summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/nv30/nv30_vbo.c
diff options
context:
space:
mode:
authorLuca Barbieri <[email protected]>2010-02-20 18:34:00 +0100
committerYounes Manton <[email protected]>2010-03-15 00:03:01 -0400
commitd673c92810636dcc6de33d3618d494ce9f5717c1 (patch)
tree489cfafd8dc702b1915fef3805688ce89e296b94 /src/gallium/drivers/nv30/nv30_vbo.c
parent1771d8f8f4256773de1883a033081f9cc4cddf00 (diff)
nouveau: s/rankine/eng3d/g; s/curie/eng3d/g
Result of running: perl -i -p -e 's/rankine/eng3d/g; s/curie/eng3d/g;' nv[34]0/*.[ch] This will allow to more easily unify nv30 and nv40.
Diffstat (limited to 'src/gallium/drivers/nv30/nv30_vbo.c')
-rw-r--r--src/gallium/drivers/nv30/nv30_vbo.c68
1 files changed, 34 insertions, 34 deletions
diff --git a/src/gallium/drivers/nv30/nv30_vbo.c b/src/gallium/drivers/nv30/nv30_vbo.c
index f3856bb5a5e..c7f119e90a0 100644
--- a/src/gallium/drivers/nv30/nv30_vbo.c
+++ b/src/gallium/drivers/nv30/nv30_vbo.c
@@ -111,7 +111,7 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
struct pipe_vertex_buffer *vb)
{
struct pipe_screen *pscreen = nv30->pipe.screen;
- struct nouveau_grobj *rankine = nv30->screen->rankine;
+ struct nouveau_grobj *eng3d = nv30->screen->eng3d;
unsigned type, ncomp;
void *map;
@@ -128,25 +128,25 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
switch (ncomp) {
case 4:
- so_method(so, rankine, NV34TCL_VTX_ATTR_4F_X(attrib), 4);
+ so_method(so, eng3d, NV34TCL_VTX_ATTR_4F_X(attrib), 4);
so_data (so, fui(v[0]));
so_data (so, fui(v[1]));
so_data (so, fui(v[2]));
so_data (so, fui(v[3]));
break;
case 3:
- so_method(so, rankine, NV34TCL_VTX_ATTR_3F_X(attrib), 3);
+ so_method(so, eng3d, NV34TCL_VTX_ATTR_3F_X(attrib), 3);
so_data (so, fui(v[0]));
so_data (so, fui(v[1]));
so_data (so, fui(v[2]));
break;
case 2:
- so_method(so, rankine, NV34TCL_VTX_ATTR_2F_X(attrib), 2);
+ so_method(so, eng3d, NV34TCL_VTX_ATTR_2F_X(attrib), 2);
so_data (so, fui(v[0]));
so_data (so, fui(v[1]));
break;
case 1:
- so_method(so, rankine, NV34TCL_VTX_ATTR_1F(attrib), 1);
+ so_method(so, eng3d, NV34TCL_VTX_ATTR_1F(attrib), 1);
so_data (so, fui(v[0]));
break;
default:
@@ -171,7 +171,7 @@ nv30_draw_arrays(struct pipe_context *pipe,
struct nv30_context *nv30 = nv30_context(pipe);
struct nv30_screen *screen = nv30->screen;
struct nouveau_channel *chan = screen->base.channel;
- struct nouveau_grobj *rankine = screen->rankine;
+ struct nouveau_grobj *eng3d = screen->eng3d;
unsigned restart = 0;
nv30_vbo_set_idxbuf(nv30, NULL, 0);
@@ -193,12 +193,12 @@ nv30_draw_arrays(struct pipe_context *pipe,
continue;
}
- BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
OUT_RING (chan, nvgl_primitive(mode));
nr = (vc & 0xff);
if (nr) {
- BEGIN_RING(chan, rankine, NV34TCL_VB_VERTEX_BATCH, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VB_VERTEX_BATCH, 1);
OUT_RING (chan, ((nr - 1) << 24) | start);
start += nr;
}
@@ -209,14 +209,14 @@ nv30_draw_arrays(struct pipe_context *pipe,
nr -= push;
- BEGIN_RING_NI(chan, rankine, NV34TCL_VB_VERTEX_BATCH, push);
+ BEGIN_RING_NI(chan, eng3d, NV34TCL_VB_VERTEX_BATCH, push);
while (push--) {
OUT_RING(chan, ((0x100 - 1) << 24) | start);
start += 0x100;
}
}
- BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
OUT_RING (chan, 0);
count -= vc;
@@ -232,7 +232,7 @@ nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
{
struct nv30_screen *screen = nv30->screen;
struct nouveau_channel *chan = screen->base.channel;
- struct nouveau_grobj *rankine = screen->rankine;
+ struct nouveau_grobj *eng3d = screen->eng3d;
while (count) {
uint8_t *elts = (uint8_t *)ib + start;
@@ -248,11 +248,11 @@ nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
}
count -= vc;
- BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
OUT_RING (chan, nvgl_primitive(mode));
if (vc & 1) {
- BEGIN_RING(chan, rankine, NV34TCL_VB_ELEMENT_U32, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VB_ELEMENT_U32, 1);
OUT_RING (chan, elts[0]);
elts++; vc--;
}
@@ -262,7 +262,7 @@ nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
push = MIN2(vc, 2047 * 2);
- BEGIN_RING_NI(chan, rankine, NV34TCL_VB_ELEMENT_U16, push >> 1);
+ BEGIN_RING_NI(chan, eng3d, NV34TCL_VB_ELEMENT_U16, push >> 1);
for (i = 0; i < push; i+=2)
OUT_RING(chan, (elts[i+1] << 16) | elts[i]);
@@ -270,7 +270,7 @@ nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
elts += push;
}
- BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
OUT_RING (chan, 0);
start = restart;
@@ -283,7 +283,7 @@ nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
{
struct nv30_screen *screen = nv30->screen;
struct nouveau_channel *chan = screen->base.channel;
- struct nouveau_grobj *rankine = screen->rankine;
+ struct nouveau_grobj *eng3d = screen->eng3d;
while (count) {
uint16_t *elts = (uint16_t *)ib + start;
@@ -299,11 +299,11 @@ nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
}
count -= vc;
- BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
OUT_RING (chan, nvgl_primitive(mode));
if (vc & 1) {
- BEGIN_RING(chan, rankine, NV34TCL_VB_ELEMENT_U32, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VB_ELEMENT_U32, 1);
OUT_RING (chan, elts[0]);
elts++; vc--;
}
@@ -313,7 +313,7 @@ nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
push = MIN2(vc, 2047 * 2);
- BEGIN_RING_NI(chan, rankine, NV34TCL_VB_ELEMENT_U16, push >> 1);
+ BEGIN_RING_NI(chan, eng3d, NV34TCL_VB_ELEMENT_U16, push >> 1);
for (i = 0; i < push; i+=2)
OUT_RING(chan, (elts[i+1] << 16) | elts[i]);
@@ -321,7 +321,7 @@ nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
elts += push;
}
- BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
OUT_RING (chan, 0);
start = restart;
@@ -334,7 +334,7 @@ nv30_draw_elements_u32(struct nv30_context *nv30, void *ib,
{
struct nv30_screen *screen = nv30->screen;
struct nouveau_channel *chan = screen->base.channel;
- struct nouveau_grobj *rankine = screen->rankine;
+ struct nouveau_grobj *eng3d = screen->eng3d;
while (count) {
uint32_t *elts = (uint32_t *)ib + start;
@@ -350,20 +350,20 @@ nv30_draw_elements_u32(struct nv30_context *nv30, void *ib,
}
count -= vc;
- BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
OUT_RING (chan, nvgl_primitive(mode));
while (vc) {
push = MIN2(vc, 2047);
- BEGIN_RING_NI(chan, rankine, NV34TCL_VB_ELEMENT_U32, push);
+ BEGIN_RING_NI(chan, eng3d, NV34TCL_VB_ELEMENT_U32, push);
OUT_RINGp (chan, elts, push);
vc -= push;
elts += push;
}
- BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
OUT_RING (chan, 0);
start = restart;
@@ -410,7 +410,7 @@ nv30_draw_elements_vbo(struct pipe_context *pipe,
struct nv30_context *nv30 = nv30_context(pipe);
struct nv30_screen *screen = nv30->screen;
struct nouveau_channel *chan = screen->base.channel;
- struct nouveau_grobj *rankine = screen->rankine;
+ struct nouveau_grobj *eng3d = screen->eng3d;
unsigned restart = 0;
while (count) {
@@ -425,12 +425,12 @@ nv30_draw_elements_vbo(struct pipe_context *pipe,
continue;
}
- BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
OUT_RING (chan, nvgl_primitive(mode));
nr = (vc & 0xff);
if (nr) {
- BEGIN_RING(chan, rankine, NV34TCL_VB_INDEX_BATCH, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VB_INDEX_BATCH, 1);
OUT_RING (chan, ((nr - 1) << 24) | start);
start += nr;
}
@@ -441,14 +441,14 @@ nv30_draw_elements_vbo(struct pipe_context *pipe,
nr -= push;
- BEGIN_RING_NI(chan, rankine, NV34TCL_VB_INDEX_BATCH, push);
+ BEGIN_RING_NI(chan, eng3d, NV34TCL_VB_INDEX_BATCH, push);
while (push--) {
OUT_RING(chan, ((0x100 - 1) << 24) | start);
start += 0x100;
}
}
- BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+ BEGIN_RING(chan, eng3d, NV34TCL_VERTEX_BEGIN_END, 1);
OUT_RING (chan, 0);
count -= vc;
@@ -485,16 +485,16 @@ static boolean
nv30_vbo_validate(struct nv30_context *nv30)
{
struct nouveau_stateobj *vtxbuf, *vtxfmt, *sattr = NULL;
- struct nouveau_grobj *rankine = nv30->screen->rankine;
+ struct nouveau_grobj *eng3d = nv30->screen->eng3d;
struct pipe_buffer *ib = nv30->idxbuf;
unsigned ib_format = nv30->idxbuf_format;
unsigned vb_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD;
int hw;
vtxbuf = so_new(3, 17, 18);
- so_method(vtxbuf, rankine, NV34TCL_VTXBUF_ADDRESS(0), nv30->vtxelt->num_elements);
+ so_method(vtxbuf, eng3d, NV34TCL_VTXBUF_ADDRESS(0), nv30->vtxelt->num_elements);
vtxfmt = so_new(1, 16, 0);
- so_method(vtxfmt, rankine, NV34TCL_VTXFMT(0), nv30->vtxelt->num_elements);
+ so_method(vtxfmt, eng3d, NV34TCL_VTXFMT(0), nv30->vtxelt->num_elements);
for (hw = 0; hw < nv30->vtxelt->num_elements; hw++) {
struct pipe_vertex_element *ve;
@@ -532,13 +532,13 @@ nv30_vbo_validate(struct nv30_context *nv30)
if (ib) {
struct nouveau_bo *bo = nouveau_bo(ib);
- so_method(vtxbuf, rankine, NV34TCL_IDXBUF_ADDRESS, 2);
+ so_method(vtxbuf, eng3d, NV34TCL_IDXBUF_ADDRESS, 2);
so_reloc (vtxbuf, bo, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
so_reloc (vtxbuf, bo, ib_format, vb_flags | NOUVEAU_BO_OR,
0, NV34TCL_IDXBUF_FORMAT_DMA1);
}
- so_method(vtxbuf, rankine, 0x1710, 1);
+ so_method(vtxbuf, eng3d, 0x1710, 1);
so_data (vtxbuf, 0);
so_ref(vtxbuf, &nv30->state.hw[NV30_STATE_VTXBUF]);