-/**************************************************************************
- *
+/*
* Copyright 2003 VMware, Inc.
* All Rights Reserved.
*
* copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
+ * distribute, sublicense, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
* IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
* ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
-
+ */
#include "main/mtypes.h"
+#include "main/blit.h"
#include "main/context.h"
#include "main/enums.h"
#include "main/colormac.h"
#define FILE_DEBUG_FLAG DEBUG_BLIT
+#define SET_TILING_XY_FAST_COPY_BLT(tiling, tr_mode, type) \
+({ \
+ switch (tiling) { \
+ case I915_TILING_X: \
+ CMD |= type ## _TILED_X; \
+ break; \
+ case I915_TILING_Y: \
+ if (tr_mode == INTEL_MIPTREE_TRMODE_YS) \
+ CMD |= type ## _TILED_64K; \
+ else \
+ CMD |= type ## _TILED_Y; \
+ break; \
+ default: \
+ unreachable("not reached"); \
+ } \
+})
+
static void
intel_miptree_set_alpha_to_one(struct brw_context *brw,
struct intel_mipmap_tree *mt,
br13_for_cpp(int cpp)
{
switch (cpp) {
+ case 16:
+ return BR13_32323232;
+ case 8:
+ return BR13_16161616;
case 4:
return BR13_8888;
- break;
case 2:
return BR13_565;
- break;
case 1:
return BR13_8;
- break;
default:
unreachable("not reached");
}
}
+static uint32_t
+get_tr_horizontal_align(uint32_t tr_mode, uint32_t cpp, bool is_src) {
+ /* Alignment tables for YF/YS tiled surfaces. */
+ const uint32_t align_2d_yf[] = {64, 64, 32, 32, 16};
+ const uint32_t bpp = cpp * 8;
+ const uint32_t shift = is_src ? 17 : 10;
+ uint32_t align;
+ int i = 0;
+
+ if (tr_mode == INTEL_MIPTREE_TRMODE_NONE)
+ return 0;
+
+ /* Compute array index. */
+ assert (bpp >= 8 && bpp <= 128 && _mesa_is_pow_two(bpp));
+ i = ffs(bpp / 8) - 1;
+
+ align = tr_mode == INTEL_MIPTREE_TRMODE_YF ?
+ align_2d_yf[i] :
+ 4 * align_2d_yf[i];
+
+ assert(_mesa_is_pow_two(align));
+
+ /* XY_FAST_COPY_BLT doesn't support horizontal alignment of 16. */
+ if (align == 16)
+ align = 32;
+
+ return (ffs(align) - 6) << shift;
+}
+
+static uint32_t
+get_tr_vertical_align(uint32_t tr_mode, uint32_t cpp, bool is_src) {
+ /* Vertical alignment tables for YF/YS tiled surfaces. */
+ const unsigned align_2d_yf[] = {64, 32, 32, 16, 16};
+ const uint32_t bpp = cpp * 8;
+ const uint32_t shift = is_src ? 15 : 8;
+ uint32_t align;
+ int i = 0;
+
+ if (tr_mode == INTEL_MIPTREE_TRMODE_NONE)
+ return 0;
+
+ /* Compute array index. */
+ assert (bpp >= 8 && bpp <= 128 && _mesa_is_pow_two(bpp));
+ i = ffs(bpp / 8) - 1;
+
+ align = tr_mode == INTEL_MIPTREE_TRMODE_YF ?
+ align_2d_yf[i] :
+ 4 * align_2d_yf[i];
+
+ assert(_mesa_is_pow_two(align));
+
+ /* XY_FAST_COPY_BLT doesn't support vertical alignments of 16 and 32. */
+ if (align == 16 || align == 32)
+ align = 64;
+
+ return (ffs(align) - 7) << shift;
+}
+
/**
* Emits the packet for switching the blitter from X to Y tiled or back.
*
* tiling state would leak into other unsuspecting applications (like the X
* server).
*/
-static void
+static uint32_t *
set_blitter_tiling(struct brw_context *brw,
- bool dst_y_tiled, bool src_y_tiled)
+ bool dst_y_tiled, bool src_y_tiled,
+ uint32_t *__map)
{
assert(brw->gen >= 6);
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));
+ return __map;
}
+#define SET_BLITTER_TILING(...) __map = set_blitter_tiling(__VA_ARGS__, __map)
-#define BEGIN_BATCH_BLT_TILED(n, dst_y_tiled, src_y_tiled) do { \
+#define BEGIN_BATCH_BLT_TILED(n, dst_y_tiled, src_y_tiled) \
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)
+ SET_BLITTER_TILING(brw, dst_y_tiled, src_y_tiled)
-#define ADVANCE_BATCH_TILED(dst_y_tiled, src_y_tiled) do { \
+#define ADVANCE_BATCH_TILED(dst_y_tiled, src_y_tiled) \
if (dst_y_tiled || src_y_tiled) \
- set_blitter_tiling(brw, false, false); \
- ADVANCE_BATCH(); \
- } while (0)
+ SET_BLITTER_TILING(brw, false, false); \
+ ADVANCE_BATCH()
static int
blt_pitch(struct intel_mipmap_tree *mt)
return pitch;
}
+bool
+intel_miptree_blit_compatible_formats(mesa_format src, mesa_format dst)
+{
+ /* The BLT doesn't handle sRGB conversion */
+ assert(src == _mesa_get_srgb_format_linear(src));
+ assert(dst == _mesa_get_srgb_format_linear(dst));
+
+ /* No swizzle or format conversions possible, except... */
+ if (src == dst)
+ return true;
+
+ /* ...we can either discard the alpha channel when going from A->X,
+ * or we can fill the alpha channel with 0xff when going from X->A
+ */
+ if (src == MESA_FORMAT_B8G8R8A8_UNORM || src == MESA_FORMAT_B8G8R8X8_UNORM)
+ return (dst == MESA_FORMAT_B8G8R8A8_UNORM ||
+ dst == MESA_FORMAT_B8G8R8X8_UNORM);
+
+ if (src == MESA_FORMAT_R8G8B8A8_UNORM || src == MESA_FORMAT_R8G8B8X8_UNORM)
+ return (dst == MESA_FORMAT_R8G8B8A8_UNORM ||
+ dst == MESA_FORMAT_R8G8B8X8_UNORM);
+
+ return false;
+}
+
/**
* Implements a rectangular block transfer (blit) of pixels between two
* miptrees.
* 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_B8G8R8A8_UNORM &&
- src_format != MESA_FORMAT_B8G8R8X8_UNORM) ||
- (dst_format != MESA_FORMAT_B8G8R8A8_UNORM &&
- dst_format != MESA_FORMAT_B8G8R8X8_UNORM))) {
+ if (!intel_miptree_blit_compatible_formats(src_format, dst_format)) {
perf_debug("%s: Can't use hardware blitter from %s to %s, "
"falling back.\n", __func__,
_mesa_get_format_name(src_format),
src_pitch,
src_mt->bo, src_mt->offset,
src_mt->tiling,
+ src_mt->tr_mode,
dst_mt->pitch,
dst_mt->bo, dst_mt->offset,
dst_mt->tiling,
+ dst_mt->tr_mode,
src_x, src_y,
dst_x, dst_y,
width, height,
return false;
}
- if (src_mt->format == MESA_FORMAT_B8G8R8X8_UNORM &&
- dst_mt->format == MESA_FORMAT_B8G8R8A8_UNORM) {
+ /* XXX This could be done in a single pass using XY_FULL_MONO_PATTERN_BLT */
+ if (_mesa_get_format_bits(src_format, GL_ALPHA_BITS) == 0 &&
+ _mesa_get_format_bits(dst_format, GL_ALPHA_BITS) > 0) {
intel_miptree_set_alpha_to_one(brw, dst_mt,
dst_x, dst_y,
width, height);
return true;
}
+static bool
+can_fast_copy_blit(struct brw_context *brw,
+ drm_intel_bo *src_buffer,
+ int16_t src_x, int16_t src_y,
+ uintptr_t src_offset, uint32_t src_pitch,
+ uint32_t src_tiling, uint32_t src_tr_mode,
+ drm_intel_bo *dst_buffer,
+ int16_t dst_x, int16_t dst_y,
+ uintptr_t dst_offset, uint32_t dst_pitch,
+ uint32_t dst_tiling, uint32_t dst_tr_mode,
+ int16_t w, int16_t h, uint32_t cpp)
+{
+ const bool dst_tiling_none = dst_tiling == I915_TILING_NONE;
+ const bool src_tiling_none = src_tiling == I915_TILING_NONE;
+
+ if (brw->gen < 9)
+ return false;
+
+ if (src_buffer->handle == dst_buffer->handle &&
+ _mesa_regions_overlap(src_x, src_y, src_x + w, src_y + h,
+ dst_x, dst_y, dst_x + w, dst_y + h))
+ return false;
+
+ /* Enable fast copy blit only if the surfaces are Yf/Ys tiled.
+ * FIXME: Based on performance data, remove this condition later to
+ * enable for all types of surfaces.
+ */
+ if (src_tr_mode == INTEL_MIPTREE_TRMODE_NONE &&
+ dst_tr_mode == INTEL_MIPTREE_TRMODE_NONE)
+ return false;
+
+ /* For all surface types buffers must be cacheline-aligned. */
+ if ((dst_offset | src_offset) & 63)
+ return false;
+
+ /* Color depth greater than 128 bits not supported. */
+ if (cpp > 16)
+ return false;
+
+ /* For Fast Copy Blits the pitch cannot be a negative number. So, bit 15
+ * of the destination pitch must be zero.
+ */
+ if ((src_pitch >> 15 & 1) != 0 || (dst_pitch >> 15 & 1) != 0)
+ return false;
+
+ /* For Linear surfaces, the pitch has to be an OWord (16byte) multiple. */
+ if ((src_tiling_none && src_pitch % 16 != 0) ||
+ (dst_tiling_none && dst_pitch % 16 != 0))
+ return false;
+
+ /* For Tiled surfaces, the pitch has to be a multiple of the Tile width
+ * (X direction width of the Tile). This means the pitch value will
+ * always be Cache Line aligned (64byte multiple).
+ */
+ if ((!dst_tiling_none && dst_pitch % 64 != 0) ||
+ (!src_tiling_none && src_pitch % 64 != 0))
+ return false;
+
+ return true;
+}
+
+static uint32_t
+xy_blit_cmd(uint32_t src_tiling, uint32_t src_tr_mode,
+ uint32_t dst_tiling, uint32_t dst_tr_mode,
+ uint32_t cpp, bool use_fast_copy_blit)
+{
+ uint32_t CMD = 0;
+
+ if (use_fast_copy_blit) {
+ CMD = XY_FAST_COPY_BLT_CMD;
+
+ if (dst_tiling != I915_TILING_NONE)
+ SET_TILING_XY_FAST_COPY_BLT(dst_tiling, dst_tr_mode, XY_FAST_DST);
+
+ if (src_tiling != I915_TILING_NONE)
+ SET_TILING_XY_FAST_COPY_BLT(src_tiling, src_tr_mode, XY_FAST_SRC);
+
+ CMD |= get_tr_horizontal_align(src_tr_mode, cpp, true /* is_src */);
+ CMD |= get_tr_vertical_align(src_tr_mode, cpp, true /* is_src */);
+
+ CMD |= get_tr_horizontal_align(dst_tr_mode, cpp, false /* is_src */);
+ CMD |= get_tr_vertical_align(dst_tr_mode, cpp, false /* is_src */);
+
+ } else {
+ assert(cpp <= 4);
+ switch (cpp) {
+ case 1:
+ case 2:
+ CMD = XY_SRC_COPY_BLT_CMD;
+ break;
+ case 4:
+ CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
+ break;
+ default:
+ unreachable("not reached");
+ }
+
+ if (dst_tiling != I915_TILING_NONE)
+ CMD |= XY_DST_TILED;
+
+ if (src_tiling != I915_TILING_NONE)
+ CMD |= XY_SRC_TILED;
+ }
+ return CMD;
+}
+
/* Copy BitBlt
*/
bool
drm_intel_bo *src_buffer,
GLuint src_offset,
uint32_t src_tiling,
+ uint32_t src_tr_mode,
GLshort dst_pitch,
drm_intel_bo *dst_buffer,
GLuint dst_offset,
uint32_t dst_tiling,
+ uint32_t dst_tr_mode,
GLshort src_x, GLshort src_y,
GLshort dst_x, GLshort dst_y,
GLshort w, GLshort h,
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 (!alignment_valid(brw, dst_offset, dst_tiling))
- return false;
- if (!alignment_valid(brw, src_offset, src_tiling))
- return false;
+ bool use_fast_copy_blit = false;
if ((dst_y_tiled || src_y_tiled) && brw->gen < 6)
return false;
- assert(!dst_y_tiled || (dst_pitch % 128) == 0);
- assert(!src_y_tiled || (src_pitch % 128) == 0);
-
/* do space check before going any further */
do {
aper_array[0] = brw->batch.bo;
src_buffer, src_pitch, src_offset, src_x, src_y,
dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h);
- /* Blit pitch must be dword-aligned. Otherwise, the hardware appears to drop
- * the low bits. Offsets must be naturally aligned.
- */
- if (src_pitch % 4 != 0 || src_offset % cpp != 0 ||
- dst_pitch % 4 != 0 || dst_offset % cpp != 0)
- return false;
+ use_fast_copy_blit = can_fast_copy_blit(brw,
+ src_buffer,
+ src_x, src_y,
+ src_offset, src_pitch,
+ src_tiling, src_tr_mode,
+ dst_buffer,
+ dst_x, dst_y,
+ dst_offset, dst_pitch,
+ dst_tiling, dst_tr_mode,
+ w, h, cpp);
+ assert(use_fast_copy_blit ||
+ (src_tr_mode == INTEL_MIPTREE_TRMODE_NONE &&
+ dst_tr_mode == INTEL_MIPTREE_TRMODE_NONE));
+
+ if (use_fast_copy_blit) {
+ /* When two sequential fast copy blits have different source surfaces,
+ * but their destinations refer to the same destination surfaces and
+ * therefore destinations overlap it is imperative that a flush be
+ * inserted between the two blits.
+ *
+ * FIXME: Figure out a way to avoid flushing when not required.
+ */
+ brw_emit_mi_flush(brw);
+
+ assert(cpp <= 16);
+ BR13 = br13_for_cpp(cpp);
+
+ if (src_tr_mode == INTEL_MIPTREE_TRMODE_YF)
+ BR13 |= XY_FAST_SRC_TRMODE_YF;
+
+ if (dst_tr_mode == INTEL_MIPTREE_TRMODE_YF)
+ BR13 |= XY_FAST_DST_TRMODE_YF;
+
+ CMD = xy_blit_cmd(src_tiling, src_tr_mode,
+ dst_tiling, dst_tr_mode,
+ cpp, use_fast_copy_blit);
+
+ /* For tiled source and destination, pitch value should be specified
+ * as a number of Dwords.
+ */
+ if (dst_tiling != I915_TILING_NONE)
+ dst_pitch /= 4;
+
+ if (src_tiling != I915_TILING_NONE)
+ src_pitch /= 4;
- /* 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;
+ } else {
+ assert(!dst_y_tiled || (dst_pitch % 128) == 0);
+ assert(!src_y_tiled || (src_pitch % 128) == 0);
+
+ /* 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;
+ }
}
- }
- BR13 = br13_for_cpp(cpp) | translate_raster_op(logic_op) << 16;
+ if (!alignment_valid(brw, dst_offset, dst_tiling))
+ return false;
+ if (!alignment_valid(brw, src_offset, src_tiling))
+ return false;
- switch (cpp) {
- case 1:
- case 2:
- CMD = XY_SRC_COPY_BLT_CMD;
- break;
- case 4:
- CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
- break;
- default:
- return false;
- }
+ /* Blit pitch must be dword-aligned. Otherwise, the hardware appears to drop
+ * the low bits. Offsets must be naturally aligned.
+ */
+ if (src_pitch % 4 != 0 || src_offset % cpp != 0 ||
+ dst_pitch % 4 != 0 || dst_offset % cpp != 0)
+ 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;
+ assert(cpp <= 4);
+ BR13 = br13_for_cpp(cpp) | translate_raster_op(logic_op) << 16;
+
+ CMD = xy_blit_cmd(src_tiling, src_tr_mode,
+ dst_tiling, dst_tr_mode,
+ cpp, use_fast_copy_blit);
+
+ if (dst_tiling != I915_TILING_NONE)
+ dst_pitch /= 4;
+
+ if (src_tiling != I915_TILING_NONE)
+ src_pitch /= 4;
}
if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
ADVANCE_BATCH_TILED(dst_y_tiled, src_y_tiled);
- intel_batchbuffer_emit_mi_flush(brw);
+ brw_emit_mi_flush(brw);
return true;
}
intel_batchbuffer_data(brw, src_bits, dwords * 4, BLT_RING);
- intel_batchbuffer_emit_mi_flush(brw);
+ brw_emit_mi_flush(brw);
return true;
}
int16_t src_x, dst_x;
bool ok;
- /* 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;
- src_x = src_offset % 64;
- dst_x = dst_offset % 64;
- ok = intelEmitCopyBlit(brw, 1,
- pitch, src_bo, src_offset - src_x, I915_TILING_NONE,
- pitch, dst_bo, dst_offset - dst_x, I915_TILING_NONE,
- src_x, 0, /* src x/y */
- dst_x, 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;
- src_x = src_offset % 64;
- dst_x = dst_offset % 64;
- size -= pitch * height;
- assert (size < (1 << 15));
- pitch = ALIGN(size, 4);
-
- if (size != 0) {
+ do {
+ /* 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) - 64), 4);
+ height = (size < pitch || pitch == 0) ? 1 : size / pitch;
+
+ src_x = src_offset % 64;
+ dst_x = dst_offset % 64;
+ pitch = ALIGN(MIN2(size, (1 << 15) - 64), 4);
+ assert(src_x + pitch < 1 << 15);
+ assert(dst_x + pitch < 1 << 15);
+
ok = intelEmitCopyBlit(brw, 1,
- pitch, src_bo, src_offset - src_x, I915_TILING_NONE,
- pitch, dst_bo, dst_offset - dst_x, I915_TILING_NONE,
- src_x, 0, /* src x/y */
- dst_x, 0, /* dst x/y */
- size, 1, /* w, h */
- GL_COPY);
- if (!ok)
- _mesa_problem(ctx, "Failed to linear blit %dx%d\n", size, 1);
- }
+ pitch, src_bo, src_offset - src_x, I915_TILING_NONE,
+ INTEL_MIPTREE_TRMODE_NONE,
+ pitch, dst_bo, dst_offset - dst_x, I915_TILING_NONE,
+ INTEL_MIPTREE_TRMODE_NONE,
+ src_x, 0, /* src x/y */
+ dst_x, 0, /* dst x/y */
+ MIN2(size, pitch), height, /* w, h */
+ GL_COPY);
+ if (!ok) {
+ _mesa_problem(ctx, "Failed to linear blit %dx%d\n",
+ MIN2(size, pitch), height);
+ return;
+ }
+
+ pitch *= height;
+ if (size <= pitch)
+ return;
+
+ src_offset += pitch;
+ dst_offset += pitch;
+ size -= pitch;
+ } while (1);
}
/**
OUT_BATCH(0xffffffff); /* white, but only alpha gets written */
ADVANCE_BATCH_TILED(dst_y_tiled, false);
- intel_batchbuffer_emit_mi_flush(brw);
+ brw_emit_mi_flush(brw);
}