# Makefile (for typhoon/GNU-style make).  Recognized gmake variables:
#
#  DEBUG=1: make debug version (includes asserts and sanity checks)
#  TEST=1: include self-test code in the binary produced
#  ASM=1: create assembly code, does not link final binary
#  EVAL_DUMP=1: make a version that can dump eval breakdowns
#  EVAL_TIME=1: make a version that counts cycles spent in eval
#  PERF_COUNTERS=1: make version with perf counters enabled
#  MOVE_ORDERING_EVIDENCE_COLLECTION=1: instrument the move-ordering
#    evidence-bucket harnesses (countermove/move-pair hit tracking,
#    CM/MP evidence-bucket tries/FH -- see MOVE_ORDERING_IDEAS.md) --
#    a dedicated flag, independent of PERF_COUNTERS, since this
#    collection (per-move IS_SAME_MOVE checks + bucket lookups on every
#    quiet move scored) is its own cost, not bundled into the general
#    perf-counter set by default.
#  BOUNDS_CHECKING=1: make version with bounds checking enabled
#  MP=1: make version for multi-processor machine
#  TREEDUMP=1: dump a flat MOVE/RESULT search-tree trace to
#    treedump.log (see TREEDUMP_FORMAT.md)
#  OSX=1: build for an Intel-based Apple Mac (omit this for FreeBSD/Linux)
#  USE_READLINE=1: link against the GNU readline library
#  SIXTYFOUR=1: make an X64 binary
#  CROUTINES=1: use the C versions of the asm routines [slower]
#  GETATTACKS_BITBOARD=1: use the bbPieces/bbPawns-backed _GetAttacksBB
#    instead of the asm/CROUTINES GetAttacks -- see
#    board_representation/MIGRATION.md section 6
#  DISABLE_BITBOARD_MOVEGEN=1: opt back into the mailbox move
#    generator/escapes/ExposesCheck/IsAttacked implementations -- the
#    nine GENERATE_*_BITBOARD/EXPOSESCHECK_BITBOARD/ISATTACKED_BITBOARD
#    toggles (board_representation/MOVEGEN_MIGRATION.md) are on by
#    default as of the 2026-09-04 correctness/no-regression sweep
#    (perft, move-set comparison harness, sd10 curated-suite parity,
#    match_play.py). The mailbox code is still fully present and
#    compiled either way -- this flag only flips which side of each
#    macro-swap wins, it doesn't remove anything.
#  GPROF=1: instrument with -pg for gprof(1)-based profiling (see
#    PROFILE.md) -- root-free on both FreeBSD and Linux, unlike
#    hardware-counter-based profiling (pmcstat/perf). Requires a
#    `gmake clean` first/after, since -pg changes codegen for every
#    translation unit.
#  EVERYTHING=1: everything everything everything everything
#
# $Id$
#
CC 		= 	clang
CXX		=	clang++
NASM		=	yasm

# Default to a 64-bit build -- 32-bit clang isn't available in this
# environment anyway (see CLAUDE.md). Still overridable in principle
# (e.g. `gmake SIXTYFOUR=` won't un-set it since ?= only fires when
# unset, but this matches how every build in this session has actually
# been invoked).
SIXTYFOUR	?=	1
ifdef OSX
 LIBRARIES	=	-pthread -lc
 ifdef SIXTYFOUR
  $(warning you will need to use yasm for now to get x64 on OSX)
  NASMFLAGS	=	-f macho64 -dOSX -d_X64_
 else
  NASMFLAGS	=	-f macho -dOSX
 endif
else
 LIBRARIES	=	-pthread -lc -lstdc++ 
 ifdef SIXTYFOUR
  NASMFLAGS	=	-f elf64 -d__GNUC__
 else
  NASMFLAGS	=	-f elf -d__GNUC__
 endif
endif
RM		= 	/bin/rm

#
# Setup commandline based on make defines
#
ifdef ASM
 ifdef SIXTYFOUR
  PROFILE	=	-S -fverbose-asm -D_X64_
 else
  PROFILE	=	-S -fverbose-asm -D_X86_
 endif
else
 ifdef SIXTYFOUR
  PROFILE 	=	-D_X64_
 else
  PROFILE 	=	-D_X86_
 endif
endif

ifdef OSX
 PROFILE	+=	-DOSX
endif

ifdef GENETIC
 PROFILE	+=	-DGENETIC
endif

ifdef GPROF
 PROFILE	+=	-pg
endif

# Scoped ISA opt-in for NNUE SIMD work (int8/int16 dot products, and
# the FC head's existing float32 loops which clang already
# autovectorizes to SSE only without this). Default ON since 2026-10-16
# -- confirmed a real, measured win for both classic HCE (~3%) and
# (once real NNUE weights are loaded) the NNUE FC head (~7%, see
# docs/NNUE.md), and the build box (Zen 2) has full-width 256-bit
# AVX2 + FMA3. Pass AVX2=0 explicitly to disable (e.g. building on a
# pre-Haswell/pre-Zen box, 2013 and earlier) -- unlike -march=native,
# this doesn't assume the BUILD box matches the RUN box, but it does
# assume AVX2 availability by default now rather than opt-in.
ifneq ($(AVX2),0)
 PROFILE	+=	-mavx2 -mfma
endif

ifdef DEBUG
 PROFILE 	+=	-g -DDEBUG
 BINARY 	=	_typhoon
else
 PROFILE	+=	-g -O3 -ffast-math -finline-functions
 BINARY 	=	typhoon
endif

ifdef USE_READLINE
 LIBRARIES	+=	-lreadline
 PROFILE	+=	-DUSE_READLINE
endif

ifdef SIXTYFOUR
 # -mpopcnt: every x86-64 chip built since ~2008 (and this box's Ryzen
 # 9 3900X) has the POPCNT instruction, but clang won't emit it for
 # __builtin_popcountll without this -- without it, the builtin lowers
 # to a ~15-instruction SWAR bit-trick sequence instead (still faster
 # than the old CDECL-call CountBits, but not the ~1-3 cycle single
 # instruction the "fast" eval.c wrappers were assumed to get).
 PROFILE	+=	-m64 -mpopcnt
 NASMFLAGS	+=	-d_X64_
else
 PROFILE	+=	-m32
 NASMFLAGS	+=	-d_X86_
endif

ifdef CROUTINES
 PROFILE	+=	-DCROUTINES
endif

ifdef GETATTACKS_BITBOARD
 PROFILE	+=	-DGETATTACKS_BITBOARD
endif

ifndef DISABLE_BITBOARD_MOVEGEN
 PROFILE	+=	-DGENERATE_KNIGHT_BITBOARD -DGENERATE_KING_BITBOARD \
			-DGENERATE_ROOK_BITBOARD -DGENERATE_BISHOP_BITBOARD \
			-DGENERATE_QUEEN_BITBOARD -DGENERATE_PAWN_BITBOARD \
			-DGENERATE_ESCAPES_KING_BITBOARD -DGENERATE_ESCAPES_BLOCK_BITBOARD \
			-DEXPOSESCHECK_BITBOARD -DISATTACKED_BITBOARD \
			-DUNBLOCKED_BITBOARD
endif

ifdef EVERYTHING
PROFILE 	+=	-DEVAL_DUMP -DEVAL_TIME -DPERF_COUNTERS -DMP -DSMP -DTEST_NULL -DTREEDUMP -fbounds-checking -DMOVE_ORDERING_EVIDENCE_COLLECTION
else
ifdef EVAL_DUMP
PROFILE		+= 	-DEVAL_DUMP
endif
ifdef EVAL_TIME
PROFILE		+= 	-DEVAL_TIME
endif
ifdef PERF_COUNTERS
PROFILE		+=	-DPERF_COUNTERS
endif
# 2026-10-10: this flag was referenced (search.c's joint-table/marginal
# collection block, #ifdef FH_TABLE_COLLECTION) but never actually wired
# to a compiler define anywhere in this makefile -- every prior
# collection run must have been built some other way (or this gap
# predates the comment claiming it's "a live, in-progress collection
# run"). Requires PERF_COUNTERS=1 too (the block this gates is nested
# inside an outer #ifdef PERF_COUNTERS).
#
# Re-added 2026-10-20: this wiring was itself silently dropped at some
# point after landing (collateral damage from the 2026-10-14 PRUNE_VERIFY/
# PRUNE_LIVE dead-flag cleanup, going by GNUmakefile.orig's surviving copy)
# -- every FH_TABLE_COLLECTION=1 build in between compiled clean but
# never actually included the instrumentation. Caught before launching
# the NNUE dual-table re-collection run; a smoke-test build+run with
# this restored confirmed -DFH_TABLE_COLLECTION actually reaches the
# compiler and the (currently 4D-alpha-only, no beta side -- see
# docs/NNUE.md) FOURAXISRAW dump fires. The original run's richer
# ALPHARAW/BETARAW/ESTRAW dump format no longer exists in this tree at
# all (not just unwired) -- that's the separate, larger restoration
# this collection run is waiting on, tracked alongside this fix.
ifdef FH_TABLE_COLLECTION
PROFILE		+=	-DFH_TABLE_COLLECTION
endif
ifdef MOVE_ORDERING_EVIDENCE_COLLECTION
PROFILE		+=	-DMOVE_ORDERING_EVIDENCE_COLLECTION
endif
# 2026-10-23, one-off measurement: does IS_CAPTURE(mvHash) (optionally
# && pi->eNodeType==NODE_CUT) meaningfully predict a leftover move's
# fail-high rate, beyond what the existing joint FH table's five axes
# already capture? Neither signal is currently baked into the table at
# all. Bucketed counters only (chess.h's u64HashCapTries/FH, [2][2] --
# [hash move is a capture][node type guess is CUT]), printed by
# script.c's _PrintSuiteReport -- cheap enough to leave compiled in for
# a --cpus 1 collection run; not wired into split.c's MP merge, same
# convention as the other single-threaded-only calibration counters
# (u64TotalCMEvidenceTries et al) in that same report.
ifdef HASHCAP_COLLECTION
PROFILE		+=	-DHASHCAP_COLLECTION
endif
ifdef BOUNDS_CHECKING
PROFILE		+=	-fbounds-checking
endif
ifdef MP
PROFILE		+=	-DMP -DSMP
endif
ifdef TEST_NULL
PROFILE		+=	-DTEST_NULL
endif
ifdef TREEDUMP
PROFILE 	+=  -DTREEDUMP
endif
ifdef CALIBRATE_POSITIONAL
PROFILE		+=	-DCALIBRATE_POSITIONAL
endif
ifdef KING_DEFECT_DUMP
PROFILE		+=	-DKING_DEFECT_DUMP
endif
ifdef CALIBRATE_BASE_MARGIN
PROFILE		+=	-DCALIBRATE_BASE_MARGIN
endif
ifdef CALIBRATE_MARGIN_SAFETY
PROFILE		+=	-DCALIBRATE_MARGIN_SAFETY
endif
ifdef DUMP_RESIDUAL_SAMPLES
PROFILE		+=	-DDUMP_RESIDUAL_SAMPLES
endif
ifdef DIAG_NO_QSEARCH_FUTILITY
PROFILE		+=	-DDIAG_NO_QSEARCH_FUTILITY
endif
ifdef CALIBRATE_QSEARCH_FUTILITY
PROFILE		+=	-DCALIBRATE_QSEARCH_FUTILITY
endif
endif # EVERYTHING
# 2026-09-28: full-coverage (non-sampled) throwaway shadow re-search,
# purely to measure the "silent miss" failure mode a reduced move's own
# real re-search hook can't see: the extra-reduced score is < beta (no
# real re-search triggered) but removing just the EXTRA plies (i.e. the
# depth HEAD's own baseline -ONE_PLY reduction would have used) finds
# a fail-high. Every extra-reduced move gets this check within a build
# with this flag on -- doubles search cost for that subset, so
# calibration-only, never part of a normal or self-play build. Opt in
# explicitly, same as the other CALIBRATE_*/DUMP_* flags above.
ifdef JOINT_EXTRA_VERIFY
PROFILE		+=	-DJOINT_EXTRA_VERIFY
endif
# 2026-09-28: audit-only EFP disable, same technique this project used
# during the joint-table collection run's self-censoring fix (force the
# prune decision FALSE, keep computing/logging everything around it) --
# isolates the joint-table-driven LMR's own behavior from EFP's
# single-axis (gap/down only) margin-based pruning, which the
# JOINT_EXTRA_VERIFY silent-miss data implicates as the dominant source
# of the measured instability, not LMR's own reduction depth. Opt-in,
# calibration-only.
ifdef DISABLE_EFP
PROFILE		+=	-DDISABLE_EFP
endif
# 2026-09-28: Trace() the FEN/move/scores whenever JOINT_EXTRA_VERIFY's
# shadow search catches a silent miss -- for looking at real, concrete
# failing cases by hand (the methodology that actually cracked the
# hanging-material bug earlier this project, rather than continuing to
# guess which other search mechanism might explain the EFP-tier-window
# silent-miss spike). Requires JOINT_EXTRA_VERIFY. Calibration-only.
ifdef SILENT_MISS_TRACE
PROFILE		+=	-DSILENT_MISS_TRACE
endif
# 2026-09-29: alternate, maximally-simple extra-reduction policy for A/B
# comparison against the axis-gated joint-table version -- skips the
# dumb/gap/late table lookup entirely and just asks "is there enough
# pre-reduction depth headroom to spare (>= 6 plies)?" If so, apply the
# full extra reduction (R=3) uniformly, for ANY depth >= 6 (not just
# exactly 6 -- 7, 8, 9, ... all get the same treatment). Tests whether
# the axis-based table is pulling any real weight once a depth floor is
# already in place, or whether depth alone does just as well.
ifdef DEPTH_ONLY_EXTRA_REDUCTION
PROFILE		+=	-DDEPTH_ONLY_EXTRA_REDUCTION
endif
# 2026-09-29: neutralize searchsup.c's null-move "history-reduced
# bailout" branch (Tord Romstad idea) for our own shadow-verification
# Search() call specifically, by temporarily zeroing our own
# pi->iExtensionAmount around it -- see search.c's NEUTRALIZE_NULL_
# BAILOUT comment. Correct in the real search path, that branch
# fabricates a "please re-search me" placeholder score that our
# bespoke, wrapper-less shadow call can't interpret and misreads as a
# genuine fail-high. Build once WITHOUT this flag to measure how much
# of our silent-miss data is this artifact (see
# u64JointExtraVerifyNullBailoutSignature), once WITH it for the
# corrected rate.
# 2026-10-03: PRUNE_VERIFY -- shadow verification for the DEEP PRUNE gate
# (ShouldLatePrune) was removed (2026-10-14): it never actually compiled
# (fVerifyFlagged/uVerifyBand/etc. were used in search.c but never
# declared -- PRUNE_VERIFY is never defined by any path in this file, so
# nothing ever caught it). Dead, broken code, not a working diagnostic;
# removed rather than fixed since reconstructing its intended semantics
# from scratch would be inventing behavior, not restoring it.

# 2026-10-03: PRUNE_LIVE=<K> was removed (2026-10-14): PruneLivePredict/
# PRUNE_LIVE_SENTINEL (search.c) were never actually called from
# anywhere, under any flag -- the wiring to make PRUNE_LIVE=<K> actually
# prune a move was never written, so this flag has never done anything
# even when it compiled. It also didn't compile: PruneVerifyPredict
# (the function PruneLivePredict aliased to) returns
# g_uJointFHThousandths[...]'s element as ULONG, but that array is
# declared USHORT in data.c -- confirmed pre-existing, unrelated to
# PRUNE_VERIFY's own removal above. Dead on every axis; removed rather
# than fixed for the same reason as PRUNE_VERIFY.
# 2026-10-03: DISABLE_LMR -- GetLMRReduction returns 0 unconditionally,
# so no move is ever reduced. Exists for ONE experiment: what is LMR
# actually worth in this engine? The time-odds calibration established
# that TIME-bought depth is ~95 Elo/ply here, but nothing has ever
# measured whether REDUCTION-bought depth converts at that rate. Two
# hypotheses fit "deeper but no stronger": (1) missed refutations, i.e.
# the reduction is too aggressive and its errors cancel the gain --
# remedy is to constrain it; or (2) nominal depth under reduction is a
# debased currency that never carried a time-ply's worth of information
# -- in which case constraining it cannot help, since less reduction
# buys less depth at the same bad rate. Removing LMR entirely and
# pricing the loss against the 95 Elo/ply yardstick separates them.
# Measurement only; never a shipping configuration.
ifdef DISABLE_LMR
PROFILE		+=	-DDISABLE_LMR
endif

# 2026-10-13: override search.c's NULLMOVE_SKIP_FH_CEILING_THOUSANDTHS
# default (300, validated -- see search.c's own comment) from the
# command line for sweeping, e.g. `gmake ... NULLMOVE_SKIP_CEILING=1000`
# -- search.c's #define is guarded with #ifndef so this wins without
# editing the source.
ifdef NULLMOVE_SKIP_CEILING
PROFILE		+=	-DNULLMOVE_SKIP_FH_CEILING_THOUSANDTHS=$(NULLMOVE_SKIP_CEILING)
endif

# 2026-10-13: override movesup.c's SelectBestWithHistory skip-
# selection-sort threshold for sweeping, e.g.
# `gmake ... SKIP_SELECT_N=6`.
ifdef SKIP_SELECT_N
PROFILE		+=	-DSKIP_SELECT_AFTER_N_MOVES=$(SKIP_SELECT_N)
endif

ifdef NEUTRALIZE_NULL_BAILOUT
PROFILE		+=	-DNEUTRALIZE_NULL_BAILOUT
endif

# Short commit hash (plus a -dirty suffix if the working tree has
# uncommitted changes) baked into the binary so a --logfile trace can
# be tied back to the exact source state, not just a build timestamp --
# distinguishing rebuilds of the same day/commit during A/B testing was
# previously only possible by diffing binaries. Placed here, after all
# the PROFILE-overwriting `=` assignments above (not `+=`), so it
# survives into CFLAGS instead of being wiped out by them.
GIT_COMMIT	:=	$(shell git rev-parse --short HEAD 2>/dev/null || echo unknown)
# Scoped to this directory: an unscoped `git diff` has to stat (and,
# with a stale index, re-hash) every tracked file in the repo,
# including the multi-GB pgn/twic.pgn, on every make invocation.
GIT_DIRTY	:=	$(shell git diff --quiet -- . 2>/dev/null || echo -dirty)
# Kept out of PROFILE itself (not folded in via +=): PROFILE gets
# stringified whole into -DPROFILE="..." below for the "Make profile
# used" trace line, and this value's embedded quotes broke that outer
# string literal when it was included there.
# -flto: several functions (e.g. _RookAttacksBB/_BishopAttacksBB,
# generate.c) are declared FORCEINLINE but non-static, exposed
# specifically so other translation units (eval.c, searchsup.c, see.c)
# can call them -- __attribute__((always_inline)) only actually forces
# inlining within the TU that has the function body, so every
# cross-file call was paying for a real call/ret instead of inlining
# without this. Measured via the `bench` command (4 reps each,
# interleaved with a non-LTO baseline): a clean, reproducible ~6.6% NPS
# improvement with zero overlap between the two distributions (baseline
# 1643k-1661k nps, LTO 1752k-1768k nps) -- not noise. -mtune=native:
# tunes instruction scheduling/costs for the actual build machine
# rather than a generic x86-64 target; measured weaker and noisier
# (~1.5%, some overlap with baseline) but real and free (no
# portability cost the way -march=native's newer-ISA-extension
# dependency would have -- see the BMI2/PEXT-on-Zen2 caveat if anyone
# reaches for -march=native or PEXT-based magic bitboards later).
CFLAGS		=	-DPROFILE="\"$(PROFILE)\"" -DGIT_COMMIT=\"$(GIT_COMMIT)$(GIT_DIRTY)\" $(PROFILE) -Wall -Wno-nan-infinity-disabled -Ifathom -flto -mtune=native
HEADERS 	=	chess.h compiler.h
FATHOM_CFLAGS	=	-std=gnu99 -O2 -Ifathom $(filter -m32 -m64,$(PROFILE))

#
# ---> .o, not .c! <---
#
OBJS    =       main.o root.o search.o searchsup.o draw.o dynamic.o \
		hash.o eval.o x86.o pawnhash.o bitboard.o \
		generate.o see.o move.o movesup.o command.o script.o \
		input.o vars.o util.o unix.o gamelist.o mersenne.o \
		sig.o piece.o ics.o san.o fen.o book.o bench.o board.o \
		data.o probe.o fathom.o recogn.o list.o x64.o \
		nnuefeatures.o nnueaccum.o nnueeval.o

ifdef TEST
# ---> .o, not .c! <---
#
OBJS    +=      testdraw.o testmove.o testfen.o testgenerate.o \
		testsan.o testics.o testeval.o testbitboard.o \
		testhash.o testsee.o testsearch.o testsup.o \
		testrecogn.o testnnue.o
PROFILE +=	-DTEST
else
ifdef EVAL_DUMP
OBJS	+=	testeval.o testsup.o
endif
endif
ifdef TREEDUMP
OBJS	+=	dumptree.o
endif
ifdef MP
OBJS	+=	split.o
endif

.SUFFIXES:	.c .o
.SUFFIXES:	.cpp .o
.SUFFIXES:	.asm .o

.PHONY:		all
all:		$(BINARY)

$(BINARY):	$(OBJS)
		$(CXX) $(PROFILE) -flto -o $(BINARY) -Wall $(LIBRARIES) \
		$(OBJS)

%.o:        %.c $(HEADERS)
		$(CC) $(CFLAGS) -c $< -o $@

fathom.o:	fathom/tbprobe.c fathom/tbchess.c fathom/tbconfig.h fathom/tbprobe.h fathom/stdendian.h
		$(CC) $(FATHOM_CFLAGS) -c fathom/tbprobe.c -o $@

%.o:        %.cpp $(HEADERS)
		$(CXX) $(CFLAGS) -c -w $< -o $@

%.o:        %.asm $(HEADERS)
		$(NASM) $(NASMFLAGS) -o $@ $<

.PHONY:		release
release:
		$(MAKE) -j5 MP=1 GENETIC=1

.PHONY:		debug
debug:
		$(MAKE) -j5 DEBUG=1 PERF_COUNTERS=1 GENETIC=1

.PHONY:		clean
clean:
		$(RM) -f *.o *.core *~ *.gmon \#*\# typhoon _typhoon