devs/foxbox
1
0
Fork 0
forked from len0rd/rockbox
Code Pull requests Activity

Submit interim version of FS#10565. Performance optimization of atrac3 decoder for ARM. Introduce ASM routines for multiplications and two synthesis loops, refactured parts of synthesis and windowing. Speeds up decoding by a factor of 2.4 on PP502x.

Browse source 
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@22548 a1c6a512-1295-4272-9138-f99709370657
This commit is contained in:
Andree Buschmann 2009-08-29 19:44:27 +00:00
parent eb3cb724e8
commit f4774bf5bf
5 changed files with 343 additions and 146 deletions
Download patch file Download diff file Expand all files Collapse all files

92
apps/codecs/libatrac/fixp_math.h
View file

@ -10,27 +10,77 @@
/* Fixed point math routines for use in atrac3.c */
static inline int32_t fixmul16(int32_t x, int32_t y)
{
int64_t temp;
temp = x;
temp *= y;
temp >>= 16;
return (int32_t)temp;
}
static inline int32_t fixmul31(int32_t x, int32_t y)
{
int64_t temp;
temp = x;
temp *= y;
temp >>= 31; //16+31-16 = 31 bits
return (int32_t)temp;
}
#if defined(CPU_ARM)
#define fixmul16(X,Y) \
({ \
int32_t low; \
int32_t high; \
asm volatile ( /* calculates: result = (X*Y)>>16 */ \
"smull %0,%1,%2,%3 \n\t" /* 64 = 32x32 multiply */ \
"mov %0, %0, lsr #16 \n\t" /* %0 = %0 >> 16 */ \
"orr %0, %0, %1, lsl #16 \n\t"/* result = %0 OR (%1 << 16) */ \
: "=&r"(low), "=&r" (high) \
: "r"(X),"r"(Y)); \
low; \
})
#define fixmul31(X,Y) \
({ \
int32_t low; \
int32_t high; \
asm volatile ( /* calculates: result = (X*Y)>>31 */ \
"smull %0,%1,%2,%3 \n\t" /* 64 = 32x32 multiply */ \
"mov %0, %0, lsr #31 \n\t" /* %0 = %0 >> 31 */ \
"orr %0, %0, %1, lsl #1 \n\t" /* result = %0 OR (%1 << 1) */ \
: "=&r"(low), "=&r" (high) \
: "r"(X),"r"(Y)); \
low; \
})
#define fixmul32(X,Y) \
({ \
int32_t low; \
int32_t high; \
asm volatile ( /* calculates: result = (X*Y)>>32 */ \
"smull %0,%1,%2,%3 \n\t" /* 64 = 32x32 multiply */ \
: "=&r"(low), "=&r" (high) \
: "r"(X),"r"(Y)); \
high; \
})
#else
static inline int32_t fixmul16(int32_t x, int32_t y)
{
int64_t temp;
temp = x;
temp *= y;
temp >>= 16;
return (int32_t)temp;
}
static inline int32_t fixmul31(int32_t x, int32_t y)
{
int64_t temp;
temp = x;
temp *= y;
temp >>= 31; //16+31-16 = 31 bits
return (int32_t)temp;
}
static inline int32_t fixmul32(int32_t x, int32_t y)
{
int64_t temp;
temp = x;
temp *= y;
temp >>= 32; //16+31-16 = 31 bits
return (int32_t)temp;
}
#endif
static inline int32_t fixdiv16(int32_t x, int32_t y)
{