From cdbccad21eccf012b1df1df33d5f0e11ce539fcf Mon Sep 17 00:00:00 2001
From: Irmen de Jong <irmen@razorvine.net>
Date: Tue, 23 May 2023 20:10:01 +0200
Subject: [PATCH] optimized gfx2 plot and horizontal_line a bit more

---
 compiler/res/prog8lib/cx16/gfx2.p8 | 46 ++++++++++++++++++++----------
 examples/cx16/testgfx2.p8          |  4 ++-
 2 files changed, 34 insertions(+), 16 deletions(-)

diff --git a/compiler/res/prog8lib/cx16/gfx2.p8 b/compiler/res/prog8lib/cx16/gfx2.p8
index f87c280aa..d98f614cd 100644
--- a/compiler/res/prog8lib/cx16/gfx2.p8
+++ b/compiler/res/prog8lib/cx16/gfx2.p8
@@ -155,6 +155,9 @@ gfx2 {
     }
 
     sub horizontal_line(uword x, uword y, uword length, ubyte color) {
+        ubyte[9] masked_ends   = [ 0, %10000000, %11000000, %11100000, %11110000, %11111000, %11111100, %11111110, %11111111]
+        ubyte[9] masked_starts = [ 0, %00000001, %00000011, %00000111, %00001111, %00011111, %00111111, %01111111, %11111111]
+
         if length==0
             return
         when active_mode {
@@ -163,12 +166,13 @@ gfx2 {
                 ubyte separate_pixels = (8-lsb(x)) & 7
                 if separate_pixels as uword > length
                     separate_pixels = lsb(length)
-                repeat separate_pixels {
-                    ; TODO optimize this by writing a masked byte in 1 go
-                    plot(x, y, color)
-                    x++
+                if separate_pixels {
+                    position(x,y)
+                    cx16.VERA_ADDR_H &= %00000111   ; vera auto-increment off
+                    cx16.VERA_DATA0 = cx16.VERA_DATA0 | masked_starts[separate_pixels]
+                    length -= separate_pixels
+                    x += separate_pixels
                 }
-                length -= separate_pixels
                 if length {
                     position(x, y)
                     separate_pixels = lsb(length) & 7
@@ -205,11 +209,9 @@ _loop                   lda  length
                         bra  _loop
 _done
                     }}
-                    repeat separate_pixels {
-                        ; TODO optimize this by writing a masked byte in 1 go
-                        plot(x, y, color)
-                        x++
-                    }
+
+                    cx16.VERA_ADDR_H &= %00000111   ; vera auto-increment off
+                    cx16.VERA_DATA0 = cx16.VERA_DATA0 | masked_ends[separate_pixels]
                 }
                 cx16.VERA_ADDR_H &= %00000111   ; vera auto-increment off again
             }
@@ -366,8 +368,14 @@ _done
                     return
                 position2(x,y,true)
                 set_both_strides(13)    ; 160 increment = 1 line in 640 px 4c mode
-                color &= 3
-                color <<= gfx2.plot.shift4c[lsb(x) & 3]         ; TODO with lookup table
+                ;; color &= 3
+                ;; color <<= gfx2.plot.shift4c[lsb(x) & 3]
+                cx16.r2L = lsb(x) & 3
+                when color & 3 {
+                    1 -> color = gfx2.plot.shiftedleft_4c_1[cx16.r2L]
+                    2 -> color = gfx2.plot.shiftedleft_4c_2[cx16.r2L]
+                    3 -> color = gfx2.plot.shiftedleft_4c_3[cx16.r2L]
+                }
                 ubyte @shared mask = gfx2.plot.mask4c[lsb(x) & 3]
                 repeat lheight {
                     %asm {{
@@ -551,6 +559,9 @@ _done
         ubyte[8] @shared bits = [128, 64, 32, 16, 8, 4, 2, 1]
         ubyte[4] @shared mask4c = [%00111111, %11001111, %11110011, %11111100]
         ubyte[4] @shared shift4c = [6,4,2,0]
+        ubyte[4] shiftedleft_4c_1 = [1<<6, 1<<4, 1<<2, 1<<0]
+        ubyte[4] shiftedleft_4c_2 = [2<<6, 2<<4, 2<<2, 2<<0]
+        ubyte[4] shiftedleft_4c_3 = [3<<6, 3<<4, 3<<2, 3<<0]
 
         when active_mode {
             1 -> {
@@ -643,8 +654,13 @@ _done
                 ; TODO also mostly usable for lores 4c?
                 void addr_mul_24_for_highres_4c(y, x)      ; 24 bits result is in r0 and r1L (highest byte)
                 cx16.r2L = lsb(x) & 3       ; xbits
-                color &= 3
-                color <<= shift4c[cx16.r2L]     ; TODO with lookup table
+                ; color &= 3
+                ; color <<= shift4c[cx16.r2L]
+                when color & 3 {
+                    1 -> color = shiftedleft_4c_1[cx16.r2L]
+                    2 -> color = shiftedleft_4c_2[cx16.r2L]
+                    3 -> color = shiftedleft_4c_3[cx16.r2L]
+                }
                 %asm {{
                     stz  cx16.VERA_CTRL
                     lda  cx16.r1L
@@ -743,7 +759,7 @@ _done
                     sta  cx16.r0L
                 }}
                 cx16.r1L = lsb(x) & 3
-                cx16.r0L >>= gfx2.plot.shift4c[cx16.r1L]        ; TODO with lookup table
+                cx16.r0L >>= gfx2.plot.shift4c[cx16.r1L]
                 return cx16.r0L & 3
             }
             else -> return 0
diff --git a/examples/cx16/testgfx2.p8 b/examples/cx16/testgfx2.p8
index 01a35c9aa..3f72fc70f 100644
--- a/examples/cx16/testgfx2.p8
+++ b/examples/cx16/testgfx2.p8
@@ -31,7 +31,9 @@ main {
         sys.wait(60)
         gfx2.fill(100,100,2)
         ;gfx2.monochrome_stipple(false)
-        gfx2.fill(182,140,4)
+        gfx2.fill(182,140,3)
+        gfx2.fill(182,40,1)
+
     }
 
     sub demo1() {