coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-07 01:06:46 +08:00
Code Issues Packages Projects Releases Wiki Activity

Lint fix for C casting for rotation code on arm

Browse Source 
instead of casting int to int64, pass the int
and use %w modifier to use the word version of the register.

TBR=kjellander@chromium.org
BUG=libyuv:706
TEST=git cl lint
R=wangcheng@google.com

Change-Id: Iee5a70f04d928903ca8efac00066b8821a465e36
Reviewed-on: https://chromium-review.googlesource.com/528381
Reviewed-by: Cheng Wang <wangcheng@google.com>
Reviewed-by: Frank Barchard <fbarchard@google.com>
This commit is contained in:
Frank Barchard 2017-06-08 17:48:40 -07:00
parent d981495b42
commit 5f94a33e0c
2 changed files with 22 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
source/rotate_neon64.cc
View File

@ -30,12 +30,11 @@ void TransposeWx8_NEON(const uint8* src,
int dst_stride, int dst_stride,
int width) { int width) {
const uint8* src_temp; const uint8* src_temp;
int64 width64 = (int64)width; // Work around clang 3.4 warning.
asm volatile ( asm volatile (
// loops are on blocks of 8. loop will stop when // loops are on blocks of 8. loop will stop when
// counter gets to or below 0. starting the counter // counter gets to or below 0. starting the counter
// at w-8 allow for this // at w-8 allow for this
"sub %3, %3, #8 \n" "sub %w3, %w3, #8 \n"
// handle 8x8 blocks. this should be the majority of the plane // handle 8x8 blocks. this should be the majority of the plane
"1: \n" "1: \n"
@ -106,19 +105,19 @@ void TransposeWx8_NEON(const uint8* src,
"add %1, %1, #8 \n" // src += 8 "add %1, %1, #8 \n" // src += 8
"add %2, %2, %6, lsl #3 \n" // dst += 8 * dst_stride "add %2, %2, %6, lsl #3 \n" // dst += 8 * dst_stride
"subs %3, %3, #8 \n" // w -= 8 "subs %w3, %w3, #8 \n" // w -= 8
"b.ge 1b \n" "b.ge 1b \n"
// add 8 back to counter. if the result is 0 there are // add 8 back to counter. if the result is 0 there are
// no residuals. // no residuals.
"adds %3, %3, #8 \n" "adds %w3, %w3, #8 \n"
"b.eq 4f \n" "b.eq 4f \n"
// some residual, so between 1 and 7 lines left to transpose // some residual, so between 1 and 7 lines left to transpose
"cmp %3, #2 \n" "cmp %w3, #2 \n"
"b.lt 3f \n" "b.lt 3f \n"
"cmp %3, #4 \n" "cmp %w3, #4 \n"
"b.lt 2f \n" "b.lt 2f \n"
// 4x8 block // 4x8 block
@ -171,12 +170,12 @@ void TransposeWx8_NEON(const uint8* src,
"add %1, %1, #4 \n" // src += 4 "add %1, %1, #4 \n" // src += 4
"add %2, %2, %6, lsl #2 \n" // dst += 4 * dst_stride "add %2, %2, %6, lsl #2 \n" // dst += 4 * dst_stride
"subs %3, %3, #4 \n" // w -= 4 "subs %w3, %w3, #4 \n" // w -= 4
"b.eq 4f \n" "b.eq 4f \n"
// some residual, check to see if it includes a 2x8 block, // some residual, check to see if it includes a 2x8 block,
// or less // or less
"cmp %3, #2 \n" "cmp %w3, #2 \n"
"b.lt 3f \n" "b.lt 3f \n"
// 2x8 block // 2x8 block
@ -211,7 +210,7 @@ void TransposeWx8_NEON(const uint8* src,
"add %1, %1, #2 \n" // src += 2 "add %1, %1, #2 \n" // src += 2
"add %2, %2, %6, lsl #1 \n" // dst += 2 * dst_stride "add %2, %2, %6, lsl #1 \n" // dst += 2 * dst_stride
"subs %3, %3, #2 \n" // w -= 2 "subs %w3, %w3, #2 \n" // w -= 2
"b.eq 4f \n" "b.eq 4f \n"
// 1x8 block // 1x8 block
@ -241,7 +240,7 @@ void TransposeWx8_NEON(const uint8* src,
: "=&r"(src_temp), // %0 : "=&r"(src_temp), // %0
"+r"(src), // %1 "+r"(src), // %1
"+r"(dst), // %2 "+r"(dst), // %2
"+r"(width64) // %3 "+r"(width) // %3
: "r"(&kVTbl4x4Transpose), // %4 : "r"(&kVTbl4x4Transpose), // %4
"r"(static_cast<ptrdiff_t>(src_stride)), // %5 "r"(static_cast<ptrdiff_t>(src_stride)), // %5
"r"(static_cast<ptrdiff_t>(dst_stride)) // %6 "r"(static_cast<ptrdiff_t>(dst_stride)) // %6
@ -262,12 +261,11 @@ void TransposeUVWx8_NEON(const uint8* src,
int dst_stride_b, int dst_stride_b,
int width) { int width) {
const uint8* src_temp; const uint8* src_temp;
int64 width64 = (int64)width; // Work around clang 3.4 warning.
asm volatile ( asm volatile (
// loops are on blocks of 8. loop will stop when // loops are on blocks of 8. loop will stop when
// counter gets to or below 0. starting the counter // counter gets to or below 0. starting the counter
// at w-8 allow for this // at w-8 allow for this
"sub %4, %4, #8 \n" "sub %w4, %w4, #8 \n"
// handle 8x8 blocks. this should be the majority of the plane // handle 8x8 blocks. this should be the majority of the plane
"1: \n" "1: \n"
@ -358,19 +356,19 @@ void TransposeUVWx8_NEON(const uint8* src,
"add %1, %1, #16 \n" // src += 8*2 "add %1, %1, #16 \n" // src += 8*2
"add %2, %2, %6, lsl #3 \n" // dst_a += 8 * dst_stride_a "add %2, %2, %6, lsl #3 \n" // dst_a += 8 * dst_stride_a
"add %3, %3, %7, lsl #3 \n" // dst_b += 8 * dst_stride_b "add %3, %3, %7, lsl #3 \n" // dst_b += 8 * dst_stride_b
"subs %4, %4, #8 \n" // w -= 8 "subs %w4, %w4, #8 \n" // w -= 8
"b.ge 1b \n" "b.ge 1b \n"
// add 8 back to counter. if the result is 0 there are // add 8 back to counter. if the result is 0 there are
// no residuals. // no residuals.
"adds %4, %4, #8 \n" "adds %w4, %w4, #8 \n"
"b.eq 4f \n" "b.eq 4f \n"
// some residual, so between 1 and 7 lines left to transpose // some residual, so between 1 and 7 lines left to transpose
"cmp %4, #2 \n" "cmp %w4, #2 \n"
"b.lt 3f \n" "b.lt 3f \n"
"cmp %4, #4 \n" "cmp %w4, #4 \n"
"b.lt 2f \n" "b.lt 2f \n"
// TODO(frkoenig): Clean this up // TODO(frkoenig): Clean this up
@ -447,12 +445,12 @@ void TransposeUVWx8_NEON(const uint8* src,
"add %1, %1, #8 \n" // src += 4 * 2 "add %1, %1, #8 \n" // src += 4 * 2
"add %2, %2, %6, lsl #2 \n" // dst_a += 4 * dst_stride_a "add %2, %2, %6, lsl #2 \n" // dst_a += 4 * dst_stride_a
"add %3, %3, %7, lsl #2 \n" // dst_b += 4 * dst_stride_b "add %3, %3, %7, lsl #2 \n" // dst_b += 4 * dst_stride_b
"subs %4, %4, #4 \n" // w -= 4 "subs %w4, %w4, #4 \n" // w -= 4
"b.eq 4f \n" "b.eq 4f \n"
// some residual, check to see if it includes a 2x8 block, // some residual, check to see if it includes a 2x8 block,
// or less // or less
"cmp %4, #2 \n" "cmp %w4, #2 \n"
"b.lt 3f \n" "b.lt 3f \n"
// 2x8 block // 2x8 block
@ -497,7 +495,7 @@ void TransposeUVWx8_NEON(const uint8* src,
"add %1, %1, #4 \n" // src += 2 * 2 "add %1, %1, #4 \n" // src += 2 * 2
"add %2, %2, %6, lsl #1 \n" // dst_a += 2 * dst_stride_a "add %2, %2, %6, lsl #1 \n" // dst_a += 2 * dst_stride_a
"add %3, %3, %7, lsl #1 \n" // dst_b += 2 * dst_stride_b "add %3, %3, %7, lsl #1 \n" // dst_b += 2 * dst_stride_b
"subs %4, %4, #2 \n" // w -= 2 "subs %w4, %w4, #2 \n" // w -= 2
"b.eq 4f \n" "b.eq 4f \n"
// 1x8 block // 1x8 block
@ -530,7 +528,7 @@ void TransposeUVWx8_NEON(const uint8* src,
"+r"(src), // %1 "+r"(src), // %1
"+r"(dst_a), // %2 "+r"(dst_a), // %2
"+r"(dst_b), // %3 "+r"(dst_b), // %3
"+r"(width64) // %4 "+r"(width) // %4
: "r"(static_cast<ptrdiff_t>(src_stride)), // %5 : "r"(static_cast<ptrdiff_t>(src_stride)), // %5
"r"(static_cast<ptrdiff_t>(dst_stride_a)), // %6 "r"(static_cast<ptrdiff_t>(dst_stride_a)), // %6
"r"(static_cast<ptrdiff_t>(dst_stride_b)), // %7 "r"(static_cast<ptrdiff_t>(dst_stride_b)), // %7

9
source/scale_neon64.cc
View File

@ -622,7 +622,6 @@ void ScaleFilterCols_NEON(uint8* dst_ptr,
int dx_offset[4] = {0, 1, 2, 3}; int dx_offset[4] = {0, 1, 2, 3};
int* tmp = dx_offset; int* tmp = dx_offset;
const uint8* src_tmp = src_ptr; const uint8* src_tmp = src_ptr;
int64 dst_width64 = (int64)dst_width; // Work around ios 64 bit warning.
int64 x64 = (int64)x; int64 x64 = (int64)x;
int64 dx64 = (int64)dx; int64 dx64 = (int64)dx;
asm volatile ( asm volatile (
@ -669,7 +668,7 @@ void ScaleFilterCols_NEON(uint8* dst_ptr,
"b.gt 1b \n" "b.gt 1b \n"
: "+r"(dst_ptr), // %0 : "+r"(dst_ptr), // %0
"+r"(src_ptr), // %1 "+r"(src_ptr), // %1
"+r"(dst_width64), // %2 "+r"(dst_width), // %2
"+r"(x64), // %3 "+r"(x64), // %3
"+r"(dx64), // %4 "+r"(dx64), // %4
"+r"(tmp), // %5 "+r"(tmp), // %5
@ -979,7 +978,6 @@ void ScaleARGBCols_NEON(uint8* dst_argb,
int x, int x,
int dx) { int dx) {
const uint8* src_tmp = src_argb; const uint8* src_tmp = src_argb;
int64 dst_width64 = (int64)dst_width; // Work around ios 64 bit warning.
int64 x64 = (int64)x; int64 x64 = (int64)x;
int64 dx64 = (int64)dx; int64 dx64 = (int64)dx;
int64 tmp64; int64 tmp64;
@ -1000,7 +998,7 @@ void ScaleARGBCols_NEON(uint8* dst_argb,
"b.gt 1b \n" "b.gt 1b \n"
: "+r"(dst_argb), // %0 : "+r"(dst_argb), // %0
"+r"(src_argb), // %1 "+r"(src_argb), // %1
"+r"(dst_width64), // %2 "+r"(dst_width), // %2
"+r"(x64), // %3 "+r"(x64), // %3
"+r"(dx64), // %4 "+r"(dx64), // %4
"=&r"(tmp64), // %5 "=&r"(tmp64), // %5
@ -1031,7 +1029,6 @@ void ScaleARGBFilterCols_NEON(uint8* dst_argb,
int dx_offset[4] = {0, 1, 2, 3}; int dx_offset[4] = {0, 1, 2, 3};
int* tmp = dx_offset; int* tmp = dx_offset;
const uint8* src_tmp = src_argb; const uint8* src_tmp = src_argb;
int64 dst_width64 = (int64)dst_width; // Work around ios 64 bit warning.
int64 x64 = (int64)x; int64 x64 = (int64)x;
int64 dx64 = (int64)dx; int64 dx64 = (int64)dx;
asm volatile ( asm volatile (
@ -1077,7 +1074,7 @@ void ScaleARGBFilterCols_NEON(uint8* dst_argb,
"b.gt 1b \n" "b.gt 1b \n"
: "+r"(dst_argb), // %0 : "+r"(dst_argb), // %0
"+r"(src_argb), // %1 "+r"(src_argb), // %1
"+r"(dst_width64), // %2 "+r"(dst_width), // %2
"+r"(x64), // %3 "+r"(x64), // %3
"+r"(dx64), // %4 "+r"(dx64), // %4
"+r"(tmp), // %5 "+r"(tmp), // %5