coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-06 08:46:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

[AArch64] Optimize ScaleARGBRowDown2Linear_NEON

Browse Source 
Replace LD4 with a pair of LD2 instructions to avoid needing an ST2
instruction for storing the result, since ST2 instructions are known to
be slow on some micro-architectures.

Observed reduction in runtimes compared to the existing Neon code:

 Cortex-A55: -23.3%
Cortex-A510: -49.6%
Cortex-A520: -31.1%
 Cortex-A76: -44.5%
Cortex-A715: -45.8%
Cortex-A720: -46.0%
  Cortex-X1: -74.5%
  Cortex-X2: -72.4%
  Cortex-X3: -76.8%
  Cortex-X4: -39.5%

Co-authored-by: Cosmina Dunca <cosmina.dunca@arm.com>
Bug: libyuv:976
Change-Id: Iab9e802d0784d69b7e970dcc8f1f4036985cd2e1
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/5790972
Reviewed-by: Justin Green <greenjustin@google.com>
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
This commit is contained in:
George Steed 2024-05-16 17:13:30 +01:00 committed by Frank Barchard
parent 00886670bb
commit e6297afd14
1 changed files with 16 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
source/scale_neon64.cc
View File

@ -1181,23 +1181,24 @@ void ScaleARGBRowDown2Linear_NEON(const uint8_t* src_argb,
uint8_t* dst_argb,
int dst_width) {
(void)src_stride;
asm volatile (
const uint8_t* src_argb1 = src_argb + 32;
asm volatile(
"1: \n"
// load 16 ARGB pixels with even pixels into q0/q2, odd into q1/q3
"ld4 {v0.4s,v1.4s,v2.4s,v3.4s}, [%0], #64 \n"
"subs %w2, %w2, #8 \n" // 8 processed per loop
"urhadd v0.16b, v0.16b, v1.16b \n" // rounding half add
"prfm pldl1keep, [%0, 448] \n" // prefetch 7 lines ahead
"urhadd v1.16b, v2.16b, v3.16b \n"
"st2 {v0.4s,v1.4s}, [%1], #32 \n" // store 8 pixels
"b.gt 1b \n"
: "+r"(src_argb), // %0
"+r"(dst_argb), // %1
"+r"(dst_width) // %2
"ld2 {v0.4s, v1.4s}, [%[src]] \n"
"add %[src], %[src], #64 \n"
"ld2 {v2.4s, v3.4s}, [%[src1]] \n"
"add %[src1], %[src1], #64 \n"
"urhadd v0.16b, v0.16b, v1.16b \n"
"urhadd v1.16b, v2.16b, v3.16b \n"
"subs %w[width], %w[width], #8 \n"
"st1 {v0.16b, v1.16b}, [%[dst]], #32 \n"
"b.gt 1b \n"
: [src] "+r"(src_argb), // %[src]
[src1] "+r"(src_argb1), // %[src1]
[dst] "+r"(dst_argb), // %[dst]
[width] "+r"(dst_width) // %[width]
:
: "memory", "cc", "v0", "v1", "v2", "v3" // Clobber List
);
: "memory", "cc", "v0", "v1", "v2", "v3");
}
void ScaleARGBRowDown2Box_NEON(const uint8_t* src_ptr,