#368 GBA sound samples and music playback optimization

src/fixed/game.h

@@ -130,14 +130,6 @@ struct Game
         rooms[roomIndex].add(lara);
     }
 
-    int32 getAmbientTrack() // TODO
-    {
-        extern int32 gLevelID;
-        if (gLevelID == 0)
-            return -1;
-        return 5;
-    }
-
     void updateItems()
     {
         ItemObj* item = ItemObj::sFirstActive;

src/fixed/level.h

@@ -170,4 +170,12 @@ void updateLevel(int32 frames)
     }
 }
 
+int32 getAmbientTrack()
+{
+    extern int32 gLevelID;
+    if (gLevelID == 0)
+        return -1;
+    return 5;
+}
+
 #endif

src/platform/gba/packer/main.cpp

@@ -3079,6 +3079,35 @@ struct LevelPC
 
         //f.writeArray(demoData, demoDataSize);
 
+        for (int32 i = 0; i < soundOffsetsCount; i++)
+        {
+            uint8* ptr = soundData + soundOffsets[i];
+            int32 size = *(int32*)(ptr + 40);
+            uint8* src = ptr + 44;
+            uint8* dst = ptr;
+
+            while ((dst - soundData) % 4 != 0) {
+                dst++;
+            }
+            dst += 4;
+
+            for (int32 j = 0; j < size; j++)
+            {
+                dst[j] = src[j];
+            }
+
+            while ((size % 4) != 0)
+            {
+                dst[size] = dst[size - 1];
+                size++;
+            }
+
+            dst -= 4;
+            *(int32*)dst = size;
+
+            soundOffsets[i] = dst - soundData;
+        }
+
         header.soundMap = f.align4();
         f.write(soundMap, 256);
 

src/platform/gba/render.iwram.cpp

@@ -1039,19 +1039,25 @@ void renderShadow(int32 x, int32 z, int32 sx, int32 sz)
         return;
     }
 
-    int32 sx2 = sx << 1;
-    int32 sz2 = sz << 1;
-
-    MeshVertex v[8] = {
-        { x - sx,  0, z + sz2 }, // 0
-        { x + sx,  0, z + sz2 }, // 1
-        { x + sx2, 0, z + sz  }, // 2
-        { x + sx2, 0, z - sz  }, // 3
-        { x + sx,  0, z - sz2 }, // 4
-        { x - sx,  0, z - sz2 }, // 5
-        { x - sx2, 0, z - sz  }, // 6
-        { x - sx2, 0, z + sz  }  // 7
-    };
+    int16 xns1 = x - sx;
+    int16 xps1 = x + sx;
+    int16 xns2 = xns1 - sx;
+    int16 xps2 = xps1 + sx;
+
+    int16 zns1 = z - sz;
+    int16 zps1 = z + sz;
+    int16 zns2 = zns1 - sz;
+    int16 zps2 = zps1 + sz;
+
+    MeshVertex v[8];
+    v[0].x = xns1; v[0].y = 0; v[0].z = zps2;
+    v[1].x = xps1; v[1].y = 0; v[1].z = zps2;
+    v[2].x = xps2; v[2].y = 0; v[2].z = zps1;
+    v[3].x = xps2; v[3].y = 0; v[3].z = zns1;
+    v[4].x = xps1; v[4].y = 0; v[4].z = zns2;
+    v[5].x = xns1; v[5].y = 0; v[5].z = zns2;
+    v[6].x = xns2; v[6].y = 0; v[6].z = zns1;
+    v[7].x = xns2; v[7].y = 0; v[7].z = zps1;
 
     transformMesh(v, 8, 0);
     faceAddMeshQuads(gShadowQuads, 3);
@@ -1153,79 +1159,3 @@ void renderGlyph(int32 vx, int32 vy, int32 index)
 {
     //
 }
-
-extern int16 IMA_STEP[89];
-
-#define DECODE_IMA_4(n)\
-    step = IMA_STEP[idx];\
-    index = n & 7;\
-    step += index * step << 1;\
-    if (index < 4) {\
-        idx = X_MAX(idx - 1, 0);\
-    } else {\
-        idx = X_MIN(idx + ((index - 3) << 1), X_COUNT(IMA_STEP) - 1);\
-    }\
-    if (n & 8) {\
-        smp -= step >> 3;\
-    } else {\
-        smp += step >> 3;\
-    }\
-    *buffer++ = smp >> (16 - (8 + SND_VOL_SHIFT));
-
-void decodeIMA(IMA_STATE &state, const uint8* data, int32* buffer, int32 size)
-{
-    uint32 step, index;
-
-    int32 idx = state.idx;
-    int32 smp = state.smp;
-
-    for (int32 i = 0; i < size; i++)
-    {
-        uint32 n = *data++;
-        DECODE_IMA_4(n);
-        n >>= 4;
-        DECODE_IMA_4(n);
-    }
-
-    state.idx = idx;
-    state.smp = smp;
-}
-
-/* TODO OUT OF IWRAM!
-#define DECODE_IMA_4_sample(n)\
-    step = IMA_STEP[idx];\
-    index = n & 7;\
-    step += index * step << 1;\
-    if (index < 4) {\
-        idx = X_MAX(idx - 1, 0);\
-    } else {\
-        idx = X_MIN(idx + ((index - 3) << 1), X_COUNT(IMA_STEP) - 1);\
-    }\
-    if (n & 8) {\
-        smp -= step >> 3;\
-    } else {\
-        smp += step >> 3;\
-    }\
-    *buffer++ += smp * volume >> (16 - (8 + SND_VOL_SHIFT));
-
-void decodeIMA_sample(IMA_STATE &state, const uint8* data, int32* buffer, int32 size, int32 inc, int32 volume)
-{
-    uint32 step, index;
-
-    int32 idx = state.idx;
-    int32 smp = state.smp;
-
-    for (int32 i = 0; i < size; i++)
-    {
-        uint32 n = *data;
-        DECODE_IMA_4_sample(n);
-        n >>= 4;
-        DECODE_IMA_4_sample(n);
-
-        data += inc;
-    }
-
-    state.idx = idx;
-    state.smp = smp;
-}
-*/

src/platform/gba/sndIMA.s

@@ -0,0 +1,64 @@
+#include "common_asm.inc"
+
+state   .req r0
+buffer  .req r1
+data    .req r2
+size    .req r3
+smp     .req r4
+idx     .req r5
+stepLUT .req r6
+step    .req r7
+n       .req r8
+index   .req r9
+outA    .req r12
+outB    .req lr
+tmp     .req outB
+
+IMA_STEP_SIZE = 88
+
+.macro decode4 n, out
+    ldr step, [stepLUT, idx, lsl #2]
+
+    and index, \n, #7
+    mov tmp, step, lsl #1
+    mla step, index, tmp, step
+    tst \n, #8
+    subne smp, smp, step, lsr #3
+    addeq smp, smp, step, lsr #3
+
+    subs index, #3
+    suble idx, idx, #1
+    bicle idx, idx, idx, asr #31
+    addgt idx, idx, index, lsl #1
+    cmpgt idx, #IMA_STEP_SIZE
+    movgt idx, #IMA_STEP_SIZE
+
+    mov \out, smp, asr #2
+.endm
+
+.global sndIMA_asm
+sndIMA_asm:
+    stmfd sp!, {r4-r9, lr}
+
+    ldmia state, {smp, idx}
+
+    ldr stepLUT, =IMA_STEP
+
+.loop:
+    ldrb n, [data], #1
+
+    decode4 n, outA
+
+    mov n, n, lsr #4
+
+    decode4 n, outB
+
+    stmia buffer!, {outA, outB}
+
+    subs size, #1
+    bne .loop
+
+    stmia state, {smp, idx}
+
+    ldmfd sp!, {r4-r9, lr}
+    bx lr

src/platform/gba/sndPCM.s

@@ -0,0 +1,48 @@
+#include "common_asm.inc"
+
+pos     .req r0
+inc     .req r1
+size    .req r2
+volume  .req r3
+
+data    .req r4
+buffer  .req r5
+count   .req r6
+ampA    .req r7
+ampB    .req r8
+outA    .req r9
+outB    .req r12
+last    .req count
+tmp     .req outB
+
+.global sndPCM_asm
+sndPCM_asm:
+    mov tmp, sp
+    stmfd sp!, {r4-r9}
+
+    ldmia tmp, {data, buffer, count}
+
+    mla last, inc, count, pos
+    cmp last, size
+    movgt last, size
+
+.loop:
+    ldrb ampA, [data, pos, lsr #8]
+    add pos, pos, inc
+    ldrb ampB, [data, pos, lsr #8]
+    add pos, pos, inc
+    cmp pos, last
+
+    sub ampA, ampA, #128
+    sub ampB, ampB, #128
+
+    ldmia buffer, {outA, outB}
+    mla outA, volume, ampA, outA
+    mla outB, volume, ampB, outB
+    stmia buffer!, {outA, outB}
+
+    blt .loop
+
+.done:
+    ldmfd sp!, {r4-r9}
+    bx lr

src/platform/gba/sndWrite.s

@@ -0,0 +1,44 @@
+#include "common_asm.inc"
+
+buffer  .req r0
+count   .req r1
+data    .req r2
+vA      .req r3
+vB      .req r4
+vC      .req r5
+vD      .req r12
+
+SND_VOL_SHIFT = 6
+
+.macro encode amp
+    mov \amp, \amp, asr #SND_VOL_SHIFT
+    cmp \amp, #-128
+    movlt \amp, #-128
+    cmp \amp, #127
+    movgt \amp, #127
+.endm
+
+.global sndWrite_asm
+sndWrite_asm:
+    stmfd sp!, {r4-r5}
+.loop:
+    ldmia data!, {vA, vB, vC, vD}
+
+    encode vA
+    encode vB
+    encode vC
+    encode vD
+
+    and vA, vA, #0xFF
+    and vB, vB, #0xFF
+    and vC, vC, #0xFF
+    orr vA, vA, vB, lsl #8
+    orr vA, vA, vC, lsl #16
+    orr vA, vA, vD, lsl #24
+    str vA, [buffer], #4
+
+    subs count, #4
+    bne .loop
+
+    ldmfd sp!, {r4-r5}
+    bx lr