From af525a26fe2a78538d427da6c87a7258d6eae362 Mon Sep 17 00:00:00 2001
From: blondie7575 <qd@quinndunki.com>
Date: Sun, 6 Aug 2017 16:33:56 -0700
Subject: [PATCH] Optimization

---
 gscats.2mg | Bin 819264 -> 819264 bytes
 gscats.s   |   3 +-
 terrain.s  | 105 +++++++++++++++++++++++++++++++++++++++++++----------
 3 files changed, 87 insertions(+), 21 deletions(-)

diff --git a/gscats.2mg b/gscats.2mg
index acf3d7f427deb0bc36c421e285ed71a31345c4eb..08807cf519dab4f523e6e02d12d78929a523db28 100644
GIT binary patch
delta 1474
zcma))drVtZ9LLY?<8t5a?SnqYI+%{)V->r}%CfcUlp#Yf8gYwB{lh@BKg4LFF2**x
zD|NIX^qOqTKpVAW7p<U;5`r!jMIjLwSz;oZh2U$xlA20PW=uVoUkiKqhfTic^ZkC$
zIrrS2+<dGtA8Q=%(VKZRY(r$7%ZSjYh*FJ44ryj`L6Rm3Lhj@y<S<v2qxkk`G!dKJ
zb==#SpjUlk-uw#v$p9XweaZz&A{B%@<MbP~wVuBy&mWBEEb~s0@+{GwJ}SE=k00N?
zqe#tlphD5<OctrwZ}z{Zec2*a_%0Ao^S5eN2$Wotd8uyDr)zk2)ibs$zmlA%6*Pd&
zSPu>hH0>L}X4=!7T%?uKkSjfl<j~3c<*ZaC&eQ6`>;er-vWxPJ)Zzah4|*#y$=BR@
z_knxvZ8_L_U)!PfTuAdL5p}hBnM@=Qn43ifofL`vvg*lRcLwe13M(B*q7+e-y3;0<
zAti@$gh@WaWJqT8qlo+CqwX1MyyjerdAXYzecr&f9bsAzmP$_h<CH|j<C8Biv5V`o
z8HyQIbm~^T$|8Re_eNWm596!F_p2UJemdUJ^n=5<wXmgSggn8Fo*It0FJ0=vV}^7(
z=N>=>K_X-G>JgG79*->^yWA5_Bc%N7kQ|3kc}wW*B)`QQjUF3x%;wYR{I6kz-eV`&
zbL{8rBKtM_JzHS^VOh?}MH{$g?hqH|p5|WRUgtjK7Py}{f_L*x{6qXxe2Raczre5Y
zc44pZn9wJ@D|{uC1h3&iL(Gshd}CmY0pn5Qu<^W6C+-oS5J$v|qRrH4N|<I%dh=c8
z7tPaVt>u6vZu!z8S|7H)X8qB+v--K}&#H~KM{RH0E=O&x_A~b1?Lo(@j*;%=wk`&b
z6>M<RT;Z;?SNW^MHR+nzzIoVk<yU@*O2oXr`L7Ch@{|2$%{nXqE1()+1K0r$fD_;X
zxB)dpm%P+pqlTsat$=N~3k_~lLujxTPzTr!xB+k@U<aTc-~}`QZUWp4xCL-4U?;!_
zxDC(<xE<gJ>;mit+yQ6;>;VJ-&49gtefXacseNi7Qu_g+^)9GYZ;?m^b?U9FMqW9g
zgDKh9uP?h^Ne{b%a<4>g+Nz56CFqT%bfruu*m7A&aFyvKUv?b{q0&i0WlOJQsMugq
z+?W?DWj<lvD9u{@o#Z-)wrg|sr;8dnapr?ucxb=t&HB^COa7@o{}h3x!O~*su=H3Y
RRuvY7MPo4&{wbE5`4^wKPmBNn

delta 1405
zcmajfZD><h7zglkZ`1U?oSXNgc~QK!+L~^&vDU7O8*PWyw{%2=&We+Mk=mh9Kj>sQ
zl3}lAORu%P2*sk5)s|i{v=Bi`rKIf&Z58dlZ1}<`9fNMlxYG^aKGbvbq*1A0;5onl
zbIx;5NC-(q%Tm#@p+mWLj7~U_v9M4;XaX^Mpp_bWGFfe8jZ}`Z8(&u%abA)UzW9dT
z;#Bsw`b&Rc=L3WO)Eqn3PS){&I?G5#6(KRf{!~yP{>O+z+*OScZ?j@6lX$EoTPun+
zW;%?tCtk0tG0{ivm*ati#%O;8gL?l!?^O!Pj}?nlSQJJGotPH~H>Kv}8CFH@Bn%!R
z%}tgKw39F{ZkBH;4blPY467d|ZnChQxTQ=<;fl)lKh0il9$xVILZ5HhTDxugj-9(g
zb;%m~e=1VU^)&~B(^E(@NX^l|=f%Xb_h5Z{8`Ut97}aW(ntE)g&nS7+22TEx8$Qid
z{^<#<(F!J+p>?AdxYJGCm6Nx*Py@Gf$Ks;beUOouSd5HCf7g4-6N^P>(sO+bcSWW1
z^jpXS%4n=4(u0sX;E_E0y8Mf1V0gjnk3_!h@Jyw8(CouTgj)FH{CB*>pXIywUVf0j
z!O!q_`M>xVJjSc=M!X##z{jwR`*8~2$4{|CC=+%IM}<z|n(&+OQV>nGrX!~Frcu*^
z$!@MPA2wewPn(~cODy{>?Usp%<(Xx@^`P~l^;fIKw##<L_Op$)*Vw<ekJ}B7ZH~BO
z*5Pt~<?M6*<E(U@a@}#Q$p1QjIG=Uzb6<5Yy7zeco9>0`IkGO1^>uh7y%|Z4q@TQ(
zer2wK=B+uh_4eWNe%Zh~M&7ZNo4#84C2prf^#ljt1YAHq;08Q^7w`cqfR(^1s$Tgw
zE|PVuL)5p%I#vS(Kq0UOSPK*Z#lSki50n5O03QM$0qcQMAOLIt%7BePIj{++06qpP
zflq)Suo?Ii_zb8bpNs0OPOnut<clX=l|lb<^`K}!8sJC~<GGpC5;ra-$*A9=54K2o
zuOpL<ZCqw_PUe?V8_uMFEM(G<Y|5l_ZRSjBk<8g^B+F8a%C_veEwkAvJF<I@9P;BL
z{|wbsvab6Ht%T_!>SFTfr3qiMx~r7#t?Q;vm5-b$AE8Lmq;g0ZNad0;lFB2+kYY)3
K*ULwEJoySFl}Hf)

diff --git a/gscats.s b/gscats.s
index 7a9218f..2331ab5 100644
--- a/gscats.s
+++ b/gscats.s
@@ -30,6 +30,7 @@ mainBank2:
 	jsr colorFill
 
 	jsr generateTerrain
+	jsr computeScrollClipBoundaries
 
 mainGameLoop:
 	ldy mapScrollPos
@@ -107,7 +108,7 @@ kbdScanSpace:
 basePalette:
 	.word $0800,$0080,$0000,$000F,$0000,$0000,$0000,$0000,$0000,$0000,$0000,$0000,$0000,$0000,$0000,$0000
 mapScrollPos:	; 4-pixel columns distance from right terrain edge
-	.word 80
+	.word 0
 quitRequested:
 	.word $0000
 
diff --git a/terrain.s b/terrain.s
index 38fa51c..7f88811 100644
--- a/terrain.s
+++ b/terrain.s
@@ -13,6 +13,9 @@ MAXTERRAINHEIGHT = 100	; In pixels
 ;
 ; No stack operations permitted here!
 ;
+; Initial implementation: 535 cycles per row
+; Current implementation: 443 cycles per row
+;
 renderTerrain:
 	lda #199*2
 	sta <ROWINDEX
@@ -62,15 +65,14 @@ renderClippedSpanChain:
 	lda #renderClippedSpanChainRenderNext	; 2
 	sta renderSpanComplete+1		; 4
 
-	; Find right edge of screen within span chains
-	; = 27 cycles per skipped span
+	; Look up right edge clipping data from precalculated table
+	; = 15 cycles per row
 renderClippedSpanChainLoop:
-
-	lda spanChain,y		; 4
-	sec					; 2
-	sbc <RIGHTEDGE		; 3
-	bmi renderClippedSpanChainNextSpan	; 2/3
-	beq renderClippedSpanChainNextSpan	; 2/3
+	lda <MAPSCROLLPOS		; 3
+	asl						; 2
+	tax						; 2
+	ldy scrollClipIndices,x	; 4
+	lda	scrollClipIndices+2,x ; 4
 
 renderClippedSpanChainLoop2:
 	; Now render spans until left edge of screen
@@ -91,7 +93,6 @@ renderSpanComplete:
 	; unrolled span rendering blocks
 	jmp renderClippedSpanChainRenderNext	; 3
 
-
 renderClippedSpanChainRenderNext:
 	; Track remaining words until left edge
 	; = 26 cycles per span rendered
@@ -108,16 +109,6 @@ renderClippedSpanChainRenderNext:
 	lda spanChain,y		; 5
 	bra renderClippedSpanChainLoop2	; 3
 
-renderClippedSpanChainNextSpan:
-	; Track remaining distance from right edge and
-	; continue searching for visible right edge
-	eor #$ffff			; 2
-	inc					; 2
-	sta <RIGHTEDGE		; 3
-	dey					; 2
-	dey					; 2
-	bra renderClippedSpanChainLoop	; 3
-
 renderClippedSpanChainLastSpan:
 	; Render visible portion of last visible span
 	; = 26 cycles per row
@@ -151,6 +142,75 @@ spanColors:
 spanChainIndex:
 	.word 0
 
+
+;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
+; computeScrollClipBoundaries
+;
+; Generates a table of the clipped span indices for the right edge
+; at every possible scroll position
+;
+computeScrollClipBoundaries:
+	SAVE_AXY
+
+	lda #0
+	sta <MAPSCROLLPOS
+
+computeScrollClipBoundariesOuterLoop:
+	; Prepare our state
+	lda #80					; 2
+	sta <XLEFT				; 3
+	ldy #spanChainEnd-spanChain-2	; 2
+	lda <MAPSCROLLPOS				; 3
+	sta <RIGHTEDGE					; 3
+
+	; Find right edge of screen within span chains
+computeScrollClipBoundariesInnerLoop:
+
+	lda spanChain,y		; 4
+	sec					; 2
+	sbc <RIGHTEDGE		; 3
+	bmi computeScrollClipBoundariesNextSpan	; 2/3
+	beq computeScrollClipBoundariesNextSpan	; 2/3
+
+	; Y now contains the pointer to span that will be clipped at this scroll position
+	; A now contains remaining words of clipped span to render
+	; Write these values into our tuple data structure
+	pha
+	ldx <MAPSCROLLPOS
+	txa
+	asl
+	tax
+	tya
+    sta scrollClipIndices,x
+	inx
+	inx
+	pla
+	sta scrollClipIndices,x
+
+	; On to the next scroll position
+	lda <MAPSCROLLPOS
+	inc
+	inc
+	cmp #(TERRAINWIDTH-320)/4+2
+	beq computeScrollClipBoundariesDone
+	sta <MAPSCROLLPOS
+	bra computeScrollClipBoundariesOuterLoop
+
+computeScrollClipBoundariesNextSpan:
+	; Track remaining distance from right edge and
+	; continue searching for visible right edge
+	eor #$ffff			; 2
+	inc					; 2
+	sta <RIGHTEDGE		; 3
+	dey					; 2
+	dey					; 2
+	bra computeScrollClipBoundariesInnerLoop	; 3
+
+computeScrollClipBoundariesDone:
+	RESTORE_AXY
+	rts
+
+
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 ; generateTerrain
 ;
@@ -186,9 +246,14 @@ generateTerrainLoop:
 
 terrainData:
 	.repeat TERRAINWIDTH/2
-	.byte 0
+	.word 0
 	.endrepeat
 
+scrollClipIndices:
+	; Tuples: (Clipped span, Words to Render)
+	.repeat (TERRAINWIDTH-320)/4+2	; Always scroll by two words
+	.word 0
+	.endrepeat
 
 .if 0
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;