performance improvement

.gitignore

@@ -2,3 +2,4 @@
 test*
 obj/
 hexview
+profile/

Makefile

@@ -11,20 +11,37 @@ LDFLAGS+= -lfmt
 PROJ = hexview
 
 OBJ = $(addprefix obj/,$(SRC:.cpp=.o))
+PROF_OBJ = $(addprefix profile/,$(SRC:.cpp=.o))
 
 .PHONY: all
 all: dirs $(PROJ)
 
 dirs:
 	@mkdir -p obj/src
+	@mkdir -p profile/src
 
 obj/%.o: %.cpp
 	$(CXX) -c $< $(_HX_FEAT) $(CXXFLAGS) -o $@ $(LDFLAGS)
 
+profile/%.o: %.cpp
+	$(CXX) -c $< $(_HX_FEAT) $(CXXFLAGS) -fprofile-generate -o $@ $(LDFLAGS)
+
 $(PROJ): $(OBJ)
 	$(CXX) $^ $(CXXFLAGS) -o $@ $(LDFLAGS)
 	strip $@
 
+pgo-generate: $(PROF_OBJ)
+	$(CXX) $^ $(CXXFLAGS) -fprofile-generate -o $@ $(LDFLAGS)
+
+pgo-profile: pgo-generate
+	for i in {1..32}; do \
+		dd if=/dev/urandom of=./pgo-test bs=1024 count=1024 \
+		./pgo-generate ./pgo-test
+	done
+	rm ./pgo-test
+
 clean:
 	rm -f $(PROJ)
 	rm -rf obj
+	rm -rf profile
+	rm -f pgo-generate

src/hex.cpp

@@ -53,12 +53,18 @@ inline __attribute__((always_inline)) void prints(const S&... strings)
 	(fputs((const char*)strings, stdout), ...);
 }
 
-inline void hex02(std::uint8_t byte, char buffer[3])
+inline void hex02(std::uint8_t byte, char buffer[2])
 {
 	buffer[0] =hex_map[byte][0];
 	buffer[1] =hex_map[byte][1];
 }
 
+template<int N>
+inline __attribute__((always_inline)) void print_exact(const char (&ar)[N])
+{
+	fwrite(ar, N, 1, stdout);
+}
+
 namespace hv {
 	void print_screen(const span<unsigned char> memory, unsigned long offset)
 	{
@@ -71,28 +77,34 @@ namespace hv {
 		fmt::print("0x{:016x}  ", offset);
 		char hxbuf[3];
 		std::size_t i=0;
-		char ascii[1 + ROW_SZ];
+		char r_ascii[2 + ROW_SZ];
+		r_ascii[0] = ' ';
+		r_ascii[1] = ' ';
+		char* ascii = r_ascii + 2;
 			
-		ascii[ROW_SZ] = 0;
-		hxbuf[2] = 0;
+		hxbuf[2] = ' ';
 		for(;i<memory.size();i++) {
 			if (i && i % ROW_SZ == 0) {
-				prints("  ", S ascii);
+				print_exact(r_ascii);
 				fmt::print("\n0x{:016x}  ", i+offset);
 			}
 			unsigned char idx = memory[i];
 			hex02(idx, hxbuf);
-			prints(hxbuf, " ");
+			print_exact(hxbuf);
 			ascii[i%ROW_SZ] = ascii_map[idx];
 		}
 		if (memory.size() % ROW_SZ != 0)
 		{
 			auto rest = memory.size() % ROW_SZ;
 			ascii[rest] = 0;
+			constexpr const char output[3] = { ' ', ' ', ' ' };
 			for(std::size_t j=0;j< ROW_SZ - rest;j++)
-				prints("   ");
+				print_exact(output);
 
+			prints(r_ascii);
+		}
+		else {
+			print_exact(r_ascii);
 		}
-		fmt::print("  {}", S ascii);    
 	}
 }