Commit 0d41aee2 authored by fbarchard@google.com's avatar fbarchard@google.com

Port compare functions to Nacl

BUG=253
TEST=none
R=nfullagar@google.com

Review URL: https://webrtc-codereview.appspot.com/1998004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@752 16f28f9a-4ce2-e073-06de-1de4eb20be90
parent 341b17a2
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 751
Version: 752
License: BSD
License File: LICENSE
......
......@@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 751
#define LIBYUV_VERSION 752
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT
......@@ -31,7 +31,6 @@ uint32 HashDjb2_C(const uint8* src, int count, uint32 seed);
// This module is for Visual C x86
#if !defined(LIBYUV_DISABLE_X86) && \
!(defined(__native_client__) && defined(__x86_64__)) && \
(defined(_M_IX86) || \
(defined(__x86_64__) || (defined(__i386__) && !defined(__pic__))))
#define HAS_HASHDJB2_SSE41
......@@ -76,7 +75,6 @@ uint32 SumSquareError_C(const uint8* src_a, const uint8* src_b, int count);
uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count);
#endif
#if !defined(LIBYUV_DISABLE_X86) && \
!(defined(__native_client__) && defined(__x86_64__)) && \
(defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
#define HAS_SUMSQUAREERROR_SSE2
uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count);
......
......@@ -16,21 +16,28 @@ namespace libyuv {
extern "C" {
#endif
#if !defined(LIBYUV_DISABLE_X86) && \
!(defined(__native_client__) && defined(__x86_64__)) && \
(defined(__x86_64__) || defined(__i386__))
#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
#if defined(__native_client__) && defined(__x86_64__)
#define MEMACCESS(base) "%%nacl:(%%r15,%q" #base ")"
#define MEMLEA(offset, base) #offset "(%q" #base ")"
#else
#define MEMACCESS(base) "(%" #base ")"
#define MEMLEA(offset, base) #offset "(%" #base ")"
#endif
uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count) {
uint32 sse;
asm volatile (
asm volatile ( // NOLINT
"pxor %%xmm0,%%xmm0 \n"
"pxor %%xmm5,%%xmm5 \n"
"sub %0,%1 \n"
".p2align 4 \n"
"1: \n"
"movdqa (%0),%%xmm1 \n"
"movdqa (%0,%1,1),%%xmm2 \n"
"lea 0x10(%0),%0 \n"
"movdqa "MEMACCESS(0)",%%xmm1 \n"
"lea "MEMLEA(0x10, 0)",%0 \n"
"movdqa "MEMACCESS(1)",%%xmm2 \n"
"lea "MEMLEA(0x10, 1)",%1 \n"
"sub $0x10,%2 \n"
"movdqa %%xmm1,%%xmm3 \n"
"psubusb %%xmm2,%%xmm1 \n"
......@@ -60,14 +67,13 @@ uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count) {
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
#endif
);
); // NOLINT
return sse;
}
#endif // defined(__x86_64__) || defined(__i386__)
#if !defined(LIBYUV_DISABLE_X86) && \
!(defined(__native_client__) && defined(__x86_64__)) && \
(defined(__x86_64__) || (defined(__i386__) && !defined(__pic__)))
#define HAS_HASHDJB2_SSE41
static uvec32 kHash16x33 = { 0x92d9e201, 0, 0, 0 }; // 33 ^ 16
......@@ -98,14 +104,14 @@ static uvec32 kHashMul3 = {
uint32 HashDjb2_SSE41(const uint8* src, int count, uint32 seed) {
uint32 hash;
asm volatile (
asm volatile ( // NOLINT
"movd %2,%%xmm0 \n"
"pxor %%xmm7,%%xmm7 \n"
"movdqa %4,%%xmm6 \n"
".p2align 4 \n"
"1: \n"
"movdqu (%0),%%xmm1 \n"
"lea 0x10(%0),%0 \n"
"movdqu "MEMACCESS(0)",%%xmm1 \n"
"lea "MEMLEA(0x10, 0)",%0 \n"
"pmulld %%xmm6,%%xmm0 \n"
"movdqa %5,%%xmm5 \n"
"movdqa %%xmm1,%%xmm2 \n"
......@@ -149,7 +155,7 @@ uint32 HashDjb2_SSE41(const uint8* src, int count, uint32 seed) {
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7"
#endif
);
); // NOLINT
return hash;
}
#endif // defined(__x86_64__) || (defined(__i386__) && !defined(__pic__)))
......@@ -158,3 +164,4 @@ uint32 HashDjb2_SSE41(const uint8* src, int count, uint32 seed) {
} // extern "C"
} // namespace libyuv
#endif
......@@ -26,13 +26,13 @@ uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count) {
mov ecx, [esp + 12] // count
pxor xmm0, xmm0
pxor xmm5, xmm5
sub edx, eax
align 16
wloop:
movdqa xmm1, [eax]
movdqa xmm2, [eax + edx]
lea eax, [eax + 16]
movdqa xmm2, [edx]
lea edx, [edx + 16]
sub ecx, 16
movdqa xmm3, xmm1 // abs trick
psubusb xmm1, xmm2
......
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