need to finally make a metaprogram for offset math...

code/duffle/gte.h

@@ -5,6 +5,7 @@
 #	include "mips.h"
 #endif
 
+#pragma region ASM DSL
 /* ============================================================================
  *  gte.h — Geometry Transformation Engine (COP2) for the PS1
  * ============================================================================
@@ -332,7 +333,7 @@ enum { _C2_OPS_ = 0
    | enc_gte_cmd(cmd) \
 )
 
-/* Pre-baked GTE command words for the common cases.
+/* GTE command words for the common cases.
  *
  * These are pure compile-time integer constants — the C compiler
  * constant-folds them into `.word` directives in .rodata. Use them
@@ -379,12 +380,17 @@ enum { _C2_OPS_ = 0
 #define gte_cmdw_op     (gte_cmd_base | enc_gte_cmd(gte_cmd_op   ))
 #define gte_cmdw_mvmva  (gte_cmd_base | enc_gte_cmd(gte_cmd_mvmva))
 
+#define gte_cmdw_rotate_translate_perspective_triple gte_cmdw_rtpt
+
 /* PsyQ compatibility bits for AVSZ3 (Bits 20, 22, 24 must be set) */
 #define gte_cmdw_psyq_avsz3_compat (0x15 << 20)
 #define gte_cmd_avsz3               0x2D
 
 #define gte_cmdw_avsz3 (gte_cmd_base | enc_gte_cmd(gte_cmd_avsz3) | gte_cmdw_psyq_avsz3_compat)
 
+// Takes the three screen-space Z values of the triangle → averages them → writes the result into the OTZ register.
+#define gte_avg_sort_z3 gte_cmdw_avsz3
+
 /* AVSZ4 — average Z of 4 vertices (for quads) */
 #define gte_cmd_avsz4 0x2E
 #define gte_cmdw_avsz4 (gte_cmd_base | enc_gte_cmd(gte_cmd_avsz4) | gte_cmdw_psyq_avsz3_compat)
@@ -403,7 +409,7 @@ enum { _C2_OPS_ = 0
  *   asm_gte_load_v0(svector_ptr);
  */
 
-/* Pre-baked lwc2 encoding helpers parameterized on the base GPR.
+/* lwc2 encoding helpers parameterized on the base GPR.
  *
  * gte_lwc2_v0(base)  → lwc2 $0,  0(base)   ; C2_VXY0
  * gte_lwc2_v0z(base) → lwc2 $1,  4(base)   ; C2_VZ0
@@ -656,3 +662,11 @@ enum {
 		, r_use(r0)                         \
 		asm_clobber: clb_system, rlit(R_T4), rlit(R_T5), rlit(R_T6) \
 	)
+
+#pragma region ASM DSL
+
+#pragma region Reserved
+
+
+
+#pragma endregion Reserved

code/duffle/lottes_tape.h

@@ -15,6 +15,7 @@
  *  The C compiler is completely unaware of these bindings.
  * ---------------------------------------------------------------------------*/
 enum {
+	R_AtomJmp  = R_T9,
 	R_TapePtr  = R_T8,  /* The Instruction Stream Pointer */
 	R_PrimCur  = R_T7,  /* VRAM output cursor (primitive buffer) */
 	R_FaceCur  = R_T4,  /* Input data cursor (indices/faces) */
@@ -78,33 +79,30 @@ FI_ Slice_U4 tb_slice(TapeBuilder  tb) {                             return (Sli
  * Loads the next pointer from the tape, advances the tape, and jumps. 
  * Cost: ~ 4 cycles */
 #define mac_yield()          \
-	  load_word(R_T9, R_TapePtr, 0) \
+	  load_word(R_AtomJmp, R_TapePtr, 0) \
-	, add_ui_1(       R_TapePtr, 4) \
+	, add_ui_1(            R_TapePtr, 4) \
-	, jump_reg( R_T9)               \
+	, jump_reg( R_AtomJmp)               \
 	, nop
 
-/* Loads 3 16-bit indices from the face array */
+/* Words: 3; Loads 3 S2 indices from the face array */
 #define mac_load_tri_indices(rId_0, rId_1, rId_2) \
 	  load_half_u(rId_0, R_FaceCur, 0) \
 	, load_half_u(rId_1, R_FaceCur, 2) \
 	, load_half_u(rId_2, R_FaceCur, 4)
 
-/* Translates indices to vertex addresses and pushes them to GTE */
+/* Words: 18; Translates indices to vertex addresses and pushes them to GTE 
+	R_AT  = rId_[#] << 3; 
+	R_AT += R_VertBase;
+	R_V0  = R_AT[0];
+	gte_mt(R_V0, V.xy[#]);
+	gte_mt(R_V1, V.z [#]);
+*/
 #define mac_load_tri_verts(rId_0, rId_1, rId_2) \
 	  shift_ll(R_AT, rId_0, 3), add_u(R_AT, R_AT, R_VertBase), load_word(R_V0, R_AT, 0), load_word(R_V1, R_AT, 4), gte_mt(R_V0, C2_VXY0), gte_mt(R_V1, C2_VZ0) \
 	, shift_ll(R_AT, rId_1, 3), add_u(R_AT, R_AT, R_VertBase), load_word(R_V0, R_AT, 0), load_word(R_V1, R_AT, 4), gte_mt(R_V0, C2_VXY1), gte_mt(R_V1, C2_VZ1) \
 	, shift_ll(R_AT, rId_2, 3), add_u(R_AT, R_AT, R_VertBase), load_word(R_V0, R_AT, 0), load_word(R_V1, R_AT, 4), gte_mt(R_V0, C2_VXY2), gte_mt(R_V1, C2_VZ2)
 
-/* Formats the primitive memory layout (Tag + Color + Coordinates) */
+/* Words: 11; Correctly inserts a primitive into the Ordering Table linked list */
-#define mac_format_prim_f3(color_hi, color_lo)          \
-	  store_word(R_0, R_PrimCur,  0)                      \
-	, load_ui(R_AT, color_hi), or_i(R_AT, R_AT, color_lo) \
-	, store_word(R_AT, R_PrimCur, 4)                      \
-	, gte_sw(C2_SXY0, R_PrimCur, 8)                       \
-	, gte_sw(C2_SXY1, R_PrimCur, 12)                      \
-	, gte_sw(C2_SXY2, R_PrimCur, 16)
-
-/* Correctly inserts a primitive into the Ordering Table linked list */
 #define mac_insert_ot_tag(r_otz, prim_length)                               \
 	  shift_ll(  R_T1, r_otz, 2)                                              \
 	, add_u(     R_T1, R_T1, R_OtBase)   /* T1 = &OrderingTable[OTZ] */       \
@@ -136,7 +134,7 @@ FI_ void atombuilder_unroll(MipsAtomBuilder_R ab, Slice_MipsCode_R code) {
 	mem_copy(ab->start, u4_(code->ptr), code->len);
 	mem_bump(ab->start, ab->capacity, & ab->used, code->len);
 }
-#define atombuilder_unroll_mac(ab, mac) 
+#define atombuilder_unroll_mac(ab, mac) atombuilder_unroll(ab, slice_arg_from_array(Slice_MipsCode, mac))
 
 // When done authoring, utilize this to cap-off the atom
 FI_ void atombuilder_end(MipsAtomBuilder_R ab) {
@@ -152,14 +150,19 @@ FI_ void atombuilder_end(MipsAtomBuilder_R ab) {
 #pragma region Baked Mips Atoms
 // These atoms are resolved at compile time and are (usually) statically linked readonly data.
 
+typedef Struct_(Binds_SyncPrimCursor) {
+	U4 PrimtiveArena_Used;
+	U4 PrimtiveBase;
+};
 internal MipsAtom_(sync_prim_cursor) {
 	/* Pop the C-struct address and base address from the tape */
-	load_word(R_AT, R_TapePtr, 0), /* AT = &pa->used */
+	// Note(Ed): Argument shuffle....
-	load_word(R_T0, R_TapePtr, 4), /* T0 = prim_base */
+	load_word(R_AT, R_TapePtr, O_(Binds_SyncPrimCursor,PrimtiveArena_Used)),
-	add_ui_1(       R_TapePtr, 8),
+	load_word(R_T0, R_TapePtr, O_(Binds_SyncPrimCursor,PrimtiveBase)),      
+	add_ui_1(       R_TapePtr, S_(Binds_SyncPrimCursor)),
 	/* Calculate byte offset and store directly back to RAM */
-	sub_u(R_T0, R_PrimCur, R_T0),
+	sub_u(R_T0, R_PrimCur, R_T0), // R_T0 = PrimitiveArea_Used(R_AT) - R_PrimCur
-	store_word(R_T0, R_AT, 0),
+	store_word(R_T0, R_AT, 0),    // PrimitiveBase(R_AT)[0] = R_T0
 	mac_yield()
 };
 
@@ -195,6 +198,8 @@ internal MipsAtom_(rbind_cube_tri) {
 	load_word(R_VertBase, R_TapePtr, O_(Binds_CubeTri,VertBase)), 
 	load_word(R_OtBase,   R_TapePtr, O_(Binds_CubeTri,OtBase)), 
 	add_ui_1(             R_TapePtr, S_(Binds_CubeTri)),
+	// Note(Ed): This entire thing is argument shuffle?
+	// TODO(Ed): Eliminate
 	mac_yield()
 };
 
@@ -345,46 +350,11 @@ internal MipsAtom_(rbind_floor_tri) {
 	load_word(R_VertBase, R_TapePtr, O_(Binds_FloorTri,VertBase)), 
 	load_word(R_OtBase,   R_TapePtr, O_(Binds_FloorTri,OtBase)), 
 	add_ui_1(             R_TapePtr, S_(Binds_FloorTri)),
+	// Note(Ed): This entire thing is argument shuffle?
+	// TODO(Ed): Eliminate
 	mac_yield()
 };
 
-internal MipsAtom_(floor_tri) {
-	mac_load_tri_indices(R_T0, R_T1, R_T2),
-	mac_load_tri_verts(  R_T0, R_T1, R_T2),
-
-	/* 3. Execute Math */
-	nop, nop, gte_cmdw_rtpt,
-	nop, nop, gte_cmdw_nclip,
-	nop, nop, 
-
-	/* 4. Culling (Branch forward 29 instructions if Backface) */
-	gte_mf(R_T0, C2_MAC0),
-	nop,                      
-	branch_le_zero(R_T0, 29), 
-	nop,                      
-	
-	/* 5. Format Primitive */
-	mac_format_prim_f3(0x20FF, 0xFFFF), /* High: 0x20/B, Low: G/R */
-	
-	/* 6. Calculate Depth */
-	nop, nop, gte_cmdw_avsz3,            
-	nop, nop, 
-	gte_mf(R_T1, C2_OTZ),      
-	
-	/* 7. Bounds Check OTZ < 2048 (Branch forward 13 instructions to skip insertion) */
-	add_ui(      R_AT, R_0,  2048),   
-	slt_u(       R_AT, R_T1, R_AT), 
-	branch_equal(R_AT, R_0,  13),   
-	nop, 
-	
-	/* 8. Insert into Ordering Table Linked List */
-	mac_insert_ot_tag(R_T1, 0x0400), /* Length = 4 words */
-
-	add_ui(R_PrimCur, R_PrimCur, 20), /* Advance Prim Cursor (5 words) */
-	/* 9. Advance Input Cursor & Yield (Both branch targets land here) */
-	add_ui(R_FaceCur, R_FaceCur, 8),  /* Advance Face Cursor (4 * S2 = 8 bytes) */
-	mac_yield()
-};
 
 /* DIAGNOSTIC 1: Pure tape loop test */
 internal MipsAtom_(diag_yield) { mac_yield() };

code/duffle/mips.h

@@ -69,6 +69,7 @@ enum {
 /* Semantic Aliases for MIPS Registers (O32 ABI) */
 
 	, rdiscard      = R_0  /* Hardwired to 0 */
+	, rasm_tmp      = R_AT /* Assembler temporary (destroyed by some assembler pseudoinstructions!) */
 	, rret_0        = R_V0 /* Function return value */
 	, rret_1        = R_V1 /* Second return value (e.g., 64-bit) */
 	, rarg_0        = R_A0 /* First function argument */
@@ -112,21 +113,30 @@ enum {
 	, op_xori    = 0x0E /* XOR Immediate */
 	, op_lui     = 0x0F /* Load Upper Immediate */
 	, op_cop0    = 0x10 /* Coprocessor 0 (System) */
+	, op_cop1    = 0x11 /* Coprocessor 1 (Reserved, FP Unit, Omitted by Sony) */
 	, op_cop2    = 0x12 /* Coprocessor 2 (GTE) */
-	, op_la      = 0
+	, op_cop3    = 0x13 /* Coprocessor 3 (Reserved, Unused)*/
-	, op_li      = 0
+	/* 14-1F: N/A */
 	, op_lb      = 0x20 /* Load Byte */
 	, op_lh      = 0x21 /* Load Halfword */
+	, op_lwl     = 0x22 /* Load Word (Left Bits) */
 	, op_lw      = 0x23 /* Load Word */
 	, op_lbu     = 0x24 /* Load Byte Unsigned */
 	, op_lhu     = 0x25 /* Load Halfword Unsigned */
+	, op_lwr     = 0x26 /* Load Word (Right Bits) */
+	/* 27: N/A */
 	, op_sb      = 0x28 /* Store Byte */
 	, op_sh      = 0x29 /* Store Halfword */
+	, op_swl     = 0x2A /* Store Word (Left Bits) */
 	, op_sw      = 0x2B /* Store Word */
+	/* 2C-2D: N/A */
+	, op_swr     = 0x2E /* Store Word (Right Bits) */
+	/* 2F: N/A */
+	// , op_lwc0
 
 
-	, op_load_addr  = op_la
+	// , op_load_addr  = op_la
-	, op_load_imm   = op_li
+	// , op_load_imm   = op_li
 	, op_jump       = op_j
 	, op_jump_nlink = op_jal
 
@@ -246,6 +256,10 @@ enum { _BitOffsets = 0
 #define xor_i(rt, rs, imm)         enc_i(op_xori,  (rs),   (rt), (imm))
 #define load_ui(rt, imm)           enc_i(op_lui,   R_0,    (rt), (imm))
 
+#define load_u1 load_byte_u
+#define load_u2 load_half_u
+#define load_u4 load_word
+
 // Ergonomic add to the same register.
 #define add_ui_1(rt_rs, imm)       enc_i(op_addiu, (rt_rs), (rt_rs), (imm))
 

code/gte_hello/hello_gte.c

@@ -14,11 +14,7 @@
 #include "duffle/gte.h"
 #include "duffle/lottes_tape.h"
 #include "hello_gte.h"
-
+#include "hello_gte_tape.c"
-enum {
-	PrimitiveBuff_Len = 4096,
-	OrderingTbl_Len   = 2048
-};
 
 typedef U4 OrderingTable_Buffer[OrderingTbl_Len];
 typedef Array_(OrderingTable_Buffer, 2);
@@ -104,6 +100,8 @@ typedef Struct_(SMemory) {
 
 	Ent_Cube  cube;
 	Ent_Floor floor;
+
+
 };
 global SMemory static_mem;
 extern SMemory static_mem;
@@ -245,10 +243,10 @@ void update(PrimitiveArena* pa, U4* ordering_buf)
 		LP_ U4 mem_temp_tape[512]; FArena tape_arena; farena_init(& tape_arena, slice_ut_arr(mem_temp_tape));
 		TapeBuilder tb = tb_make(&tape_arena); tb_scope(& tb) {
 			tb_emit(& tb, code_rbind_cube_tri);
-			tb_data(& tb, prim_cursor);
+				tb_data(& tb, prim_cursor);
-			tb_data(& tb, u4_(static_mem.cube.faces));
+				tb_data(& tb, u4_(static_mem.cube.faces));
-			tb_data(& tb, u4_(static_mem.cube.verts));
+				tb_data(& tb, u4_(static_mem.cube.verts));
-			tb_data(& tb, u4_(ordering_buf));
+				tb_data(& tb, u4_(ordering_buf));
 
 			for (U4 i = 0; i < Cube_num_faces; i++) {
 				// Two triangles per quad face: (x,y,z) and (x,z,w)
@@ -256,8 +254,8 @@ void update(PrimitiveArena* pa, U4* ordering_buf)
 			}
 
 			tb_emit(& tb, code_sync_prim_cursor);
-			tb_data(& tb, u4_(& pa->used));
+				tb_data(& tb, u4_(& pa->used));
-			tb_data(& tb, prim_base);
+				tb_data(& tb, prim_base);
 		}
 		tape_run(tb_slice(tb));
 
@@ -334,21 +332,24 @@ void update(PrimitiveArena* pa, U4* ordering_buf)
 		LP_ U4 mem_temp_tape[512]; FArena tape_arena; farena_init(& tape_arena, slice_ut_arr(mem_temp_tape));
 		TapeBuilder tb = tb_make(&tape_arena); tb_scope(& tb) {
 			// Push "Protocol" to tape
-			tb_emit_(& tb, rbind_floor_tri);
+			tb_emit(& tb, code_rbind_floor_tri);
+				// Note(Ed): This is technically argument shuffle and would be better if we did a single reference at most to a global batch context.
+				// Note(Ed): We can technically allocate a single ptr with the global offset to the working context instead of utilizing the tape for this stack of refs.
 				tb_data(& tb, prim_cursor);
 				tb_data(& tb, u4_(static_mem.floor.faces));
 				tb_data(& tb, u4_(static_mem.floor.verts));
 				tb_data(& tb, u4_(ordering_buf));
+
 			tb_emit(& tb, code_set_gte_world);
-			tb_data(& tb, u4_(& static_mem.tform_world));
+				tb_data(& tb, u4_(& static_mem.tform_world));
-            
+
 			for (U4 i = 0; i < Floor_num_faces; i++) {
 				tb_emit(& tb, code_floor_tri);
 			}
 			
 			tb_emit(& tb, code_sync_prim_cursor);
-			tb_data(& tb, u4_(& pa->used));
+				tb_data(& tb, u4_(& pa->used));
-			tb_data(& tb, prim_base);
+				tb_data(& tb, prim_base);
 		}
 
 		// Fire off the tape.

code/gte_hello/hello_gte.h

@@ -5,6 +5,11 @@
 #	include "duffle/gp.h"
 #endif
 
+enum {
+	PrimitiveBuff_Len = 4096,
+	OrderingTbl_Len   = 2048
+};
+
 typedef Struct_(DrawEnv_Packed) { U4 tag; U4 code[15]; };
 typedef Struct_(DrawEnv) {
 	Rect_S2 clip_area;

code/gte_hello/hello_gte_tape.c

@@ -0,0 +1,61 @@
+#ifdef INTELLISENSE_DIRECTIVES
+#	include "duffle/lottes_tape.h"
+#	include "hello_gte.h"
+#endif
+
+#pragma region MACs
+
+/* Words: 3; High: 0x20/B, Low: G/R */
+#define mac_format_f3_color(color_hi, color_lo) \
+	  load_ui(R_AT, color_hi), or_i(R_AT, R_AT, color_lo) \
+	, store_word(R_AT, R_PrimCur, O_(Poly_F3,color))   \
+
+/* Words: 3 */
+#define mac_gte_store_f3() \
+	  gte_sw(C2_SXY0, R_PrimCur, O_(Poly_F3,p0)) \
+	, gte_sw(C2_SXY1, R_PrimCur, O_(Poly_F3,p1)) \
+	, gte_sw(C2_SXY2, R_PrimCur, O_(Poly_F3,p2))
+
+#pragma endregion MACs
+
+#pragma region Baked Atoms
+
+internal MipsAtom_(floor_tri) {
+	// T0-T2 allocated
+	mac_load_tri_indices(R_T0, R_T1, R_T2),
+	mac_load_tri_verts(  R_T0, R_T1, R_T2),
+
+	/* 3. Execute Math */
+	nop, nop, gte_cmdw_rotate_translate_perspective_triple,
+	nop, nop, gte_cmdw_nclip,
+	nop, nop, 
+
+	/* 4. Culling (Branch forward 29 instructions if Backface) */
+	gte_mf(R_T0, C2_MAC0),
+	nop, branch_le_zero(R_T0, 29), 
+	nop,                      
+	
+	/* 5. Format Primitive */
+	mac_format_f3_color(0x20FF, 0xFFFF), 
+	mac_gte_store_f3(),
+	
+	/* 6. Calculate Depth */
+	nop, nop, gte_avg_sort_z3,            
+	nop, nop, gte_mf(R_T1, C2_OTZ),      
+	
+	/* 7. Bounds Check OTZ < 2048 (Branch forward 13 instructions to skip insertion) */
+	add_ui(      R_AT, R_0,  OrderingTbl_Len),   
+	slt_u(       R_AT, R_T1, R_AT), 
+	branch_equal(R_AT, R_0,  13),   
+	nop, 
+	
+	/* 8. Insert into Ordering Table Linked List */
+	mac_insert_ot_tag(R_T1, 0x0400),
+	add_ui(R_PrimCur, R_PrimCur, S_(Poly_F3)), /* Advance Prim Cursor (5 words) */
+
+	/* 9. Advance Input Cursor & Yield (Both branch targets land here) */
+	add_ui(R_FaceCur, R_FaceCur, S_(S2) * 4),  /* Advance Face Cursor (4 * S2 = 8 bytes) */
+	mac_yield()
+};
+
+#pragma endregion Baked Atoms