#include "main/teximage.h"
#include "main/fbobject.h"
-
-#include "glsl/ralloc.h"
+#include "main/renderbuffer.h"
#include "intel_fbo.h"
#include "brw_blorp.h"
#include "brw_context.h"
-#include "brw_eu.h"
+#include "brw_blorp_blit_eu.h"
#include "brw_state.h"
+#include "brw_meta_util.h"
-
-/**
- * Helper function for handling mirror image blits.
- *
- * If coord0 > coord1, swap them and invert the "mirror" boolean.
- */
-static inline void
-fixup_mirroring(bool &mirror, GLint &coord0, GLint &coord1)
-{
- if (coord0 > coord1) {
- mirror = !mirror;
- GLint tmp = coord0;
- coord0 = coord1;
- coord1 = tmp;
- }
-}
-
-
-/**
- * Adjust {src,dst}_x{0,1} to account for clipping and scissoring of
- * destination coordinates.
- *
- * Return true if there is still blitting to do, false if all pixels got
- * rejected by the clip and/or scissor.
- *
- * For clarity, the nomenclature of this function assumes we are clipping and
- * scissoring the X coordinate; the exact same logic applies for Y
- * coordinates.
- *
- * Note: this function may also be used to account for clipping of source
- * coordinates, by swapping the roles of src and dst.
- */
-static inline bool
-clip_or_scissor(bool mirror, GLint &src_x0, GLint &src_x1, GLint &dst_x0,
- GLint &dst_x1, GLint fb_xmin, GLint fb_xmax)
-{
- /* If we are going to scissor everything away, stop. */
- if (!(fb_xmin < fb_xmax &&
- dst_x0 < fb_xmax &&
- fb_xmin < dst_x1 &&
- dst_x0 < dst_x1)) {
- return false;
- }
-
- /* Clip the destination rectangle, and keep track of how many pixels we
- * clipped off of the left and right sides of it.
- */
- GLint pixels_clipped_left = 0;
- GLint pixels_clipped_right = 0;
- if (dst_x0 < fb_xmin) {
- pixels_clipped_left = fb_xmin - dst_x0;
- dst_x0 = fb_xmin;
- }
- if (fb_xmax < dst_x1) {
- pixels_clipped_right = dst_x1 - fb_xmax;
- dst_x1 = fb_xmax;
- }
-
- /* If we are mirrored, then before applying pixels_clipped_{left,right} to
- * the source coordinates, we need to flip them to account for the
- * mirroring.
- */
- if (mirror) {
- GLint tmp = pixels_clipped_left;
- pixels_clipped_left = pixels_clipped_right;
- pixels_clipped_right = tmp;
- }
-
- /* Adjust the source rectangle to remove the pixels corresponding to those
- * that were clipped/scissored out of the destination rectangle.
- */
- src_x0 += pixels_clipped_left;
- src_x1 -= pixels_clipped_right;
-
- return true;
-}
-
+#define FILE_DEBUG_FLAG DEBUG_BLORP
static struct intel_mipmap_tree *
find_miptree(GLbitfield buffer_bit, struct intel_renderbuffer *irb)
return mt;
}
+
+/**
+ * Note: if the src (or dst) is a 2D multisample array texture on Gen7+ using
+ * INTEL_MSAA_LAYOUT_UMS or INTEL_MSAA_LAYOUT_CMS, src_layer (dst_layer) is
+ * the physical layer holding sample 0. So, for example, if
+ * src_mt->num_samples == 4, then logical layer n corresponds to src_layer ==
+ * 4*n.
+ */
void
-brw_blorp_blit_miptrees(struct intel_context *intel,
+brw_blorp_blit_miptrees(struct brw_context *brw,
struct intel_mipmap_tree *src_mt,
unsigned src_level, unsigned src_layer,
+ mesa_format src_format,
struct intel_mipmap_tree *dst_mt,
unsigned dst_level, unsigned dst_layer,
- int src_x0, int src_y0,
- int dst_x0, int dst_y0,
- int dst_x1, int dst_y1,
- bool mirror_x, bool mirror_y)
+ mesa_format dst_format,
+ float src_x0, float src_y0,
+ float src_x1, float src_y1,
+ float dst_x0, float dst_y0,
+ float dst_x1, float dst_y1,
+ GLenum filter, bool mirror_x, bool mirror_y)
{
- brw_blorp_blit_params params(brw_context(&intel->ctx),
- src_mt, src_level, src_layer,
- dst_mt, dst_level, dst_layer,
+ /* Get ready to blit. This includes depth resolving the src and dst
+ * buffers if necessary. Note: it's not necessary to do a color resolve on
+ * the destination buffer because we use the standard render path to render
+ * to destination color buffers, and the standard render path is
+ * fast-color-aware.
+ */
+ intel_miptree_resolve_color(brw, src_mt);
+ intel_miptree_slice_resolve_depth(brw, src_mt, src_level, src_layer);
+ intel_miptree_slice_resolve_depth(brw, dst_mt, dst_level, dst_layer);
+
+ DBG("%s from %dx %s mt %p %d %d (%f,%f) (%f,%f)"
+ "to %dx %s mt %p %d %d (%f,%f) (%f,%f) (flip %d,%d)\n",
+ __func__,
+ src_mt->num_samples, _mesa_get_format_name(src_mt->format), src_mt,
+ src_level, src_layer, src_x0, src_y0, src_x1, src_y1,
+ dst_mt->num_samples, _mesa_get_format_name(dst_mt->format), dst_mt,
+ dst_level, dst_layer, dst_x0, dst_y0, dst_x1, dst_y1,
+ mirror_x, mirror_y);
+
+ brw_blorp_blit_params params(brw,
+ src_mt, src_level, src_layer, src_format,
+ dst_mt, dst_level, dst_layer, dst_format,
src_x0, src_y0,
+ src_x1, src_y1,
dst_x0, dst_y0,
dst_x1, dst_y1,
- mirror_x, mirror_y);
- brw_blorp_exec(intel, ¶ms);
+ filter, mirror_x, mirror_y);
+ brw_blorp_exec(brw, ¶ms);
+
+ intel_miptree_slice_set_needs_hiz_resolve(dst_mt, dst_level, dst_layer);
}
static void
-do_blorp_blit(struct intel_context *intel, GLbitfield buffer_bit,
- struct intel_renderbuffer *src_irb,
- struct intel_renderbuffer *dst_irb,
- GLint srcX0, GLint srcY0,
- GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
- bool mirror_x, bool mirror_y)
+do_blorp_blit(struct brw_context *brw, GLbitfield buffer_bit,
+ struct intel_renderbuffer *src_irb, mesa_format src_format,
+ struct intel_renderbuffer *dst_irb, mesa_format dst_format,
+ GLfloat srcX0, GLfloat srcY0, GLfloat srcX1, GLfloat srcY1,
+ GLfloat dstX0, GLfloat dstY0, GLfloat dstX1, GLfloat dstY1,
+ GLenum filter, bool mirror_x, bool mirror_y)
{
/* Find source/dst miptrees */
struct intel_mipmap_tree *src_mt = find_miptree(buffer_bit, src_irb);
struct intel_mipmap_tree *dst_mt = find_miptree(buffer_bit, dst_irb);
- /* Get ready to blit. This includes depth resolving the src and dst
- * buffers if necessary.
- */
- intel_renderbuffer_resolve_depth(intel, src_irb);
- intel_renderbuffer_resolve_depth(intel, dst_irb);
-
/* Do the blit */
- brw_blorp_blit_miptrees(intel,
+ brw_blorp_blit_miptrees(brw,
src_mt, src_irb->mt_level, src_irb->mt_layer,
+ src_format,
dst_mt, dst_irb->mt_level, dst_irb->mt_layer,
- srcX0, srcY0, dstX0, dstY0, dstX1, dstY1,
- mirror_x, mirror_y);
-
- intel_renderbuffer_set_needs_hiz_resolve(dst_irb);
- intel_renderbuffer_set_needs_downsample(dst_irb);
-}
+ dst_format,
+ srcX0, srcY0, srcX1, srcY1,
+ dstX0, dstY0, dstX1, dstY1,
+ filter, mirror_x, mirror_y);
-
-static bool
-formats_match(GLbitfield buffer_bit, struct intel_renderbuffer *src_irb,
- struct intel_renderbuffer *dst_irb)
-{
- /* Note: don't just check gl_renderbuffer::Format, because in some cases
- * multiple gl_formats resolve to the same native type in the miptree (for
- * example MESA_FORMAT_X8_Z24 and MESA_FORMAT_S8_Z24), and we can blit
- * between those formats.
- */
- return find_miptree(buffer_bit, src_irb)->format ==
- find_miptree(buffer_bit, dst_irb)->format;
+ dst_irb->need_downsample = true;
}
-
static bool
-try_blorp_blit(struct intel_context *intel,
- GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
- GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+try_blorp_blit(struct brw_context *brw,
+ const struct gl_framebuffer *read_fb,
+ const struct gl_framebuffer *draw_fb,
+ GLfloat srcX0, GLfloat srcY0, GLfloat srcX1, GLfloat srcY1,
+ GLfloat dstX0, GLfloat dstY0, GLfloat dstX1, GLfloat dstY1,
GLenum filter, GLbitfield buffer_bit)
{
- struct gl_context *ctx = &intel->ctx;
+ struct gl_context *ctx = &brw->ctx;
/* Sync up the state of window system buffers. We need to do this before
* we go looking for the buffers.
*/
- intel_prepare_render(intel);
-
- const struct gl_framebuffer *read_fb = ctx->ReadBuffer;
- const struct gl_framebuffer *draw_fb = ctx->DrawBuffer;
-
- /* Detect if the blit needs to be mirrored */
- bool mirror_x = false, mirror_y = false;
- fixup_mirroring(mirror_x, srcX0, srcX1);
- fixup_mirroring(mirror_x, dstX0, dstX1);
- fixup_mirroring(mirror_y, srcY0, srcY1);
- fixup_mirroring(mirror_y, dstY0, dstY1);
-
- /* Make sure width and height match */
- if (srcX1 - srcX0 != dstX1 - dstX0) return false;
- if (srcY1 - srcY0 != dstY1 - dstY0) return false;
-
- /* If the destination rectangle needs to be clipped or scissored, do so.
- */
- if (!(clip_or_scissor(mirror_x, srcX0, srcX1, dstX0, dstX1,
- draw_fb->_Xmin, draw_fb->_Xmax) &&
- clip_or_scissor(mirror_y, srcY0, srcY1, dstY0, dstY1,
- draw_fb->_Ymin, draw_fb->_Ymax))) {
- /* Everything got clipped/scissored away, so the blit was successful. */
- return true;
- }
+ intel_prepare_render(brw);
- /* If the source rectangle needs to be clipped or scissored, do so. */
- if (!(clip_or_scissor(mirror_x, dstX0, dstX1, srcX0, srcX1,
- 0, read_fb->Width) &&
- clip_or_scissor(mirror_y, dstY0, dstY1, srcY0, srcY1,
- 0, read_fb->Height))) {
- /* Everything got clipped/scissored away, so the blit was successful. */
+ bool mirror_x, mirror_y;
+ if (brw_meta_mirror_clip_and_scissor(ctx, read_fb, draw_fb,
+ &srcX0, &srcY0, &srcX1, &srcY1,
+ &dstX0, &dstY0, &dstX1, &dstY1,
+ &mirror_x, &mirror_y))
return true;
- }
-
- /* Account for the fact that in the system framebuffer, the origin is at
- * the lower left.
- */
- if (_mesa_is_winsys_fbo(read_fb)) {
- GLint tmp = read_fb->Height - srcY0;
- srcY0 = read_fb->Height - srcY1;
- srcY1 = tmp;
- mirror_y = !mirror_y;
- }
- if (_mesa_is_winsys_fbo(draw_fb)) {
- GLint tmp = draw_fb->Height - dstY0;
- dstY0 = draw_fb->Height - dstY1;
- dstY1 = tmp;
- mirror_y = !mirror_y;
- }
/* Find buffers */
struct intel_renderbuffer *src_irb;
struct intel_renderbuffer *dst_irb;
+ struct intel_mipmap_tree *src_mt;
+ struct intel_mipmap_tree *dst_mt;
switch (buffer_bit) {
case GL_COLOR_BUFFER_BIT:
src_irb = intel_renderbuffer(read_fb->_ColorReadBuffer);
- for (unsigned i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; ++i) {
- dst_irb = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[i]);
- if (dst_irb && !formats_match(buffer_bit, src_irb, dst_irb))
- return false;
- }
- for (unsigned i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; ++i) {
- dst_irb = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[i]);
- do_blorp_blit(intel, buffer_bit, src_irb, dst_irb, srcX0, srcY0,
- dstX0, dstY0, dstX1, dstY1, mirror_x, mirror_y);
+ for (unsigned i = 0; i < draw_fb->_NumColorDrawBuffers; ++i) {
+ dst_irb = intel_renderbuffer(draw_fb->_ColorDrawBuffers[i]);
+ if (dst_irb)
+ do_blorp_blit(brw, buffer_bit,
+ src_irb, src_irb->Base.Base.Format,
+ dst_irb, dst_irb->Base.Base.Format,
+ srcX0, srcY0, srcX1, srcY1,
+ dstX0, dstY0, dstX1, dstY1,
+ filter, mirror_x, mirror_y);
}
break;
case GL_DEPTH_BUFFER_BIT:
intel_renderbuffer(read_fb->Attachment[BUFFER_DEPTH].Renderbuffer);
dst_irb =
intel_renderbuffer(draw_fb->Attachment[BUFFER_DEPTH].Renderbuffer);
- if (!formats_match(buffer_bit, src_irb, dst_irb))
+ src_mt = find_miptree(buffer_bit, src_irb);
+ dst_mt = find_miptree(buffer_bit, dst_irb);
+
+ /* We can't handle format conversions between Z24 and other formats
+ * since we have to lie about the surface format. See the comments in
+ * brw_blorp_surface_info::set().
+ */
+ if ((src_mt->format == MESA_FORMAT_Z24_UNORM_X8_UINT) !=
+ (dst_mt->format == MESA_FORMAT_Z24_UNORM_X8_UINT))
return false;
- do_blorp_blit(intel, buffer_bit, src_irb, dst_irb, srcX0, srcY0,
- dstX0, dstY0, dstX1, dstY1, mirror_x, mirror_y);
+
+ do_blorp_blit(brw, buffer_bit, src_irb, MESA_FORMAT_NONE,
+ dst_irb, MESA_FORMAT_NONE, srcX0, srcY0,
+ srcX1, srcY1, dstX0, dstY0, dstX1, dstY1,
+ filter, mirror_x, mirror_y);
break;
case GL_STENCIL_BUFFER_BIT:
src_irb =
intel_renderbuffer(read_fb->Attachment[BUFFER_STENCIL].Renderbuffer);
dst_irb =
intel_renderbuffer(draw_fb->Attachment[BUFFER_STENCIL].Renderbuffer);
- if (!formats_match(buffer_bit, src_irb, dst_irb))
- return false;
- do_blorp_blit(intel, buffer_bit, src_irb, dst_irb, srcX0, srcY0,
- dstX0, dstY0, dstX1, dstY1, mirror_x, mirror_y);
+ do_blorp_blit(brw, buffer_bit, src_irb, MESA_FORMAT_NONE,
+ dst_irb, MESA_FORMAT_NONE, srcX0, srcY0,
+ srcX1, srcY1, dstX0, dstY0, dstX1, dstY1,
+ filter, mirror_x, mirror_y);
break;
default:
- assert(false);
+ unreachable("not reached");
+ }
+
+ return true;
+}
+
+bool
+brw_blorp_copytexsubimage(struct brw_context *brw,
+ struct gl_renderbuffer *src_rb,
+ struct gl_texture_image *dst_image,
+ int slice,
+ int srcX0, int srcY0,
+ int dstX0, int dstY0,
+ int width, int height)
+{
+ struct gl_context *ctx = &brw->ctx;
+ struct intel_renderbuffer *src_irb = intel_renderbuffer(src_rb);
+ struct intel_texture_image *intel_image = intel_texture_image(dst_image);
+
+ /* No pixel transfer operations (zoom, bias, mapping), just a blit */
+ if (brw->ctx._ImageTransferState)
+ return false;
+
+ /* Sync up the state of window system buffers. We need to do this before
+ * we go looking at the src renderbuffer's miptree.
+ */
+ intel_prepare_render(brw);
+
+ struct intel_mipmap_tree *src_mt = src_irb->mt;
+ struct intel_mipmap_tree *dst_mt = intel_image->mt;
+
+ /* BLORP is only supported for Gen6-7. */
+ if (brw->gen < 6 || brw->gen > 7)
+ return false;
+
+ if (_mesa_get_format_base_format(src_rb->Format) !=
+ _mesa_get_format_base_format(dst_image->TexFormat)) {
+ return false;
+ }
+
+ /* We can't handle format conversions between Z24 and other formats since
+ * we have to lie about the surface format. See the comments in
+ * brw_blorp_surface_info::set().
+ */
+ if ((src_mt->format == MESA_FORMAT_Z24_UNORM_X8_UINT) !=
+ (dst_mt->format == MESA_FORMAT_Z24_UNORM_X8_UINT)) {
+ return false;
+ }
+
+ if (!brw->format_supported_as_render_target[dst_image->TexFormat])
+ return false;
+
+ /* Source clipping shouldn't be necessary, since copytexsubimage (in
+ * src/mesa/main/teximage.c) calls _mesa_clip_copytexsubimage() which
+ * takes care of it.
+ *
+ * Destination clipping shouldn't be necessary since the restrictions on
+ * glCopyTexSubImage prevent the user from specifying a destination rectangle
+ * that falls outside the bounds of the destination texture.
+ * See error_check_subtexture_dimensions().
+ */
+
+ int srcY1 = srcY0 + height;
+ int srcX1 = srcX0 + width;
+ int dstX1 = dstX0 + width;
+ int dstY1 = dstY0 + height;
+
+ /* Account for the fact that in the system framebuffer, the origin is at
+ * the lower left.
+ */
+ bool mirror_y = false;
+ if (_mesa_is_winsys_fbo(ctx->ReadBuffer)) {
+ GLint tmp = src_rb->Height - srcY0;
+ srcY0 = src_rb->Height - srcY1;
+ srcY1 = tmp;
+ mirror_y = true;
+ }
+
+ /* Account for face selection and texture view MinLayer */
+ int dst_slice = slice + dst_image->TexObject->MinLayer + dst_image->Face;
+ int dst_level = dst_image->Level + dst_image->TexObject->MinLevel;
+
+ brw_blorp_blit_miptrees(brw,
+ src_mt, src_irb->mt_level, src_irb->mt_layer,
+ src_rb->Format,
+ dst_mt, dst_level, dst_slice,
+ dst_image->TexFormat,
+ srcX0, srcY0, srcX1, srcY1,
+ dstX0, dstY0, dstX1, dstY1,
+ GL_NEAREST, false, mirror_y);
+
+ /* If we're copying to a packed depth stencil texture and the source
+ * framebuffer has separate stencil, we need to also copy the stencil data
+ * over.
+ */
+ src_rb = ctx->ReadBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
+ if (_mesa_get_format_bits(dst_image->TexFormat, GL_STENCIL_BITS) > 0 &&
+ src_rb != NULL) {
+ src_irb = intel_renderbuffer(src_rb);
+ src_mt = src_irb->mt;
+
+ if (src_mt->stencil_mt)
+ src_mt = src_mt->stencil_mt;
+ if (dst_mt->stencil_mt)
+ dst_mt = dst_mt->stencil_mt;
+
+ if (src_mt != dst_mt) {
+ brw_blorp_blit_miptrees(brw,
+ src_mt, src_irb->mt_level, src_irb->mt_layer,
+ src_mt->format,
+ dst_mt, dst_level, dst_slice,
+ dst_mt->format,
+ srcX0, srcY0, srcX1, srcY1,
+ dstX0, dstY0, dstX1, dstY1,
+ GL_NEAREST, false, mirror_y);
+ }
}
return true;
}
+
GLbitfield
-brw_blorp_framebuffer(struct intel_context *intel,
+brw_blorp_framebuffer(struct brw_context *brw,
+ struct gl_framebuffer *readFb,
+ struct gl_framebuffer *drawFb,
GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
GLbitfield mask, GLenum filter)
{
/* BLORP is not supported before Gen6. */
- if (intel->gen < 6)
+ if (brw->gen < 6 || brw->gen >= 8)
return mask;
static GLbitfield buffer_bits[] = {
for (unsigned int i = 0; i < ARRAY_SIZE(buffer_bits); ++i) {
if ((mask & buffer_bits[i]) &&
- try_blorp_blit(intel,
+ try_blorp_blit(brw, readFb, drawFb,
srcX0, srcY0, srcX1, srcY1,
dstX0, dstY0, dstX1, dstY1,
filter, buffer_bits[i])) {
* (In these formulas, pitch is the number of bytes occupied by a single row
* of samples).
*/
-class brw_blorp_blit_program
+class brw_blorp_blit_program : public brw_blorp_eu_emitter
{
public:
brw_blorp_blit_program(struct brw_context *brw,
- const brw_blorp_blit_prog_key *key);
- ~brw_blorp_blit_program();
+ const brw_blorp_blit_prog_key *key, bool debug_flag);
const GLuint *compile(struct brw_context *brw, GLuint *program_size);
void translate_tiling(bool old_tiled_w, bool new_tiled_w);
void encode_msaa(unsigned num_samples, intel_msaa_layout layout);
void decode_msaa(unsigned num_samples, intel_msaa_layout layout);
- void kill_if_outside_dst_rect();
void translate_dst_to_src();
+ void clamp_tex_coords(struct brw_reg regX, struct brw_reg regY,
+ struct brw_reg clampX0, struct brw_reg clampY0,
+ struct brw_reg clampX1, struct brw_reg clampY1);
void single_to_blend();
- void manual_blend(unsigned num_samples);
+ void manual_blend_average(unsigned num_samples);
+ void manual_blend_bilinear(unsigned num_samples);
void sample(struct brw_reg dst);
void texel_fetch(struct brw_reg dst);
void mcs_fetch();
- void expand_to_32_bits(struct brw_reg src, struct brw_reg dst);
- void texture_lookup(struct brw_reg dst, GLuint msg_type,
+ void texture_lookup(struct brw_reg dst, enum opcode op,
const sampler_message_arg *args, int num_args);
void render_target_write();
*/
static const unsigned LOG2_MAX_BLEND_SAMPLES = 3;
- void *mem_ctx;
struct brw_context *brw;
const brw_blorp_blit_prog_key *key;
- struct brw_compile func;
/* Thread dispatch header */
struct brw_reg R0;
struct brw_reg dst_x1;
struct brw_reg dst_y0;
struct brw_reg dst_y1;
+ /* Top right coordinates of the rectangular grid used for scaled blitting */
+ struct brw_reg rect_grid_x1;
+ struct brw_reg rect_grid_y1;
struct {
struct brw_reg multiplier;
struct brw_reg offset;
*/
struct brw_reg y_coords[2];
+ /* X, Y coordinates of the pixel from which we need to fetch the specific
+ * sample. These are used for multisample scaled blitting.
+ */
+ struct brw_reg x_sample_coords;
+ struct brw_reg y_sample_coords;
+
+ /* Fractional parts of the x and y coordinates, used as bilinear interpolation coefficients */
+ struct brw_reg x_frac;
+ struct brw_reg y_frac;
+
/* Which element of x_coords and y_coords is currently in use.
*/
int xy_coord_index;
brw_blorp_blit_program::brw_blorp_blit_program(
struct brw_context *brw,
- const brw_blorp_blit_prog_key *key)
- : mem_ctx(ralloc_context(NULL)),
+ const brw_blorp_blit_prog_key *key,
+ bool debug_flag)
+ : brw_blorp_eu_emitter(brw, debug_flag),
brw(brw),
key(key)
{
- brw_init_compile(brw, &func, mem_ctx);
-}
-
-brw_blorp_blit_program::~brw_blorp_blit_program()
-{
- ralloc_free(mem_ctx);
}
const GLuint *
memset(&prog_data, 0, sizeof(prog_data));
prog_data.persample_msaa_dispatch = key->persample_msaa_dispatch;
- brw_set_compression_control(&func, BRW_COMPRESSION_NONE);
-
alloc_regs();
compute_frag_coords();
*/
if (key->use_kill)
- kill_if_outside_dst_rect();
+ emit_kill_if_outside_rect(x_coords[xy_coord_index],
+ y_coords[xy_coord_index],
+ dst_x0, dst_x1, dst_y0, dst_y1);
/* Next, apply a translation to obtain coordinates in the source image. */
translate_dst_to_src();
* that we want to texture from. Exception: if we are blending, then S is
* irrelevant, because we are going to fetch all samples.
*/
- if (key->blend) {
- if (brw->intel.gen == 6) {
+ if (key->blend && !key->blit_scaled) {
+ if (brw->gen == 6) {
/* Gen6 hardware an automatically blend using the SAMPLE message */
single_to_blend();
sample(texture_data[0]);
} else {
/* Gen7+ hardware doesn't automaticaly blend. */
- manual_blend(key->src_samples);
+ manual_blend_average(key->src_samples);
}
+ } else if(key->blend && key->blit_scaled) {
+ manual_blend_bilinear(key->src_samples);
} else {
/* We aren't blending, which means we just want to fetch a single sample
* from the source surface. The address that we want to fetch from is
* the same as the configuration of the texture, then we need to adjust
* the coordinates to compensate for the difference.
*/
- if (tex_tiled_w != key->src_tiled_w ||
- key->tex_samples != key->src_samples ||
- key->tex_layout != key->src_layout) {
+ if ((tex_tiled_w != key->src_tiled_w ||
+ key->tex_samples != key->src_samples ||
+ key->tex_layout != key->src_layout) &&
+ !key->bilinear_filter) {
encode_msaa(key->src_samples, key->src_layout);
/* Now (X, Y, S) = detile(src_tiling, offset) */
translate_tiling(key->src_tiled_w, tex_tiled_w);
decode_msaa(key->tex_samples, key->tex_layout);
}
- /* Now (X, Y, S) = decode_msaa(tex_samples, detile(tex_tiling, offset)).
- *
- * In other words: X, Y, and S now contain values which, when passed to
- * the texturing unit, will cause data to be read from the correct
- * memory location. So we can fetch the texel now.
- */
- if (key->tex_layout == INTEL_MSAA_LAYOUT_CMS)
- mcs_fetch();
- texel_fetch(texture_data[0]);
+ if (key->bilinear_filter) {
+ sample(texture_data[0]);
+ }
+ else {
+ /* Now (X, Y, S) = decode_msaa(tex_samples, detile(tex_tiling, offset)).
+ *
+ * In other words: X, Y, and S now contain values which, when passed to
+ * the texturing unit, will cause data to be read from the correct
+ * memory location. So we can fetch the texel now.
+ */
+ if (key->tex_layout == INTEL_MSAA_LAYOUT_CMS)
+ mcs_fetch();
+ texel_fetch(texture_data[0]);
+ }
}
/* Finally, write the fetched (or blended) value to the render target and
* terminate the thread.
*/
render_target_write();
- return brw_get_program(&func, program_size);
+
+ return get_program(program_size);
}
void
brw_blorp_blit_program::alloc_push_const_regs(int base_reg)
{
#define CONST_LOC(name) offsetof(brw_blorp_wm_push_constants, name)
-#define ALLOC_REG(name) \
- this->name = \
- brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, base_reg, CONST_LOC(name) / 2)
-
- ALLOC_REG(dst_x0);
- ALLOC_REG(dst_x1);
- ALLOC_REG(dst_y0);
- ALLOC_REG(dst_y1);
- ALLOC_REG(x_transform.multiplier);
- ALLOC_REG(x_transform.offset);
- ALLOC_REG(y_transform.multiplier);
- ALLOC_REG(y_transform.offset);
+#define ALLOC_REG(name, type) \
+ this->name = \
+ retype(brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, \
+ base_reg + CONST_LOC(name) / 32, \
+ (CONST_LOC(name) % 32) / 4), type)
+
+ ALLOC_REG(dst_x0, BRW_REGISTER_TYPE_UD);
+ ALLOC_REG(dst_x1, BRW_REGISTER_TYPE_UD);
+ ALLOC_REG(dst_y0, BRW_REGISTER_TYPE_UD);
+ ALLOC_REG(dst_y1, BRW_REGISTER_TYPE_UD);
+ ALLOC_REG(rect_grid_x1, BRW_REGISTER_TYPE_F);
+ ALLOC_REG(rect_grid_y1, BRW_REGISTER_TYPE_F);
+ ALLOC_REG(x_transform.multiplier, BRW_REGISTER_TYPE_F);
+ ALLOC_REG(x_transform.offset, BRW_REGISTER_TYPE_F);
+ ALLOC_REG(y_transform.multiplier, BRW_REGISTER_TYPE_F);
+ ALLOC_REG(y_transform.offset, BRW_REGISTER_TYPE_F);
#undef CONST_LOC
#undef ALLOC_REG
}
}
this->mcs_data =
retype(brw_vec8_grf(reg, 0), BRW_REGISTER_TYPE_UD); reg += 8;
+
for (int i = 0; i < 2; ++i) {
this->x_coords[i]
- = vec16(retype(brw_vec8_grf(reg++, 0), BRW_REGISTER_TYPE_UW));
+ = retype(brw_vec8_grf(reg, 0), BRW_REGISTER_TYPE_UD);
+ reg += 2;
this->y_coords[i]
- = vec16(retype(brw_vec8_grf(reg++, 0), BRW_REGISTER_TYPE_UW));
+ = retype(brw_vec8_grf(reg, 0), BRW_REGISTER_TYPE_UD);
+ reg += 2;
}
+
+ if (key->blit_scaled && key->blend) {
+ this->x_sample_coords = brw_vec8_grf(reg, 0);
+ reg += 2;
+ this->y_sample_coords = brw_vec8_grf(reg, 0);
+ reg += 2;
+ this->x_frac = brw_vec8_grf(reg, 0);
+ reg += 2;
+ this->y_frac = brw_vec8_grf(reg, 0);
+ reg += 2;
+ }
+
this->xy_coord_index = 0;
this->sample_index
- = vec16(retype(brw_vec8_grf(reg++, 0), BRW_REGISTER_TYPE_UW));
- this->t1 = vec16(retype(brw_vec8_grf(reg++, 0), BRW_REGISTER_TYPE_UW));
- this->t2 = vec16(retype(brw_vec8_grf(reg++, 0), BRW_REGISTER_TYPE_UW));
+ = retype(brw_vec8_grf(reg, 0), BRW_REGISTER_TYPE_UD);
+ reg += 2;
+ this->t1 = retype(brw_vec8_grf(reg, 0), BRW_REGISTER_TYPE_UD);
+ reg += 2;
+ this->t2 = retype(brw_vec8_grf(reg, 0), BRW_REGISTER_TYPE_UD);
+ reg += 2;
/* Make sure we didn't run out of registers */
assert(reg <= GEN7_MRF_HACK_START);
* Then, we need to add the repeating sequence (0, 1, 0, 1, ...) to the
* result, since pixels n+1 and n+3 are in the right half of the subspan.
*/
- brw_ADD(&func, X, stride(suboffset(R1, 4), 2, 4, 0), brw_imm_v(0x10101010));
+ emit_add(vec16(retype(X, BRW_REGISTER_TYPE_UW)),
+ stride(suboffset(R1, 4), 2, 4, 0), brw_imm_v(0x10101010));
/* Similarly, Y coordinates for subspans come from R1.2[31:16] through
* R1.5[31:16], so to get pixel Y coordinates we need to start at the 5th
* And we need to add the repeating sequence (0, 0, 1, 1, ...), since
* pixels n+2 and n+3 are in the bottom half of the subspan.
*/
- brw_ADD(&func, Y, stride(suboffset(R1, 5), 2, 4, 0), brw_imm_v(0x11001100));
+ emit_add(vec16(retype(Y, BRW_REGISTER_TYPE_UW)),
+ stride(suboffset(R1, 5), 2, 4, 0), brw_imm_v(0x11001100));
+
+ /* Move the coordinates to UD registers. */
+ emit_mov(vec16(Xp), retype(X, BRW_REGISTER_TYPE_UW));
+ emit_mov(vec16(Yp), retype(Y, BRW_REGISTER_TYPE_UW));
+ SWAP_XY_AND_XPYP();
if (key->persample_msaa_dispatch) {
switch (key->rt_samples) {
- case 4:
+ case 4: {
/* The WM will be run in MSDISPMODE_PERSAMPLE with num_samples == 4.
* Therefore, subspan 0 will represent sample 0, subspan 1 will
* represent sample 1, and so on.
* populate a temporary variable with the sequence (0, 1, 2, 3), and
* then copy from it using vstride=1, width=4, hstride=0.
*/
- brw_MOV(&func, t1, brw_imm_v(0x3210));
- brw_MOV(&func, S, stride(t1, 1, 4, 0));
+ struct brw_reg t1_uw1 = retype(t1, BRW_REGISTER_TYPE_UW);
+ emit_mov(vec16(t1_uw1), brw_imm_v(0x3210));
+ /* Move to UD sample_index register. */
+ emit_mov_8(S, stride(t1_uw1, 1, 4, 0));
+ emit_mov_8(offset(S, 1), suboffset(stride(t1_uw1, 1, 4, 0), 2));
break;
+ }
case 8: {
/* The WM will be run in MSDISPMODE_PERSAMPLE with num_samples == 8.
* Therefore, subspan 0 will represent sample N (where N is 0 or 4),
* using vstride=1, width=4, hstride=0.
*/
struct brw_reg t1_ud1 = vec1(retype(t1, BRW_REGISTER_TYPE_UD));
+ struct brw_reg t2_uw1 = retype(t2, BRW_REGISTER_TYPE_UW);
struct brw_reg r0_ud1 = vec1(retype(R0, BRW_REGISTER_TYPE_UD));
- brw_AND(&func, t1_ud1, r0_ud1, brw_imm_ud(0xc0));
- brw_SHR(&func, t1_ud1, t1_ud1, brw_imm_ud(5));
- brw_MOV(&func, t2, brw_imm_v(0x3210));
- brw_ADD(&func, S, retype(t1_ud1, BRW_REGISTER_TYPE_UW),
- stride(t2, 1, 4, 0));
+ emit_and(t1_ud1, r0_ud1, brw_imm_ud(0xc0));
+ emit_shr(t1_ud1, t1_ud1, brw_imm_ud(5));
+ emit_mov(vec16(t2_uw1), brw_imm_v(0x3210));
+ emit_add(vec16(S), retype(t1_ud1, BRW_REGISTER_TYPE_UW),
+ stride(t2_uw1, 1, 4, 0));
+ emit_add_8(offset(S, 1),
+ retype(t1_ud1, BRW_REGISTER_TYPE_UW),
+ suboffset(stride(t2_uw1, 1, 4, 0), 2));
break;
}
default:
- assert(!"Unrecognized sample count in "
- "brw_blorp_blit_program::compute_frag_coords()");
- break;
+ unreachable("Unrecognized sample count in "
+ "brw_blorp_blit_program::compute_frag_coords()");
}
s_is_zero = false;
} else {
* X' = (X & ~0b1011) >> 1 | (Y & 0b1) << 2 | X & 0b1 (4)
* Y' = (Y & ~0b1) << 1 | (X & 0b1000) >> 2 | (X & 0b10) >> 1
*/
- brw_AND(&func, t1, X, brw_imm_uw(0xfff4)); /* X & ~0b1011 */
- brw_SHR(&func, t1, t1, brw_imm_uw(1)); /* (X & ~0b1011) >> 1 */
- brw_AND(&func, t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- brw_SHL(&func, t2, t2, brw_imm_uw(2)); /* (Y & 0b1) << 2 */
- brw_OR(&func, t1, t1, t2); /* (X & ~0b1011) >> 1 | (Y & 0b1) << 2 */
- brw_AND(&func, t2, X, brw_imm_uw(1)); /* X & 0b1 */
- brw_OR(&func, Xp, t1, t2);
- brw_AND(&func, t1, Y, brw_imm_uw(0xfffe)); /* Y & ~0b1 */
- brw_SHL(&func, t1, t1, brw_imm_uw(1)); /* (Y & ~0b1) << 1 */
- brw_AND(&func, t2, X, brw_imm_uw(8)); /* X & 0b1000 */
- brw_SHR(&func, t2, t2, brw_imm_uw(2)); /* (X & 0b1000) >> 2 */
- brw_OR(&func, t1, t1, t2); /* (Y & ~0b1) << 1 | (X & 0b1000) >> 2 */
- brw_AND(&func, t2, X, brw_imm_uw(2)); /* X & 0b10 */
- brw_SHR(&func, t2, t2, brw_imm_uw(1)); /* (X & 0b10) >> 1 */
- brw_OR(&func, Yp, t1, t2);
+ emit_and(t1, X, brw_imm_uw(0xfff4)); /* X & ~0b1011 */
+ emit_shr(t1, t1, brw_imm_uw(1)); /* (X & ~0b1011) >> 1 */
+ emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
+ emit_shl(t2, t2, brw_imm_uw(2)); /* (Y & 0b1) << 2 */
+ emit_or(t1, t1, t2); /* (X & ~0b1011) >> 1 | (Y & 0b1) << 2 */
+ emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
+ emit_or(Xp, t1, t2);
+ emit_and(t1, Y, brw_imm_uw(0xfffe)); /* Y & ~0b1 */
+ emit_shl(t1, t1, brw_imm_uw(1)); /* (Y & ~0b1) << 1 */
+ emit_and(t2, X, brw_imm_uw(8)); /* X & 0b1000 */
+ emit_shr(t2, t2, brw_imm_uw(2)); /* (X & 0b1000) >> 2 */
+ emit_or(t1, t1, t2); /* (Y & ~0b1) << 1 | (X & 0b1000) >> 2 */
+ emit_and(t2, X, brw_imm_uw(2)); /* X & 0b10 */
+ emit_shr(t2, t2, brw_imm_uw(1)); /* (X & 0b10) >> 1 */
+ emit_or(Yp, t1, t2);
SWAP_XY_AND_XPYP();
} else {
/* Applying the same logic as above, but in reverse, we obtain the
* X' = (X & ~0b101) << 1 | (Y & 0b10) << 2 | (Y & 0b1) << 1 | X & 0b1
* Y' = (Y & ~0b11) >> 1 | (X & 0b100) >> 2
*/
- brw_AND(&func, t1, X, brw_imm_uw(0xfffa)); /* X & ~0b101 */
- brw_SHL(&func, t1, t1, brw_imm_uw(1)); /* (X & ~0b101) << 1 */
- brw_AND(&func, t2, Y, brw_imm_uw(2)); /* Y & 0b10 */
- brw_SHL(&func, t2, t2, brw_imm_uw(2)); /* (Y & 0b10) << 2 */
- brw_OR(&func, t1, t1, t2); /* (X & ~0b101) << 1 | (Y & 0b10) << 2 */
- brw_AND(&func, t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- brw_SHL(&func, t2, t2, brw_imm_uw(1)); /* (Y & 0b1) << 1 */
- brw_OR(&func, t1, t1, t2); /* (X & ~0b101) << 1 | (Y & 0b10) << 2
+ emit_and(t1, X, brw_imm_uw(0xfffa)); /* X & ~0b101 */
+ emit_shl(t1, t1, brw_imm_uw(1)); /* (X & ~0b101) << 1 */
+ emit_and(t2, Y, brw_imm_uw(2)); /* Y & 0b10 */
+ emit_shl(t2, t2, brw_imm_uw(2)); /* (Y & 0b10) << 2 */
+ emit_or(t1, t1, t2); /* (X & ~0b101) << 1 | (Y & 0b10) << 2 */
+ emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
+ emit_shl(t2, t2, brw_imm_uw(1)); /* (Y & 0b1) << 1 */
+ emit_or(t1, t1, t2); /* (X & ~0b101) << 1 | (Y & 0b10) << 2
| (Y & 0b1) << 1 */
- brw_AND(&func, t2, X, brw_imm_uw(1)); /* X & 0b1 */
- brw_OR(&func, Xp, t1, t2);
- brw_AND(&func, t1, Y, brw_imm_uw(0xfffc)); /* Y & ~0b11 */
- brw_SHR(&func, t1, t1, brw_imm_uw(1)); /* (Y & ~0b11) >> 1 */
- brw_AND(&func, t2, X, brw_imm_uw(4)); /* X & 0b100 */
- brw_SHR(&func, t2, t2, brw_imm_uw(2)); /* (X & 0b100) >> 2 */
- brw_OR(&func, Yp, t1, t2);
+ emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
+ emit_or(Xp, t1, t2);
+ emit_and(t1, Y, brw_imm_uw(0xfffc)); /* Y & ~0b11 */
+ emit_shr(t1, t1, brw_imm_uw(1)); /* (Y & ~0b11) >> 1 */
+ emit_and(t2, X, brw_imm_uw(4)); /* X & 0b100 */
+ emit_shr(t2, t2, brw_imm_uw(2)); /* (X & 0b100) >> 2 */
+ emit_or(Yp, t1, t2);
SWAP_XY_AND_XPYP();
}
}
/* We can't compensate for compressed layout since at this point in the
* program we haven't read from the MCS buffer.
*/
- assert(!"Bad layout in encode_msaa");
- break;
+ unreachable("Bad layout in encode_msaa");
case INTEL_MSAA_LAYOUT_UMS:
/* No translation necessary. */
break;
* where X' = (X & ~0b1) << 1 | (S & 0b1) << 1 | (X & 0b1)
* Y' = (Y & ~0b1) << 1 | (S & 0b10) | (Y & 0b1)
*/
- brw_AND(&func, t1, X, brw_imm_uw(0xfffe)); /* X & ~0b1 */
+ emit_and(t1, X, brw_imm_uw(0xfffe)); /* X & ~0b1 */
if (!s_is_zero) {
- brw_AND(&func, t2, S, brw_imm_uw(1)); /* S & 0b1 */
- brw_OR(&func, t1, t1, t2); /* (X & ~0b1) | (S & 0b1) */
+ emit_and(t2, S, brw_imm_uw(1)); /* S & 0b1 */
+ emit_or(t1, t1, t2); /* (X & ~0b1) | (S & 0b1) */
}
- brw_SHL(&func, t1, t1, brw_imm_uw(1)); /* (X & ~0b1) << 1
+ emit_shl(t1, t1, brw_imm_uw(1)); /* (X & ~0b1) << 1
| (S & 0b1) << 1 */
- brw_AND(&func, t2, X, brw_imm_uw(1)); /* X & 0b1 */
- brw_OR(&func, Xp, t1, t2);
- brw_AND(&func, t1, Y, brw_imm_uw(0xfffe)); /* Y & ~0b1 */
- brw_SHL(&func, t1, t1, brw_imm_uw(1)); /* (Y & ~0b1) << 1 */
+ emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
+ emit_or(Xp, t1, t2);
+ emit_and(t1, Y, brw_imm_uw(0xfffe)); /* Y & ~0b1 */
+ emit_shl(t1, t1, brw_imm_uw(1)); /* (Y & ~0b1) << 1 */
if (!s_is_zero) {
- brw_AND(&func, t2, S, brw_imm_uw(2)); /* S & 0b10 */
- brw_OR(&func, t1, t1, t2); /* (Y & ~0b1) << 1 | (S & 0b10) */
+ emit_and(t2, S, brw_imm_uw(2)); /* S & 0b10 */
+ emit_or(t1, t1, t2); /* (Y & ~0b1) << 1 | (S & 0b10) */
}
- brw_AND(&func, t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- brw_OR(&func, Yp, t1, t2);
+ emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
+ emit_or(Yp, t1, t2);
break;
case 8:
/* encode_msaa(8, IMS, X, Y, S) = (X', Y', 0)
* | (X & 0b1)
* Y' = (Y & ~0b1) << 1 | (S & 0b10) | (Y & 0b1)
*/
- brw_AND(&func, t1, X, brw_imm_uw(0xfffe)); /* X & ~0b1 */
- brw_SHL(&func, t1, t1, brw_imm_uw(2)); /* (X & ~0b1) << 2 */
+ emit_and(t1, X, brw_imm_uw(0xfffe)); /* X & ~0b1 */
+ emit_shl(t1, t1, brw_imm_uw(2)); /* (X & ~0b1) << 2 */
if (!s_is_zero) {
- brw_AND(&func, t2, S, brw_imm_uw(4)); /* S & 0b100 */
- brw_OR(&func, t1, t1, t2); /* (X & ~0b1) << 2 | (S & 0b100) */
- brw_AND(&func, t2, S, brw_imm_uw(1)); /* S & 0b1 */
- brw_SHL(&func, t2, t2, brw_imm_uw(1)); /* (S & 0b1) << 1 */
- brw_OR(&func, t1, t1, t2); /* (X & ~0b1) << 2 | (S & 0b100)
+ emit_and(t2, S, brw_imm_uw(4)); /* S & 0b100 */
+ emit_or(t1, t1, t2); /* (X & ~0b1) << 2 | (S & 0b100) */
+ emit_and(t2, S, brw_imm_uw(1)); /* S & 0b1 */
+ emit_shl(t2, t2, brw_imm_uw(1)); /* (S & 0b1) << 1 */
+ emit_or(t1, t1, t2); /* (X & ~0b1) << 2 | (S & 0b100)
| (S & 0b1) << 1 */
}
- brw_AND(&func, t2, X, brw_imm_uw(1)); /* X & 0b1 */
- brw_OR(&func, Xp, t1, t2);
- brw_AND(&func, t1, Y, brw_imm_uw(0xfffe)); /* Y & ~0b1 */
- brw_SHL(&func, t1, t1, brw_imm_uw(1)); /* (Y & ~0b1) << 1 */
+ emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
+ emit_or(Xp, t1, t2);
+ emit_and(t1, Y, brw_imm_uw(0xfffe)); /* Y & ~0b1 */
+ emit_shl(t1, t1, brw_imm_uw(1)); /* (Y & ~0b1) << 1 */
if (!s_is_zero) {
- brw_AND(&func, t2, S, brw_imm_uw(2)); /* S & 0b10 */
- brw_OR(&func, t1, t1, t2); /* (Y & ~0b1) << 1 | (S & 0b10) */
+ emit_and(t2, S, brw_imm_uw(2)); /* S & 0b10 */
+ emit_or(t1, t1, t2); /* (Y & ~0b1) << 1 | (S & 0b10) */
}
- brw_AND(&func, t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- brw_OR(&func, Yp, t1, t2);
+ emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
+ emit_or(Yp, t1, t2);
break;
}
SWAP_XY_AND_XPYP();
/* We can't compensate for compressed layout since at this point in the
* program we don't have access to the MCS buffer.
*/
- assert(!"Bad layout in encode_msaa");
- break;
+ unreachable("Bad layout in encode_msaa");
case INTEL_MSAA_LAYOUT_UMS:
/* No translation necessary. */
break;
* Y' = (Y & ~0b11) >> 1 | (Y & 0b1)
* S = (Y & 0b10) | (X & 0b10) >> 1
*/
- brw_AND(&func, t1, X, brw_imm_uw(0xfffc)); /* X & ~0b11 */
- brw_SHR(&func, t1, t1, brw_imm_uw(1)); /* (X & ~0b11) >> 1 */
- brw_AND(&func, t2, X, brw_imm_uw(1)); /* X & 0b1 */
- brw_OR(&func, Xp, t1, t2);
- brw_AND(&func, t1, Y, brw_imm_uw(0xfffc)); /* Y & ~0b11 */
- brw_SHR(&func, t1, t1, brw_imm_uw(1)); /* (Y & ~0b11) >> 1 */
- brw_AND(&func, t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- brw_OR(&func, Yp, t1, t2);
- brw_AND(&func, t1, Y, brw_imm_uw(2)); /* Y & 0b10 */
- brw_AND(&func, t2, X, brw_imm_uw(2)); /* X & 0b10 */
- brw_SHR(&func, t2, t2, brw_imm_uw(1)); /* (X & 0b10) >> 1 */
- brw_OR(&func, S, t1, t2);
+ emit_and(t1, X, brw_imm_uw(0xfffc)); /* X & ~0b11 */
+ emit_shr(t1, t1, brw_imm_uw(1)); /* (X & ~0b11) >> 1 */
+ emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
+ emit_or(Xp, t1, t2);
+ emit_and(t1, Y, brw_imm_uw(0xfffc)); /* Y & ~0b11 */
+ emit_shr(t1, t1, brw_imm_uw(1)); /* (Y & ~0b11) >> 1 */
+ emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
+ emit_or(Yp, t1, t2);
+ emit_and(t1, Y, brw_imm_uw(2)); /* Y & 0b10 */
+ emit_and(t2, X, brw_imm_uw(2)); /* X & 0b10 */
+ emit_shr(t2, t2, brw_imm_uw(1)); /* (X & 0b10) >> 1 */
+ emit_or(S, t1, t2);
break;
case 8:
/* decode_msaa(8, IMS, X, Y, 0) = (X', Y', S)
* Y' = (Y & ~0b11) >> 1 | (Y & 0b1)
* S = (X & 0b100) | (Y & 0b10) | (X & 0b10) >> 1
*/
- brw_AND(&func, t1, X, brw_imm_uw(0xfff8)); /* X & ~0b111 */
- brw_SHR(&func, t1, t1, brw_imm_uw(2)); /* (X & ~0b111) >> 2 */
- brw_AND(&func, t2, X, brw_imm_uw(1)); /* X & 0b1 */
- brw_OR(&func, Xp, t1, t2);
- brw_AND(&func, t1, Y, brw_imm_uw(0xfffc)); /* Y & ~0b11 */
- brw_SHR(&func, t1, t1, brw_imm_uw(1)); /* (Y & ~0b11) >> 1 */
- brw_AND(&func, t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- brw_OR(&func, Yp, t1, t2);
- brw_AND(&func, t1, X, brw_imm_uw(4)); /* X & 0b100 */
- brw_AND(&func, t2, Y, brw_imm_uw(2)); /* Y & 0b10 */
- brw_OR(&func, t1, t1, t2); /* (X & 0b100) | (Y & 0b10) */
- brw_AND(&func, t2, X, brw_imm_uw(2)); /* X & 0b10 */
- brw_SHR(&func, t2, t2, brw_imm_uw(1)); /* (X & 0b10) >> 1 */
- brw_OR(&func, S, t1, t2);
+ emit_and(t1, X, brw_imm_uw(0xfff8)); /* X & ~0b111 */
+ emit_shr(t1, t1, brw_imm_uw(2)); /* (X & ~0b111) >> 2 */
+ emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
+ emit_or(Xp, t1, t2);
+ emit_and(t1, Y, brw_imm_uw(0xfffc)); /* Y & ~0b11 */
+ emit_shr(t1, t1, brw_imm_uw(1)); /* (Y & ~0b11) >> 1 */
+ emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
+ emit_or(Yp, t1, t2);
+ emit_and(t1, X, brw_imm_uw(4)); /* X & 0b100 */
+ emit_and(t2, Y, brw_imm_uw(2)); /* Y & 0b10 */
+ emit_or(t1, t1, t2); /* (X & 0b100) | (Y & 0b10) */
+ emit_and(t2, X, brw_imm_uw(2)); /* X & 0b10 */
+ emit_shr(t2, t2, brw_imm_uw(1)); /* (X & 0b10) >> 1 */
+ emit_or(S, t1, t2);
break;
}
s_is_zero = false;
}
/**
- * Emit code that kills pixels whose X and Y coordinates are outside the
- * boundary of the rectangle defined by the push constants (dst_x0, dst_y0,
- * dst_x1, dst_y1).
+ * Emit code to translate from destination (X, Y) coordinates to source (X, Y)
+ * coordinates.
*/
void
-brw_blorp_blit_program::kill_if_outside_dst_rect()
+brw_blorp_blit_program::translate_dst_to_src()
{
- struct brw_reg f0 = brw_flag_reg();
- struct brw_reg g1 = retype(brw_vec1_grf(1, 7), BRW_REGISTER_TYPE_UW);
- struct brw_reg null16 = vec16(retype(brw_null_reg(), BRW_REGISTER_TYPE_UW));
-
- brw_CMP(&func, null16, BRW_CONDITIONAL_GE, X, dst_x0);
- brw_CMP(&func, null16, BRW_CONDITIONAL_GE, Y, dst_y0);
- brw_CMP(&func, null16, BRW_CONDITIONAL_L, X, dst_x1);
- brw_CMP(&func, null16, BRW_CONDITIONAL_L, Y, dst_y1);
-
- brw_set_predicate_control(&func, BRW_PREDICATE_NONE);
- brw_push_insn_state(&func);
- brw_set_mask_control(&func, BRW_MASK_DISABLE);
- brw_AND(&func, g1, f0, g1);
- brw_pop_insn_state(&func);
+ struct brw_reg X_f = retype(X, BRW_REGISTER_TYPE_F);
+ struct brw_reg Y_f = retype(Y, BRW_REGISTER_TYPE_F);
+ struct brw_reg Xp_f = retype(Xp, BRW_REGISTER_TYPE_F);
+ struct brw_reg Yp_f = retype(Yp, BRW_REGISTER_TYPE_F);
+
+ /* Move the UD coordinates to float registers. */
+ emit_mov(Xp_f, X);
+ emit_mov(Yp_f, Y);
+ /* Scale and offset */
+ emit_mad(X_f, x_transform.offset, Xp_f, x_transform.multiplier);
+ emit_mad(Y_f, y_transform.offset, Yp_f, y_transform.multiplier);
+ if (key->blit_scaled && key->blend) {
+ /* Translate coordinates to lay out the samples in a rectangular grid
+ * roughly corresponding to sample locations.
+ */
+ emit_mul(X_f, X_f, brw_imm_f(key->x_scale));
+ emit_mul(Y_f, Y_f, brw_imm_f(key->y_scale));
+ /* Adjust coordinates so that integers represent pixel centers rather
+ * than pixel edges.
+ */
+ emit_add(X_f, X_f, brw_imm_f(-0.5));
+ emit_add(Y_f, Y_f, brw_imm_f(-0.5));
+
+ /* Clamp the X, Y texture coordinates to properly handle the sampling of
+ * texels on texture edges.
+ */
+ clamp_tex_coords(X_f, Y_f,
+ brw_imm_f(0.0), brw_imm_f(0.0),
+ rect_grid_x1, rect_grid_y1);
+
+ /* Store the fractional parts to be used as bilinear interpolation
+ * coefficients.
+ */
+ emit_frc(x_frac, X_f);
+ emit_frc(y_frac, Y_f);
+
+ /* Round the float coordinates down to nearest integer */
+ emit_rndd(Xp_f, X_f);
+ emit_rndd(Yp_f, Y_f);
+ emit_mul(X_f, Xp_f, brw_imm_f(1.0f / key->x_scale));
+ emit_mul(Y_f, Yp_f, brw_imm_f(1.0f / key->y_scale));
+ SWAP_XY_AND_XPYP();
+ } else if (!key->bilinear_filter) {
+ /* Round the float coordinates down to nearest integer by moving to
+ * UD registers.
+ */
+ emit_mov(Xp, X_f);
+ emit_mov(Yp, Y_f);
+ SWAP_XY_AND_XPYP();
+ }
}
-/**
- * Emit code to translate from destination (X, Y) coordinates to source (X, Y)
- * coordinates.
- */
void
-brw_blorp_blit_program::translate_dst_to_src()
+brw_blorp_blit_program::clamp_tex_coords(struct brw_reg regX,
+ struct brw_reg regY,
+ struct brw_reg clampX0,
+ struct brw_reg clampY0,
+ struct brw_reg clampX1,
+ struct brw_reg clampY1)
{
- brw_MUL(&func, Xp, X, x_transform.multiplier);
- brw_MUL(&func, Yp, Y, y_transform.multiplier);
- brw_ADD(&func, Xp, Xp, x_transform.offset);
- brw_ADD(&func, Yp, Yp, y_transform.offset);
- SWAP_XY_AND_XPYP();
+ emit_max(regX, regX, clampX0);
+ emit_max(regY, regY, clampY0);
+ emit_min(regX, regX, clampX1);
+ emit_min(regY, regY, clampY1);
}
/**
* that maxe up a pixel). So we need to multiply our X and Y coordinates
* each by 2 and then add 1.
*/
- brw_SHL(&func, t1, X, brw_imm_w(1));
- brw_SHL(&func, t2, Y, brw_imm_w(1));
- brw_ADD(&func, Xp, t1, brw_imm_w(1));
- brw_ADD(&func, Yp, t2, brw_imm_w(1));
+ emit_shl(t1, X, brw_imm_w(1));
+ emit_shl(t2, Y, brw_imm_w(1));
+ emit_add(Xp, t1, brw_imm_w(1));
+ emit_add(Yp, t2, brw_imm_w(1));
SWAP_XY_AND_XPYP();
}
*/
inline int count_trailing_one_bits(unsigned value)
{
-#if defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
+#ifdef HAVE___BUILTIN_CTZ
return __builtin_ctz(~value);
#else
return _mesa_bitcount(value & ~(value + 1));
void
-brw_blorp_blit_program::manual_blend(unsigned num_samples)
+brw_blorp_blit_program::manual_blend_average(unsigned num_samples)
{
if (key->tex_layout == INTEL_MSAA_LAYOUT_CMS)
mcs_fetch();
* For integer formats, we replace the add operations with average
* operations and skip the final division.
*/
- typedef struct brw_instruction *(*brw_op2_ptr)(struct brw_compile *,
- struct brw_reg,
- struct brw_reg,
- struct brw_reg);
- brw_op2_ptr combine_op =
- key->texture_data_type == BRW_REGISTER_TYPE_F ? brw_ADD : brw_AVG;
unsigned stack_depth = 0;
for (unsigned i = 0; i < num_samples; ++i) {
assert(stack_depth == _mesa_bitcount(i)); /* Loop invariant */
s_is_zero = true;
} else {
s_is_zero = false;
- brw_MOV(&func, S, brw_imm_uw(i));
+ emit_mov(vec16(S), brw_imm_ud(i));
}
texel_fetch(texture_data[stack_depth++]);
* Since we have already sampled from sample 0, all we need to do is
* skip the remaining fetches and averaging if MCS is zero.
*/
- brw_CMP(&func, vec16(brw_null_reg()), BRW_CONDITIONAL_NZ,
- mcs_data, brw_imm_ud(0));
- brw_IF(&func, BRW_EXECUTE_16);
+ emit_cmp_if(BRW_CONDITIONAL_NZ, mcs_data, brw_imm_ud(0));
}
/* Do count_trailing_one_bits(i) times */
/* TODO: should use a smaller loop bound for non_RGBA formats */
for (int k = 0; k < 4; ++k) {
- combine_op(&func, offset(texture_data[stack_depth - 1], 2*k),
- offset(vec8(texture_data[stack_depth - 1]), 2*k),
- offset(vec8(texture_data[stack_depth]), 2*k));
+ emit_combine(key->texture_data_type == BRW_REGISTER_TYPE_F ?
+ BRW_OPCODE_ADD : BRW_OPCODE_AVG,
+ offset(texture_data[stack_depth - 1], 2*k),
+ offset(vec8(texture_data[stack_depth - 1]), 2*k),
+ offset(vec8(texture_data[stack_depth]), 2*k));
}
}
}
/* Scale the result down by a factor of num_samples */
/* TODO: should use a smaller loop bound for non-RGBA formats */
for (int j = 0; j < 4; ++j) {
- brw_MUL(&func, offset(texture_data[0], 2*j),
+ emit_mul(offset(texture_data[0], 2*j),
offset(vec8(texture_data[0]), 2*j),
- brw_imm_f(1.0/num_samples));
+ brw_imm_f(1.0f / num_samples));
}
}
if (key->tex_layout == INTEL_MSAA_LAYOUT_CMS)
- brw_ENDIF(&func);
+ emit_endif();
+}
+
+void
+brw_blorp_blit_program::manual_blend_bilinear(unsigned num_samples)
+{
+ /* We do this computation by performing the following operations:
+ *
+ * In case of 4x, 8x MSAA:
+ * - Compute the pixel coordinates and sample numbers (a, b, c, d)
+ * which are later used for interpolation
+ * - linearly interpolate samples a and b in X
+ * - linearly interpolate samples c and d in X
+ * - linearly interpolate the results of last two operations in Y
+ *
+ * result = lrp(lrp(a + b) + lrp(c + d))
+ */
+ struct brw_reg Xp_f = retype(Xp, BRW_REGISTER_TYPE_F);
+ struct brw_reg Yp_f = retype(Yp, BRW_REGISTER_TYPE_F);
+ struct brw_reg t1_f = retype(t1, BRW_REGISTER_TYPE_F);
+ struct brw_reg t2_f = retype(t2, BRW_REGISTER_TYPE_F);
+
+ for (unsigned i = 0; i < 4; ++i) {
+ assert(i < ARRAY_SIZE(texture_data));
+ s_is_zero = false;
+
+ /* Compute pixel coordinates */
+ emit_add(vec16(x_sample_coords), Xp_f,
+ brw_imm_f((float)(i & 0x1) * (1.0f / key->x_scale)));
+ emit_add(vec16(y_sample_coords), Yp_f,
+ brw_imm_f((float)((i >> 1) & 0x1) * (1.0f / key->y_scale)));
+ emit_mov(vec16(X), x_sample_coords);
+ emit_mov(vec16(Y), y_sample_coords);
+
+ /* The MCS value we fetch has to match up with the pixel that we're
+ * sampling from. Since we sample from different pixels in each
+ * iteration of this "for" loop, the call to mcs_fetch() should be
+ * here inside the loop after computing the pixel coordinates.
+ */
+ if (key->tex_layout == INTEL_MSAA_LAYOUT_CMS)
+ mcs_fetch();
+
+ /* Compute sample index and map the sample index to a sample number.
+ * Sample index layout shows the numbering of slots in a rectangular
+ * grid of samples with in a pixel. Sample number layout shows the
+ * rectangular grid of samples roughly corresponding to the real sample
+ * locations with in a pixel.
+ * In case of 4x MSAA, layout of sample indices matches the layout of
+ * sample numbers:
+ * ---------
+ * | 0 | 1 |
+ * ---------
+ * | 2 | 3 |
+ * ---------
+ *
+ * In case of 8x MSAA the two layouts don't match.
+ * sample index layout : --------- sample number layout : ---------
+ * | 0 | 1 | | 5 | 2 |
+ * --------- ---------
+ * | 2 | 3 | | 4 | 6 |
+ * --------- ---------
+ * | 4 | 5 | | 0 | 3 |
+ * --------- ---------
+ * | 6 | 7 | | 7 | 1 |
+ * --------- ---------
+ */
+ emit_frc(vec16(t1_f), x_sample_coords);
+ emit_frc(vec16(t2_f), y_sample_coords);
+ emit_mul(vec16(t1_f), t1_f, brw_imm_f(key->x_scale));
+ emit_mul(vec16(t2_f), t2_f, brw_imm_f(key->x_scale * key->y_scale));
+ emit_add(vec16(t1_f), t1_f, t2_f);
+ emit_mov(vec16(S), t1_f);
+
+ if (num_samples == 8) {
+ /* Map the sample index to a sample number */
+ emit_cmp_if(BRW_CONDITIONAL_L, S, brw_imm_d(4));
+ {
+ emit_mov(vec16(t2), brw_imm_d(5));
+ emit_if_eq_mov(S, 1, vec16(t2), 2);
+ emit_if_eq_mov(S, 2, vec16(t2), 4);
+ emit_if_eq_mov(S, 3, vec16(t2), 6);
+ }
+ emit_else();
+ {
+ emit_mov(vec16(t2), brw_imm_d(0));
+ emit_if_eq_mov(S, 5, vec16(t2), 3);
+ emit_if_eq_mov(S, 6, vec16(t2), 7);
+ emit_if_eq_mov(S, 7, vec16(t2), 1);
+ }
+ emit_endif();
+ emit_mov(vec16(S), t2);
+ }
+ texel_fetch(texture_data[i]);
+ }
+
+#define SAMPLE(x, y) offset(texture_data[x], y)
+ for (int index = 3; index > 0; ) {
+ /* Since we're doing SIMD16, 4 color channels fits in to 8 registers.
+ * Counter value of 8 in 'for' loop below is used to interpolate all
+ * the color components.
+ */
+ for (int k = 0; k < 8; k += 2)
+ emit_lrp(vec8(SAMPLE(index - 1, k)),
+ x_frac,
+ vec8(SAMPLE(index, k)),
+ vec8(SAMPLE(index - 1, k)));
+ index -= 2;
+ }
+ for (int k = 0; k < 8; k += 2)
+ emit_lrp(vec8(SAMPLE(0, k)),
+ y_frac,
+ vec8(SAMPLE(2, k)),
+ vec8(SAMPLE(0, k)));
+#undef SAMPLE
}
/**
SAMPLER_MESSAGE_ARG_V_FLOAT
};
- texture_lookup(dst, GEN5_SAMPLER_MESSAGE_SAMPLE, args, ARRAY_SIZE(args));
+ texture_lookup(dst, SHADER_OPCODE_TEX, args, ARRAY_SIZE(args));
}
/**
SAMPLER_MESSAGE_ARG_V_INT
};
- switch (brw->intel.gen) {
+ switch (brw->gen) {
case 6:
- texture_lookup(dst, GEN5_SAMPLER_MESSAGE_SAMPLE_LD, gen6_args,
- s_is_zero ? 2 : 5);
+ texture_lookup(dst, SHADER_OPCODE_TXF, gen6_args, s_is_zero ? 2 : 5);
break;
case 7:
switch (key->tex_layout) {
* INTEL_MSAA_LAYOUT_CMS.
*/
case INTEL_MSAA_LAYOUT_CMS:
- texture_lookup(dst, GEN7_SAMPLER_MESSAGE_SAMPLE_LD2DMS,
+ texture_lookup(dst, SHADER_OPCODE_TXF_CMS,
gen7_ld2dms_args, ARRAY_SIZE(gen7_ld2dms_args));
break;
case INTEL_MSAA_LAYOUT_UMS:
- texture_lookup(dst, GEN7_SAMPLER_MESSAGE_SAMPLE_LD2DSS,
+ texture_lookup(dst, SHADER_OPCODE_TXF_UMS,
gen7_ld2dss_args, ARRAY_SIZE(gen7_ld2dss_args));
break;
case INTEL_MSAA_LAYOUT_NONE:
assert(s_is_zero);
- texture_lookup(dst, GEN5_SAMPLER_MESSAGE_SAMPLE_LD, gen7_ld_args,
+ texture_lookup(dst, SHADER_OPCODE_TXF, gen7_ld_args,
ARRAY_SIZE(gen7_ld_args));
break;
}
break;
default:
- assert(!"Should not get here.");
- break;
+ unreachable("Should not get here.");
};
}
SAMPLER_MESSAGE_ARG_U_INT,
SAMPLER_MESSAGE_ARG_V_INT
};
- texture_lookup(vec16(mcs_data), GEN7_SAMPLER_MESSAGE_SAMPLE_LD_MCS,
+ texture_lookup(vec16(mcs_data), SHADER_OPCODE_TXF_MCS,
gen7_ld_mcs_args, ARRAY_SIZE(gen7_ld_mcs_args));
}
-void
-brw_blorp_blit_program::expand_to_32_bits(struct brw_reg src,
- struct brw_reg dst)
-{
- brw_MOV(&func, vec8(dst), vec8(src));
- brw_set_compression_control(&func, BRW_COMPRESSION_2NDHALF);
- brw_MOV(&func, offset(vec8(dst), 1), suboffset(vec8(src), 8));
- brw_set_compression_control(&func, BRW_COMPRESSION_NONE);
-}
-
void
brw_blorp_blit_program::texture_lookup(struct brw_reg dst,
- GLuint msg_type,
+ enum opcode op,
const sampler_message_arg *args,
int num_args)
{
for (int arg = 0; arg < num_args; ++arg) {
switch (args[arg]) {
case SAMPLER_MESSAGE_ARG_U_FLOAT:
- expand_to_32_bits(X, retype(mrf, BRW_REGISTER_TYPE_F));
+ if (key->bilinear_filter)
+ emit_mov(retype(mrf, BRW_REGISTER_TYPE_F),
+ retype(X, BRW_REGISTER_TYPE_F));
+ else
+ emit_mov(retype(mrf, BRW_REGISTER_TYPE_F), X);
break;
case SAMPLER_MESSAGE_ARG_V_FLOAT:
- expand_to_32_bits(Y, retype(mrf, BRW_REGISTER_TYPE_F));
+ if (key->bilinear_filter)
+ emit_mov(retype(mrf, BRW_REGISTER_TYPE_F),
+ retype(Y, BRW_REGISTER_TYPE_F));
+ else
+ emit_mov(retype(mrf, BRW_REGISTER_TYPE_F), Y);
break;
case SAMPLER_MESSAGE_ARG_U_INT:
- expand_to_32_bits(X, mrf);
+ emit_mov(mrf, X);
break;
case SAMPLER_MESSAGE_ARG_V_INT:
- expand_to_32_bits(Y, mrf);
+ emit_mov(mrf, Y);
break;
case SAMPLER_MESSAGE_ARG_SI_INT:
/* Note: on Gen7, this code may be reached with s_is_zero==true
* appropriate message register.
*/
if (s_is_zero)
- brw_MOV(&func, mrf, brw_imm_ud(0));
+ emit_mov(mrf, brw_imm_ud(0));
else
- expand_to_32_bits(S, mrf);
+ emit_mov(mrf, S);
break;
case SAMPLER_MESSAGE_ARG_MCS_INT:
switch (key->tex_layout) {
case INTEL_MSAA_LAYOUT_CMS:
- brw_MOV(&func, mrf, mcs_data);
+ emit_mov(mrf, mcs_data);
break;
case INTEL_MSAA_LAYOUT_IMS:
/* When sampling from an IMS surface, MCS data is not relevant,
}
break;
case SAMPLER_MESSAGE_ARG_ZERO_INT:
- brw_MOV(&func, mrf, brw_imm_ud(0));
+ emit_mov(mrf, brw_imm_ud(0));
break;
}
mrf.nr += 2;
}
- brw_SAMPLE(&func,
- retype(dst, BRW_REGISTER_TYPE_UW) /* dest */,
- base_mrf /* msg_reg_nr */,
- brw_message_reg(base_mrf) /* src0 */,
- BRW_BLORP_TEXTURE_BINDING_TABLE_INDEX,
- 0 /* sampler */,
- WRITEMASK_XYZW,
- msg_type,
- 8 /* response_length. TODO: should be smaller for non-RGBA formats? */,
- mrf.nr - base_mrf /* msg_length */,
- 0 /* header_present */,
- BRW_SAMPLER_SIMD_MODE_SIMD16,
- BRW_SAMPLER_RETURN_FORMAT_FLOAT32);
+ emit_texture_lookup(retype(dst, BRW_REGISTER_TYPE_UW) /* dest */,
+ op,
+ base_mrf,
+ mrf.nr - base_mrf /* msg_length */);
}
#undef X
bool use_header = key->use_kill;
if (use_header) {
/* Copy R0/1 to MRF */
- brw_MOV(&func, retype(mrf_rt_write, BRW_REGISTER_TYPE_UD),
- retype(R0, BRW_REGISTER_TYPE_UD));
+ emit_mov(retype(mrf_rt_write, BRW_REGISTER_TYPE_UD),
+ retype(R0, BRW_REGISTER_TYPE_UD));
mrf_offset += 2;
}
/* Copy texture data to MRFs */
for (int i = 0; i < 4; ++i) {
/* E.g. mov(16) m2.0<1>:f r2.0<8;8,1>:f { Align1, H1 } */
- brw_MOV(&func, offset(mrf_rt_write, mrf_offset),
- offset(vec8(texture_data[0]), 2*i));
+ emit_mov(offset(mrf_rt_write, mrf_offset),
+ offset(vec8(texture_data[0]), 2*i));
mrf_offset += 2;
}
/* Now write to the render target and terminate the thread */
- brw_fb_WRITE(&func,
- 16 /* dispatch_width */,
- base_mrf /* msg_reg_nr */,
- mrf_rt_write /* src0 */,
- BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE,
- BRW_BLORP_RENDERBUFFER_BINDING_TABLE_INDEX,
- mrf_offset /* msg_length. TODO: Should be smaller for non-RGBA formats. */,
- 0 /* response_length */,
- true /* eot */,
- use_header);
+ emit_render_target_write(
+ mrf_rt_write,
+ base_mrf,
+ mrf_offset /* msg_length. TODO: Should be smaller for non-RGBA formats. */,
+ use_header);
}
void
-brw_blorp_coord_transform_params::setup(GLuint src0, GLuint dst0, GLuint dst1,
+brw_blorp_coord_transform_params::setup(GLfloat src0, GLfloat src1,
+ GLfloat dst0, GLfloat dst1,
bool mirror)
{
+ float scale = (src1 - src0) / (dst1 - dst0);
if (!mirror) {
/* When not mirroring a coordinate (say, X), we need:
- * x' - src_x0 = x - dst_x0
+ * src_x - src_x0 = (dst_x - dst_x0 + 0.5) * scale
* Therefore:
- * x' = 1*x + (src_x0 - dst_x0)
+ * src_x = src_x0 + (dst_x - dst_x0 + 0.5) * scale
+ *
+ * blorp program uses "round toward zero" to convert the
+ * transformed floating point coordinates to integer coordinates,
+ * whereas the behaviour we actually want is "round to nearest",
+ * so 0.5 provides the necessary correction.
*/
- multiplier = 1;
- offset = src0 - dst0;
+ multiplier = scale;
+ offset = src0 + (-dst0 + 0.5f) * scale;
} else {
/* When mirroring X we need:
- * x' - src_x0 = dst_x1 - x - 1
+ * src_x - src_x0 = dst_x1 - dst_x - 0.5
* Therefore:
- * x' = -1*x + (src_x0 + dst_x1 - 1)
+ * src_x = src_x0 + (dst_x1 -dst_x - 0.5) * scale
*/
- multiplier = -1;
- offset = src0 + dst1 - 1;
+ multiplier = -scale;
+ offset = src0 + (dst1 - 0.5f) * scale;
}
}
}
/* Prior to Gen7, all MSAA surfaces use IMS layout. */
- if (brw->intel.gen == 6) {
+ if (brw->gen == 6) {
assert(true_layout == INTEL_MSAA_LAYOUT_IMS);
}
brw_blorp_blit_params::brw_blorp_blit_params(struct brw_context *brw,
struct intel_mipmap_tree *src_mt,
unsigned src_level, unsigned src_layer,
+ mesa_format src_format,
struct intel_mipmap_tree *dst_mt,
unsigned dst_level, unsigned dst_layer,
- GLuint src_x0, GLuint src_y0,
- GLuint dst_x0, GLuint dst_y0,
- GLuint dst_x1, GLuint dst_y1,
+ mesa_format dst_format,
+ GLfloat src_x0, GLfloat src_y0,
+ GLfloat src_x1, GLfloat src_y1,
+ GLfloat dst_x0, GLfloat dst_y0,
+ GLfloat dst_x1, GLfloat dst_y1,
+ GLenum filter,
bool mirror_x, bool mirror_y)
{
- src.set(brw, src_mt, src_level, src_layer);
- dst.set(brw, dst_mt, dst_level, dst_layer);
+ src.set(brw, src_mt, src_level, src_layer, src_format, false);
+ dst.set(brw, dst_mt, dst_level, dst_layer, dst_format, true);
+
+ /* Even though we do multisample resolves at the time of the blit, OpenGL
+ * specification defines them as if they happen at the time of rendering,
+ * which means that the type of averaging we do during the resolve should
+ * only depend on the source format; the destination format should be
+ * ignored. But, specification doesn't seem to be strict about it.
+ *
+ * It has been observed that mulitisample resolves produce slightly better
+ * looking images when averaging is done using destination format. NVIDIA's
+ * proprietary OpenGL driver also follow this approach. So, we choose to
+ * follow it in our driver.
+ *
+ * When multisampling, if the source and destination formats are equal
+ * (aside from the color space), we choose to blit in sRGB space to get
+ * this higher quality image.
+ */
+ if (src.num_samples > 1 &&
+ _mesa_get_format_color_encoding(dst_mt->format) == GL_SRGB &&
+ _mesa_get_srgb_format_linear(src_mt->format) ==
+ _mesa_get_srgb_format_linear(dst_mt->format)) {
+ assert(brw->format_supported_as_render_target[dst_mt->format]);
+ dst.brw_surfaceformat = brw->render_target_format[dst_mt->format];
+ src.brw_surfaceformat = brw_format_for_mesa_format(dst_mt->format);
+ }
+
+ /* When doing a multisample resolve of a GL_LUMINANCE32F or GL_INTENSITY32F
+ * texture, the above code configures the source format for L32_FLOAT or
+ * I32_FLOAT, and the destination format for R32_FLOAT. On Sandy Bridge,
+ * the SAMPLE message appears to handle multisampled L32_FLOAT and
+ * I32_FLOAT textures incorrectly, resulting in blocky artifacts. So work
+ * around the problem by using a source format of R32_FLOAT. This
+ * shouldn't affect rendering correctness, since the destination format is
+ * R32_FLOAT, so only the contents of the red channel matters.
+ */
+ if (brw->gen == 6 && src.num_samples > 1 && dst.num_samples <= 1 &&
+ src_mt->format == dst_mt->format &&
+ dst.brw_surfaceformat == BRW_SURFACEFORMAT_R32_FLOAT) {
+ src.brw_surfaceformat = dst.brw_surfaceformat;
+ }
use_wm_prog = true;
memset(&wm_prog_key, 0, sizeof(wm_prog_key));
wm_prog_key.texture_data_type = BRW_REGISTER_TYPE_F;
break;
case GL_UNSIGNED_INT:
- if (src_mt->format == MESA_FORMAT_S8) {
+ if (src_mt->format == MESA_FORMAT_S_UINT8) {
/* We process stencil as though it's an unsigned normalized color */
wm_prog_key.texture_data_type = BRW_REGISTER_TYPE_F;
} else {
wm_prog_key.texture_data_type = BRW_REGISTER_TYPE_D;
break;
default:
- assert(!"Unrecognized blorp format");
- break;
+ unreachable("Unrecognized blorp format");
}
- if (brw->intel.gen > 6) {
+ if (brw->gen > 6) {
/* Gen7's rendering hardware only supports the IMS layout for depth and
* stencil render targets. Blorp always maps its destination surface as
* a color render target (even if it's actually a depth or stencil
wm_prog_key.persample_msaa_dispatch = true;
}
- /* The render path must be configured to use the same number of samples as
- * the destination buffer.
+ /* Scaled blitting or not. */
+ wm_prog_key.blit_scaled =
+ ((dst_x1 - dst_x0) == (src_x1 - src_x0) &&
+ (dst_y1 - dst_y0) == (src_y1 - src_y0)) ? false : true;
+
+ /* Scaling factors used for bilinear filtering in multisample scaled
+ * blits.
*/
- num_samples = dst.num_samples;
+ wm_prog_key.x_scale = 2.0f;
+ wm_prog_key.y_scale = src_mt->num_samples / 2.0f;
+
+ if (filter == GL_LINEAR && src.num_samples <= 1 && dst.num_samples <= 1)
+ wm_prog_key.bilinear_filter = true;
GLenum base_format = _mesa_get_format_base_format(src_mt->format);
if (base_format != GL_DEPTH_COMPONENT && /* TODO: what about depth/stencil? */
wm_prog_key.src_tiled_w = src.map_stencil_as_y_tiled;
wm_prog_key.dst_tiled_w = dst.map_stencil_as_y_tiled;
- x0 = wm_push_consts.dst_x0 = dst_x0;
- y0 = wm_push_consts.dst_y0 = dst_y0;
- x1 = wm_push_consts.dst_x1 = dst_x1;
- y1 = wm_push_consts.dst_y1 = dst_y1;
- wm_push_consts.x_transform.setup(src_x0, dst_x0, dst_x1, mirror_x);
- wm_push_consts.y_transform.setup(src_y0, dst_y0, dst_y1, mirror_y);
+ /* Round floating point values to nearest integer to avoid "off by one texel"
+ * kind of errors when blitting.
+ */
+ x0 = wm_push_consts.dst_x0 = roundf(dst_x0);
+ y0 = wm_push_consts.dst_y0 = roundf(dst_y0);
+ x1 = wm_push_consts.dst_x1 = roundf(dst_x1);
+ y1 = wm_push_consts.dst_y1 = roundf(dst_y1);
+ wm_push_consts.rect_grid_x1 = (minify(src_mt->logical_width0, src_level) *
+ wm_prog_key.x_scale - 1.0f);
+ wm_push_consts.rect_grid_y1 = (minify(src_mt->logical_height0, src_level) *
+ wm_prog_key.y_scale - 1.0f);
+
+ wm_push_consts.x_transform.setup(src_x0, src_x1, dst_x0, dst_x1, mirror_x);
+ wm_push_consts.y_transform.setup(src_y0, src_y1, dst_y0, dst_y1, mirror_y);
if (dst.num_samples <= 1 && dst_mt->num_samples > 1) {
/* We must expand the rectangle we send through the rendering pipeline,
y1 = ALIGN(y1 * 2, 4);
break;
default:
- assert(!"Unrecognized sample count in brw_blorp_blit_params ctor");
- break;
+ unreachable("Unrecognized sample count in brw_blorp_blit_params ctor");
}
wm_prog_key.use_kill = true;
}
if (dst.map_stencil_as_y_tiled) {
/* We must modify the rectangle we send through the rendering pipeline
- * (and the size of the destination surface), to account for the fact
- * that we are mapping it as Y-tiled when it is in fact W-tiled.
+ * (and the size and x/y offset of the destination surface), to account
+ * for the fact that we are mapping it as Y-tiled when it is in fact
+ * W-tiled.
*
* Both Y tiling and W tiling can be understood as organizations of
* 32-byte sub-tiles; within each 32-byte sub-tile, the layout of pixels
* coordinates of its edges are multiples of 4 (the W sub-tile height).
* Then we need to scale the X and Y coordinates of the rectangle to
* account for the differences in aspect ratio between the Y and W
- * sub-tiles.
+ * sub-tiles. We need to modify the layer width and height similarly.
+ *
+ * A correction needs to be applied when MSAA is in use: since
+ * INTEL_MSAA_LAYOUT_IMS uses an interleaving pattern whose height is 4,
+ * we need to align the Y coordinates to multiples of 8, so that when
+ * they are divided by two they are still multiples of 4.
+ *
+ * Note: Since the x/y offset of the surface will be applied using the
+ * SURFACE_STATE command packet, it will be invisible to the swizzling
+ * code in the shader; therefore it needs to be in a multiple of the
+ * 32-byte sub-tile size. Fortunately it is, since the sub-tile is 8
+ * pixels wide and 4 pixels high (when viewed as a W-tiled stencil
+ * buffer), and the miplevel alignment used for stencil buffers is 8
+ * pixels horizontally and either 4 or 8 pixels vertically (see
+ * intel_horizontal_texture_alignment_unit() and
+ * intel_vertical_texture_alignment_unit()).
+ *
+ * Note: Also, since the SURFACE_STATE command packet can only apply
+ * offsets that are multiples of 4 pixels horizontally and 2 pixels
+ * vertically, it is important that the offsets will be multiples of
+ * these sizes after they are converted into Y-tiled coordinates.
+ * Fortunately they will be, since we know from above that the offsets
+ * are a multiple of the 32-byte sub-tile size, and in Y-tiled
+ * coordinates the sub-tile is 16 pixels wide and 2 pixels high.
*
* TODO: what if this makes the coordinates (or the texture size) too
* large?
*/
- const unsigned x_align = 8, y_align = 4;
+ const unsigned x_align = 8, y_align = dst.num_samples != 0 ? 8 : 4;
x0 = ROUND_DOWN_TO(x0, x_align) * 2;
y0 = ROUND_DOWN_TO(y0, y_align) / 2;
x1 = ALIGN(x1, x_align) * 2;
y1 = ALIGN(y1, y_align) / 2;
- dst.width *= 2;
- dst.height /= 2;
+ dst.width = ALIGN(dst.width, x_align) * 2;
+ dst.height = ALIGN(dst.height, y_align) / 2;
+ dst.x_offset *= 2;
+ dst.y_offset /= 2;
wm_prog_key.use_kill = true;
}
if (src.map_stencil_as_y_tiled) {
- /* We must modify the size of the source surface to account for the fact
- * that we are mapping it as Y-tiled when it is in fact W tiled.
+ /* We must modify the size and x/y offset of the source surface to
+ * account for the fact that we are mapping it as Y-tiled when it is in
+ * fact W tiled.
+ *
+ * See the comments above concerning x/y offset alignment for the
+ * destination surface.
*
* TODO: what if this makes the texture size too large?
*/
- src.width *= 2;
- src.height /= 2;
+ const unsigned x_align = 8, y_align = src.num_samples != 0 ? 8 : 4;
+ src.width = ALIGN(src.width, x_align) * 2;
+ src.height = ALIGN(src.height, y_align) / 2;
+ src.x_offset *= 2;
+ src.y_offset /= 2;
}
}
brw_blorp_blit_params::get_wm_prog(struct brw_context *brw,
brw_blorp_prog_data **prog_data) const
{
- uint32_t prog_offset;
- if (!brw_search_cache(&brw->cache, BRW_BLORP_BLIT_PROG,
+ uint32_t prog_offset = 0;
+ if (!brw_search_cache(&brw->cache, BRW_CACHE_BLORP_BLIT_PROG,
&this->wm_prog_key, sizeof(this->wm_prog_key),
&prog_offset, prog_data)) {
- brw_blorp_blit_program prog(brw, &this->wm_prog_key);
+ brw_blorp_blit_program prog(brw, &this->wm_prog_key,
+ INTEL_DEBUG & DEBUG_BLORP);
GLuint program_size;
const GLuint *program = prog.compile(brw, &program_size);
- brw_upload_cache(&brw->cache, BRW_BLORP_BLIT_PROG,
+ brw_upload_cache(&brw->cache, BRW_CACHE_BLORP_BLIT_PROG,
&this->wm_prog_key, sizeof(this->wm_prog_key),
program, program_size,
&prog.prog_data, sizeof(prog.prog_data),
projects / mesa.git / blobdiff