generic-library/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
77753f874a
ffmpeg/libavcodec/armv4l/jrevdct_arm.S
Måns Rullgård 77753f874a ARM: fix j_rev_dct_ARM 
This is a bugfix for ARMv4 assembly implementation of 'j_rev_dct'
function.

The problem was in the incorrect partially empty row detection. Even
if the first two coefficients in the row were nonzero, it handled this
just like the case with only the first nonzero coefficient.

Now this function produces exactly the same output as the stripped
down reference C version of 'j_rev_dct' (with the nested checks like
'if (d6) { if (d2) { ...' always evaluated as true, avoiding shortcut
branches).

Originally committed as revision 15805 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-11-12 20:23:36 +00:00

389 lines
14 KiB
ArmAsm
Raw Blame History

/*
C-like prototype :
void j_rev_dct_ARM(DCTBLOCK data)
With DCTBLOCK being a pointer to an array of 64 'signed shorts'
Copyright (c) 2001 Lionel Ulmer (lionel.ulmer@free.fr / bbrox@bbrox.org)
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
#include "asm.S"
#define FIX_0_298631336 2446
#define FIX_0_541196100 4433
#define FIX_0_765366865 6270
#define FIX_1_175875602 9633
#define FIX_1_501321110 12299
#define FIX_2_053119869 16819
#define FIX_3_072711026 25172
#define FIX_M_0_390180644 -3196
#define FIX_M_0_899976223 -7373
#define FIX_M_1_847759065 -15137
#define FIX_M_1_961570560 -16069
#define FIX_M_2_562915447 -20995
#define FIX_0xFFFF 0xFFFF
#define FIX_0_298631336_ID 0
#define FIX_0_541196100_ID 4
#define FIX_0_765366865_ID 8
#define FIX_1_175875602_ID 12
#define FIX_1_501321110_ID 16
#define FIX_2_053119869_ID 20
#define FIX_3_072711026_ID 24
#define FIX_M_0_390180644_ID 28
#define FIX_M_0_899976223_ID 32
#define FIX_M_1_847759065_ID 36
#define FIX_M_1_961570560_ID 40
#define FIX_M_2_562915447_ID 44
#define FIX_0xFFFF_ID 48
.text
.align
function j_rev_dct_ARM, export=1
stmdb sp!, { r4 - r12, lr } @ all callee saved regs
sub sp, sp, #4 @ reserve some space on the stack
str r0, [ sp ] @ save the DCT pointer to the stack
mov lr, r0 @ lr = pointer to the current row
mov r12, #8 @ r12 = row-counter
add r11, pc, #(const_array-.-8) @ r11 = base pointer to the constants array
row_loop:
ldrsh r0, [lr, # 0] @ r0 = 'd0'
ldrsh r2, [lr, # 2] @ r2 = 'd2'
@ Optimization for row that have all items except the first set to 0
@ (this works as the DCTELEMS are always 4-byte aligned)
ldr r5, [lr, # 0]
ldr r6, [lr, # 4]
ldr r3, [lr, # 8]
ldr r4, [lr, #12]
orr r3, r3, r4
orr r3, r3, r6
orrs r5, r3, r5
beq end_of_row_loop @ nothing to be done as ALL of them are '0'
orrs r3, r3, r2
beq empty_row
ldrsh r1, [lr, # 8] @ r1 = 'd1'
ldrsh r4, [lr, # 4] @ r4 = 'd4'
ldrsh r6, [lr, # 6] @ r6 = 'd6'
ldr r3, [r11, #FIX_0_541196100_ID]
add r7, r2, r6
ldr r5, [r11, #FIX_M_1_847759065_ID]
mul r7, r3, r7 @ r7 = z1
ldr r3, [r11, #FIX_0_765366865_ID]
mla r6, r5, r6, r7 @ r6 = tmp2
add r5, r0, r4 @ r5 = tmp0
mla r2, r3, r2, r7 @ r2 = tmp3
sub r3, r0, r4 @ r3 = tmp1
add r0, r2, r5, lsl #13 @ r0 = tmp10
rsb r2, r2, r5, lsl #13 @ r2 = tmp13
add r4, r6, r3, lsl #13 @ r4 = tmp11
rsb r3, r6, r3, lsl #13 @ r3 = tmp12
stmdb sp!, { r0, r2, r3, r4 } @ save on the stack tmp10, tmp13, tmp12, tmp11
ldrsh r3, [lr, #10] @ r3 = 'd3'
ldrsh r5, [lr, #12] @ r5 = 'd5'
ldrsh r7, [lr, #14] @ r7 = 'd7'
add r0, r3, r5 @ r0 = 'z2'
add r2, r1, r7 @ r2 = 'z1'
add r4, r3, r7 @ r4 = 'z3'
add r6, r1, r5 @ r6 = 'z4'
ldr r9, [r11, #FIX_1_175875602_ID]
add r8, r4, r6 @ r8 = z3 + z4
ldr r10, [r11, #FIX_M_0_899976223_ID]
mul r8, r9, r8 @ r8 = 'z5'
ldr r9, [r11, #FIX_M_2_562915447_ID]
mul r2, r10, r2 @ r2 = 'z1'
ldr r10, [r11, #FIX_M_1_961570560_ID]
mul r0, r9, r0 @ r0 = 'z2'
ldr r9, [r11, #FIX_M_0_390180644_ID]
mla r4, r10, r4, r8 @ r4 = 'z3'
ldr r10, [r11, #FIX_0_298631336_ID]
mla r6, r9, r6, r8 @ r6 = 'z4'
ldr r9, [r11, #FIX_2_053119869_ID]
mla r7, r10, r7, r2 @ r7 = tmp0 + z1
ldr r10, [r11, #FIX_3_072711026_ID]
mla r5, r9, r5, r0 @ r5 = tmp1 + z2
ldr r9, [r11, #FIX_1_501321110_ID]
mla r3, r10, r3, r0 @ r3 = tmp2 + z2
add r7, r7, r4 @ r7 = tmp0
mla r1, r9, r1, r2 @ r1 = tmp3 + z1
add r5, r5, r6 @ r5 = tmp1
add r3, r3, r4 @ r3 = tmp2
add r1, r1, r6 @ r1 = tmp3
ldmia sp!, { r0, r2, r4, r6 } @ r0 = tmp10 / r2 = tmp13 / r4 = tmp12 / r6 = tmp11
@ r1 = tmp3 / r3 = tmp2 / r5 = tmp1 / r7 = tmp0
@ Compute DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS)
add r8, r0, r1
add r8, r8, #(1<<10)
mov r8, r8, asr #11
strh r8, [lr, # 0]
@ Compute DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS)
sub r8, r0, r1
add r8, r8, #(1<<10)
mov r8, r8, asr #11
strh r8, [lr, #14]
@ Compute DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS)
add r8, r6, r3
add r8, r8, #(1<<10)
mov r8, r8, asr #11
strh r8, [lr, # 2]
@ Compute DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS)
sub r8, r6, r3
add r8, r8, #(1<<10)
mov r8, r8, asr #11
strh r8, [lr, #12]
@ Compute DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS)
add r8, r4, r5
add r8, r8, #(1<<10)
mov r8, r8, asr #11
strh r8, [lr, # 4]
@ Compute DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS)
sub r8, r4, r5
add r8, r8, #(1<<10)
mov r8, r8, asr #11
strh r8, [lr, #10]
@ Compute DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS)
add r8, r2, r7
add r8, r8, #(1<<10)
mov r8, r8, asr #11
strh r8, [lr, # 6]
@ Compute DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS)
sub r8, r2, r7
add r8, r8, #(1<<10)
mov r8, r8, asr #11
strh r8, [lr, # 8]
@ End of row loop
add lr, lr, #16
subs r12, r12, #1
bne row_loop
beq start_column_loop
empty_row:
ldr r1, [r11, #FIX_0xFFFF_ID]
mov r0, r0, lsl #2
and r0, r0, r1
add r0, r0, r0, lsl #16
str r0, [lr, # 0]
str r0, [lr, # 4]
str r0, [lr, # 8]
str r0, [lr, #12]
end_of_row_loop:
@ End of loop
add lr, lr, #16
subs r12, r12, #1
bne row_loop
start_column_loop:
@ Start of column loop
ldr lr, [ sp ]
mov r12, #8
column_loop:
ldrsh r0, [lr, #( 0*8)] @ r0 = 'd0'
ldrsh r2, [lr, #( 4*8)] @ r2 = 'd2'
ldrsh r4, [lr, #( 8*8)] @ r4 = 'd4'
ldrsh r6, [lr, #(12*8)] @ r6 = 'd6'
ldr r3, [r11, #FIX_0_541196100_ID]
add r1, r2, r6
ldr r5, [r11, #FIX_M_1_847759065_ID]
mul r1, r3, r1 @ r1 = z1
ldr r3, [r11, #FIX_0_765366865_ID]
mla r6, r5, r6, r1 @ r6 = tmp2
add r5, r0, r4 @ r5 = tmp0
mla r2, r3, r2, r1 @ r2 = tmp3
sub r3, r0, r4 @ r3 = tmp1
add r0, r2, r5, lsl #13 @ r0 = tmp10
rsb r2, r2, r5, lsl #13 @ r2 = tmp13
add r4, r6, r3, lsl #13 @ r4 = tmp11
rsb r6, r6, r3, lsl #13 @ r6 = tmp12
ldrsh r1, [lr, #( 2*8)] @ r1 = 'd1'
ldrsh r3, [lr, #( 6*8)] @ r3 = 'd3'
ldrsh r5, [lr, #(10*8)] @ r5 = 'd5'
ldrsh r7, [lr, #(14*8)] @ r7 = 'd7'
@ Check for empty odd column (happens about 20 to 25 % of the time according to my stats)
orr r9, r1, r3
orr r10, r5, r7
orrs r10, r9, r10
beq empty_odd_column
stmdb sp!, { r0, r2, r4, r6 } @ save on the stack tmp10, tmp13, tmp12, tmp11
add r0, r3, r5 @ r0 = 'z2'
add r2, r1, r7 @ r2 = 'z1'
add r4, r3, r7 @ r4 = 'z3'
add r6, r1, r5 @ r6 = 'z4'
ldr r9, [r11, #FIX_1_175875602_ID]
add r8, r4, r6
ldr r10, [r11, #FIX_M_0_899976223_ID]
mul r8, r9, r8 @ r8 = 'z5'
ldr r9, [r11, #FIX_M_2_562915447_ID]
mul r2, r10, r2 @ r2 = 'z1'
ldr r10, [r11, #FIX_M_1_961570560_ID]
mul r0, r9, r0 @ r0 = 'z2'
ldr r9, [r11, #FIX_M_0_390180644_ID]
mla r4, r10, r4, r8 @ r4 = 'z3'
ldr r10, [r11, #FIX_0_298631336_ID]
mla r6, r9, r6, r8 @ r6 = 'z4'
ldr r9, [r11, #FIX_2_053119869_ID]
mla r7, r10, r7, r2 @ r7 = tmp0 + z1
ldr r10, [r11, #FIX_3_072711026_ID]
mla r5, r9, r5, r0 @ r5 = tmp1 + z2
ldr r9, [r11, #FIX_1_501321110_ID]
mla r3, r10, r3, r0 @ r3 = tmp2 + z2
add r7, r7, r4 @ r7 = tmp0
mla r1, r9, r1, r2 @ r1 = tmp3 + z1
add r5, r5, r6 @ r5 = tmp1
add r3, r3, r4 @ r3 = tmp2
add r1, r1, r6 @ r1 = tmp3
ldmia sp!, { r0, r2, r4, r6 } @ r0 = tmp10 / r2 = tmp13 / r4 = tmp11 / r6 = tmp12
@ r1 = tmp3 / r3 = tmp2 / r5 = tmp1 / r7 = tmp0
@ Compute DESCALE(tmp10 + tmp3, CONST_BITS+PASS1_BITS+3)
add r8, r0, r1
add r8, r8, #(1<<17)
mov r8, r8, asr #18
strh r8, [lr, #( 0*8)]
@ Compute DESCALE(tmp10 - tmp3, CONST_BITS+PASS1_BITS+3)
sub r8, r0, r1
add r8, r8, #(1<<17)
mov r8, r8, asr #18
strh r8, [lr, #(14*8)]
@ Compute DESCALE(tmp11 + tmp2, CONST_BITS+PASS1_BITS+3)
add r8, r4, r3
add r8, r8, #(1<<17)
mov r8, r8, asr #18
strh r8, [lr, #( 2*8)]
@ Compute DESCALE(tmp11 - tmp2, CONST_BITS+PASS1_BITS+3)
sub r8, r4, r3
add r8, r8, #(1<<17)
mov r8, r8, asr #18
strh r8, [lr, #(12*8)]
@ Compute DESCALE(tmp12 + tmp1, CONST_BITS+PASS1_BITS+3)
add r8, r6, r5
add r8, r8, #(1<<17)
mov r8, r8, asr #18
strh r8, [lr, #( 4*8)]
@ Compute DESCALE(tmp12 - tmp1, CONST_BITS+PASS1_BITS+3)
sub r8, r6, r5
add r8, r8, #(1<<17)
mov r8, r8, asr #18
strh r8, [lr, #(10*8)]
@ Compute DESCALE(tmp13 + tmp0, CONST_BITS+PASS1_BITS+3)
add r8, r2, r7
add r8, r8, #(1<<17)
mov r8, r8, asr #18
strh r8, [lr, #( 6*8)]
@ Compute DESCALE(tmp13 - tmp0, CONST_BITS+PASS1_BITS+3)
sub r8, r2, r7
add r8, r8, #(1<<17)
mov r8, r8, asr #18
strh r8, [lr, #( 8*8)]
@ End of row loop
add lr, lr, #2
subs r12, r12, #1
bne column_loop
beq the_end
empty_odd_column:
@ Compute DESCALE(tmp10 + tmp3, CONST_BITS+PASS1_BITS+3)
@ Compute DESCALE(tmp10 - tmp3, CONST_BITS+PASS1_BITS+3)
add r0, r0, #(1<<17)
mov r0, r0, asr #18
strh r0, [lr, #( 0*8)]
strh r0, [lr, #(14*8)]
@ Compute DESCALE(tmp11 + tmp2, CONST_BITS+PASS1_BITS+3)
@ Compute DESCALE(tmp11 - tmp2, CONST_BITS+PASS1_BITS+3)
add r4, r4, #(1<<17)
mov r4, r4, asr #18
strh r4, [lr, #( 2*8)]
strh r4, [lr, #(12*8)]
@ Compute DESCALE(tmp12 + tmp1, CONST_BITS+PASS1_BITS+3)
@ Compute DESCALE(tmp12 - tmp1, CONST_BITS+PASS1_BITS+3)
add r6, r6, #(1<<17)
mov r6, r6, asr #18
strh r6, [lr, #( 4*8)]
strh r6, [lr, #(10*8)]
@ Compute DESCALE(tmp13 + tmp0, CONST_BITS+PASS1_BITS+3)
@ Compute DESCALE(tmp13 - tmp0, CONST_BITS+PASS1_BITS+3)
add r2, r2, #(1<<17)
mov r2, r2, asr #18
strh r2, [lr, #( 6*8)]
strh r2, [lr, #( 8*8)]
@ End of row loop
add lr, lr, #2
subs r12, r12, #1
bne column_loop
the_end:
@ The end....
add sp, sp, #4
ldmia sp!, { r4 - r12, pc } @ restore callee saved regs and return
const_array:
.align
.word FIX_0_298631336
.word FIX_0_541196100
.word FIX_0_765366865
.word FIX_1_175875602
.word FIX_1_501321110
.word FIX_2_053119869
.word FIX_3_072711026
.word FIX_M_0_390180644
.word FIX_M_0_899976223
.word FIX_M_1_847759065
.word FIX_M_1_961570560
.word FIX_M_2_562915447
.word FIX_0xFFFF
Reference in New Issue View Git Blame Copy Permalink