#include "main/context.h"
#include "main/objectlabel.h"
#include "main/shaderapi.h"
+#include "main/shaderobj.h"
#include "main/arrayobj.h"
#include "main/bufferobj.h"
#include "main/buffers.h"
#include "main/varray.h"
#include "main/uniforms.h"
#include "main/fbobject.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
#include "main/texobj.h"
#include "main/api_validate.h"
#include "main/state.h"
+#include "util/format_srgb.h"
+
#include "vbo/vbo_context.h"
#include "drivers/common/meta.h"
#include "brw_defines.h"
#include "brw_context.h"
#include "brw_draw.h"
+#include "brw_state.h"
#include "intel_fbo.h"
#include "intel_batchbuffer.h"
#include "brw_blorp.h"
+#include "brw_meta_util.h"
struct brw_fast_clear_state {
struct gl_buffer_object *buf_obj;
struct gl_vertex_array_object *array_obj;
+ struct gl_shader_program *shader_prog;
GLuint vao;
- GLuint shader_prog;
GLint color_location;
};
const char *vs_source =
"#extension GL_AMD_vertex_shader_layer : enable\n"
"#extension GL_ARB_draw_instanced : enable\n"
- "attribute vec4 position;\n"
+ "#extension GL_ARB_explicit_attrib_location : enable\n"
+ "layout(location = 0) in vec4 position;\n"
"uniform int layer;\n"
"void main()\n"
"{\n"
" gl_FragColor = color;\n"
"}\n";
- GLuint vs, fs;
struct brw_fast_clear_state *clear = brw->fast_clear_state;
struct gl_context *ctx = &brw->ctx;
if (clear->shader_prog) {
- _mesa_UseProgram(clear->shader_prog);
+ _mesa_meta_use_program(ctx, clear->shader_prog);
_mesa_Uniform4fv(clear->color_location, 1, color);
return;
}
- vs = _mesa_meta_compile_shader_with_debug(ctx, GL_VERTEX_SHADER, vs_source);
- fs = _mesa_meta_compile_shader_with_debug(ctx, GL_FRAGMENT_SHADER, fs_source);
-
- clear->shader_prog = _mesa_CreateProgram();
- _mesa_AttachShader(clear->shader_prog, fs);
- _mesa_DeleteShader(fs);
- _mesa_AttachShader(clear->shader_prog, vs);
- _mesa_DeleteShader(vs);
- _mesa_BindAttribLocation(clear->shader_prog, 0, "position");
- _mesa_ObjectLabel(GL_PROGRAM, clear->shader_prog, -1, "meta repclear");
- _mesa_LinkProgram(clear->shader_prog);
+ _mesa_meta_compile_and_link_program(ctx, vs_source, fs_source,
+ "meta repclear",
+ &clear->shader_prog);
clear->color_location =
- _mesa_GetUniformLocation(clear->shader_prog, "color");
+ _mesa_program_resource_location(clear->shader_prog, GL_UNIFORM, "color");
- _mesa_UseProgram(clear->shader_prog);
+ _mesa_meta_use_program(ctx, clear->shader_prog);
_mesa_Uniform4fv(clear->color_location, 1, color);
}
_mesa_DeleteVertexArrays(1, &clear->vao);
_mesa_reference_buffer_object(&brw->ctx, &clear->buf_obj, NULL);
- _mesa_DeleteProgram(clear->shader_prog);
+ _mesa_reference_shader_program(&brw->ctx, &clear->shader_prog, NULL);
free(clear);
if (old_context)
}
struct rect {
- int x0, y0, x1, y1;
+ unsigned x0, y0, x1, y1;
};
static void
NULL, 0, NULL);
}
-static void
-get_fast_clear_rect(struct brw_context *brw, struct gl_framebuffer *fb,
- struct intel_renderbuffer *irb, struct rect *rect)
+void
+brw_get_fast_clear_rect(const struct brw_context *brw,
+ const struct gl_framebuffer *fb,
+ const struct intel_mipmap_tree* mt,
+ unsigned *x0, unsigned *y0,
+ unsigned *x1, unsigned *y1)
{
unsigned int x_align, y_align;
unsigned int x_scaledown, y_scaledown;
- if (irb->mt->msaa_layout == INTEL_MSAA_LAYOUT_NONE) {
+ /* Only single sampled surfaces need to (and actually can) be resolved. */
+ if (mt->msaa_layout == INTEL_MSAA_LAYOUT_NONE ||
+ intel_miptree_is_lossless_compressed(brw, mt)) {
/* From the Ivy Bridge PRM, Vol2 Part1 11.7 "MCS Buffer for Render
* Target(s)", beneath the "Fast Color Clear" bullet (p327):
*
* alignment size returned by intel_get_non_msrt_mcs_alignment(), but
* with X alignment multiplied by 16 and Y alignment multiplied by 32.
*/
- intel_get_non_msrt_mcs_alignment(irb->mt, &x_align, &y_align);
+ intel_get_non_msrt_mcs_alignment(mt, &x_align, &y_align);
x_align *= 16;
/* SKL+ line alignment requirement for Y-tiled are half those of the prior
* vertically and either 4 or 16 horizontally, and the scaledown
* factor is 2 vertically and either 2 or 8 horizontally.
*/
- switch (irb->mt->num_samples) {
+ switch (mt->num_samples) {
case 2:
case 4:
x_scaledown = 8;
y_align = y_scaledown * 2;
}
- rect->x0 = fb->_Xmin;
- rect->x1 = fb->_Xmax;
+ *x0 = fb->_Xmin;
+ *x1 = fb->_Xmax;
if (fb->Name != 0) {
- rect->y0 = fb->_Ymin;
- rect->y1 = fb->_Ymax;
+ *y0 = fb->_Ymin;
+ *y1 = fb->_Ymax;
} else {
- rect->y0 = fb->Height - fb->_Ymax;
- rect->y1 = fb->Height - fb->_Ymin;
+ *y0 = fb->Height - fb->_Ymax;
+ *y1 = fb->Height - fb->_Ymin;
}
- rect->x0 = ROUND_DOWN_TO(rect->x0, x_align) / x_scaledown;
- rect->y0 = ROUND_DOWN_TO(rect->y0, y_align) / y_scaledown;
- rect->x1 = ALIGN(rect->x1, x_align) / x_scaledown;
- rect->y1 = ALIGN(rect->y1, y_align) / y_scaledown;
+ *x0 = ROUND_DOWN_TO(*x0, x_align) / x_scaledown;
+ *y0 = ROUND_DOWN_TO(*y0, y_align) / y_scaledown;
+ *x1 = ALIGN(*x1, x_align) / x_scaledown;
+ *y1 = ALIGN(*y1, y_align) / y_scaledown;
}
-static void
-get_buffer_rect(const struct gl_framebuffer *fb, struct rect *rect)
+void
+brw_meta_get_buffer_rect(const struct gl_framebuffer *fb,
+ unsigned *x0, unsigned *y0,
+ unsigned *x1, unsigned *y1)
{
- rect->x0 = fb->_Xmin;
- rect->x1 = fb->_Xmax;
+ *x0 = fb->_Xmin;
+ *x1 = fb->_Xmax;
if (fb->Name != 0) {
- rect->y0 = fb->_Ymin;
- rect->y1 = fb->_Ymax;
+ *y0 = fb->_Ymin;
+ *y1 = fb->_Ymax;
} else {
- rect->y0 = fb->Height - fb->_Ymax;
- rect->y1 = fb->Height - fb->_Ymin;
+ *y0 = fb->Height - fb->_Ymax;
+ *y1 = fb->Height - fb->_Ymin;
}
}
* Fast color clear can only clear to color values of 1.0 or 0.0. At the
* moment we only support floating point, unorm, and snorm buffers.
*/
-static bool
-is_color_fast_clear_compatible(struct brw_context *brw,
- mesa_format format,
- const union gl_color_union *color)
+bool
+brw_is_color_fast_clear_compatible(struct brw_context *brw,
+ const struct intel_mipmap_tree *mt,
+ const union gl_color_union *color)
{
+ const struct gl_context *ctx = &brw->ctx;
+
+ /* If we're mapping the render format to a different format than the
+ * format we use for texturing then it is a bit questionable whether it
+ * should be possible to use a fast clear. Although we only actually
+ * render using a renderable format, without the override workaround it
+ * wouldn't be possible to have a non-renderable surface in a fast clear
+ * state so the hardware probably legitimately doesn't need to support
+ * this case. At least on Gen9 this really does seem to cause problems.
+ */
+ if (brw->gen >= 9 &&
+ brw_format_for_mesa_format(mt->format) !=
+ brw->render_target_format[mt->format])
+ return false;
+
+ /* Gen9 doesn't support fast clear on single-sampled SRGB buffers. When
+ * GL_FRAMEBUFFER_SRGB is enabled any color renderbuffers will be
+ * resolved in intel_update_state. In that case it's pointless to do a
+ * fast clear because it's very likely to be immediately resolved.
+ */
+ if (brw->gen >= 9 &&
+ mt->num_samples <= 1 &&
+ ctx->Color.sRGBEnabled &&
+ _mesa_get_srgb_format_linear(mt->format) != mt->format)
+ return false;
+
+ const mesa_format format = _mesa_get_render_format(ctx, mt->format);
if (_mesa_is_format_integer_color(format)) {
if (brw->gen >= 8) {
perf_debug("Integer fast clear not enabled for (%s)",
* Convert the given color to a bitfield suitable for ORing into DWORD 7 of
* SURFACE_STATE (DWORD 12-15 on SKL+).
*/
-static void
-set_fast_clear_color(struct brw_context *brw,
- struct intel_mipmap_tree *mt,
- const union gl_color_union *color)
+void
+brw_meta_set_fast_clear_color(struct brw_context *brw,
+ struct intel_mipmap_tree *mt,
+ const union gl_color_union *color)
{
union gl_color_union override_color = *color;
override_color.f[3] = 1.0f;
}
+ /* Handle linear→SRGB conversion */
+ if (brw->ctx.Color.sRGBEnabled &&
+ _mesa_get_srgb_format_linear(mt->format) != mt->format) {
+ for (int i = 0; i < 3; i++) {
+ override_color.f[i] =
+ util_format_linear_to_srgb_float(override_color.f[i]);
+ }
+ }
+
if (brw->gen >= 9) {
mt->gen9_fast_clear_color = override_color;
} else {
uint32_t fast_clear_buffers,
struct rect fast_clear_rect)
{
- assert(brw->gen >= 9);
struct gl_context *ctx = &brw->ctx;
+ const bool srgb_enabled = ctx->Color.sRGBEnabled;
+
+ assert(brw->gen >= 9);
+
+ /* Make sure the GL_FRAMEBUFFER_SRGB is disabled during fast clear so that
+ * the surface state will always be uploaded with a linear buffer. SRGB
+ * buffers are not supported on Gen9 because they are not marked as
+ * losslessly compressible. This shouldn't matter for the fast clear
+ * because the color is not written to the framebuffer yet so the hardware
+ * doesn't need to do any SRGB conversion.
+ */
+ if (srgb_enabled)
+ _mesa_set_framebuffer_srgb(ctx, GL_FALSE);
brw_bind_rep_write_shader(brw, (float *) fast_clear_color);
_mesa_meta_drawbuffers_from_bitfield(1 << index);
- brw_draw_rectlist(ctx, &fast_clear_rect, MAX2(1, fb->MaxNumLayers));
+ brw_draw_rectlist(brw, &fast_clear_rect, MAX2(1, fb->MaxNumLayers));
/* Now set the mcs we cleared to INTEL_FAST_CLEAR_STATE_CLEAR so we'll
* resolve them eventually.
}
set_fast_clear_op(brw, 0);
+
+ if (srgb_enabled)
+ _mesa_set_framebuffer_srgb(ctx, GL_TRUE);
}
bool
GLbitfield buffers, bool partial_clear)
{
struct gl_context *ctx = &brw->ctx;
- mesa_format format;
enum { FAST_CLEAR, REP_CLEAR, PLAIN_CLEAR } clear_type;
GLbitfield plain_clear_buffers, meta_save, rep_clear_buffers, fast_clear_buffers;
struct rect fast_clear_rect, clear_rect;
if (brw->gen < 7)
clear_type = REP_CLEAR;
- /* Certain formats have unresolved issues with sampling from the MCS
- * buffer on Gen9. This disables fast clears altogether for MSRTs until
- * we can figure out what's going on.
- */
- if (brw->gen >= 9 && irb->mt->num_samples > 1)
- clear_type = REP_CLEAR;
-
if (irb->mt->fast_clear_state == INTEL_FAST_CLEAR_STATE_NO_MCS)
clear_type = REP_CLEAR;
/* Fast clear is only supported for colors where all components are
* either 0 or 1.
*/
- format = _mesa_get_render_format(ctx, irb->mt->format);
- if (!is_color_fast_clear_compatible(brw, format, &ctx->Color.ClearColor))
+ if (!brw_is_color_fast_clear_compatible(brw, irb->mt,
+ &ctx->Color.ClearColor))
clear_type = REP_CLEAR;
/* From the SNB PRM (Vol4_Part1):
GLubyte *color_mask = ctx->Color.ColorMask[buf];
for (int i = 0; i < 4; i++) {
if (_mesa_format_has_color_component(irb->mt->format, i) &&
+ !(i == 3 && irb->Base.Base._BaseFormat == GL_RGB) &&
!color_mask[i]) {
perf_debug("Falling back to plain clear on %dx%d buffer because of color mask\n",
irb->mt->logical_width0, irb->mt->logical_height0);
switch (clear_type) {
case FAST_CLEAR:
- set_fast_clear_color(brw, irb->mt, &ctx->Color.ClearColor);
+ brw_meta_set_fast_clear_color(brw, irb->mt, &ctx->Color.ClearColor);
irb->need_downsample = true;
/* If the buffer is already in INTEL_FAST_CLEAR_STATE_CLEAR, the
irb->mt->fast_clear_state = INTEL_FAST_CLEAR_STATE_RESOLVED;
irb->need_downsample = true;
fast_clear_buffers |= 1 << index;
- get_fast_clear_rect(brw, fb, irb, &fast_clear_rect);
+ brw_get_fast_clear_rect(brw, fb, irb->mt,
+ &fast_clear_rect.x0, &fast_clear_rect.y0,
+ &fast_clear_rect.x1, &fast_clear_rect.y1);
break;
case REP_CLEAR:
rep_clear_buffers |= 1 << index;
- get_buffer_rect(fb, &clear_rect);
+ brw_meta_get_buffer_rect(fb,
+ &clear_rect.x0, &clear_rect.y0,
+ &clear_rect.x1, &clear_rect.y1);
break;
case PLAIN_CLEAR:
plain_clear_buffers |= 1 << index;
- get_buffer_rect(fb, &clear_rect);
+ brw_meta_get_buffer_rect(fb,
+ &clear_rect.x0, &clear_rect.y0,
+ &clear_rect.x1, &clear_rect.y1);
continue;
}
}
return true;
}
-static void
-get_resolve_rect(struct brw_context *brw,
- struct intel_mipmap_tree *mt, struct rect *rect)
+void
+brw_get_resolve_rect(const struct brw_context *brw,
+ const struct intel_mipmap_tree *mt,
+ unsigned *x0, unsigned *y0,
+ unsigned *x1, unsigned *y1)
{
unsigned x_align, y_align;
unsigned x_scaledown, y_scaledown;
x_scaledown = x_align / 2;
y_scaledown = y_align / 2;
}
- rect->x0 = rect->y0 = 0;
- rect->x1 = ALIGN(mt->logical_width0, x_scaledown) / x_scaledown;
- rect->y1 = ALIGN(mt->logical_height0, y_scaledown) / y_scaledown;
+ *x0 = *y0 = 0;
+ *x1 = ALIGN(mt->logical_width0, x_scaledown) / x_scaledown;
+ *y1 = ALIGN(mt->logical_height0, y_scaledown) / y_scaledown;
}
void
struct intel_mipmap_tree *mt)
{
struct gl_context *ctx = &brw->ctx;
- GLuint fbo, rbo;
+ struct gl_framebuffer *drawFb;
+ struct gl_renderbuffer *rb;
struct rect rect;
brw_emit_mi_flush(brw);
+ drawFb = ctx->Driver.NewFramebuffer(ctx, 0xDEADBEEF);
+ if (drawFb == NULL) {
+ _mesa_error(ctx, GL_OUT_OF_MEMORY, "in %s", __func__);
+ return;
+ }
+
_mesa_meta_begin(ctx, MESA_META_ALL);
- _mesa_GenFramebuffers(1, &fbo);
- rbo = brw_get_rb_for_slice(brw, mt, 0, 0, false);
+ rb = brw_get_rb_for_slice(brw, mt, 0, 0, false);
- _mesa_BindFramebuffer(GL_DRAW_FRAMEBUFFER, fbo);
- _mesa_FramebufferRenderbuffer(GL_DRAW_FRAMEBUFFER,
- GL_COLOR_ATTACHMENT0,
- GL_RENDERBUFFER, rbo);
+ _mesa_bind_framebuffers(ctx, drawFb, ctx->ReadBuffer);
+ _mesa_framebuffer_renderbuffer(ctx, ctx->DrawBuffer, GL_COLOR_ATTACHMENT0,
+ rb);
_mesa_DrawBuffer(GL_COLOR_ATTACHMENT0);
brw_fast_clear_init(brw);
* bits to let us select the type of resolve. For fast clear resolves, it
* turns out we can use the same value as pre-SKL though.
*/
- set_fast_clear_op(brw, GEN7_PS_RENDER_TARGET_RESOLVE_ENABLE);
+ if (intel_miptree_is_lossless_compressed(brw, mt))
+ set_fast_clear_op(brw, GEN9_PS_RENDER_TARGET_RESOLVE_FULL);
+ else
+ set_fast_clear_op(brw, GEN7_PS_RENDER_TARGET_RESOLVE_ENABLE);
mt->fast_clear_state = INTEL_FAST_CLEAR_STATE_RESOLVED;
- get_resolve_rect(brw, mt, &rect);
+ brw_get_resolve_rect(brw, mt, &rect.x0, &rect.y0, &rect.x1, &rect.y1);
brw_draw_rectlist(brw, &rect, 1);
set_fast_clear_op(brw, 0);
use_rectlist(brw, false);
- _mesa_DeleteRenderbuffers(1, &rbo);
- _mesa_DeleteFramebuffers(1, &fbo);
+ _mesa_reference_renderbuffer(&rb, NULL);
+ _mesa_reference_framebuffer(&drawFb, NULL);
_mesa_meta_end(ctx);