initial commit

start: day1p1

Fortune for aoc2021's current commit: Half blessing − 半吉

.gitignore

@@ -0,0 +1,3 @@
+obj/
+day*-*
+day*/input

day1/Makefile

@@ -0,0 +1,90 @@
+# Vectorised rot13ing data
+# Contains targets for `release', `debug', and `clean'.
+
+PROJECT=day1
+AUTHOR=Avril (Flanchan) <flanchan@cumallover.me>
+
+SRC_C   = $(wildcard src/*.c)
+SRC_CXX = $(wildcard src/*.cpp)
+
+INCLUDE=include
+
+COMMON_FLAGS+= -W -Wall -fno-strict-aliasing $(addprefix -I,$(INCLUDE))
+
+COMMON_FLAGS+=-msse -msse2 -msse3
+
+OPT_FLAGS?= -march=native -fgraphite -fopenmp -floop-parallelize-all -ftree-parallelize-loops=4 \
+	    -floop-interchange -ftree-loop-distribution -floop-strip-mine -floop-block \
+	    -fno-stack-check
+
+CXX_OPT_FLAGS?= $(OPT_FLAGS) -felide-constructors
+
+CFLAGS   += $(COMMON_FLAGS) --std=gnu17
+CXXFLAGS += $(COMMON_FLAGS) --std=gnu++20
+LDFLAGS  += -ltracem
+
+STRIP=strip
+
+RELEASE_COMMON_FLAGS+= -Werror
+DEBUG_COMMON_FLAGS+= -fanalyzer 
+
+ifneq ($(TARGET_SPEC_FLAGS),no)
+	RELEASE_CFLAGS?=   -O3 -flto $(OPT_FLAGS)
+	RELEASE_CXXFLAGS?= -O3 -flto $(CXX_OPT_FLAGS)
+	RELEASE_LDFLAGS?=  -Wl,-O3 -Wl,-flto
+
+	DEBUG_CFLAGS?=	-Og -g
+	DEBUG_CXXFLAGS?=-Og -g
+	DEBUG_LDFLAGS?=
+endif
+
+DEBUG_CFLAGS+=-DDEBUG $(DEBUG_COMMON_FLAGS)
+DEBUG_CXXFLAGS+=-DDEBUG $(DEBUG_COMMON_FLAGS)
+
+RELEASE_CFLAGS+=-DRELEASE $(RELEASE_COMMON_FLAGS)
+RELEASE_CXXFLAGS+=-DRELEASE $(RELEASE_COMMON_FLAGS)
+
+# Objects
+
+OBJ_C   = $(addprefix obj/c/,$(SRC_C:.c=.o))
+OBJ_CXX = $(addprefix obj/cxx/,$(SRC_CXX:.cpp=.o))
+OBJ = $(OBJ_C) $(OBJ_CXX)
+
+# Phonies
+
+.PHONY: release
+release: | dirs $(PROJECT)-release
+
+.PHONY: debug
+debug: | dirs $(PROJECT)-debug
+
+# Targets
+
+dirs:
+	@mkdir -p obj/c{,xx}/src
+
+obj/c/%.o: %.c
+	$(CC) -c $< $(CFLAGS) -o $@ $(LDFLAGS)
+
+obj/cxx/%.o: %.cpp
+	$(CXX) -c $< $(CXXFLAGS) -o $@ $(LDFLAGS)
+
+$(PROJECT)-release: CFLAGS+= $(RELEASE_CFLAGS)
+$(PROJECT)-release: CXXFLAGS += $(RELEASE_CXXFLAGS)
+$(PROJECT)-release: LDFLAGS += $(RELEASE_LDFLAGS)
+$(PROJECT)-release: $(OBJ)
+	$(CXX) $^ $(CXXFLAGS) -o $@ $(LDFLAGS)
+	$(STRIP) $@
+
+$(PROJECT)-debug: CFLAGS+= $(DEBUG_CFLAGS)
+$(PROJECT)-debug: CXXFLAGS += $(DEBUG_CXXFLAGS)
+$(PROJECT)-debug: LDFLAGS += $(DEBUG_LDFLAGS)
+$(PROJECT)-debug: $(OBJ)
+	$(CXX) $^ $(CXXFLAGS) -o $@ $(LDFLAGS)
+
+clean-rebuild:
+	rm -rf obj
+
+clean: clean-rebuild
+	rm -f $(PROJECT)-{release,debug,pgo}
+

day1/include/input.h

@@ -0,0 +1,22 @@
+#ifndef _INPUT_H
+#define _INPUT_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stddef.h>
+#include <stdint.h>
+
+typedef uint16_t input_t;
+
+#define INPUT_VECTOR_STEP (16 / sizeof(input_t))
+
+extern const size_t  INPUT_SIZE;
+extern const input_t INPUT[];
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _INPUT_H */

day1/include/vector.h

@@ -0,0 +1,71 @@
+#ifndef _VECTOR_H
+#define _VECTOR_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stdint.h>
+
+#include <mmintrin.h>
+#include <immintrin.h>
+#include <tmmintrin.h>
+#include <xmmintrin.h>
+#include <emmintrin.h>
+
+#define _VECTOR_SIZE 16
+
+// Vector128 int 8
+typedef uint8_t  v8_u128n __attribute__((vector_size(_VECTOR_SIZE)));
+typedef int8_t   v8_i128n __attribute__((vector_size(_VECTOR_SIZE)));
+
+// Vector128 int 16
+typedef uint16_t v16_u128n __attribute__((vector_size(_VECTOR_SIZE)));
+typedef int16_t  v16_i128n __attribute__((vector_size(_VECTOR_SIZE)));
+
+// Vector128 int 32 
+typedef uint32_t v32_u128n __attribute__((vector_size(_VECTOR_SIZE)));
+typedef int32_t  v32_i128n __attribute__((vector_size(_VECTOR_SIZE)));
+
+// Vector128 int 64
+typedef uint64_t v64_u128n __attribute__((vector_size(_VECTOR_SIZE)));
+typedef int64_t  v64_i128n __attribute__((vector_size(_VECTOR_SIZE)));
+
+typedef union v128 {
+	// Intrinsics
+	__m128i mm; // long int
+	__m128d mmd; // double
+	__m128  mmf; // float //XXX: Should we have the halfs
+
+	// Vectorised integers
+	union {
+		v8_u128n u8;
+		v8_i128n i8;
+
+		v16_u128n u16;
+		v16_i128n i16;
+
+		v32_u128n u32;
+		v32_i128n i32;
+
+		v64_u128n u64;
+		v64_i128n i64;
+	} as;
+
+	// Native 128 bit integers
+	unsigned __int128 u128;
+	signed   __int128  i128;
+
+	// Raw bytes
+	uint8_t bytes[_VECTOR_SIZE];
+} v128_t;
+
+#ifdef __cplusplus
+}
+static_assert(sizeof(union v128)==_VECTOR_SIZE, "invalid vector size");
+static_assert(alignof(union v128)==alignof(__m128i), "invalid vector alignment");
+#else
+_Static_assert(sizeof(union v128)==_VECTOR_SIZE, "invalid vector size");
+_Static_assert(_Alignof(union v128)==_Alignof(__m128i), "invalid vector alignment");
+#endif
+#endif /* _VECTOR_H */

day1/mkinput

@@ -0,0 +1,15 @@
+#!/bin/bash
+
+INPUT=${INPUT:-input}
+
+NUM=$(wc -l < ${INPUT})
+
+echo "#include <input.h>"
+echo ""
+echo "const input_t INPUT[${NUM}] = {"
+while IFS= read -r line; do
+	echo "	$line, "
+done < $INPUT
+unset line
+echo "};"
+echo "const size_t INPUT_SIZE=${NUM};"

day1/src/vcomp.c

@@ -0,0 +1,69 @@
+#include <assert.h>
+
+
+#include <input.h>
+#include <vector.h>
+
+#include <tracem/ints.h>
+
+_Static_assert(sizeof(input_t)*INPUT_VECTOR_STEP == sizeof(v128_t), "invalid vector size mul");
+
+#define INPUT_VPROC_STEP 15
+
+__attribute__((pure))
+u64 vproc(usize len, const input_t input[const restrict len])
+{
+	static const v128_t SINGLE = { .as.i16 = { -1, -1, -1, -1, -1, -1, -1, -1 } };
+	static const v128_t ONES   = { .as.i16 = {  1,  1,  1,  1,  1,  1,  1,  1 } };
+#if DEBUG
+	assert(len % INPUT_VPROC_STEP == 0);
+#endif
+
+#define LAST(j) input[i +  (j)]
+#define NEXT(j) input[i + ((j) * 2)]
+	__m128i end = _mm_setzero_si128();
+	for(usize i = 0;i<len;i+=INPUT_VPROC_STEP)
+	{
+		// load the vectors
+		v128_t prev = { .as.u16 = {
+				input[i+0],
+				input[i+2],
+
+				input[i+4],
+				input[i+6],
+
+				input[i+8],
+				input[i+10],
+
+				input[i+12],
+				input[i+14],
+			}
+		};
+		v128_t next = { .as.u16 = {
+				input[i+1],
+				input[i+3],
+
+				input[i+5],
+				input[i+7],
+
+				input[i+9],
+				input[i+11],
+
+				input[i+13],
+				input[i+15],
+			}
+		};
+		//__m128i res = _mm_subs_epi16(prev.mm, next.mm);
+		__m128i pltn = _mm_cmplt_epi16(prev.mm, next.mm); // prev < next
+		pltn = _mm_sign_epi16(pltn, SINGLE.mm);
+		pltn = _mm_sign_epi16(ONES.mm, pltn);
+		end = _mm_add_epi16(end, pltn);
+		//__m128i pr =  _mm_sign_epi16(FULL_MASK.mm, res);
+	}
+	
+	v16_u128n _ev = ((union v128)end).as.u16;
+	return (u64) (    _ev[0] + _ev[1] 
+			+ _ev[2] + _ev[3]
+			+ _ev[4] + _ev[5]
+			+ _ev[6] + _ev[7] );
+}