static void
gen7_blorp_emit_depth_disable(struct brw_context *brw,
const brw_blorp_params *params)
{
intel_emit_depth_stall_flushes(brw);
BEGIN_BATCH(7);
OUT_BATCH(GEN7_3DSTATE_DEPTH_BUFFER << 16 | (7 - 2));
OUT_BATCH(BRW_DEPTHFORMAT_D32_FLOAT << 18 | (BRW_SURFACE_NULL << 29));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_HIER_DEPTH_BUFFER << 16 | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_STENCIL_BUFFER << 16 | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
void
gen7_emit_depth_stencil_hiz(struct brw_context *brw,
struct intel_mipmap_tree *depth_mt,
uint32_t depth_offset, uint32_t depthbuffer_format,
uint32_t depth_surface_type,
struct intel_mipmap_tree *stencil_mt,
bool hiz, bool separate_stencil,
uint32_t width, uint32_t height,
uint32_t tile_x, uint32_t tile_y)
{
struct gl_context *ctx = &brw->ctx;
const uint8_t mocs = GEN7_MOCS_L3;
struct gl_framebuffer *fb = ctx->DrawBuffer;
uint32_t surftype;
unsigned int depth = 1;
unsigned int min_array_element;
GLenum gl_target = GL_TEXTURE_2D;
unsigned int lod;
const struct intel_mipmap_tree *mt = depth_mt ? depth_mt : stencil_mt;
const struct intel_renderbuffer *irb = NULL;
const struct gl_renderbuffer *rb = NULL;
intel_emit_depth_stall_flushes(brw);
irb = intel_get_renderbuffer(fb, BUFFER_DEPTH);
if (!irb)
irb = intel_get_renderbuffer(fb, BUFFER_STENCIL);
rb = (struct gl_renderbuffer*) irb;
if (rb) {
depth = MAX2(rb->Depth, 1);
if (rb->TexImage)
gl_target = rb->TexImage->TexObject->Target;
}
switch (gl_target) {
case GL_TEXTURE_CUBE_MAP_ARRAY:
case GL_TEXTURE_CUBE_MAP:
/* The PRM claims that we should use BRW_SURFACE_CUBE for this
* situation, but experiments show that gl_Layer doesn't work when we do
* this. So we use BRW_SURFACE_2D, since for rendering purposes this is
* equivalent.
*/
surftype = BRW_SURFACE_2D;
depth *= 6;
break;
default:
surftype = translate_tex_target(gl_target);
break;
}
if (fb->Layered || !irb) {
min_array_element = 0;
} else if (irb->mt->num_samples > 1) {
/* Convert physical layer to logical layer. */
min_array_element = irb->mt_layer / irb->mt->num_samples;
} else {
min_array_element = irb->mt_layer;
}
lod = irb ? irb->mt_level - irb->mt->first_level : 0;
if (mt) {
width = mt->physical_width0;
height = mt->physical_height0;
}
/* _NEW_DEPTH, _NEW_STENCIL, _NEW_BUFFERS */
BEGIN_BATCH(7);
/* 3DSTATE_DEPTH_BUFFER dw0 */
OUT_BATCH(GEN7_3DSTATE_DEPTH_BUFFER << 16 | (7 - 2));
/* 3DSTATE_DEPTH_BUFFER dw1 */
OUT_BATCH((depth_mt ? depth_mt->region->pitch - 1 : 0) |
(depthbuffer_format << 18) |
((hiz ? 1 : 0) << 22) |
((stencil_mt != NULL && ctx->Stencil._WriteEnabled) << 27) |
((ctx->Depth.Mask != 0) << 28) |
(surftype << 29));
/* 3DSTATE_DEPTH_BUFFER dw2 */
if (depth_mt) {
OUT_RELOC(depth_mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
0);
} else {
OUT_BATCH(0);
}
/* 3DSTATE_DEPTH_BUFFER dw3 */
OUT_BATCH(((width - 1) << 4) |
((height - 1) << 18) |
lod);
/* 3DSTATE_DEPTH_BUFFER dw4 */
OUT_BATCH(((depth - 1) << 21) |
(min_array_element << 10) |
mocs);
/* 3DSTATE_DEPTH_BUFFER dw5 */
OUT_BATCH(0);
/* 3DSTATE_DEPTH_BUFFER dw6 */
OUT_BATCH((depth - 1) << 21);
ADVANCE_BATCH();
if (!hiz) {
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_HIER_DEPTH_BUFFER << 16 | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
struct intel_mipmap_tree *hiz_mt = depth_mt->hiz_mt;
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_HIER_DEPTH_BUFFER << 16 | (3 - 2));
OUT_BATCH((mocs << 25) |
(hiz_mt->region->pitch - 1));
OUT_RELOC(hiz_mt->region->bo,
I915_GEM_DOMAIN_RENDER,
I915_GEM_DOMAIN_RENDER,
0);
ADVANCE_BATCH();
}
if (stencil_mt == NULL) {
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_STENCIL_BUFFER << 16 | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
const int enabled = brw->is_haswell ? HSW_STENCIL_ENABLED : 0;
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_STENCIL_BUFFER << 16 | (3 - 2));
/* The stencil buffer has quirky pitch requirements. From the
* Sandybridge PRM, Volume 2 Part 1, page 329 (3DSTATE_STENCIL_BUFFER
* dword 1 bits 16:0 - Surface Pitch):
*
* The pitch must be set to 2x the value computed based on width, as
* the stencil buffer is stored with two rows interleaved.
*
* While the Ivybridge PRM lacks this comment, the BSpec contains the
* same text, and experiments indicate that this is necessary.
*/
OUT_BATCH(enabled |
mocs << 25 |
(2 * stencil_mt->region->pitch - 1));
OUT_RELOC(stencil_mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
0);
ADVANCE_BATCH();
}
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_CLEAR_PARAMS << 16 | (3 - 2));
OUT_BATCH(depth_mt ? depth_mt->depth_clear_value : 0);
OUT_BATCH(1);
ADVANCE_BATCH();
}
static void
gen6_blorp_emit_depth_stencil_config(struct brw_context *brw,
const brw_blorp_params *params)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
uint32_t draw_x = params->depth.x_offset;
uint32_t draw_y = params->depth.y_offset;
uint32_t tile_mask_x, tile_mask_y;
brw_get_depthstencil_tile_masks(params->depth.mt,
params->depth.level,
params->depth.layer,
NULL,
&tile_mask_x, &tile_mask_y);
/* 3DSTATE_DEPTH_BUFFER */
{
uint32_t tile_x = draw_x & tile_mask_x;
uint32_t tile_y = draw_y & tile_mask_y;
uint32_t offset =
intel_region_get_aligned_offset(params->depth.mt->region,
draw_x & ~tile_mask_x,
draw_y & ~tile_mask_y, false);
/* According to the Sandy Bridge PRM, volume 2 part 1, pp326-327
* (3DSTATE_DEPTH_BUFFER dw5), in the documentation for "Depth
* Coordinate Offset X/Y":
*
* "The 3 LSBs of both offsets must be zero to ensure correct
* alignment"
*
* We have no guarantee that tile_x and tile_y are correctly aligned,
* since they are determined by the mipmap layout, which is only aligned
* to multiples of 4.
*
* So, to avoid hanging the GPU, just smash the low order 3 bits of
* tile_x and tile_y to 0. This is a temporary workaround until we come
* up with a better solution.
*/
WARN_ONCE((tile_x & 7) || (tile_y & 7),
"Depth/stencil buffer needs alignment to 8-pixel boundaries.\n"
"Truncating offset, bad rendering may occur.\n");
tile_x &= ~7;
tile_y &= ~7;
intel_emit_post_sync_nonzero_flush(intel);
intel_emit_depth_stall_flushes(intel);
BEGIN_BATCH(7);
OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (7 - 2));
OUT_BATCH((params->depth.mt->region->pitch - 1) |
params->depth_format << 18 |
1 << 21 | /* separate stencil enable */
1 << 22 | /* hiz enable */
BRW_TILEWALK_YMAJOR << 26 |
1 << 27 | /* y-tiled */
BRW_SURFACE_2D << 29);
OUT_RELOC(params->depth.mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
offset);
OUT_BATCH(BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1 |
(params->depth.width + tile_x - 1) << 6 |
(params->depth.height + tile_y - 1) << 19);
OUT_BATCH(0);
OUT_BATCH(tile_x |
tile_y << 16);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_HIER_DEPTH_BUFFER */
{
struct intel_region *hiz_region = params->depth.mt->hiz_mt->region;
uint32_t hiz_offset =
intel_region_get_aligned_offset(hiz_region,
draw_x & ~tile_mask_x,
(draw_y & ~tile_mask_y) / 2, false);
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
OUT_BATCH(hiz_region->pitch - 1);
OUT_RELOC(hiz_region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
hiz_offset);
ADVANCE_BATCH();
}
/* 3DSTATE_STENCIL_BUFFER */
{
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
}
static void
gen7_blorp_emit_depth_stencil_config(struct brw_context *brw,
const brw_blorp_params *params)
{
const uint8_t mocs = GEN7_MOCS_L3;
uint32_t surfwidth, surfheight;
uint32_t surftype;
unsigned int depth = MAX2(params->depth.mt->logical_depth0, 1);
unsigned int min_array_element;
GLenum gl_target = params->depth.mt->target;
unsigned int lod;
switch (gl_target) {
case GL_TEXTURE_CUBE_MAP_ARRAY:
case GL_TEXTURE_CUBE_MAP:
/* The PRM claims that we should use BRW_SURFACE_CUBE for this
* situation, but experiments show that gl_Layer doesn't work when we do
* this. So we use BRW_SURFACE_2D, since for rendering purposes this is
* equivalent.
*/
surftype = BRW_SURFACE_2D;
depth *= 6;
break;
default:
surftype = translate_tex_target(gl_target);
break;
}
min_array_element = params->depth.layer;
if (params->depth.mt->num_samples > 1) {
/* Convert physical layer to logical layer. */
min_array_element /= params->depth.mt->num_samples;
}
lod = params->depth.level - params->depth.mt->first_level;
if (params->hiz_op != GEN6_HIZ_OP_NONE && lod == 0) {
/* HIZ ops for lod 0 may set the width & height a little
* larger to allow the fast depth clear to fit the hardware
* alignment requirements. (8x4)
*/
surfwidth = params->depth.width;
surfheight = params->depth.height;
} else {
surfwidth = params->depth.mt->logical_width0;
surfheight = params->depth.mt->logical_height0;
}
/* 3DSTATE_DEPTH_BUFFER */
{
intel_emit_depth_stall_flushes(brw);
BEGIN_BATCH(7);
OUT_BATCH(GEN7_3DSTATE_DEPTH_BUFFER << 16 | (7 - 2));
OUT_BATCH((params->depth.mt->region->pitch - 1) |
params->depth_format << 18 |
1 << 22 | /* hiz enable */
1 << 28 | /* depth write */
surftype << 29);
OUT_RELOC(params->depth.mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
0);
OUT_BATCH((surfwidth - 1) << 4 |
(surfheight - 1) << 18 |
lod);
OUT_BATCH(((depth - 1) << 21) |
(min_array_element << 10) |
mocs);
OUT_BATCH(0);
OUT_BATCH((depth - 1) << 21);
ADVANCE_BATCH();
}
/* 3DSTATE_HIER_DEPTH_BUFFER */
{
struct intel_region *hiz_region = params->depth.mt->hiz_mt->region;
BEGIN_BATCH(3);
OUT_BATCH((GEN7_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
OUT_BATCH((mocs << 25) |
(hiz_region->pitch - 1));
OUT_RELOC(hiz_region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
0);
ADVANCE_BATCH();
}
/* 3DSTATE_STENCIL_BUFFER */
{
BEGIN_BATCH(3);
OUT_BATCH((GEN7_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
}
static void emit_depthbuffer(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
struct gl_framebuffer *fb = ctx->DrawBuffer;
/* _NEW_BUFFERS */
struct intel_renderbuffer *depth_irb = intel_get_renderbuffer(fb, BUFFER_DEPTH);
struct intel_renderbuffer *stencil_irb = intel_get_renderbuffer(fb, BUFFER_STENCIL);
struct intel_mipmap_tree *depth_mt = brw->depthstencil.depth_mt;
struct intel_mipmap_tree *stencil_mt = brw->depthstencil.stencil_mt;
struct intel_mipmap_tree *hiz_mt = brw->depthstencil.hiz_mt;
uint32_t tile_x = brw->depthstencil.tile_x;
uint32_t tile_y = brw->depthstencil.tile_y;
unsigned int len;
bool separate_stencil = false;
if (stencil_mt && stencil_mt->format == MESA_FORMAT_S8)
separate_stencil = true;
/* 3DSTATE_DEPTH_BUFFER, 3DSTATE_STENCIL_BUFFER are both
* non-pipelined state that will need the PIPE_CONTROL workaround.
*/
if (intel->gen == 6) {
intel_emit_post_sync_nonzero_flush(intel);
intel_emit_depth_stall_flushes(intel);
}
/* If there's a packed depth/stencil bound to stencil only, we need to
* emit the packed depth/stencil buffer packet.
*/
if (!depth_irb && stencil_irb && !separate_stencil) {
depth_irb = stencil_irb;
depth_mt = stencil_mt;
}
if (intel->gen >= 6)
len = 7;
else if (intel->is_g4x || intel->gen == 5)
len = 6;
else
len = 5;
if (!depth_irb && !separate_stencil) {
BEGIN_BATCH(len);
OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (len - 2));
OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) |
(BRW_SURFACE_NULL << 29));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
if (intel->is_g4x || intel->gen >= 5)
OUT_BATCH(0);
if (intel->gen >= 6)
OUT_BATCH(0);
ADVANCE_BATCH();
} else if (!depth_irb && separate_stencil) {
/*
* There exists a separate stencil buffer but no depth buffer.
*
* The stencil buffer inherits most of its fields from
* 3DSTATE_DEPTH_BUFFER: namely the tile walk, surface type, width, and
* height.
*
* Enable the hiz bit because it and the separate stencil bit must have
* the same value. From Section 2.11.5.6.1.1 3DSTATE_DEPTH_BUFFER, Bit
* 1.21 "Separate Stencil Enable":
* [DevIL]: If this field is enabled, Hierarchical Depth Buffer
* Enable must also be enabled.
*
* [DevGT]: This field must be set to the same value (enabled or
* disabled) as Hierarchical Depth Buffer Enable
*
* The tiled bit must be set. From the Sandybridge PRM, Volume 2, Part 1,
* Section 7.5.5.1.1 3DSTATE_DEPTH_BUFFER, Bit 1.27 Tiled Surface:
* [DevGT+]: This field must be set to TRUE.
*/
assert(intel->has_separate_stencil);
BEGIN_BATCH(len);
OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (len - 2));
OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) |
(1 << 21) | /* separate stencil enable */
(1 << 22) | /* hiz enable */
(BRW_TILEWALK_YMAJOR << 26) |
(1 << 27) | /* tiled surface */
(BRW_SURFACE_2D << 29));
OUT_BATCH(0);
OUT_BATCH(((stencil_irb->Base.Base.Width + tile_x - 1) << 6) |
(stencil_irb->Base.Base.Height + tile_y - 1) << 19);
OUT_BATCH(0);
if (intel->is_g4x || intel->gen >= 5)
OUT_BATCH(tile_x | (tile_y << 16));
else
assert(tile_x == 0 && tile_y == 0);
if (intel->gen >= 6)
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
struct intel_region *region = depth_mt->region;
/* If using separate stencil, hiz must be enabled. */
assert(!separate_stencil || hiz_mt);
assert(intel->gen < 6 || region->tiling == I915_TILING_Y);
assert(!hiz_mt || region->tiling == I915_TILING_Y);
BEGIN_BATCH(len);
OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (len - 2));
OUT_BATCH((region->pitch - 1) |
(brw_depthbuffer_format(brw) << 18) |
((hiz_mt ? 1 : 0) << 21) | /* separate stencil enable */
((hiz_mt ? 1 : 0) << 22) | /* hiz enable */
(BRW_TILEWALK_YMAJOR << 26) |
((region->tiling != I915_TILING_NONE) << 27) |
(BRW_SURFACE_2D << 29));
OUT_RELOC(region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
brw->depthstencil.depth_offset);
OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) |
(((depth_irb->Base.Base.Width + tile_x) - 1) << 6) |
(((depth_irb->Base.Base.Height + tile_y) - 1) << 19));
OUT_BATCH(0);
if (intel->is_g4x || intel->gen >= 5)
OUT_BATCH(tile_x | (tile_y << 16));
else
assert(tile_x == 0 && tile_y == 0);
if (intel->gen >= 6)
OUT_BATCH(0);
ADVANCE_BATCH();
}
if (hiz_mt || separate_stencil) {
/*
* In the 3DSTATE_DEPTH_BUFFER batch emitted above, the 'separate
* stencil enable' and 'hiz enable' bits were set. Therefore we must
* emit 3DSTATE_HIER_DEPTH_BUFFER and 3DSTATE_STENCIL_BUFFER. Even if
* there is no stencil buffer, 3DSTATE_STENCIL_BUFFER must be emitted;
* failure to do so causes hangs on gen5 and a stall on gen6.
*/
/* Emit hiz buffer. */
if (hiz_mt) {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
OUT_BATCH(hiz_mt->region->pitch - 1);
OUT_RELOC(hiz_mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
brw->depthstencil.hiz_offset);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* Emit stencil buffer. */
if (separate_stencil) {
struct intel_region *region = stencil_mt->region;
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
/* The stencil buffer has quirky pitch requirements. From Vol 2a,
* 11.5.6.2.1 3DSTATE_STENCIL_BUFFER, field "Surface Pitch":
* The pitch must be set to 2x the value computed based on width, as
* the stencil buffer is stored with two rows interleaved.
*/
OUT_BATCH(2 * region->pitch - 1);
OUT_RELOC(region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
brw->depthstencil.stencil_offset);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
}
/*
* On Gen >= 6, emit clear params for safety. If using hiz, then clear
* params must be emitted.
*
* From Section 2.11.5.6.4.1 3DSTATE_CLEAR_PARAMS:
* 3DSTATE_CLEAR_PARAMS packet must follow the DEPTH_BUFFER_STATE packet
* when HiZ is enabled and the DEPTH_BUFFER_STATE changes.
*/
if (intel->gen >= 6 || hiz_mt) {
if (intel->gen == 6)
intel_emit_post_sync_nonzero_flush(intel);
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_CLEAR_PARAMS << 16 |
GEN5_DEPTH_CLEAR_VALID |
(2 - 2));
OUT_BATCH(depth_irb ? depth_irb->mt->depth_clear_value : 0);
ADVANCE_BATCH();
}
}
void
brw_emit_depth_stencil_hiz(struct brw_context *brw,
struct intel_mipmap_tree *depth_mt,
uint32_t depth_offset, uint32_t depthbuffer_format,
uint32_t depth_surface_type,
struct intel_mipmap_tree *stencil_mt,
struct intel_mipmap_tree *hiz_mt,
bool separate_stencil, uint32_t width,
uint32_t height, uint32_t tile_x, uint32_t tile_y)
{
struct intel_context *intel = &brw->intel;
/* Enable the hiz bit if we're doing separate stencil, because it and the
* separate stencil bit must have the same value. From Section 2.11.5.6.1.1
* 3DSTATE_DEPTH_BUFFER, Bit 1.21 "Separate Stencil Enable":
* [DevIL]: If this field is enabled, Hierarchical Depth Buffer
* Enable must also be enabled.
*
* [DevGT]: This field must be set to the same value (enabled or
* disabled) as Hierarchical Depth Buffer Enable
*/
bool enable_hiz_ss = hiz_mt || separate_stencil;
/* 3DSTATE_DEPTH_BUFFER, 3DSTATE_STENCIL_BUFFER are both
* non-pipelined state that will need the PIPE_CONTROL workaround.
*/
if (intel->gen == 6) {
intel_emit_post_sync_nonzero_flush(intel);
intel_emit_depth_stall_flushes(intel);
}
unsigned int len;
if (intel->gen >= 6)
len = 7;
else if (intel->is_g4x || intel->gen == 5)
len = 6;
else
len = 5;
BEGIN_BATCH(len);
OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (len - 2));
OUT_BATCH((depth_mt ? depth_mt->region->pitch - 1 : 0) |
(depthbuffer_format << 18) |
((enable_hiz_ss ? 1 : 0) << 21) | /* separate stencil enable */
((enable_hiz_ss ? 1 : 0) << 22) | /* hiz enable */
(BRW_TILEWALK_YMAJOR << 26) |
((depth_mt ? depth_mt->region->tiling != I915_TILING_NONE : 1)
<< 27) |
(depth_surface_type << 29));
if (depth_mt) {
OUT_RELOC(depth_mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
depth_offset);
} else {
OUT_BATCH(0);
}
OUT_BATCH(((width + tile_x - 1) << 6) |
((height + tile_y - 1) << 19));
OUT_BATCH(0);
if (intel->is_g4x || intel->gen >= 5)
OUT_BATCH(tile_x | (tile_y << 16));
else
assert(tile_x == 0 && tile_y == 0);
if (intel->gen >= 6)
OUT_BATCH(0);
ADVANCE_BATCH();
if (hiz_mt || separate_stencil) {
/*
* In the 3DSTATE_DEPTH_BUFFER batch emitted above, the 'separate
* stencil enable' and 'hiz enable' bits were set. Therefore we must
* emit 3DSTATE_HIER_DEPTH_BUFFER and 3DSTATE_STENCIL_BUFFER. Even if
* there is no stencil buffer, 3DSTATE_STENCIL_BUFFER must be emitted;
* failure to do so causes hangs on gen5 and a stall on gen6.
*/
/* Emit hiz buffer. */
if (hiz_mt) {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
OUT_BATCH(hiz_mt->region->pitch - 1);
OUT_RELOC(hiz_mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
brw->depthstencil.hiz_offset);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* Emit stencil buffer. */
if (separate_stencil) {
struct intel_region *region = stencil_mt->region;
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
/* The stencil buffer has quirky pitch requirements. From Vol 2a,
* 11.5.6.2.1 3DSTATE_STENCIL_BUFFER, field "Surface Pitch":
* The pitch must be set to 2x the value computed based on width, as
* the stencil buffer is stored with two rows interleaved.
*/
OUT_BATCH(2 * region->pitch - 1);
OUT_RELOC(region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
brw->depthstencil.stencil_offset);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
}
/*
* On Gen >= 6, emit clear params for safety. If using hiz, then clear
* params must be emitted.
*
* From Section 2.11.5.6.4.1 3DSTATE_CLEAR_PARAMS:
* 3DSTATE_CLEAR_PARAMS packet must follow the DEPTH_BUFFER_STATE packet
* when HiZ is enabled and the DEPTH_BUFFER_STATE changes.
*/
if (intel->gen >= 6 || hiz_mt) {
if (intel->gen == 6)
intel_emit_post_sync_nonzero_flush(intel);
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_CLEAR_PARAMS << 16 |
GEN5_DEPTH_CLEAR_VALID |
(2 - 2));
OUT_BATCH(depth_mt ? depth_mt->depth_clear_value : 0);
ADVANCE_BATCH();
}
}
static void
gen7_blorp_emit_depth_stencil_config(struct brw_context *brw,
const brw_blorp_params *params)
{
struct intel_context *intel = &brw->intel;
uint32_t draw_x = params->depth.x_offset;
uint32_t draw_y = params->depth.y_offset;
uint32_t tile_mask_x, tile_mask_y;
gen6_blorp_compute_tile_masks(params, &tile_mask_x, &tile_mask_y);
/* 3DSTATE_DEPTH_BUFFER */
{
uint32_t tile_x = draw_x & tile_mask_x;
uint32_t tile_y = draw_y & tile_mask_y;
uint32_t offset =
intel_region_get_aligned_offset(params->depth.mt->region,
draw_x & ~tile_mask_x,
draw_y & ~tile_mask_y, false);
/* According to the Sandy Bridge PRM, volume 2 part 1, pp326-327
* (3DSTATE_DEPTH_BUFFER dw5), in the documentation for "Depth
* Coordinate Offset X/Y":
*
* "The 3 LSBs of both offsets must be zero to ensure correct
* alignment"
*
* We have no guarantee that tile_x and tile_y are correctly aligned,
* since they are determined by the mipmap layout, which is only aligned
* to multiples of 4.
*
* So, to avoid hanging the GPU, just smash the low order 3 bits of
* tile_x and tile_y to 0. This is a temporary workaround until we come
* up with a better solution.
*/
tile_x &= ~7;
tile_y &= ~7;
intel_emit_depth_stall_flushes(intel);
BEGIN_BATCH(7);
OUT_BATCH(GEN7_3DSTATE_DEPTH_BUFFER << 16 | (7 - 2));
uint32_t pitch_bytes =
params->depth.mt->region->pitch * params->depth.mt->region->cpp;
OUT_BATCH((pitch_bytes - 1) |
params->depth_format << 18 |
1 << 22 | /* hiz enable */
1 << 28 | /* depth write */
BRW_SURFACE_2D << 29);
OUT_RELOC(params->depth.mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
offset);
OUT_BATCH((params->depth.width + tile_x - 1) << 4 |
(params->depth.height + tile_y - 1) << 18);
OUT_BATCH(0);
OUT_BATCH(tile_x |
tile_y << 16);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_HIER_DEPTH_BUFFER */
{
struct intel_region *hiz_region = params->depth.mt->hiz_mt->region;
uint32_t hiz_offset =
intel_region_get_aligned_offset(hiz_region,
draw_x & ~tile_mask_x,
(draw_y & ~tile_mask_y) / 2, false);
BEGIN_BATCH(3);
OUT_BATCH((GEN7_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
OUT_BATCH(hiz_region->pitch * hiz_region->cpp - 1);
OUT_RELOC(hiz_region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
hiz_offset);
ADVANCE_BATCH();
}
/* 3DSTATE_STENCIL_BUFFER */
{
BEGIN_BATCH(3);
OUT_BATCH((GEN7_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
}
/**
* Helper function to emit depth related command packets.
*/
static void
emit_depth_packets(struct brw_context *brw,
struct intel_mipmap_tree *depth_mt,
uint32_t depthbuffer_format,
uint32_t depth_surface_type,
bool depth_writable,
struct intel_mipmap_tree *stencil_mt,
bool stencil_writable,
uint32_t stencil_offset,
bool hiz,
uint32_t width,
uint32_t height,
uint32_t depth,
uint32_t lod,
uint32_t min_array_element)
{
/* Skip repeated NULL depth/stencil emits (think 2D rendering). */
if (!depth_mt && !stencil_mt && brw->no_depth_or_stencil) {
assert(brw->hw_ctx);
return;
}
intel_emit_depth_stall_flushes(brw);
/* _NEW_BUFFERS, _NEW_DEPTH, _NEW_STENCIL */
BEGIN_BATCH(8);
OUT_BATCH(GEN7_3DSTATE_DEPTH_BUFFER << 16 | (8 - 2));
OUT_BATCH(depth_surface_type << 29 |
(depth_writable ? (1 << 28) : 0) |
(stencil_mt != NULL && stencil_writable) << 27 |
(hiz ? 1 : 0) << 22 |
depthbuffer_format << 18 |
(depth_mt ? depth_mt->region->pitch - 1 : 0));
if (depth_mt) {
OUT_RELOC64(depth_mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, 0);
} else {
OUT_BATCH(0);
OUT_BATCH(0);
}
OUT_BATCH(((width - 1) << 4) | ((height - 1) << 18) | lod);
OUT_BATCH(((depth - 1) << 21) | (min_array_element << 10) | BDW_MOCS_WB);
OUT_BATCH(0);
OUT_BATCH(depth_mt ? depth_mt->qpitch >> 2 : 0);
ADVANCE_BATCH();
if (!hiz) {
BEGIN_BATCH(5);
OUT_BATCH(GEN7_3DSTATE_HIER_DEPTH_BUFFER << 16 | (5 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(5);
OUT_BATCH(GEN7_3DSTATE_HIER_DEPTH_BUFFER << 16 | (5 - 2));
OUT_BATCH((depth_mt->hiz_mt->region->pitch - 1) | BDW_MOCS_WB << 25);
OUT_RELOC64(depth_mt->hiz_mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, 0);
OUT_BATCH(depth_mt->hiz_mt->qpitch >> 2);
ADVANCE_BATCH();
}
if (stencil_mt == NULL) {
BEGIN_BATCH(5);
OUT_BATCH(GEN7_3DSTATE_STENCIL_BUFFER << 16 | (5 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(5);
OUT_BATCH(GEN7_3DSTATE_STENCIL_BUFFER << 16 | (5 - 2));
/* The stencil buffer has quirky pitch requirements. From the Graphics
* BSpec: vol2a.11 3D Pipeline Windower > Early Depth/Stencil Processing
* > Depth/Stencil Buffer State > 3DSTATE_STENCIL_BUFFER [DevIVB+],
* field "Surface Pitch":
*
* The pitch must be set to 2x the value computed based on width, as
* the stencil buffer is stored with two rows interleaved.
*
* (Note that it is not 100% clear whether this intended to apply to
* Gen7; the BSpec flags this comment as "DevILK,DevSNB" (which would
* imply that it doesn't), however the comment appears on a "DevIVB+"
* page (which would imply that it does). Experiments with the hardware
* indicate that it does.
*/
OUT_BATCH(HSW_STENCIL_ENABLED | BDW_MOCS_WB << 22 |
(2 * stencil_mt->region->pitch - 1));
OUT_RELOC64(stencil_mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
stencil_offset);
OUT_BATCH(stencil_mt ? stencil_mt->qpitch >> 2 : 0);
ADVANCE_BATCH();
}
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_CLEAR_PARAMS << 16 | (3 - 2));
OUT_BATCH(depth_mt ? depth_mt->depth_clear_value : 0);
OUT_BATCH(1);
ADVANCE_BATCH();
brw->no_depth_or_stencil = !depth_mt && !stencil_mt;
}
static void emit_depthbuffer(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
struct gl_framebuffer *fb = ctx->DrawBuffer;
/* _NEW_BUFFERS */
struct intel_renderbuffer *depth_irb = intel_get_renderbuffer(fb, BUFFER_DEPTH);
struct intel_renderbuffer *stencil_irb = intel_get_renderbuffer(fb, BUFFER_STENCIL);
struct intel_mipmap_tree *stencil_mt = NULL;
struct intel_region *hiz_region = NULL;
unsigned int len;
bool separate_stencil = false;
/* Amount by which drawing should be offset in order to draw to the
* appropriate miplevel/zoffset/cubeface. We will extract these values
* from depth_irb or stencil_irb once we determine which is present.
*/
uint32_t draw_x = 0, draw_y = 0;
/* Masks used to determine how much of the draw_x and draw_y offsets should
* be performed using the fine adjustment of "depth coordinate offset X/Y"
* (dw5 of 3DSTATE_DEPTH_BUFFER). Any remaining coarse adjustment will be
* performed by changing the base addresses of the buffers.
*
* Since the HiZ, depth, and stencil buffers all use the same "depth
* coordinate offset X/Y" values, we need to make sure that the coarse
* adjustment will be possible to apply to all three buffers. Since coarse
* adjustment can only be applied in multiples of the tile size, we will OR
* together the tile masks of all the buffers to determine which offsets to
* perform as fine adjustments.
*/
uint32_t tile_mask_x = 0, tile_mask_y = 0;
if (depth_irb) {
intel_region_get_tile_masks(depth_irb->mt->region,
&tile_mask_x, &tile_mask_y);
}
if (depth_irb &&
depth_irb->mt &&
depth_irb->mt->hiz_mt) {
hiz_region = depth_irb->mt->hiz_mt->region;
uint32_t hiz_tile_mask_x, hiz_tile_mask_y;
intel_region_get_tile_masks(hiz_region,
&hiz_tile_mask_x, &hiz_tile_mask_y);
/* Each HiZ row represents 2 rows of pixels */
hiz_tile_mask_y = hiz_tile_mask_y << 1 | 1;
tile_mask_x |= hiz_tile_mask_x;
tile_mask_y |= hiz_tile_mask_y;
}
/* 3DSTATE_DEPTH_BUFFER, 3DSTATE_STENCIL_BUFFER are both
* non-pipelined state that will need the PIPE_CONTROL workaround.
*/
if (intel->gen == 6) {
intel_emit_post_sync_nonzero_flush(intel);
intel_emit_depth_stall_flushes(intel);
}
/* Find the real separate stencil mt if present. */
if (stencil_irb) {
stencil_mt = stencil_irb->mt;
if (stencil_mt->stencil_mt)
stencil_mt = stencil_mt->stencil_mt;
if (stencil_mt->format == MESA_FORMAT_S8) {
separate_stencil = true;
/* Separate stencil buffer uses 64x64 tiles. */
tile_mask_x |= 63;
tile_mask_y |= 63;
} else {
uint32_t stencil_tile_mask_x, stencil_tile_mask_y;
intel_region_get_tile_masks(stencil_mt->region,
&stencil_tile_mask_x,
&stencil_tile_mask_y);
tile_mask_x |= stencil_tile_mask_x;
tile_mask_y |= stencil_tile_mask_y;
}
}
/* If there's a packed depth/stencil bound to stencil only, we need to
* emit the packed depth/stencil buffer packet.
*/
if (!depth_irb && stencil_irb && !separate_stencil)
depth_irb = stencil_irb;
if (intel->gen >= 6)
len = 7;
else if (intel->is_g4x || intel->gen == 5)
len = 6;
else
len = 5;
if (!depth_irb && !separate_stencil) {
BEGIN_BATCH(len);
OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (len - 2));
OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) |
(BRW_SURFACE_NULL << 29));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
if (intel->is_g4x || intel->gen >= 5)
OUT_BATCH(0);
if (intel->gen >= 6)
OUT_BATCH(0);
ADVANCE_BATCH();
} else if (!depth_irb && separate_stencil) {
uint32_t tile_x, tile_y;
/*
* There exists a separate stencil buffer but no depth buffer.
*
* The stencil buffer inherits most of its fields from
* 3DSTATE_DEPTH_BUFFER: namely the tile walk, surface type, width, and
* height.
*
* Enable the hiz bit because it and the separate stencil bit must have
* the same value. From Section 2.11.5.6.1.1 3DSTATE_DEPTH_BUFFER, Bit
* 1.21 "Separate Stencil Enable":
* [DevIL]: If this field is enabled, Hierarchical Depth Buffer
* Enable must also be enabled.
*
* [DevGT]: This field must be set to the same value (enabled or
* disabled) as Hierarchical Depth Buffer Enable
*
* The tiled bit must be set. From the Sandybridge PRM, Volume 2, Part 1,
* Section 7.5.5.1.1 3DSTATE_DEPTH_BUFFER, Bit 1.27 Tiled Surface:
* [DevGT+]: This field must be set to TRUE.
*/
assert(intel->has_separate_stencil);
draw_x = stencil_irb->draw_x;
draw_y = stencil_irb->draw_y;
tile_x = draw_x & tile_mask_x;
tile_y = draw_y & tile_mask_y;
/* According to the Sandy Bridge PRM, volume 2 part 1, pp326-327
* (3DSTATE_DEPTH_BUFFER dw5), in the documentation for "Depth
* Coordinate Offset X/Y":
*
* "The 3 LSBs of both offsets must be zero to ensure correct
* alignment"
*
* We have no guarantee that tile_x and tile_y are correctly aligned,
* since they are determined by the mipmap layout, which is only aligned
* to multiples of 4.
*
* So, to avoid hanging the GPU, just smash the low order 3 bits of
* tile_x and tile_y to 0. This is a temporary workaround until we come
* up with a better solution.
*/
tile_x &= ~7;
tile_y &= ~7;
BEGIN_BATCH(len);
OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (len - 2));
OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) |
(1 << 21) | /* separate stencil enable */
(1 << 22) | /* hiz enable */
(BRW_TILEWALK_YMAJOR << 26) |
(1 << 27) | /* tiled surface */
(BRW_SURFACE_2D << 29));
OUT_BATCH(0);
OUT_BATCH(((stencil_irb->Base.Base.Width + tile_x - 1) << 6) |
(stencil_irb->Base.Base.Height + tile_y - 1) << 19);
OUT_BATCH(0);
if (intel->is_g4x || intel->gen >= 5)
OUT_BATCH(tile_x | (tile_y << 16));
else
assert(tile_x == 0 && tile_y == 0);
if (intel->gen >= 6)
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
struct intel_region *region = depth_irb->mt->region;
uint32_t tile_x, tile_y, offset;
/* If using separate stencil, hiz must be enabled. */
assert(!separate_stencil || hiz_region);
assert(intel->gen < 6 || region->tiling == I915_TILING_Y);
assert(!hiz_region || region->tiling == I915_TILING_Y);
draw_x = depth_irb->draw_x;
draw_y = depth_irb->draw_y;
tile_x = draw_x & tile_mask_x;
tile_y = draw_y & tile_mask_y;
/* According to the Sandy Bridge PRM, volume 2 part 1, pp326-327
* (3DSTATE_DEPTH_BUFFER dw5), in the documentation for "Depth
* Coordinate Offset X/Y":
*
* "The 3 LSBs of both offsets must be zero to ensure correct
* alignment"
*
* We have no guarantee that tile_x and tile_y are correctly aligned,
* since they are determined by the mipmap layout, which is only aligned
* to multiples of 4.
*
* So, to avoid hanging the GPU, just smash the low order 3 bits of
* tile_x and tile_y to 0. This is a temporary workaround until we come
* up with a better solution.
*/
tile_x &= ~7;
tile_y &= ~7;
offset = intel_region_get_aligned_offset(region,
draw_x & ~tile_mask_x,
draw_y & ~tile_mask_y);
BEGIN_BATCH(len);
OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (len - 2));
OUT_BATCH(((region->pitch * region->cpp) - 1) |
(brw_depthbuffer_format(brw) << 18) |
((hiz_region ? 1 : 0) << 21) | /* separate stencil enable */
((hiz_region ? 1 : 0) << 22) | /* hiz enable */
(BRW_TILEWALK_YMAJOR << 26) |
((region->tiling != I915_TILING_NONE) << 27) |
(BRW_SURFACE_2D << 29));
OUT_RELOC(region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
offset);
OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) |
(((depth_irb->Base.Base.Width + tile_x) - 1) << 6) |
(((depth_irb->Base.Base.Height + tile_y) - 1) << 19));
OUT_BATCH(0);
if (intel->is_g4x || intel->gen >= 5)
OUT_BATCH(tile_x | (tile_y << 16));
else
assert(tile_x == 0 && tile_y == 0);
if (intel->gen >= 6)
OUT_BATCH(0);
ADVANCE_BATCH();
}
if (hiz_region || separate_stencil) {
/*
* In the 3DSTATE_DEPTH_BUFFER batch emitted above, the 'separate
* stencil enable' and 'hiz enable' bits were set. Therefore we must
* emit 3DSTATE_HIER_DEPTH_BUFFER and 3DSTATE_STENCIL_BUFFER. Even if
* there is no stencil buffer, 3DSTATE_STENCIL_BUFFER must be emitted;
* failure to do so causes hangs on gen5 and a stall on gen6.
*/
/* Emit hiz buffer. */
if (hiz_region) {
uint32_t hiz_offset =
intel_region_get_aligned_offset(hiz_region,
draw_x & ~tile_mask_x,
(draw_y & ~tile_mask_y) / 2);
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
OUT_BATCH(hiz_region->pitch * hiz_region->cpp - 1);
OUT_RELOC(hiz_region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
hiz_offset);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* Emit stencil buffer. */
if (separate_stencil) {
struct intel_region *region = stencil_mt->region;
/* Note: we can't compute the stencil offset using
* intel_region_get_aligned_offset(), because stencil_region claims
* that the region is untiled; in fact it's W tiled.
*/
uint32_t stencil_offset =
(draw_y & ~tile_mask_y) * region->pitch +
(draw_x & ~tile_mask_x) * 64;
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
/* The stencil buffer has quirky pitch requirements. From Vol 2a,
* 11.5.6.2.1 3DSTATE_STENCIL_BUFFER, field "Surface Pitch":
* The pitch must be set to 2x the value computed based on width, as
* the stencil buffer is stored with two rows interleaved.
*/
OUT_BATCH(2 * region->pitch * region->cpp - 1);
OUT_RELOC(region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
stencil_offset);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
}
/*
* On Gen >= 6, emit clear params for safety. If using hiz, then clear
* params must be emitted.
*
* From Section 2.11.5.6.4.1 3DSTATE_CLEAR_PARAMS:
* 3DSTATE_CLEAR_PARAMS packet must follow the DEPTH_BUFFER_STATE packet
* when HiZ is enabled and the DEPTH_BUFFER_STATE changes.
*/
if (intel->gen >= 6 || hiz_region) {
if (intel->gen == 6)
intel_emit_post_sync_nonzero_flush(intel);
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_CLEAR_PARAMS << 16 |
GEN5_DEPTH_CLEAR_VALID |
(2 - 2));
OUT_BATCH(depth_irb ? depth_irb->mt->depth_clear_value : 0);
ADVANCE_BATCH();
}
}
/**
* \copydoc gen6_hiz_exec()
*/
static void
gen7_hiz_exec(struct intel_context *intel,
struct intel_mipmap_tree *mt,
unsigned int level,
unsigned int layer,
enum gen6_hiz_op op)
{
struct gl_context *ctx = &intel->ctx;
struct brw_context *brw = brw_context(ctx);
assert(op != GEN6_HIZ_OP_DEPTH_CLEAR); /* Not implemented yet. */
assert(mt->hiz_mt != NULL);
intel_miptree_check_level_layer(mt, level, layer);
uint32_t depth_format;
switch (mt->format) {
case MESA_FORMAT_Z16: depth_format = BRW_DEPTHFORMAT_D16_UNORM; break;
case MESA_FORMAT_Z32_FLOAT: depth_format = BRW_DEPTHFORMAT_D32_FLOAT; break;
case MESA_FORMAT_X8_Z24: depth_format = BRW_DEPTHFORMAT_D24_UNORM_X8_UINT; break;
default: assert(0); break;
}
gen6_hiz_emit_batch_head(brw);
gen6_hiz_emit_vertices(brw, mt, level, layer);
/* 3DSTATE_URB_VS
* 3DSTATE_URB_HS
* 3DSTATE_URB_DS
* 3DSTATE_URB_GS
*
* If the 3DSTATE_URB_VS is emitted, than the others must be also. From the
* BSpec, Volume 2a "3D Pipeline Overview", Section 1.7.1 3DSTATE_URB_VS:
* 3DSTATE_URB_HS, 3DSTATE_URB_DS, and 3DSTATE_URB_GS must also be
* programmed in order for the programming of this state to be
* valid.
*/
{
/* The minimum valid value is 32. See 3DSTATE_URB_VS,
* Dword 1.15:0 "VS Number of URB Entries".
*/
int num_vs_entries = 32;
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_URB_VS << 16 | (2 - 2));
OUT_BATCH(1 << GEN7_URB_ENTRY_SIZE_SHIFT |
0 << GEN7_URB_STARTING_ADDRESS_SHIFT |
num_vs_entries);
ADVANCE_BATCH();
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_URB_GS << 16 | (2 - 2));
OUT_BATCH(0);
ADVANCE_BATCH();
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_URB_HS << 16 | (2 - 2));
OUT_BATCH(0);
ADVANCE_BATCH();
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_URB_DS << 16 | (2 - 2));
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_DEPTH_STENCIL_STATE_POINTERS
*
* The offset is relative to CMD_STATE_BASE_ADDRESS.DynamicStateBaseAddress.
*/
{
uint32_t depthstencil_offset;
gen6_hiz_emit_depth_stencil_state(brw, op, &depthstencil_offset);
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_DEPTH_STENCIL_STATE_POINTERS << 16 | (2 - 2));
OUT_BATCH(depthstencil_offset | 1);
ADVANCE_BATCH();
}
/* 3DSTATE_VS
*
* Disable vertex shader.
*/
{
BEGIN_BATCH(6);
OUT_BATCH(_3DSTATE_VS << 16 | (6 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_HS
*
* Disable the hull shader.
*/
{
BEGIN_BATCH(7);
OUT_BATCH(_3DSTATE_HS << 16 | (7 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_TE
*
* Disable the tesselation engine.
*/
{
BEGIN_BATCH(4);
OUT_BATCH(_3DSTATE_TE << 16 | (4 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_DS
*
* Disable the domain shader.
*/
{
BEGIN_BATCH(6);
OUT_BATCH(_3DSTATE_DS << 16 | (6 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_GS
*
* Disable the geometry shader.
*/
{
BEGIN_BATCH(7);
OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_STREAMOUT
*
* Disable streamout.
*/
{
BEGIN_BATCH(3);
OUT_BATCH(_3DSTATE_STREAMOUT << 16 | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_CLIP
*
* Disable the clipper.
*
* The HiZ op emits a rectangle primitive, which requires clipping to
* be disabled. From page 10 of the Sandy Bridge PRM Volume 2 Part 1
* Section 1.3 "3D Primitives Overview":
* RECTLIST:
* Either the CLIP unit should be DISABLED, or the CLIP unit's Clip
* Mode should be set to a value other than CLIPMODE_NORMAL.
*
* Also disable perspective divide. This doesn't change the clipper's
* output, but does spare a few electrons.
*/
{
BEGIN_BATCH(4);
OUT_BATCH(_3DSTATE_CLIP << 16 | (4 - 2));
OUT_BATCH(0);
OUT_BATCH(GEN6_CLIP_PERSPECTIVE_DIVIDE_DISABLE);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_SF
*
* Disable ViewportTransformEnable (dw1.1)
*
* From the SandyBridge PRM, Volume 2, Part 1, Section 1.3, "3D
* Primitives Overview":
* RECTLIST: Viewport Mapping must be DISABLED (as is typical with the
* use of screen- space coordinates).
*
* A solid rectangle must be rendered, so set FrontFaceFillMode (dw1.6:5)
* and BackFaceFillMode (dw1.4:3) to SOLID(0).
*
* From the Sandy Bridge PRM, Volume 2, Part 1, Section
* 6.4.1.1 3DSTATE_SF, Field FrontFaceFillMode:
* SOLID: Any triangle or rectangle object found to be front-facing
* is rendered as a solid object. This setting is required when
* (rendering rectangle (RECTLIST) objects.
*/
{
BEGIN_BATCH(7);
OUT_BATCH(_3DSTATE_SF << 16 | (7 - 2));
OUT_BATCH(depth_format << GEN7_SF_DEPTH_BUFFER_SURFACE_FORMAT_SHIFT);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_SBE */
{
BEGIN_BATCH(14);
OUT_BATCH(_3DSTATE_SBE << 16 | (14 - 2));
OUT_BATCH((1 - 1) << GEN7_SBE_NUM_OUTPUTS_SHIFT | /* only position */
1 << GEN7_SBE_URB_ENTRY_READ_LENGTH_SHIFT |
0 << GEN7_SBE_URB_ENTRY_READ_OFFSET_SHIFT);
for (int i = 0; i < 12; ++i)
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_WM
*
* Disable PS thread dispatch (dw1.29) and enable the HiZ op.
*/
{
uint32_t dw1 = 0;
switch (op) {
case GEN6_HIZ_OP_DEPTH_CLEAR:
assert(!"not implemented");
dw1 |= GEN7_WM_DEPTH_CLEAR;
break;
case GEN6_HIZ_OP_DEPTH_RESOLVE:
dw1 |= GEN7_WM_DEPTH_RESOLVE;
break;
case GEN6_HIZ_OP_HIZ_RESOLVE:
dw1 |= GEN7_WM_HIERARCHICAL_DEPTH_RESOLVE;
break;
default:
assert(0);
break;
}
BEGIN_BATCH(3);
OUT_BATCH(_3DSTATE_WM << 16 | (3 - 2));
OUT_BATCH(dw1);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_PS
*
* Pixel shader dispatch is disabled above in 3DSTATE_WM, dw1.29. Despite
* that, thread dispatch info must still be specified.
* - Maximum Number of Threads (dw4.24:31) must be nonzero, as the BSpec
* states that the valid range for this field is [0x3, 0x2f].
* - A dispatch mode must be given; that is, at least one of the
* "N Pixel Dispatch Enable" (N=8,16,32) fields must be set. This was
* discovered through simulator error messages.
*/
{
BEGIN_BATCH(8);
OUT_BATCH(_3DSTATE_PS << 16 | (8 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(((brw->max_wm_threads - 1) << IVB_PS_MAX_THREADS_SHIFT) |
GEN7_PS_32_DISPATCH_ENABLE);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_DEPTH_BUFFER */
{
uint32_t width = mt->level[level].width;
uint32_t height = mt->level[level].height;
uint32_t tile_x;
uint32_t tile_y;
uint32_t offset;
{
/* Construct a dummy renderbuffer just to extract tile offsets. */
struct intel_renderbuffer rb;
rb.mt = mt;
rb.mt_level = level;
rb.mt_layer = layer;
intel_renderbuffer_set_draw_offset(&rb);
offset = intel_renderbuffer_tile_offsets(&rb, &tile_x, &tile_y);
}
intel_emit_depth_stall_flushes(intel);
BEGIN_BATCH(7);
OUT_BATCH(GEN7_3DSTATE_DEPTH_BUFFER << 16 | (7 - 2));
OUT_BATCH(((mt->region->pitch * mt->region->cpp) - 1) |
depth_format << 18 |
1 << 22 | /* hiz enable */
1 << 28 | /* depth write */
BRW_SURFACE_2D << 29);
OUT_RELOC(mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
offset);
OUT_BATCH((width + tile_x - 1) << 4 |
(height + tile_y - 1) << 18);
OUT_BATCH(0);
OUT_BATCH(tile_x |
tile_y << 16);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_HIER_DEPTH_BUFFER */
{
struct intel_region *hiz_region = mt->hiz_mt->region;
BEGIN_BATCH(3);
OUT_BATCH((GEN7_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
OUT_BATCH(hiz_region->pitch * hiz_region->cpp - 1);
OUT_RELOC(hiz_region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
0);
ADVANCE_BATCH();
}
/* 3DSTATE_STENCIL_BUFFER */
{
BEGIN_BATCH(3);
OUT_BATCH((GEN7_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_CLEAR_PARAMS
*
* From the BSpec, Volume 2a.11 Windower, Section 1.5.6.3.2
* 3DSTATE_CLEAR_PARAMS:
* [DevIVB] 3DSTATE_CLEAR_PARAMS must always be programmed in the along
* with the other Depth/Stencil state commands(i.e. 3DSTATE_DEPTH_BUFFER,
* 3DSTATE_STENCIL_BUFFER, or 3DSTATE_HIER_DEPTH_BUFFER).
*/
{
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_CLEAR_PARAMS << 16 | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DSTATE_DRAWING_RECTANGLE */
{
BEGIN_BATCH(4);
OUT_BATCH(_3DSTATE_DRAWING_RECTANGLE << 16 | (4 - 2));
OUT_BATCH(0);
OUT_BATCH(((mt->level[level].width - 1) & 0xffff) |
((mt->level[level].height - 1) << 16));
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* 3DPRIMITIVE */
{
BEGIN_BATCH(7);
OUT_BATCH(CMD_3D_PRIM << 16 | (7 - 2));
OUT_BATCH(GEN7_3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL |
_3DPRIM_RECTLIST);
OUT_BATCH(3); /* vertex count per instance */
OUT_BATCH(0);
OUT_BATCH(1); /* instance count */
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* See comments above at first invocation of intel_flush() in
* gen6_hiz_emit_batch_head().
*/
intel_flush(ctx);
/* Be safe. */
brw->state.dirty.brw = ~0;
brw->state.dirty.cache = ~0;
}
void
gen7_emit_depth_stencil_hiz(struct brw_context *brw,
struct intel_mipmap_tree *depth_mt,
uint32_t depth_offset, uint32_t depthbuffer_format,
uint32_t depth_surface_type,
struct intel_mipmap_tree *stencil_mt,
struct intel_mipmap_tree *hiz_mt,
bool separate_stencil, uint32_t width,
uint32_t height, uint32_t tile_x, uint32_t tile_y)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
intel_emit_depth_stall_flushes(intel);
/* _NEW_DEPTH, _NEW_STENCIL */
BEGIN_BATCH(7);
OUT_BATCH(GEN7_3DSTATE_DEPTH_BUFFER << 16 | (7 - 2));
OUT_BATCH((depth_mt ? depth_mt->region->pitch - 1 : 0) |
(depthbuffer_format << 18) |
((hiz_mt ? 1 : 0) << 22) |
((stencil_mt != NULL && ctx->Stencil._WriteEnabled) << 27) |
((ctx->Depth.Mask != 0) << 28) |
(depth_surface_type << 29));
if (depth_mt) {
OUT_RELOC(depth_mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
depth_offset);
} else {
OUT_BATCH(0);
}
OUT_BATCH(((width + tile_x - 1) << 4) |
((height + tile_y - 1) << 18));
OUT_BATCH(0);
OUT_BATCH(tile_x | (tile_y << 16));
OUT_BATCH(0);
ADVANCE_BATCH();
if (hiz_mt == NULL) {
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_HIER_DEPTH_BUFFER << 16 | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_HIER_DEPTH_BUFFER << 16 | (3 - 2));
OUT_BATCH(hiz_mt->region->pitch - 1);
OUT_RELOC(hiz_mt->region->bo,
I915_GEM_DOMAIN_RENDER,
I915_GEM_DOMAIN_RENDER,
brw->depthstencil.hiz_offset);
ADVANCE_BATCH();
}
if (stencil_mt == NULL) {
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_STENCIL_BUFFER << 16 | (3 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
const int enabled = intel->is_haswell ? HSW_STENCIL_ENABLED : 0;
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_STENCIL_BUFFER << 16 | (3 - 2));
/* The stencil buffer has quirky pitch requirements. From the Graphics
* BSpec: vol2a.11 3D Pipeline Windower > Early Depth/Stencil Processing
* > Depth/Stencil Buffer State > 3DSTATE_STENCIL_BUFFER [DevIVB+],
* field "Surface Pitch":
*
* The pitch must be set to 2x the value computed based on width, as
* the stencil buffer is stored with two rows interleaved.
*
* (Note that it is not 100% clear whether this intended to apply to
* Gen7; the BSpec flags this comment as "DevILK,DevSNB" (which would
* imply that it doesn't), however the comment appears on a "DevIVB+"
* page (which would imply that it does). Experiments with the hardware
* indicate that it does.
*/
OUT_BATCH(enabled |
(2 * stencil_mt->region->pitch - 1));
OUT_RELOC(stencil_mt->region->bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
brw->depthstencil.stencil_offset);
ADVANCE_BATCH();
}
BEGIN_BATCH(3);
OUT_BATCH(GEN7_3DSTATE_CLEAR_PARAMS << 16 | (3 - 2));
OUT_BATCH(depth_mt ? depth_mt->depth_clear_value : 0);
OUT_BATCH(1);
ADVANCE_BATCH();
}