diff --git a/Makefile b/Makefile
index 2254f7a..145eee5 100644
--- a/Makefile
+++ b/Makefile
@@ -1,4 +1,9 @@
-all:
+all: $(patsubst %.cpp,%.elf,$(wildcard example/*.cpp))
+
+phony:
+
+example/arm/%.bin: phony
+	make -C example/arm $*.bin
 
 include common.mk
 
@@ -7,3 +12,5 @@ geometry/%.hpp: geometry/%.obj
 	mv $@.tmp $@
 
 include example/example.mk
+
+.PHONY: phony
diff --git a/base.mk b/base.mk
index f28e5a6..2b555ea 100644
--- a/base.mk
+++ b/base.mk
@@ -4,7 +4,8 @@ AFLAGS += --fatal-warnings
 
 CFLAGS += -falign-functions=4 -ffunction-sections -fdata-sections -fshort-enums -ffreestanding -nostdlib
 CFLAGS += -Wall -Werror -Wfatal-errors
-CFLAGS += -Wno-error=narrowing -Wno-error=unused-variable -Wno-error=array-bounds= -Wno-array-bounds
+CFLAGS += -Wno-array-bounds
+#CFLAGS += -Wno-error=narrowing -Wno-error=unused-variable -Wno-error=array-bounds=
 
 CXXFLAGS += -fno-exceptions -fno-non-call-exceptions -fno-rtti -fno-threadsafe-statics
 
diff --git a/common.mk b/common.mk
index 4e7986e..6dba415 100644
--- a/common.mk
+++ b/common.mk
@@ -79,7 +79,9 @@ sine.pcm: common.mk
 		/1ST_READ.BIN=./$< \
 		/=./COPYRIGH.TXT \
 		/=./ABSTRACT.TXT \
-		/=./BIBLIOGR.TXT
+		/=./BIBLIOGR.TXT \
+		/DUNE.PCM=./dune.pcm \
+		/REIGN.PCM=./reign.pcm
 
 %.cdi: %.iso
 	./cdi4dc $< $@ >/dev/null
diff --git a/example/aica_gdrom.cpp b/example/aica_gdrom.cpp
index 803a672..dc8232e 100644
--- a/example/aica_gdrom.cpp
+++ b/example/aica_gdrom.cpp
@@ -1,4 +1,6 @@
 #include "memorymap.hpp"
+#include "sh7091/sh7091.hpp"
+#include "sh7091/sh7091_bits.hpp"
 #include "sh7091/serial.hpp"
 #include "systembus.hpp"
 #include "systembus_bits.hpp"
@@ -73,6 +75,10 @@ void aica_init(uint32_t& chunk_index, const uint32_t * src_chunk)
   chunk_index = (chunk_index + 1) % 2;
 
   aica_wait_write(); aica_sound.common.vreg_armrst = aica::vreg_armrst::ARMRST(0);
+
+  { // send arm interrupt
+    aica_sound.common.scipd = scipd__arm_interrupt;
+  }
 }
 
 void aica_step(uint32_t& chunk_index, const uint32_t * src_chunk)
@@ -83,6 +89,8 @@ void aica_step(uint32_t& chunk_index, const uint32_t * src_chunk)
     aica_wait_write(); aica_sound.common.mcire = mcipd__sh4_interrupt;
   }
 
+  const uint32_t step_start = sh7091.TMU.TCNT0;
+
   { // fill the requested chunk
     aica_fill_chunk(&(*chunk)[chunk_index][0],
                     src_chunk,
@@ -90,6 +98,16 @@ void aica_step(uint32_t& chunk_index, const uint32_t * src_chunk)
 
     chunk_index = (chunk_index + 1) % 2;
   }
+
+  { // send arm interrupt
+    aica_sound.common.scipd = scipd__arm_interrupt;
+  }
+
+
+  const uint32_t step_end = sh7091.TMU.TCNT0;
+  const uint32_t step_time = step_start - step_end;
+  serial::string("step: ");
+  serial::integer<uint32_t>(step_time);
 }
 
 // gdrom
@@ -175,19 +193,21 @@ uint32_t gdrom_cd_read2(uint16_t * buf,
   //serial::integer<uint32_t>(next_address);
   gdrom_pio_data(packet._data());
 
+  while ((gdrom::status::bsy(gdrom_if.status)) != 0); // wait for drive to become not-busy
+
   uint32_t length = 0;
   while ((gdrom::status::drq(gdrom_if.status)) != 0) {
     const uint32_t byte_count = gdrom_if.byte_count();
     length += byte_count;
     gdrom_read_data(buf, byte_count);
 
-    serial::string("read status: ");
+    serial::string("status: ");
     serial::integer<uint8_t>(gdrom_if.status);
 
     while ((gdrom::status::bsy(gdrom_if.status)) != 0); // wait for drive to become not-busy
   }
 
-  serial::string("read length: ");
+  serial::string("length: ");
   serial::integer<uint32_t>(length);
   return length;
 }
@@ -237,11 +257,11 @@ struct extent gdrom_find_file()
 
   const uint32_t primary_volume_descriptor = fad + 16;
   uint16_t buf[2048 / 2];
-  const uint32_t length0 = gdrom_cd_read2(buf,
-                                          primary_volume_descriptor,     // starting address
-                                          1,                             // one sector; 2048 bytes
-                                          primary_volume_descriptor + 1  // next address
-                                          );
+  gdrom_cd_read2(buf,
+		 primary_volume_descriptor,     // starting address
+		 1,                             // one sector; 2048 bytes
+		 primary_volume_descriptor + 1  // next address
+		 );
   serial::character('\n');
 
   auto pvd = reinterpret_cast<const iso9660::primary_volume_descriptor *>(&buf[0]);
@@ -260,11 +280,11 @@ struct extent gdrom_find_file()
   serial::character('\n');
 
   const uint32_t root_directory_extent = root_dr->location_of_extent.get();
-  const uint32_t length1 = gdrom_cd_read2(buf,
-                                          root_directory_extent + 150, // 150?
-                                          1, // one sector; 2048 bytes
-                                          root_directory_extent + 151  // 150?
-                                          );
+  gdrom_cd_read2(buf,
+		 root_directory_extent + 150, // 150?
+		 1, // one sector; 2048 bytes
+		 root_directory_extent + 151  // 150?
+		 );
   serial::character('\n');
 
   auto buf8 = reinterpret_cast<const uint8_t *>(buf);
@@ -311,11 +331,18 @@ struct extent gdrom_find_file()
 
 void gdrom_read_chunk(uint32_t * buf, const uint32_t extent, const uint32_t num_extents)
 {
-  const uint32_t length1 = gdrom_cd_read2(reinterpret_cast<uint16_t *>(buf),
-                                          extent + 150, // 150?
-                                          num_extents,  // one sector; 2048 bytes
-                                          extent + 150 + num_extents  // 150?
-                                          );
+  const uint32_t gdrom_start = sh7091.TMU.TCNT0;
+
+  gdrom_cd_read2(reinterpret_cast<uint16_t *>(buf),
+		 extent + 150, // 150?
+		 num_extents,  // one sector; 2048 bytes
+		 extent + 150 + num_extents  // 150?
+		 );
+
+  const uint32_t gdrom_end = sh7091.TMU.TCNT0;
+  const uint32_t gdrom_time = gdrom_start - gdrom_end;
+  serial::string("time: ");
+  serial::integer<uint32_t>(gdrom_time);
 }
 
 void next_segment(const struct extent& extent, uint32_t& segment_index)
@@ -327,6 +354,13 @@ void next_segment(const struct extent& extent, uint32_t& segment_index)
 
 void main()
 {
+  sh7091.TMU.TSTR = 0; // stop all timers
+  sh7091.TMU.TOCR = tmu::tocr::tcoe::tclk_is_external_clock_or_input_capture;
+  sh7091.TMU.TCR0 = tmu::tcr0::tpsc::p_phi_256; // 256 / 50MHz = 5.12 μs ; underflows in ~1 hour
+  sh7091.TMU.TCOR0 = 0xffff'ffff;
+  sh7091.TMU.TCNT0 = 0xffff'ffff;
+  sh7091.TMU.TSTR = tmu::tstr::str0::counter_start;
+
   uint32_t chunk_index = 0;
   uint32_t segment_index = 0;
 
@@ -338,13 +372,12 @@ void main()
 
   aica_init(chunk_index, &gdrom_buf[0]);
 
-  serial::string("aica wave memory:\n");
-  while (aica_wave_memory[0] == 0xeaffffff) { aica_wait_read(); };
-  aica_wait_read(); serial::integer<uint32_t>(aica_wave_memory[0]);
-  aica_wait_read(); serial::integer<uint32_t>(aica_wave_memory[1]);
+  //serial::string("aica wave memory:\n");
+  //while (aica_wave_memory[0] == 0xeaffffff) { aica_wait_read(); };
+  //aica_wait_read(); serial::integer<uint32_t>(aica_wave_memory[0]);
+  //aica_wait_read(); serial::integer<uint32_t>(aica_wave_memory[1]);
 
   while (1) {
-    //serial::integer<uint8_t>(chunk_index);
     gdrom_read_chunk(gdrom_buf, extent.location + segment_index, sectors_per_chunk);
     next_segment(extent, segment_index);
 
diff --git a/example/arm/sh4_interrupt.cpp b/example/arm/sh4_interrupt.cpp
index 04cb4ad..66c3703 100644
--- a/example/arm/sh4_interrupt.cpp
+++ b/example/arm/sh4_interrupt.cpp
@@ -39,20 +39,25 @@ void main()
   aica_sound.channel[0].LPCTL(1);
   aica_sound.channel[0].PCMS(0);
   aica_sound.channel[0].LSA(0);
-  aica_sound.channel[0].LEA(chunk_size / 2);
+  aica_sound.channel[0].LEA((chunk_size / 2) * 2);
   aica_sound.channel[0].D2R(0x0);
   aica_sound.channel[0].D1R(0x0);
-  aica_sound.channel[0].RR(0x0);
+  aica_sound.channel[0].RR(0x1f);
   aica_sound.channel[0].AR(0x1f);
 
-  aica_sound.channel[0].OCT(0);
-  aica_sound.channel[0].FNS(0);
+  aica_sound.channel[0].OCT(0xf);
+  aica_sound.channel[0].FNS(0x7d);
   aica_sound.channel[0].DISDL(0xf);
   aica_sound.channel[0].DIPAN(0x0);
 
   aica_sound.channel[0].Q(0b00100);
   aica_sound.channel[0].TL(0);
-  aica_sound.channel[0].LPOFF(1);
+  aica_sound.channel[0].LPOFF(0);
+  aica_sound.channel[0].FLV0(0x1ff8);
+  aica_sound.channel[0].FLV1(0x1ff8);
+  aica_sound.channel[0].FLV2(0x1ff8);
+  aica_sound.channel[0].FLV3(0x1ff8);
+  aica_sound.channel[0].FLV4(0x1ff8);
 
   aica_sound.common.mono_mem8mb_dac18b_ver_mvol =
       aica::mono_mem8mb_dac18b_ver_mvol::MONO(0)   // enable panpots
@@ -61,33 +66,51 @@ void main()
     | aica::mono_mem8mb_dac18b_ver_mvol::MVOL(0xf) // 15/15 volume
     ;
 
-  constexpr uint32_t tactl = 7;        // increment once every 1 samples
-  constexpr uint32_t tima = 256 - 128; // interrupt after 128 samples
+  //constexpr uint32_t tactl = 6;        // increment once every 1 samples
+  //constexpr uint32_t tima = 0; // interrupt after 128 samples
 
   dram[0] = reinterpret_cast<uint32_t>(&chunk[0][0]);
   dram[1] = reinterpret_cast<uint32_t>(&chunk[1][0]);
 
-  uint32_t next_chunk = 0;
-  aica_sound.channel[0].SA(reinterpret_cast<const uint32_t>(&chunk[next_chunk][0]));
+  aica_sound.channel[0].SA(reinterpret_cast<const uint32_t>(&chunk[0][0]));
   aica_sound.channel[0].KYONEX(1);
-  next_chunk = (next_chunk + 1) % 2;
-  request_chunk();
 
-  constexpr uint32_t timer_a_interrupt = (1 << 6);
-  aica_sound.common.tactl_tima = aica::tactl_tima::TACTL(tactl)
-			       | aica::tactl_tima::TIMA(tima);
-  aica_sound.common.scire = timer_a_interrupt;
+  //constexpr uint32_t scipd__arm_interrupt = (1 << 5);
+  //constexpr uint32_t timer_a_interrupt = (1 << 6);
+
+  uint8_t next_chunk = 1;
 
-  uint32_t sample = 0;
   while (1) {
-    if (aica_sound.common.SCIPD() & timer_a_interrupt) {
+    // detect buffer underrun
+    /*
+    if (!(aica_sound.common.SCIPD() & scipd__arm_interrupt)) {
+      //aica_sound.channel[0].KYONB(0);
+      aica_sound.channel[0].KYONEX(1);
+      while (!(aica_sound.common.SCIPD() & scipd__arm_interrupt));
+      aica_sound.channel[0].KYONB(1);
       aica_sound.channel[0].SA(reinterpret_cast<const uint32_t>(&chunk[next_chunk][0]));
-      aica_sound.common.tactl_tima = aica::tactl_tima::TACTL(tactl)
-        | aica::tactl_tima::TIMA(tima);
-      next_chunk = (next_chunk + 1) % 2;
-      request_chunk();
+      aica_sound.channel[0].KYONEX(1);
+      aica_sound.channel[0].SA(reinterpret_cast<const uint32_t>(&chunk[0][0]));
+    }
+    aica_sound.common.scire = scipd__arm_interrupt;
+    */
 
-      aica_sound.common.scire = timer_a_interrupt;
+    next_chunk = !next_chunk;
+    request_chunk();
+
+    uint32_t sample = 0;
+    const uint32_t target = 16384 + (16384 * 10 / 12);
+    constexpr uint32_t sample_interval = (1 << 10);
+    while (sample < target) {
+      //aica_sound.common.tactl_tima = aica::tactl_tima::TACTL(tactl)
+      //                           | aica::tactl_tima::TIMA(tima);
+
+      //while (!(aica_sound.common.SCIPD() & timer_a_interrupt));
+      //aica_sound.common.scire = timer_a_interrupt;
+      while (!(aica_sound.common.SCIPD() & sample_interval));
+      aica_sound.common.scire = sample_interval;
+
+      sample++;
     }
   }
 }
diff --git a/example/dump_ram.cpp b/example/dump_ram.cpp
index 985e78f..2eacbd0 100644
--- a/example/dump_ram.cpp
+++ b/example/dump_ram.cpp
@@ -7,7 +7,7 @@
 void dump_ram(const volatile uint32_t * mem, const uint32_t len)
 {
   uint32_t sum = 0;
-  for (uint32_t i = 0; i < ; i++) {
+  for (uint32_t i = 0; i < len; i++) {
     uint8_t n = mem[i];
     sum += n;
     serial::hexlify(n);
diff --git a/example/gdrom_iso9660.cpp b/example/gdrom_iso9660.cpp
index e03e5cd..5998d06 100644
--- a/example/gdrom_iso9660.cpp
+++ b/example/gdrom_iso9660.cpp
@@ -116,10 +116,10 @@ void main()
 
   const uint32_t primary_volume_descriptor = fad + 16;
   uint16_t buf[2048 / 2];
-  const uint32_t length0 = cd_read(buf,
-                                   primary_volume_descriptor,
-                                   1 // one sector; 2048 bytes
-                                   );
+  cd_read(buf,
+	  primary_volume_descriptor,
+	  1 // one sector; 2048 bytes
+	  );
   serial::character('\n');
 
   auto pvd = reinterpret_cast<const iso9660::primary_volume_descriptor *>(&buf[0]);
@@ -148,10 +148,10 @@ void main()
   serial::character('\n');
 
   const uint32_t root_directory_extent = root_dr->location_of_extent.get();
-  const uint32_t length1 = cd_read(buf,
-                                   root_directory_extent + 150, // 150?
-                                   1 // one sector; 2048 bytes
-                                   );
+  cd_read(buf,
+	  root_directory_extent + 150, // 150?
+	  1 // one sector; 2048 bytes
+	  );
   serial::character('\n');
 
   auto buf8 = reinterpret_cast<const uint8_t *>(buf);
@@ -182,10 +182,10 @@ void main()
 	const uint32_t extent = dr->location_of_extent.get();
 
         uint16_t buf2[2048 / 2];
-        const uint32_t length1 = cd_read(buf2,
-                                         extent + 150, // 150?
-                                         1 // one sector; 2048 bytes
-                                         );
+	cd_read(buf2,
+		extent + 150, // 150?
+		1 // one sector; 2048 bytes
+		);
 
 	auto file = reinterpret_cast<const uint8_t *>(&buf2[0]);
 	serial::string("---begin file content---\n");
diff --git a/example/gdrom_test.cpp b/example/gdrom_test.cpp
index 26dc028..9f92f19 100644
--- a/example/gdrom_test.cpp
+++ b/example/gdrom_test.cpp
@@ -60,9 +60,9 @@ void pio_data(const uint8_t * data)
   serial::string("status2: ");
   serial::integer<uint8_t>(gdrom_if.status);
 
-  serial::string("byte_control: ");
-  serial::integer<uint8_t>(gdrom_if.byte_control_high, ' ');
-  serial::integer<uint8_t>(gdrom_if.byte_control_low);
+  serial::string("byte_count: ");
+  serial::integer<uint8_t>(gdrom_if.byte_count_high, ' ');
+  serial::integer<uint8_t>(gdrom_if.byte_count_low);
 }
 
 void read_data(uint32_t length)
@@ -148,7 +148,7 @@ void get_toc()
 
   serial::string("\nget_toc\n");
   pio_data(data);
-  const uint32_t length = (gdrom_if.byte_control_high << 8) | (gdrom_if.byte_control_low << 0);
+  const uint32_t length = (gdrom_if.byte_count_high << 8) | (gdrom_if.byte_count_low << 0);
   // 102 entries ; 4 bytes per entry, 408 bytes (0x0198)
   read_data(length);
 }
@@ -210,14 +210,14 @@ void cd_read2()
     serial::string("offset: ");
     serial::integer<uint32_t>(read);
 
-    const uint32_t length = (gdrom_if.byte_control_high << 8) | (gdrom_if.byte_control_low << 0);
+    const uint32_t length = (gdrom_if.byte_count_high << 8) | (gdrom_if.byte_count_low << 0);
     // 102 entries ; 4 bytes per entry, 408 bytes (0x0198)
     read_data(length);
     read += length;
 
-    serial::string("byte_control: ");
-    serial::integer<uint8_t>(gdrom_if.byte_control_high, ' ');
-    serial::integer<uint8_t>(gdrom_if.byte_control_low);
+    serial::string("byte_count: ");
+    serial::integer<uint8_t>(gdrom_if.byte_count_high, ' ');
+    serial::integer<uint8_t>(gdrom_if.byte_count_low);
   }
   serial::string("read bytes: ");
   serial::integer<uint32_t>(read);
diff --git a/example/interrupt.cpp b/example/interrupt.cpp
index 29c9e0f..ca2756b 100644
--- a/example/interrupt.cpp
+++ b/example/interrupt.cpp
@@ -1,9 +1,9 @@
 #include <cstdint>
 
-#include "serial.hpp"
+#include "sh7091/serial.hpp"
 
-#include "sh7095.hpp"
-#include "sh7095_bits.hpp"
+#include "sh7091/sh7091.hpp"
+#include "sh7091/sh7091_bits.hpp"
 
 void main()
 {
diff --git a/example/maple_controller.cpp b/example/maple_controller.cpp
index 3d02c66..78f277f 100644
--- a/example/maple_controller.cpp
+++ b/example/maple_controller.cpp
@@ -1,8 +1,9 @@
 #include <bit>
 
-#include "holly/video_output.hpp"
 #include "align.hpp"
-
+#include "holly/video_output.hpp"
+#include "holly/core_bits.hpp"
+#include "holly/holly.hpp"
 #include "maple/maple.hpp"
 #include "maple/maple_impl.hpp"
 #include "maple/maple_bus_bits.hpp"
@@ -85,8 +86,8 @@ void do_device_request()
                    receive_buf, maple::sizeof_command(host_response));
 
   for (uint8_t port = 0; port < 4; port++) {
-    auto& bus_data = response[port].bus_data;
-    auto& data_fields = response[port].bus_data.data_fields;
+    auto& bus_data = host_response[port].bus_data;
+    auto& data_fields = host_response[port].bus_data.data_fields;
     if (bus_data.command_code != device_status::command_code) {
       // the controller is disconnected
     } else {
@@ -110,7 +111,8 @@ void main()
   video_output::set_mode_vga();
 
   while (1) {
-    v_sync_in();
+    while (!spg_status::vsync(holly.SPG_STATUS));
+    while (spg_status::vsync(holly.SPG_STATUS));
     do_device_request();
   };
 }
diff --git a/example/maple_device_request.cpp b/example/maple_device_request.cpp
index 15826c2..9781b9c 100644
--- a/example/maple_device_request.cpp
+++ b/example/maple_device_request.cpp
@@ -29,7 +29,7 @@ void main()
     serial::integer<uint8_t>(port);
 
     for (uint32_t i = 0; i < host_response_size; i++) {
-      serial::integer<uint8_t>(buf[port * command_response_size + i]);
+      serial::integer<uint8_t>(buf[port * host_response_size + i]);
     }
     serial::character('\n');
   }
diff --git a/example/maple_vibrator.cpp b/example/maple_vibrator.cpp
index 16f2d34..c86f8f9 100644
--- a/example/maple_vibrator.cpp
+++ b/example/maple_vibrator.cpp
@@ -1,8 +1,9 @@
 #include <bit>
 
-#include "holly/video_output.hpp"
 #include "align.hpp"
-
+#include "holly/video_output.hpp"
+#include "holly/core_bits.hpp"
+#include "holly/holly.hpp"
 #include "maple/maple.hpp"
 #include "maple/maple_impl.hpp"
 #include "maple/maple_bus_bits.hpp"
@@ -188,8 +189,8 @@ void main()
 
   while (1) {
     for (int i = 0; i < 120; i++) {
-      v_sync_out();
-      v_sync_in();
+      while (!spg_status::vsync(holly.SPG_STATUS));
+      while (spg_status::vsync(holly.SPG_STATUS));
     }
     do_device_request();
   };
diff --git a/example/sprite.cpp b/example/sprite.cpp
index aca0ab2..8670980 100644
--- a/example/sprite.cpp
+++ b/example/sprite.cpp
@@ -114,8 +114,6 @@ void main()
 				       , .punch_through = 0
 				       };
 
-  constexpr uint32_t tiles = (640 / 32) * (480 / 32);
-
   holly.SOFTRESET = softreset::pipeline_soft_reset
 		  | softreset::ta_soft_reset;
   holly.SOFTRESET = 0;
diff --git a/holly/core.cpp b/holly/core.cpp
index d54a911..65e35d0 100644
--- a/holly/core.cpp
+++ b/holly/core.cpp
@@ -82,8 +82,6 @@ void core_start_render(uint32_t frame_ix, uint32_t num_frames)
 		    frame_ix, num_frames);
 }
 
-static bool flycast_bug = 0;
-
 void core_wait_end_of_render_video()
 {
   /*