From: kub <derkub@gmail.com>
Date: Wed, 21 Sep 2022 23:26:48 +0000 (+0000)
Subject: sms, add fast renderer, remove 1st column (8 px) if blanked
X-Git-Tag: v2.00~296
X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=96948bdfc88d5f66c3e7047b7fa3256813fd657d;p=picodrive.git

sms, add fast renderer, remove 1st column (8 px) if blanked
---

diff --git a/pico/draw.c b/pico/draw.c
index bf4ef38e..38820e8f 100644
--- a/pico/draw.c
+++ b/pico/draw.c
@@ -1646,13 +1646,17 @@ void FinalizeLine555(int sh, int line, struct PicoEState *est)
 
   PicoDrawUpdateHighPal();
 
-  if ((PicoIn.AHW & PAHW_SMS) && (Pico.m.hardware & 0x3) == 0x3)
-                                        len = 160;
-  else if (Pico.video.reg[12]&1)        len = 320;
-  else                                  len = 256;
+  len = 256;
+  if ((PicoIn.AHW & PAHW_SMS) && (Pico.m.hardware & (PMS_HW_GG|PMS_HW_LCD)) == (PMS_HW_GG|PMS_HW_LCD))
+    len = 160;
+  else if (!(PicoIn.AHW & PAHW_SMS) && (Pico.video.reg[12]&1))
+    len = 320;
+  if ((PicoIn.AHW & PAHW_SMS) && (Pico.video.reg[0] & 0x20) && len == 256)
+    len -= 8, ps += 8;
 
   if ((*est->PicoOpt & POPT_EN_SOFTSCALE) && len < 320) {
-    if (len == 256) {
+    if (len >= 240 && len <= 256) {
+      pd += (256-len)>>1;
       switch (PicoIn.filter) {
       case 3: h_upscale_bl4_4_5(pd, 320, ps, 256, len, f_pal); break;
       case 2: h_upscale_bl2_4_5(pd, 320, ps, 256, len, f_pal); break;
@@ -1706,10 +1710,13 @@ void FinalizeLine8bit(int sh, int line, struct PicoEState *est)
     Pico.m.dirtyPal = 2;
   }
 
-  if ((PicoIn.AHW & PAHW_SMS) && (Pico.m.hardware & 0x3) == 0x3)
-                                        len = 160;
-  else if (Pico.video.reg[12]&1)        len = 320;
-  else                                  len = 256;
+  len = 256;
+  if ((PicoIn.AHW & PAHW_SMS) && (Pico.m.hardware & (PMS_HW_GG|PMS_HW_LCD)) == (PMS_HW_GG|PMS_HW_LCD))
+    len = 160;
+  else if (!(PicoIn.AHW & PAHW_SMS) && (Pico.video.reg[12]&1))
+    len = 320;
+  if ((PicoIn.AHW & PAHW_SMS) && (Pico.video.reg[0] & 0x20) && len == 256)
+    len -= 8, ps += 8;
 
   if (DrawLineDestIncrement == 0)
     pd = est->HighCol+8;
@@ -1720,7 +1727,7 @@ void FinalizeLine8bit(int sh, int line, struct PicoEState *est)
     if (!sh && (est->rendstatus & PDRAW_SONIC_MODE))
       pal = est->SonicPalCount*0x40;
     // Smoothing can't be used with CLUT, hence it's always Nearest Neighbour.
-    if (len == 256)
+    if (len >= 240)
       // use reverse version since src and dest ptr may be the same.
       rh_upscale_nn_4_5(pd, 320, ps, 256, len, f_or);
     else
diff --git a/pico/draw_arm.S b/pico/draw_arm.S
index c5997025..962811ea 100644
--- a/pico/draw_arm.S
+++ b/pico/draw_arm.S
@@ -1651,34 +1651,47 @@ FinalizeLine555:
     ldr     r4, [r5]
     ldr     r7, [r5, #OFS_PicoIn_AHW-OFS_PicoIn_opt]
     ldrb    r12,[r8, #OFS_Pico_video_reg+12]
+    ldrb    r6, [r8, #OFS_Pico_video_reg+0]
     ldr     r2, [r8, #OFS_Pico_m_hardware]
     add     r1, r1, #8
 
-    tst     r7, #0x10
-    beq     .fl_no20colRGB555
+    tst     r7, #0x10   	@ SMS ?
+    beq     .fl_noSMS
+
     and     r7, r2, #0x3
     cmp     r7, #0x3		@ Game Gear, LCD?
-    bne     .fl_no20colRGB555
+    beq     .fl_gg20col
+
+    tst     r6, #0x20
+    movne   r2, #248/8  	@ len = 248
+    addne   r1, r1, #8  	@ ps += 8
+    moveq   r2, #256/8		@ len = 256
+    b       .fl_check32scaling
 
+.fl_gg20col:
     mov     r2, #160/8		@ len = 160
     tst     r4, #0x4000		@ EN_SOFTSCALE?
     bne     .fl_20scale_RGB555	@ scale 160->320
-    beq     .fl_checkborder
+    b       .fl_checkborder
 
-.fl_no20colRGB555:
+.fl_noSMS:
     tst     r12, #1		@ h32?
     movne   r2, #320/8          @ len = 320
-    bne     .fl_no32colRGB555
-    moveq   r2, #256/8		@ len = 256
+    bne     .fl_40colRGB555
+    mov     r2, #256/8		@ len = 256
+
+.fl_check32scaling:
     tst     r4, #0x4000		@ EN_SOFTSCALE?
+    rsbne   r7, r2, #256/8
+    addne   r0, r0, r7, lsl #3	@ pd += (256-len)>>1
     bne     .fl_32scale_RGB555	@ scale 256->320
 
 .fl_checkborder:
     tst     r4, #0x0100		@ DIS_32C_BORDER?
-    rsbeq   r4, r2, #320/8	@ pd += (320-len)/2
-    addeq   r0, r0, r4, lsl #3
+    rsbeq   r7, r2, #320/8	@ pd += (320-len)/2
+    addeq   r0, r0, r7, lsl #3
 
-.fl_no32colRGB555:
+.fl_40colRGB555:
 #ifdef UNALIGNED_DRAWLINEDEST
     @ this is basically for Gizmondo, which has unaligned odd lines in the framebuffer
     tst     r0, #2
@@ -2250,24 +2263,51 @@ blockcpy:
     cmp     r0, r1
     bhs     blockcpyhi
 
-    mov     r2, r2, lsr #4
+    subs    r2, r2, #16
+    blt     blockcpy2
 blockcpy_loop:
     ldmia   r1!, {r3-r5,r12}
-    subs    r2, r2, #1
+    subs    r2, r2, #16
     stmia   r0!, {r3-r5,r12}
-    bne     blockcpy_loop
+    bge     blockcpy_loop
+
+blockcpy2:
+    adds    r2, r2, #16-4
+    ldmltfd sp!, {r4,r5}
+    bxlt    lr
+
+blockcpy_loop2:
+    ldr     r3, [r1], #4
+    subs    r2, r2, #4
+    str     r3, [r0], #4
+    bge     blockcpy_loop2
+
     ldmfd   sp!, {r4,r5}
     bx      lr
 
 blockcpyhi:
     add     r0, r0, r2
     add     r1, r1, r2
-    mov     r2, r2, lsr #4
+
+    subs    r2, r2, #16
+    blt     blockcpyhi2
 blockcpyhi_loop:
     ldmdb   r1!, {r3-r5,r12}
-    subs    r2, r2, #1
+    subs    r2, r2, #16
     stmdb   r0!, {r3-r5,r12}
-    bne     blockcpyhi_loop
+    bge     blockcpyhi_loop
+
+blockcpyhi2:
+    adds    r2, r2, #16-4
+    ldmltfd sp!, {r4,r5}
+    bxlt    lr
+
+blockcpyhi_loop2:
+    ldr     r3, [r1, #-4]!
+    subs    r2, r2, #4
+    str     r3, [r0, #-4]!
+    bge     blockcpyhi_loop2
+
     ldmfd   sp!, {r4,r5}
     bx      lr
 
@@ -2281,32 +2321,61 @@ blockcpy_or:
     cmp     r0, r1
     bhs     blockcpyhi_or
 
-    mov     r2, r2, lsr #4
+    subs    r2, r2, #16
+    blt     blockcpy_or2
 blockcpy_loop_or:
     ldmia   r1!, {r4-r6,r12}
-    subs    r2, r2, #1
+    subs    r2, r2, #16
     orr     r4, r4, r3
     orr     r5, r5, r3
     orr     r6, r6, r3
     orr     r12,r12,r3
     stmia   r0!, {r4-r6,r12}
-    bne     blockcpy_loop_or
+    bge     blockcpy_loop_or
+
+blockcpy_or2:
+    adds    r2, r2, #16-4
+    ldmltfd sp!, {r4-r6}
+    bxlt    lr
+
+blockcpy_loop_or2:
+    ldr     r4, [r1], #4
+    subs    r2, r2, #4
+    orr     r4, r4, r3
+    str     r4, [r0], #4
+    bge     blockcpy_loop_or2
+
     ldmfd   sp!, {r4-r6}
     bx      lr
 
 blockcpyhi_or:
     add     r0, r0, r2
     add     r1, r1, r2
-    mov     r2, r2, lsr #4
+
+    subs    r2, r2, #16
+    blt     blockcpyhi_or2
 blockcpyhi_loop_or:
     ldmdb   r1!, {r4-r6,r12}
-    subs    r2, r2, #1
+    subs    r2, r2, #16
     orr     r4, r4, r3
     orr     r5, r5, r3
     orr     r6, r6, r3
     orr     r12,r12,r3
     stmdb   r0!, {r4-r6,r12}
-    bne     blockcpyhi_loop_or
+    bge     blockcpyhi_loop_or
+
+blockcpyhi_or2:
+    adds    r2, r2, #16-4
+    ldmltfd sp!, {r4-r6}
+    bxlt    lr
+
+blockcpyhi_loop_or2:
+    ldr     r4, [r1, #-4]!
+    subs    r2, r2, #4
+    orr     r4, r4, r3
+    str     r4, [r0, #-4]!
+    bge     blockcpyhi_loop_or2
+
     ldmfd   sp!, {r4-r6}
     bx      lr
 
diff --git a/pico/mode4.c b/pico/mode4.c
index 451bddfb..d28b1f57 100644
--- a/pico/mode4.c
+++ b/pico/mode4.c
@@ -154,15 +154,14 @@ static void ParseSpritesM4(int scanline)
 {
   struct PicoVideo *pv = &Pico.video;
   u8 *sat;
-  int xoff = 8; // relative to HighCol, which is (screen - 8)
+  int xoff = line_offset;
   int sprite_base, addr_mask;
   int zoomed = pv->reg[1] & 0x1; // zoomed sprites, e.g. Earthworm Jim
   unsigned int pack;
   int i, s, h, m;
 
   if (pv->reg[0] & 8)
-    xoff = 0;
-  xoff += line_offset;
+    xoff -= 8;  // sprite shift
   if ((Pico.m.hardware & (PMS_HW_GG|PMS_HW_LCD)) == (PMS_HW_GG|PMS_HW_LCD))
     xoff -= 48; // GG LCD, adjust to center 160 px
 
@@ -204,6 +203,9 @@ static void ParseSpritesM4(int scanline)
         // make sprite pixel map by merging the 4 bitplanes
         pack = ((pack | (pack>>16)) | ((pack | (pack>>16))>>8)) & 0xff;
         if (!m) m = CollisionDetect(sprites_map, sprites_x[s], pack, zoomed);
+        // no collision detection in 1st column if it's masked
+        if (pv->reg[0] & 0x20)
+          sprites_map[1] = 0;
       }
       s++;
     }
@@ -290,28 +292,26 @@ static void DrawDisplayM4(int scanline)
   if (scanline < 16 && (pv->reg[0] & 0x40))
     dx = 0; // hscroll disabled for top 2 rows (e.g. Fantasy Zone II)
 
-  tilex = ((-dx >> 3) + cellskip) & 0x1f;
+  tilex = (32 - (dx >> 3) + cellskip) & 0x1f;
   ty = (line & 7) << 1; // Y-Offset into tile
   cells = maxcells - cellskip;
 
-  dx = ((dx - 1) & 7) + 1;
-  if (dx != 8)
-    cells++; // have hscroll, need to draw 1 cell more
+  dx = (dx & 7);
   dx += cellskip << 3;
   dx += line_offset;
 
   // tiles
   if (!(pv->debug_p & PVD_KILL_B)) {
     if ((Pico.m.hardware & (PMS_HW_GG|PMS_HW_LCD)) == (PMS_HW_GG|PMS_HW_LCD)) {
-      // on GG render only the center 160 px
-      DrawStripM4(nametab , dx | ((cells-12)<< 16),(tilex+6) | (ty  << 16));
+      // on GG render only the center 160 px, but mind hscroll
+      DrawStripM4(nametab , (dx-8) | ((cells-11)<< 16),(tilex+5) | (ty  << 16));
     } else if (pv->reg[0] & 0x80) {
       // vscroll disabled for rightmost 8 columns (e.g. Gauntlet)
       int dx2 = dx + (cells-8)*8, tilex2 = tilex + (cells-8), ty2 = scanline&7;
-      DrawStripM4(nametab,  dx | ((cells-8) << 16), tilex  | (ty  << 16));
-      DrawStripM4(nametab2, dx2 |       (8  << 16), tilex2 | (ty2 << 17));
+      DrawStripM4(nametab,   dx    | ((cells-8) << 16), tilex    | (ty  << 16));
+      DrawStripM4(nametab2,  dx2   |        (8  << 16), tilex2   | (ty2 << 17));
     } else
-      DrawStripM4(nametab , dx | ( cells    << 16), tilex  | (ty  << 16));
+      DrawStripM4(nametab ,  dx    | ( cells    << 16), tilex    | (ty  << 16));
   }
 
   // sprites
@@ -320,9 +320,9 @@ static void DrawDisplayM4(int scanline)
 
   if ((pv->reg[0] & 0x20) && (Pico.m.hardware & (PMS_HW_GG|PMS_HW_LCD)) != (PMS_HW_GG|PMS_HW_LCD)) {
     // first column masked with background, caculate offset to start of line
-    dx = (dx&~0x1f) / 4;
+    dx = line_offset / 4;
     ty = ((pv->reg[7]&0x0f)|0x10) * 0x01010101;
-    ((u32 *)Pico.est.HighCol)[dx+2] = ((u32 *)Pico.est.HighCol)[dx+3] = ty;
+    ((u32 *)Pico.est.HighCol)[dx] = ((u32 *)Pico.est.HighCol)[dx+1] = ty;
   }
 }
 
@@ -432,12 +432,12 @@ static void ParseSpritesTMS(int scanline)
   struct PicoVideo *pv = &Pico.video;
   unsigned int pack;
   u8 *sat;
-  int xoff = 8; // relative to HighCol, which is (screen - 8)
+  int xoff;
   int sprite_base, addr_mask;
   int zoomed = pv->reg[1] & 0x1; // zoomed sprites
   int i, s, h, m;
 
-  xoff += line_offset;
+  xoff = line_offset;
 
   sat = (u8 *)PicoMem.vramb + ((pv->reg[5] & 0x7e) << 7);
   if (pv->reg[1] & 2) {
@@ -561,7 +561,7 @@ static void DrawDisplayM1(int scanline)
 
   tilex = cellskip & 0x1f;
   cells = maxcells - cellskip;
-  dx = (cellskip << 3) + line_offset + 8;
+  dx = (cellskip << 3) + line_offset;
 
   // tiles
   if (!(pv->debug_p & PVD_KILL_B))
@@ -607,7 +607,7 @@ static void DrawDisplayM2(int scanline)
 
   tilex = cellskip & 0x1f;
   cells = maxcells - cellskip;
-  dx = (cellskip << 3) + line_offset + 8;
+  dx = (cellskip << 3) + line_offset;
 
   // tiles
   if (!(pv->debug_p & PVD_KILL_B))
@@ -658,7 +658,7 @@ static void DrawDisplayM3(int scanline)
 
   tilex = cellskip & 0x1f;
   cells = maxcells - cellskip;
-  dx = (cellskip << 3) + line_offset + 8;
+  dx = (cellskip << 3) + line_offset;
 
   // tiles
   if (!(pv->debug_p & PVD_KILL_B))
@@ -708,7 +708,7 @@ static void DrawDisplayM0(int scanline)
 
   tilex = cellskip & 0x1f;
   cells = maxcells - cellskip;
-  dx = (cellskip << 3) + line_offset + 8;
+  dx = (cellskip << 3) + line_offset;
 
   // tiles
   if (!(pv->debug_p & PVD_KILL_B))
@@ -729,6 +729,7 @@ static void FinalizeLine8bitSMS(int line);
 void PicoFrameStartSMS(void)
 {
   int lines = 192, columns = 256, loffs, coffs;
+
   skip_next_line = 0;
   loffs = screen_offset = 24; // 192 lines is really 224 with top/bottom bars
   Pico.est.rendstatus = PDRAW_32_COLS;
@@ -741,9 +742,14 @@ void PicoFrameStartSMS(void)
 
   // Copy LCD enable flag for easier handling
   Pico.m.hardware &= ~PMS_HW_LCD;
-  if (PicoIn.opt & POPT_EN_GG_LCD)
+  if ((PicoIn.opt & POPT_EN_GG_LCD) && (Pico.m.hardware & PMS_HW_GG))
     Pico.m.hardware |= PMS_HW_LCD;
 
+  if (!(Pico.m.hardware & PMS_HW_LCD) && (mode & 4) && (Pico.video.reg[0] & 0x20)) {
+    // SMS mode 4 with 1st column blanked
+    columns = 248;
+    Pico.est.rendstatus |= PDRAW_SMS_BLANK_1;
+  }
   if ((Pico.m.hardware & (PMS_HW_GG|PMS_HW_LCD)) == (PMS_HW_GG|PMS_HW_LCD)) {
     // GG LCD always has 160x144 regardless of settings
     screen_offset = 24; // nonetheless the vdp timing has 224 lines
@@ -761,15 +767,20 @@ void PicoFrameStartSMS(void)
       lines = 224;
       break;
   }
-  if (PicoIn.opt & POPT_EN_SOFTSCALE) {
+
+  line_offset = 8; // FinalizeLine requires HighCol+8
+  // ugh... nonetheless has offset in 8-bit fast mode if 1st col blanked!
+  coffs = (FinalizeLineSMS == NULL && columns == 248 ? 8 : 0);
+  if (FinalizeLineSMS != NULL && (PicoIn.opt & POPT_EN_SOFTSCALE)) {
+    // softscaling always generates 320px, but no scaling in 8bit fast
     coffs = 0;
     columns = 320;
-  } else
-    coffs = PicoIn.opt & POPT_DIS_32C_BORDER ? 0:(320-columns)/2;
-  line_offset = (FinalizeLineSMS == NULL ? coffs : 0);
-
-  if (FinalizeLineSMS == FinalizeLineRGB555SMS)
-    line_offset = 0 /* done in FinalizeLine */;
+  } else if (!(PicoIn.opt & POPT_DIS_32C_BORDER)) {
+    line_offset -= coffs;
+    coffs = (320-columns) / 2;
+    if (FinalizeLineSMS == NULL)
+      line_offset += coffs; // ... else centering done in FinalizeLine
+  }
 
   if (Pico.est.rendstatus != rendstatus_old || lines != rendlines) {
     emu_video_mode_change(loffs, lines, coffs, columns);
@@ -836,13 +847,16 @@ norender:
 /* Palette for TMS9918 mode, see https://www.smspower.org/Development/Palette */
 // RGB values: #000000 #000000 #21c842 #5edc78 #5455ed #7d76fc #d4524d #42ebf5
 //             #fc5554 #ff7978 #d4c154 #e6ce80 #21b03b #c95bba #cccccc #ffffff
-static u16 tmspal[32] = {
+static u16 tmspal[] = {
   // SMS palette
   0x0000, 0x0000, 0x00a0, 0x00f0, 0x0a00, 0x0f00, 0x0005, 0x0ff0,
   0x000a, 0x000f, 0x00aa, 0x00ff, 0x0050, 0x0f0f, 0x0aaa, 0x0fff,
-  // GG palette
+  // TMS palette
   0x0000, 0x0000, 0x04c2, 0x07d6, 0x0e55, 0x0f77, 0x055c, 0x0ee4,
   0x055f, 0x077f, 0x05bc, 0x08ce, 0x03a2, 0x0b5c, 0x0ccc, 0x0fff,
+  // SMS palette, closer to the TMS one
+  0x0000, 0x0000, 0x05f0, 0x05f5, 0x0a50, 0x0f55, 0x055a, 0x0ff0,
+  0x055f, 0x0aaf, 0x05aa, 0x05af, 0x00a0, 0x0f5f, 0x0aaa, 0x0fff,
 };
 
 void PicoDoHighPal555SMS(void)
@@ -853,11 +867,11 @@ void PicoDoHighPal555SMS(void)
   unsigned int t;
   int i, j;
  
-  if (FinalizeLineSMS != FinalizeLine8bitSMS || Pico.m.dirtyPal == 2)
+  if (FinalizeLineSMS == FinalizeLineRGB555SMS || Pico.m.dirtyPal == 2)
     Pico.m.dirtyPal = 0;
 
-  // use hardware palette for 16bit accurate mode
-  if (FinalizeLineSMS == FinalizeLineRGB555SMS)
+  // use hardware palette if not in 8bit accurate mode
+  if (FinalizeLineSMS != FinalizeLine8bitSMS)
     spal = (void *)PicoMem.cram;
 
   /* SMS 6 bit cram data was already converted to MD/GG format by vdp write,
@@ -906,8 +920,7 @@ void PicoDrawSetOutputSMS(pdso_t which)
   {
     case PDF_8BIT:   FinalizeLineSMS = FinalizeLine8bitSMS; break;
     case PDF_RGB555: FinalizeLineSMS = FinalizeLineRGB555SMS; break;
-    // there's no fast renderer yet, just treat it like PDF_8BIT
-    default:         FinalizeLineSMS = FinalizeLine8bitSMS;
+    default:         FinalizeLineSMS = NULL; // no multiple palettes, no scaling
                      PicoDrawSetInternalBuf(Pico.est.Draw2FB, 328); break;
   }
   rendstatus_old = -1;
diff --git a/pico/pico.h b/pico/pico.h
index 9b8df316..3675f96f 100644
--- a/pico/pico.h
+++ b/pico/pico.h
@@ -227,6 +227,7 @@ void PicoDoHighPal555(int sh, int line, struct PicoEState *est);
 #define PDRAW_SKIP_FRAME   (1<<10) // frame is skipped
 #define PDRAW_30_ROWS      (1<<11) // 30 rows mode (240 lines)
 #define PDRAW_32X_SCALE    (1<<12) // scale CLUT layer for 32X
+#define PDRAW_SMS_BLANK_1  (1<<13) // 1st column blanked
 extern int rendstatus_old;
 extern int rendlines;
 
diff --git a/pico/sms.c b/pico/sms.c
index eb8da7e5..91d61e5d 100644
--- a/pico/sms.c
+++ b/pico/sms.c
@@ -461,7 +461,7 @@ static void write_bank_xor(unsigned short a, unsigned char d)
   if ((a&0x6000) != 0x2000) return;
   if (Pico.ms.mapper != PMS_MAP_XOR && Pico.ms.mapper) return;
 
-  elprintf(EL_Z80BNK, "bank 32k %04x %02x @ %04x", a, d, z80_pc());
+  elprintf(EL_Z80BNK, "bank xor %04x %02x @ %04x", a, d, z80_pc());
   Pico.ms.mapper = PMS_MAP_XOR;
 
   Pico.ms.carthw[0] = d;
@@ -544,7 +544,7 @@ static void xwrite(unsigned int a, unsigned char d)
   }
 }
 
-// TMR product codes and hardware type for know 50Hz-only games
+// TMR product codes and hardware type for known 50Hz-only games
 static u32 region_pal[] = { // cf. GX+, core/cart_hw/sms_cartc.c
   0x40207067 /* Addams Family */, 0x40207020 /* Back.Future 3 */,
   0x40207058 /* Battlemaniacs */, 0x40007105 /* Cal.Games 2 */,
diff --git a/pico/sound/resampler.c b/pico/sound/resampler.c
index ebf50914..5761bc54 100644
--- a/pico/sound/resampler.c
+++ b/pico/sound/resampler.c
@@ -46,6 +46,10 @@
 #include "../pico_types.h"
 #include "resampler.h"
 
+#ifndef M_PI
+#define M_PI            3.14159265358979323846
+#endif
+
 static double besseli0(double x)
 {
    unsigned i;
diff --git a/platform/common/arm_utils.h b/platform/common/arm_utils.h
index 628daf65..213186e5 100644
--- a/platform/common/arm_utils.h
+++ b/platform/common/arm_utils.h
@@ -2,8 +2,9 @@
 void bgr444_to_rgb32(void *to, void *from, unsigned entries);
 void bgr444_to_rgb32_sh(void *to, void *from);
 
-void vidcpy_m2(void *dest, void *src, int m32col, int with_32c_border);
-void vidcpy_m2_rot(void *dest, void *src, int m32col, int with_32c_border);
+void vidcpy_8bit(void *dest, void *src, int x_y, int w_h);
+void vidcpy_8bit_rot(void *dest, void *src, int x_y, int w_h);
+
 void spend_cycles(int c); // utility
 
 void rotated_blit8 (void *dst, void *linesx4, int y, int is_32col);
diff --git a/platform/common/arm_utils.s b/platform/common/arm_utils.s
index b4f739d9..68ef3e41 100644
--- a/platform/common/arm_utils.s
+++ b/platform/common/arm_utils.s
@@ -124,72 +124,138 @@ bgr444_to_rgb32_sh:
 
 @ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
 
-
-@ mode2 blitter
-.global vidcpy_m2 @ void *dest, void *src, int m32col, int with_32c_border
-vidcpy_m2:
+.global vidcpy_8bit @ void *dest, void *src, int x_y, int w_h
+vidcpy_8bit:
     stmfd   sp!, {r4-r6,lr}
 
-    mov     r12, #224       @ lines
-    add     r0, r0, #320*8
+    mov     r12, r2, lsl #16    @ y
+
+    mov     r4, r12, lsr #16-8  @ 320*y = 256*y+64*y
+    add     r4, r4, r12, lsr #16-6
+    add     r0, r0, r4          @ pd += 320*y + x
+    add     r0, r0, r2, lsr #16
+
+    add     r4, r4, r12, lsr #16-3 @ 328*y = 320*y + 8*y
+    add     r1, r1, r4          @ ps += 328*y + x + 8
+    add     r1, r1, r2, lsr #16
     add     r1, r1, #8
-    mov     lr, #0
 
-    tst     r2, r2
-    movne   lr, #64
-    tstne   r3, r3
-    addne   r0, r0, #32
-    addne   r1, r1, #32
+    mov     lr, r3, lsr #16     @ w
+    mov     r12, r3, lsl #16    @ h
 
-vidCpyM2_loop_out:
-    mov     r6, #10
-    sub     r6, r6, lr, lsr #5	@ -= 2 in 32col mode
-vidCpyM2_loop:
+vidCpy8bit_loop_out:
+    lsrs    r6, lr, #5
+@   beq     vidCpy8bit_loop_end
+vidCpy8bit_loop:
     subs    r6, r6, #1
     ldmia   r1!, {r2-r5}
     stmia   r0!, {r2-r5}
     ldmia   r1!, {r2-r5}
     stmia   r0!, {r2-r5}
-    bne     vidCpyM2_loop
-
-    subs    r12,r12,#1
-    add     r0, r0, lr
-    add     r1, r1, #8
-    add     r1, r1, lr
-    bne     vidCpyM2_loop_out
+    bne     vidCpy8bit_loop
+
+    ands    r6, lr, #0x0018
+    beq     vidCpy8bit_loop_end
+vidCpy8bit_loop2:
+    ldmia   r1!, {r2-r3}
+    subs    r6, r6, #8
+    stmia   r0!, {r2-r3}
+    bne     vidCpy8bit_loop2
+
+vidCpy8bit_loop_end:
+    subs    r12,r12,#1<<16
+    add     r0, r0, #320
+    sub     r0, r0, lr
+    add     r1, r1, #328
+    sub     r1, r1, lr
+    bne     vidCpy8bit_loop_out
 
     ldmfd   sp!, {r4-r6,pc}
 
 
-.global vidcpy_m2_rot @ void *dest, void *src, int m32col, int with_32c_border
-vidcpy_m2_rot:
-    stmfd   sp!,{r4-r8,lr}
+.global vidcpy_8bit_rot @ void *dest, void *src, int x_y, int w_h
+vidcpy_8bit_rot:
+    stmfd   sp!, {r4-r10,lr}
+
+    mov     r12, r2, lsl #16    @ y
+
+    add     r0, r0, r12, lsr #16 @ pd += y + (319-x)*240
+    mov     r4, #320
+    sub     r4, r4, #1
+    sub     r4, r4, r2, lsr #16 @    (319-x)
+    add     r0, r0, r4, lsl #8
+    sub     r0, r0, r4, lsl #4
+
+    mov     r4, r12, lsr #16-8  @ 328*y = 256*y + 64*y + 8*y
+    add     r4, r4, r12, lsr #16-6
+    add     r4, r4, r12, lsr #16-3
+    add     r1, r1, r4          @ ps += 328*y + x + 8
+    add     r1, r1, r2, lsr #16
     add     r1, r1, #8
-    tst     r2, r2
-    subne   r1, r1, #32		@ adjust
-
-    mov     r4, r0
-    mov     r5, r1
-    mov     r6, r2
-    mov     r7, #8+4
-
-vidcpy_m2_rot_loop:
-    @ a bit lame but oh well..
-    mov     r0, r4
-    mov     r1, r5
-    mov     r2, r7
-    mov     r3, r6
+
+    mov     lr, r3, lsr #16     @ w
+    mov     r12, r3, lsl #16    @ h
+
     mov     r8, #328
-    adr     lr, after_rot_blit8
-    stmfd   sp!,{r4-r8,lr}
-    b       rotated_blit8_2
-
-after_rot_blit8:
-    add     r5, r5, #328*4
-    add     r7, r7, #4
-    cmp     r7, #224+8+4
-    ldmgefd sp!,{r4-r8,pc}
-    b       vidcpy_m2_rot_loop
+vidCpy8bitrot_loop_out:
+    mov     r10, r0
+    lsrs    r9, lr, #2
+@   beq     vidCpy8bitrot_loop_end
+vidCpy8bitrot_loop:
+    mov     r6, r1
+    ldr     r2, [r6], r8
+    ldr     r3, [r6], r8
+    ldr     r4, [r6], r8
+    ldr     r5, [r6], r8
+
+    mov     r6, r2, lsl #24
+    mov     r6, r6, lsr #8
+    orr     r6, r6, r3, lsl #24
+    mov     r6, r6, lsr #8
+    orr     r6, r6, r4, lsl #24
+    mov     r6, r6, lsr #8
+    orr     r6, r6, r5, lsl #24
+    str     r6, [r0], #-240
+
+    and     r6, r3, #0xff00
+    and     r7, r2, #0xff00
+    orr     r6, r6, r7, lsr #8
+    and     r7, r4, #0xff00
+    orr     r6, r6, r7, lsl #8
+    and     r7, r5, #0xff00
+    orr     r6, r6, r7, lsl #16
+    str     r6, [r0], #-240
+
+    and     r6, r4, #0xff0000
+    and     r7, r2, #0xff0000
+    orr     r6, r6, r7, lsr #16
+    and     r7, r3, #0xff0000
+    orr     r6, r6, r7, lsr #8
+    and     r7, r5, #0xff0000
+    orr     r6, r6, r7, lsl #8
+    str     r6, [r0], #-240
+
+    mov     r6, r5, lsr #24
+    mov     r6, r6, lsl #8
+    orr     r6, r6, r4, lsr #24
+    mov     r6, r6, lsl #8
+    orr     r6, r6, r3, lsr #24
+    mov     r6, r6, lsl #8
+    orr     r6, r6, r2, lsr #24
+    str     r6, [r0], #-240
+
+    subs    r9, r9, #1
+    add     r1, r1, #4
+    bne     vidCpy8bitrot_loop
+
+vidCpy8bitrot_loop_end:
+    subs    r12,r12,#4<<16
+    add     r0, r10, #4
+    sub     r1, r1, lr
+    add     r1, r1, #4*328
+    bne     vidCpy8bitrot_loop_out
+
+    ldmfd   sp!, {r4-r10,pc}
 
 
 .global rotated_blit8 @ void *dst, void *linesx4, u32 y, int is_32col
diff --git a/platform/common/menu_pico.c b/platform/common/menu_pico.c
index a72953f1..39c7fafe 100644
--- a/platform/common/menu_pico.c
+++ b/platform/common/menu_pico.c
@@ -146,12 +146,6 @@ static void copy_bg(int dir)
 	}
 }
 
-static void menu_enter(int is_rom_loaded)
-{
-	plat_video_menu_enter(is_rom_loaded);
-	menu_w = menu_h = 0;
-}
-
 static void menu_draw_prep(void)
 {
 	if (menu_w == g_menuscreen_w && menu_h == g_menuscreen_h)
@@ -198,6 +192,13 @@ static void draw_savestate_bg(int slot)
 	PicoTmpStateRestore(tmp_state);
 }
 
+static void menu_enter(int is_rom_loaded)
+{
+	plat_video_menu_enter(is_rom_loaded);
+	menu_w = menu_h = 0;
+	menu_draw_prep();
+}
+
 // --------- loading ROM screen ----------
 
 static int cdload_called = 0;
diff --git a/platform/gp2x/emu.c b/platform/gp2x/emu.c
index 92ea2ec6..c6a708f4 100644
--- a/platform/gp2x/emu.c
+++ b/platform/gp2x/emu.c
@@ -85,8 +85,6 @@ static int get_renderer(void)
 {
 	if (doing_bg_frame)
 		return RT_16BIT;
-	if ((PicoIn.AHW & PAHW_SMS) && currentConfig.renderer == RT_8BIT_FAST)
-		return RT_8BIT_ACC;	// 8bpp fast is not there (yet?)
 	if (PicoIn.AHW & PAHW_32X)
 		return currentConfig.renderer32x;
 	else
@@ -217,6 +215,26 @@ static void draw_pico_ptr(void)
 	p[pitch*2] ^= 0xffff;
 }
 
+static void clear_1st_column(int firstcol, int firstline, int linecount)
+{
+	int size = is_16bit_mode() ? 2 : 1;
+	int black = is_16bit_mode() ? 0 : 0xe0;
+	int i;
+
+	// SMS 1st column blanked, replace with black
+	if ((currentConfig.EmuOpt & EOPT_WIZ_TEAR_FIX) && !doing_bg_frame) {
+		int pitch = 240*size;
+		char *p = (char *)g_screen_ptr + (319-(firstcol-8))*pitch;
+		for (i = 0; i < 8; i++, p -= pitch)
+			memset(p+(firstline)*size, black, linecount*size);
+	} else {
+		int pitch = 320*size;
+		char *p = (char *)g_screen_ptr + (firstline)*pitch;
+		for (i = 0; i < linecount; i++, p += pitch)
+			memset(p+(firstcol-8)*size, black, 8*size);
+	}
+}
+
 /* rot thing for Wiz */
 static unsigned char __attribute__((aligned(4))) rot_buff[320*4*2];
 
@@ -246,7 +264,7 @@ static int EmuScanEnd8_rot(unsigned int num)
 	if ((num & 3) != 3)
 		return 0;
 	rotated_blit8(g_screen_ptr, rot_buff, num + 1,
-		!(Pico.video.reg[12] & 1));
+		!(Pico.video.reg[12] & 1) && !(PicoIn.opt & POPT_EN_SOFTSCALE));
 	return 0;
 }
 
@@ -296,7 +314,7 @@ static int EmuScanEnd16_ld(unsigned int num)
 static int localPal[0x100];
 static int localPalSize;
 
-static void (*vidcpyM2)(void *dest, void *src, int m32col, int with_32c_border);
+static void (*vidcpy8bit)(void *dest, void *src, int x_y, int w_h);
 static int (*make_local_pal)(int fast_mode);
 
 static int make_local_pal_md(int fast_mode)
@@ -351,6 +369,11 @@ static int make_local_pal_sms(int fast_mode)
 			bgr444_to_rgb32(localPal+i*0x40, tmspal, 32);
 			memcpy(localPal+i*0x40+0x20, localPal+i*0x40, 0x20*4);
 		}
+	} else if (fast_mode) {
+		for (i = 0;i >= 0; i--) {
+			bgr444_to_rgb32(localPal+i*0x40, PicoMem.cram+i*0x40, 32);
+			memcpy(localPal+i*0x40+0x20, localPal+i*0x40, 0x20*4);
+		}
 	} else {
 		for (i = Pico.est.SonicPalCount; i >= 0; i--) {
 			bgr444_to_rgb32(localPal+i*0x40, Pico.est.SonicPal+i*0x40, 32);
@@ -362,11 +385,15 @@ static int make_local_pal_sms(int fast_mode)
 	return (Pico.est.SonicPalCount+1)*0x40;
 }
 
+static int is_1stblanked;
+static int firstline, linecount;
+static int firstcol, colcount;
+
 void pemu_finalize_frame(const char *fps, const char *notice)
 {
 	int emu_opt = currentConfig.EmuOpt;
 
-	if (PicoIn.AHW & PAHW_32X)
+	if (is_16bit_mode())
 		localPalSize = 0; // nothing to do
 	else if (get_renderer() == RT_8BIT_FAST)
 	{
@@ -377,13 +404,14 @@ void pemu_finalize_frame(const char *fps, const char *notice)
 		if (PicoIn.AHW & PAHW_SVP)
 			memset32((int *)(Pico.est.Draw2FB+328*8+328*223), 0xe0e0e0e0, 328/4);
 		// clear top and bottom of overlap trash
-		if (!(Pico.est.rendstatus & PDRAW_30_ROWS)) {
-			memset32((int *)(Pico.est.Draw2FB+8*(224+8)), 0xe0e0e0e0, 328*8/4);
-			memset32((int *)(Pico.est.Draw2FB), 0xe0e0e0e0, 328*8/4);
+		if (firstline >= 8) {
+			unsigned char *p = Pico.est.Draw2FB + firstline*328;
+			memset32((int *)(p-        8*328), 0xe0e0e0e0, 328*8/4);
+			memset32((int *)(p+linecount*328), 0xe0e0e0e0, 328*8/4);
 		}
 		// do actual copy
-		vidcpyM2(g_screen_ptr, Pico.est.Draw2FB+328*8,
-			!(Pico.video.reg[12] & 1), !(PicoIn.opt & POPT_DIS_32C_BORDER));
+		vidcpy8bit(g_screen_ptr, Pico.est.Draw2FB,
+			(firstcol << 16) | firstline, (colcount << 16) | linecount);
 	}
 	else if (get_renderer() == RT_8BIT_ACC)
 	{
@@ -391,7 +419,9 @@ void pemu_finalize_frame(const char *fps, const char *notice)
 		if (Pico.m.dirtyPal)
 			localPalSize = make_local_pal(0);
 	}
-	else	localPalSize = 0; // no palette in 16bit mode
+
+	if (is_1stblanked)
+		clear_1st_column(firstcol, firstline, linecount);
 
 	if (notice)
 		osd_text(4, osd_y, notice);
@@ -522,7 +552,7 @@ static void vid_reset_mode(void)
 	case RT_8BIT_FAST:
 		PicoIn.opt |= POPT_ALT_RENDERER;
 		PicoDrawSetOutFormat(PDF_NONE, 0);
-		vidcpyM2 = vidcpy_m2;
+		vidcpy8bit = vidcpy_8bit;
 		gp2x_mode = 8;
 		break;
 	default:
@@ -546,13 +576,15 @@ static void vid_reset_mode(void)
 		if ((PicoIn.AHW & PAHW_32X) || renderer == RT_16BIT) {
 			emu_scan_begin = EmuScanBegin16_rot;
 			emu_scan_end = EmuScanEnd16_rot;
+			memset(rot_buff, 0, 320*4*2);
 		}
 		else if (renderer == RT_8BIT_ACC) {
 			emu_scan_begin = EmuScanBegin8_rot;
 			emu_scan_end = EmuScanEnd8_rot;
+			memset(rot_buff, 0xe0, 320*4);
 		}
 		else if (renderer == RT_8BIT_FAST)
-			vidcpyM2 = vidcpy_m2_rot;
+			vidcpy8bit = vidcpy_8bit_rot;
 	}
 
 	PicoDrawSetCallbacks(emu_scan_begin, emu_scan_end);
@@ -588,18 +620,26 @@ void emu_video_mode_change(int start_line, int line_count, int start_col, int co
 	int scalex = 320, scaley = 240;
 	int ln_offs = 0;
 
-	/* NTSC always has 224 visible lines, anything smaller has bars */
-	if (line_count < 224 && line_count > 144) {
-		start_line -= (224-line_count) /2;
-		line_count = 224;
-	}
+	if (currentConfig.vscaling != EOPT_SCALE_NONE &&
+			(is_16bit_mode() || get_renderer() != RT_8BIT_FAST)) {
+		/* NTSC always has 224 visible lines, anything smaller has bars */
+		if (line_count < 224 && line_count > 144) {
+			start_line -= (224-line_count) /2;
+			line_count = 224;
+		}
 
-	/* line doubling for swscaling, also needed for bg frames */
-	if (currentConfig.vscaling == EOPT_SCALE_SW && line_count < 240) {
-		ld_lines = ld_left = 2*line_count / (240 - line_count);
-		PicoDrawSetCallbacks(EmuScanBegin16_ld, EmuScanEnd16_ld);
+		/* line doubling for swscaling, also needed for bg frames */
+		if (currentConfig.vscaling == EOPT_SCALE_SW && line_count < 240) {
+			ld_lines = ld_left = 2*line_count / (240 - line_count);
+			PicoDrawSetCallbacks(EmuScanBegin16_ld,EmuScanEnd16_ld);
+		}
 	}
 
+	/* blanking for SMS with 1st tile blanked */
+	is_1stblanked = (col_count == 248);
+	firstline = start_line; linecount = line_count;
+	firstcol = start_col; colcount = col_count;
+
 	if (doing_bg_frame)
 		return;
 
@@ -754,11 +794,16 @@ void pemu_forced_frame(int no_scale, int do_emu)
 	Pico.m.dirtyPal = 1;
 	PicoIn.opt &= ~POPT_DIS_32C_BORDER;
 	gp2x_current_bpp = 16;
+	// always render in screen 3 since menu uses 0-2
+	g_screen_ptr = gp2x_screens[3];
 
 	if (!no_scale)
 		no_scale = currentConfig.scaling == EOPT_SCALE_NONE;
 	emu_cmn_forced_frame(no_scale, do_emu, g_screen_ptr);
 
+	if (is_1stblanked)
+		clear_1st_column(firstcol, firstline, linecount);
+
 	g_menubg_src_ptr = g_screen_ptr;
 	doing_bg_frame = 0;
 }
@@ -791,6 +836,12 @@ void pemu_loop_end(void)
 {
 	pemu_sound_stop();
 
+	if (g_screen_ptr == gp2x_screens[0]) {
+		/* currently on screen 3, which is needed for forced_frame */
+		int size = gp2x_current_bpp / 8;
+		gp2x_memcpy_all_buffers(g_screen_ptr, 0, 320*240 * size);
+		gp2x_video_flip();
+	}
 	/* do one more frame for menu bg */
 	pemu_forced_frame(0, 1);
 }
diff --git a/platform/libretro/libretro.c b/platform/libretro/libretro.c
index b213b6af..f20cf861 100644
--- a/platform/libretro/libretro.c
+++ b/platform/libretro/libretro.c
@@ -535,7 +535,7 @@ void emu_video_mode_change(int start_line, int line_count, int start_col, int co
    // calculate the borders of the real image inside the picodrive image
    vout_width = (vout_16bit ? VOUT_MAX_WIDTH : VOUT_8BIT_WIDTH);
    vout_height = (vout_16bit ? VOUT_MAX_HEIGHT : VOUT_8BIT_HEIGHT);
-   vout_offset = (vout_16bit ? 0 : 8); // 8bit has 8 px overlap area on the left
+   vout_offset = (vout_16bit ? 0 : col_count == 248 ? 16 : 8); // 8bit has overlap area on the left
    padding = (struct retro_hw_ps2_insets){start_line, vout_offset, vout_height - line_count - start_line, vout_width - col_count - vout_offset};
 
    int pxsz = (vout_16bit ? 2 : 1); // pixel size: RGB = 16 bits, CLUT = 8 bits
@@ -2247,6 +2247,9 @@ void retro_run(void)
       int x;
       if (Pico.m.dirtyPal)
          PicoDrawUpdateHighPal();
+      /* 8 bit renderers have an extra offset for SMS wíth 1st tile blanked */
+      if (vout_width == 248)
+         ps += 8;
       /* Copy, and skip the leftmost 8 columns again */
       for (i = 0; i < vout_height; i++, ps += 8) {
          for (x = 0; x < vout_width; x+=4) {
diff --git a/platform/linux/blit.c b/platform/linux/blit.c
index b70262eb..ee2b6d0e 100644
--- a/platform/linux/blit.c
+++ b/platform/linux/blit.c
@@ -6,6 +6,8 @@
  * See COPYING file in the top-level directory.
  */
 
+#include <string.h>
+
 // Convert 0000bbb0 ggg0rrr0 0000bbb0 ggg0rrr0
 // to      00000000 rrr00000 ggg00000 bbb00000 ...
 // TODO: rm when gp2x/emu.c is no longer used
@@ -52,40 +54,88 @@ void bgr444_to_rgb32_sh(void *to, void *from)
 	}
 }
 
-void vidcpy_m2(void *dest, void *src, int m32col, int with_32c_border)
+#define X (x_y >> 16)
+#define Y (x_y & 0xffff)
+#define W (w_h >> 16)
+#define H (w_h & 0xffff)
+
+// gp2x:   0-> X    wiz: Y <-0
+//         |                 |
+//         v                 v
+//
+//         Y                 X
+
+void vidcpy_8bit(void *dest, void *src, int x_y, int w_h)
 {
 	unsigned char *pd = dest, *ps = src;
-	int i, u;
+	int i;
 
-	if (m32col) {
-		for (i = 0; i < 224; i++)
-		{
-			ps += 8;
-			ps += 32;
-			pd += 32;
-			for (u = 0; u < 256; u++)
-				*pd++ = *ps++;
-			ps += 32;
-			pd += 32;
-		}
-	} else {
-		for (i = 0; i < 224; i++)
-		{
-			ps += 8;
-			for (u = 0; u < 320; u++)
-				*pd++ = *ps++;
-		}
+	pd += X + Y*320;
+	ps += X + Y*328 + 8;
+	for (i = 0; i < H; i++) {
+		memcpy(pd, ps, W);
+		ps += 328; pd += 320;
 	}
 }
 
-void vidcpy_m2_rot(void *dest, void *src, int m32col, int with_32c_border)
+void vidcpy_8bit_rot(void *dest, void *src, int x_y, int w_h)
 {
+	unsigned char *pd = dest, *ps = src;
+	int i, u;
+
+	pd += Y + (319-X)*240;
+	ps += X + Y*328 + 8;
+	for (i = 0; i < H; i += 4) {
+		unsigned char *p = (void *)ps;
+		unsigned int  *q = (void *)pd;
+		for (u = 0; u < W; u++) {
+			*q = (p[3*328]<<24) + (p[2*328]<<16) + (p[1*328]<<8) + p[0*328];
+			p += 1;
+			q -= 240/4;
+		}
+		ps += 4*328; pd += 4;
+	}
 }
 
 void rotated_blit8 (void *dst, void *linesx4, int y, int is_32col)
 {
+	unsigned char *pd = dst, *ps = linesx4;
+	int x, w, u;
+
+	x = (is_32col ? 32 : 0);
+	w = (is_32col ? 256 : 320);
+	y -= 4;
+
+	pd += y + (319-x)*240;
+	ps += x;
+
+	unsigned char *p = (void *)ps;
+	unsigned int  *q = (void *)pd;
+	for (u = 0; u < w; u++) {
+		*q = (p[3*328]<<24) + (p[2*328]<<16) + (p[1*328]<<8) + p[0*328];
+		p += 1;
+		q -= 240/4;
+	}
 }
 
 void rotated_blit16(void *dst, void *linesx4, int y, int is_32col)
 {
+	unsigned short *pd = dst, *ps = linesx4;
+	int x, w, u;
+
+	x = (is_32col ? 32 : 0);
+	w = (is_32col ? 256 : 320);
+	y -= 4;
+
+	pd += y + (319-x)*240;
+	ps += x;
+
+	unsigned short *p = (void *)ps;
+	unsigned int   *q = (void *)pd;
+	for (u = 0; u < w; u++) {
+		q[0] = (p[1*328]<<16) + p[0*328];
+		q[1] = (p[3*328]<<16) + p[2*328];
+		p += 1;
+		q -= 2*240/4;
+	}
 }
diff --git a/platform/linux/emu.c b/platform/linux/emu.c
index 0a05cdd8..ee796f57 100644
--- a/platform/linux/emu.c
+++ b/platform/linux/emu.c
@@ -138,13 +138,13 @@ void screen_blit(u16 *pd, int pp, u8* ps, int ss, u16 *pal)
 	if (currentConfig.scaling == EOPT_SCALE_SW && out_w <= 256) {
 	    if (currentConfig.vscaling == EOPT_SCALE_SW && out_h <= 224)
 		// h+v scaling
-		upscale = out_w >= 256 ? upscale_256_224_hv: upscale_160_144_hv;
+		upscale = out_w >= 240 ? upscale_256_224_hv: upscale_160_144_hv;
 	    else
 		// h scaling
-		upscale = out_w >= 256 ? upscale_256_____h : upscale_160_____h;
+		upscale = out_w >= 240 ? upscale_256_____h : upscale_160_____h;
 	} else if (currentConfig.vscaling == EOPT_SCALE_SW && out_h <= 224)
 		// v scaling
-		upscale = out_w >= 256 ? upscale_____224_v : upscale_____144_v;
+		upscale = out_w >= 240 ? upscale_____224_v : upscale_____144_v;
 	if (!upscale) {
 		// no scaling
 		for (y = 0; y < out_h; y++)
@@ -165,6 +165,8 @@ void pemu_finalize_frame(const char *fps, const char *notice)
 
 		PicoDrawUpdateHighPal();
 
+		if (out_w == 248 && currentConfig.scaling == EOPT_SCALE_SW)
+			pd += (320 - out_w*320/256) / 2; // SMS with 1st tile blanked, recenter
 		screen_blit(pd, g_screen_ppitch, ps, 328, Pico.est.HighPal);
 	}
 
@@ -205,16 +207,13 @@ void plat_video_set_buffer(void *buf)
 
 static void apply_renderer(void)
 {
-	PicoIn.opt &= ~(POPT_ALT_RENDERER|POPT_EN_SOFTSCALE|POPT_DIS_32C_BORDER);
+	PicoIn.opt |= POPT_DIS_32C_BORDER;
+	PicoIn.opt &= ~(POPT_ALT_RENDERER|POPT_EN_SOFTSCALE);
 	if (is_16bit_mode()) {
 		if (currentConfig.scaling == EOPT_SCALE_SW)
 			PicoIn.opt |= POPT_EN_SOFTSCALE;
-		else if (currentConfig.scaling == EOPT_SCALE_HW)
-			// hw scaling, render without any padding
-			PicoIn.opt |= POPT_DIS_32C_BORDER;
 		PicoIn.filter = currentConfig.filter;
-	} else
-		PicoIn.opt |= POPT_DIS_32C_BORDER;
+	}
 
 	switch (get_renderer()) {
 	case RT_16BIT:
@@ -374,8 +373,9 @@ void emu_video_mode_change(int start_line, int line_count, int start_col, int co
 
 	switch (currentConfig.scaling) {
 	case EOPT_SCALE_HW:
-		screen_w = out_w;
-		screen_x = 0;
+		// mind aspect ratio for SMS with 1st column blanked
+		screen_w = (out_w == 248 ? 256 : out_w);
+		screen_x = (screen_w - out_w)/2;
 		break;
 	case EOPT_SCALE_SW:
 		screen_x = (screen_w - 320)/2;
diff --git a/platform/psp/emu.c b/platform/psp/emu.c
index 6747b07d..a76d5839 100644
--- a/platform/psp/emu.c
+++ b/platform/psp/emu.c
@@ -189,7 +189,8 @@ static void set_scaling_params(void)
 	g_vertices[1].y = fbimg_yoffs + fbimg_height;
 	if (!is_16bit_mode()) {
 		// 8-bit modes have an 8 px overlap area on the left
-		g_vertices[0].u += 8; g_vertices[1].u += 8;
+		int offs = out_w == 248 ? 16 : 8;
+		g_vertices[0].u += offs; g_vertices[1].u += offs;
 	}
 	if (border_hack) {
 		g_vertices[0].u++;    g_vertices[1].u--;
@@ -210,12 +211,16 @@ static void do_pal_update_sms(void)
 		// SMS palette
 		0x0000, 0x0000, 0x00a0, 0x00f0, 0x0500, 0x0f00, 0x0005, 0x0ff0,
 		0x000a, 0x000f, 0x0055, 0x00ff, 0x0050, 0x0f0f, 0x0555, 0x0fff,
+		// TMS palette
+		0x0000, 0x0000, 0x04c2, 0x07d6, 0x0e55, 0x0f77, 0x055c, 0x0ee4,
+		0x055f, 0x077f, 0x05bc, 0x08ce, 0x03a2, 0x0b5c, 0x0ccc, 0x0fff,
 	};
 	int i;
 	
 	if (!(Pico.video.reg[0] & 0x4)) {
+		int sg = !!(Pico.m.hardware & PMS_HW_SG);
 		for (i = Pico.est.SonicPalCount; i >= 0; i--)
-			do_pal_convert(localPal+i*0x40, tmspal, currentConfig.gamma, currentConfig.gamma2);
+			do_pal_convert(localPal+i*0x40, tmspal+sg*0x10, currentConfig.gamma, currentConfig.gamma2);
 	} else {
 		for (i = Pico.est.SonicPalCount; i >= 0; i--)
 			do_pal_convert(localPal+i*0x40, Pico.est.SonicPal+i*0x40, currentConfig.gamma, currentConfig.gamma2);
diff --git a/platform/psp/menu.c b/platform/psp/menu.c
index 42081edc..81356b29 100644
--- a/platform/psp/menu.c
+++ b/platform/psp/menu.c
@@ -13,6 +13,8 @@ static const char *men_filter_opts[] = { "nearest", "bilinear" };
 
 #define MENU_OPTIONS_ADV
 
+static menu_entry e_menu_sms_options[];
+
 void psp_menu_init(void)
 {
 	me_enable(e_menu_sms_options, MA_SMSOPT_GHOSTING, 0);