**************************************************************************/
-#include <stdio.h>
-#include <errno.h>
+#include "main/mtypes.h"
+#include "main/context.h"
+#include "main/enums.h"
+#include "main/colormac.h"
+#include "main/fbobject.h"
-#include "mtypes.h"
-#include "context.h"
-#include "enums.h"
-#include "vblank.h"
-
-#include "intel_reg.h"
-#include "intel_batchbuffer.h"
-#include "intel_context.h"
+#include "brw_context.h"
#include "intel_blit.h"
+#include "intel_buffers.h"
+#include "intel_fbo.h"
+#include "intel_reg.h"
#include "intel_regions.h"
-#include "intel_structs.h"
-
-#include "bufmgr.h"
-
-
-
-
-/*
- * Copy the back buffer to the front buffer.
- */
-void intelCopyBuffer( const __DRIdrawablePrivate *dPriv,
- const drm_clip_rect_t *rect )
-{
- struct intel_context *intel;
- GLboolean missed_target;
- int64_t ust;
-
- DBG("%s\n", __FUNCTION__);
-
- assert(dPriv);
- assert(dPriv->driContextPriv);
- assert(dPriv->driContextPriv->driverPrivate);
-
- intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate;
- intelFlush( &intel->ctx );
-
-
- bmFinishFenceLock(intel, intel->last_swap_fence);
-
- /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets
- * should work regardless.
- */
- LOCK_HARDWARE( intel );
-
- if (!rect)
- {
- UNLOCK_HARDWARE( intel );
- driWaitForVBlank( dPriv, &dPriv->vblSeq, dPriv->vblFlags,
- &missed_target );
- LOCK_HARDWARE( intel );
- }
-
- {
- intelScreenPrivate *intelScreen = intel->intelScreen;
- __DRIdrawablePrivate *dPriv = intel->driDrawable;
- int nbox = dPriv->numClipRects;
- drm_clip_rect_t *pbox = dPriv->pClipRects;
- int cpp = intelScreen->cpp;
- struct intel_region *src, *dst;
- int BR13, CMD;
- int i;
- int src_pitch, dst_pitch;
-
- if (intel->sarea->pf_current_page == 0) {
- dst = intel->front_region;
- src = intel->back_region;
- }
- else {
- assert(0);
- src = intel->front_region;
- dst = intel->back_region;
- }
-
- src_pitch = src->pitch * src->cpp;
- dst_pitch = dst->pitch * dst->cpp;
-
- if (cpp == 2) {
- BR13 = (0xCC << 16) | (1<<24);
- CMD = XY_SRC_COPY_BLT_CMD;
- }
- else {
- BR13 = (0xCC << 16) | (1<<24) | (1<<25);
- CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
- }
-
- if (src->tiled) {
- CMD |= XY_SRC_TILED;
- src_pitch /= 4;
- }
-
- if (dst->tiled) {
- CMD |= XY_DST_TILED;
- dst_pitch /= 4;
- }
-
- for (i = 0 ; i < nbox; i++, pbox++)
- {
- drm_clip_rect_t tmp = *pbox;
-
- if (rect) {
- if (!intel_intersect_cliprects(&tmp, &tmp, rect))
- continue;
- }
-
-
- if (tmp.x1 > tmp.x2 ||
- tmp.y1 > tmp.y2 ||
- tmp.x2 > intelScreen->width ||
- tmp.y2 > intelScreen->height)
- continue;
-
- BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
- OUT_BATCH( CMD );
- OUT_BATCH( dst_pitch | BR13 );
- OUT_BATCH( (tmp.y1 << 16) | tmp.x1 );
- OUT_BATCH( (tmp.y2 << 16) | tmp.x2 );
- OUT_RELOC( dst->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, 0 );
- OUT_BATCH( (tmp.y1 << 16) | tmp.x1 );
- OUT_BATCH( src_pitch );
- OUT_RELOC( src->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, 0 );
- ADVANCE_BATCH();
- }
- }
-
- intel_batchbuffer_flush( intel->batch );
- intel->second_last_swap_fence = intel->last_swap_fence;
- intel->last_swap_fence = bmSetFenceLock( intel );
- UNLOCK_HARDWARE( intel );
-
- if (!rect)
- {
- intel->swap_count++;
- (*dri_interface->getUST)(&ust);
- if (missed_target) {
- intel->swap_missed_count++;
- intel->swap_missed_ust = ust - intel->swap_ust;
- }
-
- intel->swap_ust = ust;
- }
-
-}
-
-
-
-
-void intelEmitFillBlit( struct intel_context *intel,
- GLuint cpp,
- GLshort dst_pitch,
- struct buffer *dst_buffer,
- GLuint dst_offset,
- GLboolean dst_tiled,
- GLshort x, GLshort y,
- GLshort w, GLshort h,
- GLuint color )
-{
- GLuint BR13, CMD;
- BATCH_LOCALS;
-
- dst_pitch *= cpp;
-
- switch(cpp) {
- case 1:
- case 2:
- case 3:
- BR13 = (0xF0 << 16) | (1<<24);
- CMD = XY_COLOR_BLT_CMD;
- break;
- case 4:
- BR13 = (0xF0 << 16) | (1<<24) | (1<<25);
- CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
- break;
- default:
- return;
- }
+#include "intel_batchbuffer.h"
+#include "intel_mipmap_tree.h"
- if (dst_tiled) {
- CMD |= XY_DST_TILED;
- dst_pitch /= 4;
- }
+#define FILE_DEBUG_FLAG DEBUG_BLIT
- BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS);
- OUT_BATCH( CMD );
- OUT_BATCH( dst_pitch | BR13 );
- OUT_BATCH( (y << 16) | x );
- OUT_BATCH( ((y+h) << 16) | (x+w) );
- OUT_RELOC( dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, dst_offset );
- OUT_BATCH( color );
- ADVANCE_BATCH();
-}
+static void
+intel_miptree_set_alpha_to_one(struct brw_context *brw,
+ struct intel_mipmap_tree *mt,
+ int x, int y, int width, int height);
static GLuint translate_raster_op(GLenum logicop)
{
}
}
-
-/* Copy BitBlt
- */
-void intelEmitCopyBlit( struct intel_context *intel,
- GLuint cpp,
- GLshort src_pitch,
- struct buffer *src_buffer,
- GLuint src_offset,
- GLboolean src_tiled,
- GLshort dst_pitch,
- struct buffer *dst_buffer,
- GLuint dst_offset,
- GLboolean dst_tiled,
- GLshort src_x, GLshort src_y,
- GLshort dst_x, GLshort dst_y,
- GLshort w, GLshort h,
- GLenum logic_op )
+static uint32_t
+br13_for_cpp(int cpp)
{
- GLuint CMD, BR13;
- int dst_y2 = dst_y + h;
- int dst_x2 = dst_x + w;
- BATCH_LOCALS;
-
-
- DBG("%s src:buf(%d)/%d %d,%d dst:buf(%d)/%d %d,%d sz:%dx%d op:%d\n",
- __FUNCTION__,
- src_buffer, src_pitch, src_x, src_y,
- dst_buffer, dst_pitch, dst_x, dst_y,
- w,h,logic_op);
-
- assert( logic_op - GL_CLEAR >= 0 );
- assert( logic_op - GL_CLEAR < 0x10 );
-
- src_pitch *= cpp;
- dst_pitch *= cpp;
-
- switch(cpp) {
- case 1:
- case 2:
- case 3:
- BR13 = (translate_raster_op(logic_op) << 16) | (1<<24);
- CMD = XY_SRC_COPY_BLT_CMD;
- break;
+ switch (cpp) {
case 4:
- BR13 = (translate_raster_op(logic_op) << 16) | (1<<24) |
- (1<<25);
- CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
+ return BR13_8888;
+ break;
+ case 2:
+ return BR13_565;
+ break;
+ case 1:
+ return BR13_8;
break;
default:
- return;
+ assert(0);
+ return 0;
}
+}
- if (src_tiled) {
- CMD |= XY_SRC_TILED;
- src_pitch /= 4;
- }
-
- if (dst_tiled) {
- CMD |= XY_DST_TILED;
- dst_pitch /= 4;
- }
+/**
+ * Emits the packet for switching the blitter from X to Y tiled or back.
+ *
+ * This has to be called in a single BEGIN_BATCH_BLT_TILED() /
+ * ADVANCE_BATCH_TILED(). This is because BCS_SWCTRL is saved and restored as
+ * part of the power context, not a render context, and if the batchbuffer was
+ * to get flushed between setting and blitting, or blitting and restoring, our
+ * tiling state would leak into other unsuspecting applications (like the X
+ * server).
+ */
+static void
+set_blitter_tiling(struct brw_context *brw,
+ bool dst_y_tiled, bool src_y_tiled)
+{
+ assert(brw->gen >= 6);
+
+ /* Idle the blitter before we update how tiling is interpreted. */
+ OUT_BATCH(MI_FLUSH_DW);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+
+ OUT_BATCH(MI_LOAD_REGISTER_IMM | (3 - 2));
+ OUT_BATCH(BCS_SWCTRL);
+ OUT_BATCH((BCS_SWCTRL_DST_Y | BCS_SWCTRL_SRC_Y) << 16 |
+ (dst_y_tiled ? BCS_SWCTRL_DST_Y : 0) |
+ (src_y_tiled ? BCS_SWCTRL_SRC_Y : 0));
+}
- if (dst_y2 < dst_y ||
- dst_x2 < dst_x) {
- return;
- }
+#define BEGIN_BATCH_BLT_TILED(n, dst_y_tiled, src_y_tiled) do { \
+ BEGIN_BATCH_BLT(n + ((dst_y_tiled || src_y_tiled) ? 14 : 0)); \
+ if (dst_y_tiled || src_y_tiled) \
+ set_blitter_tiling(brw, dst_y_tiled, src_y_tiled); \
+ } while (0)
+
+#define ADVANCE_BATCH_TILED(dst_y_tiled, src_y_tiled) do { \
+ if (dst_y_tiled || src_y_tiled) \
+ set_blitter_tiling(brw, false, false); \
+ ADVANCE_BATCH(); \
+ } while (0)
+
+/**
+ * Implements a rectangular block transfer (blit) of pixels between two
+ * miptrees.
+ *
+ * Our blitter can operate on 1, 2, or 4-byte-per-pixel data, with generous,
+ * but limited, pitches and sizes allowed.
+ *
+ * The src/dst coordinates are relative to the given level/slice of the
+ * miptree.
+ *
+ * If @src_flip or @dst_flip is set, then the rectangle within that miptree
+ * will be inverted (including scanline order) when copying. This is common
+ * in GL when copying between window system and user-created
+ * renderbuffers/textures.
+ */
+bool
+intel_miptree_blit(struct brw_context *brw,
+ struct intel_mipmap_tree *src_mt,
+ int src_level, int src_slice,
+ uint32_t src_x, uint32_t src_y, bool src_flip,
+ struct intel_mipmap_tree *dst_mt,
+ int dst_level, int dst_slice,
+ uint32_t dst_x, uint32_t dst_y, bool dst_flip,
+ uint32_t width, uint32_t height,
+ GLenum logicop)
+{
+ /* No sRGB decode or encode is done by the hardware blitter, which is
+ * consistent with what we want in the callers (glCopyTexSubImage(),
+ * glBlitFramebuffer(), texture validation, etc.).
+ */
+ gl_format src_format = _mesa_get_srgb_format_linear(src_mt->format);
+ gl_format dst_format = _mesa_get_srgb_format_linear(dst_mt->format);
- dst_pitch &= 0xffff;
- src_pitch &= 0xffff;
+ /* The blitter doesn't support doing any format conversions. We do also
+ * support blitting ARGB8888 to XRGB8888 (trivial, the values dropped into
+ * the X channel don't matter), and XRGB8888 to ARGB8888 by setting the A
+ * channel to 1.0 at the end.
+ */
+ if (src_format != dst_format &&
+ ((src_format != MESA_FORMAT_ARGB8888 &&
+ src_format != MESA_FORMAT_XRGB8888) ||
+ (dst_format != MESA_FORMAT_ARGB8888 &&
+ dst_format != MESA_FORMAT_XRGB8888))) {
+ perf_debug("%s: Can't use hardware blitter from %s to %s, "
+ "falling back.\n", __FUNCTION__,
+ _mesa_get_format_name(src_format),
+ _mesa_get_format_name(dst_format));
+ return false;
+ }
- /* Initial y values don't seem to work with negative pitches. If
- * we adjust the offsets manually (below), it seems to work fine.
+ /* According to the Ivy Bridge PRM, Vol1 Part4, section 1.2.1.2 (Graphics
+ * Data Size Limitations):
+ *
+ * The BLT engine is capable of transferring very large quantities of
+ * graphics data. Any graphics data read from and written to the
+ * destination is permitted to represent a number of pixels that
+ * occupies up to 65,536 scan lines and up to 32,768 bytes per scan line
+ * at the destination. The maximum number of pixels that may be
+ * represented per scan line’s worth of graphics data depends on the
+ * color depth.
+ *
+ * Furthermore, intelEmitCopyBlit (which is called below) uses a signed
+ * 16-bit integer to represent buffer pitch, so it can only handle buffer
+ * pitches < 32k.
*
- * On the other hand, if we always adjust, the hardware doesn't
- * know which blit directions to use, so overlapping copypixels get
- * the wrong result.
+ * As a result of these two limitations, we can only use the blitter to do
+ * this copy when the region's pitch is less than 32k.
*/
- if (dst_pitch > 0 && src_pitch > 0) {
- BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
- OUT_BATCH( CMD );
- OUT_BATCH( dst_pitch | BR13 );
- OUT_BATCH( (dst_y << 16) | dst_x );
- OUT_BATCH( (dst_y2 << 16) | dst_x2 );
- OUT_RELOC( dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE,
- dst_offset );
- OUT_BATCH( (src_y << 16) | src_x );
- OUT_BATCH( src_pitch );
- OUT_RELOC( src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ,
- src_offset );
- ADVANCE_BATCH();
+ if (src_mt->region->pitch > 32768 ||
+ dst_mt->region->pitch > 32768) {
+ perf_debug("Falling back due to >32k pitch\n");
+ return false;
}
- else {
- BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
- OUT_BATCH( CMD );
- OUT_BATCH( (dst_pitch & 0xffff) | BR13 );
- OUT_BATCH( (0 << 16) | dst_x );
- OUT_BATCH( (h << 16) | dst_x2 );
- OUT_RELOC( dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE,
- dst_offset + dst_y * dst_pitch );
- OUT_BATCH( (src_pitch & 0xffff) );
- OUT_RELOC( src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ,
- src_offset + src_y * src_pitch );
- ADVANCE_BATCH();
+
+ /* The blitter has no idea about HiZ or fast color clears, so we need to
+ * resolve the miptrees before we do anything.
+ */
+ intel_miptree_slice_resolve_depth(brw, src_mt, src_level, src_slice);
+ intel_miptree_slice_resolve_depth(brw, dst_mt, dst_level, dst_slice);
+ intel_miptree_resolve_color(brw, src_mt);
+ intel_miptree_resolve_color(brw, dst_mt);
+
+ if (src_flip)
+ src_y = src_mt->level[src_level].height - src_y - height;
+
+ if (dst_flip)
+ dst_y = dst_mt->level[dst_level].height - dst_y - height;
+
+ int src_pitch = src_mt->region->pitch;
+ if (src_flip != dst_flip)
+ src_pitch = -src_pitch;
+
+ uint32_t src_image_x, src_image_y;
+ intel_miptree_get_image_offset(src_mt, src_level, src_slice,
+ &src_image_x, &src_image_y);
+ src_x += src_image_x;
+ src_y += src_image_y;
+
+ uint32_t dst_image_x, dst_image_y;
+ intel_miptree_get_image_offset(dst_mt, dst_level, dst_slice,
+ &dst_image_x, &dst_image_y);
+ dst_x += dst_image_x;
+ dst_y += dst_image_y;
+
+ if (!intelEmitCopyBlit(brw,
+ src_mt->cpp,
+ src_pitch,
+ src_mt->region->bo, src_mt->offset,
+ src_mt->region->tiling,
+ dst_mt->region->pitch,
+ dst_mt->region->bo, dst_mt->offset,
+ dst_mt->region->tiling,
+ src_x, src_y,
+ dst_x, dst_y,
+ width, height,
+ logicop)) {
+ return false;
}
-}
+ if (src_mt->format == MESA_FORMAT_XRGB8888 &&
+ dst_mt->format == MESA_FORMAT_ARGB8888) {
+ intel_miptree_set_alpha_to_one(brw, dst_mt,
+ dst_x, dst_y,
+ width, height);
+ }
+ return true;
+}
-void intelClearWithBlit(GLcontext *ctx, GLbitfield flags)
+/* Copy BitBlt
+ */
+bool
+intelEmitCopyBlit(struct brw_context *brw,
+ GLuint cpp,
+ GLshort src_pitch,
+ drm_intel_bo *src_buffer,
+ GLuint src_offset,
+ uint32_t src_tiling,
+ GLshort dst_pitch,
+ drm_intel_bo *dst_buffer,
+ GLuint dst_offset,
+ uint32_t dst_tiling,
+ GLshort src_x, GLshort src_y,
+ GLshort dst_x, GLshort dst_y,
+ GLshort w, GLshort h,
+ GLenum logic_op)
{
- struct intel_context *intel = intel_context( ctx );
- intelScreenPrivate *intelScreen = intel->intelScreen;
- GLuint clear_depth, clear_color;
- GLint cx, cy, cw, ch;
- GLint cpp = intelScreen->cpp;
- GLboolean all;
- GLint i;
- struct intel_region *front = intel->front_region;
- struct intel_region *back = intel->back_region;
- struct intel_region *depth = intel->depth_region;
- GLuint BR13, FRONT_CMD, BACK_CMD, DEPTH_CMD;
- GLuint front_pitch;
- GLuint back_pitch;
- GLuint depth_pitch;
- BATCH_LOCALS;
-
-
- clear_color = intel->ClearColor;
- clear_depth = 0;
-
- if (flags & BUFFER_BIT_DEPTH) {
- clear_depth = (GLuint)(ctx->Depth.Clear * intel->ClearDepth);
+ GLuint CMD, BR13, pass = 0;
+ int dst_y2 = dst_y + h;
+ int dst_x2 = dst_x + w;
+ drm_intel_bo *aper_array[3];
+ bool dst_y_tiled = dst_tiling == I915_TILING_Y;
+ bool src_y_tiled = src_tiling == I915_TILING_Y;
+
+ if (dst_tiling != I915_TILING_NONE) {
+ if (dst_offset & 4095)
+ return false;
}
+ if (src_tiling != I915_TILING_NONE) {
+ if (src_offset & 4095)
+ return false;
+ }
+ if ((dst_y_tiled || src_y_tiled) && brw->gen < 6)
+ return false;
+
+ /* do space check before going any further */
+ do {
+ aper_array[0] = brw->batch.bo;
+ aper_array[1] = dst_buffer;
+ aper_array[2] = src_buffer;
+
+ if (dri_bufmgr_check_aperture_space(aper_array, 3) != 0) {
+ intel_batchbuffer_flush(brw);
+ pass++;
+ } else
+ break;
+ } while (pass < 2);
+
+ if (pass >= 2)
+ return false;
+
+ intel_batchbuffer_require_space(brw, 8 * 4, true);
+ DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
+ __FUNCTION__,
+ src_buffer, src_pitch, src_offset, src_x, src_y,
+ dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h);
- if (flags & BUFFER_BIT_STENCIL) {
- clear_depth |= (ctx->Stencil.Clear & 0xff) << 24;
+ /* Blit pitch must be dword-aligned. Otherwise, the hardware appears to drop
+ * the low bits.
+ */
+ if (src_pitch % 4 != 0 || dst_pitch % 4 != 0)
+ return false;
+
+ /* For big formats (such as floating point), do the copy using 16 or 32bpp
+ * and multiply the coordinates.
+ */
+ if (cpp > 4) {
+ if (cpp % 4 == 2) {
+ dst_x *= cpp / 2;
+ dst_x2 *= cpp / 2;
+ src_x *= cpp / 2;
+ cpp = 2;
+ } else {
+ assert(cpp % 4 == 0);
+ dst_x *= cpp / 4;
+ dst_x2 *= cpp / 4;
+ src_x *= cpp / 4;
+ cpp = 4;
+ }
}
- switch(cpp) {
- case 2:
- BR13 = (0xF0 << 16) | (1<<24);
- BACK_CMD = FRONT_CMD = XY_COLOR_BLT_CMD;
- DEPTH_CMD = XY_COLOR_BLT_CMD;
+ BR13 = br13_for_cpp(cpp) | translate_raster_op(logic_op) << 16;
+
+ switch (cpp) {
+ case 1:
+ case 2:
+ CMD = XY_SRC_COPY_BLT_CMD;
break;
case 4:
- BR13 = (0xF0 << 16) | (1<<24) | (1<<25);
- BACK_CMD = FRONT_CMD = XY_COLOR_BLT_CMD |
- XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
- DEPTH_CMD = XY_COLOR_BLT_CMD;
- if (flags & BUFFER_BIT_DEPTH) DEPTH_CMD |= XY_BLT_WRITE_RGB;
- if (flags & BUFFER_BIT_STENCIL) DEPTH_CMD |= XY_BLT_WRITE_ALPHA;
+ CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
break;
default:
- return;
+ return false;
}
+ if (dst_tiling != I915_TILING_NONE) {
+ CMD |= XY_DST_TILED;
+ dst_pitch /= 4;
+ }
+ if (src_tiling != I915_TILING_NONE) {
+ CMD |= XY_SRC_TILED;
+ src_pitch /= 4;
+ }
+ if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
+ return true;
+ }
- intelFlush( &intel->ctx );
- LOCK_HARDWARE( intel );
- {
- /* get clear bounds after locking */
- cx = ctx->DrawBuffer->_Xmin;
- cy = ctx->DrawBuffer->_Ymin;
- ch = ctx->DrawBuffer->_Ymax - ctx->DrawBuffer->_Ymin;
- cw = ctx->DrawBuffer->_Xmax - ctx->DrawBuffer->_Xmin;
- all = (cw == ctx->DrawBuffer->Width && ch == ctx->DrawBuffer->Height);
-
- /* flip top to bottom */
- cy = intel->driDrawable->h - cy - ch;
- cx = cx + intel->drawX;
- cy += intel->drawY;
-
- /* adjust for page flipping */
- if ( intel->sarea->pf_current_page == 0 ) {
- front = intel->front_region;
- back = intel->back_region;
- }
- else {
- back = intel->front_region;
- front = intel->back_region;
- }
-
- front_pitch = front->pitch * front->cpp;
- back_pitch = back->pitch * back->cpp;
- depth_pitch = depth->pitch * depth->cpp;
-
- if (front->tiled) {
- FRONT_CMD |= XY_DST_TILED;
- front_pitch /= 4;
- }
+ assert(dst_x < dst_x2);
+ assert(dst_y < dst_y2);
- if (back->tiled) {
- BACK_CMD |= XY_DST_TILED;
- back_pitch /= 4;
- }
+ BEGIN_BATCH_BLT_TILED(8, dst_y_tiled, src_y_tiled);
- if (depth->tiled) {
- DEPTH_CMD |= XY_DST_TILED;
- depth_pitch /= 4;
- }
+ OUT_BATCH(CMD | (8 - 2));
+ OUT_BATCH(BR13 | (uint16_t)dst_pitch);
+ OUT_BATCH((dst_y << 16) | dst_x);
+ OUT_BATCH((dst_y2 << 16) | dst_x2);
+ OUT_RELOC_FENCED(dst_buffer,
+ I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+ dst_offset);
+ OUT_BATCH((src_y << 16) | src_x);
+ OUT_BATCH((uint16_t)src_pitch);
+ OUT_RELOC_FENCED(src_buffer,
+ I915_GEM_DOMAIN_RENDER, 0,
+ src_offset);
- for (i = 0 ; i < intel->numClipRects ; i++)
- {
- drm_clip_rect_t *box = &intel->pClipRects[i];
- drm_clip_rect_t b;
-
- if (!all) {
- GLint x = box->x1;
- GLint y = box->y1;
- GLint w = box->x2 - x;
- GLint h = box->y2 - y;
-
- if (x < cx) w -= cx - x, x = cx;
- if (y < cy) h -= cy - y, y = cy;
- if (x + w > cx + cw) w = cx + cw - x;
- if (y + h > cy + ch) h = cy + ch - y;
- if (w <= 0) continue;
- if (h <= 0) continue;
-
- b.x1 = x;
- b.y1 = y;
- b.x2 = x + w;
- b.y2 = y + h;
- } else {
- b = *box;
- }
-
-
- if (b.x1 > b.x2 ||
- b.y1 > b.y2 ||
- b.x2 > intelScreen->width ||
- b.y2 > intelScreen->height)
- continue;
-
- if ( flags & BUFFER_BIT_FRONT_LEFT ) {
- BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS);
- OUT_BATCH( FRONT_CMD );
- OUT_BATCH( front_pitch | BR13 );
- OUT_BATCH( (b.y1 << 16) | b.x1 );
- OUT_BATCH( (b.y2 << 16) | b.x2 );
- OUT_RELOC( front->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE,
- 0 );
- OUT_BATCH( clear_color );
- ADVANCE_BATCH();
- }
-
- if ( flags & BUFFER_BIT_BACK_LEFT ) {
- BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS);
- OUT_BATCH( BACK_CMD );
- OUT_BATCH( back_pitch | BR13 );
- OUT_BATCH( (b.y1 << 16) | b.x1 );
- OUT_BATCH( (b.y2 << 16) | b.x2 );
- OUT_RELOC( back->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE,
- 0 );
- OUT_BATCH( clear_color );
- ADVANCE_BATCH();
- }
-
- if ( flags & (BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH) ) {
- BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS);
- OUT_BATCH( DEPTH_CMD );
- OUT_BATCH( depth_pitch | BR13 );
- OUT_BATCH( (b.y1 << 16) | b.x1 );
- OUT_BATCH( (b.y2 << 16) | b.x2 );
- OUT_RELOC( depth->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE,
- 0 );
- OUT_BATCH( clear_depth );
- ADVANCE_BATCH();
- }
- }
- }
- intel_batchbuffer_flush( intel->batch );
- UNLOCK_HARDWARE( intel );
-}
+ ADVANCE_BATCH_TILED(dst_y_tiled, src_y_tiled);
+ intel_batchbuffer_emit_mi_flush(brw);
-void
-intelEmitImmediateColorExpandBlit(struct intel_context *intel,
+ return true;
+}
+
+bool
+intelEmitImmediateColorExpandBlit(struct brw_context *brw,
GLuint cpp,
GLubyte *src_bits, GLuint src_size,
GLuint fg_color,
GLshort dst_pitch,
- struct buffer *dst_buffer,
+ drm_intel_bo *dst_buffer,
GLuint dst_offset,
- GLboolean dst_tiled,
- GLshort x, GLshort y,
+ uint32_t dst_tiling,
+ GLshort x, GLshort y,
GLshort w, GLshort h,
GLenum logic_op)
{
- struct xy_text_immediate_blit text;
int dwords = ALIGN(src_size, 8) / 4;
- uint32_t opcode, br13;
+ uint32_t opcode, br13, blit_cmd;
+
+ if (dst_tiling != I915_TILING_NONE) {
+ if (dst_offset & 4095)
+ return false;
+ if (dst_tiling == I915_TILING_Y)
+ return false;
+ }
assert( logic_op - GL_CLEAR >= 0 );
assert( logic_op - GL_CLEAR < 0x10 );
+ assert(dst_pitch > 0);
- if (w < 0 || h < 0)
- return;
-
- dst_pitch *= cpp;
-
- if (dst_tiled)
- dst_pitch /= 4;
+ if (w < 0 || h < 0)
+ return true;
DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d, %d bytes %d dwords\n",
__FUNCTION__,
dst_buffer, dst_pitch, dst_offset, x, y, w, h, src_size, dwords);
- memset(&text, 0, sizeof(text));
- text.dw0.client = CLIENT_2D;
- text.dw0.opcode = OPCODE_XY_TEXT_IMMEDIATE_BLT;
- text.dw0.pad0 = 0;
- text.dw0.byte_packed = 1; /* ?maybe? */
- text.dw0.pad1 = 0;
- text.dw0.dst_tiled = dst_tiled;
- text.dw0.pad2 = 0;
- text.dw0.length = (sizeof(text)/sizeof(int)) - 2 + dwords;
- text.dw1.dest_y1 = y; /* duplicates info in setup blit */
- text.dw1.dest_x1 = x;
- text.dw2.dest_y2 = y + h;
- text.dw2.dest_x2 = x + w;
-
- intel_batchbuffer_require_space( intel->batch,
- (8 * 4) +
- sizeof(text) +
- dwords,
- INTEL_BATCH_NO_CLIPRECTS );
+ intel_batchbuffer_require_space(brw, (8 * 4) + (3 * 4) + dwords * 4, true);
opcode = XY_SETUP_BLT_CMD;
if (cpp == 4)
opcode |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
- if (dst_tiled)
+ if (dst_tiling != I915_TILING_NONE) {
opcode |= XY_DST_TILED;
+ dst_pitch /= 4;
+ }
br13 = dst_pitch | (translate_raster_op(logic_op) << 16) | (1 << 29);
- if (cpp == 2)
- br13 |= BR13_565;
- else
- br13 |= BR13_8888;
+ br13 |= br13_for_cpp(cpp);
+
+ blit_cmd = XY_TEXT_IMMEDIATE_BLIT_CMD | XY_TEXT_BYTE_PACKED; /* packing? */
+ if (dst_tiling != I915_TILING_NONE)
+ blit_cmd |= XY_DST_TILED;
- BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
- OUT_BATCH(opcode);
+ BEGIN_BATCH_BLT(8 + 3);
+ OUT_BATCH(opcode | (8 - 2));
OUT_BATCH(br13);
OUT_BATCH((0 << 16) | 0); /* clip x1, y1 */
OUT_BATCH((100 << 16) | 100); /* clip x2, y2 */
- OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, dst_offset);
+ OUT_RELOC_FENCED(dst_buffer,
+ I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+ dst_offset);
OUT_BATCH(0); /* bg */
OUT_BATCH(fg_color); /* fg */
OUT_BATCH(0); /* pattern base addr */
+
+ OUT_BATCH(blit_cmd | ((3 - 2) + dwords));
+ OUT_BATCH((y << 16) | x);
+ OUT_BATCH(((y + h) << 16) | (x + w));
ADVANCE_BATCH();
- intel_batchbuffer_data( intel->batch,
- &text,
- sizeof(text),
- INTEL_BATCH_NO_CLIPRECTS );
+ intel_batchbuffer_data(brw, src_bits, dwords * 4, true);
+
+ intel_batchbuffer_emit_mi_flush(brw);
+
+ return true;
+}
+
+/* We don't have a memmove-type blit like some other hardware, so we'll do a
+ * rectangular blit covering a large space, then emit 1-scanline blit at the
+ * end to cover the last if we need.
+ */
+void
+intel_emit_linear_blit(struct brw_context *brw,
+ drm_intel_bo *dst_bo,
+ unsigned int dst_offset,
+ drm_intel_bo *src_bo,
+ unsigned int src_offset,
+ unsigned int size)
+{
+ struct gl_context *ctx = &brw->ctx;
+ GLuint pitch, height;
+ bool ok;
- intel_batchbuffer_data( intel->batch,
- src_bits,
- dwords * 4,
- INTEL_BATCH_NO_CLIPRECTS );
+ /* The pitch given to the GPU must be DWORD aligned, and
+ * we want width to match pitch. Max width is (1 << 15 - 1),
+ * rounding that down to the nearest DWORD is 1 << 15 - 4
+ */
+ pitch = ROUND_DOWN_TO(MIN2(size, (1 << 15) - 1), 4);
+ height = (pitch == 0) ? 1 : size / pitch;
+ ok = intelEmitCopyBlit(brw, 1,
+ pitch, src_bo, src_offset, I915_TILING_NONE,
+ pitch, dst_bo, dst_offset, I915_TILING_NONE,
+ 0, 0, /* src x/y */
+ 0, 0, /* dst x/y */
+ pitch, height, /* w, h */
+ GL_COPY);
+ if (!ok)
+ _mesa_problem(ctx, "Failed to linear blit %dx%d\n", pitch, height);
+
+ src_offset += pitch * height;
+ dst_offset += pitch * height;
+ size -= pitch * height;
+ assert (size < (1 << 15));
+ pitch = ALIGN(size, 4);
+ if (size != 0) {
+ ok = intelEmitCopyBlit(brw, 1,
+ pitch, src_bo, src_offset, I915_TILING_NONE,
+ pitch, dst_bo, dst_offset, I915_TILING_NONE,
+ 0, 0, /* src x/y */
+ 0, 0, /* dst x/y */
+ size, 1, /* w, h */
+ GL_COPY);
+ if (!ok)
+ _mesa_problem(ctx, "Failed to linear blit %dx%d\n", size, 1);
+ }
}
+/**
+ * Used to initialize the alpha value of an ARGB8888 miptree after copying
+ * into it from an XRGB8888 source.
+ *
+ * This is very common with glCopyTexImage2D(). Note that the coordinates are
+ * relative to the start of the miptree, not relative to a slice within the
+ * miptree.
+ */
+static void
+intel_miptree_set_alpha_to_one(struct brw_context *brw,
+ struct intel_mipmap_tree *mt,
+ int x, int y, int width, int height)
+{
+ struct intel_region *region = mt->region;
+ uint32_t BR13, CMD;
+ int pitch, cpp;
+ drm_intel_bo *aper_array[2];
+
+ pitch = region->pitch;
+ cpp = region->cpp;
+
+ DBG("%s dst:buf(%p)/%d %d,%d sz:%dx%d\n",
+ __FUNCTION__, region->bo, pitch, x, y, width, height);
+
+ BR13 = br13_for_cpp(cpp) | 0xf0 << 16;
+ CMD = XY_COLOR_BLT_CMD;
+ CMD |= XY_BLT_WRITE_ALPHA;
+
+ if (region->tiling != I915_TILING_NONE) {
+ CMD |= XY_DST_TILED;
+ pitch /= 4;
+ }
+ BR13 |= pitch;
+
+ /* do space check before going any further */
+ aper_array[0] = brw->batch.bo;
+ aper_array[1] = region->bo;
+
+ if (drm_intel_bufmgr_check_aperture_space(aper_array,
+ ARRAY_SIZE(aper_array)) != 0) {
+ intel_batchbuffer_flush(brw);
+ }
+
+ bool dst_y_tiled = region->tiling == I915_TILING_Y;
+
+ BEGIN_BATCH_BLT_TILED(6, dst_y_tiled, false);
+ OUT_BATCH(CMD | (6 - 2));
+ OUT_BATCH(BR13);
+ OUT_BATCH((y << 16) | x);
+ OUT_BATCH(((y + height) << 16) | (x + width));
+ OUT_RELOC_FENCED(region->bo,
+ I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+ 0);
+ OUT_BATCH(0xffffffff); /* white, but only alpha gets written */
+ ADVANCE_BATCH_TILED(dst_y_tiled, false);
+
+ intel_batchbuffer_emit_mi_flush(brw);
+}