#368 GBA more conditional muls, matrixSetIdentity_asm optimization

src/platform/gba/asm/common_asm.inc

@@ -110,13 +110,14 @@
     ldrsh vx2, [vp2, #VERTEX_X]
     ldrsh vy1, [vp1, #VERTEX_Y]
     rsb vx2, vx2, vx0, asr #16      // reverse order for mla
-    sub vy1, vy1, vy0, asr #16
-    mul vy1, vx2, vy1
+    subs vy1, vy1, vy0, asr #16
+    mulne vy1, vx2, vy1
     ldrsh vx1, [vp1, #VERTEX_X]
     sub vx0, vx1, vx0, asr #16
     ldrsh vy2, [vp2, #VERTEX_Y]
-    sub vy0, vy2, vy0, asr #16
-    mlas vy1, vx0, vy0, vy1
+    subs vy0, vy2, vy0, asr #16
+    mulne vy0, vx0, vy0
+    adds vy1, vy0
     ble \skip
 .endm
 

src/platform/gba/asm/matrixSetIdentity.s

@@ -4,12 +4,15 @@ e0  .req r0
 e1  .req r1
 e2  .req r2
 e3  .req r3
-e4  .req r12
-m   .req lr
+// FIQ regs
+e4  .req r8
+e5  .req r9
+e6  .req r10
+m   .req r11
 
 .global matrixSetIdentity_asm
 matrixSetIdentity_asm:
-    stmfd sp!, {lr}
+    fiq_on
     ldr m, =gMatrixPtr
     ldr m, [m]
     mov e0, #0x4000
@@ -17,15 +20,16 @@ matrixSetIdentity_asm:
     mov e2, #0
     mov e3, #0
     mov e4, #0
+    mov e5, #0x4000
+    mov e6, #0
 
     // row-major
     // e0 e1 e2 e3
-    // e4 e0 e1 e2
-    // e3 e4 e0 e1
+    // e4 e5 e6 e2
+    // e3 e4 e5 e6
 
-    stmia m!, {e0, e1, e2, e3, e4}
-    stmia m!, {e0, e1, e2, e3, e4}
-    stmia m!, {e0, e1}
+    stmia m!, {e0, e1, e2, e3, e4, e5, e6}
+    stmia m!, {e2, e3, e4, e5, e6}
 
-    ldmfd sp!, {lr}
+    fiq_off
     bx lr

src/platform/gba/asm/sndPCM.s

@@ -29,8 +29,8 @@ out     .req size
 .macro pcm_sample_fetch
     ldrb out, [data, pos, lsr #SND_FIXED_SHIFT]
     add pos, inc
-    sub out, #128
-    mul out, volume
+    subs out, #128
+    mulne out, volume
 .endm
 
 .macro pcm_sample_fill