nve0: initial exa/xv acceleration for kepler chipsets

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

9 files changed, 582 insertions, 39 deletions

diff --git a/src/nouveau_exa.c b/src/nouveau_exa.c
index 6ad1790..9aaa3f2 100644
--- a/src/nouveau_exa.c
+++ b/src/nouveau_exa.c
@@ -47,6 +47,11 @@ NVAccelM2MF(NVPtr pNv, int w, int h, int cpp, uint32_t srcoff, uint32_t dstoff,
 	    struct nouveau_bo *src, int sd, int sp, int sh, int sx, int sy,
 	    struct nouveau_bo *dst, int dd, int dp, int dh, int dx, int dy)
 {
+	if (pNv->Architecture >= NV_ARCH_E0)
+		return NVE0EXARectCopy(pNv, w, h, cpp,
+				       src, srcoff, sd, sp, sh, sx, sy,
+				       dst, dstoff, dd, dp, dh, dx, dy);
+	else
 	if (pNv->Architecture >= NV_ARCH_C0)
 		return NVC0EXARectM2MF(pNv, w, h, cpp,
 				       src, srcoff, sd, sp, sh, sx, sy,
@@ -413,6 +418,7 @@ nouveau_exa_init(ScreenPtr pScreen)
 		exa->DoneComposite    = NV50EXADoneComposite;
 		break;
 	case NV_ARCH_C0:
+	case NV_ARCH_E0:
 		exa->CheckComposite   = NVC0EXACheckComposite;
 		exa->PrepareComposite = NVC0EXAPrepareComposite;
 		exa->Composite        = NVC0EXAComposite;
diff --git a/src/nouveau_xv.c b/src/nouveau_xv.c
index a43225c..9ada296 100644
--- a/src/nouveau_xv.c
+++ b/src/nouveau_xv.c
@@ -256,7 +256,7 @@ nouveau_xv_bo_realloc(ScrnInfoPtr pScrn, unsigned flags, unsigned size,
 		if (pNv->Architecture == NV_ARCH_50)
 			config.nv50.memtype = 0x70;
 		else
-		if (pNv->Architecture == NV_ARCH_C0)
+		if (pNv->Architecture >= NV_ARCH_C0)
 			config.nvc0.memtype = 0xfe;
 	}
 	flags |= NOUVEAU_BO_MAP;
diff --git a/src/nv_accel_common.c b/src/nv_accel_common.c
index 09272b9..32cae49 100644
--- a/src/nv_accel_common.c
+++ b/src/nv_accel_common.c
@@ -623,11 +623,15 @@ NVAccelCommonInit(ScrnInfoPtr pScrn)
 	if (pNv->Architecture < NV_ARCH_C0)
 		INIT_CONTEXT_OBJECT(M2MF_NV50);
 	else
+	if (pNv->Architecture < NV_ARCH_E0)
 		INIT_CONTEXT_OBJECT(M2MF_NVC0);
+	else
+		INIT_CONTEXT_OBJECT(P2MF_NVE0);
 
 	/* 3D init */
 	switch (pNv->Architecture) {
 	case NV_ARCH_C0:
+	case NV_ARCH_E0:
 		INIT_CONTEXT_OBJECT(3D_NVC0);
 		break;
 	case NV_ARCH_50:
diff --git a/src/nv_proto.h b/src/nv_proto.h
index 88cd870..f3aa3be 100644
--- a/src/nv_proto.h
+++ b/src/nv_proto.h
@@ -148,6 +148,7 @@ Bool NVAccelInitNV50TCL(ScrnInfoPtr pScrn);
 
 /* in nvc0_accel.c */
 Bool NVAccelInitM2MF_NVC0(ScrnInfoPtr pScrn);
+Bool NVAccelInitP2MF_NVE0(ScrnInfoPtr pScrn);
 Bool NVAccelInit2D_NVC0(ScrnInfoPtr pScrn);
 Bool NVAccelInit3D_NVC0(ScrnInfoPtr pScrn);
 
@@ -191,6 +192,9 @@ Bool NVC0EXAUploadSIFC(const char *src, int src_pitch,
 Bool NVC0EXARectM2MF(NVPtr pNv, int, int, int,
 		     struct nouveau_bo *, uint32_t, int, int, int, int, int,
 		     struct nouveau_bo *, uint32_t, int, int, int, int, int);
+Bool NVE0EXARectCopy(NVPtr pNv, int, int, int,
+		     struct nouveau_bo *, uint32_t, int, int, int, int, int,
+		     struct nouveau_bo *, uint32_t, int, int, int, int, int);
 
 /* nv50_xv.c */
 int nv50_xv_image_put(ScrnInfoPtr, struct nouveau_bo *, int, int, int, int,
diff --git a/src/nvc0_accel.c b/src/nvc0_accel.c
index 2fc7ef9..ad34554 100644
--- a/src/nvc0_accel.c
+++ b/src/nvc0_accel.c
@@ -23,6 +23,7 @@
 #include "nv_include.h"
 #include "nvc0_accel.h"
 #include "nvc0_shader.h"
+#include "nve0_shader.h"
 
 Bool
 NVAccelInitM2MF_NVC0(ScrnInfoPtr pScrn)
@@ -52,6 +53,30 @@ NVAccelInitM2MF_NVC0(ScrnInfoPtr pScrn)
 }
 
 Bool
+NVAccelInitP2MF_NVE0(ScrnInfoPtr pScrn)
+{
+	NVPtr pNv = NVPTR(pScrn);
+	struct nouveau_pushbuf *push = pNv->pushbuf;
+	int ret;
+
+	ret = nouveau_bo_new(pNv->dev, NOUVEAU_BO_VRAM, 128 * 1024, 0x20000,
+			     NULL, &pNv->tesla_scratch);
+	if (ret)
+		return FALSE;
+
+	ret = nouveau_object_new(pNv->channel, 0x0000a040, 0xa040,
+				 NULL, 0, &pNv->NvMemFormat);
+	if (ret)
+		return FALSE;
+
+	BEGIN_NVC0(push, NV01_SUBC(P2MF, OBJECT), 1);
+	PUSH_DATA (push, pNv->NvMemFormat->handle);
+	BEGIN_NVC0(push, NV01_SUBC(COPY, OBJECT), 1);
+	PUSH_DATA (push, 0x0000a0b5);
+	return TRUE;
+}
+
+Bool
 NVAccelInit2D_NVC0(ScrnInfoPtr pScrn)
 {
 	NVPtr pNv = NVPTR(pScrn);
@@ -104,9 +129,15 @@ NVAccelInit3D_NVC0(ScrnInfoPtr pScrn)
 	NVPtr pNv = NVPTR(pScrn);
 	struct nouveau_pushbuf *push = pNv->pushbuf;
 	struct nouveau_bo *bo = pNv->tesla_scratch;
+	uint32_t class;
 	int ret;
 
-	ret = nouveau_object_new(pNv->channel, 0x00009097, 0x9097,
+	if (pNv->Architecture < NV_ARCH_E0)
+		class = 0x9097;
+	else
+		class = 0xa097;
+
+	ret = nouveau_object_new(pNv->channel, class, class,
 				 NULL, 0, &pNv->Nv3D);
 	if (ret)
 		return FALSE;
@@ -155,11 +186,25 @@ NVAccelInit3D_NVC0(ScrnInfoPtr pScrn)
 	PUSH_DATA (push, 0);
 	BEGIN_NVC0(push, NVC0_3D(LINKED_TSC), 1);
 	PUSH_DATA (push, 1);
-	BEGIN_NVC0(push, NVC0_3D(TEX_LIMITS(4)), 1);
-	PUSH_DATA (push, 0x54);
-	BEGIN_NIC0(push, NVC0_3D(BIND_TIC(4)), 2);
-	PUSH_DATA (push, (0 << 9) | (0 << 1) | NVC0_3D_BIND_TIC_ACTIVE);
-	PUSH_DATA (push, (1 << 9) | (1 << 1) | NVC0_3D_BIND_TIC_ACTIVE);
+	if (pNv->Architecture < NV_ARCH_E0) {
+		BEGIN_NVC0(push, NVC0_3D(TEX_LIMITS(4)), 1);
+		PUSH_DATA (push, 0x54);
+		BEGIN_NIC0(push, NVC0_3D(BIND_TIC(4)), 2);
+		PUSH_DATA (push, (0 << 9) | (0 << 1) | NVC0_3D_BIND_TIC_ACTIVE);
+		PUSH_DATA (push, (1 << 9) | (1 << 1) | NVC0_3D_BIND_TIC_ACTIVE);
+	} else {
+		BEGIN_NVC0(push, NVC0_3D(CB_SIZE), 6);
+		PUSH_DATA (push, 256);
+		PUSH_DATA (push, (bo->offset + TB_OFFSET) >> 32);
+		PUSH_DATA (push, (bo->offset + TB_OFFSET));
+		PUSH_DATA (push, 0);
+		PUSH_DATA (push, 0x00000000);
+		PUSH_DATA (push, 0x00000001);
+		BEGIN_NVC0(push, NVC0_3D(CB_BIND(4)), 1);
+		PUSH_DATA (push, 0x11);
+		BEGIN_NVC0(push, SUBC_3D(0x2608), 1);
+		PUSH_DATA (push, 1);
+	}
 
 	BEGIN_NVC0(push, NVC0_3D(VERTEX_QUARANTINE_ADDRESS_HIGH), 3);
 	PUSH_DATA (push, (bo->offset + MISC_OFFSET) >> 32);
@@ -169,18 +214,28 @@ NVAccelInit3D_NVC0(ScrnInfoPtr pScrn)
 	BEGIN_NVC0(push, NVC0_3D(CODE_ADDRESS_HIGH), 2);
 	PUSH_DATA (push, (bo->offset + CODE_OFFSET) >> 32);
 	PUSH_DATA (push, (bo->offset + CODE_OFFSET));
+	if (pNv->Architecture < NV_ARCH_E0) {
+		NVC0PushProgram(pNv, PVP_PASS, NVC0VP_Passthrough);
+		NVC0PushProgram(pNv, PFP_S, NVC0FP_Source);
+		NVC0PushProgram(pNv, PFP_C, NVC0FP_Composite);
+		NVC0PushProgram(pNv, PFP_CCA, NVC0FP_CAComposite);
+		NVC0PushProgram(pNv, PFP_CCASA, NVC0FP_CACompositeSrcAlpha);
+		NVC0PushProgram(pNv, PFP_S_A8, NVC0FP_Source_A8);
+		NVC0PushProgram(pNv, PFP_C_A8, NVC0FP_Composite_A8);
+		NVC0PushProgram(pNv, PFP_NV12, NVC0FP_NV12);
 
-	NVC0PushProgram(pNv, PVP_PASS, NVC0VP_Passthrough);
-	NVC0PushProgram(pNv, PFP_S, NVC0FP_Source);
-	NVC0PushProgram(pNv, PFP_C, NVC0FP_Composite);
-	NVC0PushProgram(pNv, PFP_CCA, NVC0FP_CAComposite);
-	NVC0PushProgram(pNv, PFP_CCASA, NVC0FP_CACompositeSrcAlpha);
-	NVC0PushProgram(pNv, PFP_S_A8, NVC0FP_Source_A8);
-	NVC0PushProgram(pNv, PFP_C_A8, NVC0FP_Composite_A8);
-	NVC0PushProgram(pNv, PFP_NV12, NVC0FP_NV12);
-
-	BEGIN_NVC0(push, NVC0_3D(MEM_BARRIER), 1);
-	PUSH_DATA (push, 0x1111);
+		BEGIN_NVC0(push, NVC0_3D(MEM_BARRIER), 1);
+		PUSH_DATA (push, 0x1111);
+	} else {
+		NVC0PushProgram(pNv, PVP_PASS, NVE0VP_Passthrough);
+		NVC0PushProgram(pNv, PFP_S, NVE0FP_Source);
+		NVC0PushProgram(pNv, PFP_C, NVE0FP_Composite);
+		NVC0PushProgram(pNv, PFP_CCA, NVE0FP_CAComposite);
+		NVC0PushProgram(pNv, PFP_CCASA, NVE0FP_CACompositeSrcAlpha);
+		NVC0PushProgram(pNv, PFP_S_A8, NVE0FP_Source_A8);
+		NVC0PushProgram(pNv, PFP_C_A8, NVE0FP_Composite_A8);
+		NVC0PushProgram(pNv, PFP_NV12, NVE0FP_NV12);
+	}
 
 	BEGIN_NVC0(push, NVC0_3D(SP_SELECT(1)), 4);
 	PUSH_DATA (push, NVC0_3D_SP_SELECT_PROGRAM_VP_B |
diff --git a/src/nvc0_accel.h b/src/nvc0_accel.h
index 9a7475a..3b838ea 100644
--- a/src/nvc0_accel.h
+++ b/src/nvc0_accel.h
@@ -13,10 +13,12 @@
 #define SUBC_3D(mthd)    0, (mthd)
 #define NVC0_3D(mthd)    SUBC_3D(NVC0_3D_##mthd)
 #define SUBC_M2MF(mthd)  2, (mthd)
+#define SUBC_P2MF(mthd)  2, (mthd)
 #define NVC0_M2MF(mthd)  SUBC_M2MF(NVC0_M2MF_##mthd)
 #define SUBC_2D(mthd)    3, (mthd)
 #define NV50_2D(mthd)    SUBC_2D(NV50_2D_##mthd)
 #define NVC0_2D(mthd)    SUBC_2D(NVC0_2D_##mthd)
+#define SUBC_COPY(mthd)  4, (mthd)
 #define SUBC_NVSW(mthd)  5, (mthd)
 
 /* scratch buffer offsets */
@@ -26,21 +28,23 @@
 #define NTFY_OFFSET 0x08000
 #define MISC_OFFSET 0x10000
 
-/* fragment programs */
-#define PFP_S     0x0000 /* (src) */
-#define PFP_C     0x0100 /* (src IN mask) */
-#define PFP_CCA   0x0200 /* (src IN mask) component-alpha */
-#define PFP_CCASA 0x0300 /* (src IN mask) component-alpha src-alpha */
-#define PFP_S_A8  0x0400 /* (src) a8 rt */
-#define PFP_C_A8  0x0500 /* (src IN mask) a8 rt - same for CA and CA_SA */
-#define PFP_NV12  0x0600 /* NV12 YUV->RGB */
-
-/* vertex programs */
-#define PVP_PASS  0x0700 /* vertex pass-through shader */
+/* vertex/fragment programs */
+#define SPO       ((pNv->Architecture < NV_ARCH_E0) ? 0x0000 : 0x0030)
+#define PVP_PASS  (0x0000 + SPO) /* vertex pass-through shader */
+#define PFP_S     (0x0200 + SPO) /* (src) */
+#define PFP_C     (0x0400 + SPO) /* (src IN mask) */
+#define PFP_CCA   (0x0600 + SPO) /* (src IN mask) component-alpha */
+#define PFP_CCASA (0x0800 + SPO) /* (src IN mask) component-alpha src-alpha */
+#define PFP_S_A8  (0x0a00 + SPO) /* (src) a8 rt */
+#define PFP_C_A8  (0x0c00 + SPO) /* (src IN mask) a8 rt - same for CCA/CCASA */
+#define PFP_NV12  (0x0e00 + SPO) /* NV12 YUV->RGB */
 
 /* shader constants */
 #define CB_OFFSET 0x1000
 
+/* texture bindings (kepler) */
+#define TB_OFFSET 0x1800
+
 #define VTX_ATTR(a, c, t, s)				\
 	((NVC0_3D_VTX_ATTR_DEFINE_TYPE_##t) |		\
 	 ((a) << NVC0_3D_VTX_ATTR_DEFINE_ATTR__SHIFT) |	\
@@ -98,15 +102,25 @@ static __inline__ void
 PUSH_DATAu(struct nouveau_pushbuf *push, struct nouveau_bo *bo,
 	   unsigned delta, unsigned dwords)
 {
-	BEGIN_NVC0(push, NVC0_M2MF(OFFSET_OUT_HIGH), 2);
-	PUSH_DATA (push, (bo->offset + delta) >> 32);
-	PUSH_DATA (push, (bo->offset + delta));
-	BEGIN_NVC0(push, NVC0_M2MF(LINE_LENGTH_IN), 2);
-	PUSH_DATA (push, dwords * 4);
-	PUSH_DATA (push, 1);
-	BEGIN_NVC0(push, NVC0_M2MF(EXEC), 1);
-	PUSH_DATA (push, 0x100111);
-	BEGIN_NIC0(push, NVC0_M2MF(DATA), dwords);
+	if (push->client->device->chipset < 0xe0) {
+		BEGIN_NVC0(push, NVC0_M2MF(OFFSET_OUT_HIGH), 2);
+		PUSH_DATA (push, (bo->offset + delta) >> 32);
+		PUSH_DATA (push, (bo->offset + delta));
+		BEGIN_NVC0(push, NVC0_M2MF(LINE_LENGTH_IN), 2);
+		PUSH_DATA (push, dwords * 4);
+		PUSH_DATA (push, 1);
+		BEGIN_NVC0(push, NVC0_M2MF(EXEC), 1);
+		PUSH_DATA (push, 0x100111);
+		BEGIN_NIC0(push, NVC0_M2MF(DATA), dwords);
+	} else {
+		BEGIN_NVC0(push, SUBC_P2MF(0x0180), 4);
+		PUSH_DATA (push, dwords * 4);
+		PUSH_DATA (push, 1);
+		PUSH_DATA (push, (bo->offset + delta) >> 32);
+		PUSH_DATA (push, (bo->offset + delta));
+		BEGIN_1IC0(push, SUBC_P2MF(0x01b0), 1 + dwords);
+		PUSH_DATA (push, 0x001001);
+	}
 }
 
 #endif
diff --git a/src/nvc0_exa.c b/src/nvc0_exa.c
index 4ba862d..8b6d51e 100644
--- a/src/nvc0_exa.c
+++ b/src/nvc0_exa.c
@@ -1006,3 +1006,60 @@ NVC0EXARectM2MF(NVPtr pNv, int w, int h, int cpp,
 
 	return TRUE;
 }
+
+Bool
+NVE0EXARectCopy(NVPtr pNv, int w, int h, int cpp,
+		struct nouveau_bo *src, uint32_t src_off, int src_dom,
+		int src_pitch, int src_h, int src_x, int src_y,
+		struct nouveau_bo *dst, uint32_t dst_off, int dst_dom,
+		int dst_pitch, int dst_h, int dst_x, int dst_y)
+{
+	struct nouveau_pushbuf *push = pNv->pushbuf;
+	struct nouveau_pushbuf_refn refs[] = {
+		{ src, src_dom | NOUVEAU_BO_RD },
+		{ dst, dst_dom | NOUVEAU_BO_WR },
+	};
+	unsigned exec;
+
+	if (nouveau_pushbuf_space(push, 64, 0, 0) ||
+	    nouveau_pushbuf_refn (push, refs, 2))
+		return FALSE;
+
+	exec = 0x00000206;
+	if (!src->config.nvc0.memtype) {
+		src_off += src_y * src_pitch + src_x * cpp;
+		exec |= 0x00000080;
+	}
+	if (!dst->config.nvc0.memtype) {
+		dst_off += dst_y * dst_pitch + dst_x * cpp;
+		exec |= 0x00000100;
+	}
+
+	BEGIN_NVC0(push, SUBC_COPY(0x0728), 6);
+	PUSH_DATA (push, 0x00001000 | src->config.nvc0.tile_mode);
+	PUSH_DATA (push, src_pitch);
+	PUSH_DATA (push, src_h);
+	PUSH_DATA (push, 1);
+	PUSH_DATA (push, 0);
+	PUSH_DATA (push, (src_y << 16) | src_x * cpp);
+	BEGIN_NVC0(push, SUBC_COPY(0x070c), 6);
+	PUSH_DATA (push, 0x000001000 | dst->config.nvc0.tile_mode);
+	PUSH_DATA (push, dst_pitch);
+	PUSH_DATA (push, dst_h);
+	PUSH_DATA (push, 1);
+	PUSH_DATA (push, 0);
+	PUSH_DATA (push, (dst_y << 16) | dst_x * cpp);
+	BEGIN_NVC0(push, SUBC_COPY(0x0400), 8);
+	PUSH_DATA (push, (src->offset + src_off) >> 32);
+	PUSH_DATA (push, (src->offset + src_off));
+	PUSH_DATA (push, (dst->offset + dst_off) >> 32);
+	PUSH_DATA (push, (dst->offset + dst_off));
+	PUSH_DATA (push, src_pitch);
+	PUSH_DATA (push, dst_pitch);
+	PUSH_DATA (push, w * cpp);
+	PUSH_DATA (push, h);
+	BEGIN_NVC0(push, SUBC_COPY(0x0300), 1);
+	PUSH_DATA (push, exec);
+
+	return TRUE;
+}
diff --git a/src/nvc0_shader.h b/src/nvc0_shader.h
index f97b74b..2084a6d 100644
--- a/src/nvc0_shader.h
+++ b/src/nvc0_shader.h
@@ -5,7 +5,6 @@
 	const unsigned size = sizeof(code) / sizeof(code[0]);                  \
 	PUSH_DATAu((pNv)->pushbuf, (pNv)->tesla_scratch, (addr), size);        \
 	PUSH_DATAp((pNv)->pushbuf, (code), size);                              \
-	ErrorF("20 + %d * 2\n", (size - 20) / 2); \
 } while(0)
 
 static uint32_t
diff --git a/src/nve0_shader.h b/src/nve0_shader.h
new file mode 100644
index 0000000..8f36279
--- /dev/null
+++ b/src/nve0_shader.h
@@ -0,0 +1,404 @@
+#ifndef __NVE0_SHADER_H__
+#define __NVE0_SHADER_H__
+
+static uint32_t
+NVE0VP_Passthrough[] = {
+	0x00020461,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x000ff000,
+	0x00000000, /* VP_ATTR_EN[0x000] */
+	0x0001033f, /* VP_ATTR_EN[0x080] */
+	0x00000000, /* VP_ATTR_EN[0x100] */
+	0x00000000,
+	0x00000000, /* VP_ATTR_EN[0x200] */
+	0x00000000,
+	0x00000000, /* VP_ATTR_EN[0x300] */
+	0x00000000,
+	0x0033f000, /* VP_EXPORT_EN[0x040] */
+	0x00000000, /* VP_EXPORT_EN[0x0c0] */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000, /* VP_EXPORT_EN[0x2c0] */
+	0x00000000,
+	0xfff01c66,
+	0x06000080, /* vfetch { $r0,1,2,3 } b128 a[0x80] */
+	0xfff11c26,
+	0x06000090, /* vfetch { $r4,5 } b64 a[0x90] */
+	0xfff19c26,
+	0x060000a0, /* vfetch { $r6,7 } b64 a[0xa0] */
+	0x03f01c66,
+	0x0a7e0070, /* export v[0x70] { $r0 $r1 $r2 $r3 } */
+	0x13f01c26,
+	0x0a7e0080, /* export v[0x80] { $r4 $r5 } */
+	0x1bf01c26,
+	0x0a7e0090, /* export v[0x90] { $r6 $r7 } */
+	0x00001de7,
+	0x80000000, /* exit */
+};
+
+static uint32_t
+NVE0FP_Source[] = {
+	0x00021462,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x80000000,
+	0x0000000a,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x0000000f,
+	0x00000000,
+	0xfff01c00,
+	0xc07e007c, /* linterp f32 $r0 v[$r63+0x7c] */
+	0x10001c00,
+	0xc8000000, /* rcp f32 $r0 $r0 */
+	0x03f05c40,
+	0xc07e0084, /* pinterp f32 $r1 $r0 v[$r63+0x84] */
+	0x03f01c40,
+	0xc07e0080, /* pinterp f32 $r0 $r0 v[$r63+0x80] */
+	0xfc001e86,
+	0x8013c000, /* tex { $r0,1,2,3 } $t0 { $r0,1 } */
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x00001de7,
+	0x80000000, /* exit */
+};
+
+static uint32_t
+NVE0FP_Composite[] = {
+	0x00021462,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x80000000,
+	0x00000a0a,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x0000000f,
+	0x00000000,
+	0xfff01c00,
+	0xc07e007c, /* linterp f32 $r0 v[$r63+0x7c] */
+	0x10001c00,
+	0xc8000000, /* rcp f32 $r0 $r0 */
+	0x03f0dc40,
+	0xc07e0094, /* pinterp f32 $r3 $r0 v[$r63+0x94] */
+	0x03f09c40,
+	0xc07e0090, /* pinterp f32 $r2 $r0 v[$r63+0x90] */
+	0xfc211e86,
+	0x80120001, /* tex { _,_,_,$r4 } $t1 { $r2,3 } */
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x03f05c40,
+	0xc07e0084, /* pinterp f32 $r1 $r0 v[$r63+0x84] */
+	0x03f01c40,
+	0xc07e0080, /* pinterp f32 $r0 $r0 v[$r63+0x80] */
+	0xfc001e86,
+	0x8013c000, /* tex { $r0,1,2,3 } $t0 { $r0,1 } */
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x1030dc40,
+	0x58000000, /* mul ftz rn f32 $r3 $r3 $r4 */
+	0x10209c40,
+	0x58000000, /* mul ftz rn f32 $r2 $r2 $r4 */
+	0x10105c40,
+	0x58000000, /* mul ftz rn f32 $r1 $r1 $r4 */
+	0x10001c40,
+	0x58000000, /* mul ftz rn f32 $r0 $r0 $r4 */
+	0x00001de7,
+	0x80000000, /* exit */
+};
+
+static uint32_t
+NVE0FP_CAComposite[] = {
+	0x00021462, /* 0x0000c000 = USES_KIL, MULTI_COLORS */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x80000000, /* FRAG_COORD_UMASK = 0x8 */
+	0x00000a0a, /* FP_INTERP[0x080], 0022 0022 */
+	0x00000000, /* FP_INTERP[0x0c0], 0 = OFF */
+	0x00000000, /* FP_INTERP[0x100], 1 = FLAT */
+	0x00000000, /* FP_INTERP[0x140], 2 = PERSPECTIVE */
+	0x00000000, /* FP_INTERP[0x180], 3 = LINEAR */
+	0x00000000, /* FP_INTERP[0x1c0] */
+	0x00000000, /* FP_INTERP[0x200] */
+	0x00000000, /* FP_INTERP[0x240] */
+	0x00000000, /* FP_INTERP[0x280] */
+	0x00000000, /* FP_INTERP[0x2c0] */
+	0x00000000, /* FP_INTERP[0x300] */
+	0x00000000,
+	0x0000000f, /* FP_RESULT_MASK (0x8000 Face ?) */
+	0x00000000, /* 0x2 = FragDepth, 0x1 = SampleMask */
+	0xfff01c00,
+	0xc07e007c, /* linterp f32 $r0 v[$r63+0x7c] */
+	0x10001c00,
+	0xc8000000, /* rcp f32 $r0 $r0 */
+	0x03f0dc40,
+	0xc07e0094, /* pinterp f32 $r3 $r0 v[$r63+0x94] */
+	0x03f09c40,
+	0xc07e0090, /* pinterp f32 $r2 $r0 v[$r63+0x90] */
+	0xfc211e86,
+	0x8013c001, /* tex { $r4,5,6,7 } $t1 { $r2,3 } */
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x03f05c40,
+	0xc07e0084, /* pinterp f32 $r1 $r0 v[$r63+0x84] */
+	0x03f01c40,
+	0xc07e0080, /* pinterp f32 $r0 $r0 v[$r63+0x80] */
+	0xfc001e86,
+	0x8013c000, /* tex { $r0,1,2,3 } $t0 { $r0,1 } */
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x1c30dc40,
+	0x58000000, /* mul ftz rn f32 $r3 $r3 $r7 */
+	0x18209c40,
+	0x58000000, /* mul ftz rn f32 $r2 $r2 $r6 */
+	0x14105c40,
+	0x58000000, /* mul ftz rn f32 $r1 $r1 $r5 */
+	0x10001c40,
+	0x58000000, /* mul ftz rn f32 $r0 $r0 $r4 */
+	0x00001de7,
+	0x80000000, /* exit */
+};
+
+static uint32_t
+NVE0FP_CACompositeSrcAlpha[] = {
+	0x00021462,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x80000000,
+	0x00000a0a,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x0000000f,
+	0x00000000,
+	0xfff01c00,
+	0xc07e007c, /* linterp f32 $r0 v[$r63+0x7c] */
+	0x10001c00,
+	0xc8000000, /* rcp f32 $r0 $r0 */
+	0x03f0dc40,
+	0xc07e0084, /* pinterp f32 $r3 $r0 v[$r63+0x84] */
+	0x03f09c40,
+	0xc07e0080, /* pinterp f32 $r2 $r0 v[$r63+0x80] */
+	0xfc211e86,
+	0x80120000, /* tex { _,_,_,$r4 } $t0 { $r2,3 } */
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x03f05c40,
+	0xc07e0094, /* pinterp f32 $r1 $r0 v[$r63+0x94] */
+	0x03f01c40,
+	0xc07e0090, /* pinterp f32 $r0 $r0 v[$r63+0x90] */
+	0xfc001e86,
+	0x8013c001, /* tex { $r0,1,2,3 } $t1 { $r0,1 } */
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x1030dc40,
+	0x58000000, /* mul ftz rn f32 $r3 $r3 $r4 */
+	0x10209c40,
+	0x58000000, /* mul ftz rn f32 $r2 $r2 $r4 */
+	0x10105c40,
+	0x58000000, /* mul ftz rn f32 $r1 $r1 $r4 */
+	0x10001c40,
+	0x58000000, /* mul ftz rn f32 $r0 $r0 $r4 */
+	0x00001de7,
+	0x80000000, /* exit */
+};
+
+static uint32_t
+NVE0FP_Source_A8[] = {
+	0x00021462,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x80000000,
+	0x0000000a,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x0000000f,
+	0x00000000,
+	0xfff01c00,
+	0xc07e007c, /* linterp f32 $r0 v[$r63+0x7c] */
+	0x10001c00,
+	0xc8000000, /* rcp f32 $r0 $r0 */
+	0x03f05c40,
+	0xc07e0084, /* pinterp f32 $r1 $r0 v[$r63+0x84] */
+	0x03f01c40,
+	0xc07e0080, /* pinterp f32 $r0 $r0 v[$r63+0x80] */
+	0xfc001e86,
+	0x80120000, /* tex { _ _ _ $r0 } $t0 { $r0 $r1 } */
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x0000dde4,
+	0x28000000, /* mov b32 $r3 $r0 */
+	0x00009de4,
+	0x28000000, /* mov b32 $r2 $r0 */
+	0x00005de4,
+	0x28000000, /* mov b32 $r1 $r0 */
+	0x00001de7,
+	0x80000000, /* exit */
+};
+
+static uint32_t
+NVE0FP_Composite_A8[] = {
+	0x00021462,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x80000000,
+	0x00000a0a,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x0000000f,
+	0x00000000,
+	0xfff01c00,
+	0xc07e007c, /* linterp f32 $r0 v[$r63+0x7c] */
+	0x10001c00,
+	0xc8000000, /* rcp f32 $r0 $r0 */
+	0x03f0dc40,
+	0xc07e0094, /* pinterp f32 $r3 $r0 v[$r63+0x94] */
+	0x03f09c40,
+	0xc07e0090, /* pinterp f32 $r2 $r0 v[$r63+0x90] */
+	0xfc205e86,
+	0x80120001, /* tex { _ _ _ $r1 } $t1 { $r2 $r3 } */
+	0x03f0dc40,
+	0xc07e0084, /* pinterp f32 $r3 $r0 v[$r63+0x84] */
+	0x03f09c40,
+	0xc07e0080, /* pinterp f32 $r2 $r0 v[$r63+0x80] */
+	0xfc201e86,
+	0x80120000, /* tex { _ _ _ $r0 } $t0 { $r2 $r3 } */
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x0400dc40,
+	0x58000000, /* mul ftz rn f32 $r3 $r0 $r1 */
+	0x0c009de4,
+	0x28000000, /* mov b32 $r2 $r3 */
+	0x0c005de4,
+	0x28000000, /* mov b32 $r1 $r3 */
+	0x0c001de4,
+	0x28000000, /* mov b32 $r0 $r3 */
+	0x00001de7,
+	0x80000000, /* exit */
+};
+
+static uint32_t
+NVE0FP_NV12[] = {
+	0x00021462,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x80000000,
+	0x00000a0a,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x0000000f,
+	0x00000000,
+	0xfff09c00,
+	0xc07e007c,
+	0x10209c00,
+	0xc8000000,
+	0x0bf01c40,
+	0xc07e0080,
+	0x0bf05c40,
+	0xc07e0084,
+	0xfc001e86,
+	0x80120000,
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x00015c40,
+	0x58004000,
+	0x1050dc20,
+	0x50004000,
+	0x20511c20,
+	0x50004000,
+	0x30515c20,
+	0x50004000,
+	0x0bf01c40,
+	0xc07e0090,
+	0x0bf05c40,
+	0xc07e0094,
+	0xfc001e86,
+	0x80130001,
+	0x00001de6,
+	0xf0000000, /* texbar */
+	0x4000dc40,
+	0x30064000,
+	0x50011c40,
+	0x30084000,
+	0x60015c40,
+	0x300a4000,
+	0x70101c40,
+	0x30064000,
+	0x90109c40,
+	0x300a4000,
+	0x80105c40,
+	0x30084000,
+	0x00001de7,
+	0x80000000,
+};
+
+#endif