Mesa (nvc0): nvc0: support user clip planes
Christoph Bumiller
chrisbmr at kemper.freedesktop.org
Sat Dec 11 15:25:04 UTC 2010
Module: Mesa
Branch: nvc0
Commit: 5138ac033ad3708e2b82f2beebc887f65a77309e
URL: http://cgit.freedesktop.org/mesa/mesa/commit/?id=5138ac033ad3708e2b82f2beebc887f65a77309e
Author: Christoph Bumiller <e0425955 at student.tuwien.ac.at>
Date: Sat Dec 11 16:23:43 2010 +0100
nvc0: support user clip planes
---
src/gallium/drivers/nvc0/nvc0_program.c | 7 +++
src/gallium/drivers/nvc0/nvc0_program.h | 2 +
src/gallium/drivers/nvc0/nvc0_screen.c | 12 +++++-
src/gallium/drivers/nvc0/nvc0_shader_state.c | 8 ++++
src/gallium/drivers/nvc0/nvc0_state.c | 5 ++
src/gallium/drivers/nvc0/nvc0_state_validate.c | 17 ++++++++
src/gallium/drivers/nvc0/nvc0_tgsi_to_nc.c | 48 +++++++++++++++++++++++-
7 files changed, 96 insertions(+), 3 deletions(-)
diff --git a/src/gallium/drivers/nvc0/nvc0_program.c b/src/gallium/drivers/nvc0/nvc0_program.c
index 7aa6ef0..e159b71 100644
--- a/src/gallium/drivers/nvc0/nvc0_program.c
+++ b/src/gallium/drivers/nvc0/nvc0_program.c
@@ -418,6 +418,8 @@ nvc0_vp_gen_header(struct nvc0_program *vp, struct nvc0_translation_info *ti)
vp->hdr[0] = 0x20461;
vp->hdr[4] = 0xff000;
+ vp->hdr[18] = (1 << vp->vp.num_ucps) - 1;
+
return nvc0_vp_gp_gen_header(vp, ti);
}
@@ -605,6 +607,9 @@ nvc0_program_translate(struct nvc0_program *prog)
ti->edgeflag_out = PIPE_MAX_SHADER_OUTPUTS;
+ if (prog->type == PIPE_SHADER_VERTEX && prog->vp.num_ucps)
+ ti->append_ucp = TRUE;
+
ret = nvc0_prog_scan(ti);
if (ret) {
NOUVEAU_ERR("unsupported shader program\n");
@@ -646,5 +651,7 @@ nvc0_program_destroy(struct nvc0_context *nvc0, struct nvc0_program *prog)
if (prog->relocs)
FREE(prog->relocs);
+ memset(prog->hdr, 0, sizeof(prog->hdr));
+
prog->translated = FALSE;
}
diff --git a/src/gallium/drivers/nvc0/nvc0_program.h b/src/gallium/drivers/nvc0/nvc0_program.h
index 42d9be3..1271303 100644
--- a/src/gallium/drivers/nvc0/nvc0_program.h
+++ b/src/gallium/drivers/nvc0/nvc0_program.h
@@ -27,6 +27,7 @@ struct nvc0_program {
struct {
uint8_t edgeflag;
+ uint8_t num_ucps;
} vp;
void *relocs;
@@ -71,6 +72,7 @@ struct nvc0_translation_info {
ubyte edgeflag_out;
struct nvc0_subroutine *subr;
unsigned num_subrs;
+ boolean append_ucp;
struct tgsi_shader_info scan;
};
diff --git a/src/gallium/drivers/nvc0/nvc0_screen.c b/src/gallium/drivers/nvc0/nvc0_screen.c
index fb100e5..f0d06be 100644
--- a/src/gallium/drivers/nvc0/nvc0_screen.c
+++ b/src/gallium/drivers/nvc0/nvc0_screen.c
@@ -457,11 +457,21 @@ nvc0_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
nouveau_resource_init(&screen->text_heap, 0, 1 << 20);
- ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 12, 5 << 16,
+ ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 12, 6 << 16,
&screen->uniforms);
if (ret)
goto fail;
+ /* auxiliary constants (6 user clip planes, base instance id) */
+ BEGIN_RING(chan, RING_3D(CB_SIZE), 3);
+ OUT_RING (chan, 256);
+ OUT_RELOCh(chan, screen->uniforms, 5 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+ OUT_RELOCl(chan, screen->uniforms, 5 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+ for (i = 0; i < 5; ++i) {
+ BEGIN_RING(chan, RING_3D(CB_BIND(i)), 1);
+ OUT_RING (chan, (15 << 4) | 1);
+ }
+
screen->tls_size = 4 * 4 * 32 * 128 * 4;
ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 17,
screen->tls_size, &screen->tls);
diff --git a/src/gallium/drivers/nvc0/nvc0_shader_state.c b/src/gallium/drivers/nvc0/nvc0_shader_state.c
index a16fa6c..a6595c5 100644
--- a/src/gallium/drivers/nvc0/nvc0_shader_state.c
+++ b/src/gallium/drivers/nvc0/nvc0_shader_state.c
@@ -67,6 +67,14 @@ nvc0_vertprog_validate(struct nvc0_context *nvc0)
struct nouveau_channel *chan = nvc0->screen->base.channel;
struct nvc0_program *vp = nvc0->vertprog;
+ if (nvc0->clip.nr > vp->vp.num_ucps) {
+ assert(nvc0->clip.nr <= 6);
+ vp->vp.num_ucps = 6;
+
+ if (vp->translated)
+ nvc0_program_destroy(nvc0, vp);
+ }
+
if (!nvc0_program_validate(nvc0, vp))
return;
diff --git a/src/gallium/drivers/nvc0/nvc0_state.c b/src/gallium/drivers/nvc0/nvc0_state.c
index c7a8c4b..8d29323 100644
--- a/src/gallium/drivers/nvc0/nvc0_state.c
+++ b/src/gallium/drivers/nvc0/nvc0_state.c
@@ -699,8 +699,13 @@ nvc0_set_clip_state(struct pipe_context *pipe,
const struct pipe_clip_state *clip)
{
struct nvc0_context *nvc0 = nvc0_context(pipe);
+ const unsigned size = clip->nr * sizeof(clip->ucp[0]);
+
+ memcpy(&nvc0->clip.ucp[0][0], &clip->ucp[0][0], size);
+ nvc0->clip.nr = clip->nr;
nvc0->clip.depth_clamp = clip->depth_clamp;
+
nvc0->dirty |= NVC0_NEW_CLIP;
}
diff --git a/src/gallium/drivers/nvc0/nvc0_state_validate.c b/src/gallium/drivers/nvc0/nvc0_state_validate.c
index 2b38ebc..ded461b 100644
--- a/src/gallium/drivers/nvc0/nvc0_state_validate.c
+++ b/src/gallium/drivers/nvc0/nvc0_state_validate.c
@@ -193,6 +193,23 @@ nvc0_validate_clip(struct nvc0_context *nvc0)
BEGIN_RING(chan, RING_3D(VIEW_VOLUME_CLIP_CTRL), 1);
OUT_RING (chan, clip);
+
+ if (nvc0->clip.nr) {
+ struct nouveau_bo *bo = nvc0->screen->uniforms;
+
+ BEGIN_RING(chan, RING_3D(CB_SIZE), 3);
+ OUT_RING (chan, 256);
+ OUT_RELOCh(chan, bo, 5 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+ OUT_RELOCl(chan, bo, 5 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+ BEGIN_RING_1I(chan, RING_3D(CB_POS), nvc0->clip.nr * 4 + 1);
+ OUT_RING (chan, 0);
+ OUT_RINGp (chan, &nvc0->clip.ucp[0][0], nvc0->clip.nr * 4);
+
+ BEGIN_RING(chan, RING_3D(VP_CLIP_DISTANCE_ENABLE), 1);
+ OUT_RING (chan, (1 << nvc0->clip.nr) - 1);
+ } else {
+ INLIN_RING(chan, RING_3D(VP_CLIP_DISTANCE_ENABLE), 0);
+ }
}
static void
diff --git a/src/gallium/drivers/nvc0/nvc0_tgsi_to_nc.c b/src/gallium/drivers/nvc0/nvc0_tgsi_to_nc.c
index 8a20983..d0c8275 100644
--- a/src/gallium/drivers/nvc0/nvc0_tgsi_to_nc.c
+++ b/src/gallium/drivers/nvc0/nvc0_tgsi_to_nc.c
@@ -111,6 +111,7 @@ struct bld_context {
struct bld_register ovs[BLD_MAX_OUTPS][4]; /* TGSI_FILE_OUTPUT, FP only */
uint32_t outputs_written[(PIPE_MAX_SHADER_OUTPUTS + 7) / 8];
+ int hpos_index;
struct nv_value *zero;
struct nv_value *frag_coord[4];
@@ -904,6 +905,38 @@ bld_is_output_written(struct bld_context *bld, int i, int c)
}
static void
+bld_append_vp_ucp(struct bld_context *bld)
+{
+ struct nv_value *res[6];
+ struct nv_value *ucp, *vtx, *out;
+ struct nv_instruction *insn;
+ int i, c;
+
+ assert(bld->ti->prog->vp.num_ucps <= 6);
+
+ for (c = 0; c < 4; ++c) {
+ vtx = bld_fetch_global(bld, &bld->ovs[bld->hpos_index][c]);
+
+ for (i = 0; i < bld->ti->prog->vp.num_ucps; ++i) {
+ ucp = new_value(bld->pc, NV_FILE_MEM_C(15), 4);
+ ucp->reg.address = i * 16 + c * 4;
+
+ if (c == 0)
+ res[i] = bld_insn_2(bld, NV_OP_MUL_F32, vtx, ucp);
+ else
+ res[i] = bld_insn_3(bld, NV_OP_MAD_F32, vtx, ucp, res[i]);
+ }
+ }
+
+ for (i = 0; i < bld->ti->prog->vp.num_ucps; ++i) {
+ (out = new_value(bld->pc, NV_FILE_MEM_V, 4))->reg.address = 0x2c0 + i * 4;
+ (insn = new_instruction(bld->pc, NV_OP_EXPORT))->fixed = 1;
+ nv_reference(bld->pc, insn, 0, out);
+ nv_reference(bld->pc, insn, 1, res[i]);
+ }
+}
+
+static void
bld_export_fp_outputs(struct bld_context *bld)
{
struct nv_value *vals[4];
@@ -1755,11 +1788,13 @@ bld_instruction(struct bld_context *bld,
/* VP outputs are exported in-place as scalars, optimization later */
if (bld->pc->is_fragprog)
bld_export_fp_outputs(bld);
- break;
+ if (bld->ti->append_ucp)
+ bld_append_vp_ucp(bld);
+ return;
default:
NOUVEAU_ERR("unhandled opcode %u\n", insn->Instruction.Opcode);
abort();
- break;
+ return;
}
if (insn->Dst[0].Register.File == TGSI_FILE_OUTPUT &&
@@ -1767,6 +1802,15 @@ bld_instruction(struct bld_context *bld,
struct nv_instruction *mi = NULL;
uint size;
+ if (bld->ti->append_ucp) {
+ if (bld->ti->output_loc[insn->Dst[0].Register.Index][0] == 0x70) {
+ bld->hpos_index = insn->Dst[0].Register.Index;
+ for (c = 0; c < 4; ++c)
+ if (mask & (1 << c))
+ STORE_OUTP(insn->Dst[0].Register.Index, c, dst0[c]);
+ }
+ }
+
for (c = 0; c < 4; ++c)
if ((mask & (1 << c)) &&
((dst0[c]->reg.file == NV_FILE_IMM) ||
More information about the mesa-commit
mailing list