diff options
| author | Justin Green <greenjustin@google.com> | 2023-03-14 10:23:17 -0400 |
|---|---|---|
| committer | libyuv LUCI CQ <libyuv-scoped@luci-project-accounts.iam.gserviceaccount.com> | 2023-03-14 14:59:26 +0000 |
| commit | 76468711d5c8302431a900499ff73d34fdfc146b (patch) | |
| tree | a91e3f09e5b124c688eb809bb0b288db42eb8d46 /source/row_neon64.cc | |
| parent | f9b23b9cc0ca3bd27b9acc07ea0450cd5097175d (diff) | |
| download | libyuv-76468711d5c8302431a900499ff73d34fdfc146b.tar.gz | |
M2T2 Unpack fixes
Fix the algorithm for unpacking the lower 2 bits of M2T2 pixels.
Bug: b:258474032
Change-Id: Iea1d63f26e3f127a70ead26bc04ea3d939e793e3
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4337978
Commit-Queue: Justin Green <greenjustin@google.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Diffstat (limited to 'source/row_neon64.cc')
| -rw-r--r-- | source/row_neon64.cc | 70 |
1 files changed, 30 insertions, 40 deletions
diff --git a/source/row_neon64.cc b/source/row_neon64.cc index 7f04b606..3afb5a20 100644 --- a/source/row_neon64.cc +++ b/source/row_neon64.cc @@ -752,49 +752,39 @@ void DetileToYUY2_NEON(const uint8_t* src_y, // Unpack MT2T into tiled P010 64 pixels at a time. See // tinyurl.com/mtk-10bit-video-format for format documentation. void UnpackMT2T_NEON(const uint8_t* src, uint16_t* dst, size_t size) { - const uint8_t* src_lower_bits = src; - const uint8_t* src_upper_bits = src + 16; asm volatile( - "1: \n" - "ld4 {v0.8b, v1.8b, v2.8b, v3.8b}, [%1], #32 \n" - "ld1 {v7.8b}, [%0], #8 \n" - "shl v6.8b, v7.8b, #2 \n" - "shl v5.8b, v7.8b, #4 \n" - "shl v4.8b, v7.8b, #6 \n" - "zip1 v0.16b, v4.16b, v0.16b \n" - "zip1 v1.16b, v5.16b, v1.16b \n" - "zip1 v2.16b, v6.16b, v2.16b \n" - "zip1 v3.16b, v7.16b, v3.16b \n" - "sri v0.8h, v0.8h, #10 \n" - "sri v1.8h, v1.8h, #10 \n" - "sri v2.8h, v2.8h, #10 \n" - "sri v3.8h, v3.8h, #10 \n" - "st4 {v0.8h, v1.8h, v2.8h, v3.8h}, [%2], #64 \n" - "ld4 {v0.8b, v1.8b, v2.8b, v3.8b}, [%1], #32 \n" - "ld1 {v7.8b}, [%0], #8 \n" - "shl v6.8b, v7.8b, #2 \n" - "shl v5.8b, v7.8b, #4 \n" - "shl v4.8b, v7.8b, #6 \n" - "zip1 v0.16b, v4.16b, v0.16b \n" - "zip1 v1.16b, v5.16b, v1.16b \n" - "zip1 v2.16b, v6.16b, v2.16b \n" - "zip1 v3.16b, v7.16b, v3.16b \n" - "sri v0.8h, v0.8h, #10 \n" - "sri v1.8h, v1.8h, #10 \n" - "sri v2.8h, v2.8h, #10 \n" - "sri v3.8h, v3.8h, #10 \n" - "st4 {v0.8h, v1.8h, v2.8h, v3.8h}, [%2], #64 \n" - "mov %0, %1 \n" - "add %1, %0, #16 \n" - "subs %3, %3, #80 \n" - "b.gt 1b \n" - : "+r"(src_lower_bits), // %0 - "+r"(src_upper_bits), // %1 - "+r"(dst), // %2 - "+r"(size) // %3 + "1: \n" + "ld1 {v7.16b}, [%0], #16 \n" + "ld1 {v0.16b-v3.16b}, [%0], #64 \n" + "shl v4.16b, v7.16b, #6 \n" + "shl v5.16b, v7.16b, #4 \n" + "shl v6.16b, v7.16b, #2 \n" + "subs %2, %2, #80 \n" + "zip1 v16.16b, v4.16b, v0.16b \n" + "zip1 v18.16b, v5.16b, v1.16b \n" + "zip1 v20.16b, v6.16b, v2.16b \n" + "zip1 v22.16b, v7.16b, v3.16b \n" + "zip2 v17.16b, v4.16b, v0.16b \n" + "zip2 v19.16b, v5.16b, v1.16b \n" + "zip2 v21.16b, v6.16b, v2.16b \n" + "zip2 v23.16b, v7.16b, v3.16b \n" + "sri v16.8h, v16.8h, #10 \n" + "sri v17.8h, v17.8h, #10 \n" + "sri v18.8h, v18.8h, #10 \n" + "sri v19.8h, v19.8h, #10 \n" + "st1 {v16.8h-v19.8h}, [%1], #64 \n" + "sri v20.8h, v20.8h, #10 \n" + "sri v21.8h, v21.8h, #10 \n" + "sri v22.8h, v22.8h, #10 \n" + "sri v23.8h, v23.8h, #10 \n" + "st1 {v20.8h-v23.8h}, [%1], #64 \n" + "b.gt 1b \n" + : "+r"(src), // %0 + "+r"(dst), // %1 + "+r"(size) // %2 : : "cc", "memory", "w0", "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", - "v8", "v9", "v10", "v11", "v12"); + "v16", "v17", "v18", "v19", "v20", "v21", "v22", "v23"); } #if LIBYUV_USE_ST2 |