Skip to content

L
libyuv
Commit a62a97f1 authored by fbarchard@google.com's avatar fbarchard@google.com
Browse files
Options

Change branch notation to clang compatible b dot cc 

BUG=357
TESTED=local ios a64 build
R=yunqingwang@google.com

Review URL: https://webrtc-codereview.appspot.com/25549004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1084 16f28f9a-4ce2-e073-06de-1de4eb20be90
parent 8cbfc5d4
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 38 additions and 38 deletions
source/compare_neon.cc
View file @ a62a97f1
...@@ -80,7 +80,7 @@ uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count) { ...@@ -80,7 +80,7 @@ uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count) {
"smlal v17.4s, v3.4h, v3.4h \n" "smlal v17.4s, v3.4h, v3.4h \n"
"smlal2 v18.4s, v2.8h, v2.8h \n" "smlal2 v18.4s, v2.8h, v2.8h \n"
"smlal2 v19.4s, v3.8h, v3.8h \n" "smlal2 v19.4s, v3.8h, v3.8h \n"
"bgt 1b \n" "b.gt 1b \n"
"add v16.4s, v16.4s, v17.4s \n" "add v16.4s, v16.4s, v17.4s \n"
"add v18.4s, v18.4s, v19.4s \n" "add v18.4s, v18.4s, v19.4s \n"
......
source/rotate_neon64.cc
View file @ a62a97f1
...@@ -104,19 +104,19 @@ void TransposeWx8_NEON(const uint8* src, int src_stride, ...@@ -104,19 +104,19 @@ void TransposeWx8_NEON(const uint8* src, int src_stride,
"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 %3, %3, #8 \n" // w -= 8
"bge 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 %3, %3, #8 \n"
"beq 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 %3, #2 \n"
"blt 3f \n" "b.lt 3f \n"
"cmp %3, #4 \n" "cmp %3, #4 \n"
"blt 2f \n" "b.lt 2f \n"
// 4x8 block // 4x8 block
"mov %0, %1 \n" "mov %0, %1 \n"
...@@ -169,12 +169,12 @@ void TransposeWx8_NEON(const uint8* src, int src_stride, ...@@ -169,12 +169,12 @@ void TransposeWx8_NEON(const uint8* src, int src_stride,
"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 %3, %3, #4 \n" // w -= 4
"beq 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 %3, #2 \n"
"blt 3f \n" "b.lt 3f \n"
// 2x8 block // 2x8 block
"2: \n" "2: \n"
...@@ -209,7 +209,7 @@ void TransposeWx8_NEON(const uint8* src, int src_stride, ...@@ -209,7 +209,7 @@ void TransposeWx8_NEON(const uint8* src, int src_stride,
"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 %3, %3, #2 \n" // w -= 2
"beq 4f \n" "b.eq 4f \n"
// 1x8 block // 1x8 block
"3: \n" "3: \n"
...@@ -352,19 +352,19 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride, ...@@ -352,19 +352,19 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride,
"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 %4, %4, #8 \n" // w -= 8
"bge 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 %4, %4, #8 \n"
"beq 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 %4, #2 \n"
"blt 3f \n" "b.lt 3f \n"
"cmp %4, #4 \n" "cmp %4, #4 \n"
"blt 2f \n" "b.lt 2f \n"
// TODO(frkoenig): Clean this up // TODO(frkoenig): Clean this up
// 4x8 block // 4x8 block
...@@ -441,12 +441,12 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride, ...@@ -441,12 +441,12 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride,
"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 %4, %4, #4 \n" // w -= 4
"beq 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 %4, #2 \n"
"blt 3f \n" "b.lt 3f \n"
// 2x8 block // 2x8 block
"2: \n" "2: \n"
...@@ -491,7 +491,7 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride, ...@@ -491,7 +491,7 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride,
"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 %4, %4, #2 \n" // w -= 2
"beq 4f \n" "b.eq 4f \n"
// 1x8 block // 1x8 block
"3: \n" "3: \n"
......
source/row_neon64.cc
View file @ a62a97f1
This diff is collapsed.
source/scale_neon64.cc
View file @ a62a97f1
...@@ -32,7 +32,7 @@ void ScaleRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride, ...@@ -32,7 +32,7 @@ void ScaleRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
"subs %2, %2, #16 \n" // 16 processed per loop "subs %2, %2, #16 \n" // 16 processed per loop
MEMACCESS(1) MEMACCESS(1)
"st1 {v1.16b}, [%1], #16 \n" // store odd pixels "st1 {v1.16b}, [%1], #16 \n" // store odd pixels
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_ptr), // %0 : "+r"(src_ptr), // %0
"+r"(dst), // %1 "+r"(dst), // %1
"+r"(dst_width) // %2 "+r"(dst_width) // %2
...@@ -63,7 +63,7 @@ void ScaleRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride, ...@@ -63,7 +63,7 @@ void ScaleRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
"rshrn2 v0.16b, v1.8h, #2 \n" "rshrn2 v0.16b, v1.8h, #2 \n"
MEMACCESS(2) MEMACCESS(2)
"st1 {v0.16b}, [%2], #16 \n" "st1 {v0.16b}, [%2], #16 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_ptr), // %0 : "+r"(src_ptr), // %0
"+r"(src_stride), // %1 "+r"(src_stride), // %1
"+r"(dst), // %2 "+r"(dst), // %2
...@@ -84,7 +84,7 @@ void ScaleRowDown4_NEON(const uint8* src_ptr, ptrdiff_t src_stride, ...@@ -84,7 +84,7 @@ void ScaleRowDown4_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
"subs %2, %2, #8 \n" // 8 processed per loop "subs %2, %2, #8 \n" // 8 processed per loop
MEMACCESS(1) MEMACCESS(1)
"st1 {v2.8b}, [%1], #8 \n" "st1 {v2.8b}, [%1], #8 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_ptr), // %0 : "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1 "+r"(dst_ptr), // %1
"+r"(dst_width) // %2 "+r"(dst_width) // %2
...@@ -119,7 +119,7 @@ asm volatile ( ...@@ -119,7 +119,7 @@ asm volatile (
"rshrn v0.8b, v0.8h, #4 \n" // divide by 16 w/rounding "rshrn v0.8b, v0.8h, #4 \n" // divide by 16 w/rounding
MEMACCESS(1) MEMACCESS(1)
"st1 {v0.s}[0], [%1], #4 \n" "st1 {v0.s}[0], [%1], #4 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_ptr), // %0 : "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1 "+r"(dst_ptr), // %1
"+r"(dst_width), // %2 "+r"(dst_width), // %2
...@@ -147,7 +147,7 @@ void ScaleRowDown34_NEON(const uint8* src_ptr, ...@@ -147,7 +147,7 @@ void ScaleRowDown34_NEON(const uint8* src_ptr,
"mov v2.8b, v3.8b \n" // order v0, v1, v2 "mov v2.8b, v3.8b \n" // order v0, v1, v2
MEMACCESS(1) MEMACCESS(1)
"st3 {v0.8b-v2.8b}, [%1], #24 \n" "st3 {v0.8b-v2.8b}, [%1], #24 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_ptr), // %0 : "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1 "+r"(dst_ptr), // %1
"+r"(dst_width) // %2 "+r"(dst_width) // %2
...@@ -207,7 +207,7 @@ void ScaleRowDown34_0_Box_NEON(const uint8* src_ptr, ...@@ -207,7 +207,7 @@ void ScaleRowDown34_0_Box_NEON(const uint8* src_ptr,
MEMACCESS(1) MEMACCESS(1)
"st3 {v0.8b-v2.8b}, [%1], #24 \n" "st3 {v0.8b-v2.8b}, [%1], #24 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_ptr), // %0 : "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1 "+r"(dst_ptr), // %1
"+r"(dst_width), // %2 "+r"(dst_width), // %2
...@@ -253,7 +253,7 @@ void ScaleRowDown34_1_Box_NEON(const uint8* src_ptr, ...@@ -253,7 +253,7 @@ void ScaleRowDown34_1_Box_NEON(const uint8* src_ptr,
MEMACCESS(1) MEMACCESS(1)
"st3 {v0.8b-v2.8b}, [%1], #24 \n" "st3 {v0.8b-v2.8b}, [%1], #24 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_ptr), // %0 : "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1 "+r"(dst_ptr), // %1
"+r"(dst_width), // %2 "+r"(dst_width), // %2
...@@ -292,7 +292,7 @@ void ScaleRowDown38_NEON(const uint8* src_ptr, ...@@ -292,7 +292,7 @@ void ScaleRowDown38_NEON(const uint8* src_ptr,
"st1 {v2.8b}, [%1], #8 \n" "st1 {v2.8b}, [%1], #8 \n"
MEMACCESS(1) MEMACCESS(1)
"st1 {v2.s}[2], [%1], #4 \n" "st1 {v2.s}[2], [%1], #4 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_ptr), // %0 : "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1 "+r"(dst_ptr), // %1
"+r"(dst_width) // %2 "+r"(dst_width) // %2
...@@ -417,7 +417,7 @@ void OMITFP ScaleRowDown38_3_Box_NEON(const uint8* src_ptr, ...@@ -417,7 +417,7 @@ void OMITFP ScaleRowDown38_3_Box_NEON(const uint8* src_ptr,
"st1 {v3.8b}, [%1], #8 \n" "st1 {v3.8b}, [%1], #8 \n"
MEMACCESS(1) MEMACCESS(1)
"st1 {v3.s}[2], [%1], #4 \n" "st1 {v3.s}[2], [%1], #4 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_ptr), // %0 : "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1 "+r"(dst_ptr), // %1
"+r"(dst_width), // %2 "+r"(dst_width), // %2
...@@ -527,7 +527,7 @@ void ScaleRowDown38_2_Box_NEON(const uint8* src_ptr, ...@@ -527,7 +527,7 @@ void ScaleRowDown38_2_Box_NEON(const uint8* src_ptr,
"st1 {v3.8b}, [%1], #8 \n" "st1 {v3.8b}, [%1], #8 \n"
MEMACCESS(1) MEMACCESS(1)
"st1 {v3.s}[2], [%1], #4 \n" "st1 {v3.s}[2], [%1], #4 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_ptr), // %0 : "+r"(src_ptr), // %0
"+r"(dst_ptr), // %1 "+r"(dst_ptr), // %1
"+r"(dst_width), // %2 "+r"(dst_width), // %2
...@@ -547,14 +547,14 @@ void ScaleFilterRows_NEON(uint8* dst_ptr, ...@@ -547,14 +547,14 @@ void ScaleFilterRows_NEON(uint8* dst_ptr,
int y_fraction = 256 - source_y_fraction; int y_fraction = 256 - source_y_fraction;
asm volatile ( asm volatile (
"cmp %4, #0 \n" "cmp %4, #0 \n"
"beq 100f \n" "b.eq 100f \n"
"add %2, %2, %1 \n" "add %2, %2, %1 \n"
"cmp %4, #64 \n" "cmp %4, #64 \n"
"beq 75f \n" "b.eq 75f \n"
"cmp %4, #128 \n" "cmp %4, #128 \n"
"beq 50f \n" "b.eq 50f \n"
"cmp %4, #192 \n" "cmp %4, #192 \n"
"beq 25f \n" "b.eq 25f \n"
"dup v5.8b, %w4 \n" "dup v5.8b, %w4 \n"
"dup v4.8b, %w5 \n" "dup v4.8b, %w5 \n"
...@@ -573,7 +573,7 @@ void ScaleFilterRows_NEON(uint8* dst_ptr, ...@@ -573,7 +573,7 @@ void ScaleFilterRows_NEON(uint8* dst_ptr,
"rshrn2 v0.16b, v7.8h, #8 \n" "rshrn2 v0.16b, v7.8h, #8 \n"
MEMACCESS(0) MEMACCESS(0)
"st1 {v0.16b}, [%0], #16 \n" "st1 {v0.16b}, [%0], #16 \n"
"bgt 1b \n" "b.gt 1b \n"
"b 99f \n" "b 99f \n"
// Blend 25 / 75. // Blend 25 / 75.
...@@ -587,7 +587,7 @@ void ScaleFilterRows_NEON(uint8* dst_ptr, ...@@ -587,7 +587,7 @@ void ScaleFilterRows_NEON(uint8* dst_ptr,
"urhadd v0.16b, v0.16b, v1.16b \n" "urhadd v0.16b, v0.16b, v1.16b \n"
MEMACCESS(0) MEMACCESS(0)
"st1 {v0.16b}, [%0], #16 \n" "st1 {v0.16b}, [%0], #16 \n"
"bgt 25b \n" "b.gt 25b \n"
"b 99f \n" "b 99f \n"
// Blend 50 / 50. // Blend 50 / 50.
...@@ -600,7 +600,7 @@ void ScaleFilterRows_NEON(uint8* dst_ptr, ...@@ -600,7 +600,7 @@ void ScaleFilterRows_NEON(uint8* dst_ptr,
"urhadd v0.16b, v0.16b, v1.16b \n" "urhadd v0.16b, v0.16b, v1.16b \n"
MEMACCESS(0) MEMACCESS(0)
"st1 {v0.16b}, [%0], #16 \n" "st1 {v0.16b}, [%0], #16 \n"
"bgt 50b \n" "b.gt 50b \n"
"b 99f \n" "b 99f \n"
// Blend 75 / 25. // Blend 75 / 25.
...@@ -614,7 +614,7 @@ void ScaleFilterRows_NEON(uint8* dst_ptr, ...@@ -614,7 +614,7 @@ void ScaleFilterRows_NEON(uint8* dst_ptr,
"urhadd v0.16b, v0.16b, v1.16b \n" "urhadd v0.16b, v0.16b, v1.16b \n"
MEMACCESS(0) MEMACCESS(0)
"st1 {v0.16b}, [%0], #16 \n" "st1 {v0.16b}, [%0], #16 \n"
"bgt 75b \n" "b.gt 75b \n"
"b 99f \n" "b 99f \n"
// Blend 100 / 0 - Copy row unchanged. // Blend 100 / 0 - Copy row unchanged.
...@@ -624,7 +624,7 @@ void ScaleFilterRows_NEON(uint8* dst_ptr, ...@@ -624,7 +624,7 @@ void ScaleFilterRows_NEON(uint8* dst_ptr,
"subs %3, %3, #16 \n" "subs %3, %3, #16 \n"
MEMACCESS(0) MEMACCESS(0)
"st1 {v0.16b}, [%0], #16 \n" "st1 {v0.16b}, [%0], #16 \n"
"bgt 100b \n" "b.gt 100b \n"
"99: \n" "99: \n"
MEMACCESS(0) MEMACCESS(0)
...@@ -655,7 +655,7 @@ void ScaleARGBRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride, ...@@ -655,7 +655,7 @@ void ScaleARGBRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
"st1 {v1.16b}, [%1], #16 \n" // store odd pixels "st1 {v1.16b}, [%1], #16 \n" // store odd pixels
MEMACCESS (1) MEMACCESS (1)
"st1 {v3.16b}, [%1], #16 \n" "st1 {v3.16b}, [%1], #16 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r" (src_ptr), // %0 : "+r" (src_ptr), // %0
"+r" (dst), // %1 "+r" (dst), // %1
"+r" (dst_width) // %2 "+r" (dst_width) // %2
...@@ -691,7 +691,7 @@ void ScaleARGBRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride, ...@@ -691,7 +691,7 @@ void ScaleARGBRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
"rshrn v3.8b, v3.8h, #2 \n" "rshrn v3.8b, v3.8h, #2 \n"
MEMACCESS (2) MEMACCESS (2)
"st4 {v0.8b - v3.8b}, [%2], #32 \n" "st4 {v0.8b - v3.8b}, [%2], #32 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r" (src_ptr), // %0 : "+r" (src_ptr), // %0
"+r" (src_stride), // %1 "+r" (src_stride), // %1
"+r" (dst), // %2 "+r" (dst), // %2
...@@ -720,7 +720,7 @@ void ScaleARGBRowDownEven_NEON(const uint8* src_argb, ptrdiff_t src_stride, ...@@ -720,7 +720,7 @@ void ScaleARGBRowDownEven_NEON(const uint8* src_argb, ptrdiff_t src_stride,
"subs %2, %2, #4 \n" // 4 pixels per loop. "subs %2, %2, #4 \n" // 4 pixels per loop.
MEMACCESS(1) MEMACCESS(1)
"st1 {v0.16b}, [%1], #16 \n" "st1 {v0.16b}, [%1], #16 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_argb), // %0 : "+r"(src_argb), // %0
"+r"(dst_argb), // %1 "+r"(dst_argb), // %1
"+r"(dst_width) // %2 "+r"(dst_width) // %2
...@@ -774,7 +774,7 @@ void ScaleARGBRowDownEvenBox_NEON(const uint8* src_argb, ptrdiff_t src_stride, ...@@ -774,7 +774,7 @@ void ScaleARGBRowDownEvenBox_NEON(const uint8* src_argb, ptrdiff_t src_stride,
"subs %3, %3, #4 \n" // 4 pixels per loop. "subs %3, %3, #4 \n" // 4 pixels per loop.
MEMACCESS(2) MEMACCESS(2)
"st1 {v0.16b}, [%2], #16 \n" "st1 {v0.16b}, [%2], #16 \n"
"bgt 1b \n" "b.gt 1b \n"
: "+r"(src_argb), // %0 : "+r"(src_argb), // %0
"+r"(src_stride), // %1 "+r"(src_stride), // %1
"+r"(dst_argb), // %2 "+r"(dst_argb), // %2
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment