FilterRows_NEON - remove unused function - same as InterpolateRow_NEON

- Bump version to 1872
- Add scale_rvv to build files

Bug: libyuv:956
Change-Id: Ib9e9fd840a0774bd35bcdcca55a2596f33272383
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4608519
Reviewed-by: Mirko Bonadei <mbonadei@chromium.org>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
This commit is contained in:
Frank Barchard 2023-06-13 07:56:05 -07:00 committed by libyuv LUCI CQ
parent 873eaa3bbf
commit 2a5d7e2fbc
5 changed files with 4 additions and 97 deletions

View File

@ -72,6 +72,7 @@ cc_library {
"source/scale_neon.cc", "source/scale_neon.cc",
"source/scale_neon64.cc", "source/scale_neon64.cc",
"source/scale_rgb.cc", "source/scale_rgb.cc",
"source/scale_rvv.cc",
"source/scale_uv.cc", "source/scale_uv.cc",
"source/video_common.cc", "source/video_common.cc",
], ],

View File

@ -142,6 +142,7 @@ static_library("libyuv_internal") {
"source/scale_common.cc", "source/scale_common.cc",
"source/scale_gcc.cc", "source/scale_gcc.cc",
"source/scale_rgb.cc", "source/scale_rgb.cc",
"source/scale_rvv.cc",
"source/scale_uv.cc", "source/scale_uv.cc",
"source/scale_win.cc", "source/scale_win.cc",
"source/video_common.cc", "source/video_common.cc",

View File

@ -1,6 +1,6 @@
Name: libyuv Name: libyuv
URL: http://code.google.com/p/libyuv/ URL: http://code.google.com/p/libyuv/
Version: 1871 Version: 1872
License: BSD License: BSD
License File: LICENSE License File: LICENSE

View File

@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ #ifndef INCLUDE_LIBYUV_VERSION_H_
#define INCLUDE_LIBYUV_VERSION_H_ #define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 1871 #define LIBYUV_VERSION 1872
#endif // INCLUDE_LIBYUV_VERSION_H_ #endif // INCLUDE_LIBYUV_VERSION_H_

View File

@ -1118,101 +1118,6 @@ void ScaleFilterCols_NEON(uint8_t* dst_ptr,
#undef LOAD2_DATA8_LANE #undef LOAD2_DATA8_LANE
// 16x2 -> 16x1
void ScaleFilterRows_NEON(uint8_t* dst_ptr,
const uint8_t* src_ptr,
ptrdiff_t src_stride,
int dst_width,
int source_y_fraction) {
int y_fraction = 256 - source_y_fraction;
asm volatile(
"cmp %w4, #0 \n"
"b.eq 100f \n"
"add %2, %2, %1 \n"
"cmp %w4, #64 \n"
"b.eq 75f \n"
"cmp %w4, #128 \n"
"b.eq 50f \n"
"cmp %w4, #192 \n"
"b.eq 25f \n"
"dup v5.8b, %w4 \n"
"dup v4.8b, %w5 \n"
// General purpose row blend.
"1: \n"
"ld1 {v0.16b}, [%1], #16 \n"
"ld1 {v1.16b}, [%2], #16 \n"
"subs %w3, %w3, #16 \n"
"umull v6.8h, v0.8b, v4.8b \n"
"umull2 v7.8h, v0.16b, v4.16b \n"
"prfm pldl1keep, [%1, 448] \n" // prefetch 7 lines ahead
"umlal v6.8h, v1.8b, v5.8b \n"
"umlal2 v7.8h, v1.16b, v5.16b \n"
"prfm pldl1keep, [%2, 448] \n"
"rshrn v0.8b, v6.8h, #8 \n"
"rshrn2 v0.16b, v7.8h, #8 \n"
"st1 {v0.16b}, [%0], #16 \n"
"b.gt 1b \n"
"b 99f \n"
// Blend 25 / 75.
"25: \n"
"ld1 {v0.16b}, [%1], #16 \n"
"ld1 {v1.16b}, [%2], #16 \n"
"subs %w3, %w3, #16 \n"
"urhadd v0.16b, v0.16b, v1.16b \n"
"prfm pldl1keep, [%1, 448] \n" // prefetch 7 lines ahead
"urhadd v0.16b, v0.16b, v1.16b \n"
"prfm pldl1keep, [%2, 448] \n"
"st1 {v0.16b}, [%0], #16 \n"
"b.gt 25b \n"
"b 99f \n"
// Blend 50 / 50.
"50: \n"
"ld1 {v0.16b}, [%1], #16 \n"
"ld1 {v1.16b}, [%2], #16 \n"
"subs %w3, %w3, #16 \n"
"prfm pldl1keep, [%1, 448] \n" // prefetch 7 lines ahead
"urhadd v0.16b, v0.16b, v1.16b \n"
"prfm pldl1keep, [%2, 448] \n"
"st1 {v0.16b}, [%0], #16 \n"
"b.gt 50b \n"
"b 99f \n"
// Blend 75 / 25.
"75: \n"
"ld1 {v1.16b}, [%1], #16 \n"
"ld1 {v0.16b}, [%2], #16 \n"
"subs %w3, %w3, #16 \n"
"urhadd v0.16b, v0.16b, v1.16b \n"
"prfm pldl1keep, [%1, 448] \n" // prefetch 7 lines ahead
"urhadd v0.16b, v0.16b, v1.16b \n"
"prfm pldl1keep, [%2, 448] \n"
"st1 {v0.16b}, [%0], #16 \n"
"b.gt 75b \n"
"b 99f \n"
// Blend 100 / 0 - Copy row unchanged.
"100: \n"
"ld1 {v0.16b}, [%1], #16 \n"
"subs %w3, %w3, #16 \n"
"prfm pldl1keep, [%1, 448] \n" // prefetch 7 lines ahead
"st1 {v0.16b}, [%0], #16 \n"
"b.gt 100b \n"
"99: \n"
"st1 {v0.b}[15], [%0] \n"
: "+r"(dst_ptr), // %0
"+r"(src_ptr), // %1
"+r"(src_stride), // %2
"+r"(dst_width), // %3
"+r"(source_y_fraction), // %4
"+r"(y_fraction) // %5
:
: "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "memory", "cc");
}
void ScaleARGBRowDown2_NEON(const uint8_t* src_ptr, void ScaleARGBRowDown2_NEON(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst, uint8_t* dst,