# Chunky2Planar algorithm, originally by James McCoull # Modified by Peter McGavin for variable size and depth # and "compare buffer" (hope I didn't slow it down too much) # # Ported by Frank Wille (phx) to PowerPC (04-Jan-98). # Using optimized merge-macro by Tim Boescke (phx,26-Jan-98) # AGA-bug and EHB-palette-change-bug fixed (phx,31-Jan-98) # # Modified for interleaved and system friendly by Álmos Rajnai (04-Jul-99) # # Included to the Flamingo project on 13-Jun-99 # Now it is being set to 8 bitplanes (for Flamingo) # # Cpu only solution VERSION 2 # Not optimized - directly ported from 040-source... (phx) # bitplanes are assumed interleaved # analyse instruction offsets to check performance #void c2p_8_ppc (__reg("r3") UBYTE *chunky_data, # __reg("r4") PLANEPTR raster, # __reg("r5") UBYTE *compare_buffer, # __reg("r22") ULONG bytesperrow, # __reg("r23") ULONG rowlength # __reg("r24") ULONG chunkysize) # r3 -> width*height chunky pixels # r4 -> contiguous bitplanes # r5 -> compare buffer # r22 = bytes per row # r23 = rowlength in pixels .macro merge # in1,in2,tmp3,tmp4,mask,shift # \1 = abqr # \2 = ijyz srwi \3,\2,\6 lis \4,(\5&0xffff0000)>>16 ori \4,\4,\5&0xffff xor \3,\3,\1 and \3,\3,\4 xor \1,\1,\3 slwi \3,\3,\6 xor \2,\2,\3 .endm .text .global c2p_8_ppc c2p_8_ppc: stwu r1,-80(r1) stmw r14,4(r1) # save all non-volatile registers # r3 = chunky buffer # r4 = output area # r5 = compare buffer # r22 = bytesperrow # r23 = rowlength # r24 = chunkysize mr r12,r22 # r12 = length of a plane-row srwi r31,r23,3 # r31 = a rowlength in bytes slwi r27,r22,3 # r27 = skip at the end of a row sub r27,r27,r31 # minus a row srwi r28,r23,5 # 32 pixel a cycle # r28 = rowlength in cycles add r30,r3,r24 # calc end of screen address mr r29,r28 first_loop: lwz r6,0(r3) lwz r7,0(r5) cmpw r6,r7 bne first_case lwz r6,4(r3) lwz r7,4(r5) cmpw r6,r7 bne first_case lwz r6,8(r3) lwz r7,8(r5) cmpw r6,r7 bne first_case lwz r6,12(r3) lwz r7,12(r5) cmpw r6,r7 bne first_case lwz r6,16(r3) lwz r7,16(r5) cmpw r6,r7 bne first_case lwz r6,20(r3) lwz r7,20(r5) cmpw r6,r7 bne first_case lwz r6,24(r3) lwz r7,24(r5) cmpw r6,r7 bne first_case lwz r6,28(r3) lwz r7,28(r5) cmpw r6,r7 bne first_case addi r4,r4,4 # skip 32 pixels on output addi r3,r3,32 addi r5,r5,32 subi r29,r29,1 cmpwi r29,0 bne+ .next add r4,r4,r27 # skip extra bytes at rowend mr r29,r28 .next: cmplw r3,r30 blt first_loop b exit # exit if no changes found first_case: # hint from M68k: d0-d3 -> r8-r11, d4-d7 -> r14-r17 lwz r9,0(r3) lwz r11,4(r3) lwz r8,8(r3) lwz r10,12(r3) lwz r14,2(r3) lwz r15,10(r3) lwz r16,6(r3) lwz r17,14(r3) stw r9,0(r5) stw r11,4(r5) stw r8,8(r5) stw r10,12(r5) lwz r6,16(r3) rlwimi r9,r6,16,16,31 rlwimi r14,r6,0,16,31 lwz r6,20(r3) rlwimi r11,r6,16,16,31 rlwimi r16,r6,0,16,31 lwz r6,24(r3) rlwimi r8,r6,16,16,31 rlwimi r15,r6,0,16,31 lwz r6,28(r3) rlwimi r10,r6,16,16,31 rlwimi r17,r6,0,16,31 sth r9,16(r5) sth r8,24(r5) sth r11,20(r5) sth r10,28(r5) sth r14,18(r5) sth r15,26(r5) sth r16,22(r5) sth r17,30(r5) merge r9,r8,r6,r7,0x00ff00ff,8 merge r11,r10,r6,r7,0x00ff00ff,8 merge r9,r11,r6,r7,0x0f0f0f0f,4 merge r8,r10,r6,r7,0x0f0f0f0f,4 merge r14,r15,r6,r7,0x00ff00ff,8 merge r16,r17,r6,r7,0x00ff00ff,8 merge r14,r16,r6,r7,0x0f0f0f0f,4 merge r15,r17,r6,r7,0x0f0f0f0f,4 merge r11,r16,r6,r7,0x33333333,2 merge r10,r17,r6,r7,0x33333333,2 merge r11,r10,r6,r7,0x55555555,1 merge r16,r17,r6,r7,0x55555555,1 mr r18,r17 # plane0 mr r19,r16 # plane1 mr r20,r10 # plane2 mr r21,r11 # plane3 merge r9,r14,r6,r7,0x33333333,2 merge r8,r15,r6,r7,0x33333333,2 merge r9,r8,r6,r7,0x55555555,1 merge r14,r15,r6,r7,0x55555555,1 mr r22,r15 # plane4 mr r23,r14 # plane5 mr r24,r8 # plane6 mr r25,r9 # plane7 addi r3,r3,32 addi r5,r5,32 mr r26,r4 # save output address addi r4,r4,4 # skip 32 pixels on output subi r29,r29,1 cmpwi r29,0 bne+ .next2 add r4,r4,r27 # skip extra bytes at rowend mr r29,r28 .next2: cmplw r3,r30 bge final_case main_loop: lwz r6,0(r3) # compare next 32 pixels lwz r7,0(r5) cmpw r6,r7 bne main_case lwz r6,4(r3) lwz r7,4(r5) cmpw r6,r7 bne main_case lwz r6,8(r3) lwz r7,8(r5) cmpw r6,r7 bne main_case lwz r6,12(r3) lwz r7,12(r5) cmpw r6,r7 bne main_case lwz r6,16(r3) lwz r7,16(r5) cmpw r6,r7 bne main_case lwz r6,20(r3) lwz r7,20(r5) cmpw r6,r7 bne main_case lwz r6,24(r3) lwz r7,24(r5) cmpw r6,r7 bne main_case lwz r6,28(r3) lwz r7,28(r5) cmpw r6,r7 bne main_case addi r4,r4,4 # skip 32 pixels on output addi r3,r3,32 addi r5,r5,32 subi r29,r29,1 cmpwi r29,0 bne+ .next3 add r4,r4,r27 # skip extra bytes at rowend mr r29,r28 .next3: cmplw r3,r30 blt main_loop b final_case # exit if no more changes found main_case: lwz r9,0(r3) lwz r11,4(r3) lwz r8,8(r3) lwz r10,12(r3) lwz r14,2(r3) lwz r15,10(r3) lwz r16,6(r3) lwz r17,14(r3) stw r9,0(r5) stw r11,4(r5) stw r8,8(r5) stw r10,12(r5) lwz r6,16(r3) rlwimi r9,r6,16,16,31 rlwimi r14,r6,0,16,31 lwz r6,20(r3) rlwimi r11,r6,16,16,31 rlwimi r16,r6,0,16,31 lwz r6,24(r3) rlwimi r8,r6,16,16,31 rlwimi r15,r6,0,16,31 lwz r6,28(r3) rlwimi r10,r6,16,16,31 rlwimi r17,r6,0,16,31 sth r9,16(r5) sth r8,24(r5) sth r11,20(r5) sth r10,28(r5) sth r14,18(r5) sth r15,26(r5) sth r16,22(r5) sth r17,30(r5) sub r26,r26,r12 stwux r18,r26,r12 # store plane0 (r26 += plsiz) merge r9,r8,r6,r7,0x00ff00ff,8 merge r11,r10,r6,r7,0x00ff00ff,8 stwux r19,r26,r12 # store plane1 (r26 += plsiz) merge r9,r11,r6,r7,0x0f0f0f0f,4 merge r8,r10,r6,r7,0x0f0f0f0f,4 stwux r20,r26,r12 # store plane2 (r26 += plsiz) merge r14,r15,r6,r7,0x00ff00ff,8 merge r16,r17,r6,r7,0x00ff00ff,8 stwux r21,r26,r12 # store plane3 (r26 += plsiz) merge r14,r16,r6,r7,0x0f0f0f0f,4 merge r15,r17,r6,r7,0x0f0f0f0f,4 stwux r22,r26,r12 # store plane4 (r26 += plsiz) merge r11,r16,r6,r7,0x33333333,2 merge r10,r17,r6,r7,0x33333333,2 stwux r23,r26,r12 # store plane5 (r26 += plsiz) merge r11,r10,r6,r7,0x55555555,1 merge r16,r17,r6,r7,0x55555555,1 mr r18,r17 # plane0 mr r19,r16 # plane1 mr r20,r10 # plane2 mr r21,r11 # plane3 stwux r24,r26,r12 # store plane6 (r26 += plsiz) merge r9,r14,r6,r7,0x33333333,2 merge r8,r15,r6,r7,0x33333333,2 stwux r25,r26,r12 # store plane7 (r26 += plsiz) merge r9,r8,r6,r7,0x55555555,1 merge r14,r15,r6,r7,0x55555555,1 mr r22,r15 # plane4 mr r23,r14 # plane5 mr r24,r8 # plane6 mr r25,r9 # plane7 addi r3,r3,32 addi r5,r5,32 mr r26,r4 # save output address addi r4,r4,4 # skip 32 pixels on output subi r29,r29,1 cmpwi r29,0 bne+ .next4 add r4,r4,r27 # skip extra bytes at rowend mr r29,r28 .next4: cmplw r3,r30 blt main_loop final_case: sub r26,r26,r12 stwux r18,r26,r12 # store plane0 (r26 += plsiz) stwux r19,r26,r12 # store plane1 (r26 += plsiz) stwux r20,r26,r12 # store plane2 (r26 += plsiz) stwux r21,r26,r12 # store plane3 (r26 += plsiz) stwux r22,r26,r12 # store plane4 (r26 += plsiz) stwux r23,r26,r12 # store plane5 (r26 += plsiz) stwux r24,r26,r12 # store plane6 (r26 += plsiz) stwux r25,r26,r12 # store plane7 (r26 += plsiz) exit: lmw r14,4(r1) # restore non-volatile registers addi r1,r1,80 blr