optimized some float/int conversions, added -fomit-frame-pointer flag

Makefile.gc

@@ -26,7 +26,7 @@ INCLUDES	:=	source source/m68k source/z80 source/sound source/sound/SRC source/n
 # options for code generation
 #---------------------------------------------------------------------------------
 
-CFLAGS	= -O3 -Wall $(MACHDEP) $(INCLUDE) -DWORDS_BIGENDIAN -DNGC="1" -DHW_DOL
+CFLAGS	= -O3 -fomit-frame-pointer -Wall $(MACHDEP) $(INCLUDE) -DWORDS_BIGENDIAN -DNGC="1" -DHW_DOL
 CXXFLAGS	=	$(CFLAGS)
 
 LDFLAGS	=	$(MACHDEP) -Wl,-Map,$(notdir $@).map

Makefile.wii

@@ -26,7 +26,7 @@ INCLUDES	:=	source source/m68k source/z80 source/sound source/sound/SRC source/n
 # options for code generation
 #---------------------------------------------------------------------------------
 
-CFLAGS	= -O3 -mrvl -Wall $(MACHDEP) -Wno-strict-aliasing $(INCLUDE) -DWORDS_BIGENDIAN -DNGC="1" -DHW_RVL
+CFLAGS	= -O3 -fomit-frame-pointer -mrvl -Wall $(MACHDEP) -Wno-strict-aliasing $(INCLUDE) -DWORDS_BIGENDIAN -DNGC="1" -DHW_RVL
 CXXFLAGS	=	$(CFLAGS)
 
 LDFLAGS	=	$(MACHDEP) -Wl,-Map,$(notdir $@).map

source/ngc/config.h

@@ -31,8 +31,8 @@
 typedef struct 
 {
   char version[15];
-  double psg_preamp;
+  int32 psg_preamp;
-  double fm_preamp;
+  int32 fm_preamp;
   uint8 boost;
   uint8 filter;
   uint8 hq_fm;

source/sound/fm.c

@@ -1804,8 +1804,16 @@ void YM2612UpdateOne(int **buffer, int length)
     Limit(rt,MAXOUT,MINOUT);
 
     /* buffering */
-    bufL[i] = lt;
+    if (src_buffer)
-    bufR[i] = rt;
+    {
+      src_buffer[i*2]      = (float) lt / (8.0 * 0x10000000);
+      src_buffer[i*2 + 1]  = (float) rt / (8.0 * 0x10000000);
+    }
+    else
+    {
+      bufL[i] = lt;
+      bufR[i] = rt;
+    }
 
     /* timer A control */
     INTERNAL_TIMER_A();

source/sound/sn76489.c

@@ -48,7 +48,7 @@ static SN76489_Context SN76489[MAX_SN76489];
 void SN76489_Init(int which, int PSGClockValue, int SamplingRate)
 {
   SN76489_Context *p = &SN76489[which];
-  p->dClock=(float)PSGClockValue/16/SamplingRate;
+  p->dClock=(float)(PSGClockValue)/16.0/(float)SamplingRate;
   SN76489_Config(which, MUTE_ALLON, VOL_FULL, FB_SEGAVDP, SRW_SEGAVDP, 1);
   SN76489_Reset(which);
 }

source/sound/sound.c

@@ -22,7 +22,6 @@
  ****************************************************************************************/
 
 #include "shared.h"
-#include "samplerate.h"
 
 #define CLOCK_NTSC 53693175
 #define CLOCK_PAL  53203424
@@ -34,13 +33,10 @@ void (*_YM2612_Update)(int **buf, int length);
 int (*_YM2612_Reset)(void);
 
 /* cycle-accurate samples */
-static double m68cycles_per_sample[2];
+static int m68cycles_per_sample[2];
-static double z80cycles_per_sample[2];
 
-/* libsamplerate buffers (max. is 488 cycles per line x 313 lines / 144) */
+/* pointer to current SRC buffer */
-static SRC_DATA src_data;
+float *src_buffer;
-static float src_in[1061*2];
-static int src_buffer[2][1061];
 
 /* YM2612 register arrays */
 int fm_reg[2][0x100];
@@ -48,14 +44,14 @@ int fm_reg[2][0x100];
 /* return the number of samples that should have been rendered so far */
 static inline uint32 fm_sample_cnt(uint8 is_z80)
 {
-  if (is_z80) return (uint32) ((double)(count_z80 + current_z80 - z80_ICount) / z80cycles_per_sample[0]);
+  if (is_z80) return ((count_z80 + current_z80 - z80_ICount) * 15) / (7 * m68cycles_per_sample[0]);
-  else return (uint32) ((double) count_m68k / m68cycles_per_sample[0]);
+  else return count_m68k / m68cycles_per_sample[0];
 }
 
 static inline uint32 psg_sample_cnt(uint8 is_z80)
 {
-  if (is_z80) return (uint32) ((double)(count_z80 + current_z80 - z80_ICount) / z80cycles_per_sample[1]);
+  if (is_z80) return ((count_z80 + current_z80 - z80_ICount) * 15) / (7 * m68cycles_per_sample[1]);
-  else return (uint32) ((double) count_m68k / m68cycles_per_sample[1]);
+  else return count_m68k / m68cycles_per_sample[1];
 }
 
 /* update FM samples */
@@ -64,16 +60,12 @@ static inline void fm_update()
   if(snd.fm.curStage - snd.fm.lastStage > 0)
   {
     int *tempBuffer[2];
-    
+    tempBuffer[0] = snd.fm.buffer[0] + snd.fm.lastStage;
-    if (config.hq_fm && !config.fm_core)
+    tempBuffer[1] = snd.fm.buffer[1] + snd.fm.lastStage;
+
+    if (src_buffer)
     {
-      tempBuffer[0] = src_buffer[0] + snd.fm.lastStage;
+      src_buffer = src_data.data_in + (snd.fm.lastStage * 2);
-      tempBuffer[1] = src_buffer[1] + snd.fm.lastStage;
-    }
-    else
-    {
-      tempBuffer[0] = snd.fm.buffer[0] + snd.fm.lastStage;
-      tempBuffer[1] = snd.fm.buffer[1] + snd.fm.lastStage;
     }
 
     _YM2612_Update(tempBuffer, snd.fm.curStage - snd.fm.lastStage);
@@ -97,28 +89,20 @@ void sound_init(int rate)
   double vclk = (vdp_pal ? (double)CLOCK_PAL : (double)CLOCK_NTSC) / 7.0;  /* 68000 and YM2612 clock */
   double zclk = (vdp_pal ? (double)CLOCK_PAL : (double)CLOCK_NTSC) / 15.0; /* Z80 and SN76489 clock  */
 
-  /* cycle-accurate FM samples */
+  /* cycle-accurate samples */
-  if (config.hq_fm && !config.fm_core)
+  m68cycles_per_sample[0] = (m68cycles_per_line * lines_per_frame * vdp_rate) / rate;
-  {
+  m68cycles_per_sample[1] = (m68cycles_per_line * lines_per_frame * vdp_rate) / rate;
-    m68cycles_per_sample[0] = 144.0;
-    z80cycles_per_sample[0] = (144.0 * 7.0) / 15.0;
 
-    /* initialize samplerate converter data */
+  /* YM2612 is emulated at the original frequency */
-    src_data.data_in        = src_in;
+  src_buffer = 0;
-    src_data.data_out       = snd.fm.src_out;
+  if (src_data.data_in)
-    src_data.input_frames   = (int)(((double)m68cycles_per_line * (double)lines_per_frame / 144.0) + 0.5);
-    src_data.output_frames  = rate / vdp_rate;
-    src_data.src_ratio      = (double)src_data.output_frames  / (double)src_data.input_frames;
-  }
-  else
   {
-    m68cycles_per_sample[0] = ((double)m68cycles_per_line * (double)lines_per_frame) / (double) (rate / vdp_rate);
+    /* YM2612 original frequency is VCLK/144 */
-    z80cycles_per_sample[0] = ((double)z80cycles_per_line * (double)lines_per_frame) / (double) (rate / vdp_rate);
+    m68cycles_per_sample[0] = 144;
-  }
 
-  /* cycle-accurate PSG samples */
+    /* Initialize SRC buffer */
-  m68cycles_per_sample[1] = ((double)m68cycles_per_line * (double)lines_per_frame) / (double) (rate / vdp_rate);
+    src_buffer = src_data.data_in;
-  z80cycles_per_sample[1] = ((double)z80cycles_per_line * (double)lines_per_frame) / (double) (rate / vdp_rate);
+  }
 
   /* initialize sound chips */
   SN76489_Init(0, (int)zclk, rate);
@@ -145,7 +129,7 @@ void sound_init(int rate)
 void sound_update(void)
 {
   /* finalize sound buffers */
-  snd.fm.curStage  = (config.hq_fm && !config.fm_core) ? src_data.input_frames : snd.buffer_size;
+  snd.fm.curStage  = (src_data.data_in) ? src_data.input_frames : snd.buffer_size;
   snd.psg.curStage = snd.buffer_size;
 
   /* update last samples (if needed) */
@@ -153,45 +137,34 @@ void sound_update(void)
   psg_update();
 
   /* Resampling */
-  if (config.hq_fm && !config.fm_core)
+  if (src_data.data_in)
   {
-    double scaled_value ;
-    int len = src_data.input_frames;
-
-    /* this is basically libsamplerate "src_int_to_float_array" function, adapted to interlace samples */
-    while (len)
-    {
-      len -- ;
-      src_in[len*2]      = (float) (src_buffer[0] [len] / (8.0 * 0x10000000));
-      src_in[len*2 + 1]  = (float) (src_buffer[1] [len] / (8.0 * 0x10000000));
-    }
-
     /* samplerate conversion */
     src_simple (&src_data, (config.hq_fm&1) ? SRC_LINEAR : SRC_SINC_FASTEST, 2);
 
     /* this is basically libsamplerate "src_float_to_int_array" function, adapted to interlace samples */
-    len = snd.buffer_size;
+    double scaled_value;
+    int len = snd.buffer_size;
     while (len)
     {
       len -- ;
-
+      scaled_value = src_data.data_out[len*2] * (8.0 * 0x10000000);
-      scaled_value = snd.fm.src_out[len*2] * (8.0 * 0x10000000);
       if (scaled_value >= (1.0 * 0x7FFFFFFF))
         snd.fm.buffer[0][len] = 0x7fffffff;
       else if (scaled_value <= (-8.0 * 0x10000000))
         snd.fm.buffer[0][len] = -1 - 0x7fffffff;
       else
-        snd.fm.buffer[0][len] = (long)scaled_value;
+        snd.fm.buffer[0][len] = lrint(scaled_value);
 
-      scaled_value = snd.fm.src_out[len*2+1] * (8.0 * 0x10000000);
+      scaled_value = src_data.data_out[len*2+1] * (8.0 * 0x10000000);
       if (scaled_value >= (1.0 * 0x7FFFFFFF))
         snd.fm.buffer[1][len] = 0x7fffffff;
       else if (scaled_value <= (-8.0 * 0x10000000))
         snd.fm.buffer[1][len] = -1 - 0x7fffffff;
       else
-        snd.fm.buffer[1][len] = (long)scaled_value;
+        snd.fm.buffer[1][len] = lrint(scaled_value);
-      }
     }
+  }
 
   /* reset samples count */
   snd.fm.curStage   = 0;

source/sound/sound.h

@@ -28,6 +28,7 @@
 extern int fm_reg[2][0x100];
 extern double fm_timera_tab[0x400];
 extern double fm_timerb_tab[0x100];
+extern float *src_buffer;
 
 /* Function prototypes */
 extern void sound_init(int rate);

source/system.c

@@ -35,6 +35,7 @@ uint32 count_z80;
 uint32 line_z80;
 int32 current_z80;
 uint8 system_hw;
+SRC_DATA src_data;
 
 static inline void audio_update (void);
 
@@ -65,9 +66,13 @@ void system_reset (void)
   SN76489_Reset(0);
 
   /* Sound Buffers */
-  memset (snd.psg.buffer, 0, SND_SIZE);
+  if (snd.psg.buffer)     memset(snd.psg.buffer,   0, SND_SIZE);
-  memset (snd.fm.buffer[0], 0, SND_SIZE*2);
+  if (snd.fm.buffer[0])   memset(snd.fm.buffer[0], 0, SND_SIZE*2);
-  memset (snd.fm.buffer[1], 0, SND_SIZE*2);
+  if (snd.fm.buffer[1])   memset(snd.fm.buffer[1], 0, SND_SIZE*2);
+
+  /* SRC */
+  if (src_data.data_in)   memset(src_data.data_in, 0, src_data.input_frames  * 2 * sizeof(float));
+  if (src_data.data_out)  memset(src_data.data_out,0, src_data.output_frames * 2 * sizeof(float));
 }
 
 /****************************************************************
@@ -275,28 +280,30 @@ int audio_init (int rate)
   snd.buffer[0] = (int16 *) malloc(SND_SIZE);
   snd.buffer[1] = (int16 *) malloc(SND_SIZE);
   if (!snd.buffer[0] || !snd.buffer[1]) return (-1);
-  memset (snd.buffer[0], 0, SND_SIZE);
-  memset (snd.buffer[1], 0, SND_SIZE);
 #endif
 
   /* YM2612 stream buffers */
   snd.fm.buffer[0] = (int *)malloc (SND_SIZE*2);
   snd.fm.buffer[1] = (int *)malloc (SND_SIZE*2);
   if (!snd.fm.buffer[0] || !snd.fm.buffer[1]) return (-1);
-  memset (snd.fm.buffer[0], 0, SND_SIZE*2);
-  memset (snd.fm.buffer[1], 0, SND_SIZE*2);
 
-  /* SRC buffers */
+  /* YM2612 resampling */
+  src_data.data_in  = NULL;
+  src_data.data_out = NULL;
   if (config.hq_fm && !config.fm_core)
   {
-    snd.fm.src_out = (float *) malloc(snd.buffer_size*2*sizeof(float));
+    /* initialize SRC */
-    if (!snd.fm.src_out) return (-1);
+    src_data.input_frames   = (int)(((double)m68cycles_per_line * (double)lines_per_frame / 144.0) + 0.5);
+    src_data.output_frames  = snd.buffer_size;
+    src_data.data_in        = (float *)malloc(src_data.input_frames  * 2 * sizeof(float));
+    src_data.data_out       = (float *)malloc(src_data.output_frames * 2 * sizeof(float));
+    src_data.src_ratio      = (double)src_data.output_frames  / (double)src_data.input_frames;
+    if (!src_data.data_in || !src_data.data_out) return (-1);
   }
 
   /* SN76489 stream buffers */
   snd.psg.buffer = (int16 *)malloc (SND_SIZE);
   if (!snd.psg.buffer) return (-1);
-  memset (snd.psg.buffer, 0, SND_SIZE);
 
   /* Set audio enable flag */
   snd.enabled = 1;
@@ -310,12 +317,13 @@ int audio_init (int rate)
 void audio_shutdown(void)
 {
   /* free sound buffers */
-  if (snd.buffer[0])    free(snd.buffer[0]);
+  if (snd.buffer[0])      free(snd.buffer[0]);
-  if (snd.buffer[1])    free(snd.buffer[1]);
+  if (snd.buffer[1])      free(snd.buffer[1]);
-  if (snd.fm.buffer[0]) free(snd.fm.buffer[0]);
+  if (snd.fm.buffer[0])   free(snd.fm.buffer[0]);
-  if (snd.fm.buffer[1]) free(snd.fm.buffer[1]);
+  if (snd.fm.buffer[1])   free(snd.fm.buffer[1]);
-  if (snd.fm.src_out)   free(snd.fm.src_out);
+  if (snd.psg.buffer)     free(snd.psg.buffer);
-  if (snd.psg.buffer)   free(snd.psg.buffer);
+  if (src_data.data_in)   free(src_data.data_in);
+  if (src_data.data_out)  free(src_data.data_out);
 }
 
 static int ll, rr;
@@ -324,8 +332,8 @@ static inline void audio_update (void)
 {
   int i;
   int l, r;
-  double psg_preamp = config.psg_preamp;
+  int psg_preamp = config.psg_preamp;
-  double fm_preamp  = config.fm_preamp;
+  int fm_preamp  = config.fm_preamp;
   int boost  = config.boost;
   int filter = config.filter;
 
@@ -339,9 +347,9 @@ static inline void audio_update (void)
   /* mix samples */
   for (i = 0; i < snd.buffer_size; i ++)
   {
-    l = r = (int) ((double)snd.psg.buffer[i] * psg_preamp);
+    l = r = (snd.psg.buffer[i] * psg_preamp) / 100;
-    l += (int) ((double)snd.fm.buffer[0][i] * fm_preamp);
+    l += ((snd.fm.buffer[0][i] * fm_preamp) / 100);
-    r += (int) ((double)snd.fm.buffer[1][i] * fm_preamp);
+    r += ((snd.fm.buffer[1][i] * fm_preamp) / 100);
     snd.fm.buffer[0][i] = 0;
     snd.fm.buffer[1][i] = 0;
     snd.psg.buffer[i] = 0;

source/system.h

@@ -24,6 +24,8 @@
 #ifndef _SYSTEM_H_
 #define _SYSTEM_H_
 
+#include "samplerate.h"
+
 #define SYSTEM_GENESIS    0
 #define SYSTEM_MEGADRIVE  1
 #define SYSTEM_PICO       2
@@ -66,7 +68,6 @@ typedef struct
     int curStage;
     int lastStage;
     int *buffer[2];
-    float *src_out; /* SRC conversion buffer */
   } fm;
   struct
   {
@@ -86,6 +87,7 @@ extern uint32 count_z80;
 extern uint32 line_z80;
 extern int32 current_z80;
 extern uint8 system_hw;
+extern SRC_DATA src_data;
 
 /* Function prototypes */
 extern void system_init (void);

source/vdp.c

@@ -91,9 +91,6 @@ static uint16 sat_base_mask;      /* Base bits of SAT */
 static uint16 sat_addr_mask;      /* Index bits of SAT */
 static uint32 dma_endCycles;      /* 68k cycles to DMA end */
 static uint8 dma_type;            /* Type of DMA */
-static double vdp_timings[4][4];  /* DMA timings */
-
-static inline void vdp_reg_w(unsigned int r, unsigned int d);
 
 /* DMA Timings
 
@@ -122,31 +119,22 @@ static inline void vdp_reg_w(unsigned int r, unsigned int d);
  CRAM or VSRAM for a 68K > VDP transfer, in which case it is in words.
 
 */
-static const uint8 dma_rates[16] = {
+static const uint32 dma_rates[16] = {
-  8,  9, 83 , 102,  /* 68K to VRAM */
+  8,   83,  9, 102, /* 68K to VRAM (1 word = 2 bytes) */
-  16, 18, 167, 205, /* 68K to CRAM or VSRAM */
+  16, 167, 18, 205, /* 68K to CRAM or VSRAM */
-  15, 17, 166, 204, /* DMA fill */
+  15, 166, 17, 204, /* DMA fill */
-  8,  9, 83 , 102,  /* DMA Copy */
+  8,   83,  9, 102, /* DMA Copy */
 };
 
 /* Function prototypes */
 static inline void data_write(unsigned int data);
+static inline void vdp_reg_w(unsigned int r, unsigned int d);
 
 /*--------------------------------------------------------------------------*/
 /* Init, reset, shutdown functions                                          */
 /*--------------------------------------------------------------------------*/
 void vdp_init(void)
 {
-  /* reinitialize DMA timings table */
-  int i;
-  for (i=0; i<4; i++)
-  {
-    vdp_timings[0][i] = ((double)m68cycles_per_line) / ((double) dma_rates[i]);
-    vdp_timings[1][i] = ((double)m68cycles_per_line) / ((double) dma_rates[i +  4]);
-    vdp_timings[2][i] = ((double)m68cycles_per_line) / ((double) dma_rates[i +  8]);
-    vdp_timings[3][i] = ((double)m68cycles_per_line) / ((double) dma_rates[i + 12]);
-  }
-
   /* PAL/NTSC timings */
   vdp_rate        = vdp_pal ? 50 : 60;
   lines_per_frame = vdp_pal ? 313 : 262;
@@ -280,24 +268,27 @@ void vdp_restore(uint8 *vdp_regs)
 /* Update DMA timings (this is call on start of DMA and then at the start of each scanline) */
 void dma_update()
 {
-  int32 left_cycles;
+  int dma_cycles = 0;
-  uint32 dma_cycles, dma_bytes;
-  uint8 index = 0;
 
-  /* get the appropriate tranfer rate (bytes/line) for this DMA operation */
+  /* DMA timings table index */
-  if ((status&8) || !(reg[1] & 0x40)) index = 2; /* VBLANK or Display OFF */
+  int index = (4 * dma_type) + ((reg[12] & 1)*2);
-  index += (reg[12] & 1);               /* 32 or 40 Horizontal Cells    */
+  if ((status&8) || !(reg[1] & 0x40)) index++;
 
-  /* calculate transfer quantity for the remaining 68k cycles */
+  /* DMA transfer rate */
-  left_cycles = line_m68k + m68cycles_per_line - count_m68k;
+  int rate = dma_rates[index];
+
+  /* 68k cycles left */
+  int left_cycles = (line_m68k + m68cycles_per_line) - count_m68k;
   if (left_cycles < 0) left_cycles = 0;
-  dma_bytes = (uint32)(((double)left_cycles / vdp_timings[dma_type][index]) + 0.5);
+
+  /* DMA bytes left */
+  int dma_bytes = (left_cycles * rate) / m68cycles_per_line;
 
   /* determinate DMA length in CPU cycles */
   if (dma_length < dma_bytes)
   {
     /* DMA will be finished during this line */
-    dma_cycles = (uint32)(((double)dma_length * vdp_timings[dma_type][index]) + 0.5);
+    dma_cycles = (dma_length * m68cycles_per_line) / rate;
     dma_length = 0;
   }
   else
@@ -307,6 +298,7 @@ void dma_update()
     dma_length -= dma_bytes;
   }
 
+  /* update 68k cycles counter */
   if (dma_type < 2)
   {
     /* 68K COPY to V-RAM */