static void
nv10_zclear(struct gl_context *ctx, GLbitfield *buffers)
{
/*
* Pre-nv17 cards don't have native support for fast Z clears,
* but in some cases we can still "clear" the Z buffer without
* actually blitting to it if we're willing to sacrifice a few
* bits of depth precision.
*
* Each time a clear is requested we modify the viewport
* transform in such a way that the old contents of the depth
* buffer are clamped to the requested clear value when
* they're read by the GPU.
*/
struct nouveau_context *nctx = to_nouveau_context(ctx);
struct gl_framebuffer *fb = ctx->DrawBuffer;
struct nouveau_framebuffer *nfb = to_nouveau_framebuffer(fb);
struct nouveau_surface *s = &to_nouveau_renderbuffer(
fb->Attachment[BUFFER_DEPTH].Renderbuffer)->surface;
if (nv10_use_viewport_zclear(ctx)) {
int x, y, w, h;
float z = ctx->Depth.Clear;
uint32_t value = pack_zs_f(s->format, z, 0);
get_scissors(fb, &x, &y, &w, &h);
*buffers &= ~BUFFER_BIT_DEPTH;
if (use_fast_zclear(ctx, *buffers)) {
if (nfb->hierz.clear_value != value) {
/* Don't fast clear if we're changing
* the depth value. */
nfb->hierz.clear_value = value;
} else if (z == 0.0) {
nctx->hierz.clear_seq++;
context_dirty(ctx, ZCLEAR);
if ((nctx->hierz.clear_seq & 7) != 0 &&
nctx->hierz.clear_seq != 1)
/* We didn't wrap around -- no need to
* clear the depth buffer for real. */
return;
} else if (z == 1.0) {
nctx->hierz.clear_seq--;
context_dirty(ctx, ZCLEAR);
if ((nctx->hierz.clear_seq & 7) != 7)
/* No wrap around */
return;
}
}
value = pack_zs_f(s->format,
(z + (nctx->hierz.clear_seq & 7)) / 8, 0);
context_drv(ctx)->surface_fill(ctx, s, ~0, value, x, y, w, h);
}
}
void
nv04_emit_scissor(struct gl_context *ctx, int emit)
{
struct nouveau_pushbuf *push = context_push(ctx);
int x, y, w, h;
get_scissors(ctx->DrawBuffer, &x, &y, &w, &h);
BEGIN_NV04(push, NV04_SF3D(CLIP_HORIZONTAL), 2);
PUSH_DATA (push, w << 16 | x);
PUSH_DATA (push, h << 16 | y);
}
void
nouveau_clear(GLcontext *ctx, GLbitfield buffers)
{
struct gl_framebuffer *fb = ctx->DrawBuffer;
int x, y, w, h;
int i, buf;
nouveau_validate_framebuffer(ctx);
get_scissors(fb, &x, &y, &w, &h);
for (i = 0; i < BUFFER_COUNT; i++) {
struct nouveau_surface *s;
unsigned mask, value;
buf = buffers & (1 << i);
if (!buf)
continue;
s = &to_nouveau_renderbuffer(
fb->Attachment[i].Renderbuffer->Wrapped)->surface;
if (buf & BUFFER_BITS_COLOR) {
mask = pack_rgba_i(s->format, ctx->Color.ColorMask[0]);
value = pack_rgba_f(s->format, ctx->Color.ClearColor);
if (mask)
context_drv(ctx)->surface_fill(
ctx, s, mask, value, x, y, w, h);
buffers &= ~buf;
} else if (buf & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) {
mask = pack_zs_i(s->format,
(buffers & BUFFER_BIT_DEPTH &&
ctx->Depth.Mask) ? ~0 : 0,
(buffers & BUFFER_BIT_STENCIL ?
ctx->Stencil.WriteMask[0] : 0));
value = pack_zs_f(s->format,
ctx->Depth.Clear,
ctx->Stencil.Clear);
if (mask)
context_drv(ctx)->surface_fill(
ctx, s, mask, value, x, y, w, h);
buffers &= ~(BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL);
}
}
if (buffers)
_mesa_meta_Clear(ctx, buffers);
}
void
nv10_emit_scissor(struct gl_context *ctx, int emit)
{
struct nouveau_channel *chan = context_chan(ctx);
struct nouveau_grobj *celsius = context_eng3d(ctx);
int x, y, w, h;
get_scissors(ctx->DrawBuffer, &x, &y, &w, &h);
BEGIN_RING(chan, celsius, NV10_3D_RT_HORIZ, 2);
OUT_RING(chan, w << 16 | x);
OUT_RING(chan, h << 16 | y);
}
void
nv04_emit_scissor(GLcontext *ctx, int emit)
{
struct nouveau_channel *chan = context_chan(ctx);
struct nouveau_hw_state *hw = &to_nouveau_context(ctx)->hw;
struct nouveau_grobj *surf3d = hw->surf3d;
int x, y, w, h;
get_scissors(ctx->DrawBuffer, &x, &y, &w, &h);
BEGIN_RING(chan, surf3d, NV04_CONTEXT_SURFACES_3D_CLIP_HORIZONTAL, 2);
OUT_RING(chan, w << 16 | x);
OUT_RING(chan, h << 16 | y);
/* Messing with surf3d invalidates some engine state. */
context_dirty(ctx, CONTROL);
context_dirty(ctx, BLEND);
}
static void
setup_hierz_buffer(struct gl_context *ctx)
{
struct nouveau_pushbuf *push = context_push(ctx);
struct gl_framebuffer *fb = ctx->DrawBuffer;
struct nouveau_framebuffer *nfb = to_nouveau_framebuffer(fb);
unsigned pitch = align(fb->Width, 128),
height = align(fb->Height, 2),
size = pitch * height;
if (!nfb->hierz.bo || nfb->hierz.bo->size != size) {
union nouveau_bo_config config = {
.nv04.surf_flags = NV04_BO_ZETA,
.nv04.surf_pitch = 0
};
nouveau_bo_ref(NULL, &nfb->hierz.bo);
nouveau_bo_new(context_dev(ctx), NOUVEAU_BO_VRAM, 0, size,
&config, &nfb->hierz.bo);
}
PUSH_SPACE(push, 11);
BEGIN_NV04(push, NV17_3D(HIERZ_OFFSET), 1);
PUSH_MTHDl(push, NV17_3D(HIERZ_OFFSET), BUFCTX_FB,
nfb->hierz.bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
BEGIN_NV04(push, NV17_3D(HIERZ_WINDOW_X), 4);
PUSH_DATAf(push, - 1792);
PUSH_DATAf(push, - 2304 + fb->Height);
PUSH_DATAf(push, fb->_DepthMaxF / 2);
PUSH_DATAf(push, 0);
BEGIN_NV04(push, NV17_3D(HIERZ_PITCH), 1);
PUSH_DATA (push, pitch);
BEGIN_NV04(push, NV17_3D(HIERZ_ENABLE), 1);
PUSH_DATA (push, 1);
}
void
nv10_emit_framebuffer(struct gl_context *ctx, int emit)
{
struct nouveau_pushbuf *push = context_push(ctx);
struct gl_framebuffer *fb = ctx->DrawBuffer;
struct nouveau_surface *s;
unsigned rt_format = NV10_3D_RT_FORMAT_TYPE_LINEAR;
unsigned rt_pitch = 0, zeta_pitch = 0;
unsigned bo_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR;
if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT)
return;
PUSH_RESET(push, BUFCTX_FB);
/* At least nv11 seems to get sad if we don't do this before
* swapping RTs.*/
if (context_eng3d(ctx)->oclass < NV17_3D_CLASS) {
int i;
for (i = 0; i < 6; i++) {
BEGIN_NV04(push, NV04_GRAPH(3D, NOP), 1);
PUSH_DATA (push, 0);
}
}
/* Render target */
if (fb->_ColorDrawBuffers[0]) {
s = &to_nouveau_renderbuffer(
fb->_ColorDrawBuffers[0])->surface;
rt_format |= get_rt_format(s->format);
zeta_pitch = rt_pitch = s->pitch;
BEGIN_NV04(push, NV10_3D(COLOR_OFFSET), 1);
PUSH_MTHDl(push, NV10_3D(COLOR_OFFSET), BUFCTX_FB,
s->bo, 0, bo_flags);
}
/* depth/stencil */
if (fb->Attachment[BUFFER_DEPTH].Renderbuffer) {
s = &to_nouveau_renderbuffer(
fb->Attachment[BUFFER_DEPTH].Renderbuffer)->surface;
rt_format |= get_rt_format(s->format);
zeta_pitch = s->pitch;
BEGIN_NV04(push, NV10_3D(ZETA_OFFSET), 1);
PUSH_MTHDl(push, NV10_3D(ZETA_OFFSET), BUFCTX_FB,
s->bo, 0, bo_flags);
if (context_eng3d(ctx)->oclass >= NV17_3D_CLASS) {
setup_hierz_buffer(ctx);
context_dirty(ctx, ZCLEAR);
}
}
BEGIN_NV04(push, NV10_3D(RT_FORMAT), 2);
PUSH_DATA (push, rt_format);
PUSH_DATA (push, zeta_pitch << 16 | rt_pitch);
context_dirty(ctx, VIEWPORT);
context_dirty(ctx, SCISSOR);
context_dirty(ctx, DEPTH);
}
void
nv10_emit_render_mode(struct gl_context *ctx, int emit)
{
}
void
nv10_emit_scissor(struct gl_context *ctx, int emit)
{
struct nouveau_pushbuf *push = context_push(ctx);
int x, y, w, h;
get_scissors(ctx->DrawBuffer, &x, &y, &w, &h);
BEGIN_NV04(push, NV10_3D(RT_HORIZ), 2);
PUSH_DATA (push, w << 16 | x);
PUSH_DATA (push, h << 16 | y);
}
void
nv10_emit_viewport(struct gl_context *ctx, int emit)
{
struct nouveau_pushbuf *push = context_push(ctx);
struct gl_viewport_attrib *vp = &ctx->ViewportArray[0];
struct gl_framebuffer *fb = ctx->DrawBuffer;
float a[4] = {};
get_viewport_translate(ctx, a);
a[0] -= 2048;
a[1] -= 2048;
if (nv10_use_viewport_zclear(ctx))
a[2] = nv10_transform_depth(ctx, (vp->Far + vp->Near) / 2);
BEGIN_NV04(push, NV10_3D(VIEWPORT_TRANSLATE_X), 4);
PUSH_DATAp(push, a, 4);
BEGIN_NV04(push, NV10_3D(VIEWPORT_CLIP_HORIZ(0)), 1);
PUSH_DATA (push, (fb->Width - 1) << 16 | 0x08000800);
BEGIN_NV04(push, NV10_3D(VIEWPORT_CLIP_VERT(0)), 1);
PUSH_DATA (push, (fb->Height - 1) << 16 | 0x08000800);
context_dirty(ctx, PROJECTION);
}
void
nv10_emit_zclear(struct gl_context *ctx, int emit)
{
struct nouveau_context *nctx = to_nouveau_context(ctx);
struct nouveau_pushbuf *push = context_push(ctx);
struct nouveau_framebuffer *nfb =
to_nouveau_framebuffer(ctx->DrawBuffer);
if (nfb->hierz.bo) {
BEGIN_NV04(push, NV17_3D(ZCLEAR_ENABLE), 2);
PUSH_DATAb(push, !nctx->hierz.clear_blocked);
PUSH_DATA (push, nfb->hierz.clear_value |
(nctx->hierz.clear_seq & 0xff));
} else {
BEGIN_NV04(push, NV10_3D(DEPTH_RANGE_NEAR), 2);
PUSH_DATAf(push, nv10_transform_depth(ctx, 0));
PUSH_DATAf(push, nv10_transform_depth(ctx, 1));
context_dirty(ctx, VIEWPORT);
}
}
void
nouveau_clear(struct gl_context *ctx, GLbitfield buffers)
{
struct gl_framebuffer *fb = ctx->DrawBuffer;
int x, y, w, h;
int i, buf;
nouveau_validate_framebuffer(ctx);
get_scissors(fb, &x, &y, &w, &h);
for (i = 0; i < BUFFER_COUNT; i++) {
struct nouveau_surface *s;
unsigned mask, value;
buf = buffers & (1 << i);
if (!buf)
continue;
s = &to_nouveau_renderbuffer(
fb->Attachment[i].Renderbuffer)->surface;
if (buf & BUFFER_BITS_COLOR) {
const float *color = ctx->Color.ClearColor.f;
if (fb->Attachment[i].Renderbuffer->_BaseFormat ==
GL_LUMINANCE_ALPHA)
value = pack_la_clamp_f(
s->format, color[0], color[3]);
else
value = pack_rgba_clamp_f(s->format, color);
const uint8_t colormask[4] = {
GET_COLORMASK_BIT(ctx->Color.ColorMask, 0, 0) ? 0xff : 0,
GET_COLORMASK_BIT(ctx->Color.ColorMask, 0, 1) ? 0xff : 0,
GET_COLORMASK_BIT(ctx->Color.ColorMask, 0, 2) ? 0xff : 0,
GET_COLORMASK_BIT(ctx->Color.ColorMask, 0, 3) ? 0xff : 0,
};
mask = pack_rgba_i(s->format, colormask);
if (mask)
context_drv(ctx)->surface_fill(
ctx, s, mask, value, x, y, w, h);
buffers &= ~buf;
} else if (buf & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) {
mask = pack_zs_i(s->format,
(buffers & BUFFER_BIT_DEPTH &&
ctx->Depth.Mask) ? ~0 : 0,
(buffers & BUFFER_BIT_STENCIL ?
ctx->Stencil.WriteMask[0] : 0));
value = pack_zs_f(s->format,
ctx->Depth.Clear,
ctx->Stencil.Clear);
if (mask)
context_drv(ctx)->surface_fill(
ctx, s, mask, value, x, y, w, h);
buffers &= ~(BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL);
}
}
if (buffers)
_mesa_meta_Clear(ctx, buffers);
}