pacc: initial

40 files changed, 2441 insertions, 0 deletions

diff --git a/pacc/glsl/Makefile b/pacc/glsl/Makefile
new file mode 100644
index 0000000..36aace4
--- /dev/null
+++ b/pacc/glsl/Makefile
@@ -0,0 +1,40 @@
+VSHADERS:=blit.vert
+FSHADERS:=copy.frag color.frag color_trans.frag key.frag fill.frag font.frag
+DSHEADER:=dsheader
+ESHEADER:=esheader
+INCS:=$(addsuffix .inc,$(VSHADERS) $(FSHADERS) $(DSHEADER) $(ESHEADER))
+SHADERS2:=$(addsuffix .ds.vert,$(basename $(VSHADERS)))
+SHADERS2+=$(addsuffix .es.vert,$(basename $(VSHADERS)))
+SHADERS2+=$(addsuffix .ds.frag,$(basename $(FSHADERS)))
+SHADERS2+=$(addsuffix .es.frag,$(basename $(FSHADERS)))
+VALS:=$(addsuffix .vald,$(SHADERS2))
+VALIDATOR:=glslangValidator
+VFLAGS:=
+
+all:	$(INCS)
+
+val:	$(VALS)
+
+%.es.vert:	%.vert
+	cat $(ESHEADER) $< > $@
+
+%.es.frag:	%.frag
+	cat $(ESHEADER) $< > $@
+
+%.ds.vert:	%.vert
+	cat $(DSHEADER) $< > $@
+
+%.ds.frag:	%.frag
+	cat $(DSHEADER) $< > $@
+
+%.vald:	%
+	$(VALIDATOR) $(VFLAGS) $<
+
+%.inc:	%.zt
+	xxd -i $< | sed -e '/^unsigned char /c\static const uint8_t $(shell echo $(basename $<) | sed -e 's/\./_/g')[] = {' -e '/^unsigned int /c\\' > $@
+
+%.zt:	%
+	dd bs=1 count=1 if=/dev/zero | cat $< - > $@
+
+clean:
+	rm -f $(INCS)
diff --git a/pacc/glsl/blit.vert b/pacc/glsl/blit.vert
new file mode 100644
index 0000000..4232a8d
--- /dev/null
+++ b/pacc/glsl/blit.vert
@@ -0,0 +1,7 @@
+attribute vec4 coord;
+varying vec2 texcoord;
+void main(void) {
+  gl_Position = vec4(coord.xy, 0.0, 1.0);
+  texcoord = coord.zw;
+}
+
diff --git a/pacc/glsl/blit.vert.inc b/pacc/glsl/blit.vert.inc
new file mode 100644
index 0000000..34691a9
--- /dev/null
+++ b/pacc/glsl/blit.vert.inc
@@ -0,0 +1,14 @@
+static const uint8_t blit_vert[] = {
+  0x61, 0x74, 0x74, 0x72, 0x69, 0x62, 0x75, 0x74, 0x65, 0x20, 0x76, 0x65,
+  0x63, 0x34, 0x20, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x76, 0x61,
+  0x72, 0x79, 0x69, 0x6e, 0x67, 0x20, 0x76, 0x65, 0x63, 0x32, 0x20, 0x74,
+  0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x76, 0x6f, 0x69,
+  0x64, 0x20, 0x6d, 0x61, 0x69, 0x6e, 0x28, 0x76, 0x6f, 0x69, 0x64, 0x29,
+  0x20, 0x7b, 0x0a, 0x20, 0x20, 0x67, 0x6c, 0x5f, 0x50, 0x6f, 0x73, 0x69,
+  0x74, 0x69, 0x6f, 0x6e, 0x20, 0x3d, 0x20, 0x76, 0x65, 0x63, 0x34, 0x28,
+  0x63, 0x6f, 0x6f, 0x72, 0x64, 0x2e, 0x78, 0x79, 0x2c, 0x20, 0x30, 0x2e,
+  0x30, 0x2c, 0x20, 0x31, 0x2e, 0x30, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x74,
+  0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x20, 0x3d, 0x20, 0x63, 0x6f,
+  0x6f, 0x72, 0x64, 0x2e, 0x7a, 0x77, 0x3b, 0x0a, 0x7d, 0x0a, 0x0a, 0x00
+};
+
diff --git a/pacc/glsl/color.frag b/pacc/glsl/color.frag
new file mode 100644
index 0000000..d03bee5
--- /dev/null
+++ b/pacc/glsl/color.frag
@@ -0,0 +1,13 @@
+uniform sampler2D palette;
+uniform sampler2D tex;
+varying mediump vec2 texcoord;
+uniform lowp float color;
+void main(void) {
+  lowp float index = texture2D(tex, texcoord).x;
+  if (index > (0.5/255.0)) {
+    index = color;
+  } else {
+    index = 0.5 / 256.0;
+  }
+  gl_FragColor = texture2D(palette, vec2(index, 0.0));
+}
diff --git a/pacc/glsl/color.frag.inc b/pacc/glsl/color.frag.inc
new file mode 100644
index 0000000..8ea77ef
--- /dev/null
+++ b/pacc/glsl/color.frag.inc
@@ -0,0 +1,30 @@
+static const uint8_t color_frag[] = {
+  0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73, 0x61, 0x6d, 0x70,
+  0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x70, 0x61, 0x6c, 0x65, 0x74, 0x74,
+  0x65, 0x3b, 0x0a, 0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73,
+  0x61, 0x6d, 0x70, 0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x74, 0x65, 0x78,
+  0x3b, 0x0a, 0x76, 0x61, 0x72, 0x79, 0x69, 0x6e, 0x67, 0x20, 0x6d, 0x65,
+  0x64, 0x69, 0x75, 0x6d, 0x70, 0x20, 0x76, 0x65, 0x63, 0x32, 0x20, 0x74,
+  0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x75, 0x6e, 0x69,
+  0x66, 0x6f, 0x72, 0x6d, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20, 0x66, 0x6c,
+  0x6f, 0x61, 0x74, 0x20, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x3b, 0x0a, 0x76,
+  0x6f, 0x69, 0x64, 0x20, 0x6d, 0x61, 0x69, 0x6e, 0x28, 0x76, 0x6f, 0x69,
+  0x64, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20,
+  0x66, 0x6c, 0x6f, 0x61, 0x74, 0x20, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x20,
+  0x3d, 0x20, 0x74, 0x65, 0x78, 0x74, 0x75, 0x72, 0x65, 0x32, 0x44, 0x28,
+  0x74, 0x65, 0x78, 0x2c, 0x20, 0x74, 0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72,
+  0x64, 0x29, 0x2e, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28,
+  0x69, 0x6e, 0x64, 0x65, 0x78, 0x20, 0x3e, 0x20, 0x28, 0x30, 0x2e, 0x35,
+  0x2f, 0x32, 0x35, 0x35, 0x2e, 0x30, 0x29, 0x29, 0x20, 0x7b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x20, 0x3d, 0x20, 0x63,
+  0x6f, 0x6c, 0x6f, 0x72, 0x3b, 0x0a, 0x20, 0x20, 0x7d, 0x20, 0x65, 0x6c,
+  0x73, 0x65, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x69, 0x6e, 0x64,
+  0x65, 0x78, 0x20, 0x3d, 0x20, 0x30, 0x2e, 0x35, 0x20, 0x2f, 0x20, 0x32,
+  0x35, 0x36, 0x2e, 0x30, 0x3b, 0x0a, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20,
+  0x67, 0x6c, 0x5f, 0x46, 0x72, 0x61, 0x67, 0x43, 0x6f, 0x6c, 0x6f, 0x72,
+  0x20, 0x3d, 0x20, 0x74, 0x65, 0x78, 0x74, 0x75, 0x72, 0x65, 0x32, 0x44,
+  0x28, 0x70, 0x61, 0x6c, 0x65, 0x74, 0x74, 0x65, 0x2c, 0x20, 0x76, 0x65,
+  0x63, 0x32, 0x28, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x2c, 0x20, 0x30, 0x2e,
+  0x30, 0x29, 0x29, 0x3b, 0x0a, 0x7d, 0x0a, 0x00
+};
+
diff --git a/pacc/glsl/color_trans.frag b/pacc/glsl/color_trans.frag
new file mode 100644
index 0000000..ef742a4
--- /dev/null
+++ b/pacc/glsl/color_trans.frag
@@ -0,0 +1,11 @@
+uniform sampler2D palette;
+uniform sampler2D tex;
+varying mediump vec2 texcoord;
+uniform lowp float color;
+void main(void) {
+  lowp float index = texture2D(tex, texcoord).x;
+  if (index < (0.5/255.0)) {
+    discard;
+  }
+  gl_FragColor = texture2D(palette, vec2(color, 0.0));
+}
diff --git a/pacc/glsl/color_trans.frag.inc b/pacc/glsl/color_trans.frag.inc
new file mode 100644
index 0000000..f49404c
--- /dev/null
+++ b/pacc/glsl/color_trans.frag.inc
@@ -0,0 +1,27 @@
+static const uint8_t color_trans_frag[] = {
+  0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73, 0x61, 0x6d, 0x70,
+  0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x70, 0x61, 0x6c, 0x65, 0x74, 0x74,
+  0x65, 0x3b, 0x0a, 0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73,
+  0x61, 0x6d, 0x70, 0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x74, 0x65, 0x78,
+  0x3b, 0x0a, 0x76, 0x61, 0x72, 0x79, 0x69, 0x6e, 0x67, 0x20, 0x6d, 0x65,
+  0x64, 0x69, 0x75, 0x6d, 0x70, 0x20, 0x76, 0x65, 0x63, 0x32, 0x20, 0x74,
+  0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x75, 0x6e, 0x69,
+  0x66, 0x6f, 0x72, 0x6d, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20, 0x66, 0x6c,
+  0x6f, 0x61, 0x74, 0x20, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x3b, 0x0a, 0x76,
+  0x6f, 0x69, 0x64, 0x20, 0x6d, 0x61, 0x69, 0x6e, 0x28, 0x76, 0x6f, 0x69,
+  0x64, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20,
+  0x66, 0x6c, 0x6f, 0x61, 0x74, 0x20, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x20,
+  0x3d, 0x20, 0x74, 0x65, 0x78, 0x74, 0x75, 0x72, 0x65, 0x32, 0x44, 0x28,
+  0x74, 0x65, 0x78, 0x2c, 0x20, 0x74, 0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72,
+  0x64, 0x29, 0x2e, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28,
+  0x69, 0x6e, 0x64, 0x65, 0x78, 0x20, 0x3c, 0x20, 0x28, 0x30, 0x2e, 0x35,
+  0x2f, 0x32, 0x35, 0x35, 0x2e, 0x30, 0x29, 0x29, 0x20, 0x7b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x64, 0x69, 0x73, 0x63, 0x61, 0x72, 0x64, 0x3b, 0x0a,
+  0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x67, 0x6c, 0x5f, 0x46, 0x72, 0x61,
+  0x67, 0x43, 0x6f, 0x6c, 0x6f, 0x72, 0x20, 0x3d, 0x20, 0x74, 0x65, 0x78,
+  0x74, 0x75, 0x72, 0x65, 0x32, 0x44, 0x28, 0x70, 0x61, 0x6c, 0x65, 0x74,
+  0x74, 0x65, 0x2c, 0x20, 0x76, 0x65, 0x63, 0x32, 0x28, 0x63, 0x6f, 0x6c,
+  0x6f, 0x72, 0x2c, 0x20, 0x30, 0x2e, 0x30, 0x29, 0x29, 0x3b, 0x0a, 0x7d,
+  0x0a, 0x00
+};
+
diff --git a/pacc/glsl/copy.frag b/pacc/glsl/copy.frag
new file mode 100644
index 0000000..a77ad91
--- /dev/null
+++ b/pacc/glsl/copy.frag
@@ -0,0 +1,9 @@
+uniform sampler2D palette;
+uniform sampler2D tex;
+varying mediump vec2 texcoord;
+void main(void) {
+  lowp float index = texture2D(tex, texcoord).x;
+  lowp float color = (index * 255.0 + 0.5) / 256.0;
+  gl_FragColor = texture2D(palette, vec2(color, 0.0));
+}
+
diff --git a/pacc/glsl/copy.frag.inc b/pacc/glsl/copy.frag.inc
new file mode 100644
index 0000000..fa3f388
--- /dev/null
+++ b/pacc/glsl/copy.frag.inc
@@ -0,0 +1,25 @@
+static const uint8_t copy_frag[] = {
+  0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73, 0x61, 0x6d, 0x70,
+  0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x70, 0x61, 0x6c, 0x65, 0x74, 0x74,
+  0x65, 0x3b, 0x0a, 0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73,
+  0x61, 0x6d, 0x70, 0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x74, 0x65, 0x78,
+  0x3b, 0x0a, 0x76, 0x61, 0x72, 0x79, 0x69, 0x6e, 0x67, 0x20, 0x6d, 0x65,
+  0x64, 0x69, 0x75, 0x6d, 0x70, 0x20, 0x76, 0x65, 0x63, 0x32, 0x20, 0x74,
+  0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x76, 0x6f, 0x69,
+  0x64, 0x20, 0x6d, 0x61, 0x69, 0x6e, 0x28, 0x76, 0x6f, 0x69, 0x64, 0x29,
+  0x20, 0x7b, 0x0a, 0x20, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20, 0x66, 0x6c,
+  0x6f, 0x61, 0x74, 0x20, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x20, 0x3d, 0x20,
+  0x74, 0x65, 0x78, 0x74, 0x75, 0x72, 0x65, 0x32, 0x44, 0x28, 0x74, 0x65,
+  0x78, 0x2c, 0x20, 0x74, 0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x29,
+  0x2e, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20, 0x66,
+  0x6c, 0x6f, 0x61, 0x74, 0x20, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x20, 0x3d,
+  0x20, 0x28, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x20, 0x2a, 0x20, 0x32, 0x35,
+  0x35, 0x2e, 0x30, 0x20, 0x2b, 0x20, 0x30, 0x2e, 0x35, 0x29, 0x20, 0x2f,
+  0x20, 0x32, 0x35, 0x36, 0x2e, 0x30, 0x3b, 0x0a, 0x20, 0x20, 0x67, 0x6c,
+  0x5f, 0x46, 0x72, 0x61, 0x67, 0x43, 0x6f, 0x6c, 0x6f, 0x72, 0x20, 0x3d,
+  0x20, 0x74, 0x65, 0x78, 0x74, 0x75, 0x72, 0x65, 0x32, 0x44, 0x28, 0x70,
+  0x61, 0x6c, 0x65, 0x74, 0x74, 0x65, 0x2c, 0x20, 0x76, 0x65, 0x63, 0x32,
+  0x28, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x2c, 0x20, 0x30, 0x2e, 0x30, 0x29,
+  0x29, 0x3b, 0x0a, 0x7d, 0x0a, 0x0a, 0x00
+};
+
diff --git a/pacc/glsl/dsheader b/pacc/glsl/dsheader
new file mode 100644
index 0000000..13e65b7
--- /dev/null
+++ b/pacc/glsl/dsheader
@@ -0,0 +1,5 @@
+#version 110
+#define lowp
+#define mediump
+#define highp
+
diff --git a/pacc/glsl/dsheader.inc b/pacc/glsl/dsheader.inc
new file mode 100644
index 0000000..b742cec
--- /dev/null
+++ b/pacc/glsl/dsheader.inc
@@ -0,0 +1,8 @@
+static const uint8_t dsheader[] = {
+  0x23, 0x76, 0x65, 0x72, 0x73, 0x69, 0x6f, 0x6e, 0x20, 0x31, 0x31, 0x30,
+  0x0a, 0x23, 0x64, 0x65, 0x66, 0x69, 0x6e, 0x65, 0x20, 0x6c, 0x6f, 0x77,
+  0x70, 0x0a, 0x23, 0x64, 0x65, 0x66, 0x69, 0x6e, 0x65, 0x20, 0x6d, 0x65,
+  0x64, 0x69, 0x75, 0x6d, 0x70, 0x0a, 0x23, 0x64, 0x65, 0x66, 0x69, 0x6e,
+  0x65, 0x20, 0x68, 0x69, 0x67, 0x68, 0x70, 0x0a, 0x0a, 0x00
+};
+
diff --git a/pacc/glsl/esheader b/pacc/glsl/esheader
new file mode 100644
index 0000000..43b26bc
--- /dev/null
+++ b/pacc/glsl/esheader
@@ -0,0 +1,2 @@
+#version 100
+
diff --git a/pacc/glsl/esheader.inc b/pacc/glsl/esheader.inc
new file mode 100644
index 0000000..3ca61f1
--- /dev/null
+++ b/pacc/glsl/esheader.inc
@@ -0,0 +1,5 @@
+static const uint8_t esheader[] = {
+  0x23, 0x76, 0x65, 0x72, 0x73, 0x69, 0x6f, 0x6e, 0x20, 0x31, 0x30, 0x30,
+  0x0a, 0x0a, 0x00
+};
+
diff --git a/pacc/glsl/fill.frag b/pacc/glsl/fill.frag
new file mode 100644
index 0000000..010b5dc
--- /dev/null
+++ b/pacc/glsl/fill.frag
@@ -0,0 +1,5 @@
+uniform sampler2D palette;
+varying mediump vec2 texcoord;
+void main(void) {
+  gl_FragColor = texture2D(palette, vec2(texcoord.x, 0.0));
+}
diff --git a/pacc/glsl/fill.frag.inc b/pacc/glsl/fill.frag.inc
new file mode 100644
index 0000000..8065199
--- /dev/null
+++ b/pacc/glsl/fill.frag.inc
@@ -0,0 +1,15 @@
+static const uint8_t fill_frag[] = {
+  0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73, 0x61, 0x6d, 0x70,
+  0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x70, 0x61, 0x6c, 0x65, 0x74, 0x74,
+  0x65, 0x3b, 0x0a, 0x76, 0x61, 0x72, 0x79, 0x69, 0x6e, 0x67, 0x20, 0x6d,
+  0x65, 0x64, 0x69, 0x75, 0x6d, 0x70, 0x20, 0x76, 0x65, 0x63, 0x32, 0x20,
+  0x74, 0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x76, 0x6f,
+  0x69, 0x64, 0x20, 0x6d, 0x61, 0x69, 0x6e, 0x28, 0x76, 0x6f, 0x69, 0x64,
+  0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x67, 0x6c, 0x5f, 0x46, 0x72, 0x61,
+  0x67, 0x43, 0x6f, 0x6c, 0x6f, 0x72, 0x20, 0x3d, 0x20, 0x74, 0x65, 0x78,
+  0x74, 0x75, 0x72, 0x65, 0x32, 0x44, 0x28, 0x70, 0x61, 0x6c, 0x65, 0x74,
+  0x74, 0x65, 0x2c, 0x20, 0x76, 0x65, 0x63, 0x32, 0x28, 0x74, 0x65, 0x78,
+  0x63, 0x6f, 0x6f, 0x72, 0x64, 0x2e, 0x78, 0x2c, 0x20, 0x30, 0x2e, 0x30,
+  0x29, 0x29, 0x3b, 0x0a, 0x7d, 0x0a, 0x00
+};
+
diff --git a/pacc/glsl/font.frag b/pacc/glsl/font.frag
new file mode 100644
index 0000000..8beb0a7
--- /dev/null
+++ b/pacc/glsl/font.frag
@@ -0,0 +1,15 @@
+uniform sampler2D palette;
+uniform sampler2D tex;
+varying mediump vec2 texcoord;
+uniform lowp float bg;
+uniform lowp float color;
+void main(void) {
+  gl_FragColor = vec4(1.0, 0.0, 1.0, 1.0);
+  lowp float pixel = texture2D(tex, texcoord).x;
+  lowp float index = color;
+  if (pixel < 0.5) {
+    if (bg < 0.5) discard;
+    index = 0.5 / 256.0;
+  }
+  gl_FragColor = texture2D(palette, vec2(index, 0.0));
+}
diff --git a/pacc/glsl/font.frag.inc b/pacc/glsl/font.frag.inc
new file mode 100644
index 0000000..27d143d
--- /dev/null
+++ b/pacc/glsl/font.frag.inc
@@ -0,0 +1,37 @@
+static const uint8_t font_frag[] = {
+  0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73, 0x61, 0x6d, 0x70,
+  0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x70, 0x61, 0x6c, 0x65, 0x74, 0x74,
+  0x65, 0x3b, 0x0a, 0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73,
+  0x61, 0x6d, 0x70, 0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x74, 0x65, 0x78,
+  0x3b, 0x0a, 0x76, 0x61, 0x72, 0x79, 0x69, 0x6e, 0x67, 0x20, 0x6d, 0x65,
+  0x64, 0x69, 0x75, 0x6d, 0x70, 0x20, 0x76, 0x65, 0x63, 0x32, 0x20, 0x74,
+  0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x75, 0x6e, 0x69,
+  0x66, 0x6f, 0x72, 0x6d, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20, 0x66, 0x6c,
+  0x6f, 0x61, 0x74, 0x20, 0x62, 0x67, 0x3b, 0x0a, 0x75, 0x6e, 0x69, 0x66,
+  0x6f, 0x72, 0x6d, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20, 0x66, 0x6c, 0x6f,
+  0x61, 0x74, 0x20, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x3b, 0x0a, 0x76, 0x6f,
+  0x69, 0x64, 0x20, 0x6d, 0x61, 0x69, 0x6e, 0x28, 0x76, 0x6f, 0x69, 0x64,
+  0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x67, 0x6c, 0x5f, 0x46, 0x72, 0x61,
+  0x67, 0x43, 0x6f, 0x6c, 0x6f, 0x72, 0x20, 0x3d, 0x20, 0x76, 0x65, 0x63,
+  0x34, 0x28, 0x31, 0x2e, 0x30, 0x2c, 0x20, 0x30, 0x2e, 0x30, 0x2c, 0x20,
+  0x31, 0x2e, 0x30, 0x2c, 0x20, 0x31, 0x2e, 0x30, 0x29, 0x3b, 0x0a, 0x20,
+  0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20, 0x66, 0x6c, 0x6f, 0x61, 0x74, 0x20,
+  0x70, 0x69, 0x78, 0x65, 0x6c, 0x20, 0x3d, 0x20, 0x74, 0x65, 0x78, 0x74,
+  0x75, 0x72, 0x65, 0x32, 0x44, 0x28, 0x74, 0x65, 0x78, 0x2c, 0x20, 0x74,
+  0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x29, 0x2e, 0x78, 0x3b, 0x0a,
+  0x20, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20, 0x66, 0x6c, 0x6f, 0x61, 0x74,
+  0x20, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x20, 0x3d, 0x20, 0x63, 0x6f, 0x6c,
+  0x6f, 0x72, 0x3b, 0x0a, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x70, 0x69,
+  0x78, 0x65, 0x6c, 0x20, 0x3c, 0x20, 0x30, 0x2e, 0x35, 0x29, 0x20, 0x7b,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x62, 0x67, 0x20,
+  0x3c, 0x20, 0x30, 0x2e, 0x35, 0x29, 0x20, 0x64, 0x69, 0x73, 0x63, 0x61,
+  0x72, 0x64, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x69, 0x6e, 0x64, 0x65,
+  0x78, 0x20, 0x3d, 0x20, 0x30, 0x2e, 0x35, 0x20, 0x2f, 0x20, 0x32, 0x35,
+  0x36, 0x2e, 0x30, 0x3b, 0x0a, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x67,
+  0x6c, 0x5f, 0x46, 0x72, 0x61, 0x67, 0x43, 0x6f, 0x6c, 0x6f, 0x72, 0x20,
+  0x3d, 0x20, 0x74, 0x65, 0x78, 0x74, 0x75, 0x72, 0x65, 0x32, 0x44, 0x28,
+  0x70, 0x61, 0x6c, 0x65, 0x74, 0x74, 0x65, 0x2c, 0x20, 0x76, 0x65, 0x63,
+  0x32, 0x28, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x2c, 0x20, 0x30, 0x2e, 0x30,
+  0x29, 0x29, 0x3b, 0x0a, 0x7d, 0x0a, 0x00
+};
+
diff --git a/pacc/glsl/key.frag b/pacc/glsl/key.frag
new file mode 100644
index 0000000..d6bf7e1
--- /dev/null
+++ b/pacc/glsl/key.frag
@@ -0,0 +1,12 @@
+uniform sampler2D palette;
+uniform sampler2D tex;
+varying mediump vec2 texcoord;
+uniform lowp float key;
+uniform lowp float color;
+void main(void) {
+  lowp float index = texture2D(tex, texcoord).x;
+  if (index < (key + (0.5/255.0)) || (key + (1.5/255.0)) < index) {
+    discard;
+  }
+  gl_FragColor = texture2D(palette, vec2(color, 0.0));
+}
diff --git a/pacc/glsl/key.frag.inc b/pacc/glsl/key.frag.inc
new file mode 100644
index 0000000..4a4a6c1
--- /dev/null
+++ b/pacc/glsl/key.frag.inc
@@ -0,0 +1,32 @@
+static const uint8_t key_frag[] = {
+  0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73, 0x61, 0x6d, 0x70,
+  0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x70, 0x61, 0x6c, 0x65, 0x74, 0x74,
+  0x65, 0x3b, 0x0a, 0x75, 0x6e, 0x69, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x73,
+  0x61, 0x6d, 0x70, 0x6c, 0x65, 0x72, 0x32, 0x44, 0x20, 0x74, 0x65, 0x78,
+  0x3b, 0x0a, 0x76, 0x61, 0x72, 0x79, 0x69, 0x6e, 0x67, 0x20, 0x6d, 0x65,
+  0x64, 0x69, 0x75, 0x6d, 0x70, 0x20, 0x76, 0x65, 0x63, 0x32, 0x20, 0x74,
+  0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x75, 0x6e, 0x69,
+  0x66, 0x6f, 0x72, 0x6d, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20, 0x66, 0x6c,
+  0x6f, 0x61, 0x74, 0x20, 0x6b, 0x65, 0x79, 0x3b, 0x0a, 0x75, 0x6e, 0x69,
+  0x66, 0x6f, 0x72, 0x6d, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20, 0x66, 0x6c,
+  0x6f, 0x61, 0x74, 0x20, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x3b, 0x0a, 0x76,
+  0x6f, 0x69, 0x64, 0x20, 0x6d, 0x61, 0x69, 0x6e, 0x28, 0x76, 0x6f, 0x69,
+  0x64, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x6c, 0x6f, 0x77, 0x70, 0x20,
+  0x66, 0x6c, 0x6f, 0x61, 0x74, 0x20, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x20,
+  0x3d, 0x20, 0x74, 0x65, 0x78, 0x74, 0x75, 0x72, 0x65, 0x32, 0x44, 0x28,
+  0x74, 0x65, 0x78, 0x2c, 0x20, 0x74, 0x65, 0x78, 0x63, 0x6f, 0x6f, 0x72,
+  0x64, 0x29, 0x2e, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28,
+  0x69, 0x6e, 0x64, 0x65, 0x78, 0x20, 0x3c, 0x20, 0x28, 0x6b, 0x65, 0x79,
+  0x20, 0x2b, 0x20, 0x28, 0x30, 0x2e, 0x35, 0x2f, 0x32, 0x35, 0x35, 0x2e,
+  0x30, 0x29, 0x29, 0x20, 0x7c, 0x7c, 0x20, 0x28, 0x6b, 0x65, 0x79, 0x20,
+  0x2b, 0x20, 0x28, 0x31, 0x2e, 0x35, 0x2f, 0x32, 0x35, 0x35, 0x2e, 0x30,
+  0x29, 0x29, 0x20, 0x3c, 0x20, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x29, 0x20,
+  0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x64, 0x69, 0x73, 0x63, 0x61, 0x72,
+  0x64, 0x3b, 0x0a, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x67, 0x6c, 0x5f,
+  0x46, 0x72, 0x61, 0x67, 0x43, 0x6f, 0x6c, 0x6f, 0x72, 0x20, 0x3d, 0x20,
+  0x74, 0x65, 0x78, 0x74, 0x75, 0x72, 0x65, 0x32, 0x44, 0x28, 0x70, 0x61,
+  0x6c, 0x65, 0x74, 0x74, 0x65, 0x2c, 0x20, 0x76, 0x65, 0x63, 0x32, 0x28,
+  0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x2c, 0x20, 0x30, 0x2e, 0x30, 0x29, 0x29,
+  0x3b, 0x0a, 0x7d, 0x0a, 0x00
+};
+
diff --git a/pacc/hlsl/Makefile b/pacc/hlsl/Makefile
new file mode 100644
index 0000000..7cd2f51
--- /dev/null
+++ b/pacc/hlsl/Makefile
@@ -0,0 +1,15 @@
+INCS:=copy.ps.inc color.ps.inc color_trans.ps.inc key.ps.inc fill.ps.inc font.ps.inc blit.vs.inc
+FXC:=env WINEDEBUG=-all wine $(HOME)/hlsl/fxc/fxc
+all:	$(INCS)
+
+%.inc:	%.inc.tmp
+	dos2unix < $< | sed -e 's/^const BYTE g_/static const BYTE /' -e '$$a\\' > $@
+
+%.vs.inc.tmp: %.vs.hlsl
+	$(FXC) /nologo /O3 /Tvs_2_0 /E$(basename $(basename $<)) /Fh $@ $<
+
+%.ps.inc.tmp: %.ps.hlsl
+	$(FXC) /nologo /O3 /Tps_2_0 /E$(basename $(basename $<)) /Fh $@ $<
+
+clean:
+	rm -f $(INCS)
diff --git a/pacc/hlsl/blit.vs.hlsl b/pacc/hlsl/blit.vs.hlsl
new file mode 100644
index 0000000..ea4c18d
--- /dev/null
+++ b/pacc/hlsl/blit.vs.hlsl
@@ -0,0 +1,17 @@
+struct vsinput {
+  float4 coord : POSITION;
+};
+
+struct fsinput {
+  float4 coord : POSITION;
+  float4 texcoord : TEXCOORD;
+};
+
+float2 offset: register(c0);
+
+fsinput blit(vsinput input) {
+  fsinput output;
+  output.coord = float4(input.coord.xy + offset, 0.0, 1.0);
+  output.texcoord = float4(input.coord.zw, 0.0, 1.0);
+  return output;
+}
diff --git a/pacc/hlsl/blit.vs.inc b/pacc/hlsl/blit.vs.inc
new file mode 100644
index 0000000..a2c4a64
--- /dev/null
+++ b/pacc/hlsl/blit.vs.inc
@@ -0,0 +1,71 @@
+#if 0
+//
+// Generated by Microsoft (R) HLSL Shader Compiler 9.29.952.3111
+//
+//   fxc /nologo /O3 /Tvs_2_0 /Eblit /Fh blit.vs.inc.tmp blit.vs.hlsl
+//
+//
+// Parameters:
+//
+//   float2 offset;
+//
+//
+// Registers:
+//
+//   Name         Reg   Size
+//   ------------ ----- ----
+//   offset       c0       1
+//
+
+    vs_2_0
+    def c1, 0, 1, 0, 0
+    dcl_position v0
+    add oPos.xy, v0, c0
+    mov oPos.zw, c1.xyxy
+    mad oT0, v0.zwzz, c1.yyxx, c1.xxxy
+
+// approximately 3 instruction slots used
+#endif
+
+static const BYTE vs20_blit[] =
+{
+      0,   2, 254, 255, 254, 255, 
+     33,   0,  67,  84,  65,  66, 
+     28,   0,   0,   0,  79,   0, 
+      0,   0,   0,   2, 254, 255, 
+      1,   0,   0,   0,  28,   0, 
+      0,   0,   0, 129,   0,   0, 
+     72,   0,   0,   0,  48,   0, 
+      0,   0,   2,   0,   0,   0, 
+      1,   0,   2,   0,  56,   0, 
+      0,   0,   0,   0,   0,   0, 
+    111, 102, 102, 115, 101, 116, 
+      0, 171,   1,   0,   3,   0, 
+      1,   0,   2,   0,   1,   0, 
+      0,   0,   0,   0,   0,   0, 
+    118, 115,  95,  50,  95,  48, 
+      0,  77, 105,  99, 114, 111, 
+    115, 111, 102, 116,  32,  40, 
+     82,  41,  32,  72,  76,  83, 
+     76,  32,  83, 104,  97, 100, 
+    101, 114,  32,  67, 111, 109, 
+    112, 105, 108, 101, 114,  32, 
+     57,  46,  50,  57,  46,  57, 
+     53,  50,  46,  51,  49,  49, 
+     49,   0,  81,   0,   0,   5, 
+      1,   0,  15, 160,   0,   0, 
+      0,   0,   0,   0, 128,  63, 
+      0,   0,   0,   0,   0,   0, 
+      0,   0,  31,   0,   0,   2, 
+      0,   0,   0, 128,   0,   0, 
+     15, 144,   2,   0,   0,   3, 
+      0,   0,   3, 192,   0,   0, 
+    228, 144,   0,   0, 228, 160, 
+      1,   0,   0,   2,   0,   0, 
+     12, 192,   1,   0,  68, 160, 
+      4,   0,   0,   4,   0,   0, 
+     15, 224,   0,   0, 174, 144, 
+      1,   0,   5, 160,   1,   0, 
+     64, 160, 255, 255,   0,   0
+};
+
diff --git a/pacc/hlsl/color.ps.hlsl b/pacc/hlsl/color.ps.hlsl
new file mode 100644
index 0000000..5111097
--- /dev/null
+++ b/pacc/hlsl/color.ps.hlsl
@@ -0,0 +1,23 @@
+struct psinput {
+  float4 texcoord: TEXCOORD;
+};
+
+struct psoutput {
+  float4 fragcolor: COLOR;
+};
+
+sampler palette: register(s0);
+sampler tex: register(s1);
+float uni_color: register(c0);
+
+psoutput color(psinput input) {
+  psoutput output;
+  float index = tex2D(tex, input.texcoord.xy).r;
+  if (index > (0.5/255.0)) {
+    index = uni_color;
+  } else {
+    index = 0.5 / 256.0;
+  }
+  output.fragcolor = tex1D(palette, index);
+  return output;
+}
diff --git a/pacc/hlsl/color.ps.inc b/pacc/hlsl/color.ps.inc
new file mode 100644
index 0000000..5cd1a17
--- /dev/null
+++ b/pacc/hlsl/color.ps.inc
@@ -0,0 +1,106 @@
+#if 0
+//
+// Generated by Microsoft (R) HLSL Shader Compiler 9.29.952.3111
+//
+//   fxc /nologo /O3 /Tps_2_0 /Ecolor /Fh color.ps.inc.tmp color.ps.hlsl
+//
+//
+// Parameters:
+//
+//   sampler2D palette;
+//   sampler2D tex;
+//   float uni_color;
+//
+//
+// Registers:
+//
+//   Name         Reg   Size
+//   ------------ ----- ----
+//   uni_color    c0       1
+//   palette      s0       1
+//   tex          s1       1
+//
+
+    ps_2_0
+    def c1, 0.00196078443, 0.001953125, 0, 0
+    dcl t0.xy
+    dcl_2d s0
+    dcl_2d s1
+    texld r0, t0, s1
+    add r0.x, -r0.x, c1.x
+    mov r0.y, c1.y
+    cmp r0.xy, r0.x, r0.y, c0.x
+    texld r0, r0, s0
+    mov oC0, r0
+
+// approximately 6 instruction slots used (2 texture, 4 arithmetic)
+#endif
+
+static const BYTE ps20_color[] =
+{
+      0,   2, 255, 255, 254, 255, 
+     55,   0,  67,  84,  65,  66, 
+     28,   0,   0,   0, 167,   0, 
+      0,   0,   0,   2, 255, 255, 
+      3,   0,   0,   0,  28,   0, 
+      0,   0,   0, 129,   0,   0, 
+    160,   0,   0,   0,  88,   0, 
+      0,   0,   3,   0,   0,   0, 
+      1,   0,   2,   0,  96,   0, 
+      0,   0,   0,   0,   0,   0, 
+    112,   0,   0,   0,   3,   0, 
+      1,   0,   1,   0,   6,   0, 
+    116,   0,   0,   0,   0,   0, 
+      0,   0, 132,   0,   0,   0, 
+      2,   0,   0,   0,   1,   0, 
+      2,   0, 144,   0,   0,   0, 
+      0,   0,   0,   0, 112,  97, 
+    108, 101, 116, 116, 101,   0, 
+      4,   0,  12,   0,   1,   0, 
+      1,   0,   1,   0,   0,   0, 
+      0,   0,   0,   0, 116, 101, 
+    120,   0,   4,   0,  12,   0, 
+      1,   0,   1,   0,   1,   0, 
+      0,   0,   0,   0,   0,   0, 
+    117, 110, 105,  95,  99, 111, 
+    108, 111, 114,   0, 171, 171, 
+      0,   0,   3,   0,   1,   0, 
+      1,   0,   1,   0,   0,   0, 
+      0,   0,   0,   0, 112, 115, 
+     95,  50,  95,  48,   0,  77, 
+    105,  99, 114, 111, 115, 111, 
+    102, 116,  32,  40,  82,  41, 
+     32,  72,  76,  83,  76,  32, 
+     83, 104,  97, 100, 101, 114, 
+     32,  67, 111, 109, 112, 105, 
+    108, 101, 114,  32,  57,  46, 
+     50,  57,  46,  57,  53,  50, 
+     46,  51,  49,  49,  49,   0, 
+     81,   0,   0,   5,   1,   0, 
+     15, 160, 129, 128,   0,  59, 
+      0,   0,   0,  59,   0,   0, 
+      0,   0,   0,   0,   0,   0, 
+     31,   0,   0,   2,   0,   0, 
+      0, 128,   0,   0,   3, 176, 
+     31,   0,   0,   2,   0,   0, 
+      0, 144,   0,   8,  15, 160, 
+     31,   0,   0,   2,   0,   0, 
+      0, 144,   1,   8,  15, 160, 
+     66,   0,   0,   3,   0,   0, 
+     15, 128,   0,   0, 228, 176, 
+      1,   8, 228, 160,   2,   0, 
+      0,   3,   0,   0,   1, 128, 
+      0,   0,   0, 129,   1,   0, 
+      0, 160,   1,   0,   0,   2, 
+      0,   0,   2, 128,   1,   0, 
+     85, 160,  88,   0,   0,   4, 
+      0,   0,   3, 128,   0,   0, 
+      0, 128,   0,   0,  85, 128, 
+      0,   0,   0, 160,  66,   0, 
+      0,   3,   0,   0,  15, 128, 
+      0,   0, 228, 128,   0,   8, 
+    228, 160,   1,   0,   0,   2, 
+      0,   8,  15, 128,   0,   0, 
+    228, 128, 255, 255,   0,   0
+};
+
diff --git a/pacc/hlsl/color_trans.ps.hlsl b/pacc/hlsl/color_trans.ps.hlsl
new file mode 100644
index 0000000..00999fd
--- /dev/null
+++ b/pacc/hlsl/color_trans.ps.hlsl
@@ -0,0 +1,19 @@
+struct psinput {
+  float4 texcoord: TEXCOORD;
+};
+
+struct psoutput {
+  float4 fragcolor: COLOR;
+};
+
+sampler palette: register(s0);
+sampler tex: register(s1);
+float uni_color: register(c0);
+
+psoutput color_trans(psinput input) {
+  psoutput output;
+  float index = tex2D(tex, input.texcoord.xy).r;
+  if (index < (0.5/255.0)) discard;
+  output.fragcolor = tex1D(palette, uni_color);
+  return output;
+}
diff --git a/pacc/hlsl/color_trans.ps.inc b/pacc/hlsl/color_trans.ps.inc
new file mode 100644
index 0000000..6df66f8
--- /dev/null
+++ b/pacc/hlsl/color_trans.ps.inc
@@ -0,0 +1,110 @@
+#if 0
+//
+// Generated by Microsoft (R) HLSL Shader Compiler 9.29.952.3111
+//
+//   fxc /nologo /O3 /Tps_2_0 /Ecolor_trans /Fh color_trans.ps.inc.tmp
+//    color_trans.ps.hlsl
+//
+//
+// Parameters:
+//
+//   sampler2D palette;
+//   sampler2D tex;
+//   float uni_color;
+//
+//
+// Registers:
+//
+//   Name         Reg   Size
+//   ------------ ----- ----
+//   uni_color    c0       1
+//   palette      s0       1
+//   tex          s1       1
+//
+
+    ps_2_0
+    def c1, -0.00196078443, -0, -1, 0
+    dcl t0.xy
+    dcl_2d s0
+    dcl_2d s1
+    texld r0, t0, s1
+    add r0.x, r0.x, c1.x
+    cmp r0, r0.x, c1.y, c1.z
+    texkill r0
+    mov r0.xy, c0.x
+    texld r0, r0, s0
+    mov oC0, r0
+
+// approximately 7 instruction slots used (2 texture, 5 arithmetic)
+#endif
+
+static const BYTE ps20_color_trans[] =
+{
+      0,   2, 255, 255, 254, 255, 
+     55,   0,  67,  84,  65,  66, 
+     28,   0,   0,   0, 167,   0, 
+      0,   0,   0,   2, 255, 255, 
+      3,   0,   0,   0,  28,   0, 
+      0,   0,   0, 129,   0,   0, 
+    160,   0,   0,   0,  88,   0, 
+      0,   0,   3,   0,   0,   0, 
+      1,   0,   2,   0,  96,   0, 
+      0,   0,   0,   0,   0,   0, 
+    112,   0,   0,   0,   3,   0, 
+      1,   0,   1,   0,   6,   0, 
+    116,   0,   0,   0,   0,   0, 
+      0,   0, 132,   0,   0,   0, 
+      2,   0,   0,   0,   1,   0, 
+      2,   0, 144,   0,   0,   0, 
+      0,   0,   0,   0, 112,  97, 
+    108, 101, 116, 116, 101,   0, 
+      4,   0,  12,   0,   1,   0, 
+      1,   0,   1,   0,   0,   0, 
+      0,   0,   0,   0, 116, 101, 
+    120,   0,   4,   0,  12,   0, 
+      1,   0,   1,   0,   1,   0, 
+      0,   0,   0,   0,   0,   0, 
+    117, 110, 105,  95,  99, 111, 
+    108, 111, 114,   0, 171, 171, 
+      0,   0,   3,   0,   1,   0, 
+      1,   0,   1,   0,   0,   0, 
+      0,   0,   0,   0, 112, 115, 
+     95,  50,  95,  48,   0,  77, 
+    105,  99, 114, 111, 115, 111, 
+    102, 116,  32,  40,  82,  41, 
+     32,  72,  76,  83,  76,  32, 
+     83, 104,  97, 100, 101, 114, 
+     32,  67, 111, 109, 112, 105, 
+    108, 101, 114,  32,  57,  46, 
+     50,  57,  46,  57,  53,  50, 
+     46,  51,  49,  49,  49,   0, 
+     81,   0,   0,   5,   1,   0, 
+     15, 160, 129, 128,   0, 187, 
+      0,   0,   0, 128,   0,   0, 
+    128, 191,   0,   0,   0,   0, 
+     31,   0,   0,   2,   0,   0, 
+      0, 128,   0,   0,   3, 176, 
+     31,   0,   0,   2,   0,   0, 
+      0, 144,   0,   8,  15, 160, 
+     31,   0,   0,   2,   0,   0, 
+      0, 144,   1,   8,  15, 160, 
+     66,   0,   0,   3,   0,   0, 
+     15, 128,   0,   0, 228, 176, 
+      1,   8, 228, 160,   2,   0, 
+      0,   3,   0,   0,   1, 128, 
+      0,   0,   0, 128,   1,   0, 
+      0, 160,  88,   0,   0,   4, 
+      0,   0,  15, 128,   0,   0, 
+      0, 128,   1,   0,  85, 160, 
+      1,   0, 170, 160,  65,   0, 
+      0,   1,   0,   0,  15, 128, 
+      1,   0,   0,   2,   0,   0, 
+      3, 128,   0,   0,   0, 160, 
+     66,   0,   0,   3,   0,   0, 
+     15, 128,   0,   0, 228, 128, 
+      0,   8, 228, 160,   1,   0, 
+      0,   2,   0,   8,  15, 128, 
+      0,   0, 228, 128, 255, 255, 
+      0,   0
+};
+
diff --git a/pacc/hlsl/copy.ps.hlsl b/pacc/hlsl/copy.ps.hlsl
new file mode 100644
index 0000000..df9d1eb
--- /dev/null
+++ b/pacc/hlsl/copy.ps.hlsl
@@ -0,0 +1,18 @@
+struct psinput {
+  float4 texcoord: TEXCOORD;
+};
+
+struct psoutput {
+  float4 fragcolor: COLOR;
+};
+
+sampler palette: register(s0);
+sampler tex: register(s1);
+
+psoutput copy(psinput input) {
+  psoutput output;
+  float index = tex2D(tex, input.texcoord.xy).r;
+  float color = (index * 255.0 + 0.5) / 256.0;
+  output.fragcolor = tex1D(palette, color);
+  return output;
+}
diff --git a/pacc/hlsl/copy.ps.inc b/pacc/hlsl/copy.ps.inc
new file mode 100644
index 0000000..16acf0a
--- /dev/null
+++ b/pacc/hlsl/copy.ps.inc
@@ -0,0 +1,93 @@
+#if 0
+//
+// Generated by Microsoft (R) HLSL Shader Compiler 9.29.952.3111
+//
+//   fxc /nologo /O3 /Tps_2_0 /Ecopy /Fh copy.ps.inc.tmp copy.ps.hlsl
+//
+//
+// Parameters:
+//
+//   sampler2D palette;
+//   sampler2D tex;
+//
+//
+// Registers:
+//
+//   Name         Reg   Size
+//   ------------ ----- ----
+//   palette      s0       1
+//   tex          s1       1
+//
+
+    ps_2_0
+    def c0, 255, 0.5, 0.00390625, 0
+    dcl t0.xy
+    dcl_2d s0
+    dcl_2d s1
+    texld r0, t0, s1
+    mad r0.x, r0.x, c0.x, c0.y
+    mul r0.xy, r0.x, c0.z
+    texld r0, r0, s0
+    mov oC0, r0
+
+// approximately 5 instruction slots used (2 texture, 3 arithmetic)
+#endif
+
+static const BYTE ps20_copy[] =
+{
+      0,   2, 255, 255, 254, 255, 
+     43,   0,  67,  84,  65,  66, 
+     28,   0,   0,   0, 119,   0, 
+      0,   0,   0,   2, 255, 255, 
+      2,   0,   0,   0,  28,   0, 
+      0,   0,   0, 129,   0,   0, 
+    112,   0,   0,   0,  68,   0, 
+      0,   0,   3,   0,   0,   0, 
+      1,   0,   2,   0,  76,   0, 
+      0,   0,   0,   0,   0,   0, 
+     92,   0,   0,   0,   3,   0, 
+      1,   0,   1,   0,   6,   0, 
+     96,   0,   0,   0,   0,   0, 
+      0,   0, 112,  97, 108, 101, 
+    116, 116, 101,   0,   4,   0, 
+     12,   0,   1,   0,   1,   0, 
+      1,   0,   0,   0,   0,   0, 
+      0,   0, 116, 101, 120,   0, 
+      4,   0,  12,   0,   1,   0, 
+      1,   0,   1,   0,   0,   0, 
+      0,   0,   0,   0, 112, 115, 
+     95,  50,  95,  48,   0,  77, 
+    105,  99, 114, 111, 115, 111, 
+    102, 116,  32,  40,  82,  41, 
+     32,  72,  76,  83,  76,  32, 
+     83, 104,  97, 100, 101, 114, 
+     32,  67, 111, 109, 112, 105, 
+    108, 101, 114,  32,  57,  46, 
+     50,  57,  46,  57,  53,  50, 
+     46,  51,  49,  49,  49,   0, 
+     81,   0,   0,   5,   0,   0, 
+     15, 160,   0,   0, 127,  67, 
+      0,   0,   0,  63,   0,   0, 
+    128,  59,   0,   0,   0,   0, 
+     31,   0,   0,   2,   0,   0, 
+      0, 128,   0,   0,   3, 176, 
+     31,   0,   0,   2,   0,   0, 
+      0, 144,   0,   8,  15, 160, 
+     31,   0,   0,   2,   0,   0, 
+      0, 144,   1,   8,  15, 160, 
+     66,   0,   0,   3,   0,   0, 
+     15, 128,   0,   0, 228, 176, 
+      1,   8, 228, 160,   4,   0, 
+      0,   4,   0,   0,   1, 128, 
+      0,   0,   0, 128,   0,   0, 
+      0, 160,   0,   0,  85, 160, 
+      5,   0,   0,   3,   0,   0, 
+      3, 128,   0,   0,   0, 128, 
+      0,   0, 170, 160,  66,   0, 
+      0,   3,   0,   0,  15, 128, 
+      0,   0, 228, 128,   0,   8, 
+    228, 160,   1,   0,   0,   2, 
+      0,   8,  15, 128,   0,   0, 
+    228, 128, 255, 255,   0,   0
+};
+
diff --git a/pacc/hlsl/fill.ps.hlsl b/pacc/hlsl/fill.ps.hlsl
new file mode 100644
index 0000000..9939553
--- /dev/null
+++ b/pacc/hlsl/fill.ps.hlsl
@@ -0,0 +1,15 @@
+struct psinput {
+  float4 texcoord: TEXCOORD;
+};
+
+struct psoutput {
+  float4 fragcolor: COLOR;
+};
+
+sampler palette: register(s0);
+
+psoutput fill(psinput input) {
+  psoutput output;
+  output.fragcolor = tex1D(palette, input.texcoord.x);
+  return output;
+}
diff --git a/pacc/hlsl/fill.ps.inc b/pacc/hlsl/fill.ps.inc
new file mode 100644
index 0000000..3fcc075
--- /dev/null
+++ b/pacc/hlsl/fill.ps.inc
@@ -0,0 +1,65 @@
+#if 0
+//
+// Generated by Microsoft (R) HLSL Shader Compiler 9.29.952.3111
+//
+//   fxc /nologo /O3 /Tps_2_0 /Efill /Fh fill.ps.inc.tmp fill.ps.hlsl
+//
+//
+// Parameters:
+//
+//   sampler2D palette;
+//
+//
+// Registers:
+//
+//   Name         Reg   Size
+//   ------------ ----- ----
+//   palette      s0       1
+//
+
+    ps_2_0
+    dcl t0.xy
+    dcl_2d s0
+    texld r0, t0, s0
+    mov oC0, r0
+
+// approximately 2 instruction slots used (1 texture, 1 arithmetic)
+#endif
+
+static const BYTE ps20_fill[] =
+{
+      0,   2, 255, 255, 254, 255, 
+     33,   0,  67,  84,  65,  66, 
+     28,   0,   0,   0,  79,   0, 
+      0,   0,   0,   2, 255, 255, 
+      1,   0,   0,   0,  28,   0, 
+      0,   0,   0, 129,   0,   0, 
+     72,   0,   0,   0,  48,   0, 
+      0,   0,   3,   0,   0,   0, 
+      1,   0,   2,   0,  56,   0, 
+      0,   0,   0,   0,   0,   0, 
+    112,  97, 108, 101, 116, 116, 
+    101,   0,   4,   0,  12,   0, 
+      1,   0,   1,   0,   1,   0, 
+      0,   0,   0,   0,   0,   0, 
+    112, 115,  95,  50,  95,  48, 
+      0,  77, 105,  99, 114, 111, 
+    115, 111, 102, 116,  32,  40, 
+     82,  41,  32,  72,  76,  83, 
+     76,  32,  83, 104,  97, 100, 
+    101, 114,  32,  67, 111, 109, 
+    112, 105, 108, 101, 114,  32, 
+     57,  46,  50,  57,  46,  57, 
+     53,  50,  46,  51,  49,  49, 
+     49,   0,  31,   0,   0,   2, 
+      0,   0,   0, 128,   0,   0, 
+      3, 176,  31,   0,   0,   2, 
+      0,   0,   0, 144,   0,   8, 
+     15, 160,  66,   0,   0,   3, 
+      0,   0,  15, 128,   0,   0, 
+    228, 176,   0,   8, 228, 160, 
+      1,   0,   0,   2,   0,   8, 
+     15, 128,   0,   0, 228, 128, 
+    255, 255,   0,   0
+};
+
diff --git a/pacc/hlsl/font.ps.hlsl b/pacc/hlsl/font.ps.hlsl
new file mode 100644
index 0000000..406da2e
--- /dev/null
+++ b/pacc/hlsl/font.ps.hlsl
@@ -0,0 +1,24 @@
+struct psinput {
+  float4 texcoord: TEXCOORD;
+};
+
+struct psoutput {
+  float4 fragcolor: COLOR;
+};
+
+sampler palette: register(s0);
+sampler tex: register(s1);
+float bg: register(c0);
+float color: register(c1);
+
+psoutput font(psinput input) {
+  psoutput output;
+  float pixel = tex2D(tex, input.texcoord.xy).r;
+  float index = color;
+  if (pixel < 0.5) {
+    if (bg < 0.5) discard;
+    index = 0.5 / 256.0;
+  }
+  output.fragcolor = tex1D(palette, index);
+  return output;
+}
diff --git a/pacc/hlsl/font.ps.inc b/pacc/hlsl/font.ps.inc
new file mode 100644
index 0000000..2167cf7
--- /dev/null
+++ b/pacc/hlsl/font.ps.inc
@@ -0,0 +1,126 @@
+#if 0
+//
+// Generated by Microsoft (R) HLSL Shader Compiler 9.29.952.3111
+//
+//   fxc /nologo /O3 /Tps_2_0 /Efont /Fh font.ps.inc.tmp font.ps.hlsl
+//
+//
+// Parameters:
+//
+//   float bg;
+//   float color;
+//   sampler2D palette;
+//   sampler2D tex;
+//
+//
+// Registers:
+//
+//   Name         Reg   Size
+//   ------------ ----- ----
+//   bg           c0       1
+//   color        c1       1
+//   palette      s0       1
+//   tex          s1       1
+//
+
+    ps_2_0
+    def c2, -0.5, -0, -1, 0.001953125
+    dcl t0.xy
+    dcl_2d s0
+    dcl_2d s1
+    texld r0, t0, s1
+    add r0.x, r0.x, c2.x
+    mov r1.xw, c2
+    add r0.y, r1.x, c0.x
+    cmp r0.y, r0.y, c2.y, c2.z
+    cmp r2, r0.x, c2.y, r0.y
+    texkill r2
+    cmp r0.xy, r0.x, c1.x, r1.w
+    texld r0, r0, s0
+    mov oC0, r0
+
+// approximately 10 instruction slots used (2 texture, 8 arithmetic)
+#endif
+
+static const BYTE ps20_font[] =
+{
+      0,   2, 255, 255, 254, 255, 
+     60,   0,  67,  84,  65,  66, 
+     28,   0,   0,   0, 187,   0, 
+      0,   0,   0,   2, 255, 255, 
+      4,   0,   0,   0,  28,   0, 
+      0,   0,   0, 129,   0,   0, 
+    180,   0,   0,   0, 108,   0, 
+      0,   0,   2,   0,   0,   0, 
+      1,   0,   2,   0, 112,   0, 
+      0,   0,   0,   0,   0,   0, 
+    128,   0,   0,   0,   2,   0, 
+      1,   0,   1,   0,   6,   0, 
+    112,   0,   0,   0,   0,   0, 
+      0,   0, 134,   0,   0,   0, 
+      3,   0,   0,   0,   1,   0, 
+      2,   0, 144,   0,   0,   0, 
+      0,   0,   0,   0, 160,   0, 
+      0,   0,   3,   0,   1,   0, 
+      1,   0,   6,   0, 164,   0, 
+      0,   0,   0,   0,   0,   0, 
+     98, 103,   0, 171,   0,   0, 
+      3,   0,   1,   0,   1,   0, 
+      1,   0,   0,   0,   0,   0, 
+      0,   0,  99, 111, 108, 111, 
+    114,   0, 112,  97, 108, 101, 
+    116, 116, 101,   0, 171, 171, 
+      4,   0,  12,   0,   1,   0, 
+      1,   0,   1,   0,   0,   0, 
+      0,   0,   0,   0, 116, 101, 
+    120,   0,   4,   0,  12,   0, 
+      1,   0,   1,   0,   1,   0, 
+      0,   0,   0,   0,   0,   0, 
+    112, 115,  95,  50,  95,  48, 
+      0,  77, 105,  99, 114, 111, 
+    115, 111, 102, 116,  32,  40, 
+     82,  41,  32,  72,  76,  83, 
+     76,  32,  83, 104,  97, 100, 
+    101, 114,  32,  67, 111, 109, 
+    112, 105, 108, 101, 114,  32, 
+     57,  46,  50,  57,  46,  57, 
+     53,  50,  46,  51,  49,  49, 
+     49,   0,  81,   0,   0,   5, 
+      2,   0,  15, 160,   0,   0, 
+      0, 191,   0,   0,   0, 128, 
+      0,   0, 128, 191,   0,   0, 
+      0,  59,  31,   0,   0,   2, 
+      0,   0,   0, 128,   0,   0, 
+      3, 176,  31,   0,   0,   2, 
+      0,   0,   0, 144,   0,   8, 
+     15, 160,  31,   0,   0,   2, 
+      0,   0,   0, 144,   1,   8, 
+     15, 160,  66,   0,   0,   3, 
+      0,   0,  15, 128,   0,   0, 
+    228, 176,   1,   8, 228, 160, 
+      2,   0,   0,   3,   0,   0, 
+      1, 128,   0,   0,   0, 128, 
+      2,   0,   0, 160,   1,   0, 
+      0,   2,   1,   0,   9, 128, 
+      2,   0, 228, 160,   2,   0, 
+      0,   3,   0,   0,   2, 128, 
+      1,   0,   0, 128,   0,   0, 
+      0, 160,  88,   0,   0,   4, 
+      0,   0,   2, 128,   0,   0, 
+     85, 128,   2,   0,  85, 160, 
+      2,   0, 170, 160,  88,   0, 
+      0,   4,   2,   0,  15, 128, 
+      0,   0,   0, 128,   2,   0, 
+     85, 160,   0,   0,  85, 128, 
+     65,   0,   0,   1,   2,   0, 
+     15, 128,  88,   0,   0,   4, 
+      0,   0,   3, 128,   0,   0, 
+      0, 128,   1,   0,   0, 160, 
+      1,   0, 255, 128,  66,   0, 
+      0,   3,   0,   0,  15, 128, 
+      0,   0, 228, 128,   0,   8, 
+    228, 160,   1,   0,   0,   2, 
+      0,   8,  15, 128,   0,   0, 
+    228, 128, 255, 255,   0,   0
+};
+
diff --git a/pacc/hlsl/key.ps.hlsl b/pacc/hlsl/key.ps.hlsl
new file mode 100644
index 0000000..0ca9efd
--- /dev/null
+++ b/pacc/hlsl/key.ps.hlsl
@@ -0,0 +1,20 @@
+struct psinput {
+  float4 texcoord: TEXCOORD;
+};
+
+struct psoutput {
+  float4 fragcolor: COLOR;
+};
+
+sampler palette: register(s0);
+sampler tex: register(s1);
+float uni_key: register(c0);
+float color: register(c1);
+
+psoutput key(psinput input) {
+  psoutput output;
+  float index = tex2D(tex, input.texcoord.xy).r;
+  if (index < (uni_key + (0.5/255.0)) || (uni_key + (1.5/255.0)) < index) discard;
+  output.fragcolor = tex1D(palette, color);
+  return output;
+}
diff --git a/pacc/hlsl/key.ps.inc b/pacc/hlsl/key.ps.inc
new file mode 100644
index 0000000..63d51d8
--- /dev/null
+++ b/pacc/hlsl/key.ps.inc
@@ -0,0 +1,141 @@
+#if 0
+//
+// Generated by Microsoft (R) HLSL Shader Compiler 9.29.952.3111
+//
+//   fxc /nologo /O3 /Tps_2_0 /Ekey /Fh key.ps.inc.tmp key.ps.hlsl
+//
+//
+// Parameters:
+//
+//   float color;
+//   sampler2D palette;
+//   sampler2D tex;
+//   float uni_key;
+//
+//
+// Registers:
+//
+//   Name         Reg   Size
+//   ------------ ----- ----
+//   uni_key      c0       1
+//   color        c1       1
+//   palette      s0       1
+//   tex          s1       1
+//
+
+    ps_2_0
+    def c2, 0.00196078443, 0, 1, 0.00588235306
+    dcl t0.xy
+    dcl_2d s0
+    dcl_2d s1
+    texld r0, t0, s1
+    mov r1.xw, c2
+    add r0.y, r1.w, c0.x
+    add r0.y, -r0.x, r0.y
+    cmp r0.y, r0.y, c2.y, c2.z
+    add r0.z, r1.x, c0.x
+    add r0.x, -r0.z, r0.x
+    cmp r0.x, r0.x, c2.y, c2.z
+    add r0.x, r0.y, r0.x
+    cmp r0, -r0.x, -c2.y, -c2.z
+    texkill r0
+    mov r0.xy, c1.x
+    texld r0, r0, s0
+    mov oC0, r0
+
+// approximately 14 instruction slots used (2 texture, 12 arithmetic)
+#endif
+
+static const BYTE ps20_key[] =
+{
+      0,   2, 255, 255, 254, 255, 
+     61,   0,  67,  84,  65,  66, 
+     28,   0,   0,   0, 191,   0, 
+      0,   0,   0,   2, 255, 255, 
+      4,   0,   0,   0,  28,   0, 
+      0,   0,   0, 129,   0,   0, 
+    184,   0,   0,   0, 108,   0, 
+      0,   0,   2,   0,   1,   0, 
+      1,   0,   6,   0, 116,   0, 
+      0,   0,   0,   0,   0,   0, 
+    132,   0,   0,   0,   3,   0, 
+      0,   0,   1,   0,   2,   0, 
+    140,   0,   0,   0,   0,   0, 
+      0,   0, 156,   0,   0,   0, 
+      3,   0,   1,   0,   1,   0, 
+      6,   0, 160,   0,   0,   0, 
+      0,   0,   0,   0, 176,   0, 
+      0,   0,   2,   0,   0,   0, 
+      1,   0,   2,   0, 116,   0, 
+      0,   0,   0,   0,   0,   0, 
+     99, 111, 108, 111, 114,   0, 
+    171, 171,   0,   0,   3,   0, 
+      1,   0,   1,   0,   1,   0, 
+      0,   0,   0,   0,   0,   0, 
+    112,  97, 108, 101, 116, 116, 
+    101,   0,   4,   0,  12,   0, 
+      1,   0,   1,   0,   1,   0, 
+      0,   0,   0,   0,   0,   0, 
+    116, 101, 120,   0,   4,   0, 
+     12,   0,   1,   0,   1,   0, 
+      1,   0,   0,   0,   0,   0, 
+      0,   0, 117, 110, 105,  95, 
+    107, 101, 121,   0, 112, 115, 
+     95,  50,  95,  48,   0,  77, 
+    105,  99, 114, 111, 115, 111, 
+    102, 116,  32,  40,  82,  41, 
+     32,  72,  76,  83,  76,  32, 
+     83, 104,  97, 100, 101, 114, 
+     32,  67, 111, 109, 112, 105, 
+    108, 101, 114,  32,  57,  46, 
+     50,  57,  46,  57,  53,  50, 
+     46,  51,  49,  49,  49,   0, 
+     81,   0,   0,   5,   2,   0, 
+     15, 160, 129, 128,   0,  59, 
+      0,   0,   0,   0,   0,   0, 
+    128,  63, 193, 192, 192,  59, 
+     31,   0,   0,   2,   0,   0, 
+      0, 128,   0,   0,   3, 176, 
+     31,   0,   0,   2,   0,   0, 
+      0, 144,   0,   8,  15, 160, 
+     31,   0,   0,   2,   0,   0, 
+      0, 144,   1,   8,  15, 160, 
+     66,   0,   0,   3,   0,   0, 
+     15, 128,   0,   0, 228, 176, 
+      1,   8, 228, 160,   1,   0, 
+      0,   2,   1,   0,   9, 128, 
+      2,   0, 228, 160,   2,   0, 
+      0,   3,   0,   0,   2, 128, 
+      1,   0, 255, 128,   0,   0, 
+      0, 160,   2,   0,   0,   3, 
+      0,   0,   2, 128,   0,   0, 
+      0, 129,   0,   0,  85, 128, 
+     88,   0,   0,   4,   0,   0, 
+      2, 128,   0,   0,  85, 128, 
+      2,   0,  85, 160,   2,   0, 
+    170, 160,   2,   0,   0,   3, 
+      0,   0,   4, 128,   1,   0, 
+      0, 128,   0,   0,   0, 160, 
+      2,   0,   0,   3,   0,   0, 
+      1, 128,   0,   0, 170, 129, 
+      0,   0,   0, 128,  88,   0, 
+      0,   4,   0,   0,   1, 128, 
+      0,   0,   0, 128,   2,   0, 
+     85, 160,   2,   0, 170, 160, 
+      2,   0,   0,   3,   0,   0, 
+      1, 128,   0,   0,  85, 128, 
+      0,   0,   0, 128,  88,   0, 
+      0,   4,   0,   0,  15, 128, 
+      0,   0,   0, 129,   2,   0, 
+     85, 161,   2,   0, 170, 161, 
+     65,   0,   0,   1,   0,   0, 
+     15, 128,   1,   0,   0,   2, 
+      0,   0,   3, 128,   1,   0, 
+      0, 160,  66,   0,   0,   3, 
+      0,   0,  15, 128,   0,   0, 
+    228, 128,   0,   8, 228, 160, 
+      1,   0,   0,   2,   0,   8, 
+     15, 128,   0,   0, 228, 128, 
+    255, 255,   0,   0
+};
+
diff --git a/pacc/pacc-d3d9.c b/pacc/pacc-d3d9.c
new file mode 100644
index 0000000..21f3210
--- /dev/null
+++ b/pacc/pacc-d3d9.c
@@ -0,0 +1,633 @@
+#include "pacc-win.h"
+#include <d3d9.h>
+#include <stdlib.h>
+#include <stdatomic.h>
+#include <stdint.h>
+#include <stdio.h>
+
+#include "hlsl/blit.vs.inc"
+#include "hlsl/copy.ps.inc"
+#include "hlsl/color.ps.inc"
+#include "hlsl/color_trans.ps.inc"
+
+struct pacc_ctx {
+  int w, h;
+  uint8_t pal[256*3];
+  bool pal_changed;
+  uint8_t clearpal;
+  uint8_t color;
+  uint8_t color_changed;
+  bool render_enabled;
+  enum pacc_mode curr_mode;
+  atomic_bool killthread;
+  HWND hwnd;
+  HMODULE d3d9m;
+  IDirect3D9 *d3d9;
+  IDirect3DDevice9 *d3d9d;
+  IDirect3DTexture9 *tex_pal;
+  IDirect3DVertexDeclaration9 *vdecl;
+  IDirect3DVertexShader9 *vs;
+  IDirect3DPixelShader9 *ps_modes[pacc_mode_count];
+  IDirect3DStateBlock9 *state_common;
+  IDirect3DStateBlock9 *state_modes[pacc_mode_count];
+  pacc_rendercb *rendercb;
+  void *renderptr;
+  HANDLE renderthread;
+  HANDLE rendermtx;
+};
+
+struct pacc_buf {
+  IDirect3DVertexBuffer9 *buf_obj;
+  float *buf;
+  struct pacc_tex *tex;
+  int len;
+  int buflen;
+  int bufobjlen;
+  bool changed;
+};
+
+struct pacc_tex {
+  int w, h;
+  bool changed;
+  atomic_flag flag;
+  IDirect3DTexture9 *tex_obj;
+  uint8_t *buf;
+};
+
+enum {
+  PACC_BUF_DEF_LEN = 32,
+  PRINTBUFLEN = 160,
+};
+
+static void pacc_buf_delete(struct pacc_buf *pb) {
+  if (pb) {
+    if (pb->buf_obj) pb->buf_obj->lpVtbl->Release(pb->buf_obj);
+    free(pb->buf);
+    free(pb);
+  }
+}
+
+static struct pacc_buf *pacc_gen_buf(
+    struct pacc_ctx *pc, struct pacc_tex *pt, enum pacc_buf_mode mode) {
+  struct pacc_buf *pb = malloc(sizeof(*pb));
+  if (!pb) goto err;
+  *pb = (struct pacc_buf) {
+    .buflen = PACC_BUF_DEF_LEN,
+    .bufobjlen = PACC_BUF_DEF_LEN,
+    .tex = pt,
+  };
+  pb->buf = malloc(sizeof(*pb->buf) * pb->buflen);
+  if (!pb->buf) goto err;
+  HRESULT res;
+  res = pc->d3d9d->lpVtbl->CreateVertexBuffer(
+      pc->d3d9d,
+      sizeof(*pb->buf) * pb->buflen,
+      D3DUSAGE_DYNAMIC,
+      0,
+      D3DPOOL_DEFAULT,
+      &pb->buf_obj,
+      0);
+  if (res != D3D_OK) goto err;
+  return pb;
+err:
+  MessageBox(pc->hwnd, L"Hello", L"HI", 0);
+  pacc_buf_delete(pb);
+  return 0;
+}
+
+static bool buf_reserve(struct pacc_buf *pb, int len) {
+  if (pb->len + len > pb->buflen) {
+    int newlen = pb->buflen;
+    while (pb->len + len > newlen) newlen *= 2;
+    float *newbuf = realloc(pb->buf, newlen * sizeof(pb->buf[0]));
+    if (!newbuf) return false;
+    pb->buflen = newlen;
+    pb->buf = newbuf;
+  }
+  return true;
+}
+
+static void pacc_calc_scale(float *ret, int w, int h, int wdest, int hdest) {
+  ret[0] = ((float)w) / wdest;
+  ret[1] = ((float)h) / hdest;
+}
+
+static void pacc_calc_off_tex(float *ret,
+                                     int tw, int th, int xsrc, int ysrc) {
+  ret[0] = ((float)xsrc) / tw;
+  ret[1] = ((float)ysrc) / th;
+}
+
+static void pacc_calc_off(
+    float *ret, int xdest, int ydest, int w, int h, int wdest, int hdest) {
+  ret[0] = ((float)(xdest * 2 + w - wdest)) / wdest;
+  ret[1] = ((float)(ydest * 2 + h - hdest)) / hdest;
+}
+
+static void pacc_buf_rect_off(
+    const struct pacc_ctx *pc, struct pacc_buf *pb,
+    int x, int y, int w, int h, int xoff, int yoff) {
+  float scale[2];
+  float off[2];
+  float tscale[2];
+  float toff[2];
+  pacc_calc_off(off, x, y, w, h, pc->w, pc->h);
+  pacc_calc_scale(scale, w, h, pc->w, pc->h);
+  pacc_calc_off_tex(toff, pb->tex->w, pb->tex->h, xoff, yoff);
+  pacc_calc_scale(tscale, w, h, pb->tex->w, pb->tex->h);
+  float coord[16] = {
+    -1.0f * scale[0] + off[0], -1.0f * scale[1] - off[1],
+     0.0f * tscale[0] + toff[0],  1.0f * tscale[1] + toff[1],
+
+    -1.0f * scale[0] + off[0],  1.0f * scale[1] - off[1],
+     0.0f * tscale[0] + toff[0],  0.0f * tscale[1] + toff[1],
+
+     1.0f * scale[0] + off[0], -1.0f * scale[1] - off[1],
+     1.0f * tscale[0] + toff[0],  1.0f * tscale[1] + toff[1],
+
+     1.0f * scale[0] + off[0],  1.0f * scale[1] - off[1],
+     1.0f * tscale[0] + toff[0],  0.0f * tscale[1] + toff[1],
+  };
+  if (!buf_reserve(pb, 24)) return;
+  int indices[6] = {0, 1, 2, 2, 1, 3};
+  for (int i = 0; i < 6; i++) {
+    for (int j = 0; j < 4; j++) {
+      pb->buf[pb->len+i*4+j] = coord[indices[i]*4+j];
+    }
+  }
+  pb->len += 24;
+  pb->changed = true;
+}
+
+static void pacc_buf_vprintf(
+    const struct pacc_ctx *pc, struct pacc_buf *pb,
+    int x, int y, const char *fmt, va_list ap) {
+  uint8_t printbuf[PRINTBUFLEN+1];
+  vsnprintf((char *)printbuf, sizeof(printbuf), fmt, ap);
+  int len = strlen((const char *)printbuf);
+  float scale[2];
+  float off[2];
+  int w = pb->tex->w / 256;
+  int h = pb->tex->h;
+  pacc_calc_scale(scale, w, h, pc->w, pc->h);
+  pacc_calc_off(off, x, y, w, h, pc->w, pc->h);
+  if (!buf_reserve(pb, len*24)) return;
+  float *coords = pb->buf + pb->len;
+  for (int i = 0; i < len; i++) {
+    coords[24*i+0*4+0]                      = (-1.0f + 2.0f*i) * scale[0] + off[0];
+    coords[24*i+0*4+1]                      = -1.0f * scale[1] - off[1];
+    coords[24*i+1*4+0] = coords[24*i+4*4+0] = (-1.0f + 2.0f*i) * scale[0] + off[0];
+    coords[24*i+1*4+1] = coords[24*i+4*4+1] = 1.0f * scale[1] - off[1];
+    coords[24*i+2*4+0] = coords[24*i+3*4+0] = (1.0f + 2.0f*i) * scale[0] + off[0];
+    coords[24*i+2*4+1] = coords[24*i+3*4+1] = -1.0f * scale[1] - off[1];
+    coords[24*i+5*4+0]                      = (1.0f + 2.0f*i) * scale[0] + off[0];
+    coords[24*i+5*4+1]                      = 1.0f * scale[1] - off[1];
+    coords[24*i+0*4+2]                      = ((float)printbuf[i]) / 256.0f;
+    coords[24*i+0*4+3]                      = 1.0f;
+    coords[24*i+1*4+2] = coords[24*i+4*4+2] = ((float)printbuf[i]) / 256.0f;
+    coords[24*i+1*4+3] = coords[24*i+4*4+3] = 0.0f;
+    coords[24*i+2*4+2] = coords[24*i+3*4+2] = ((float)(printbuf[i]+1)) / 256.0f;
+    coords[24*i+2*4+3] = coords[24*i+3*4+3] = 1.0f;
+    coords[24*i+5*4+2]                      = ((float)(printbuf[i]+1)) / 256.0f;
+    coords[24*i+5*4+3]                      = 0.0f;
+  }
+  pb->len += len * 24;
+  pb->changed = true;
+}
+
+static void pacc_buf_printf(
+    const struct pacc_ctx *pc, struct pacc_buf *pb,
+    int x, int y, const char *fmt, ...) {
+  va_list ap;
+  va_start(ap, fmt);
+  pacc_buf_vprintf(pc, pb, x, y, fmt, ap);
+  va_end(ap);
+}
+
+static void pacc_buf_rect(
+    const struct pacc_ctx *pc, struct pacc_buf *pb,
+    int x, int y, int w, int h) {
+  pacc_buf_rect_off(pc, pb, x, y, w, h, 0, 0);
+}
+
+static void pacc_buf_clear(struct pacc_buf *pb) {
+  pb->len = 0;
+  pb->changed = true;
+}
+
+static void pacc_palette(struct pacc_ctx *pc, const uint8_t *rgb, int colors) {
+  memcpy(pc->pal, rgb, colors*3);
+  pc->pal_changed = true;
+}
+
+static void pacc_color(struct pacc_ctx *pc, uint8_t pal) {
+  pc->color = pal;
+  pc->color_changed = true;
+}
+
+static void pacc_begin_clear(struct pacc_ctx *pc) {
+  if (pc->pal_changed) {
+    D3DLOCKED_RECT lockrect;
+    if (SUCCEEDED(pc->tex_pal->lpVtbl->LockRect(
+            pc->tex_pal, 0, &lockrect, 0, D3DLOCK_DISCARD))) {
+      uint32_t *palpixel = lockrect.pBits;
+      for (int i = 0; i < 256; i++) {
+        palpixel[i] =
+          ((uint32_t)pc->pal[i*3+0]) << 16 |
+          ((uint32_t)pc->pal[i*3+1]) << 8 |
+          pc->pal[i*3+2];
+      }
+      pc->tex_pal->lpVtbl->UnlockRect(
+          pc->tex_pal, 0);
+    }
+    pc->pal_changed = false;
+  }
+  pc->d3d9d->lpVtbl->Clear(
+      pc->d3d9d,
+      0, 0, D3DCLEAR_TARGET,
+      D3DCOLOR_RGBA(
+        pc->pal[0], pc->pal[1], pc->pal[2], 0xff
+        ), 1.0, 0);
+  pc->state_common->lpVtbl->Apply(pc->state_common);
+  float offsets[4] = {
+    -0.5f / pc->w, 0.5f / pc->h, 0.0f, 0.0f,
+  };
+  pc->d3d9d->lpVtbl->SetVertexShaderConstantF(
+      pc->d3d9d,
+      0, offsets, 1);
+  pc->curr_mode = pacc_mode_count;
+}
+
+static void pacc_draw(struct pacc_ctx *pc, struct pacc_buf *pb, enum pacc_mode mode) {
+  if (mode >= pacc_mode_count) return;
+  if (pc->curr_mode != mode) {
+    pc->state_modes[mode]->lpVtbl->Apply(pc->state_modes[mode]);
+    pc->curr_mode = mode;
+  }
+  if (mode != pacc_mode_copy && pc->color_changed) {
+    float fbuf[4] = {pc->color / 255.f};
+    pc->d3d9d->lpVtbl->SetPixelShaderConstantF(pc->d3d9d, 0, fbuf, 1);
+    pc->color_changed = false;
+  }
+  if (pb->tex->changed) {
+    D3DLOCKED_RECT lockrect;
+    if (SUCCEEDED(pb->tex->tex_obj->lpVtbl->LockRect(
+            pb->tex->tex_obj, 0, &lockrect, 0, 0))) {
+      while (atomic_flag_test_and_set_explicit(&pb->tex->flag, memory_order_acquire));
+      for (int y = 0; y < pb->tex->h; y++) {
+        memcpy(
+            (char *)lockrect.pBits + lockrect.Pitch * y,
+            pb->tex->buf + pb->tex->w * y,
+            pb->tex->w);
+      }
+      atomic_flag_clear_explicit(&pb->tex->flag, memory_order_release);
+      pb->tex->tex_obj->lpVtbl->UnlockRect(
+          pb->tex->tex_obj, 0);
+    }
+    pb->tex->changed = false;
+  }
+  if (pb->changed) {
+    if (pb->buflen > pb->bufobjlen) {
+      IDirect3DVertexBuffer9 *newbufobj;
+      HRESULT res;
+      res = pc->d3d9d->lpVtbl->CreateVertexBuffer(
+          pc->d3d9d,
+          sizeof(*pb->buf) * pb->buflen,
+          D3DUSAGE_DYNAMIC,
+          0,
+          D3DPOOL_DEFAULT,
+          &newbufobj,
+          0);
+      if (res != D3D_OK) return;
+      pb->buf_obj->lpVtbl->Release(pb->buf_obj);
+      pb->buf_obj = newbufobj;
+      pb->bufobjlen = pb->buflen;
+    }
+    float *objbuf;
+    if (SUCCEEDED(pb->buf_obj->lpVtbl->Lock(
+            pb->buf_obj,
+            0, sizeof(float)*pb->len,
+            (void **)&objbuf,
+            D3DLOCK_DISCARD))) {
+      memcpy(objbuf, pb->buf, pb->len*sizeof(pb->buf[0]));
+      pb->buf_obj->lpVtbl->Unlock(pb->buf_obj);
+    }
+    pb->changed = false;
+  }
+  DWORD texaddru, texaddrv;
+  if (pb->tex->w & (pb->tex->w - 1)) {
+    texaddru = D3DTADDRESS_CLAMP;
+  } else {
+    texaddru = D3DTADDRESS_WRAP;
+  }
+  if (pb->tex->h & (pb->tex->h - 1)) {
+    texaddrv = D3DTADDRESS_CLAMP;
+  } else {
+    texaddrv = D3DTADDRESS_WRAP;
+  }
+  pc->d3d9d->lpVtbl->SetSamplerState(
+      pc->d3d9d, 1, D3DSAMP_ADDRESSU, texaddru);
+  pc->d3d9d->lpVtbl->SetSamplerState(
+      pc->d3d9d, 1, D3DSAMP_ADDRESSV, texaddrv);
+  pc->d3d9d->lpVtbl->SetTexture(
+      pc->d3d9d, 1, (IDirect3DBaseTexture9 *)pb->tex->tex_obj);
+  pc->d3d9d->lpVtbl->SetStreamSource(
+      pc->d3d9d,
+      0,
+      pb->buf_obj,
+      0, 4*sizeof(float));
+  pc->d3d9d->lpVtbl->DrawPrimitive(
+      pc->d3d9d,
+      D3DPT_TRIANGLELIST,
+      0, pb->len/(3*4));
+}
+
+static uint8_t *pacc_tex_lock(struct pacc_tex *pt) {
+  while (atomic_flag_test_and_set_explicit(&pt->flag, memory_order_acquire));
+  return pt->buf;
+}
+
+static void pacc_tex_unlock(struct pacc_tex *pt) {
+  pt->changed = true;
+  atomic_flag_clear_explicit(&pt->flag, memory_order_release);
+}
+
+static void pacc_tex_delete(struct pacc_tex *pt) {
+  if (pt) {
+    if (pt->tex_obj) pt->tex_obj->lpVtbl->Release(pt->tex_obj);
+    free(pt->buf);
+    free(pt);
+  }
+}
+
+static struct pacc_tex *pacc_gen_tex(struct pacc_ctx *pc, int w, int h) {
+  struct pacc_tex *pt = malloc(sizeof(*pt));
+  if (!pt) goto err;
+  *pt = (struct pacc_tex) {
+    .w = w,
+    .h = h,
+    .buf = calloc(w*h, 1),
+  };
+  if (!pt->buf) goto err;
+  atomic_flag_clear_explicit(&pt->flag, memory_order_release);
+  HRESULT res;
+  res = pc->d3d9d->lpVtbl->CreateTexture(
+      pc->d3d9d,
+      w, h,
+      1,
+      D3DUSAGE_DYNAMIC,
+      D3DFMT_L8,
+      D3DPOOL_DEFAULT,
+      &pt->tex_obj,
+      0);
+  if (res != D3D_OK) goto err;
+  return pt;
+err:
+  pacc_tex_delete(pt);
+  return 0;
+}
+
+static void pacc_delete(struct pacc_ctx *pc) {
+  if (pc) {
+    if (pc->renderthread) {
+      atomic_store_explicit(&pc->killthread, true, memory_order_relaxed);
+      WaitForSingleObject(pc->renderthread, INFINITE);
+    }
+    for (int i = 0; i < pacc_mode_count; i++) {
+      if (pc->ps_modes[i]) pc->ps_modes[i]->lpVtbl->Release(pc->ps_modes[i]);
+      if (pc->state_modes[i]) pc->state_modes[i]->lpVtbl->Release(pc->state_modes[i]);
+    }
+    if (pc->state_common) pc->state_common->lpVtbl->Release(pc->state_common);
+    if (pc->vs) pc->vs->lpVtbl->Release(pc->vs);
+    if (pc->vdecl) pc->vdecl->lpVtbl->Release(pc->vdecl);
+    if (pc->tex_pal) pc->tex_pal->lpVtbl->Release(pc->tex_pal);
+    if (pc->d3d9d) pc->d3d9d->lpVtbl->Release(pc->d3d9d);
+    if (pc->d3d9) pc->d3d9->lpVtbl->Release(pc->d3d9);
+    if (pc->d3d9m) FreeLibrary(pc->d3d9m);
+    if (pc->rendermtx) CloseHandle(pc->rendermtx);
+    free(pc);
+  }
+}
+
+static DWORD WINAPI pacc_renderproc(void *ptr) {
+  struct pacc_ctx *pc = ptr;
+  while (!atomic_load_explicit(&pc->killthread, memory_order_relaxed)) {
+    if (WaitForSingleObject(pc->rendermtx, INFINITE) == WAIT_OBJECT_0) {
+      if (SUCCEEDED(pc->d3d9d->lpVtbl->BeginScene(pc->d3d9d))) {
+        if (pc->render_enabled) {
+          pc->rendercb(pc->renderptr);
+        } else {
+          pc->d3d9d->lpVtbl->Clear(
+              pc->d3d9d,
+              0, 0, D3DCLEAR_TARGET,
+              D3DCOLOR_RGBA(0x00, 0x00, 0x00, 0xff),
+              1.0, 0);
+        }
+        pc->d3d9d->lpVtbl->EndScene(pc->d3d9d);
+        pc->d3d9d->lpVtbl->Present(pc->d3d9d, 0, 0, 0, 0);
+      }
+      ReleaseMutex(pc->rendermtx);
+    }
+  }
+  return 0;
+}
+
+static void pacc_renderctrl(struct pacc_ctx *pc, bool enable) {
+  while (WaitForSingleObject(pc->rendermtx, INFINITE) != WAIT_OBJECT_0);
+  pc->render_enabled = enable;
+  ReleaseMutex(pc->rendermtx);
+}
+
+static struct pacc_vtable pacc_d3d9_vtable = {
+  .pacc_delete = pacc_delete,
+  .gen_buf = pacc_gen_buf,
+  .gen_tex = pacc_gen_tex,
+  .buf_delete = pacc_buf_delete,
+  .tex_lock = pacc_tex_lock,
+  .tex_unlock = pacc_tex_unlock,
+  .tex_delete = pacc_tex_delete,
+  .buf_rect = pacc_buf_rect,
+  .buf_rect_off = pacc_buf_rect_off,
+  .buf_vprintf = pacc_buf_vprintf,
+  .buf_printf = pacc_buf_printf,
+  .buf_clear = pacc_buf_clear,
+  .palette = pacc_palette,
+  .color = pacc_color,
+  .begin_clear = pacc_begin_clear,
+  .draw = pacc_draw,
+};
+
+struct pacc_win_vtable pacc_win_vtable = {
+  .renderctrl = pacc_renderctrl,
+};
+
+struct pacc_ctx *pacc_init_d3d9(
+    HWND hwnd,
+    pacc_rendercb *rendercb, void *renderptr,
+    struct pacc_vtable *vt, struct pacc_win_vtable *winvt) {
+  struct pacc_ctx *pc = malloc(sizeof(*pc));
+  if (!pc) goto err;
+  RECT wr;
+  if (!GetClientRect(hwnd, &wr)) goto err;
+  *pc = (struct pacc_ctx) {
+    .w = wr.right - wr.left,
+    .h = wr.bottom - wr.top,
+    .hwnd = hwnd,
+    .rendercb = rendercb,
+    .renderptr = renderptr,
+  };
+  atomic_init(&pc->killthread, false);
+  pc->rendermtx = CreateMutex(0, FALSE, 0);
+  if (!pc->rendermtx) goto err;
+  pc->d3d9m = LoadLibrary(L"d3d9");
+  if (!pc->d3d9m) goto err;
+  IDirect3D9 * WINAPI (*d3d9create)(UINT) = (IDirect3D9 * WINAPI (*)(UINT))GetProcAddress(pc->d3d9m, "Direct3DCreate9");
+  if (!d3d9create) goto err;
+  pc->d3d9 = d3d9create(D3D_SDK_VERSION);
+  if (!pc->d3d9) goto err;
+
+  D3DCAPS9 dcaps;
+  HRESULT res;
+  res = pc->d3d9->lpVtbl->GetDeviceCaps(
+      pc->d3d9,
+      D3DADAPTER_DEFAULT,
+      D3DDEVTYPE_HAL,
+      &dcaps);
+  if (res != D3D_OK) goto err;
+
+  D3DMULTISAMPLE_TYPE ms;
+#ifdef PACC_DEBUG_MSAA
+  ms = D3DMULTISAMPLE_8_SAMPLES;
+#else
+  ms = D3DMULTISAMPLE_NONE;
+#endif
+  D3DPRESENT_PARAMETERS dp = {
+    .MultiSampleType = ms,
+    .BackBufferWidth = pc->w * 2,
+    .BackBufferHeight = pc->h * 2,
+    .SwapEffect = D3DSWAPEFFECT_DISCARD,
+    .hDeviceWindow = pc->hwnd,
+    .Windowed = TRUE,
+  };
+  DWORD behavior = D3DCREATE_MULTITHREADED;
+  /* from SDL_renderer_d3d.c
+     fallback to software vertex processing when hardware vertex processing not available
+   */
+  behavior |= (dcaps.DevCaps & D3DDEVCAPS_HWTRANSFORMANDLIGHT) ?
+    D3DCREATE_HARDWARE_VERTEXPROCESSING :
+    D3DCREATE_SOFTWARE_VERTEXPROCESSING;
+  res = pc->d3d9->lpVtbl->CreateDevice(
+      pc->d3d9,
+      D3DADAPTER_DEFAULT, D3DDEVTYPE_HAL,
+      0,
+      behavior,
+      &dp, &pc->d3d9d);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->CreateTexture(
+      pc->d3d9d,
+      256, 1,
+      1,
+      D3DUSAGE_DYNAMIC,
+      D3DFMT_X8R8G8B8,
+      D3DPOOL_DEFAULT,
+      &pc->tex_pal,
+      0);
+  if (res != D3D_OK) goto err;
+
+  D3DVERTEXELEMENT9 vertexdecls[] = {
+    {
+      .Stream = 0,
+      .Offset = 0,
+      .Type = D3DDECLTYPE_FLOAT4,
+      .Method = D3DDECLMETHOD_DEFAULT,
+      .Usage = D3DDECLUSAGE_POSITION,
+      .UsageIndex = 0,
+    },
+    D3DDECL_END()
+  };
+  res = pc->d3d9d->lpVtbl->CreateVertexDeclaration(
+      pc->d3d9d,
+      vertexdecls,
+      &pc->vdecl);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->CreateVertexShader(
+      pc->d3d9d,
+      (const DWORD *)vs20_blit, &pc->vs);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->CreatePixelShader(
+      pc->d3d9d,
+      (const DWORD *)ps20_copy, &pc->ps_modes[pacc_mode_copy]);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->CreatePixelShader(
+      pc->d3d9d,
+      (const DWORD *)ps20_color, &pc->ps_modes[pacc_mode_color]);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->CreatePixelShader(
+      pc->d3d9d,
+      (const DWORD *)ps20_color_trans, &pc->ps_modes[pacc_mode_color_trans]);
+  if (res != D3D_OK) goto err;
+
+  res = pc->d3d9d->lpVtbl->BeginStateBlock(pc->d3d9d);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->SetVertexDeclaration(
+      pc->d3d9d, pc->vdecl);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->SetRenderState(
+      pc->d3d9d, D3DRS_CULLMODE, D3DCULL_NONE);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->SetVertexShader(
+      pc->d3d9d, pc->vs);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->SetSamplerState(
+      pc->d3d9d, 0, D3DSAMP_ADDRESSU, D3DTADDRESS_CLAMP);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->SetSamplerState(
+      pc->d3d9d, 0, D3DSAMP_ADDRESSV, D3DTADDRESS_CLAMP);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->SetSamplerState(
+      pc->d3d9d, 1, D3DSAMP_ADDRESSU, D3DTADDRESS_CLAMP);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->SetSamplerState(
+      pc->d3d9d, 1, D3DSAMP_ADDRESSV, D3DTADDRESS_CLAMP);
+  if (res != D3D_OK) goto err;
+#ifdef PACC_DEBUG_MSAA
+  res = pc->d3d9d->lpVtbl->SetSamplerState(
+      pc->d3d9d, 0, D3DSAMP_MAGFILTER, D3DTEXF_LINEAR);
+  if (res != D3D_OK) goto err;
+#endif
+  res = pc->d3d9d->lpVtbl->SetTexture(
+      pc->d3d9d, 0, (IDirect3DBaseTexture9 *)pc->tex_pal);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->EndStateBlock(pc->d3d9d, &pc->state_common);
+  if (res != D3D_OK) goto err;
+
+  res = pc->d3d9d->lpVtbl->BeginStateBlock(pc->d3d9d);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->SetPixelShader(pc->d3d9d, pc->ps_modes[pacc_mode_copy]);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->EndStateBlock(pc->d3d9d, &pc->state_modes[pacc_mode_copy]);
+  if (res != D3D_OK) goto err;
+
+  res = pc->d3d9d->lpVtbl->BeginStateBlock(pc->d3d9d);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->SetPixelShader(pc->d3d9d, pc->ps_modes[pacc_mode_color]);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->EndStateBlock(pc->d3d9d, &pc->state_modes[pacc_mode_color]);
+  if (res != D3D_OK) goto err;
+
+  res = pc->d3d9d->lpVtbl->BeginStateBlock(pc->d3d9d);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->SetPixelShader(pc->d3d9d, pc->ps_modes[pacc_mode_color_trans]);
+  if (res != D3D_OK) goto err;
+  res = pc->d3d9d->lpVtbl->EndStateBlock(pc->d3d9d, &pc->state_modes[pacc_mode_color_trans]);
+  if (res != D3D_OK) goto err;
+
+  pc->renderthread = CreateThread(0, 0, pacc_renderproc, pc, 0, 0);
+  if (!pc->renderthread) goto err;
+  *vt = pacc_d3d9_vtable;
+  *winvt = pacc_win_vtable;
+  return pc;
+err:
+  pacc_delete(pc);
+  return 0;
+}
diff --git a/pacc/pacc-gl-procs.inc b/pacc/pacc-gl-procs.inc
new file mode 100644
index 0000000..cd991e0
--- /dev/null
+++ b/pacc/pacc-gl-procs.inc
@@ -0,0 +1,46 @@
+#ifndef PROC_NO_GL_1_3 
+#ifndef PROC_NO_GL_1_1
+PROC(CLEAR, Clear)
+PROC(DELETETEXTURES, DeleteTextures)
+PROC(TEXIMAGE2D, TexImage2D)
+PROC(CLEARCOLOR, ClearColor)
+PROC(BINDTEXTURE, BindTexture)
+PROC(GENTEXTURES, GenTextures)
+PROC(TEXPARAMETERI, TexParameteri)
+PROC(PIXELSTOREI, PixelStorei)
+PROC(DRAWARRAYS, DrawArrays)
+#endif // PROC_NO_GL_1_1
+PROC(ACTIVETEXTURE, ActiveTexture)
+#endif // PROC_NO_GL_1_3
+PROC(DELETEPROGRAM, DeleteProgram)
+PROC(CREATESHADER, CreateShader)
+PROC(SHADERSOURCE, ShaderSource)
+PROC(COMPILESHADER, CompileShader)
+PROC(GETSHADERIV, GetShaderiv)
+#ifndef NDEBUG
+PROC(GETSHADERINFOLOG, GetShaderInfoLog)
+#endif
+PROC(DELETESHADER, DeleteShader)
+PROC(CREATEPROGRAM, CreateProgram)
+PROC(ATTACHSHADER, AttachShader)
+PROC(BINDATTRIBLOCATION, BindAttribLocation)
+PROC(LINKPROGRAM, LinkProgram)
+PROC(GETPROGRAMIV, GetProgramiv)
+#ifndef NDEBUG
+PROC(GETPROGRAMINFOLOG, GetProgramInfoLog)
+#endif
+PROC(USEPROGRAM, UseProgram)
+PROC(UNIFORM1I, Uniform1i)
+PROC(GETUNIFORMLOCATION, GetUniformLocation)
+PROC(DELETEBUFFERS, DeleteBuffers)
+PROC(GENBUFFERS, GenBuffers)
+PROC(UNIFORM1F, Uniform1f)
+PROC(BINDBUFFER, BindBuffer)
+PROC(BUFFERDATA, BufferData)
+PROC(ENABLEVERTEXATTRIBARRAY, EnableVertexAttribArray)
+PROC(VERTEXATTRIBPOINTER, VertexAttribPointer)
+#ifdef PACC_GL_3
+PROC(DELETEVERTEXARRAYS, DeleteVertexArrays)
+PROC(GENVERTEXARRAYS, GenVertexArrays)
+PROC(BINDVERTEXARRAY, BindVertexArray)
+#endif
diff --git a/pacc/pacc-gl.c b/pacc/pacc-gl.c
new file mode 100644
index 0000000..e6ed0a1
--- /dev/null
+++ b/pacc/pacc-gl.c
@@ -0,0 +1,505 @@
+#include "pacc-gl.h"
+#include <stdlib.h>
+#include <string.h>
+#include <stdbool.h>
+#include "pacc-gl-inc.h"
+
+#ifdef NDEBUG
+#define DPRINTF(fmt, ...)
+#else
+#include <stdio.h>
+#define DPRINTF(fmt, ...) fprintf(stderr, fmt, __VA_ARGS__)
+#endif
+
+/*
+  OpenGL versions:
+  OpenGL 2.0
+  OpenGL 3.2 core (#define PACC_GL_3)
+  OpenGL ES 2.0 (#define PACC_GL_ES)
+  OpenGL ES 3.0 (#define PACC_GL_ES, #define PACC_GL_3)
+
+  Shader languages:
+  GLSL 1.10 / GLSL ES 1.00
+*/
+
+#ifdef PACC_GL_ES
+#include "glsl/esheader.inc"
+#else
+#include "glsl/dsheader.inc"
+#endif
+
+#include "glsl/blit.vert.inc"
+#include "glsl/copy.frag.inc"
+#include "glsl/color.frag.inc"
+#include "glsl/color_trans.frag.inc"
+
+struct pacc_ctx {
+  int w;
+  int h;
+  uint8_t pal[256*3];
+  bool pal_changed;
+  uint8_t clearcolor[3];
+  GLuint tex_pal;
+  GLuint progs[pacc_mode_count];
+  GLint uni_color, uni_color_trans;
+  uint8_t color;
+  bool color_changed;
+  enum pacc_mode curr_mode;
+};
+
+struct pacc_buf {
+  GLuint buf_obj;
+  GLuint va_obj; // for OpenGL 3.2 Core
+  GLfloat *buf;
+  struct pacc_tex *tex;
+  int len;
+  int buflen;
+  GLenum usage;
+  bool changed;
+};
+
+struct pacc_tex {
+  GLuint tex_obj;
+  int w, h;
+  bool changed;
+  uint8_t *buf;
+};
+
+enum {
+  VAI_COORD,
+};
+
+enum {
+  PACC_BUF_DEF_LEN = 32,
+  PRINTBUFLEN = 160,
+};
+
+static void pacc_delete(struct pacc_ctx *pc) {
+  if (pc) {
+    glDeleteTextures(1, &pc->tex_pal);
+    for (int i = 0; i < pacc_mode_count; i++) {
+      glDeleteProgram(pc->progs[i]);
+    }
+    free(pc);
+  }
+}
+
+static GLuint compile_shader(const uint8_t *ss, GLenum type) {
+  GLuint s = glCreateShader(type);
+  if (!s) goto err;
+  const char *sourcelist[2] = {
+#ifdef PACC_GL_ES
+    (const char *)esheader,
+#else
+    (const char *)dsheader,
+#endif
+    (const char *)ss
+  };
+  glShaderSource(s, 2, sourcelist, 0);
+  glCompileShader(s);
+  GLint res;
+  glGetShaderiv(s, GL_COMPILE_STATUS, &res);
+  if (!res) {
+#ifndef NDEBUG
+    glGetShaderiv(s, GL_INFO_LOG_LENGTH, &res);
+    char *msgbuf = malloc(res);
+    if (msgbuf) {
+      glGetShaderInfoLog(s, res, 0, msgbuf);
+      DPRINTF("%s shader error: \n%s\n", type == GL_VERTEX_SHADER ? "vertex" : "fragment", msgbuf);
+      DPRINTF("s:\n%s\n", ss);
+      free(msgbuf);
+    }
+#endif
+    goto err;
+  }
+  return s;
+err:
+  glDeleteShader(s);
+  return 0;
+}
+
+static GLuint compile_and_link(const uint8_t *vss, const uint8_t *fss) {
+  GLuint p = 0, vs = 0, fs = 0;
+  p = glCreateProgram();
+  if (!p) goto err;
+  vs = compile_shader(vss, GL_VERTEX_SHADER);
+  if (!vs) goto err;
+  fs = compile_shader(fss, GL_FRAGMENT_SHADER);
+  if (!fs) goto err;
+  glAttachShader(p, vs);
+  glAttachShader(p, fs);
+  glBindAttribLocation(p, VAI_COORD, "coord");
+  glLinkProgram(p);
+  GLint res;
+  glGetProgramiv(p, GL_LINK_STATUS, &res);
+  if (!res) {
+#ifndef NDEBUG
+    glGetProgramiv(p, GL_INFO_LOG_LENGTH, &res);
+    char *msgbuf = malloc(res);
+    if (msgbuf) {
+      glGetProgramInfoLog(p, res, 0, msgbuf);
+      DPRINTF("program link error: \n%s\n", msgbuf);
+      DPRINTF("vs:\n%s\n", vss);
+      DPRINTF("fs:\n%s\n", fss);
+      free(msgbuf);
+    }
+#endif
+    goto err;
+  }
+  glDeleteShader(vs);
+  glDeleteShader(fs);
+  glUseProgram(p);
+  glUniform1i(glGetUniformLocation(p, "tex"), 1);
+  return p;
+err:
+  glDeleteProgram(p);
+  glDeleteShader(vs);
+  glDeleteShader(fs);
+  return 0;
+}
+
+static void pacc_buf_delete(struct pacc_buf *pb) {
+  if (pb) {
+    free(pb->buf);
+    glDeleteBuffers(1, &pb->buf_obj);
+#ifdef PACC_GL_3
+    glDeleteVertexArrays(1, &pb->va_obj);
+#endif
+    free(pb);
+  }
+}
+
+static struct pacc_buf *pacc_gen_buf(
+    struct pacc_ctx *pc, struct pacc_tex *pt, enum pacc_buf_mode mode) {
+  struct pacc_buf *pb = malloc(sizeof(*pb));
+  if (!pb) goto err;
+  *pb = (struct pacc_buf) {
+    .buflen = PACC_BUF_DEF_LEN,
+    .tex = pt,
+    .usage = (mode == pacc_buf_mode_static) ?
+      GL_STATIC_DRAW : GL_STREAM_DRAW,
+  };
+  pb->buf = malloc(sizeof(*pb->buf) * pb->buflen);
+  if (!pb->buf) goto err;
+  glGenBuffers(1, &pb->buf_obj);
+  if (!pb->buf_obj) goto err;
+#ifdef PACC_GL_3
+  glGenVertexArrays(1, &pb->va_obj);
+  if (!pb->va_obj) goto err;
+  glBindVertexArray(pb->va_obj);
+  glEnableVertexAttribArray(VAI_COORD);
+  glBindBuffer(GL_ARRAY_BUFFER, pb->buf_obj);
+  glVertexAttribPointer(VAI_COORD, 4, GL_FLOAT, GL_FALSE, 0, 0);
+#endif
+  return pb;
+err:
+  pacc_buf_delete(pb);
+  return 0;
+}
+
+static bool buf_reserve(struct pacc_buf *pb, int len) {
+  if (pb->len + len > pb->buflen) {
+    int newlen = pb->buflen;
+    while (pb->len + len > newlen) newlen *= 2;
+    GLfloat *newbuf = realloc(pb->buf, newlen * sizeof(pb->buf[0]));
+    if (!newbuf) return false;
+    pb->buflen = newlen;
+    pb->buf = newbuf;
+  }
+  return true;
+}
+
+static void pacc_calc_scale(float *ret, int w, int h, int wdest, int hdest) {
+  ret[0] = ((float)w) / wdest;
+  ret[1] = ((float)h) / hdest;
+}
+
+static void pacc_calc_off_tex(float *ret,
+                                     int tw, int th, int xsrc, int ysrc) {
+  ret[0] = ((float)xsrc) / tw;
+  ret[1] = ((float)ysrc) / th;
+}
+
+static void pacc_calc_off(
+    float *ret, int xdest, int ydest, int w, int h, int wdest, int hdest) {
+  ret[0] = ((float)(xdest * 2 + w - wdest)) / wdest;
+  ret[1] = ((float)(ydest * 2 + h - hdest)) / hdest;
+}
+
+static void pacc_buf_rect_off(
+    const struct pacc_ctx *pc, struct pacc_buf *pb,
+    int x, int y, int w, int h, int xoff, int yoff) {
+  float scale[2];
+  float off[2];
+  float tscale[2];
+  float toff[2];
+  pacc_calc_off(off, x, y, w, h, pc->w, pc->h);
+  pacc_calc_scale(scale, w, h, pc->w, pc->h);
+  pacc_calc_off_tex(toff, pb->tex->w, pb->tex->h, xoff, yoff);
+  pacc_calc_scale(tscale, w, h, pb->tex->w, pb->tex->h);
+  GLfloat coord[16] = {
+    -1.0f * scale[0] + off[0], -1.0f * scale[1] - off[1],
+     0.0f * tscale[0] + toff[0],  1.0f * tscale[1] + toff[1],
+
+    -1.0f * scale[0] + off[0],  1.0f * scale[1] - off[1],
+     0.0f * tscale[0] + toff[0],  0.0f * tscale[1] + toff[1],
+
+     1.0f * scale[0] + off[0], -1.0f * scale[1] - off[1],
+     1.0f * tscale[0] + toff[0],  1.0f * tscale[1] + toff[1],
+
+     1.0f * scale[0] + off[0],  1.0f * scale[1] - off[1],
+     1.0f * tscale[0] + toff[0],  0.0f * tscale[1] + toff[1],
+  };
+  if (!buf_reserve(pb, 24)) return;
+  int indices[6] = {0, 1, 2, 2, 1, 3};
+  for (int i = 0; i < 6; i++) {
+    for (int j = 0; j < 4; j++) {
+      pb->buf[pb->len+i*4+j] = coord[indices[i]*4+j];
+    }
+  }
+  pb->len += 24;
+  pb->changed = true;
+}
+
+static void pacc_buf_vprintf(
+    const struct pacc_ctx *pc, struct pacc_buf *pb,
+    int x, int y, const char *fmt, va_list ap) {
+  uint8_t printbuf[PRINTBUFLEN+1];
+  vsnprintf((char *)printbuf, sizeof(printbuf), fmt, ap);
+  int len = strlen((const char *)printbuf);
+  float scale[2];
+  float off[2];
+  int w = pb->tex->w / 256;
+  int h = pb->tex->h;
+  pacc_calc_scale(scale, w, h, pc->w, pc->h);
+  pacc_calc_off(off, x, y, w, h, pc->w, pc->h);
+  if (!buf_reserve(pb, len*24)) return;
+  GLfloat *coords = pb->buf + pb->len;
+  for (int i = 0; i < len; i++) {
+    coords[24*i+0*4+0]                      = (-1.0f + 2.0f*i) * scale[0] + off[0];
+    coords[24*i+0*4+1]                      = -1.0f * scale[1] - off[1];
+    coords[24*i+1*4+0] = coords[24*i+4*4+0] = (-1.0f + 2.0f*i) * scale[0] + off[0];
+    coords[24*i+1*4+1] = coords[24*i+4*4+1] = 1.0f * scale[1] - off[1];
+    coords[24*i+2*4+0] = coords[24*i+3*4+0] = (1.0f + 2.0f*i) * scale[0] + off[0];
+    coords[24*i+2*4+1] = coords[24*i+3*4+1] = -1.0f * scale[1] - off[1];
+    coords[24*i+5*4+0]                      = (1.0f + 2.0f*i) * scale[0] + off[0];
+    coords[24*i+5*4+1]                      = 1.0f * scale[1] - off[1];
+    coords[24*i+0*4+2]                      = ((float)printbuf[i]) / 256.0f;
+    coords[24*i+0*4+3]                      = 1.0f;
+    coords[24*i+1*4+2] = coords[24*i+4*4+2] = ((float)printbuf[i]) / 256.0f;
+    coords[24*i+1*4+3] = coords[24*i+4*4+3] = 0.0f;
+    coords[24*i+2*4+2] = coords[24*i+3*4+2] = ((float)(printbuf[i]+1)) / 256.0f;
+    coords[24*i+2*4+3] = coords[24*i+3*4+3] = 1.0f;
+    coords[24*i+5*4+2]                      = ((float)(printbuf[i]+1)) / 256.0f;
+    coords[24*i+5*4+3]                      = 0.0f;
+  }
+  pb->len += len * 24;
+  pb->changed = true;
+}
+
+static void pacc_buf_printf(
+    const struct pacc_ctx *pc, struct pacc_buf *pb,
+    int x, int y, const char *fmt, ...) {
+  va_list ap;
+  va_start(ap, fmt);
+  pacc_buf_vprintf(pc, pb, x, y, fmt, ap);
+  va_end(ap);
+}
+
+static void pacc_buf_rect(
+    const struct pacc_ctx *pc, struct pacc_buf *pb,
+    int x, int y, int w, int h) {
+  pacc_buf_rect_off(pc, pb, x, y, w, h, 0, 0);
+}
+
+static void pacc_buf_clear(struct pacc_buf *pb) {
+  pb->len = 0;
+  pb->changed = true;
+}
+
+static void pacc_palette(struct pacc_ctx *pc, const uint8_t *rgb, int colors) {
+  memcpy(pc->pal, rgb, colors*3);
+  pc->pal_changed = true;
+}
+
+static void pacc_color(struct pacc_ctx *pc, uint8_t pal) {
+  pc->color = pal;
+  pc->color_changed = true;
+}
+
+static void pacc_begin_clear(struct pacc_ctx *pc) {
+  if (pc->pal_changed) {
+    glActiveTexture(GL_TEXTURE0);
+    glTexImage2D(
+        GL_TEXTURE_2D,
+        0, GL_RGB,
+        256, 1,
+        0, GL_RGB,
+        GL_UNSIGNED_BYTE, pc->pal);
+    pc->pal_changed = false;
+    glActiveTexture(GL_TEXTURE1);
+  }
+  if (memcmp(pc->clearcolor, pc->pal, 3)) {
+    memcpy(pc->clearcolor, pc->pal, 3);
+    glClearColor(
+        pc->clearcolor[0] / 255.f,
+        pc->clearcolor[1] / 255.f,
+        pc->clearcolor[2] / 255.f,
+        1.0f);
+  }
+  glClear(GL_COLOR_BUFFER_BIT);
+  pc->curr_mode = pacc_mode_count;
+}
+
+static void pacc_draw(struct pacc_ctx *pc, struct pacc_buf *pb, enum pacc_mode mode) {
+  if (!pb->len) return;
+  if (mode >= pacc_mode_count) return;
+  if (pc->curr_mode != mode) {
+    glUseProgram(pc->progs[mode]);
+    pc->curr_mode = mode;
+  }
+  if (mode != pacc_mode_copy && pc->color_changed) {
+    glUniform1f(pc->uni_color, pc->color / 255.f);
+    glUniform1f(pc->uni_color_trans, pc->color / 255.f);
+    pc->color_changed = false;
+  }
+  glBindTexture(GL_TEXTURE_2D, pb->tex->tex_obj);
+  if (pb->tex->changed) {
+    GLint format;
+#if PACC_GL_3
+    format = GL_RED;
+#else
+    format = GL_LUMINANCE;
+#endif
+    glTexImage2D(
+        GL_TEXTURE_2D,
+        0, format,
+        pb->tex->w, pb->tex->h,
+        0, format, GL_UNSIGNED_BYTE,
+        pb->tex->buf);
+    pb->tex->changed = false;
+  }
+  if (pb->changed) {
+    glBindBuffer(GL_ARRAY_BUFFER, pb->buf_obj);
+    glBufferData(
+        GL_ARRAY_BUFFER,
+        pb->len * sizeof(pb->buf[0]), pb->buf,
+        pb->usage);
+    pb->changed = false;
+  }
+#ifdef PACC_GL_3
+  glBindVertexArray(pb->va_obj);
+#else
+  glEnableVertexAttribArray(VAI_COORD);
+  glBindBuffer(GL_ARRAY_BUFFER, pb->buf_obj);
+  glVertexAttribPointer(VAI_COORD, 4, GL_FLOAT, GL_FALSE, 0, 0);
+#endif
+  glDrawArrays(GL_TRIANGLES, 0, pb->len / 4);
+}
+
+static uint8_t *pacc_tex_lock(struct pacc_tex *pt) {
+  return pt->buf;
+}
+
+static void pacc_tex_unlock(struct pacc_tex *pt) {
+  pt->changed = true;
+}
+
+static void pacc_tex_delete(struct pacc_tex *pt) {
+  if (pt) {
+    glDeleteTextures(1, &pt->tex_obj);
+    free(pt->buf);
+    free(pt);
+  }
+}
+
+static struct pacc_tex *pacc_gen_tex(struct pacc_ctx *pc, int w, int h) {
+  struct pacc_tex *pt = malloc(sizeof(*pt));
+  if (!pt) goto err;
+  *pt = (struct pacc_tex) {
+    .w = w,
+    .h = h,
+    .buf = calloc(w*h, 1),
+  };
+  if (!pt->buf) goto err;
+  glGenTextures(1, &pt->tex_obj);
+  glBindTexture(GL_TEXTURE_2D, pt->tex_obj);
+  GLint format;
+#ifdef PACC_GL_3
+  format = GL_RED;
+#else
+  format = GL_LUMINANCE;
+#endif
+  glTexImage2D(
+      GL_TEXTURE_2D,
+      0, format,
+      w, h, 0,
+      format, GL_UNSIGNED_BYTE,
+      pt->buf);
+  glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+  glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+  if (w & (w - 1)) {
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
+  }
+  if (h & (h - 1)) {
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+  }
+  return pt;
+err:
+  pacc_tex_delete(pt);
+  return 0;
+}
+
+static struct pacc_vtable pacc_gl_vtable = {
+  .pacc_delete = pacc_delete,
+  .gen_buf = pacc_gen_buf,
+  .gen_tex = pacc_gen_tex,
+  .buf_delete = pacc_buf_delete,
+  .tex_lock = pacc_tex_lock,
+  .tex_unlock = pacc_tex_unlock,
+  .tex_delete = pacc_tex_delete,
+  .buf_rect = pacc_buf_rect,
+  .buf_rect_off = pacc_buf_rect_off,
+  .buf_vprintf = pacc_buf_vprintf,
+  .buf_printf = pacc_buf_printf,
+  .buf_clear = pacc_buf_clear,
+  .palette = pacc_palette,
+  .color = pacc_color,
+  .begin_clear = pacc_begin_clear,
+  .draw = pacc_draw,
+};
+
+struct pacc_ctx *pacc_init_gl(int w, int h, struct pacc_vtable *vt) {
+  struct pacc_ctx *pc = malloc(sizeof(*pc));
+  if (!pc) goto err;
+  *pc = (struct pacc_ctx) {
+    .w = w,
+    .h = h,
+  };
+  glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
+  glGenTextures(1, &pc->tex_pal);
+  glBindTexture(GL_TEXTURE_2D, pc->tex_pal);
+  glTexImage2D(
+      GL_TEXTURE_2D,
+      0, GL_RGB,
+      256, 1, 0,
+      GL_RGB, GL_UNSIGNED_BYTE,
+      pc->pal);
+  glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+  glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+  pc->progs[pacc_mode_copy] = compile_and_link(blit_vert, copy_frag);
+  if (!pc->progs[pacc_mode_copy]) goto err;
+  pc->progs[pacc_mode_color] = compile_and_link(blit_vert, color_frag);
+  if (!pc->progs[pacc_mode_color]) goto err;
+  pc->progs[pacc_mode_color_trans] = compile_and_link(blit_vert, color_trans_frag);
+  if (!pc->progs[pacc_mode_color_trans]) goto err;
+  pc->uni_color = glGetUniformLocation(pc->progs[pacc_mode_color], "color");
+  pc->uni_color_trans = glGetUniformLocation(pc->progs[pacc_mode_color_trans], "color");
+  glActiveTexture(GL_TEXTURE1);
+  *vt = pacc_gl_vtable;
+  return pc;
+err:
+  pacc_delete(pc);
+  return 0;
+}
+
diff --git a/pacc/pacc-gl.h b/pacc/pacc-gl.h
new file mode 100644
index 0000000..2f5ccff
--- /dev/null
+++ b/pacc/pacc-gl.h
@@ -0,0 +1,8 @@
+#ifndef MYON_PACC_GL_H_INCLUDED
+#define MYON_PACC_GL_H_INCLUDED
+
+#include "pacc.h"
+
+struct pacc_ctx *pacc_init_gl(int w, int h, struct pacc_vtable *vt);
+
+#endif // MYON_PACC_GL_H_INCLUDED
diff --git a/pacc/pacc-win.h b/pacc/pacc-win.h
new file mode 100644
index 0000000..47c09d1
--- /dev/null
+++ b/pacc/pacc-win.h
@@ -0,0 +1,21 @@
+#ifndef MYON_PACC_WIN_H_INCLUDED
+#define MYON_PACC_WIN_H_INCLUDED
+
+#include "pacc.h"
+
+#ifndef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN
+#endif
+#include <windows.h>
+#include <stdbool.h>
+
+typedef void pacc_rendercb(void *ptr);
+
+struct pacc_win_vtable {
+  void (*renderctrl)(struct pacc_ctx *ctx, bool enable);
+};
+
+struct pacc_ctx *pacc_init_d3d9(HWND hwnd, pacc_rendercb *rendercb, void *renderptr, struct pacc_vtable *vt, struct pacc_win_vtable *winvt);
+
+#endif // MYON_PACC_WIN_H_INCLUDED
+
diff --git a/pacc/pacc.h b/pacc/pacc.h
new file mode 100644
index 0000000..b3387ac
--- /dev/null
+++ b/pacc/pacc.h
@@ -0,0 +1,53 @@
+#ifndef MYON_PACC_H_INCLUDED
+#define MYON_PACC_H_INCLUDED
+
+#include <stdint.h>
+#include <stdarg.h>
+
+enum pacc_mode {
+  pacc_mode_copy,
+  pacc_mode_color,
+  pacc_mode_color_trans,
+  pacc_mode_count,
+};
+
+enum pacc_buf_mode {
+  pacc_buf_mode_static,
+  pacc_buf_mode_stream,
+};
+
+struct pacc_ctx;
+struct pacc_tex;
+struct pacc_buf;
+
+struct pacc_vtable {
+  void (*pacc_delete)(struct pacc_ctx *pc);
+  struct pacc_buf *(*gen_buf)(
+      struct pacc_ctx *pc, struct pacc_tex *pt, enum pacc_buf_mode mode);
+  struct pacc_tex *(*gen_tex)(struct pacc_ctx *pc, int w, int h);
+  void (*buf_delete)(struct pacc_buf *buf);
+  uint8_t *(*tex_lock)(struct pacc_tex *tex);
+  void (*tex_unlock)(struct pacc_tex *tex);
+  void (*tex_delete)(struct pacc_tex *tex);
+  void (*buf_rect)(
+      const struct pacc_ctx *ctx, struct pacc_buf *buf,
+      int x, int y, int w, int h);
+  void (*buf_rect_off)(
+      const struct pacc_ctx *ctx, struct pacc_buf *buf,
+      int x, int y, int w, int h, int xoff, int yoff);
+  void (*buf_vprintf)(
+      const struct pacc_ctx *ctx, struct pacc_buf *buf,
+      int x, int y, const char *fmt, va_list ap);
+  void (*buf_printf)(
+      const struct pacc_ctx *ctx, struct pacc_buf *buf,
+      int x, int y, const char *fmt, ...);
+  void (*buf_clear)(struct pacc_buf *buf);
+  void (*palette)(struct pacc_ctx *ctx, const uint8_t *rgb, int colors);
+  void (*color)(struct pacc_ctx *ctx, uint8_t pal);
+  void (*begin_clear)(struct pacc_ctx *ctx);
+  void (*draw)(struct pacc_ctx *ctx, struct pacc_buf *buf, enum pacc_mode mode);
+};
+
+struct pacc_ctx *pacc_init_gl(int w, int h, struct pacc_vtable *vt);
+
+#endif // MYON_PACC_H_INCLUDED