static bool compare_textures(struct pipe_context *ctx,
struct pipe_resource *tex,
struct cpu_texture *cpu, int bpp)
{
struct pipe_transfer *t;
uint8_t *map;
int y,z;
bool pass = true;
map = pipe_transfer_map_3d(ctx, tex, 0, PIPE_TRANSFER_READ,
0, 0, 0, tex->width0, tex->height0,
tex->array_size, &t);
assert(map);
for (z = 0; z < tex->array_size; z++) {
for (y = 0; y < tex->height0; y++) {
uint8_t *ptr = map + t->layer_stride*z + t->stride*y;
uint8_t *cpu_ptr = cpu->ptr +
cpu->layer_stride*z + cpu->stride*y;
if (memcmp(ptr, cpu_ptr, tex->width0 * bpp)) {
pass = false;
goto done;
}
}
}
done:
pipe_transfer_unmap(ctx, t);
return pass;
}
static void set_random_pixels(struct pipe_context *ctx,
struct pipe_resource *tex,
struct cpu_texture *cpu)
{
struct pipe_transfer *t;
uint8_t *map;
int x,y,z;
map = pipe_transfer_map_3d(ctx, tex, 0, PIPE_TRANSFER_WRITE,
0, 0, 0, tex->width0, tex->height0,
tex->array_size, &t);
assert(map);
for (z = 0; z < tex->array_size; z++) {
for (y = 0; y < tex->height0; y++) {
uint64_t *ptr = (uint64_t*)
(map + t->layer_stride*z + t->stride*y);
uint64_t *ptr_cpu = (uint64_t*)
(cpu->ptr + cpu->layer_stride*z + cpu->stride*y);
unsigned size = cpu->stride / RAND_NUM_SIZE;
assert(t->stride % RAND_NUM_SIZE == 0);
assert(cpu->stride % RAND_NUM_SIZE == 0);
for (x = 0; x < size; x++) {
*ptr++ = *ptr_cpu++ =
rand_xorshift128plus(seed_xorshift128plus);
}
}
}
pipe_transfer_unmap(ctx, t);
}
/**
* Map a texture image and return the address for a particular 2D face/slice/
* layer. The stImage indicates the cube face and mipmap level. The slice
* of the 3D texture is passed in 'zoffset'.
* \param usage one of the PIPE_TRANSFER_x values
* \param x, y, w, h the region of interest of the 2D image.
* \return address of mapping or NULL if any error
*/
GLubyte *
st_texture_image_map(struct st_context *st, struct st_texture_image *stImage,
enum pipe_transfer_usage usage,
GLuint x, GLuint y, GLuint z,
GLuint w, GLuint h, GLuint d,
struct pipe_transfer **transfer)
{
struct st_texture_object *stObj =
st_texture_object(stImage->base.TexObject);
GLuint level;
void *map;
DBG("%s \n", __func__);
if (!stImage->pt)
return NULL;
if (stObj->pt != stImage->pt)
level = 0;
else
level = stImage->base.Level;
if (stObj->base.Immutable) {
level += stObj->base.MinLevel;
z += stObj->base.MinLayer;
if (stObj->pt->array_size > 1)
d = MIN2(d, stObj->base.NumLayers);
}
z += stImage->base.Face;
map = pipe_transfer_map_3d(st->pipe, stImage->pt, level, usage,
x, y, z, w, h, d, transfer);
if (map) {
/* Enlarge the transfer array if it's not large enough. */
if (z >= stImage->num_transfers) {
unsigned new_size = z + 1;
stImage->transfer = realloc(stImage->transfer,
new_size * sizeof(struct st_texture_image_transfer));
memset(&stImage->transfer[stImage->num_transfers], 0,
(new_size - stImage->num_transfers) *
sizeof(struct st_texture_image_transfer));
stImage->num_transfers = new_size;
}
assert(!stImage->transfer[z].transfer);
stImage->transfer[z].transfer = *transfer;
}
return map;
}
/**
* Map a texture image and return the address for a particular 2D face/slice/
* layer. The stImage indicates the cube face and mipmap level. The slice
* of the 3D texture is passed in 'zoffset'.
* \param usage one of the PIPE_TRANSFER_x values
* \param x, y, w, h the region of interest of the 2D image.
* \return address of mapping or NULL if any error
*/
GLubyte *
st_texture_image_map(struct st_context *st, struct st_texture_image *stImage,
enum pipe_transfer_usage usage,
GLuint x, GLuint y, GLuint z,
GLuint w, GLuint h, GLuint d)
{
struct st_texture_object *stObj =
st_texture_object(stImage->base.TexObject);
GLuint level;
DBG("%s \n", __FUNCTION__);
if (!stImage->pt)
return NULL;
if (stObj->pt != stImage->pt)
level = 0;
else
level = stImage->base.Level;
return pipe_transfer_map_3d(st->pipe, stImage->pt, level, usage,
x, y, z + stImage->base.Face,
w, h, d, &stImage->transfer);
}
/**
* This uses a blit to copy the read buffer to a texture format which matches
* the format and type combo and then a fast read-back is done using memcpy.
* We can do arbitrary X/Y/Z/W/0/1 swizzling here as long as there is
* a format which matches the swizzling.
*
* If such a format isn't available, we fall back to _mesa_readpixels.
*
* NOTE: Some drivers use a blit to convert between tiled and linear
* texture layouts during texture uploads/downloads, so the blit
* we do here should be free in such cases.
*/
static void
st_readpixels(struct gl_context *ctx, GLint x, GLint y,
GLsizei width, GLsizei height,
GLenum format, GLenum type,
const struct gl_pixelstore_attrib *pack,
GLvoid *pixels)
{
struct st_context *st = st_context(ctx);
struct gl_renderbuffer *rb =
_mesa_get_read_renderbuffer_for_format(ctx, format);
struct st_renderbuffer *strb = st_renderbuffer(rb);
struct pipe_context *pipe = st->pipe;
struct pipe_screen *screen = pipe->screen;
struct pipe_resource *src;
struct pipe_resource *dst = NULL;
struct pipe_resource dst_templ;
enum pipe_format dst_format, src_format;
struct pipe_blit_info blit;
unsigned bind = PIPE_BIND_TRANSFER_READ;
struct pipe_transfer *tex_xfer;
ubyte *map = NULL;
/* Validate state (to be sure we have up-to-date framebuffer surfaces)
* and flush the bitmap cache prior to reading. */
st_validate_state(st);
st_flush_bitmap_cache(st);
if (!st->prefer_blit_based_texture_transfer) {
goto fallback;
}
/* This must be done after state validation. */
src = strb->texture;
/* XXX Fallback for depth-stencil formats due to an incomplete
* stencil blit implementation in some drivers. */
if (format == GL_DEPTH_STENCIL) {
goto fallback;
}
/* We are creating a texture of the size of the region being read back.
* Need to check for NPOT texture support. */
if (!screen->get_param(screen, PIPE_CAP_NPOT_TEXTURES) &&
(!util_is_power_of_two(width) ||
!util_is_power_of_two(height))) {
goto fallback;
}
/* If the base internal format and the texture format don't match, we have
* to use the slow path. */
if (rb->_BaseFormat !=
_mesa_get_format_base_format(rb->Format)) {
goto fallback;
}
/* See if the texture format already matches the format and type,
* in which case the memcpy-based fast path will likely be used and
* we don't have to blit. */
if (_mesa_format_matches_format_and_type(rb->Format, format,
type, pack->SwapBytes)) {
goto fallback;
}
if (_mesa_readpixels_needs_slow_path(ctx, format, type, GL_TRUE)) {
goto fallback;
}
/* Convert the source format to what is expected by ReadPixels
* and see if it's supported. */
src_format = util_format_linear(src->format);
src_format = util_format_luminance_to_red(src_format);
src_format = util_format_intensity_to_red(src_format);
if (!src_format ||
!screen->is_format_supported(screen, src_format, src->target,
src->nr_samples,
PIPE_BIND_SAMPLER_VIEW)) {
goto fallback;
}
if (format == GL_DEPTH_COMPONENT || format == GL_DEPTH_STENCIL)
bind |= PIPE_BIND_DEPTH_STENCIL;
else
bind |= PIPE_BIND_RENDER_TARGET;
/* Choose the destination format by finding the best match
* for the format+type combo. */
dst_format = st_choose_matching_format(screen, bind, format, type,
pack->SwapBytes);
if (dst_format == PIPE_FORMAT_NONE) {
goto fallback;
}
/* create the destination texture */
memset(&dst_templ, 0, sizeof(dst_templ));
dst_templ.target = PIPE_TEXTURE_2D;
dst_templ.format = dst_format;
dst_templ.bind = bind;
dst_templ.usage = PIPE_USAGE_STAGING;
st_gl_texture_dims_to_pipe_dims(GL_TEXTURE_2D, width, height, 1,
&dst_templ.width0, &dst_templ.height0,
&dst_templ.depth0, &dst_templ.array_size);
dst = screen->resource_create(screen, &dst_templ);
if (!dst) {
goto fallback;
}
memset(&blit, 0, sizeof(blit));
blit.src.resource = src;
blit.src.level = strb->surface->u.tex.level;
blit.src.format = src_format;
blit.dst.resource = dst;
blit.dst.level = 0;
blit.dst.format = dst->format;
blit.src.box.x = x;
blit.dst.box.x = 0;
blit.src.box.y = y;
blit.dst.box.y = 0;
blit.src.box.z = strb->surface->u.tex.first_layer;
blit.dst.box.z = 0;
blit.src.box.width = blit.dst.box.width = width;
blit.src.box.height = blit.dst.box.height = height;
blit.src.box.depth = blit.dst.box.depth = 1;
blit.mask = st_get_blit_mask(rb->_BaseFormat, format);
blit.filter = PIPE_TEX_FILTER_NEAREST;
blit.scissor_enable = FALSE;
if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
blit.src.box.y = rb->Height - blit.src.box.y;
blit.src.box.height = -blit.src.box.height;
}
/* blit */
st->pipe->blit(st->pipe, &blit);
/* map resources */
pixels = _mesa_map_pbo_dest(ctx, pack, pixels);
map = pipe_transfer_map_3d(pipe, dst, 0, PIPE_TRANSFER_READ,
0, 0, 0, width, height, 1, &tex_xfer);
if (!map) {
_mesa_unmap_pbo_dest(ctx, pack);
pipe_resource_reference(&dst, NULL);
goto fallback;
}
/* memcpy data into a user buffer */
{
const uint bytesPerRow = width * util_format_get_blocksize(dst_format);
GLuint row;
for (row = 0; row < (unsigned) height; row++) {
GLvoid *dest = _mesa_image_address3d(pack, pixels,
width, height, format,
type, 0, row, 0);
memcpy(dest, map, bytesPerRow);
map += tex_xfer->stride;
}
}
pipe_transfer_unmap(pipe, tex_xfer);
_mesa_unmap_pbo_dest(ctx, pack);
pipe_resource_reference(&dst, NULL);
return;
fallback:
_mesa_readpixels(ctx, x, y, width, height, format, type, pack, pixels);
}
/**
* Fallback for pipe->clear_stencil() function.
* sw fallback doesn't look terribly useful here.
* Plus can't use these transfer fallbacks when clearing
* multisampled surfaces for instance.
* Clears all bound layers.
*/
void
util_clear_depth_stencil(struct pipe_context *pipe,
struct pipe_surface *dst,
unsigned clear_flags,
double depth,
unsigned stencil,
unsigned dstx, unsigned dsty,
unsigned width, unsigned height)
{
enum pipe_format format = dst->format;
struct pipe_transfer *dst_trans;
ubyte *dst_map;
boolean need_rmw = FALSE;
unsigned max_layer, layer;
if ((clear_flags & PIPE_CLEAR_DEPTHSTENCIL) &&
((clear_flags & PIPE_CLEAR_DEPTHSTENCIL) != PIPE_CLEAR_DEPTHSTENCIL) &&
util_format_is_depth_and_stencil(format))
need_rmw = TRUE;
assert(dst->texture);
if (!dst->texture)
return;
max_layer = dst->u.tex.last_layer - dst->u.tex.first_layer;
dst_map = pipe_transfer_map_3d(pipe,
dst->texture,
dst->u.tex.level,
(need_rmw ? PIPE_TRANSFER_READ_WRITE :
PIPE_TRANSFER_WRITE),
dstx, dsty, dst->u.tex.first_layer,
width, height, max_layer + 1, &dst_trans);
assert(dst_map);
if (dst_map) {
unsigned dst_stride = dst_trans->stride;
uint64_t zstencil = util_pack64_z_stencil(format, depth, stencil);
ubyte *dst_layer = dst_map;
unsigned i, j;
assert(dst_trans->stride > 0);
for (layer = 0; layer <= max_layer; layer++) {
dst_map = dst_layer;
switch (util_format_get_blocksize(format)) {
case 1:
assert(format == PIPE_FORMAT_S8_UINT);
if(dst_stride == width)
memset(dst_map, (uint8_t) zstencil, height * width);
else {
for (i = 0; i < height; i++) {
memset(dst_map, (uint8_t) zstencil, width);
dst_map += dst_stride;
}
}
break;
case 2:
assert(format == PIPE_FORMAT_Z16_UNORM);
for (i = 0; i < height; i++) {
uint16_t *row = (uint16_t *)dst_map;
for (j = 0; j < width; j++)
*row++ = (uint16_t) zstencil;
dst_map += dst_stride;
}
break;
case 4:
if (!need_rmw) {
for (i = 0; i < height; i++) {
uint32_t *row = (uint32_t *)dst_map;
for (j = 0; j < width; j++)
*row++ = (uint32_t) zstencil;
dst_map += dst_stride;
}
}
else {
uint32_t dst_mask;
if (format == PIPE_FORMAT_Z24_UNORM_S8_UINT)
dst_mask = 0x00ffffff;
else {
assert(format == PIPE_FORMAT_S8_UINT_Z24_UNORM);
dst_mask = 0xffffff00;
}
if (clear_flags & PIPE_CLEAR_DEPTH)
dst_mask = ~dst_mask;
for (i = 0; i < height; i++) {
uint32_t *row = (uint32_t *)dst_map;
for (j = 0; j < width; j++) {
uint32_t tmp = *row & dst_mask;
*row++ = tmp | ((uint32_t) zstencil & ~dst_mask);
}
dst_map += dst_stride;
}
}
break;
case 8:
if (!need_rmw) {
for (i = 0; i < height; i++) {
uint64_t *row = (uint64_t *)dst_map;
for (j = 0; j < width; j++)
*row++ = zstencil;
dst_map += dst_stride;
}
}
else {
uint64_t src_mask;
if (clear_flags & PIPE_CLEAR_DEPTH)
src_mask = 0x00000000ffffffffull;
else
src_mask = 0x000000ff00000000ull;
for (i = 0; i < height; i++) {
uint64_t *row = (uint64_t *)dst_map;
for (j = 0; j < width; j++) {
uint64_t tmp = *row & ~src_mask;
*row++ = tmp | (zstencil & src_mask);
}
dst_map += dst_stride;
}
}
break;
default:
assert(0);
break;
}
dst_layer += dst_trans->layer_stride;
}
pipe->transfer_unmap(pipe, dst_trans);
}
}
/**
* Fallback for pipe->clear_render_target() function.
* XXX this looks too hackish to be really useful.
* cpp > 4 looks like a gross hack at best...
* Plus can't use these transfer fallbacks when clearing
* multisampled surfaces for instance.
* Clears all bound layers.
*/
void
util_clear_render_target(struct pipe_context *pipe,
struct pipe_surface *dst,
const union pipe_color_union *color,
unsigned dstx, unsigned dsty,
unsigned width, unsigned height)
{
struct pipe_transfer *dst_trans;
ubyte *dst_map;
union util_color uc;
unsigned max_layer;
assert(dst->texture);
if (!dst->texture)
return;
if (dst->texture->target == PIPE_BUFFER) {
/*
* The fill naturally works on the surface format, however
* the transfer uses resource format which is just bytes for buffers.
*/
unsigned dx, w;
unsigned pixstride = util_format_get_blocksize(dst->format);
dx = (dst->u.buf.first_element + dstx) * pixstride;
w = width * pixstride;
max_layer = 0;
dst_map = pipe_transfer_map(pipe,
dst->texture,
0, 0,
PIPE_TRANSFER_WRITE,
dx, 0, w, 1,
&dst_trans);
}
else {
max_layer = dst->u.tex.last_layer - dst->u.tex.first_layer;
dst_map = pipe_transfer_map_3d(pipe,
dst->texture,
dst->u.tex.level,
PIPE_TRANSFER_WRITE,
dstx, dsty, dst->u.tex.first_layer,
width, height, max_layer + 1, &dst_trans);
}
assert(dst_map);
if (dst_map) {
enum pipe_format format = dst->format;
assert(dst_trans->stride > 0);
if (util_format_is_pure_integer(format)) {
/*
* We expect int/uint clear values here, though some APIs
* might disagree (but in any case util_pack_color()
* couldn't handle it)...
*/
if (util_format_is_pure_sint(format)) {
util_format_write_4i(format, color->i, 0, &uc, 0, 0, 0, 1, 1);
}
else {
assert(util_format_is_pure_uint(format));
util_format_write_4ui(format, color->ui, 0, &uc, 0, 0, 0, 1, 1);
}
}
else {
util_pack_color(color->f, format, &uc);
}
util_fill_box(dst_map, dst->format,
dst_trans->stride, dst_trans->layer_stride,
0, 0, 0, width, height, max_layer + 1, &uc);
pipe->transfer_unmap(pipe, dst_trans);
}
}
