summaryrefslogtreecommitdiffstats
path: root/src/video_core/textures
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core/textures')
-rw-r--r--src/video_core/textures/astc.cpp58
-rw-r--r--src/video_core/textures/decoders.cpp240
-rw-r--r--src/video_core/textures/decoders.h33
3 files changed, 128 insertions, 203 deletions
diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp
index e3f3d3c5d..15b9d4182 100644
--- a/src/video_core/textures/astc.cpp
+++ b/src/video_core/textures/astc.cpp
@@ -13,7 +13,9 @@
#include <boost/container/static_vector.hpp>
+#include "common/alignment.h"
#include "common/common_types.h"
+#include "common/thread_worker.h"
#include "video_core/textures/astc.h"
class InputBitStream {
@@ -1411,7 +1413,7 @@ static void FillVoidExtentLDR(InputBitStream& strm, std::span<u32> outBuf, u32 b
static void FillError(std::span<u32> outBuf, u32 blockWidth, u32 blockHeight) {
for (u32 j = 0; j < blockHeight; j++) {
for (u32 i = 0; i < blockWidth; i++) {
- outBuf[j * blockWidth + i] = 0xFFFF00FF;
+ outBuf[j * blockWidth + i] = 0x00000000;
}
}
}
@@ -1650,29 +1652,41 @@ static void DecompressBlock(std::span<const u8, 16> inBuf, const u32 blockWidth,
void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
uint32_t block_width, uint32_t block_height, std::span<uint8_t> output) {
- u32 block_index = 0;
- std::size_t depth_offset = 0;
- for (u32 z = 0; z < depth; z++) {
- for (u32 y = 0; y < height; y += block_height) {
- for (u32 x = 0; x < width; x += block_width) {
- const std::span<const u8, 16> blockPtr{data.subspan(block_index * 16, 16)};
-
- // Blocks can be at most 12x12
- std::array<u32, 12 * 12> uncompData;
- DecompressBlock(blockPtr, block_width, block_height, uncompData);
-
- u32 decompWidth = std::min(block_width, width - x);
- u32 decompHeight = std::min(block_height, height - y);
-
- const std::span<u8> outRow = output.subspan(depth_offset + (y * width + x) * 4);
- for (u32 jj = 0; jj < decompHeight; jj++) {
- std::memcpy(outRow.data() + jj * width * 4,
- uncompData.data() + jj * block_width, decompWidth * 4);
+ const u32 rows = Common::DivideUp(height, block_height);
+ const u32 cols = Common::DivideUp(width, block_width);
+
+ Common::ThreadWorker workers{std::max(std::thread::hardware_concurrency(), 2U) / 2,
+ "ASTCDecompress"};
+
+ for (u32 z = 0; z < depth; ++z) {
+ const u32 depth_offset = z * height * width * 4;
+ for (u32 y_index = 0; y_index < rows; ++y_index) {
+ auto decompress_stride = [data, width, height, depth, block_width, block_height, output,
+ rows, cols, z, depth_offset, y_index] {
+ const u32 y = y_index * block_height;
+ for (u32 x_index = 0; x_index < cols; ++x_index) {
+ const u32 block_index = (z * rows * cols) + (y_index * cols) + x_index;
+ const u32 x = x_index * block_width;
+
+ const std::span<const u8, 16> blockPtr{data.subspan(block_index * 16, 16)};
+
+ // Blocks can be at most 12x12
+ std::array<u32, 12 * 12> uncompData;
+ DecompressBlock(blockPtr, block_width, block_height, uncompData);
+
+ u32 decompWidth = std::min(block_width, width - x);
+ u32 decompHeight = std::min(block_height, height - y);
+
+ const std::span<u8> outRow = output.subspan(depth_offset + (y * width + x) * 4);
+ for (u32 h = 0; h < decompHeight; ++h) {
+ std::memcpy(outRow.data() + h * width * 4,
+ uncompData.data() + h * block_width, decompWidth * 4);
+ }
}
- ++block_index;
- }
+ };
+ workers.QueueWork(std::move(decompress_stride));
}
- depth_offset += height * width * 4;
+ workers.WaitForRequests();
}
}
diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp
index 913f8ebcb..52d067a2d 100644
--- a/src/video_core/textures/decoders.cpp
+++ b/src/video_core/textures/decoders.cpp
@@ -35,7 +35,7 @@ void incrpdep(u32& value) {
template <bool TO_LINEAR, u32 BYTES_PER_PIXEL>
void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32 height, u32 depth,
- u32 block_height, u32 block_depth, u32 stride_alignment) {
+ u32 block_height, u32 block_depth, u32 stride) {
// The origin of the transformation can be configured here, leave it as zero as the current API
// doesn't expose it.
static constexpr u32 origin_x = 0;
@@ -45,7 +45,6 @@ void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32
// We can configure here a custom pitch
// As it's not exposed 'width * BYTES_PER_PIXEL' will be the expected pitch.
const u32 pitch = width * BYTES_PER_PIXEL;
- const u32 stride = Common::AlignUpLog2(width, stride_alignment) * BYTES_PER_PIXEL;
const u32 gobs_in_x = Common::DivCeilLog2(stride, GOB_SIZE_X_SHIFT);
const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth);
@@ -89,6 +88,69 @@ void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32
}
}
+template <bool TO_LINEAR, u32 BYTES_PER_PIXEL>
+void SwizzleSubrectImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32 height,
+ u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, u32 num_lines,
+ u32 block_height, u32 block_depth, u32 pitch_linear) {
+ // The origin of the transformation can be configured here, leave it as zero as the current API
+ // doesn't expose it.
+ static constexpr u32 origin_z = 0;
+
+ // We can configure here a custom pitch
+ // As it's not exposed 'width * BYTES_PER_PIXEL' will be the expected pitch.
+ const u32 pitch = pitch_linear;
+ const u32 stride = Common::AlignUpLog2(width * BYTES_PER_PIXEL, GOB_SIZE_X_SHIFT);
+
+ const u32 gobs_in_x = Common::DivCeilLog2(stride, GOB_SIZE_X_SHIFT);
+ const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth);
+ const u32 slice_size =
+ Common::DivCeilLog2(height, block_height + GOB_SIZE_Y_SHIFT) * block_size;
+
+ const u32 block_height_mask = (1U << block_height) - 1;
+ const u32 block_depth_mask = (1U << block_depth) - 1;
+ const u32 x_shift = GOB_SIZE_SHIFT + block_height + block_depth;
+
+ u32 unprocessed_lines = num_lines;
+ u32 extent_y = std::min(num_lines, height - origin_y);
+
+ for (u32 slice = 0; slice < depth; ++slice) {
+ const u32 z = slice + origin_z;
+ const u32 offset_z = (z >> block_depth) * slice_size +
+ ((z & block_depth_mask) << (GOB_SIZE_SHIFT + block_height));
+ const u32 lines_in_y = std::min(unprocessed_lines, extent_y);
+ for (u32 line = 0; line < lines_in_y; ++line) {
+ const u32 y = line + origin_y;
+ const u32 swizzled_y = pdep<SWIZZLE_Y_BITS>(y);
+
+ const u32 block_y = y >> GOB_SIZE_Y_SHIFT;
+ const u32 offset_y = (block_y >> block_height) * block_size +
+ ((block_y & block_height_mask) << GOB_SIZE_SHIFT);
+
+ u32 swizzled_x = pdep<SWIZZLE_X_BITS>(origin_x * BYTES_PER_PIXEL);
+ for (u32 column = 0; column < extent_x;
+ ++column, incrpdep<SWIZZLE_X_BITS, BYTES_PER_PIXEL>(swizzled_x)) {
+ const u32 x = (column + origin_x) * BYTES_PER_PIXEL;
+ const u32 offset_x = (x >> GOB_SIZE_X_SHIFT) << x_shift;
+
+ const u32 base_swizzled_offset = offset_z + offset_y + offset_x;
+ const u32 swizzled_offset = base_swizzled_offset + (swizzled_x | swizzled_y);
+
+ const u32 unswizzled_offset =
+ slice * pitch * height + line * pitch + column * BYTES_PER_PIXEL;
+
+ u8* const dst = &output[TO_LINEAR ? swizzled_offset : unswizzled_offset];
+ const u8* const src = &input[TO_LINEAR ? unswizzled_offset : swizzled_offset];
+
+ std::memcpy(dst, src, BYTES_PER_PIXEL);
+ }
+ }
+ unprocessed_lines -= lines_in_y;
+ if (unprocessed_lines == 0) {
+ return;
+ }
+ }
+}
+
template <bool TO_LINEAR>
void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width,
u32 height, u32 depth, u32 block_height, u32 block_depth, u32 stride_alignment) {
@@ -111,122 +173,39 @@ void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixe
}
}
-template <u32 BYTES_PER_PIXEL>
-void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
- u8* swizzled_data, const u8* unswizzled_data, u32 block_height_bit,
- u32 offset_x, u32 offset_y) {
- const u32 block_height = 1U << block_height_bit;
- const u32 image_width_in_gobs =
- (swizzled_width * BYTES_PER_PIXEL + (GOB_SIZE_X - 1)) / GOB_SIZE_X;
- for (u32 line = 0; line < subrect_height; ++line) {
- const u32 dst_y = line + offset_y;
- const u32 gob_address_y =
- (dst_y / (GOB_SIZE_Y * block_height)) * GOB_SIZE * block_height * image_width_in_gobs +
- ((dst_y % (GOB_SIZE_Y * block_height)) / GOB_SIZE_Y) * GOB_SIZE;
-
- const u32 swizzled_y = pdep<SWIZZLE_Y_BITS>(dst_y);
- u32 swizzled_x = pdep<SWIZZLE_X_BITS>(offset_x * BYTES_PER_PIXEL);
- for (u32 x = 0; x < subrect_width;
- ++x, incrpdep<SWIZZLE_X_BITS, BYTES_PER_PIXEL>(swizzled_x)) {
- const u32 dst_x = x + offset_x;
- const u32 gob_address =
- gob_address_y + (dst_x * BYTES_PER_PIXEL / GOB_SIZE_X) * GOB_SIZE * block_height;
- const u32 swizzled_offset = gob_address + (swizzled_x | swizzled_y);
- const u32 unswizzled_offset = line * source_pitch + x * BYTES_PER_PIXEL;
-
- const u8* const source_line = unswizzled_data + unswizzled_offset;
- u8* const dest_addr = swizzled_data + swizzled_offset;
- std::memcpy(dest_addr, source_line, BYTES_PER_PIXEL);
- }
- }
-}
-
-template <u32 BYTES_PER_PIXEL>
-void UnswizzleSubrect(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 block_height,
- u32 origin_x, u32 origin_y, u8* output, const u8* input) {
- const u32 stride = width * BYTES_PER_PIXEL;
- const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) / GOB_SIZE_X;
- const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height);
-
- const u32 block_height_mask = (1U << block_height) - 1;
- const u32 x_shift = GOB_SIZE_SHIFT + block_height;
-
- for (u32 line = 0; line < line_count; ++line) {
- const u32 src_y = line + origin_y;
- const u32 swizzled_y = pdep<SWIZZLE_Y_BITS>(src_y);
-
- const u32 block_y = src_y >> GOB_SIZE_Y_SHIFT;
- const u32 src_offset_y = (block_y >> block_height) * block_size +
- ((block_y & block_height_mask) << GOB_SIZE_SHIFT);
-
- u32 swizzled_x = pdep<SWIZZLE_X_BITS>(origin_x * BYTES_PER_PIXEL);
- for (u32 column = 0; column < line_length_in;
- ++column, incrpdep<SWIZZLE_X_BITS, BYTES_PER_PIXEL>(swizzled_x)) {
- const u32 src_x = (column + origin_x) * BYTES_PER_PIXEL;
- const u32 src_offset_x = (src_x >> GOB_SIZE_X_SHIFT) << x_shift;
-
- const u32 swizzled_offset = src_offset_y + src_offset_x + (swizzled_x | swizzled_y);
- const u32 unswizzled_offset = line * pitch + column * BYTES_PER_PIXEL;
-
- std::memcpy(output + unswizzled_offset, input + swizzled_offset, BYTES_PER_PIXEL);
- }
- }
-}
-
-template <u32 BYTES_PER_PIXEL>
-void SwizzleSliceToVoxel(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 height,
- u32 block_height, u32 block_depth, u32 origin_x, u32 origin_y, u8* output,
- const u8* input) {
- UNIMPLEMENTED_IF(origin_x > 0);
- UNIMPLEMENTED_IF(origin_y > 0);
-
- const u32 stride = width * BYTES_PER_PIXEL;
- const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) / GOB_SIZE_X;
- const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth);
-
- const u32 block_height_mask = (1U << block_height) - 1;
- const u32 x_shift = static_cast<u32>(GOB_SIZE_SHIFT) + block_height + block_depth;
-
- for (u32 line = 0; line < line_count; ++line) {
- const u32 swizzled_y = pdep<SWIZZLE_Y_BITS>(line);
- const u32 block_y = line / GOB_SIZE_Y;
- const u32 dst_offset_y =
- (block_y >> block_height) * block_size + (block_y & block_height_mask) * GOB_SIZE;
-
- u32 swizzled_x = 0;
- for (u32 x = 0; x < line_length_in; ++x, incrpdep<SWIZZLE_X_BITS, 1>(swizzled_x)) {
- const u32 dst_offset =
- ((x / GOB_SIZE_X) << x_shift) + dst_offset_y + (swizzled_x | swizzled_y);
- const u32 src_offset = x * BYTES_PER_PIXEL + line * pitch;
- std::memcpy(output + dst_offset, input + src_offset, BYTES_PER_PIXEL);
- }
- }
-}
} // Anonymous namespace
void UnswizzleTexture(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel,
u32 width, u32 height, u32 depth, u32 block_height, u32 block_depth,
u32 stride_alignment) {
+ const u32 stride = Common::AlignUpLog2(width, stride_alignment) * bytes_per_pixel;
+ const u32 new_bpp = std::min(4U, static_cast<u32>(std::countr_zero(width * bytes_per_pixel)));
+ width = (width * bytes_per_pixel) >> new_bpp;
+ bytes_per_pixel = 1U << new_bpp;
Swizzle<false>(output, input, bytes_per_pixel, width, height, depth, block_height, block_depth,
- stride_alignment);
+ stride);
}
void SwizzleTexture(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width,
u32 height, u32 depth, u32 block_height, u32 block_depth,
u32 stride_alignment) {
+ const u32 stride = Common::AlignUpLog2(width, stride_alignment) * bytes_per_pixel;
+ const u32 new_bpp = std::min(4U, static_cast<u32>(std::countr_zero(width * bytes_per_pixel)));
+ width = (width * bytes_per_pixel) >> new_bpp;
+ bytes_per_pixel = 1U << new_bpp;
Swizzle<true>(output, input, bytes_per_pixel, width, height, depth, block_height, block_depth,
- stride_alignment);
+ stride);
}
-void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
- u32 bytes_per_pixel, u8* swizzled_data, const u8* unswizzled_data,
- u32 block_height_bit, u32 offset_x, u32 offset_y) {
+void SwizzleSubrect(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width,
+ u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, u32 extent_y,
+ u32 block_height, u32 block_depth, u32 pitch_linear) {
switch (bytes_per_pixel) {
#define BPP_CASE(x) \
case x: \
- return SwizzleSubrect<x>(subrect_width, subrect_height, source_pitch, swizzled_width, \
- swizzled_data, unswizzled_data, block_height_bit, offset_x, \
- offset_y);
+ return SwizzleSubrectImpl<true, x>(output, input, width, height, depth, origin_x, \
+ origin_y, extent_x, extent_y, block_height, \
+ block_depth, pitch_linear);
BPP_CASE(1)
BPP_CASE(2)
BPP_CASE(3)
@@ -241,13 +220,15 @@ void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32
}
}
-void UnswizzleSubrect(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 bytes_per_pixel,
- u32 block_height, u32 origin_x, u32 origin_y, u8* output, const u8* input) {
+void UnswizzleSubrect(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel,
+ u32 width, u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x,
+ u32 extent_y, u32 block_height, u32 block_depth, u32 pitch_linear) {
switch (bytes_per_pixel) {
#define BPP_CASE(x) \
case x: \
- return UnswizzleSubrect<x>(line_length_in, line_count, pitch, width, block_height, \
- origin_x, origin_y, output, input);
+ return SwizzleSubrectImpl<false, x>(output, input, width, height, depth, origin_x, \
+ origin_y, extent_x, extent_y, block_height, \
+ block_depth, pitch_linear);
BPP_CASE(1)
BPP_CASE(2)
BPP_CASE(3)
@@ -262,55 +243,6 @@ void UnswizzleSubrect(u32 line_length_in, u32 line_count, u32 pitch, u32 width,
}
}
-void SwizzleSliceToVoxel(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 height,
- u32 bytes_per_pixel, u32 block_height, u32 block_depth, u32 origin_x,
- u32 origin_y, u8* output, const u8* input) {
- switch (bytes_per_pixel) {
-#define BPP_CASE(x) \
- case x: \
- return SwizzleSliceToVoxel<x>(line_length_in, line_count, pitch, width, height, \
- block_height, block_depth, origin_x, origin_y, output, \
- input);
- BPP_CASE(1)
- BPP_CASE(2)
- BPP_CASE(3)
- BPP_CASE(4)
- BPP_CASE(6)
- BPP_CASE(8)
- BPP_CASE(12)
- BPP_CASE(16)
-#undef BPP_CASE
- default:
- ASSERT_MSG(false, "Invalid bytes_per_pixel={}", bytes_per_pixel);
- }
-}
-
-void SwizzleKepler(const u32 width, const u32 height, const u32 dst_x, const u32 dst_y,
- const u32 block_height_bit, const std::size_t copy_size, const u8* source_data,
- u8* swizzle_data) {
- const u32 block_height = 1U << block_height_bit;
- const u32 image_width_in_gobs{(width + GOB_SIZE_X - 1) / GOB_SIZE_X};
- std::size_t count = 0;
- for (std::size_t y = dst_y; y < height && count < copy_size; ++y) {
- const std::size_t gob_address_y =
- (y / (GOB_SIZE_Y * block_height)) * GOB_SIZE * block_height * image_width_in_gobs +
- ((y % (GOB_SIZE_Y * block_height)) / GOB_SIZE_Y) * GOB_SIZE;
- const u32 swizzled_y = pdep<SWIZZLE_Y_BITS>(static_cast<u32>(y));
- u32 swizzled_x = pdep<SWIZZLE_X_BITS>(dst_x);
- for (std::size_t x = dst_x; x < width && count < copy_size;
- ++x, incrpdep<SWIZZLE_X_BITS, 1>(swizzled_x)) {
- const std::size_t gob_address =
- gob_address_y + (x / GOB_SIZE_X) * GOB_SIZE * block_height;
- const std::size_t swizzled_offset = gob_address + (swizzled_x | swizzled_y);
- const u8* source_line = source_data + count;
- u8* dest_addr = swizzle_data + swizzled_offset;
- count++;
-
- *dest_addr = *source_line;
- }
- }
-}
-
std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
u32 block_height, u32 block_depth) {
if (tiled) {
diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h
index 31a11708f..e70407692 100644
--- a/src/video_core/textures/decoders.h
+++ b/src/video_core/textures/decoders.h
@@ -40,7 +40,6 @@ constexpr SwizzleTable MakeSwizzleTable() {
}
return table;
}
-constexpr SwizzleTable SWIZZLE_TABLE = MakeSwizzleTable();
/// Unswizzles a block linear texture into linear memory.
void UnswizzleTexture(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel,
@@ -57,34 +56,14 @@ std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height
u32 block_height, u32 block_depth);
/// Copies an untiled subrectangle into a tiled surface.
-void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
- u32 bytes_per_pixel, u8* swizzled_data, const u8* unswizzled_data,
- u32 block_height_bit, u32 offset_x, u32 offset_y);
+void SwizzleSubrect(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width,
+ u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x, u32 extent_y,
+ u32 block_height, u32 block_depth, u32 pitch_linear);
/// Copies a tiled subrectangle into a linear surface.
-void UnswizzleSubrect(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 bytes_per_pixel,
- u32 block_height, u32 origin_x, u32 origin_y, u8* output, const u8* input);
-
-/// @brief Swizzles a 2D array of pixels into a 3D texture
-/// @param line_length_in Number of pixels per line
-/// @param line_count Number of lines
-/// @param pitch Number of bytes per line
-/// @param width Width of the swizzled texture
-/// @param height Height of the swizzled texture
-/// @param bytes_per_pixel Number of bytes used per pixel
-/// @param block_height Block height shift
-/// @param block_depth Block depth shift
-/// @param origin_x Column offset in pixels of the swizzled texture
-/// @param origin_y Row offset in pixels of the swizzled texture
-/// @param output Pointer to the pixels of the swizzled texture
-/// @param input Pointer to the 2D array of pixels used as input
-/// @pre input and output points to an array large enough to hold the number of bytes used
-void SwizzleSliceToVoxel(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 height,
- u32 bytes_per_pixel, u32 block_height, u32 block_depth, u32 origin_x,
- u32 origin_y, u8* output, const u8* input);
-
-void SwizzleKepler(u32 width, u32 height, u32 dst_x, u32 dst_y, u32 block_height,
- std::size_t copy_size, const u8* source_data, u8* swizzle_data);
+void UnswizzleSubrect(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel,
+ u32 width, u32 height, u32 depth, u32 origin_x, u32 origin_y, u32 extent_x,
+ u32 extent_y, u32 block_height, u32 block_depth, u32 pitch_linear);
/// Obtains the offset of the gob for positions 'dst_x' & 'dst_y'
u64 GetGOBOffset(u32 width, u32 height, u32 dst_x, u32 dst_y, u32 block_height,