Moved XP variant to makefile

Makefile

@@ -11,9 +11,9 @@ default: F3DEX3_BrZ F3DEX3_BrW
 # List of all compile-time options supported by the microcode source.
 ALL_OPTIONS := \
   CFG_G_BRANCH_W \
-  CFG_DEBUG_NORMALS \
   CFG_NO_OCCLUSION_PLANE \
   CFG_LEGACY_VTX_PIPE \
+  CFG_EXTRA_PRECISION \
   CFG_PROFILING_A \
   CFG_PROFILING_B \
   CFG_PROFILING_C
@@ -139,44 +139,44 @@ define rule_builder_final
   $$(eval $$(call ucode_rule))
 endef
 
-define rule_builder_dbgn
-  NAME_FINAL := $(NAME_DBGN)
-  OPTIONS_FINAL := $(OPTIONS_DBGN)
+define rule_builder_prof
+  NAME_FINAL := $(NAME_PROF)
+  OPTIONS_FINAL := $(OPTIONS_PROF)
   $$(eval $$(call rule_builder_final))
 
-  NAME_FINAL := $(NAME_DBGN)_dbgN
-  OPTIONS_FINAL := $(OPTIONS_DBGN) CFG_DEBUG_NORMALS
+  NAME_FINAL := $(NAME_PROF)_PA
+  OPTIONS_FINAL := $(OPTIONS_PROF) CFG_PROFILING_A
   $$(eval $$(call rule_builder_final))
-endef
-
-define rule_builder_prof
-  NAME_DBGN := $(NAME_PROF)
-  OPTIONS_DBGN := $(OPTIONS_PROF)
-  $$(eval $$(call rule_builder_dbgn))
-
-  NAME_DBGN := $(NAME_PROF)_PA
-  OPTIONS_DBGN := $(OPTIONS_PROF) CFG_PROFILING_A
-  $$(eval $$(call rule_builder_dbgn))
 
-  NAME_DBGN := $(NAME_PROF)_PB
-  OPTIONS_DBGN := $(OPTIONS_PROF) CFG_PROFILING_B
-  $$(eval $$(call rule_builder_dbgn))
+  NAME_FINAL := $(NAME_PROF)_PB
+  OPTIONS_FINAL := $(OPTIONS_PROF) CFG_PROFILING_B
+  $$(eval $$(call rule_builder_final))
 
-  NAME_DBGN := $(NAME_PROF)_PC
-  OPTIONS_DBGN := $(OPTIONS_PROF) CFG_PROFILING_C
-  $$(eval $$(call rule_builder_dbgn))
+  NAME_FINAL := $(NAME_PROF)_PC
+  OPTIONS_FINAL := $(OPTIONS_PROF) CFG_PROFILING_C
+  $$(eval $$(call rule_builder_final))
 endef
 
-define rule_builder_noc
-  NAME_PROF := $(NAME_NOC)
-  OPTIONS_PROF := $(OPTIONS_NOC)
+define rule_builder_xp
+  NAME_PROF := $(NAME_XP)
+  OPTIONS_PROF := $(OPTIONS_XP)
   $$(eval $$(call rule_builder_prof))
 
-  NAME_PROF := $(NAME_NOC)_NOC
-  OPTIONS_PROF := $(OPTIONS_NOC) CFG_NO_OCCLUSION_PLANE
+  NAME_PROF := $(NAME_XP)_XP
+  OPTIONS_PROF := $(OPTIONS_XP) CFG_EXTRA_PRECISION
   $$(eval $$(call rule_builder_prof))
 endef
 
+define rule_builder_noc
+  NAME_XP := $(NAME_NOC)
+  OPTIONS_XP := $(OPTIONS_NOC)
+  $$(eval $$(call rule_builder_xp))
+
+  NAME_XP := $(NAME_NOC)_NOC
+  OPTIONS_XP := $(OPTIONS_NOC) CFG_NO_OCCLUSION_PLANE
+  $$(eval $$(call rule_builder_xp))
+endef
+
 define rule_builder_lvp
   NAME_NOC := $(NAME_LVP)
   OPTIONS_NOC := $(OPTIONS_LVP)

docs/Documentation/Configuration.md

@@ -107,8 +107,18 @@ SM64), or `BrW` if the microcode is replacing F3DZEX (i.e. OoT or MM). This
 controls whether `SPBranchLessZ*` uses the vertex's W coordinate or screen Z
 coordinate.
 
+## Extra Precision (`XP`)
+
+This configuration attempts to reproduce F3DEX(1) numerical behavior for Z
+buffer coefficients, potentially improving Z fighting in some cases of decals or
+opaque surfaces intended to behave like decals.
+
 ## Debug Normals (`dbgN`)
 
+Debug Normals has been moved out of the Makefile as it is not a microcode
+version intended to be shipped. It can still be enabled by changing
+`CFG_DEBUG_NORMALS equ 0` to `1` in the microcode.
+
 To help debug lighting issues when integrating F3DEX3 into your romhack, this
 feature causes the vertex colors of any material with lighting enabled to be set
 to the transformed, normalized world space normals. The X, Y, and Z components

docs/Documentation/Design Tradeoffs.md

@@ -161,6 +161,17 @@ behavior when supplying a direct-mapped or physical address such as 0x80101000,
 segment 0 must always be 0x00000000 so that this address resolves to e.g.
 0x101000 as expected in this example.
 
+## Non-textured tris
+
+In F3DEX2, the RSP time for drawing non-textured tris was significantly lower
+than for textured tris, by skipping a chunk of computation for the texture
+coefficients if they were disabled. In F3DEX3, little to no computation is
+skipped when textures are disabled, which means that the performance gain from
+disabling textures in F3DEX2 has been mostly eliminated. (RDP time savings from
+avoiding loading a texture are unaffected of course.) However, almost all
+materials use textures, and F3DEX3 is a little faster at drawing textured tris
+than F3DEX2, so this is still a benefit overall.
+
 ## Obscure semantic differences from F3DEX2 that should never matter in practice
 
 - `SPLoadUcode*` corrupts the current M inverse transpose matrix state. If using

f3dex3.s

@@ -1335,7 +1335,6 @@ tPosLmH equ $v8
 tPosHmM equ $v11
 tPosHmL equ $v12 // not computed directly in CFG_EXTRA_PRECISION
 
-CFG_EXTRA_PRECISION equ 0
 .if CFG_EXTRA_PRECISION
 tNRPosFactor equ $v31[4] // 4
 tNRNegFactor equ $v31[0] // -4
@@ -3534,6 +3533,7 @@ lt_after_xfrm_normals:
     jal     lt_normalize
      luv    vPairLt, (ltBufOfs + 0)(curLight) // Total light level, init to ambient
     // Set up ambient occlusion: light *= (factor * (alpha - 1) + 1)
+CFG_DEBUG_NORMALS equ 0 // Can manually enable here
 .if CFG_DEBUG_NORMALS
 .warning "Debug normals visualization is enabled"
     vmudh   $v29, vOne, $v31[5] // 0x4000; middle gray