accurate_sprites performance improvement, PSP is untested

[picodrive.git] / Pico / Draw_amips.s

diff --git a/Pico/Draw_amips.s b/Pico/Draw_amips.s
index c05391c..29aa7db 100644 (file)
--- a/Pico/Draw_amips.s
+++ b/Pico/Draw_amips.s
@@ -45,6 +45,43 @@ amips_clut_loop:
     nop
 
 
+.global amips_clut_6bit
+
+amips_clut_6bit:
+    srl     $a3, 2
+    li      $t4, 0
+    li      $t5, 0
+    li      $t6, 0
+    li      $t7, 0
+amips_clut_loop6:
+    lbu     $t0, 0($a1)           # tried lw here, no improvement noticed
+    lbu     $t1, 1($a1)
+    lbu     $t2, 2($a1)
+    lbu     $t3, 3($a1)
+    ins     $t4, $t0, 1, 6
+    ins     $t5, $t1, 1, 6
+    ins     $t6, $t2, 1, 6
+    ins     $t7, $t3, 1, 6
+    addu    $t0, $t4, $a2
+    addu    $t1, $t5, $a2
+    addu    $t2, $t6, $a2
+    addu    $t3, $t7, $a2
+    lhu     $t0, 0($t0)
+    lhu     $t1, 0($t1)
+    lhu     $t2, 0($t2)
+    lhu     $t3, 0($t3)
+    ins     $t0, $t1, 16, 16      # ins rt, rs, pos, size - Insert size bits starting
+    ins     $t2, $t3, 16, 16      #  from the LSB of rs into rt starting at position pos
+    sw      $t0, 0($a0)
+    sw      $t2, 4($a0)
+    addiu   $a0, 8
+    addiu   $a3, -1
+    bnez    $a3, amips_clut_loop6
+    addiu   $a1, 4
+    jr      $ra
+    nop
+
+
 # $a0 - pd, $a1 - tile word, $a2 - pal
 # ext rt, rs, pos, size  // Extract size bits from position pos in rs and store in rt