{ Copyright (c) 2015 by the Free Pascal Development team Generates Motorola 68k inline nodes This program 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 2 of the License, or (at your option) any later version. This program 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. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. **************************************************************************** } unit n68kinl; {$i fpcdefs.inc} interface uses node,ninl,ncginl,cpubase; type t68kinlinenode = class(tcgInlineNode) function first_abs_real: tnode; override; function first_sqr_real: tnode; override; function first_sqrt_real: tnode; override; {function first_arctan_real: tnode; override; function first_ln_real: tnode; override;} function first_cos_real: tnode; override; function first_sin_real: tnode; override; procedure second_abs_real; override; procedure second_sqr_real; override; procedure second_sqrt_real; override; {procedure second_arctan_real; override; procedure second_ln_real; override;} procedure second_cos_real; override; procedure second_sin_real; override; {procedure second_prefetch; override; procedure second_abs_long; override;} private procedure second_do_operation(op: TAsmOp); end; implementation uses globtype,verbose,globals,cutils, cpuinfo,defutil,symdef,aasmdata,aasmcpu,aasmtai, cgbase,cgutils,pass_1,pass_2, ncgutil,cgobj,cgcpu,hlcgobj; {***************************************************************************** t68kinlinenode *****************************************************************************} function t68kinlinenode.first_abs_real : tnode; begin if (cs_fp_emulation in current_settings.moduleswitches) then result:=inherited first_abs_real else begin case current_settings.fputype of fpu_68881: expectloc:=LOC_FPUREGISTER; else internalerror(2015022206); end; first_abs_real:=nil; end; end; function t68kinlinenode.first_sqr_real : tnode; begin if (cs_fp_emulation in current_settings.moduleswitches) then result:=inherited first_sqr_real else begin case current_settings.fputype of fpu_68881: expectloc:=LOC_FPUREGISTER; else internalerror(2015022201); end; first_sqr_real:=nil; end; end; function t68kinlinenode.first_sqrt_real : tnode; begin if (cs_fp_emulation in current_settings.moduleswitches) then result:=inherited first_sqrt_real else begin case current_settings.fputype of fpu_68881: expectloc:=LOC_FPUREGISTER; else internalerror(2015022203); end; first_sqrt_real:=nil; end; end; function t68kinlinenode.first_sin_real : tnode; begin if (cs_fp_emulation in current_settings.moduleswitches) then result:=inherited first_sin_real else begin case current_settings.fputype of fpu_68881: expectloc:=LOC_FPUREGISTER; else internalerror(2015022203); end; first_sin_real:=nil; end; end; function t68kinlinenode.first_cos_real : tnode; begin if (cs_fp_emulation in current_settings.moduleswitches) then result:=inherited first_cos_real else begin case current_settings.fputype of fpu_68881: expectloc:=LOC_FPUREGISTER; else internalerror(2015022203); end; first_cos_real:=nil; end; end; procedure t68kinlinenode.second_abs_real; begin //current_asmdata.CurrAsmList.concat(tai_comment.create(strpnew('second_abs_real called!'))); second_do_operation(A_FABS); end; procedure t68kinlinenode.second_sqr_real; begin secondpass(left); case current_settings.fputype of fpu_68881: begin //current_asmdata.CurrAsmList.concat(tai_comment.create(strpnew('second_sqr_real called!'))); hlcg.location_force_fpureg(current_asmdata.CurrAsmList,left.location,left.resultdef,true); location_copy(location,left.location); if left.location.loc=LOC_CFPUREGISTER then begin //current_asmdata.CurrAsmList.concat(tai_comment.create(strpnew('second_srq_real called!: left was cfpuregister!'))); location.register:=cg.getfpuregister(current_asmdata.CurrAsmList,location.size); location.loc := LOC_FPUREGISTER; cg.a_loadfpu_reg_reg(current_asmdata.CurrAsmlist,OS_NO,OS_NO,left.location.register,location.register); end; current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg(A_FMUL,S_FX,left.location.register,location.register)); end; else internalerror(2015022202); end; end; procedure t68kinlinenode.second_sqrt_real; begin //current_asmdata.CurrAsmList.concat(tai_comment.create(strpnew('second_sqrt_real called!'))); second_do_operation(A_FSQRT); end; procedure t68kinlinenode.second_sin_real; begin //current_asmdata.CurrAsmList.concat(tai_comment.create(strpnew('second_sqrt_real called!'))); second_do_operation(A_FSIN); end; procedure t68kinlinenode.second_cos_real; begin //current_asmdata.CurrAsmList.concat(tai_comment.create(strpnew('second_sqrt_real called!'))); second_do_operation(A_FCOS); end; procedure t68kinlinenode.second_do_operation(op: TAsmOp); var href: TReference; begin secondpass(left); case current_settings.fputype of fpu_68881: begin location_reset(location,LOC_FPUREGISTER,left.location.size); case left.location.loc of LOC_FPUREGISTER: begin location.register:=left.location.register; current_asmdata.CurrAsmList.concat(taicpu.op_reg(op,S_FX,location.register)) end; LOC_CFPUREGISTER: begin location.register:=cg.getfpuregister(current_asmdata.CurrAsmList,location.size); current_asmdata.CurrAsmList.concat(taicpu.op_reg_reg(op,S_FX,left.location.register,location.register)); end; LOC_REFERENCE,LOC_CREFERENCE: begin location.register:=cg.getfpuregister(current_asmdata.CurrAsmList,location.size); href:=left.location.reference; tcg68k(cg).fixref(current_asmdata.CurrAsmList,href); current_asmdata.CurrAsmList.concat(taicpu.op_ref_reg(op,tcgsize2opsize[left.location.size],href,location.register)); end; else internalerror(2015022205); end; end; else internalerror(2015022204); end; end; { ideas for second_abs_long (KB) } { This is probably faster on 68000 than the generic implementation, because shifting is slow on the original 68000, maybe also on the 68020? Also needs to be tested on 040/060. This can also work on a CF. input - d0, output - d2 move.l d0,d2 btst #31,d2 sne d1 extb.l d1 (or ext.w + ext.l on 68000) eor.l d1,d2 sub.l d1,d2 } { Solution using bitfield extraction, we don't support the necessary asm construct for this yet, probably this is the fastest on 020, slower on 040/060 than the one above, doesn't work on '000 or CF. input - d0, output - d2 move.l d0,d2 bfexts d0[0:1],d1 eor.l d1,d2 sub.l d1,d2 } begin cinlinenode:=t68kinlinenode; end.