// picoChip ASM file //.file "popcounthi2.S" // // Support for 16-bit population count. // // Copyright (C) 2003, 2004, 2005, 2008, 2009 Free Software Foundation, Inc. // Contributed by picoChip Designs Ltd. // Maintained by Daniel Towner (daniel.towner@picochip.com) // // This file is free software; you can redistribute it and/or modify it // under the terms of the GNU General Public License as published by the // Free Software Foundation; either version 3, or (at your option) any // later version. // // This file is distributed in the hope that it will be useful, but // WITHOUT ANY WARRANTY; without even the implied warranty of // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU // General Public License for more details. // // Under Section 7 of GPL version 3, you are granted additional // permissions described in the GCC Runtime Library Exception, version // 3.1, as published by the Free Software Foundation. // // You should have received a copy of the GNU General Public License and // a copy of the GCC Runtime Library Exception along with this program; // see the files COPYING3 and COPYING.RUNTIME respectively. If not, see // . .section .text // The following code (taken from a newsgroup posting) was compiled, and then // hand assembled (a similar version is given in the Hacker's Delight // book, chapter 5). // // int // popcount (int value) // { // value = ((value & 0xAAAA) >> 1) + (value & 0x5555); // value = ((value & 0xCCCC) >> 2) + (value & 0x3333); // value = ((value & 0xF0F0) >> 4) + (value & 0x0F0F); // return ((value & 0xFF00) >> 8) + (value & 0x00FF); // } // // This assembly function is approx. 20x faster than a naive loop // implementation of the population count, but about 30% bigger // (45 bytes v. 34 bytes). .align 8 .global ___popcounthi2 ___popcounthi2: _picoMark_FUNCTION_BEGIN= // picoChip Function Prologue : &___popcounthi2 = 0 bytes AND.0 [LSR R0,1],21845,R0 \ AND.1 R0,21845,R5 ADD.0 R0,R5,R0 AND.0 [LSR R0,2],13107,R0 \ AND.1 R0,13107,R5 ADD.0 R0,R5,R0 \ COPY.1 1807,R2 AND.0 [LSR R0,4],R2,R0 \ AND.1 R0,3855,R5 ADD.0 R0,R5,R0 JR (R12) \ AND.0 R0, 255, R5 =-> ADD.0 [LSR R0,8],R5,R0 _picoMark_FUNCTION_END= // picoChip Function Epilogue : ___popcounthi2 //============================================================================ // All DWARF information between this marker, and the END OF DWARF // marker should be included in the source file. Search for // FUNCTION_STACK_SIZE_GOES_HERE and FUNCTION NAME GOES HERE, and // provide the relevent information. Add markers called // _picoMark_FUNCTION_BEGIN and _picoMark_FUNCTION_END around the // function in question. //============================================================================ //============================================================================ // Frame information. //============================================================================ .section .debug_frame _picoMark_DebugFrame= // Common CIE header. .unalignedInitLong _picoMark_CieEnd-_picoMark_CieBegin _picoMark_CieBegin= .unalignedInitLong 0xffffffff .initByte 0x1 // CIE Version .ascii 16#0# // CIE Augmentation .uleb128 0x1 // CIE Code Alignment Factor .sleb128 2 // CIE Data Alignment Factor .initByte 0xc // CIE RA Column .initByte 0xc // DW_CFA_def_cfa .uleb128 0xd .uleb128 0x0 .align 2 _picoMark_CieEnd= // FDE _picoMark_LSFDE0I900821033007563= .unalignedInitLong _picoMark_FdeEnd-_picoMark_FdeBegin _picoMark_FdeBegin= .unalignedInitLong _picoMark_DebugFrame // FDE CIE offset .unalignedInitWord _picoMark_FUNCTION_BEGIN // FDE initial location .unalignedInitWord _picoMark_FUNCTION_END-_picoMark_FUNCTION_BEGIN .initByte 0xe // DW_CFA_def_cfa_offset .uleb128 0x0 // <-- FUNCTION_STACK_SIZE_GOES_HERE .initByte 0x4 // DW_CFA_advance_loc4 .unalignedInitLong _picoMark_FUNCTION_END-_picoMark_FUNCTION_BEGIN .initByte 0xe // DW_CFA_def_cfa_offset .uleb128 0x0 .align 2 _picoMark_FdeEnd= //============================================================================ // Abbrevation information. //============================================================================ .section .debug_abbrev _picoMark_ABBREVIATIONS= .section .debug_abbrev .uleb128 0x1 // (abbrev code) .uleb128 0x11 // (TAG: DW_TAG_compile_unit) .initByte 0x1 // DW_children_yes .uleb128 0x10 // (DW_AT_stmt_list) .uleb128 0x6 // (DW_FORM_data4) .uleb128 0x12 // (DW_AT_high_pc) .uleb128 0x1 // (DW_FORM_addr) .uleb128 0x11 // (DW_AT_low_pc) .uleb128 0x1 // (DW_FORM_addr) .uleb128 0x25 // (DW_AT_producer) .uleb128 0x8 // (DW_FORM_string) .uleb128 0x13 // (DW_AT_language) .uleb128 0x5 // (DW_FORM_data2) .uleb128 0x3 // (DW_AT_name) .uleb128 0x8 // (DW_FORM_string) .initByte 0x0 .initByte 0x0 .uleb128 0x2 ;# (abbrev code) .uleb128 0x2e ;# (TAG: DW_TAG_subprogram) .initByte 0x0 ;# DW_children_no .uleb128 0x3 ;# (DW_AT_name) .uleb128 0x8 ;# (DW_FORM_string) .uleb128 0x11 ;# (DW_AT_low_pc) .uleb128 0x1 ;# (DW_FORM_addr) .uleb128 0x12 ;# (DW_AT_high_pc) .uleb128 0x1 ;# (DW_FORM_addr) .initByte 0x0 .initByte 0x0 .initByte 0x0 //============================================================================ // Line information. DwarfLib requires this to be present, but it can // be empty. //============================================================================ .section .debug_line _picoMark_LINES= //============================================================================ // Debug Information //============================================================================ .section .debug_info //Fixed header. .unalignedInitLong _picoMark_DEBUG_INFO_END-_picoMark_DEBUG_INFO_BEGIN _picoMark_DEBUG_INFO_BEGIN= .unalignedInitWord 0x2 .unalignedInitLong _picoMark_ABBREVIATIONS .initByte 0x2 // Compile unit information. .uleb128 0x1 // (DIE 0xb) DW_TAG_compile_unit) .unalignedInitLong _picoMark_LINES .unalignedInitWord _picoMark_FUNCTION_END .unalignedInitWord _picoMark_FUNCTION_BEGIN // Producer is `picoChip' .ascii 16#70# 16#69# 16#63# 16#6f# 16#43# 16#68# 16#69# 16#70# 16#00# .unalignedInitWord 0xcafe // ASM language .ascii 16#0# // Name. DwarfLib expects this to be present. .uleb128 0x2 ;# (DIE DW_TAG_subprogram) // FUNCTION NAME GOES HERE. Use `echo name | od -t x1' to get the hex. Each hex // digit is specified using the format 16#XX# .ascii 16#5f# 16#5f# 16#70# 16#6f# 16#70# 16#63# 16#6f# 16#75# 16#6e# 16#74# 16#68# 16#69# 16#32# 16#0# // Function name `__popcounthi2' .unalignedInitWord _picoMark_FUNCTION_BEGIN // DW_AT_low_pc .unalignedInitWord _picoMark_FUNCTION_END // DW_AT_high_pc .initByte 0x0 // end of compile unit children. _picoMark_DEBUG_INFO_END= //============================================================================ // END OF DWARF //============================================================================ .section .endFile // End of picoChip ASM file