fdim)

LOCAL_LIBM_ENTRY(__libm_error_region)

// Call error support to report possible range error

.prologue

{ .mfi

        add   GR_Parameter_Y=-32,sp            // Parameter 2 value

        nop.f 0

.save   ar.pfs,GR_SAVE_PFS

        mov  GR_SAVE_PFS=ar.pfs                // Save ar.pfs

}

{ .mfi

.fframe 64

        add sp=-64,sp                          // Create new stack

        nop.f 0

        mov GR_SAVE_GP=gp                      // Save gp

};;

{ .mmi

        stfd [GR_Parameter_Y] = f9,16          // STORE Parameter 2 on stack

        add GR_Parameter_X = 16,sp             // Parameter 1 address

.save   b0, GR_SAVE_B0

        mov GR_SAVE_B0=b0                      // Save b0

};;

.body

{ .mib

        stfd [GR_Parameter_X] = fNormX         // STORE Parameter 1 on stack

        add   GR_Parameter_RESULT = 0,GR_Parameter_Y // Parameter 3 address

        nop.b 0

}

{ .mib

        stfd [GR_Parameter_Y] = f_tmp_result   // STORE Parameter 3 on stack

        add   GR_Parameter_Y = -16,GR_Parameter_Y

        br.call.sptk b0=__libm_error_support#  // Call error handling function

};;

{ .mmi

        add   GR_Parameter_RESULT = 48,sp

        nop.m 0

        nop.i 0

};;

{ .mmi

        ldfd  f8 = [GR_Parameter_RESULT]       // Get return result off stack

.restore sp

        add   sp = 64,sp                       // Restore stack pointer

        mov   b0 = GR_SAVE_B0                  // Restore return address

};;

{ .mib

        mov   gp = GR_SAVE_GP                  // Restore gp

        mov   ar.pfs = GR_SAVE_PFS             // Restore ar.pfs

        br.ret.sptk     b0                     // Return

};;

LOCAL_LIBM_END(__libm_error_region)

.type   __libm_error_support#,@function

.global __libm_error_support#