Hack back flat ROM pointer support

It was present in prerelease alpha version, but when porting to the RP2040, it was converted to double indirection, as it worked with the shitty LRU algo that I used at the time.
However, double indirection is not only slower just by requiring an extra load instruction, it's also slower by trashing the cache. On Windows, I get a 10% performance decrease with double indirection, which is pretty wild.

config.h

@@ -29,6 +29,8 @@
 //#define CONFIG_APU_MONO
 #define CONFIG_APU_N_PER_TICK 8
 #define CONFIG_APU_N_BUFSIZE 32768
+//#define CONFIG_ENABLE_LRU
+//#define CONFIG_USE_FLAT_ROM
 //#define CONFIG_PPU_ACTION_ON_START
 
 
@@ -201,6 +203,13 @@
 #define CONFIG_ENABLE_LRU 1
 #endif
 
+#ifndef CONFIG_USE_FLAT_ROM
+#define CONFIG_USE_FLAT_ROM 0
+#else
+#undef CONFIG_USE_FLAT_ROM
+#define CONFIG_USE_FLAT_ROM 1
+#endif
+
 #ifndef CONFIG_PPU_ACTION_ON_START
 #define CONFIG_PPU_ACTION_ON_START 0
 #else

fabric.h

@@ -75,9 +75,14 @@ static inline const r8* pgf_resolve_ROM_internal(void* userdata, word addr, word
     if(ud->mb->mi->ROM != NULL)
 #endif
     {
+#if CONFIG_USE_FLAT_ROM
+        res = &ud->mb->mi->ROM[bank * 0x4000];
+        return &res[addr & 0x3FFF];
+#else
         res = ud->mb->mi->ROM[bank];
         if(res != NULL)
             return &res[addr & 0x3FFF];
+#endif
     }
     
 #if CONFIG_ENABLE_LRU

mi.h

@@ -42,7 +42,11 @@ struct mb_mi_cache
 
 struct mi_dispatch
 {
+#if CONFIG_USE_FLAT_ROM
+    const r8* __restrict ROM;
+#else
     const r8* __restrict const * __restrict ROM;
+#endif
     r8* __restrict WRAM;
     r8* __restrict VRAM;
     r8* __restrict SRAM;

microcode.c

@@ -58,18 +58,29 @@ PGB_FUNC ATTR_FORCE_NOINLINE __attribute__((optimize("Os"))) static const r8* __
         {
             if(r_addr < MICACHE_R_VALUE(0x4000))
             {
+            #if CONFIG_USE_FLAT_ROM
+                ret = &mi->ROM[0];
+                return &ret[r_addr << MICACHE_R_BITS];
+            #else
                 ret = mi->ROM[0];
                 if(ret != NULL)
                     return &ret[r_addr << MICACHE_R_BITS];
+            #endif
             }
             else
             {
+            #if CONFIG_USE_FLAT_ROM
+                r_addr &= MICACHE_R_VALUE(0x3FFF);
+                ret = &mi->ROM[mi->BANK_ROM << 14];
+                return &ret[r_addr << MICACHE_R_BITS];
+            #else
                 ret = mi->ROM[mi->BANK_ROM];
                 if(ret != NULL)
                 {
                     r_addr &= MICACHE_R_VALUE(0x3FFF);
                     return &ret[r_addr << MICACHE_R_BITS];
                 }
+            #endif
             }
         }
         

test/pgb_config.h

@@ -20,6 +20,7 @@
 #define CONFIG_APU_N_PER_TICK 8
 #define CONFIG_APU_N_BUFSIZE 32768
 //#define CONFIG_ENABLE_LRU
+//#define CONFIG_USE_FLAT_ROM
 //#define CONFIG_PPU_ACTION_ON_START
 
 // 0 - none

test/winmain.c

@@ -633,7 +633,10 @@ static struct mi_dispatch dis;
 static struct mb_state mb;
 static struct ppu_t pp;
 static struct apu_t apu;
+
+#if !CONFIG_USE_FLAT_ROM
 static const r8* __restrict rommap[512];
+#endif
 
 static pixel_t* __restrict fb_lines[145];
 
@@ -692,7 +695,10 @@ int main(int argc, char** argv)
     audbuf = malloc(sizeof(*audbuf) * apu.outbuf_size);
     apu.outbuf = audbuf;
     
+    #if !CONFIG_USE_FLAT_ROM
     memset((void*)&rommap[0], 0, sizeof(rommap));
+    #endif
+    
     pixel_t* __restrict framebuffer = malloc(sizeof(pixel_t) * 256 * 145);
     memset(framebuffer, 0, sizeof(pixel_t) * 256 * 145);
     for(i = 0; i != 145; ++i)
@@ -776,11 +782,17 @@ int main(int argc, char** argv)
             fs = ff;
         }
         
+        #if !CONFIG_USE_FLAT_ROM
         for(i = 0; i != (fs >> 14); i++)
             rommap[i] = &img[i << 14];
+        #endif
         
     #if !CONFIG_ENABLE_LRU
-        dis.ROM = rommap;
+        #if CONFIG_USE_FLAT_ROM
+            dis.ROM = img;
+        #else
+            dis.ROM = rommap;
+        #endif
     #else
         dis.ROM = 0;
         dis.dispatch_ROM_Bank = cb_ROM_LRU;
@@ -800,7 +812,7 @@ int main(int argc, char** argv)
         
         mb.mi = &dis;
         micache_invalidate(&mb.micache);
-        mi_init_params_from_header(mb.mi, rommap[0]);
+        mi_init_params_from_header(mb.mi, img);
         
         mb.IE = 0;
         mb.IF = 0;
@@ -929,7 +941,7 @@ int main(int argc, char** argv)
                 fread(&dis.WRAM[0x6000], 256, 1, f);
                 fclose(f);
                 
-                memcpy(&dis.WRAM[0x6100], &dis.ROM[0][0x100], 0x50);
+                memcpy(&dis.WRAM[0x6100], &img[0x100], 0x50);
                 
                 mb.micache.mc_read[0] = &dis.WRAM[0x6000]; 
                 mb.micache.mc_execute[0] = &dis.WRAM[0x6000];