yeah

src/engine/platform/nes.cpp

@@ -28,7 +28,7 @@ struct _nla_table nla_table;
 
 #define CHIP_DIVIDER 16
 
-#define rWrite(a,v) if (!skipRegisterWrites) {doWrite(a,v); regPool[(a)&0x7f]=v; if (dumpWrites) {addWrite(a,v);} }
+#define rWrite(a,v) if (!skipRegisterWrites) {writes.push(QueuedWrite((a),v)); if (dumpWrites) {addWrite((a),v);} }
 
 const char* regCheatSheetNES[]={
   "S0Volume", "4000",
@@ -86,10 +86,10 @@ void DivPlatformNES::doWrite(unsigned short addr, unsigned char data) {
           unsigned char next=((unsigned char)s->data8[dacPos]+0x80)>>1; \
           if (dacAntiClickOn && dacAntiClick<next) { \
             dacAntiClick+=8; \
-            rWrite(0x4011,dacAntiClick); \
+            doWrite(0x4011,dacAntiClick); \
           } else { \
             dacAntiClickOn=false; \
-            rWrite(0x4011,next); \
+            doWrite(0x4011,next); \
           } \
         } \
         dacPos++; \
@@ -109,6 +109,14 @@ void DivPlatformNES::acquire_puNES(short** buf, size_t len) {
   for (size_t i=0; i<len; i++) {
     doPCM;
 
+    if (!writes.empty()) 
+    {
+      QueuedWrite w=writes.front();
+      doWrite(w.addr,w.val);
+      regPool[w.addr&0x1f]=w.val;
+      writes.pop();
+    }
+  
     apu_tick(nes,NULL);
     nes->apu.odd_cycle=!nes->apu.odd_cycle;
     if (nes->apu.clocked) {
@@ -135,6 +143,14 @@ void DivPlatformNES::acquire_NSFPlay(short** buf, size_t len) {
   for (size_t i=0; i<len; i++) {
     doPCM;
 
+    if (!writes.empty()) 
+    {
+      QueuedWrite w=writes.front();
+      doWrite(w.addr,w.val);
+      regPool[w.addr&0x1f]=w.val;
+      writes.pop();
+    }
+  
     nes1_NP->Tick(8);
     nes2_NP->TickFrameSequence(8);
     nes2_NP->Tick(8);
@@ -400,6 +416,35 @@ void DivPlatformNES::tick(bool sysTick) {
             logV("switching bank to %d",dpcmBank);
             if (dumpWrites) addWrite(0xffff0004,dpcmBank);
           }
+          //sample custom loop point...
+
+          DivSample* lsamp = parent->getSample(dacSample);
+
+          //how it works:
+          //when the initial sample info is written (see above) and playback is launched,
+          //the parameters (start point in memory and length) are locked until sample end
+          //is reached.
+
+          //thus, if we write new data after just several APU clock cycles, it will be used only when
+          //sample finishes one full loop.
+
+          //thus we can write sample's loop point as "start address" and sample's looped part length
+          //as "full sample length".
+
+          //APU will play full sample once and then repeatedly cycle through the looped part.
+
+          //sources:
+          //https://www.nesdev.org/wiki/APU_DMC
+          //https://www.youtube.com/watch?v=vB4P8x2Am6Y
+
+          if(lsamp->loopEnd > lsamp->loopStart && goingToLoop)
+          {
+            int loop_start_addr = (sampleOffDPCM[dacSample] + lsamp->loopStart) / 8;
+            int loop_len = (lsamp->loopEnd - lsamp->loopStart) / 8;
+
+            rWrite(0x4012,(loop_start_addr >> 6)&0xff);
+            rWrite(0x4013,(loop_len >> 4)&0xff);
+          }
         }
       } else {
         if (nextDPCMFreq>=0) {
@@ -790,6 +835,7 @@ float DivPlatformNES::getPostAmp() {
 }
 
 void DivPlatformNES::reset() {
+  while (!writes.empty()) writes.pop();
   for (int i=0; i<5; i++) {
     chan[i]=DivPlatformNES::Channel();
     chan[i].std.setEngine(parent);

src/engine/platform/nes.h

@@ -24,6 +24,7 @@
 
 #include "sound/nes_nsfplay/nes_apu.h"
 #include "sound/nes_nsfplay/5e01_apu.h"
+#include "../../fixedQueue.h"
 
 class DivPlatformNES: public DivDispatch {
   struct Channel: public SharedChannel<signed char> {
@@ -44,6 +45,13 @@ class DivPlatformNES: public DivDispatch {
   Channel chan[5];
   DivDispatchOscBuffer* oscBuf[5];
   bool isMuted[5];
+  struct QueuedWrite {
+      unsigned short addr;
+      unsigned char val;
+      QueuedWrite(): addr(0), val(0) {}
+      QueuedWrite(unsigned short a, unsigned char v): addr(a), val(v) {}
+  };
+  FixedQueue<QueuedWrite,128> writes;
   int dacPeriod, dacRate, dpcmPos;
   unsigned int dacPos, dacAntiClick;
   int dacSample;

src/gui/sampleEdit.cpp

@@ -254,15 +254,26 @@ void FurnaceGUI::drawSampleEdit() {
             }
             break;
           case DIV_SYSTEM_NES:
+          {
             if (sample->loop) {
-              if (sample->loopStart!=0 || sample->loopEnd!=(int)(sample->samples)) {
-                SAMPLE_WARN(warnLoopPos,_("NES: loop point ignored on DPCM (may only loop entire sample)"));
+              if (sample->loopStart&511) {
+                int tryWith=(sample->loopStart)&(~511);
+                if (tryWith>(int)sample->samples) tryWith-=512;
+                String alignHint=fmt::sprintf(_("NES: loop start must be a multiple of 512 (try with %d)"),tryWith);
+                SAMPLE_WARN(warnLoopStart,alignHint);
+              }
+              if ((sample->loopEnd)&127) {
+                int tryWith=(sample->loopEnd + 1)&(~127); //+1 bc of how sample length is treated: https://www.nesdev.org/wiki/APU_DMC
+                if (tryWith>(int)sample->samples) tryWith-=128;
+                String alignHint=fmt::sprintf(_("NES: loop end must be a multiple of 128 (try with %d)"),tryWith);
+                SAMPLE_WARN(warnLoopEnd,alignHint);
               }
             }
             if (sample->samples>32648) {
               SAMPLE_WARN(warnLength,_("NES: maximum DPCM sample length is 32648"));
             }
             break;
+          }
           case DIV_SYSTEM_X1_010:
             if (sample->loop) {
               SAMPLE_WARN(warnLoop,_("X1-010: samples can't loop"));