diff options
| author | Frank Barchard <fbarchard@google.com> | 2022-05-23 19:15:41 -0700 |
|---|---|---|
| committer | libyuv LUCI CQ <libyuv-scoped@luci-project-accounts.iam.gserviceaccount.com> | 2022-05-24 18:07:16 +0000 |
| commit | eb2c88e49903e602cb11b796d08de74aa0a8ef79 (patch) | |
| tree | e6fe5cd940935fd71c65b2673f0d05b9ef3cf29a /source/row_neon64.cc | |
| parent | 715150b5aab187315ab75029d0973fff6ff10322 (diff) | |
| download | libyuv-eb2c88e49903e602cb11b796d08de74aa0a8ef79.tar.gz | |
Convert16To8 NEON
Pixel 3
Was C I010ToI420_Opt (749 ms)
Now NEON I010ToI420_Opt (356 ms)
Pixel 4
Was C I010ToI420_Opt (581 ms)
Now NEON I010ToI420_Opt (163 ms)
Bug: b/233233302, b/233634772
Change-Id: I60a84648a66f77d97c0a7822b29bd18b8e3a3355
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/3661401
Reviewed-by: 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 | 52 |
1 files changed, 38 insertions, 14 deletions
diff --git a/source/row_neon64.cc b/source/row_neon64.cc index 27723810..8d43d594 100644 --- a/source/row_neon64.cc +++ b/source/row_neon64.cc @@ -4070,21 +4070,19 @@ void MultiplyRow_16_NEON(const uint16_t* src_y, int scale, int width) { asm volatile( - "dup v2.8h, %w2 \n" + "dup v2.8h, %w3 \n" "1: \n" "ldp q0, q1, [%0], #32 \n" "mul v0.8h, v0.8h, v2.8h \n" "prfm pldl1keep, [%0, 448] \n" "mul v1.8h, v1.8h, v2.8h \n" - "stp q0, q1, [%1] \n" // store 16 pixels - "add %1, %1, #32 \n" - "subs %w3, %w3, #16 \n" // 16 src pixels per loop + "stp q0, q1, [%1], #32 \n" // store 16 pixels + "subs %w2, %w2, #16 \n" // 16 src pixels per loop "b.gt 1b \n" : "+r"(src_y), // %0 "+r"(dst_y), // %1 - "+r"(scale), // %2 - "+r"(width) // %3 - : + "+r"(width) // %2 + : "r"(scale) // %3 : "cc", "memory", "v0", "v1", "v2"); } @@ -4093,7 +4091,7 @@ void DivideRow_16_NEON(const uint16_t* src_y, int scale, int width) { asm volatile( - "dup v0.8h, %w2 \n" + "dup v0.8h, %w3 \n" "1: \n" "ldp q1, q2, [%0], #32 \n" "ushll v3.4s, v1.4h, #0 \n" @@ -4109,18 +4107,44 @@ void DivideRow_16_NEON(const uint16_t* src_y, "shrn v4.4h, v4.4s, #16 \n" "shrn2 v3.8h, v1.4s, #16 \n" "shrn2 v4.8h, v2.4s, #16 \n" - "stp q3, q3, [%1] \n" // store 16 pixels - "add %1, %1, #32 \n" - "subs %w3, %w3, #16 \n" // 16 src pixels per loop + "stp q3, q3, [%1], #32 \n" // store 16 pixels + "subs %w2, %w2, #16 \n" // 16 src pixels per loop "b.gt 1b \n" : "+r"(src_y), // %0 "+r"(dst_y), // %1 - "+r"(scale), // %2 - "+r"(width) // %3 - : + "+r"(width) // %2 + : "r"(scale) // %3 : "cc", "memory", "v0", "v1", "v2", "v3", "v4"); } +// Use scale to convert lsb formats to msb, depending how many bits there are: +// 32768 = 9 bits +// 16384 = 10 bits +// 4096 = 12 bits +// 256 = 16 bits +void Convert16To8Row_NEON(const uint16_t* src_y, + uint8_t* dst_y, + int scale, + int width) { + asm volatile( + "dup v2.8h, %w3 \n" + "1: \n" + "ldp q0, q1, [%0], #32 \n" + "sqdmulh v0.8h, v0.8h, v2.8h \n" + "sqdmulh v1.8h, v1.8h, v2.8h \n" + "prfm pldl1keep, [%0, 448] \n" + "subs %w2, %w2, #16 \n" // 16 src pixels per loop + "uqshrn v0.8b, v0.8h, #1 \n" + "uqshrn2 v0.16b, v1.8h, #1 \n" + "str q0, [%1], #16 \n" // store 16 pixels + "b.gt 1b \n" + : "+r"(src_y), // %0 + "+r"(dst_y), // %1 + "+r"(width) // %2 + : "r"(scale) // %3 + : "cc", "memory", "v0", "v1", "v2"); +} + #endif // !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__) #ifdef __cplusplus |