|
@@ -0,0 +1,189 @@
|
|
|
+{
|
|
|
+ $Id$
|
|
|
+ This file is part of the Free Pascal run time library.
|
|
|
+ Copyright (c) 1999 by the Free Pascal development team.
|
|
|
+
|
|
|
+ Portions Copyright (c) 2000 by Casey Duncan ([email protected])
|
|
|
+
|
|
|
+ Processor dependent implementation for the system unit for
|
|
|
+ PowerPC
|
|
|
+
|
|
|
+ See the file COPYING.FPC, included in this distribution,
|
|
|
+ for details about the copyright.
|
|
|
+
|
|
|
+ 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.
|
|
|
+
|
|
|
+ **********************************************************************}
|
|
|
+
|
|
|
+
|
|
|
+{****************************************************************************
|
|
|
+ Move / Fill
|
|
|
+****************************************************************************}
|
|
|
+
|
|
|
+{$define FPC_SYSTEM_HAS_MOVE}
|
|
|
+
|
|
|
+procedure Move(var source;var dest;count:longint);
|
|
|
+begin
|
|
|
+{ register usage:
|
|
|
+ r3 source
|
|
|
+ r4 dest
|
|
|
+ r5 count
|
|
|
+ r13 ptr to end of source
|
|
|
+ r14 ptr to end of dest
|
|
|
+ r15 counter 1
|
|
|
+ r16 counter 2
|
|
|
+ r17 addr increment
|
|
|
+ r18 ptr to current source block
|
|
|
+ r19 ptr to current dest block
|
|
|
+ r20-24 buffer
|
|
|
+ f1-4 buffer
|
|
|
+ ctr Loop counter
|
|
|
+ notes:
|
|
|
+ Move uses FPRs for increased bandwidth
|
|
|
+}
|
|
|
+ asm
|
|
|
+ { do some param checking, initialization }
|
|
|
+ cmplwi cr2,r3,0
|
|
|
+ cmplwi cr3,r4,0
|
|
|
+ cmplw cr4,r3,r4
|
|
|
+ add r13,r3,r5
|
|
|
+ add r14,r4,r5
|
|
|
+ bt cr2,.MoveEnd //end if source=nil
|
|
|
+ bt cr3,.MoveEnd //end if dest=nil
|
|
|
+ bt cr4,.MoveEnd //end if source=dest
|
|
|
+ { see if source and dest overlap }
|
|
|
+ cmplw cr2,r13,r4
|
|
|
+ cmplw cr3,r4,r3
|
|
|
+ srawi. r15,r5,$5 //r15 := count div 32
|
|
|
+ andi r16,r5,$1F //r16 := count mod 32
|
|
|
+ crand cr3,cr2,cr3
|
|
|
+ mtctr r15 //Load loop counter
|
|
|
+ bgt cr3,.MoveRL //dest overlaps source on right
|
|
|
+ li r17,$8 //Offset 8 bytes per doubleword copy
|
|
|
+ sub r18,r17,r3 //calculate the starting source
|
|
|
+ sub r19,r17,r4 // and dest ptrs
|
|
|
+ beq .MoveByByte //If count<32 skip 32 byte block copy
|
|
|
+ srawi. r15,r16,$2 //r15 := r16 div 4
|
|
|
+ andi r16,r15,$3 //r16 := r15 mod 4
|
|
|
+ cmpwi cr2,r16,0 //r16 = 0 ?
|
|
|
+ crand cr3,cr2,cr0 //r15 = 0 AND r16 = 0 ?
|
|
|
+.MoveBlockLoop: //32 Byte block copy (fully optimized)
|
|
|
+ lfdux f1,r18,r17
|
|
|
+ lfdux f2,r18,r17
|
|
|
+ lfdux f3,r18,r17
|
|
|
+ lfdux f4,r18,r17
|
|
|
+ stfdux f1,r19,r17
|
|
|
+ stfdux f2,r19,r17
|
|
|
+ stfdux f3,r19,r17
|
|
|
+ stfdux f4,r19,r17
|
|
|
+ bdnz .MoveBlockLoop
|
|
|
+
|
|
|
+ bt cr3,MoveEnd //Nothing left to do...
|
|
|
+ mtspr 1,r16 //XER := r16
|
|
|
+ beq .MoveBytes //There are fewer than 4 bytes left
|
|
|
+ mtctr r15 //load counter
|
|
|
+ andi r15,r15,$3 //r15 := r15 mod 4
|
|
|
+ srawi r17,$1 //Offset := Offset div 2
|
|
|
+.MoveWordLoop: //4 byte copy
|
|
|
+ lwzux r20,r18,r17
|
|
|
+ stwux r20,r19,r17
|
|
|
+ bdnz .WordCopyLoop
|
|
|
+
|
|
|
+ bt cr2,MoveEnd //Nothing left to do...
|
|
|
+.MoveBytes: //Copy remaining stragglers
|
|
|
+ lswx r20,r0,r18
|
|
|
+ stswx r20,r0,r19
|
|
|
+.MoveEnd:
|
|
|
+ End;
|
|
|
+End;
|
|
|
+
|
|
|
+
|
|
|
+{$define FPC_SYSTEM_HAS_FILLCHAR}
|
|
|
+
|
|
|
+Procedure FillChar(var x;count:longint;value:byte);
|
|
|
+begin
|
|
|
+ asm
|
|
|
+{ Register Usage:
|
|
|
+ r3 x
|
|
|
+ r4 count
|
|
|
+ r5 value
|
|
|
+ r13 value.value.value.value
|
|
|
+ r14 ptr to current dest char
|
|
|
+ r15 byte increment, Scratch
|
|
|
+ r16 Block count
|
|
|
+ r17 misalignment byte count
|
|
|
+}
|
|
|
+ cmpwi cr2,r4,12
|
|
|
+ mr r14,r3
|
|
|
+ andi. r17,r3,3
|
|
|
+ sub r14,r3,r17 //32 bit align
|
|
|
+ blt cr2,.FillBytes //if count<12 then fill byte by byte
|
|
|
+ sub r16,r4,r17
|
|
|
+ andi r17,r16,3
|
|
|
+ cmpwi cr2,r17,0
|
|
|
+ srwi r16,r16,2 //r16:=count div 4
|
|
|
+ subi r16,r16,2
|
|
|
+ mtctr r16 //counter:=r16
|
|
|
+ mr r13,r5 //insert
|
|
|
+ insrwi r13,r5,8,16 // value into all four bytes
|
|
|
+ insrwi r13,r13,16,0 // of r13
|
|
|
+ li r15,4
|
|
|
+ stw r13,0(r3) //fill first few bytes
|
|
|
+.FillWordLoop:
|
|
|
+ stwux r13,r14,r15
|
|
|
+ bdnz .FillWordLoop
|
|
|
+ beq cr2,FillEnd //No trailing bytes, so exit
|
|
|
+ add r14,r3,r4
|
|
|
+ stw r13,-4(r14) //fill last few bytes
|
|
|
+ b .FillEnd
|
|
|
+
|
|
|
+.FillBytes:
|
|
|
+ mtctr r4 //counter:=count
|
|
|
+ li r15,1
|
|
|
+ subi r14,r3,1
|
|
|
+.FillByteLoop:
|
|
|
+ stbux r13,r14,r15
|
|
|
+ bdnz .FillByteLoop
|
|
|
+.FillEnd:
|
|
|
+ end [r13,r14,r15,r16,r17,ctr];
|
|
|
+end;
|
|
|
+
|
|
|
+
|
|
|
+{$define FPC_SYSTEM_HAS_FILLWORD}
|
|
|
+
|
|
|
+procedure fillword(var x;count : longint;value : word);
|
|
|
+begin
|
|
|
+{ registers:
|
|
|
+ r3 x
|
|
|
+ r4 count
|
|
|
+ r5 value
|
|
|
+ r13 value.value
|
|
|
+ r14 ptr to dest word
|
|
|
+ r15 increment 1
|
|
|
+ r16 increment 2
|
|
|
+ r17 scratch
|
|
|
+ r18 scratch
|
|
|
+ f1 value.value.value.value
|
|
|
+}
|
|
|
+ asm
|
|
|
+ cmpwi cr0,r3,0
|
|
|
+ andi r17,r4,$3
|
|
|
+ srwi r18,r4,1 //r18:=count div 2
|
|
|
+ mr r13,r3
|
|
|
+ li r14,4
|
|
|
+ ble .FillWordEnd //if count<=0 Then Exit
|
|
|
+.FillWordLoop:
|
|
|
+ stwux r5,r13,r14
|
|
|
+ bdnz .FillWordLoop
|
|
|
+.FillWordEnd:
|
|
|
+ end [r13,r14,ctr]
|
|
|
+end;
|
|
|
+
|
|
|
+{
|
|
|
+ $Log$
|
|
|
+ Revision 1.1 2000-07-27 07:32:12 jonas
|
|
|
+ + initial version by Casey Duncan (not yet thoroughly debugged or complete)
|
|
|
+
|
|
|
+}
|