highperfocused/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge commit 'ab05d3934de8e932dbd77979a687e6598e67535c'

Browse Source 
* commit 'ab05d3934de8e932dbd77979a687e6598e67535c':
  arm: vc1dsp: Add commas between macro arguments

Merged-by: James Almer <jamrial@gmail.com>
This commit is contained in:
James Almer 2018-03-30 15:47:31 -03:00
commit a7109b82c4
1 changed files with 47 additions and 47 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

94
libavcodec/arm/vc1dsp_neon.S
View File

@ -410,13 +410,13 @@ function ff_vc1_inv_trans_8x8_neon, export=1
@ src[48] q14
@ src[56] q15
vc1_inv_trans_8x8_helper add=4 add1beforeshift=0 rshift=3
vc1_inv_trans_8x8_helper add=4, add1beforeshift=0, rshift=3
@ Transpose result matrix of 8x8
swap4 d17, d19, d21, d23, d24, d26, d28, d30
transpose16_4x4 q8, q9, q10, q11, q12, q13, q14, q15
vc1_inv_trans_8x8_helper add=64 add1beforeshift=1 rshift=7
vc1_inv_trans_8x8_helper add=64, add1beforeshift=1, rshift=7
vst1.64 {q8-q9}, [r0,:128]!
vst1.64 {q10-q11}, [r0,:128]!
@ -431,7 +431,7 @@ function ff_vc1_inv_trans_8x4_neon, export=1
vld1.64 {q0-q1}, [r2,:128]! @ load 8 * 4 * 2 = 64 bytes / 16 bytes per quad = 4 quad registers
vld1.64 {q2-q3}, [r2,:128]
transpose16 q0 q1 q2 q3 @ transpose rows to columns
transpose16 q0, q1, q2, q3 @ transpose rows to columns
@ At this point:
@ src[0] d0
@ -443,7 +443,7 @@ function ff_vc1_inv_trans_8x4_neon, export=1
@ src[6] d5
@ src[7] d7
vc1_inv_trans_8x4_helper add=4 add1beforeshift=0 rshift=3
vc1_inv_trans_8x4_helper add=4, add1beforeshift=0, rshift=3
@ Move output to more standardized registers
vmov d0, d16
@ -465,7 +465,7 @@ function ff_vc1_inv_trans_8x4_neon, export=1
@ dst[6] d5
@ dst[7] d7
transpose16 q0 q1 q2 q3 @ turn columns into rows
transpose16 q0, q1, q2, q3 @ turn columns into rows
@ At this point:
@ row[0] q0
@ -473,7 +473,7 @@ function ff_vc1_inv_trans_8x4_neon, export=1
@ row[2] q2
@ row[3] q3
vc1_inv_trans_4x8_helper add=64 rshift=7
vc1_inv_trans_4x8_helper add=64, rshift=7
@ At this point:
@ line[0].l d0
@ -523,7 +523,7 @@ function ff_vc1_inv_trans_4x8_neon, export=1
vld4.16 {d1[2], d3[2], d5[2], d7[2]}, [r2,:64], r12
vld4.16 {d1[3], d3[3], d5[3], d7[3]}, [r2,:64]
vc1_inv_trans_4x8_helper add=4 rshift=3
vc1_inv_trans_4x8_helper add=4, rshift=3
@ At this point:
@ dst[0] = q0
@ -531,9 +531,9 @@ function ff_vc1_inv_trans_4x8_neon, export=1
@ dst[2] = q2
@ dst[3] = q3
transpose16 q0 q1 q2 q3 @ Transpose rows (registers) into columns
transpose16 q0, q1, q2, q3 @ Transpose rows (registers) into columns
vc1_inv_trans_8x4_helper add=64 add1beforeshift=1 rshift=7
vc1_inv_trans_8x4_helper add=64, add1beforeshift=1, rshift=7
vld1.32 {d28[]}, [r0,:32], r1 @ read dest
vld1.32 {d28[1]}, [r0,:32], r1
@ -611,7 +611,7 @@ function ff_vc1_inv_trans_4x4_neon, export=1
@ src[2] = d1
@ src[3] = d3
vc1_inv_trans_4x4_helper add=4 rshift=3 @ compute t1, t2, t3, t4 and combine them into dst[0-3]
vc1_inv_trans_4x4_helper add=4, rshift=3 @ compute t1, t2, t3, t4 and combine them into dst[0-3]
@ At this point:
@ dst[0] = d0
@ -619,7 +619,7 @@ function ff_vc1_inv_trans_4x4_neon, export=1
@ dst[2] = d1
@ dst[3] = d2
transpose16 d0 d3 d1 d2 @ Transpose rows (registers) into columns
transpose16 d0, d3, d1, d2 @ Transpose rows (registers) into columns
@ At this point:
@ src[0] = d0
@ -635,7 +635,7 @@ function ff_vc1_inv_trans_4x4_neon, export=1
@ src[16] = d1
@ src[24] = d3
vc1_inv_trans_4x4_helper add=64 rshift=7 @ compute t1, t2, t3, t4 and combine them into dst[0-3]
vc1_inv_trans_4x4_helper add=64, rshift=7 @ compute t1, t2, t3, t4 and combine them into dst[0-3]
@ At this point:
@ line[0] = d0
@ -665,26 +665,26 @@ endfunc
@ The absolute value of multiplication constants from vc1_mspel_filter and vc1_mspel_{ver,hor}_filter_16bits.
@ The sign is embedded in the code below that carries out the multiplication (mspel_filter{,.16}).
#define MSPEL_MODE_1_MUL_CONSTANTS 4 53 18 3
#define MSPEL_MODE_2_MUL_CONSTANTS 1 9 9 1
#define MSPEL_MODE_3_MUL_CONSTANTS 3 18 53 4
#define MSPEL_MODE_1_MUL_CONSTANTS 4, 53, 18, 3
#define MSPEL_MODE_2_MUL_CONSTANTS 1, 9, 9, 1
#define MSPEL_MODE_3_MUL_CONSTANTS 3, 18, 53, 4
@ These constants are from reading the source code of vc1_mspel_mc and determining the value that
@ is added to `rnd` to result in the variable `r`, and the value of the variable `shift`.
#define MSPEL_MODES_11_ADDSHIFT_CONSTANTS 15 5
#define MSPEL_MODES_12_ADDSHIFT_CONSTANTS 3 3
#define MSPEL_MODES_13_ADDSHIFT_CONSTANTS 15 5
#define MSPEL_MODES_11_ADDSHIFT_CONSTANTS 15, 5
#define MSPEL_MODES_12_ADDSHIFT_CONSTANTS 3, 3
#define MSPEL_MODES_13_ADDSHIFT_CONSTANTS 15, 5
#define MSPEL_MODES_21_ADDSHIFT_CONSTANTS MSPEL_MODES_12_ADDSHIFT_CONSTANTS
#define MSPEL_MODES_22_ADDSHIFT_CONSTANTS 0 1
#define MSPEL_MODES_23_ADDSHIFT_CONSTANTS 3 3
#define MSPEL_MODES_22_ADDSHIFT_CONSTANTS 0, 1
#define MSPEL_MODES_23_ADDSHIFT_CONSTANTS 3, 3
#define MSPEL_MODES_31_ADDSHIFT_CONSTANTS MSPEL_MODES_13_ADDSHIFT_CONSTANTS
#define MSPEL_MODES_32_ADDSHIFT_CONSTANTS MSPEL_MODES_23_ADDSHIFT_CONSTANTS
#define MSPEL_MODES_33_ADDSHIFT_CONSTANTS 15 5
#define MSPEL_MODES_33_ADDSHIFT_CONSTANTS 15, 5
@ The addition and shift constants from vc1_mspel_filter.
#define MSPEL_MODE_1_ADDSHIFT_CONSTANTS 32 6
#define MSPEL_MODE_2_ADDSHIFT_CONSTANTS 8 4
#define MSPEL_MODE_3_ADDSHIFT_CONSTANTS 32 6
#define MSPEL_MODE_1_ADDSHIFT_CONSTANTS 32, 6
#define MSPEL_MODE_2_ADDSHIFT_CONSTANTS 8, 4
#define MSPEL_MODE_3_ADDSHIFT_CONSTANTS 32, 6
@ Setup constants in registers for a subsequent use of mspel_filter{,.16}.
.macro mspel_constants typesize reg_a reg_b reg_c reg_d filter_a filter_b filter_c filter_d reg_add filter_add_register
@ -818,7 +818,7 @@ T mov sp, r4
sub r1, r1, r2 @ r1 = &src[-stride] @ slide back
@ Do vertical filtering from src into tmp
mspel_constants i8 d28 d29 d30 d31 \filter_v_a \filter_v_b \filter_v_c \filter_v_d q13 r3
mspel_constants i8, d28, d29, d30, d31, \filter_v_a, \filter_v_b, \filter_v_c, \filter_v_d, q13, r3
vld1.64 {d0,d1}, [r1], r2
vld1.64 {d2,d3}, [r1], r2
@ -828,23 +828,23 @@ T mov sp, r4
subs r12, r12, #4
vld1.64 {d6,d7}, [r1], r2
mspel_filter q11 q11 d0 d2 d4 d6 \filter_v_a \filter_v_b \filter_v_c \filter_v_d d28 d29 d30 d31 q13 \filter_shift narrow=0
mspel_filter q12 q12 d1 d3 d5 d7 \filter_v_a \filter_v_b \filter_v_c \filter_v_d d28 d29 d30 d31 q13 \filter_shift narrow=0
mspel_filter q11, q11, d0, d2, d4, d6, \filter_v_a, \filter_v_b, \filter_v_c, \filter_v_d, d28, d29, d30, d31, q13, \filter_shift, narrow=0
mspel_filter q12, q12, d1, d3, d5, d7, \filter_v_a, \filter_v_b, \filter_v_c, \filter_v_d, d28, d29, d30, d31, q13, \filter_shift, narrow=0
vst1.64 {q11,q12}, [r4,:128]! @ store and increment
vld1.64 {d0,d1}, [r1], r2
mspel_filter q11 q11 d2 d4 d6 d0 \filter_v_a \filter_v_b \filter_v_c \filter_v_d d28 d29 d30 d31 q13 \filter_shift narrow=0
mspel_filter q12 q12 d3 d5 d7 d1 \filter_v_a \filter_v_b \filter_v_c \filter_v_d d28 d29 d30 d31 q13 \filter_shift narrow=0
mspel_filter q11, q11, d2, d4, d6, d0, \filter_v_a, \filter_v_b, \filter_v_c, \filter_v_d, d28, d29, d30, d31, q13, \filter_shift, narrow=0
mspel_filter q12, q12, d3, d5, d7, d1, \filter_v_a, \filter_v_b, \filter_v_c, \filter_v_d, d28, d29, d30, d31, q13, \filter_shift, narrow=0
vst1.64 {q11,q12}, [r4,:128]! @ store and increment
vld1.64 {d2,d3}, [r1], r2
mspel_filter q11 q11 d4 d6 d0 d2 \filter_v_a \filter_v_b \filter_v_c \filter_v_d d28 d29 d30 d31 q13 \filter_shift narrow=0
mspel_filter q12 q12 d5 d7 d1 d3 \filter_v_a \filter_v_b \filter_v_c \filter_v_d d28 d29 d30 d31 q13 \filter_shift narrow=0
mspel_filter q11, q11, d4, d6, d0, d2, \filter_v_a, \filter_v_b, \filter_v_c, \filter_v_d, d28, d29, d30, d31, q13, \filter_shift, narrow=0
mspel_filter q12, q12, d5, d7, d1, d3, \filter_v_a, \filter_v_b, \filter_v_c, \filter_v_d, d28, d29, d30, d31, q13, \filter_shift, narrow=0
vst1.64 {q11,q12}, [r4,:128]! @ store and increment
vld1.64 {d4,d5}, [r1], r2
mspel_filter q11 q11 d6 d0 d2 d4 \filter_v_a \filter_v_b \filter_v_c \filter_v_d d28 d29 d30 d31 q13 \filter_shift narrow=0
mspel_filter q12 q12 d7 d1 d3 d5 \filter_v_a \filter_v_b \filter_v_c \filter_v_d d28 d29 d30 d31 q13 \filter_shift narrow=0
mspel_filter q11, q11, d6, d0, d2, d4, \filter_v_a, \filter_v_b, \filter_v_c, \filter_v_d, d28, d29, d30, d31, q13, \filter_shift, narrow=0
mspel_filter q12, q12, d7, d1, d3, d5, \filter_v_a, \filter_v_b, \filter_v_c, \filter_v_d, d28, d29, d30, d31, q13, \filter_shift, narrow=0
vst1.64 {q11,q12}, [r4,:128]! @ store and increment
bne 1b
@ -854,7 +854,7 @@ T mov sp, r4
mov r4, sp @ r4 = tmp
@ Do horizontal filtering from temp to dst
mspel_constants i16 d28 d29 d30 d31 \filter_h_a \filter_h_b \filter_h_c \filter_h_d q13 r3
mspel_constants i16, d28, d29, d30, d31, \filter_h_a, \filter_h_b, \filter_h_c, \filter_h_d, q13, r3
2:
subs r12, r12, #1
@ -864,7 +864,7 @@ T mov sp, r4
vext.16 q3, q0, q1, #3
vext.16 q1, q0, q1, #1 @ do last because it writes to q1 which is read by the other vext instructions
mspel_filter.16 q11 q12 d22 d23 d21 d0 d1 d2 d3 d4 d5 d6 d7 \filter_h_a \filter_h_b \filter_h_c \filter_h_d d28 d29 d30 d31 q13 7
mspel_filter.16 q11, q12, d22, d23, d21, d0, d1, d2, d3, d4, d5, d6, d7, \filter_h_a, \filter_h_b, \filter_h_c, \filter_h_d, d28, d29, d30, d31, q13, 7
vst1.64 {d21}, [r0,:64], r2 @ store and increment dst
@ -877,9 +877,9 @@ endfunc
@ Use C preprocessor and assembler macros to expand to functions for horizontal and vertical filtering.
#define PUT_VC1_MSPEL_MC_HV(hmode, vmode) \
put_vc1_mspel_mc_hv hmode vmode \
MSPEL_MODE_ ## hmode ## _MUL_CONSTANTS \
MSPEL_MODE_ ## vmode ## _MUL_CONSTANTS \
put_vc1_mspel_mc_hv hmode, vmode, \
MSPEL_MODE_ ## hmode ## _MUL_CONSTANTS, \
MSPEL_MODE_ ## vmode ## _MUL_CONSTANTS, \
MSPEL_MODES_ ## hmode ## vmode ## _ADDSHIFT_CONSTANTS
PUT_VC1_MSPEL_MC_HV(1, 1)
@ -900,7 +900,7 @@ function ff_put_vc1_mspel_mc\hmode\()0_neon, export=1
mov r12, #8 @ loop counter
sub r1, r1, #1 @ slide back, using immediate
mspel_constants i8 d28 d29 d30 d31 \filter_a \filter_b \filter_c \filter_d q13 r3
mspel_constants i8, d28, d29, d30, d31, \filter_a, \filter_b, \filter_c, \filter_d, q13, r3
1:
subs r12, r12, #1
@ -910,7 +910,7 @@ function ff_put_vc1_mspel_mc\hmode\()0_neon, export=1
vext.8 d3, d0, d1, #3
vext.8 d1, d0, d1, #1 @ do last because it writes to d1 which is read by the other vext instructions
mspel_filter q11 d21 d0 d1 d2 d3 \filter_a \filter_b \filter_c \filter_d d28 d29 d30 d31 q13 \filter_shift
mspel_filter q11, d21, d0, d1, d2, d3, \filter_a, \filter_b, \filter_c, \filter_d, d28, d29, d30, d31, q13, \filter_shift
vst1.64 {d21}, [r0,:64], r2 @ store and increment dst
@ -922,7 +922,7 @@ endfunc
@ Use C preprocessor and assembler macros to expand to functions for horizontal only filtering.
#define PUT_VC1_MSPEL_MC_H_ONLY(hmode) \
put_vc1_mspel_mc_h_only hmode MSPEL_MODE_ ## hmode ## _MUL_CONSTANTS MSPEL_MODE_ ## hmode ## _ADDSHIFT_CONSTANTS
put_vc1_mspel_mc_h_only hmode, MSPEL_MODE_ ## hmode ## _MUL_CONSTANTS, MSPEL_MODE_ ## hmode ## _ADDSHIFT_CONSTANTS
PUT_VC1_MSPEL_MC_H_ONLY(1)
PUT_VC1_MSPEL_MC_H_ONLY(2)
@ -937,7 +937,7 @@ function ff_put_vc1_mspel_mc0\vmode\()_neon, export=1
mov r12, #8 @ loop counter
sub r1, r1, r2 @ r1 = &src[-stride] @ slide back
mspel_constants i8 d28 d29 d30 d31 \filter_a \filter_b \filter_c \filter_d q13 r3
mspel_constants i8, d28, d29, d30, d31, \filter_a, \filter_b, \filter_c, \filter_d, q13, r3
vld1.64 {d0}, [r1], r2 @ d0 = src[-stride]
vld1.64 {d1}, [r1], r2 @ d1 = src[0]
@ -947,19 +947,19 @@ function ff_put_vc1_mspel_mc0\vmode\()_neon, export=1
subs r12, r12, #4
vld1.64 {d3}, [r1], r2 @ d3 = src[stride * 2]
mspel_filter q11 d21 d0 d1 d2 d3 \filter_a \filter_b \filter_c \filter_d d28 d29 d30 d31 q13 \filter_shift
mspel_filter q11, d21, d0, d1, d2, d3, \filter_a, \filter_b, \filter_c, \filter_d, d28, d29, d30, d31, q13, \filter_shift
vst1.64 {d21}, [r0,:64], r2 @ store and increment dst
vld1.64 {d0}, [r1], r2 @ d0 = next line
mspel_filter q11 d21 d1 d2 d3 d0 \filter_a \filter_b \filter_c \filter_d d28 d29 d30 d31 q13 \filter_shift
mspel_filter q11, d21, d1, d2, d3, d0, \filter_a, \filter_b, \filter_c, \filter_d, d28, d29, d30, d31, q13, \filter_shift
vst1.64 {d21}, [r0,:64], r2 @ store and increment dst
vld1.64 {d1}, [r1], r2 @ d1 = next line
mspel_filter q11 d21 d2 d3 d0 d1 \filter_a \filter_b \filter_c \filter_d d28 d29 d30 d31 q13 \filter_shift
mspel_filter q11, d21, d2, d3, d0, d1, \filter_a, \filter_b, \filter_c, \filter_d, d28, d29, d30, d31, q13, \filter_shift
vst1.64 {d21}, [r0,:64], r2 @ store and increment dst
vld1.64 {d2}, [r1], r2 @ d2 = next line
mspel_filter q11 d21 d3 d0 d1 d2 \filter_a \filter_b \filter_c \filter_d d28 d29 d30 d31 q13 \filter_shift
mspel_filter q11, d21, d3, d0, d1, d2, \filter_a, \filter_b, \filter_c, \filter_d, d28, d29, d30, d31, q13, \filter_shift
vst1.64 {d21}, [r0,:64], r2 @ store and increment dst
bne 1b
@ -970,7 +970,7 @@ endfunc
@ Use C preprocessor and assembler macros to expand to functions for vertical only filtering.
#define PUT_VC1_MSPEL_MC_V_ONLY(vmode) \
put_vc1_mspel_mc_v_only vmode MSPEL_MODE_ ## vmode ## _MUL_CONSTANTS MSPEL_MODE_ ## vmode ## _ADDSHIFT_CONSTANTS
put_vc1_mspel_mc_v_only vmode, MSPEL_MODE_ ## vmode ## _MUL_CONSTANTS, MSPEL_MODE_ ## vmode ## _ADDSHIFT_CONSTANTS
PUT_VC1_MSPEL_MC_V_ONLY(1)
PUT_VC1_MSPEL_MC_V_ONLY(2)