dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2024-12-27 06:08:21 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

NEON/simple loopfilter: avoid q4-q7 registers

Browse Source 
very tiny speed improvement

Change-Id: I3024f120feb7275ce20bfff21af31ea8650a5a03
This commit is contained in:
James Zern 2013-09-06 20:22:14 -04:00
parent 2f09d63e30
commit fc10249b36
1 changed files with 10 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
src/dsp/dec_neon.c
View File

@ -22,7 +22,7 @@ extern "C" {
#include "../dec/vp8i.h" #include "../dec/vp8i.h"
#define QRegs "q0", "q1", "q2", "q3", "q4", "q5", "q6", "q7", \ #define QRegs "q0", "q1", "q2", "q3", \
"q8", "q9", "q10", "q11", "q12", "q13", "q14", "q15" "q8", "q9", "q10", "q11", "q12", "q13", "q14", "q15"
#define FLIP_SIGN_BIT2(a, b, s) \ #define FLIP_SIGN_BIT2(a, b, s) \
@ -101,9 +101,9 @@ static void SimpleVFilter16NEON(uint8_t* p, int stride, int thresh) {
"vld1.u8 {q1}, [%[p]], %[stride] \n" // p1 "vld1.u8 {q1}, [%[p]], %[stride] \n" // p1
"vld1.u8 {q2}, [%[p]], %[stride] \n" // p0 "vld1.u8 {q2}, [%[p]], %[stride] \n" // p0
"vld1.u8 {q3}, [%[p]], %[stride] \n" // q0 "vld1.u8 {q3}, [%[p]], %[stride] \n" // q0
"vld1.u8 {q4}, [%[p]] \n" // q1 "vld1.u8 {q12}, [%[p]] \n" // q1
DO_FILTER2(q1, q2, q3, q4, %[thresh]) DO_FILTER2(q1, q2, q3, q12, %[thresh])
"sub %[p], %[p], %[stride], lsl #1 \n" // p -= 2 * stride "sub %[p], %[p], %[stride], lsl #1 \n" // p -= 2 * stride
@ -122,18 +122,18 @@ static void SimpleHFilter16NEON(uint8_t* p, int stride, int thresh) {
"add r5, r4, %[stride] \n" // base2 = base1 + stride "add r5, r4, %[stride] \n" // base2 = base1 + stride
LOAD8x4(d2, d3, d4, d5, [r4], [r5], r6) LOAD8x4(d2, d3, d4, d5, [r4], [r5], r6)
LOAD8x4(d6, d7, d8, d9, [r4], [r5], r6) LOAD8x4(d24, d25, d26, d27, [r4], [r5], r6)
"vswp d3, d6 \n" // p1:q1 p0:q3 "vswp d3, d24 \n" // p1:q1 p0:q3
"vswp d5, d8 \n" // q0:q2 q1:q4 "vswp d5, d26 \n" // q0:q2 q1:q4
"vswp q2, q3 \n" // p1:q1 p0:q2 q0:q3 q1:q4 "vswp q2, q12 \n" // p1:q1 p0:q2 q0:q3 q1:q4
DO_FILTER2(q1, q2, q3, q4, %[thresh]) DO_FILTER2(q1, q2, q12, q13, %[thresh])
"sub %[p], %[p], #1 \n" // p - 1 "sub %[p], %[p], #1 \n" // p - 1
"vswp d5, d6 \n" "vswp d5, d24 \n"
STORE8x2(d4, d5, [%[p]], %[stride]) STORE8x2(d4, d5, [%[p]], %[stride])
STORE8x2(d6, d7, [%[p]], %[stride]) STORE8x2(d24, d25, [%[p]], %[stride])
: [p] "+r"(p) : [p] "+r"(p)
: [stride] "r"(stride), [thresh] "r"(thresh) : [stride] "r"(stride), [thresh] "r"(thresh)