# 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) # # Cpu only solution VERSION 2 # Not optimized - directly ported from 040-source... (phx) # bitplanes are assumed contiguous! # analyse instruction offsets to check performance #void c2p_6_ppc (__reg("r3") UBYTE *chunky_data, # __reg("r4") PLANEPTR raster, # __reg("r5") UBYTE *compare_buffer, # __reg("r8") UBYTE *xlate, # __reg("r6") ULONG plsiz, # __reg("r7") BOOL palette_changed) #void c2p_8_ppc (__reg("r3") UBYTE *chunky_data, # __reg("r4") PLANEPTR raster, # __reg("r5") UBYTE *compare_buffer, # __reg("r6") ULONG plsiz) # r3 -> width*height chunky pixels # r4 -> contiguous bitplanes # r5 -> compare buffer # r6 = width*height/8 (width*height must be a multiple of 32) # GNU-make has some problems with quotes (e.g. in "depth=8"), so this # is a workaround (phx) .ifdef depth8 .set depth,8 .endif .ifdef depth6 .set depth,6 .endif .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 # translate 4 8-bit pixels to 6-bit EHB .macro xlate # offs1,dest2,mem3,xlate4,tmp5 lbz \5,\1(\3) lbzx \5,\4,\5 rlwimi \2,\5,24,0,7 lbz \5,\1+8(\3) lbzx \5,\4,\5 rlwimi \2,\5,16,8,15 lbz \5,\1+16(\3) lbzx \5,\4,\5 rlwimi \2,\5,8,16,23 lbz \5,\1+24(\3) lbzx \5,\4,\5 rlwimi \2,\5,0,24,31 .endm .text .ifeq depth-8 .global c2p_8_ppc c2p_8_ppc: .else .ifeq depth-6 .global c2p_6_ppc c2p_6_ppc: .else .fail "Unsupported depth! Try 6 or 8." .endif .endif stwu r1,-80(r1) stmw r14,4(r1) # save all non-volatile registers .ifle depth-6 mr r31,r7 # video_palette_changed mr r29,r8 # xlate .endif # r3 = chunky buffer # r4 = output area # r5 = compare buffer # r6 = plsiz mr r12,r6 # r12 = plsiz slwi r6,r6,3 add r30,r3,r6 # r30 = end of chunky data first_loop: .ifle depth-6 cmpwi r31,0 # palette_changed? bne first_case .endif 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 cmplw r3,r30 blt first_loop b exit # exit if no changes found first_case: .ifgt depth-6 # depth 8 code --- no need to xlate pixels # 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 .else # depth 6 code, xlate from 8-bit to 6-bit EHB lwz r6,0(r3) # copy to compare buffer stw r6,0(r5) lwz r6,4(r3) stw r6,4(r5) lwz r6,8(r3) stw r6,8(r5) lwz r6,12(r3) stw r6,12(r5) lwz r6,16(r3) stw r6,16(r5) lwz r6,20(r3) stw r6,20(r5) lwz r6,24(r3) stw r6,24(r5) lwz r6,28(r3) stw r6,28(r5) xlate 0,r9,r3,r29,r7 # does 8-bit to EHB colour translate xlate 1,r8,r3,r29,r7 # 4 pixels at a time xlate 4,r11,r3,r29,r7 xlate 5,r10,r3,r29,r7 merge r9,r11,r6,r7,0x0f0f0f0f,4 merge r8,r10,r6,r7,0x0f0f0f0f,4 xlate 2,r14,r3,r29,r7 xlate 3,r15,r3,r29,r7 xlate 6,r16,r3,r29,r7 xlate 7,r17,r3,r29,r7 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 r14,r15,r6,r7,0x55555555,1 mr r22,r15 # plane4 mr r23,r14 # plane5 .endif addi r3,r3,32 addi r5,r5,32 mr r26,r4 # save output address addi r4,r4,4 # skip 32 pixels on output cmplw r3,r30 bge final_case main_loop: .ifle depth-6 cmpwi r31,0 # palette_changed? bne main_case .endif 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 cmplw r3,r30 blt main_loop b final_case # exit if no more changes found main_case: .ifgt depth-6 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 .else # depth 6 code, xlate from 8-bit to 6-bit EHB lwz r6,0(r3) stw r6,0(r5) lwz r6,4(r3) stw r6,4(r5) lwz r6,8(r3) stw r6,8(r5) lwz r6,12(r3) stw r6,12(r5) lwz r6,16(r3) stw r6,16(r5) lwz r6,20(r3) stw r6,20(r5) lwz r6,24(r3) stw r6,24(r5) lwz r6,28(r3) stw r6,28(r5) xlate 0,r9,r3,r29,r7 # does 8-bit to EHB colour translate xlate 1,r8,r3,r29,r7 # 4 pixels at a time xlate 4,r11,r3,r29,r7 xlate 5,r10,r3,r29,r7 sub r26,r26,r12 stwux r18,r26,r12 # store plane0 (r26 += plsiz) merge r9,r11,r6,r7,0x0f0f0f0f,4 merge r8,r10,r6,r7,0x0f0f0f0f,4 xlate 2,r14,r3,r29,r7 xlate 3,r15,r3,r29,r7 xlate 6,r16,r3,r29,r7 xlate 7,r17,r3,r29,r7 stwux r19,r26,r12 # store plane1 (r26 += plsiz) merge r14,r16,r6,r7,0x0f0f0f0f,4 merge r15,r17,r6,r7,0x0f0f0f0f,4 stwux r20,r26,r12 # store plane2 (r26 += plsiz) merge r11,r16,r6,r7,0x33333333,2 merge r10,r17,r6,r7,0x33333333,2 stwux r21,r26,r12 # store plane3 (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 r22,r26,r12 # store plane4 (r26 += plsiz) merge r9,r14,r6,r7,0x33333333,2 stwux r23,r26,r12 # store plane5 (r26 += plsiz) merge r8,r15,r6,r7,0x33333333,2 merge r14,r15,r6,r7,0x55555555,1 mr r22,r15 # plane4 mr r23,r14 # plane5 .endif addi r3,r3,32 addi r5,r5,32 mr r26,r4 # save output address addi r4,r4,4 # skip 32 pixels on output 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) .ifgt depth-6 stwux r24,r26,r12 # store plane6 (r26 += plsiz) stwux r25,r26,r12 # store plane7 (r26 += plsiz) .endif exit: lmw r14,4(r1) # restore non-volatile registers addi r1,r1,80 blr