bool
intelEmitImmediateColorExpandBlit(struct intel_context *intel,
GLuint cpp,
GLubyte *src_bits, GLuint src_size,
GLuint fg_color,
GLshort dst_pitch,
drm_intel_bo *dst_buffer,
GLuint dst_offset,
uint32_t dst_tiling,
GLshort x, GLshort y,
GLshort w, GLshort h,
GLenum logic_op)
{
int dwords = ALIGN(src_size, 8) / 4;
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) && (logic_op <= (GL_CLEAR + 0x0f)));
assert(dst_pitch > 0);
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);
intel_batchbuffer_require_space(intel,
(8 * 4) +
(3 * 4) +
dwords * 4);
opcode = XY_SETUP_BLT_CMD;
if (cpp == 4)
opcode |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
br13 = dst_pitch | (translate_raster_op(logic_op) << 16) | (1 << 29);
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 + 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_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, src_bits, dwords * 4);
intel_batchbuffer_emit_mi_flush(intel);
return true;
}
/* Copy BitBlt
*/
void brw_copy_blit(struct brw_context *brw,
unsigned do_flip,
unsigned cpp,
short src_pitch,
struct pipe_buffer *src_buffer,
unsigned src_offset,
boolean src_tiled,
short dst_pitch,
struct pipe_buffer *dst_buffer,
unsigned dst_offset,
boolean dst_tiled,
short src_x, short src_y,
short dst_x, short dst_y,
short w, short h,
unsigned logic_op)
{
unsigned 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 - PIPE_LOGICOP_CLEAR >= 0 );
assert( logic_op - PIPE_LOGICOP_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;
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;
break;
default:
return;
}
if (src_tiled) {
CMD |= XY_SRC_TILED;
src_pitch /= 4;
}
if (dst_tiled) {
CMD |= XY_DST_TILED;
dst_pitch /= 4;
}
if (dst_y2 < dst_y ||
dst_x2 < dst_x) {
return;
}
dst_pitch &= 0xffff;
src_pitch &= 0xffff;
/* Initial y values don't seem to work with negative pitches. If
* we adjust the offsets manually (below), it seems to work fine.
*
* 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.
*/
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, BRW_BUFFER_ACCESS_WRITE,
dst_offset );
OUT_BATCH( (src_y << 16) | src_x );
OUT_BATCH( src_pitch );
OUT_RELOC( src_buffer, BRW_BUFFER_ACCESS_READ,
src_offset );
ADVANCE_BATCH();
}
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, BRW_BUFFER_ACCESS_WRITE,
dst_offset + dst_y * dst_pitch );
OUT_BATCH( (src_pitch & 0xffff) );
OUT_RELOC( src_buffer, BRW_BUFFER_ACCESS_READ,
src_offset + src_y * src_pitch );
ADVANCE_BATCH();
}
}
/* Copy BitBlt
*/
bool
intelEmitCopyBlit(struct intel_context *intel,
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)
{
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;
BATCH_LOCALS;
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)
return false;
/* do space check before going any further */
do {
aper_array[0] = intel->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(intel);
pass++;
} else
break;
} while (pass < 2);
if (pass >= 2)
return false;
intel_batchbuffer_require_space(intel, 8 * 4);
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);
/* 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;
}
}
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:
CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
break;
default:
return false;
}
if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
return true;
}
assert(dst_x < dst_x2);
assert(dst_y < dst_y2);
BEGIN_BATCH(8);
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);
ADVANCE_BATCH();
intel_batchbuffer_emit_mi_flush(intel);
return true;
}
/* Copy BitBlt
*/
void
intelEmitCopyBlit(struct intel_context *intel,
GLuint cpp,
GLshort src_pitch,
dri_bo *src_buffer,
GLuint src_offset,
GLboolean src_tiled,
GLshort dst_pitch,
dri_bo *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)
{
GLuint CMD, BR13;
int dst_y2 = dst_y + h;
int dst_x2 = dst_x + w;
BATCH_LOCALS;
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);
src_pitch *= cpp;
dst_pitch *= cpp;
BR13 = (translate_raster_op(logic_op) << 16);
switch (cpp) {
case 1:
case 2:
case 3:
BR13 |= (1 << 24);
CMD = XY_SRC_COPY_BLT_CMD;
break;
case 4:
BR13 |= (1 << 24) | (1 << 25);
CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
break;
default:
return;
}
#ifndef I915
if (dst_tiled) {
CMD |= XY_DST_TILED;
dst_pitch /= 4;
}
if (src_tiled) {
CMD |= XY_SRC_TILED;
src_pitch /= 4;
}
#endif
if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
return;
}
/* Initial y values don't seem to work with negative pitches. If
* we adjust the offsets manually (below), it seems to work fine.
*
* 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.
*/
if (dst_pitch > 0 && src_pitch > 0) {
assert(dst_x < dst_x2);
assert(dst_y < dst_y2);
BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
OUT_BATCH(CMD);
OUT_BATCH(BR13 | dst_pitch);
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();
}
else {
assert(dst_x < dst_x2);
assert(h > 0);
BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
OUT_BATCH(CMD);
OUT_BATCH(BR13 | dst_pitch);
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((0 << 16) | src_x);
OUT_BATCH(src_pitch);
OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ,
src_offset + src_y * src_pitch);
ADVANCE_BATCH();
}
}
drm_intel_gem_bo_unmap_gtt(src_buffer);
drm_intel_gem_bo_unmap_gtt(dst_buffer);
return GL_TRUE;
}
intel_batchbuffer_require_space(intel->batch, 8 * 4);
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);
src_pitch *= cpp;
dst_pitch *= cpp;
BR13 = translate_raster_op(logic_op) << 16;
switch (cpp) {
case 1:
CMD = XY_SRC_COPY_BLT_CMD;
break;
case 2:
BR13 |= BR13_565;
CMD = XY_SRC_COPY_BLT_CMD;
break;
case 4:
BR13 |= BR13_8888;
CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
break;
default:
return GL_FALSE;
void
intelEmitImmediateColorExpandBlit(struct intel_context *intel,
GLuint cpp,
GLubyte *src_bits, GLuint src_size,
GLuint fg_color,
GLshort dst_pitch,
dri_bo *dst_buffer,
GLuint dst_offset,
GLboolean dst_tiled,
GLshort x, GLshort y,
GLshort w, GLshort h,
GLenum logic_op)
{
int dwords = ALIGN(src_size, 8) / 4;
uint32_t opcode, br13, blit_cmd;
assert( logic_op - GL_CLEAR >= 0 );
assert( logic_op - GL_CLEAR < 0x10 );
if (w < 0 || h < 0)
return;
dst_pitch *= cpp;
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);
intel_batchbuffer_require_space( intel->batch,
(8 * 4) +
(3 * 4) +
dwords,
NO_LOOP_CLIPRECTS );
opcode = XY_SETUP_BLT_CMD;
if (cpp == 4)
opcode |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
#ifndef I915
if (dst_tiled) {
opcode |= XY_DST_TILED;
dst_pitch /= 4;
}
#endif
br13 = dst_pitch | (translate_raster_op(logic_op) << 16) | (1 << 29);
if (cpp == 2)
br13 |= BR13_565;
else
br13 |= BR13_8888;
blit_cmd = XY_TEXT_IMMEDIATE_BLIT_CMD | XY_TEXT_BYTE_PACKED; /* packing? */
if (dst_tiled)
blit_cmd |= XY_DST_TILED;
BEGIN_BATCH(8 + 3, NO_LOOP_CLIPRECTS);
OUT_BATCH(opcode);
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_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,
src_bits,
dwords * 4,
NO_LOOP_CLIPRECTS );
}
bool
intelEmitImmediateColorExpandBlit(struct brw_context *brw,
GLuint cpp,
GLubyte *src_bits, GLuint src_size,
GLuint fg_color,
GLshort dst_pitch,
struct brw_bo *dst_buffer,
GLuint dst_offset,
enum isl_tiling dst_tiling,
GLshort x, GLshort y,
GLshort w, GLshort h,
enum gl_logicop_mode logic_op)
{
const struct gen_device_info *devinfo = &brw->screen->devinfo;
int dwords = ALIGN(src_size, 8) / 4;
uint32_t opcode, br13, blit_cmd;
if (dst_tiling != ISL_TILING_LINEAR) {
if (dst_offset & 4095)
return false;
if (dst_tiling == ISL_TILING_Y0)
return false;
}
assert((unsigned) logic_op <= 0x0f);
assert(dst_pitch > 0);
if (w < 0 || h < 0)
return true;
DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d, %d bytes %d dwords\n",
__func__,
dst_buffer, dst_pitch, dst_offset, x, y, w, h, src_size, dwords);
unsigned xy_setup_blt_length = devinfo->gen >= 8 ? 10 : 8;
intel_batchbuffer_require_space(brw, (xy_setup_blt_length * 4) +
(3 * 4) + dwords * 4);
opcode = XY_SETUP_BLT_CMD;
if (cpp == 4)
opcode |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
if (dst_tiling != ISL_TILING_LINEAR) {
opcode |= XY_DST_TILED;
dst_pitch /= 4;
}
br13 = dst_pitch | (translate_raster_op(logic_op) << 16) | (1 << 29);
br13 |= br13_for_cpp(cpp);
blit_cmd = XY_TEXT_IMMEDIATE_BLIT_CMD | XY_TEXT_BYTE_PACKED; /* packing? */
if (dst_tiling != ISL_TILING_LINEAR)
blit_cmd |= XY_DST_TILED;
BEGIN_BATCH_BLT(xy_setup_blt_length + 3);
OUT_BATCH(opcode | (xy_setup_blt_length - 2));
OUT_BATCH(br13);
OUT_BATCH((0 << 16) | 0); /* clip x1, y1 */
OUT_BATCH((100 << 16) | 100); /* clip x2, y2 */
if (devinfo->gen >= 8) {
OUT_RELOC64(dst_buffer, RELOC_WRITE, dst_offset);
} else {
OUT_RELOC(dst_buffer, RELOC_WRITE, dst_offset);
}
OUT_BATCH(0); /* bg */
OUT_BATCH(fg_color); /* fg */
OUT_BATCH(0); /* pattern base addr */
if (devinfo->gen >= 8)
OUT_BATCH(0);
OUT_BATCH(blit_cmd | ((3 - 2) + dwords));
OUT_BATCH(SET_FIELD(y, BLT_Y) | SET_FIELD(x, BLT_X));
OUT_BATCH(SET_FIELD(y + h, BLT_Y) | SET_FIELD(x + w, BLT_X));
ADVANCE_BATCH();
intel_batchbuffer_data(brw, src_bits, dwords * 4);
brw_emit_mi_flush(brw);
return true;
}
/* Copy BitBlt
*/
static bool
emit_copy_blit(struct brw_context *brw,
GLuint cpp,
int32_t src_pitch,
struct brw_bo *src_buffer,
GLuint src_offset,
enum isl_tiling src_tiling,
int32_t dst_pitch,
struct brw_bo *dst_buffer,
GLuint dst_offset,
enum isl_tiling dst_tiling,
GLshort src_x, GLshort src_y,
GLshort dst_x, GLshort dst_y,
GLshort w, GLshort h,
enum gl_logicop_mode logic_op)
{
const struct gen_device_info *devinfo = &brw->screen->devinfo;
GLuint CMD, BR13;
int dst_y2 = dst_y + h;
int dst_x2 = dst_x + w;
bool dst_y_tiled = dst_tiling == ISL_TILING_Y0;
bool src_y_tiled = src_tiling == ISL_TILING_Y0;
uint32_t src_tile_w, src_tile_h;
uint32_t dst_tile_w, dst_tile_h;
if ((dst_y_tiled || src_y_tiled) && devinfo->gen < 6)
return false;
const unsigned bo_sizes = dst_buffer->size + src_buffer->size;
/* do space check before going any further */
if (!brw_batch_has_aperture_space(brw, bo_sizes))
intel_batchbuffer_flush(brw);
if (!brw_batch_has_aperture_space(brw, bo_sizes))
return false;
unsigned length = devinfo->gen >= 8 ? 10 : 8;
intel_batchbuffer_require_space(brw, length * 4);
DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
__func__,
src_buffer, src_pitch, src_offset, src_x, src_y,
dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h);
intel_get_tile_dims(src_tiling, cpp, &src_tile_w, &src_tile_h);
intel_get_tile_dims(dst_tiling, cpp, &dst_tile_w, &dst_tile_h);
/* For Tiled surfaces, the pitch has to be a multiple of the Tile width
* (X direction width of the Tile). This is ensured while allocating the
* buffer object.
*/
assert(src_tiling == ISL_TILING_LINEAR || (src_pitch % src_tile_w) == 0);
assert(dst_tiling == ISL_TILING_LINEAR || (dst_pitch % dst_tile_w) == 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;
}
}
if (!alignment_valid(brw, dst_offset, dst_tiling))
return false;
if (!alignment_valid(brw, src_offset, src_tiling))
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;
assert(cpp <= 4);
BR13 = br13_for_cpp(cpp) | translate_raster_op(logic_op) << 16;
CMD = xy_blit_cmd(src_tiling, dst_tiling, cpp);
/* For tiled source and destination, pitch value should be specified
* as a number of Dwords.
*/
if (dst_tiling != ISL_TILING_LINEAR)
dst_pitch /= 4;
if (src_tiling != ISL_TILING_LINEAR)
src_pitch /= 4;
if (dst_y2 <= dst_y || dst_x2 <= dst_x)
return true;
assert(dst_x < dst_x2);
assert(dst_y < dst_y2);
BEGIN_BATCH_BLT_TILED(length, dst_y_tiled, src_y_tiled);
OUT_BATCH(CMD | (length - 2));
OUT_BATCH(BR13 | (uint16_t)dst_pitch);
OUT_BATCH(SET_FIELD(dst_y, BLT_Y) | SET_FIELD(dst_x, BLT_X));
OUT_BATCH(SET_FIELD(dst_y2, BLT_Y) | SET_FIELD(dst_x2, BLT_X));
if (devinfo->gen >= 8) {
OUT_RELOC64(dst_buffer, RELOC_WRITE, dst_offset);
} else {
OUT_RELOC(dst_buffer, RELOC_WRITE, dst_offset);
}
OUT_BATCH(SET_FIELD(src_y, BLT_Y) | SET_FIELD(src_x, BLT_X));
OUT_BATCH((uint16_t)src_pitch);
if (devinfo->gen >= 8) {
OUT_RELOC64(src_buffer, 0, src_offset);
} else {
OUT_RELOC(src_buffer, 0, src_offset);
}
ADVANCE_BATCH_TILED(dst_y_tiled, src_y_tiled);
brw_emit_mi_flush(brw);
return true;
}
void
intelEmitImmediateColorExpandBlit(struct intel_context *intel,
GLuint cpp,
GLubyte *src_bits, GLuint src_size,
GLuint fg_color,
GLshort dst_pitch,
struct buffer *dst_buffer,
GLuint dst_offset,
GLboolean dst_tiled,
GLshort x, GLshort y,
GLshort w, GLshort h,
GLenum logic_op)
{
struct xy_setup_blit setup;
struct xy_text_immediate_blit text;
int dwords = ((src_size + 7) & ~7) / 4;
assert( logic_op - GL_CLEAR >= 0 );
assert( logic_op - GL_CLEAR < 0x10 );
if (w < 0 || h < 0)
return;
dst_pitch *= cpp;
if (dst_tiled)
dst_pitch /= 4;
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(&setup, 0, sizeof(setup));
setup.br0.client = CLIENT_2D;
setup.br0.opcode = OPCODE_XY_SETUP_BLT;
setup.br0.write_alpha = (cpp == 4);
setup.br0.write_rgb = (cpp == 4);
setup.br0.dst_tiled = dst_tiled;
setup.br0.length = (sizeof(setup) / sizeof(int)) - 2;
setup.br13.dest_pitch = dst_pitch;
setup.br13.rop = translate_raster_op(logic_op);
setup.br13.color_depth = (cpp == 4) ? BR13_8888 : BR13_565;
setup.br13.clipping_enable = 0;
setup.br13.mono_source_transparency = 1;
setup.dw2.clip_y1 = 0;
setup.dw2.clip_x1 = 0;
setup.dw3.clip_y2 = 100;
setup.dw3.clip_x2 = 100;
setup.dest_base_addr = bmBufferOffset(intel, dst_buffer) + dst_offset;
setup.background_color = 0;
setup.foreground_color = fg_color;
setup.pattern_base_addr = 0;
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,
sizeof(setup) +
sizeof(text) +
dwords,
INTEL_BATCH_NO_CLIPRECTS );
intel_batchbuffer_data( intel->batch,
&setup,
sizeof(setup),
INTEL_BATCH_NO_CLIPRECTS );
intel_batchbuffer_data( intel->batch,
&text,
sizeof(text),
INTEL_BATCH_NO_CLIPRECTS );
intel_batchbuffer_data( intel->batch,
src_bits,
dwords * 4,
INTEL_BATCH_NO_CLIPRECTS );
}
/* Copy BitBlt
*/
GLboolean
intelEmitCopyBlit(struct intel_context *intel,
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)
{
GLuint CMD, BR13, pass = 0;
int dst_y2 = dst_y + h;
int dst_x2 = dst_x + w;
drm_intel_bo *aper_array[3];
BATCH_LOCALS;
if (dst_tiling != I915_TILING_NONE) {
if (dst_offset & 4095)
return GL_FALSE;
if (dst_tiling == I915_TILING_Y)
return GL_FALSE;
}
if (src_tiling != I915_TILING_NONE) {
if (src_offset & 4095)
return GL_FALSE;
if (src_tiling == I915_TILING_Y)
return GL_FALSE;
}
/* do space check before going any further */
do {
aper_array[0] = intel->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(intel);
pass++;
} else
break;
} while (pass < 2);
if (pass >= 2)
return GL_FALSE;
intel_batchbuffer_require_space(intel, 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);
src_pitch *= cpp;
dst_pitch *= cpp;
/* For big formats (such as floating point), do the copy using 32bpp and
* multiply the coordinates.
*/
if (cpp > 4) {
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;
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 GL_FALSE;
}
#ifndef I915
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;
}
#endif
if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
return GL_TRUE;
}
assert(dst_x < dst_x2);
assert(dst_y < dst_y2);
BEGIN_BATCH_BLT(8);
OUT_BATCH(CMD);
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);
ADVANCE_BATCH();
intel_batchbuffer_emit_mi_flush(intel);
return GL_TRUE;
}
