diff --git a/base.mk b/base.mk
index 6d99611..b6e3f3f 100644
--- a/base.mk
+++ b/base.mk
@@ -103,10 +103,10 @@ endef
 	$(AS) $(AARCH) $(AFLAGS) $(DEBUG) $< -o $@
 
 %.o: %.c
-	$(CC) $(CARCH) $(CFLAGS) -std=c23 $(OPT) $(DEBUG) $(DEPFLAGS) -MF ${<}.d -c $< -o $@
+	$(CC) $(CARCH) $(CFLAGS) -std=gnu23 $(OPT) $(DEBUG) $(DEPFLAGS) -MF ${<}.d -c $< -o $@
 
 %.o: %.cpp
-	$(CXX) $(CARCH) $(CFLAGS) $(CXXFLAGS) $(OPT) $(DEBUG) $(DEPFLAGS) -MF ${<}.d -c $< -o $@
+	$(CXX) $(CARCH) $(CFLAGS) -std=c++23 $(CXXFLAGS) $(OPT) $(DEBUG) $(DEPFLAGS) -MF ${<}.d -c $< -o $@
 
 %.elf:
 	$(LD) $(LDFLAGS) -L $(LIB) -T $(LDSCRIPT) $^ -o $@
diff --git a/common.mk b/common.mk
index cebe704..03ad6bd 100644
--- a/common.mk
+++ b/common.mk
@@ -3,11 +3,9 @@ GENERATED ?=
 
 AARCH = --isa=sh4 --little
 
-CARCH = -m4-single-only -ml
+CARCH ?= -m4-single-only -ml
 CFLAGS += -mfsca -funsafe-math-optimizations -ffast-math
 
-CXXFLAGS += -std=c++23
-
 OBJARCH = -O elf32-shl -B sh4
 
 TARGET = sh4-none-elf-
diff --git a/example/example.mk b/example/example.mk
index 5c44e08..89d71d5 100644
--- a/example/example.mk
+++ b/example/example.mk
@@ -388,9 +388,9 @@ SERIAL_TRANSFER_OBJ = \
 	example/serial_transfer.o \
 	sh7091/serial.o \
 	serial_load.o \
-	maple/maple.o \
-	font/portfolio_6x8/portfolio_6x8.data.o \
 	crc32.o
+#	maple/maple.o \
+#	font/portfolio_6x8/portfolio_6x8.data.o \
 
 example/serial_transfer.elf: LDSCRIPT = $(LIB)/loader.lds
 example/serial_transfer.elf: $(START_OBJ) $(SERIAL_TRANSFER_OBJ)
diff --git a/example/serial_transfer.cpp b/example/serial_transfer.cpp
index 406d739..529f8d0 100644
--- a/example/serial_transfer.cpp
+++ b/example/serial_transfer.cpp
@@ -405,11 +405,13 @@ void main()
   constexpr uint32_t serial_speed = 0;
   serial_load::init(serial_speed);
 
+  /*
   const uint8_t * font = reinterpret_cast<const uint8_t *>(&_binary_font_portfolio_6x8);
   auto renderer0 = maple::display::font_renderer(font);
   framebuffer[0] = renderer0.fb;
   auto renderer1 = maple::display::font_renderer(font);
   framebuffer[1] = renderer1.fb;
+  */
 
   // reset serial status
   sh7091.SCIF.SCFSR2 = 0;
@@ -463,10 +465,11 @@ void main()
       }
     }
 
+    //serial::string(".");
     serial_load::tick(poll_state);
-    render(renderer0, renderer1);
+    //render(renderer0, renderer1);
 
-    poll_state.want_start = 1;
-    handle_maple(poll_state);
+    //poll_state.want_start = 1;
+    //handle_maple(poll_state);
   }
 }
diff --git a/serial_load.cpp b/serial_load.cpp
index fde73b9..0eb6da5 100644
--- a/serial_load.cpp
+++ b/serial_load.cpp
@@ -2,10 +2,14 @@
 
 #include "sh7091/serial.hpp"
 #include "sh7091/serial_dma.hpp"
+#include "sh7091/cache.hpp"
 
 #include "serial_load.hpp"
 #include "crc32.h"
 
+#include "align.hpp"
+#include "memory.hpp"
+
 namespace serial_load {
 
 struct state state;
@@ -162,7 +166,7 @@ void recv(struct maple_poll_state& poll_state, uint8_t c)
 	}
       }
       // invalid command
-      move(&state.buf.u8[0], &state.buf.u8[1], state.len - 1);
+      memory::move(&state.buf.u8[0], &state.buf.u8[1], state.len - 1);
       state.len -= 1;
       break;
     }
@@ -185,6 +189,16 @@ void tick(struct maple_poll_state& poll_state)
       if (dar1 > state.reply_crc.offset) {
 	uint32_t len = dar1 - state.reply_crc.offset;
 	const uint8_t * buf = reinterpret_cast<const uint8_t *>(state.reply_crc.offset);
+        const uint8_t * buf32 = reinterpret_cast<const uint8_t *>((state.reply_crc.offset / 32) * 32); // round down
+
+        // purge operand cache blocks for the data written by DMA, rounding up twice
+        for (uint32_t i = 0; i < align_32byte(len) + 32; i += 32) {
+          asm volatile ("ocbp @%0"
+                        :                                             // output
+                        : "r" (reinterpret_cast<uint32_t>(&buf32[i])) // input
+                        );
+        }
+
 	state.reply_crc.value = crc32_update(state.reply_crc.value, buf, len);
 	state.reply_crc.offset += len;
       }
@@ -248,6 +262,9 @@ void tick(struct maple_poll_state& poll_state)
       if ((sh7091.SCIF.SCFSR2 & transmission_end) != transmission_end)
 	return;
 
+      // clear the cache before jumping
+      cache::init();
+
       const uint32_t dest = state.buf.arg[0];
       jump_to_func(dest);
 
diff --git a/sh7091/cache.cpp b/sh7091/cache.cpp
index ee12709..a99bf57 100644
--- a/sh7091/cache.cpp
+++ b/sh7091/cache.cpp
@@ -12,10 +12,12 @@ namespace cache {
 void init()
 {
   for (int i = 0; i < 256; i++) {
+    // clear all V bits of the IC address array
     sh7091_ic_a[i][0] = 0;
   }
 
   for (int i = 0; i < 512; i++) {
+    // clear all V bits of the OC address array
     sh7091_oc_a[i][0] = 0;
   }