gcc/config/picochip/libgccExtras/divmod15.asm

   1 // picoChip ASM file
   2 //
   3 //   Support for 16-bit unsigned division/modulus.
   4 //
   5 //   Copyright (C) 2003, 2004, 2005, 2008  Free Software Foundation, Inc.
   6 //   Contributed by picoChip Designs Ltd.
   7 //   Maintained by Daniel Towner (daniel.towner@picochip.com)
   8 //
   9 //   This file is free software; you can redistribute it and/or modify it
  10 //   under the terms of the GNU General Public License as published by the
  11 //   Free Software Foundation; either version 2, or (at your option) any
  12 //   later version.
  13 //
  14 //   In addition to the permissions in the GNU General Public License, the
  15 //   Free Software Foundation gives you unlimited permission to link the
  16 //   compiled version of this file into combinations with other programs,
  17 //   and to distribute those combinations without any restriction coming
  18 //   from the use of this file.  (The General Public License restrictions
  19 //   do apply in other respects; for example, they cover modification of
  20 //   the file, and distribution when not linked into a combine
  21 //   executable.)
  22 //
  23 //   This file is distributed in the hope that it will be useful, but
  24 //   WITHOUT ANY WARRANTY; without even the implied warranty of
  25 //   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  26 //   General Public License for more details.
  27 //
  28 //   You should have received a copy of the GNU General Public License
  29 //   along with this program; see the file COPYING.  If not, write to
  30 //   the Free Software Foundation, 51 Franklin Street, Fifth Floor,
  31 //   Boston, MA 02110-1301, USA.
  32
  33 .section .text
  34
  35 .global __divmod15
  36 __divmod15:
  37 _picoMark_FUNCTION_BEGIN=
  38
  39 // picoChip Function Prologue : &__divmod15 = 0 bytes
  40
  41 __divmod15:
  42
  43         // The picoChip instruction set has a divstep instruction which
  44         // is used to perform one iteration of a binary division algorithm.
  45         // The instruction allows 16-bit signed division to be implemented.
  46         // It does not directly allow 16-bit unsigned division to be
  47         // implemented. Thus, this function pulls out the common division
  48         // iteration for 15-bits unsigned, and then special wrappers
  49         // provide the logic to change this into a 16-bit signed or
  50         // unsigned division, as appropriate. This allows the two
  51         // versions of division to share a common implementation, reducing
  52         // code size when the two are used together. It also reduces
  53         // the maintenance overhead.
  54
  55         // Input:
  56         //      r0 - dividend
  57         //      r1 - divisor
  58         // Output:
  59         //      r0 - quotient
  60         //      r1 - remainder
  61         // R5 is unused
  62
  63         // Check for special cases. The emphasis is on detecting these as
  64         // quickly as possible, so that the main division can be started. If
  65         // the user requests division by one, division by self, and so on
  66         // then they will just have to accept that this won't be particularly
  67         // quick (relatively), whereas a real division (e.g., dividing a
  68         // large value by a small value) will run as fast as possible
  69         // (i.e., special case detection should not slow down the common case)
  70         //
  71         // Special cases to consider:
  72         //
  73         //      Division by zero.
  74         //      Division of zero.
  75         //      Inputs are equal
  76         //      Divisor is bigger than dividend
  77         //      Division by power of two (can be shifted instead).
  78         //      Division by 1 (special case of power of two division)
  79         //
  80         // Division/modulus by zero is undefined (ISO C:6.5.5), so
  81         // don't bother handling this special case.
  82         //
  83         // The special cases of division by a power of 2 are ignored, since
  84         // they cause the general case to slow down. Omitting these
  85         // special cases also reduces code size considerably.
  86
  87         // Handle divisor >= dividend separately. Note that this also handles
  88         // the case where the dividend is zero. Note that the flags must be
  89         // preserved, since they are also used at the branch destination.
  90         sub.0 r1,r0,r15
  91         sbc r0,r2 \ bge divisorGeDividend
  92 =->     sbc r1,r4
  93
  94         // Compute the shift count. The amount by which the divisor
  95         // must be shifted left to be aligned with the dividend.
  96         sub.0 r4,r2,r3
  97
  98         // Align the divisor to the dividend. Execute a divstep (since at
  99         // least one will always be executed). Skip the remaining loop
 100         // if the shift count is zero.
 101         lsl.0 r1,r3,r1 \ beq skipLoop
 102 =->     divstep r0,r1 \ add.1 r3,1,r2
 103
 104         // Execute the divstep loop until temp is 0. This assumes that the
 105         // loop count is at least one.
 106         sub.0 r3,1,r4
 107 divLoop:
 108         divstep r0,r1 \ bne divLoop
 109 =->     sub.0 r4,1,r4
 110
 111 skipLoop:
 112
 113         // The top bits of the result are the remainder. The bottom
 114         // bits are the quotient.
 115         lsr.0 r0,r2,r1 \ sub.1 16,r2,r4
 116         jr (lr ) \ lsl.0 r0,r4,r0
 117 =->     lsr.0 r0,r4,r0
 118
 119 // Special case.
 120
 121 divisorGeDividend:
 122         // The divisor is greater than or equal to the dividend. The flags
 123         // indicate which of these alternatives it is. The COPYNE can be used
 124         // to set the result appropriately, without introducing any more
 125         // branches.
 126         copy.0 r0,r1 \ copy.1 0,r0
 127         jr (lr) \ copyeq r0,r1
 128 =->     copyeq 1,r0
 129
 130 _picoMark_FUNCTION_END=
 131 // picoChip Function Epilogue : __divmod15
 132
 133
 134 //============================================================================
 135 // All DWARF information between this marker, and the END OF DWARF
 136 // marker should be included in the source file. Search for
 137 // FUNCTION_STACK_SIZE_GOES_HERE and FUNCTION NAME GOES HERE, and
 138 // provide the relevent information. Add markers called
 139 // _picoMark_FUNCTION_BEGIN and _picoMark_FUNCTION_END around the
 140 // function in question.
 141 //============================================================================
 142
 143 //============================================================================
 144 // Frame information.
 145 //============================================================================
 146
 147 .section .debug_frame
 148 _picoMark_DebugFrame=
 149
 150 // Common CIE header.
 151 .unalignedInitLong _picoMark_CieEnd-_picoMark_CieBegin
 152 _picoMark_CieBegin=
 153 .unalignedInitLong 0xffffffff
 154 .initByte 0x1   // CIE Version
 155 .ascii 16#0#    // CIE Augmentation
 156 .uleb128 0x1    // CIE Code Alignment Factor
 157 .sleb128 2      // CIE Data Alignment Factor
 158 .initByte 0xc   // CIE RA Column
 159 .initByte 0xc   // DW_CFA_def_cfa
 160 .uleb128 0xd
 161 .uleb128 0x0
 162 .align 2
 163 _picoMark_CieEnd=
 164
 165 // FDE
 166 _picoMark_LSFDE0I900821033007563=
 167 .unalignedInitLong _picoMark_FdeEnd-_picoMark_FdeBegin
 168 _picoMark_FdeBegin=
 169 .unalignedInitLong _picoMark_DebugFrame // FDE CIE offset
 170 .unalignedInitWord _picoMark_FUNCTION_BEGIN     // FDE initial location
 171 .unalignedInitWord _picoMark_FUNCTION_END-_picoMark_FUNCTION_BEGIN
 172 .initByte 0xe   // DW_CFA_def_cfa_offset
 173 .uleb128 0x0    // <-- FUNCTION_STACK_SIZE_GOES_HERE
 174 .initByte 0x4   // DW_CFA_advance_loc4
 175 .unalignedInitLong _picoMark_FUNCTION_END-_picoMark_FUNCTION_BEGIN
 176 .initByte 0xe   // DW_CFA_def_cfa_offset
 177 .uleb128 0x0
 178 .align 2
 179 _picoMark_FdeEnd=
 180
 181 //============================================================================
 182 // Abbrevation information.
 183 //============================================================================
 184
 185 .section .debug_abbrev
 186 _picoMark_ABBREVIATIONS=
 187
 188 .section .debug_abbrev
 189         .uleb128 0x1    // (abbrev code)
 190         .uleb128 0x11   // (TAG: DW_TAG_compile_unit)
 191         .initByte 0x1   // DW_children_yes
 192         .uleb128 0x10   // (DW_AT_stmt_list)
 193         .uleb128 0x6    // (DW_FORM_data4)
 194         .uleb128 0x12   // (DW_AT_high_pc)
 195         .uleb128 0x1    // (DW_FORM_addr)
 196         .uleb128 0x11   // (DW_AT_low_pc)
 197         .uleb128 0x1    // (DW_FORM_addr)
 198         .uleb128 0x25   // (DW_AT_producer)
 199         .uleb128 0x8    // (DW_FORM_string)
 200         .uleb128 0x13   // (DW_AT_language)
 201         .uleb128 0x5    // (DW_FORM_data2)
 202         .uleb128 0x3    // (DW_AT_name)
 203         .uleb128 0x8    // (DW_FORM_string)
 204 .initByte 0x0
 205 .initByte 0x0
 206
 207         .uleb128 0x2    ;# (abbrev code)
 208         .uleb128 0x2e   ;# (TAG: DW_TAG_subprogram)
 209 .initByte 0x0   ;# DW_children_no
 210         .uleb128 0x3    ;# (DW_AT_name)
 211         .uleb128 0x8    ;# (DW_FORM_string)
 212         .uleb128 0x11   ;# (DW_AT_low_pc)
 213         .uleb128 0x1    ;# (DW_FORM_addr)
 214         .uleb128 0x12   ;# (DW_AT_high_pc)
 215         .uleb128 0x1    ;# (DW_FORM_addr)
 216 .initByte 0x0
 217 .initByte 0x0
 218
 219 .initByte 0x0
 220
 221 //============================================================================
 222 // Line information. DwarfLib requires this to be present, but it can
 223 // be empty.
 224 //============================================================================
 225
 226 .section .debug_line
 227 _picoMark_LINES=
 228
 229 //============================================================================
 230 // Debug Information
 231 //============================================================================
 232 .section .debug_info
 233
 234 //Fixed header.
 235 .unalignedInitLong _picoMark_DEBUG_INFO_END-_picoMark_DEBUG_INFO_BEGIN
 236 _picoMark_DEBUG_INFO_BEGIN=
 237 .unalignedInitWord 0x2
 238 .unalignedInitLong _picoMark_ABBREVIATIONS
 239 .initByte 0x2
 240
 241 // Compile unit information.
 242 .uleb128 0x1    // (DIE 0xb) DW_TAG_compile_unit)
 243 .unalignedInitLong _picoMark_LINES
 244 .unalignedInitWord _picoMark_FUNCTION_END
 245 .unalignedInitWord _picoMark_FUNCTION_BEGIN
 246 // Producer is `picoChip'
 247 .ascii 16#70# 16#69# 16#63# 16#6f# 16#43# 16#68# 16#69# 16#70# 16#00#
 248 .unalignedInitWord 0xcafe // ASM language
 249 .ascii 16#0# // Name. DwarfLib expects this to be present.
 250
 251 .uleb128 0x2    ;# (DIE DW_TAG_subprogram)
 252
 253 // FUNCTION NAME GOES HERE. Use `echo name | od -t x1' to get the hex. Each hex
 254 // digit is specified using the format 16#XX#
 255 .ascii 16#5f# 16#64# 16#69# 16#76# 16#6d# 16#6f# 16#64# 16#31# 16#35# 16#0# // Function name `_divmod15'
 256 .unalignedInitWord _picoMark_FUNCTION_BEGIN     // DW_AT_low_pc
 257 .unalignedInitWord _picoMark_FUNCTION_END       // DW_AT_high_pc
 258
 259 .initByte 0x0   // end of compile unit children.
 260
 261 _picoMark_DEBUG_INFO_END=
 262
 263 //============================================================================
 264 // END OF DWARF
 265 //============================================================================
 266
 267 .section .endFile
 268 // End of picoChip ASM file