diff --git a/f3dex3.s b/f3dex3.s
index 536529a..5a5b1d7 100644
--- a/f3dex3.s
+++ b/f3dex3.s
@@ -516,6 +516,7 @@ jumpTableEntry G_MOVEMEM_end  // G_MOVEMEM, G_MTX (load)
 
 // RDP/Immediate Command Mini Table
 // 1 byte per entry, after << 2 points to an addr in first 1/4 of IMEM
+miniTableEntry G_FLUSH_handler
 miniTableEntry G_MEMSET_handler
 miniTableEntry G_DMA_IO_handler
 miniTableEntry G_TEXTURE_handler
@@ -1726,7 +1727,7 @@ tri_return_from_decal_fix_z:
     bltz    $8, return_and_end_mat      // Return if rdpCmdBufPtr < end+1 i.e. ptr <= end
      slv    $v10[12], 0x00($10)   // ZI:F
      // 156 cycles
-flush_rdp_buffer: // $8 = rdpCmdBufPtr - rdpCmdBufEndP1
+flush_rdp_buffer: // Prereq: $8 = rdpCmdBufPtr - rdpCmdBufEndP1
     mfc0    $11, SP_DMA_BUSY                 // Check if any DMA is in flight
     lw      cmd_w1_dram, rdpFifoPos          // FIFO pointer = end of RDP read, start of RSP write
     lw      $10, OSTask + OSTask_output_buff_size // Load FIFO "size" (actually end addr)
diff --git a/gbi.h b/gbi.h
index 7b722f1..27d710f 100644
--- a/gbi.h
+++ b/gbi.h
@@ -55,7 +55,8 @@ of warnings if you use -Wpedantic. */
  */
 /*#define G_SPECIAL_3       0xD3  no-op in F3DEX2 */
 /*#define G_SPECIAL_2       0xD4  no-op in F3DEX2 */
-/*#define G_SPECIAL_1       0xD5  triggered MVP recalculation, not supported in F3DEX3 */
+/*#define G_SPECIAL_1       0xD5  triggered MVP recalculation in F3DEX2 for debug */
+#define G_FLUSH             0xD4
 #define G_MEMSET            0xD5
 #define G_DMA_IO            0xD6
 #define G_TEXTURE           0xD7
@@ -2463,22 +2464,31 @@ _DW({                                                   \
     gsDma0p(G_MEMSET, (dram), ((size) & 0xFFFFF0))
 
 /**
- * RSP short command (no DMA required) macros
+ * Flush the internal DMEM buffer of RDP commands to the RDP FIFO in DRAM,
+ * causing the RDP to immediately begin executing any previous commands.
+ * Without SPFlush, the RDP may not begin executing any given command until up
+ * to 46 more RDP commands after that have been processed by the RSP (or the
+ * final end of the display list for the frame).
+ * 
+ * The primary use case is if your frame's display list begins with clearing
+ * the framebuffer and/or Z buffer, and then proceeds to things which take
+ * significant time on the RSP before emitting many RDP commands, such as
+ * matrix and lighting for drawing a character model. You should insert SPFlush
+ * after the first large buffer clear to cause the RDP to begin executing those
+ * long operations immediately while the RSP is continuing to work. If you are
+ * clearing both the framebuffer and Z buffer, you would usually only need one
+ * SPFlush after the first of these two DPFillRect commands.
  */
-#define gImmp0(pkt, c)                  \
-_DW({                                   \
-    Gfx *_g = (Gfx *)(pkt);             \
-                                        \
-    _g->words.w0 = _SHIFTL((c), 24, 8); \
-})
+#define gSPFlush(pkt)   g1Word(pkt, G_FLUSH, 0)
 
 /**
- * @copydetails gImmp0
+ * @copydetails gSPFlush
+ */
+#define gsSPFlush()    gs1Word(     G_FLUSH, 0)
+
+/*
+ * RSP short command (no DMA required) macros
  */
-#define gsImmp0(c)      \
-{                       \
-    _SHIFTL((c), 24, 8) \
-}
 
 #define gImmp1(pkt, c, p0)              \
 _DW({                                   \
@@ -2494,58 +2504,6 @@ _DW({                                   \
     (unsigned int)(p0)      \
 }
 
-#define gImmp2(pkt, c, p0, p1)              \
-_DW({                                       \
-    Gfx *_g = (Gfx *)(pkt);                 \
-                                            \
-    _g->words.w0 = _SHIFTL((c), 24, 8);     \
-    _g->words.w1 = (_SHIFTL((p0), 16, 16) | \
-                    _SHIFTL((p1),  8,  8)); \
-})
-
-#define gsImmp2(c, p0, p1)  \
-{                           \
-    _SHIFTL((c), 24, 8),    \
-   (_SHIFTL((p0), 16, 16) | \
-    _SHIFTL((p1),  8,  8))  \
-}
-
-#define gImmp3(pkt, c, p0, p1, p2)          \
-_DW({                                       \
-    Gfx *_g = (Gfx *)(pkt);                 \
-                                            \
-    _g->words.w0 = _SHIFTL((c), 24, 8);     \
-    _g->words.w1 = (_SHIFTL((p0), 16, 16) | \
-                    _SHIFTL((p1),  8,  8) | \
-                    _SHIFTL((p2),  0,  8)); \
-})
-
-#define gsImmp3(c, p0, p1, p2)  \
-{                               \
-    _SHIFTL((c), 24, 8),        \
-   (_SHIFTL((p0), 16, 16) |     \
-    _SHIFTL((p1),  8,  8) |     \
-    _SHIFTL((p2),  0,  8))      \
-}
-
-#define gImmp21(pkt, c, p0, p1, dat)        \
-_DW({                                       \
-    Gfx *_g = (Gfx *)(pkt);                 \
-                                            \
-    _g->words.w0 = (_SHIFTL((c),  24,  8) | \
-                    _SHIFTL((p0),  8, 16) | \
-                    _SHIFTL((p1),  0,  8)); \
-    _g->words.w1 = (unsigned int) (dat);    \
-})
-
-#define gsImmp21(c, p0, p1, dat)    \
-{                                   \
-   (_SHIFTL((c),  24,  8) |         \
-    _SHIFTL((p0),  8, 16) |         \
-    _SHIFTL((p1),  0,  8)),         \
-    (unsigned int) (dat)            \
-}
-
 #define gMoveWd(pkt, index, offset, data) \
     gDma1p((pkt), G_MOVEWORD, data, (offset & 0xFFF), index)
 #define gsMoveWd(    index, offset, data) \
@@ -5252,7 +5210,7 @@ _DW({                                               \
 }
 
 #define gDPNoParam(pkt, cmd)   g1Word(pkt, cmd, 0)
-#define gsDPNoParam(cmd)      gs1Word(pkt, cmd, 0)
+#define gsDPNoParam(cmd)      gs1Word(     cmd, 0)
 
 #define gDPParam(pkt, cmd, param)       \
 _DW({                                   \
diff --git a/rsp/gbi.inc b/rsp/gbi.inc
index c81d536..60ac4ea 100644
--- a/rsp/gbi.inc
+++ b/rsp/gbi.inc
@@ -31,6 +31,7 @@ G_SHADING_SMOOTH        equ 0x00200000
 
 G_TRI_FILL          equ 0xc8 // not a GBI command
 
+G_FLUSH             equ 0xd4
 G_MEMSET            equ 0xd5
 G_DMA_IO            equ 0xd6
 G_TEXTURE           equ 0xd7