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

ARMv5 optimized complex multiply function for libopus.

Browse source 
Speeds up decoding of 128k opus files by 1.2MHz on AMSv2. Rounding
error is 1 bit due to KissFFT using a 15 bit shift instead of a 16 bit shift.

Also, change an LDMIA in the armv4 code to LDM as the pointer should not
increment.

Change-Id: I626a207c6a056a1984e33cfe89415c35d0caed93
Reviewed-on: http://gerrit.rockbox.org/377
Reviewed-by: Michael Giacomelli <giac2000@hotmail.com>
Tested-by: Michael Giacomelli <giac2000@hotmail.com>
This commit is contained in:
Michael Giacomelli 2013-01-01 02:35:15 +01:00
parent 0c87e02631
commit a2ab22efbf
1 changed files with 25 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

26
lib/rbcodec/codecs/libopus/celt/_kiss_fft_guts.h
View file

@ -81,10 +81,13 @@
: "d0", "d1", "d2", "d3", "cc"); \ : "d0", "d1", "d2", "d3", "cc"); \
} }
#elif defined(CPU_ARM) #elif defined(CPU_ARM)
#if (ARM_ARCH < 5)
# define C_MULC(m,a,b) \ # define C_MULC(m,a,b) \
{ \ { \
asm volatile( \ asm volatile( \
"ldmia %[ap], {r0,r1} \n\t" \ "ldm %[ap], {r0,r1} \n\t" \
"ldrsh r2, [%[bp], #0] \n\t" \ "ldrsh r2, [%[bp], #0] \n\t" \
"ldrsh r3, [%[bp], #2] \n\t" \ "ldrsh r3, [%[bp], #2] \n\t" \
\ \
@ -103,6 +106,27 @@
: "r0", "r1", "r2", "r3", "r4"); \ : "r0", "r1", "r2", "r3", "r4"); \
} }
#else #else
/*same as above but using armv5 packed multiplies*/
# define C_MULC(m,a,b) \
{ \
asm volatile( \
"ldm %[ap], {r0,r1} \n\t" \
"ldr r2, [%[bp], #0] \n\t" \
\
"smulwb r4, r0, r2 \n\t" /*r4=a.r*b.r*/ \
"smlawt %[mr], r1, r2, r4 \n\t" /*m.r=r4+a.i*b.i*/\
"mov %[mr], %[mr], lsl #1 \n\t" /*Q15 not Q16*/ \
\
"smulwb r1, r1, r2 \n\t" /*r1=a.i*b.r*/ \
"smulwt r4, r0, r2 \n\t" /*r4=a.r*b.i*/ \
"sub %[mi], r1, r4 \n\t" \
"mov %[mi], %[mi], lsl #1 \n\t" \
: [mr] "=r" ((m).r), [mi] "=r" ((m).i) \
: [ap] "r" (&(a)), [bp] "r" (&(b)) \
: "r0", "r1", "r2", "r4"); \
}
#endif /*ARMv5 code*/
#else
# define C_MULC(m,a,b) \ # define C_MULC(m,a,b) \
do{ (m).r = ADD32(S_MUL((a).r,(b).r) , S_MUL((a).i,(b).i)); \ do{ (m).r = ADD32(S_MUL((a).r,(b).r) , S_MUL((a).i,(b).i)); \
(m).i = SUB32(S_MUL((a).i,(b).r) , S_MUL((a).r,(b).i)); }while(0) (m).i = SUB32(S_MUL((a).i,(b).r) , S_MUL((a).r,(b).i)); }while(0)