|
 |
f76d5d |
Patch downloaded from
|
|
|
f76d5d |
http://bugs.gentoo.org/show_bug.cgi?id=121871
|
|
|
f76d5d |
http://bugs.gentoo.org/attachment.cgi?id=98094
|
|
|
f76d5d |
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/asm_common.S
|
|
|
f76d5d |
+++ libdv-0.104/libdv/asm_common.S
|
|
|
f76d5d |
@@ -0,0 +1,29 @@
|
|
|
f76d5d |
+/* public domain, do what you want */
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+# define MUNG(sym) sym##@GOTOFF(%ebp)
|
|
|
f76d5d |
+# define MUNG_ARR(sym, args...) sym##@GOTOFF(%ebp,##args)
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
+# define MUNG(sym) sym
|
|
|
f76d5d |
+# define MUNG_ARR(sym, args...) sym(,##args)
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+# undef __i686 /* gcc define gets in our way */
|
|
|
f76d5d |
+# define LOAD_PIC_REG(reg) \
|
|
|
f76d5d |
+ .ifndef __i686.get_pc_thunk.reg; \
|
|
|
f76d5d |
+ .section .gnu.linkonce.t.__i686.get_pc_thunk.reg,"ax",@progbits; \
|
|
|
f76d5d |
+ .global __i686.get_pc_thunk.reg; \
|
|
|
f76d5d |
+ .hidden __i686.get_pc_thunk.reg; \
|
|
|
f76d5d |
+ .type __i686.get_pc_thunk.reg,@function; \
|
|
|
f76d5d |
+ __i686.get_pc_thunk.reg: \
|
|
|
f76d5d |
+ movl (%esp), %e##reg; \
|
|
|
f76d5d |
+ ret; \
|
|
|
f76d5d |
+ .size __i686.get_pc_thunk.reg,.-__i686.get_pc_thunk.reg; \
|
|
|
f76d5d |
+ .previous; \
|
|
|
f76d5d |
+ .endif; \
|
|
|
f76d5d |
+ call __i686.get_pc_thunk.reg; \
|
|
|
f76d5d |
+ addl $_GLOBAL_OFFSET_TABLE_, %e##reg
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
+# define LOAD_PIC_REG(reg)
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/dct_block_mmx.S
|
|
|
f76d5d |
+++ libdv-0.104/libdv/dct_block_mmx.S
|
|
|
f76d5d |
@@ -55,19 +55,22 @@ scratch2: .quad 0
|
|
|
f76d5d |
|
|
|
f76d5d |
.section .note.GNU-stack, "", @progbits
|
|
|
f76d5d |
|
|
|
f76d5d |
+#include "asm_common.S"
|
|
|
f76d5d |
+
|
|
|
f76d5d |
.text
|
|
|
f76d5d |
|
|
|
f76d5d |
.align 8
|
|
|
f76d5d |
.global _dv_dct_88_block_mmx
|
|
|
f76d5d |
.hidden _dv_dct_88_block_mmx
|
|
|
f76d5d |
.type _dv_dct_88_block_mmx,@function
|
|
|
f76d5d |
_dv_dct_88_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp, %ebp
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
|
|
|
f76d5d |
- movl 8(%ebp), %esi # source
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl 12(%esp), %esi # source
|
|
|
f76d5d |
|
|
|
f76d5d |
# column 0
|
|
|
f76d5d |
movq 16*0(%esi), %mm0 # v0
|
|
|
f76d5d |
@@ -88,22 +93,22 @@ _dv_dct_88_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
movq 16*3(%esi), %mm5 # v3
|
|
|
f76d5d |
movq 16*4(%esi), %mm7 # v4
|
|
|
f76d5d |
- movq %mm7, scratch1 # scratch1: v4 ;
|
|
|
f76d5d |
+ movq %mm7, MUNG(scratch1) # scratch1: v4 ;
|
|
|
f76d5d |
movq %mm5, %mm7 # duplicate v3
|
|
|
f76d5d |
- paddw scratch1, %mm5 # v03: v3+v4
|
|
|
f76d5d |
- psubw scratch1, %mm7 # v04: v3-v4
|
|
|
f76d5d |
- movq %mm5, scratch2 # scratch2: v03
|
|
|
f76d5d |
+ paddw MUNG(scratch1), %mm5 # v03: v3+v4
|
|
|
f76d5d |
+ psubw MUNG(scratch1), %mm7 # v04: v3-v4
|
|
|
f76d5d |
+ movq %mm5, MUNG(scratch2) # scratch2: v03
|
|
|
f76d5d |
movq %mm0, %mm5 # mm5: v00
|
|
|
f76d5d |
|
|
|
f76d5d |
- paddw scratch2, %mm0 # v10: v00+v03
|
|
|
f76d5d |
- psubw scratch2, %mm5 # v13: v00-v03
|
|
|
f76d5d |
- movq %mm3, scratch3 # scratch3: v02
|
|
|
f76d5d |
+ paddw MUNG(scratch2), %mm0 # v10: v00+v03
|
|
|
f76d5d |
+ psubw MUNG(scratch2), %mm5 # v13: v00-v03
|
|
|
f76d5d |
+ movq %mm3, MUNG(scratch3) # scratch3: v02
|
|
|
f76d5d |
movq %mm1, %mm3 # duplicate v01
|
|
|
f76d5d |
|
|
|
f76d5d |
- paddw scratch3, %mm1 # v11: v01+v02
|
|
|
f76d5d |
- psubw scratch3, %mm3 # v12: v01-v02
|
|
|
f76d5d |
+ paddw MUNG(scratch3), %mm1 # v11: v01+v02
|
|
|
f76d5d |
+ psubw MUNG(scratch3), %mm3 # v12: v01-v02
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq %mm6, scratch4 # scratch4: v05
|
|
|
f76d5d |
+ movq %mm6, MUNG(scratch4) # scratch4: v05
|
|
|
f76d5d |
movq %mm0, %mm6 # duplicate v10
|
|
|
f76d5d |
|
|
|
f76d5d |
paddw %mm1, %mm0 # v10+v11
|
|
|
f76d5d |
@@ -113,10 +118,10 @@ _dv_dct_88_block_mmx:
|
|
|
f76d5d |
movq %mm6, 16*4(%esi) # out4: v10-v11
|
|
|
f76d5d |
|
|
|
f76d5d |
movq %mm4, %mm0 # mm0: v06
|
|
|
f76d5d |
- paddw scratch4, %mm4 # v15: v05+v06
|
|
|
f76d5d |
+ paddw MUNG(scratch4), %mm4 # v15: v05+v06
|
|
|
f76d5d |
paddw %mm2, %mm0 # v16: v07+v06
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmulhw WA3, %mm4 # v35~: WA3*v15
|
|
|
f76d5d |
+ pmulhw MUNG(WA3), %mm4 # v35~: WA3*v15
|
|
|
f76d5d |
psllw $1, %mm4 # v35: compensate the coeefient scale
|
|
|
f76d5d |
|
|
|
f76d5d |
movq %mm4, %mm6 # duplicate v35
|
|
|
f76d5d |
@@ -125,7 +130,7 @@ _dv_dct_88_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
paddw %mm5, %mm3 # v22: v12+v13
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmulhw WA1, %mm3 # v32~: WA1*v22
|
|
|
f76d5d |
+ pmulhw MUNG(WA1), %mm3 # v32~: WA1*v22
|
|
|
f76d5d |
psllw $16-NSHIFT, %mm3 # v32: compensate the coeefient scale
|
|
|
f76d5d |
movq %mm5, %mm6 # duplicate v13
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -136,13 +141,13 @@ _dv_dct_88_block_mmx:
|
|
|
f76d5d |
movq %mm6, 16*6(%esi) # out6: v13-v32
|
|
|
f76d5d |
|
|
|
f76d5d |
|
|
|
f76d5d |
- paddw scratch4, %mm7 # v14n: v04+v05
|
|
|
f76d5d |
+ paddw MUNG(scratch4), %mm7 # v14n: v04+v05
|
|
|
f76d5d |
movq %mm0, %mm5 # duplicate v16
|
|
|
f76d5d |
|
|
|
f76d5d |
psubw %mm7, %mm0 # va1: v16-v14n
|
|
|
f76d5d |
- pmulhw WA5, %mm0 # va0~: va1*WA5
|
|
|
f76d5d |
- pmulhw WA4, %mm5 # v36~~: v16*WA4
|
|
|
f76d5d |
- pmulhw WA2, %mm7 # v34~~: v14n*WA2
|
|
|
f76d5d |
+ pmulhw MUNG(WA5), %mm0 # va0~: va1*WA5
|
|
|
f76d5d |
+ pmulhw MUNG(WA4), %mm5 # v36~~: v16*WA4
|
|
|
f76d5d |
+ pmulhw MUNG(WA2), %mm7 # v34~~: v14n*WA2
|
|
|
f76d5d |
psllw $16-WA4_SHIFT, %mm5 # v36: compensate the coeefient scale
|
|
|
f76d5d |
psllw $16-NSHIFT, %mm7 # v34: compensate the coeefient scale
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -190,22 +195,22 @@ _dv_dct_88_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
movq 16*3(%esi), %mm5 # v3
|
|
|
f76d5d |
movq 16*4(%esi), %mm7 # v4
|
|
|
f76d5d |
- movq %mm7, scratch1 # scratch1: v4 ;
|
|
|
f76d5d |
+ movq %mm7, MUNG(scratch1) # scratch1: v4 ;
|
|
|
f76d5d |
movq %mm5, %mm7 # duplicate v3
|
|
|
f76d5d |
- paddw scratch1, %mm5 # v03: v3+v4
|
|
|
f76d5d |
- psubw scratch1, %mm7 # v04: v3-v4
|
|
|
f76d5d |
- movq %mm5, scratch2 # scratch2: v03
|
|
|
f76d5d |
+ paddw MUNG(scratch1), %mm5 # v03: v3+v4
|
|
|
f76d5d |
+ psubw MUNG(scratch1), %mm7 # v04: v3-v4
|
|
|
f76d5d |
+ movq %mm5, MUNG(scratch2) # scratch2: v03
|
|
|
f76d5d |
movq %mm0, %mm5 # mm5: v00
|
|
|
f76d5d |
|
|
|
f76d5d |
- paddw scratch2, %mm0 # v10: v00+v03
|
|
|
f76d5d |
- psubw scratch2, %mm5 # v13: v00-v03
|
|
|
f76d5d |
- movq %mm3, scratch3 # scratc3: v02
|
|
|
f76d5d |
+ paddw MUNG(scratch2), %mm0 # v10: v00+v03
|
|
|
f76d5d |
+ psubw MUNG(scratch2), %mm5 # v13: v00-v03
|
|
|
f76d5d |
+ movq %mm3, MUNG(scratch3) # scratc3: v02
|
|
|
f76d5d |
movq %mm1, %mm3 # duplicate v01
|
|
|
f76d5d |
|
|
|
f76d5d |
- paddw scratch3, %mm1 # v11: v01+v02
|
|
|
f76d5d |
- psubw scratch3, %mm3 # v12: v01-v02
|
|
|
f76d5d |
+ paddw MUNG(scratch3), %mm1 # v11: v01+v02
|
|
|
f76d5d |
+ psubw MUNG(scratch3), %mm3 # v12: v01-v02
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq %mm6, scratch4 # scratc4: v05
|
|
|
f76d5d |
+ movq %mm6, MUNG(scratch4) # scratc4: v05
|
|
|
f76d5d |
movq %mm0, %mm6 # duplicate v10
|
|
|
f76d5d |
|
|
|
f76d5d |
paddw %mm1, %mm0 # v10+v11
|
|
|
f76d5d |
@@ -215,10 +220,10 @@ _dv_dct_88_block_mmx:
|
|
|
f76d5d |
movq %mm6, 16*4(%esi) # out4: v10-v11
|
|
|
f76d5d |
|
|
|
f76d5d |
movq %mm4, %mm0 # mm0: v06
|
|
|
f76d5d |
- paddw scratch4, %mm4 # v15: v05+v06
|
|
|
f76d5d |
+ paddw MUNG(scratch4), %mm4 # v15: v05+v06
|
|
|
f76d5d |
paddw %mm2, %mm0 # v16: v07+v06
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmulhw WA3, %mm4 # v35~: WA3*v15
|
|
|
f76d5d |
+ pmulhw MUNG(WA3), %mm4 # v35~: WA3*v15
|
|
|
f76d5d |
psllw $16-NSHIFT, %mm4 # v35: compensate the coeefient scale
|
|
|
f76d5d |
|
|
|
f76d5d |
movq %mm4, %mm6 # duplicate v35
|
|
|
f76d5d |
@@ -227,7 +232,7 @@ _dv_dct_88_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
paddw %mm5, %mm3 # v22: v12+v13
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmulhw WA1, %mm3 # v32~: WA3*v15
|
|
|
f76d5d |
+ pmulhw MUNG(WA1), %mm3 # v32~: WA3*v15
|
|
|
f76d5d |
psllw $16-NSHIFT, %mm3 # v32: compensate the coeefient scale
|
|
|
f76d5d |
movq %mm5, %mm6 # duplicate v13
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -237,13 +242,13 @@ _dv_dct_88_block_mmx:
|
|
|
f76d5d |
movq %mm5, 16*2(%esi) # out2: v13+v32
|
|
|
f76d5d |
movq %mm6, 16*6(%esi) # out6: v13-v32
|
|
|
f76d5d |
|
|
|
f76d5d |
- paddw scratch4, %mm7 # v14n: v04+v05
|
|
|
f76d5d |
+ paddw MUNG(scratch4), %mm7 # v14n: v04+v05
|
|
|
f76d5d |
movq %mm0, %mm5 # duplicate v16
|
|
|
f76d5d |
|
|
|
f76d5d |
psubw %mm7, %mm0 # va1: v16-v14n
|
|
|
f76d5d |
- pmulhw WA2, %mm7 # v34~~: v14n*WA2
|
|
|
f76d5d |
- pmulhw WA5, %mm0 # va0~: va1*WA5
|
|
|
f76d5d |
- pmulhw WA4, %mm5 # v36~~: v16*WA4
|
|
|
f76d5d |
+ pmulhw MUNG(WA2), %mm7 # v34~~: v14n*WA2
|
|
|
f76d5d |
+ pmulhw MUNG(WA5), %mm0 # va0~: va1*WA5
|
|
|
f76d5d |
+ pmulhw MUNG(WA4), %mm5 # v36~~: v16*WA4
|
|
|
f76d5d |
psllw $16-NSHIFT, %mm7
|
|
|
f76d5d |
psllw $16-WA4_SHIFT, %mm5 # v36: compensate the coeffient
|
|
|
f76d5d |
# scale note that WA4 is shifted 1 bit less than the others
|
|
|
f76d5d |
@@ -755,11 +762,12 @@ _dv_dct_block_mmx_postscale_88:
|
|
|
f76d5d |
_dv_dct_248_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp, %ebp
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
|
|
|
f76d5d |
- movl 8(%ebp), %esi # source
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl 16(%esp), %esi # source
|
|
|
f76d5d |
|
|
|
f76d5d |
# column 0
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -781,7 +791,7 @@ _dv_dct_248_block_mmx:
|
|
|
f76d5d |
paddw %mm1, %mm0 # v20: v10+v11
|
|
|
f76d5d |
psubw %mm1, %mm3 # v21: v10-v11
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmulhw WA1, %mm5 # v32~: WA1*v22
|
|
|
f76d5d |
+ pmulhw MUNG(WA1), %mm5 # v32~: WA1*v22
|
|
|
f76d5d |
movq %mm4, %mm2
|
|
|
f76d5d |
psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -820,7 +830,7 @@ _dv_dct_248_block_mmx:
|
|
|
f76d5d |
paddw %mm1, %mm0 # v20: v10+v11
|
|
|
f76d5d |
psubw %mm1, %mm3 # v21: v10-v11
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmulhw WA1, %mm5 # v32~: WA1*v22
|
|
|
f76d5d |
+ pmulhw MUNG(WA1), %mm5 # v32~: WA1*v22
|
|
|
f76d5d |
movq %mm4, %mm2
|
|
|
f76d5d |
psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -857,7 +867,7 @@ _dv_dct_248_block_mmx:
|
|
|
f76d5d |
paddw %mm1, %mm0 # v20: v10+v11
|
|
|
f76d5d |
psubw %mm1, %mm3 # v21: v10-v11
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmulhw WA1, %mm5 # v32~: WA1*v22
|
|
|
f76d5d |
+ pmulhw MUNG(WA1), %mm5 # v32~: WA1*v22
|
|
|
f76d5d |
movq %mm4, %mm2
|
|
|
f76d5d |
psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -894,7 +904,7 @@ _dv_dct_248_block_mmx:
|
|
|
f76d5d |
paddw %mm1, %mm0 # v20: v10+v11
|
|
|
f76d5d |
psubw %mm1, %mm3 # v21: v10-v11
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmulhw WA1, %mm5 # v32~: WA1*v22
|
|
|
f76d5d |
+ pmulhw MUNG(WA1), %mm5 # v32~: WA1*v22
|
|
|
f76d5d |
movq %mm4, %mm2
|
|
|
f76d5d |
psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
|
|
|
f76d5d |
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/dv.c
|
|
|
f76d5d |
+++ libdv-0.104/libdv/dv.c
|
|
|
f76d5d |
@@ -205,6 +205,9 @@ dv_reconfigure(int clamp_luma, int clamp
|
|
|
f76d5d |
} /* dv_reconfigure */
|
|
|
f76d5d |
|
|
|
f76d5d |
|
|
|
f76d5d |
+extern uint8_t dv_quant_offset[4];
|
|
|
f76d5d |
+extern uint8_t dv_quant_shifts[22][4];
|
|
|
f76d5d |
+
|
|
|
f76d5d |
static inline void
|
|
|
f76d5d |
dv_decode_macroblock(dv_decoder_t *dv, dv_macroblock_t *mb, unsigned int quality) {
|
|
|
f76d5d |
int i;
|
|
|
f76d5d |
@@ -218,7 +221,7 @@ dv_decode_macroblock(dv_decoder_t *dv, d
|
|
|
f76d5d |
dv_idct_248 (co248, mb->b[i].coeffs);
|
|
|
f76d5d |
} else {
|
|
|
f76d5d |
#if ARCH_X86
|
|
|
f76d5d |
- _dv_quant_88_inverse_x86(mb->b[i].coeffs,mb->qno,mb->b[i].class_no);
|
|
|
f76d5d |
+ _dv_quant_88_inverse_x86(mb->b[i].coeffs,mb->qno,mb->b[i].class_no,dv_quant_offset,dv_quant_shifts);
|
|
|
f76d5d |
_dv_idct_88(mb->b[i].coeffs);
|
|
|
f76d5d |
#elif ARCH_X86_64
|
|
|
f76d5d |
_dv_quant_88_inverse_x86_64(mb->b[i].coeffs,mb->qno,mb->b[i].class_no);
|
|
|
f76d5d |
@@ -250,7 +253,7 @@ dv_decode_video_segment(dv_decoder_t *dv
|
|
|
f76d5d |
dv_idct_248 (co248, mb->b[b].coeffs);
|
|
|
f76d5d |
} else {
|
|
|
f76d5d |
#if ARCH_X86
|
|
|
f76d5d |
- _dv_quant_88_inverse_x86(bl->coeffs,mb->qno,bl->class_no);
|
|
|
f76d5d |
+ _dv_quant_88_inverse_x86(bl->coeffs,mb->qno,bl->class_no,dv_quant_offset,dv_quant_shifts);
|
|
|
f76d5d |
_dv_weight_88_inverse(bl->coeffs);
|
|
|
f76d5d |
_dv_idct_88(bl->coeffs);
|
|
|
f76d5d |
#elif ARCH_X86_64
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/encode.c
|
|
|
f76d5d |
+++ libdv-0.104/libdv/encode.c
|
|
|
f76d5d |
@@ -521,7 +521,8 @@ static void reorder_block(dv_block_t *bl
|
|
|
f76d5d |
}
|
|
|
f76d5d |
|
|
|
f76d5d |
extern unsigned long _dv_vlc_encode_block_mmx(dv_coeff_t* coeffs,
|
|
|
f76d5d |
- dv_vlc_entry_t ** out);
|
|
|
f76d5d |
+ dv_vlc_entry_t ** out,
|
|
|
f76d5d |
+ dv_vlc_entry_t * lookup);
|
|
|
f76d5d |
|
|
|
f76d5d |
extern unsigned long _dv_vlc_encode_block_mmx_x86_64(dv_coeff_t* coeffs,
|
|
|
f76d5d |
dv_vlc_entry_t ** out);
|
|
|
f76d5d |
@@ -558,7 +559,7 @@ static unsigned long vlc_encode_block(dv
|
|
|
f76d5d |
#elif ARCH_X86
|
|
|
f76d5d |
int num_bits;
|
|
|
f76d5d |
|
|
|
f76d5d |
- num_bits = _dv_vlc_encode_block_mmx(coeffs, &o);
|
|
|
f76d5d |
+ num_bits = _dv_vlc_encode_block_mmx(coeffs, &o, vlc_encode_lookup);
|
|
|
f76d5d |
emms();
|
|
|
f76d5d |
#else
|
|
|
f76d5d |
int num_bits;
|
|
|
f76d5d |
@@ -574,7 +575,7 @@ static unsigned long vlc_encode_block(dv
|
|
|
f76d5d |
return num_bits;
|
|
|
f76d5d |
}
|
|
|
f76d5d |
|
|
|
f76d5d |
-extern unsigned long _dv_vlc_num_bits_block_x86(dv_coeff_t* coeffs);
|
|
|
f76d5d |
+extern unsigned long _dv_vlc_num_bits_block_x86(dv_coeff_t* coeffs, unsigned char* lookup);
|
|
|
f76d5d |
extern unsigned long _dv_vlc_num_bits_block_x86_64(dv_coeff_t* coeffs);
|
|
|
f76d5d |
|
|
|
f76d5d |
extern unsigned long _dv_vlc_num_bits_block(dv_coeff_t* coeffs)
|
|
|
f76d5d |
@@ -600,7 +601,7 @@ extern unsigned long _dv_vlc_num_bits_bl
|
|
|
f76d5d |
#elif ARCH_X86_64
|
|
|
f76d5d |
return _dv_vlc_num_bits_block_x86_64(coeffs);
|
|
|
f76d5d |
#else
|
|
|
f76d5d |
- return _dv_vlc_num_bits_block_x86(coeffs);
|
|
|
f76d5d |
+ return _dv_vlc_num_bits_block_x86(coeffs, vlc_num_bits_lookup);
|
|
|
f76d5d |
#endif
|
|
|
f76d5d |
}
|
|
|
f76d5d |
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/encode_x86.S
|
|
|
f76d5d |
+++ libdv-0.104/libdv/encode_x86.S
|
|
|
f76d5d |
@@ -23,9 +23,6 @@
|
|
|
f76d5d |
* The libdv homepage is http://libdv.sourceforge.net/.
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
|
|
|
f76d5d |
-.data
|
|
|
f76d5d |
-ALLONE: .word 1,1,1,1
|
|
|
f76d5d |
-VLCADDMASK: .byte 255,0,0,0,255,0,0,0
|
|
|
f76d5d |
|
|
|
f76d5d |
|
|
|
f76d5d |
.section .note.GNU-stack, "", @progbits
|
|
|
f76d5d |
@@ -49,11 +47,14 @@ _dv_vlc_encode_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
movl $63, %ecx
|
|
|
f76d5d |
|
|
|
f76d5d |
- movl vlc_encode_lookup, %esi
|
|
|
f76d5d |
+ movl 4+4*4+8(%esp), %esi # vlc_encode_lookup
|
|
|
f76d5d |
|
|
|
f76d5d |
pxor %mm0, %mm0
|
|
|
f76d5d |
pxor %mm2, %mm2
|
|
|
f76d5d |
- movq VLCADDMASK, %mm1
|
|
|
f76d5d |
+ pushl $0x000000FF # these four lines
|
|
|
f76d5d |
+ pushl $0x000000FF # load VLCADDMASK
|
|
|
f76d5d |
+ movq (%esp), %mm1 # into %mm1 off the stack
|
|
|
f76d5d |
+ addl $8, %esp # --> no TEXTRELs
|
|
|
f76d5d |
xorl %ebp, %ebp
|
|
|
f76d5d |
subl $8, %edx
|
|
|
f76d5d |
vlc_encode_block_mmx_loop:
|
|
|
f76d5d |
@@ -125,7 +128,7 @@ _dv_vlc_num_bits_block_x86:
|
|
|
f76d5d |
addl $2, %edi
|
|
|
f76d5d |
|
|
|
f76d5d |
movl $63, %ecx
|
|
|
f76d5d |
- movl vlc_num_bits_lookup, %esi
|
|
|
f76d5d |
+ movl 4+4*4+4(%esp), %esi # vlc_num_bits_lookup
|
|
|
f76d5d |
|
|
|
f76d5d |
vlc_num_bits_block_x86_loop:
|
|
|
f76d5d |
movw (%edi), %ax
|
|
|
f76d5d |
@@ -583,8 +594,11 @@ _dv_need_dct_248_mmx_rows:
|
|
|
f76d5d |
paddw %mm5, %mm1
|
|
|
f76d5d |
|
|
|
f76d5d |
paddw %mm1, %mm0
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- pmaddwd ALLONE, %mm0
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ pushl $0x00010001 # these four lines
|
|
|
f76d5d |
+ pushl $0x00010001 # load ALLONE
|
|
|
f76d5d |
+ pmaddwd (%esp), %mm0 # into %mm0 off the stack
|
|
|
f76d5d |
+ addl $8, %esp # --> no TEXTRELs
|
|
|
f76d5d |
movq %mm0, %mm1
|
|
|
f76d5d |
psrlq $32, %mm1
|
|
|
f76d5d |
paddd %mm1, %mm0
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/idct_block_mmx.S
|
|
|
f76d5d |
+++ libdv-0.104/libdv/idct_block_mmx.S
|
|
|
f76d5d |
@@ -8,17 +8,21 @@
|
|
|
f76d5d |
|
|
|
f76d5d |
.section .note.GNU-stack, "", @progbits
|
|
|
f76d5d |
|
|
|
f76d5d |
+#include "asm_common.S"
|
|
|
f76d5d |
+
|
|
|
f76d5d |
.text
|
|
|
f76d5d |
.align 4
|
|
|
f76d5d |
.global _dv_idct_block_mmx
|
|
|
f76d5d |
.hidden _dv_idct_block_mmx
|
|
|
f76d5d |
.type _dv_idct_block_mmx,@function
|
|
|
f76d5d |
_dv_idct_block_mmx:
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp,%ebp
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
- leal preSC, %ecx
|
|
|
f76d5d |
- movl 8(%ebp),%esi /* source matrix */
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ leal MUNG(preSC), %ecx
|
|
|
f76d5d |
+ movl 12(%esp),%esi /* source matrix */
|
|
|
f76d5d |
|
|
|
f76d5d |
/*
|
|
|
f76d5d |
* column 0: even part
|
|
|
f76d5d |
@@ -34,7 +40,7 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
movq %mm1, %mm2 /* added 11/1/96 */
|
|
|
f76d5d |
pmulhw 8*8(%esi),%mm5 /* V8 */
|
|
|
f76d5d |
psubsw %mm0, %mm1 /* V16 */
|
|
|
f76d5d |
- pmulhw x5a825a825a825a82, %mm1 /* 23170 ->V18 */
|
|
|
f76d5d |
+ pmulhw MUNG(x5a825a825a825a82), %mm1 /* 23170 ->V18 */
|
|
|
f76d5d |
paddsw %mm0, %mm2 /* V17 */
|
|
|
f76d5d |
movq %mm2, %mm0 /* duplicate V17 */
|
|
|
f76d5d |
psraw $1, %mm2 /* t75=t82 */
|
|
|
f76d5d |
@@ -75,7 +81,7 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
paddsw %mm0, %mm3 /* V29 ; free mm0 */
|
|
|
f76d5d |
movq %mm7, %mm1 /* duplicate V26 */
|
|
|
f76d5d |
psraw $1, %mm3 /* t91=t94 */
|
|
|
f76d5d |
- pmulhw x539f539f539f539f,%mm7 /* V33 */
|
|
|
f76d5d |
+ pmulhw MUNG(x539f539f539f539f),%mm7 /* V33 */
|
|
|
f76d5d |
psraw $1, %mm1 /* t96 */
|
|
|
f76d5d |
movq %mm5, %mm0 /* duplicate V2 */
|
|
|
f76d5d |
psraw $2, %mm4 /* t85=t87 */
|
|
|
f76d5d |
@@ -83,15 +89,15 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
psubsw %mm4, %mm0 /* V28 ; free mm4 */
|
|
|
f76d5d |
movq %mm0, %mm2 /* duplicate V28 */
|
|
|
f76d5d |
psraw $1, %mm5 /* t90=t93 */
|
|
|
f76d5d |
- pmulhw x4546454645464546,%mm0 /* V35 */
|
|
|
f76d5d |
+ pmulhw MUNG(x4546454645464546),%mm0 /* V35 */
|
|
|
f76d5d |
psraw $1, %mm2 /* t97 */
|
|
|
f76d5d |
movq %mm5, %mm4 /* duplicate t90=t93 */
|
|
|
f76d5d |
psubsw %mm2, %mm1 /* V32 ; free mm2 */
|
|
|
f76d5d |
- pmulhw x61f861f861f861f8,%mm1 /* V36 */
|
|
|
f76d5d |
+ pmulhw MUNG(x61f861f861f861f8),%mm1 /* V36 */
|
|
|
f76d5d |
psllw $1, %mm7 /* t107 */
|
|
|
f76d5d |
paddsw %mm3, %mm5 /* V31 */
|
|
|
f76d5d |
psubsw %mm3, %mm4 /* V30 ; free mm3 */
|
|
|
f76d5d |
- pmulhw x5a825a825a825a82,%mm4 /* V34 */
|
|
|
f76d5d |
+ pmulhw MUNG(x5a825a825a825a82),%mm4 /* V34 */
|
|
|
f76d5d |
nop
|
|
|
f76d5d |
psubsw %mm1, %mm0 /* V38 */
|
|
|
f76d5d |
psubsw %mm7, %mm1 /* V37 ; free mm7 */
|
|
|
f76d5d |
@@ -158,7 +164,7 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
psubsw %mm7, %mm1 /* V50 */
|
|
|
f76d5d |
pmulhw 8*9(%esi), %mm5 /* V9 */
|
|
|
f76d5d |
paddsw %mm7, %mm2 /* V51 */
|
|
|
f76d5d |
- pmulhw x5a825a825a825a82, %mm1 /* 23170 ->V52 */
|
|
|
f76d5d |
+ pmulhw MUNG(x5a825a825a825a82), %mm1 /* 23170 ->V52 */
|
|
|
f76d5d |
movq %mm2, %mm6 /* duplicate V51 */
|
|
|
f76d5d |
psraw $1, %mm2 /* t138=t144 */
|
|
|
f76d5d |
movq %mm3, %mm4 /* duplicate V1 */
|
|
|
f76d5d |
@@ -199,11 +205,11 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
* even more by doing the correction step in a later stage when the number
|
|
|
f76d5d |
* is actually multiplied by 16
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
- paddw x0005000200010001, %mm4
|
|
|
f76d5d |
+ paddw MUNG(x0005000200010001), %mm4
|
|
|
f76d5d |
psubsw %mm6, %mm3 /* V60 ; free mm6 */
|
|
|
f76d5d |
psraw $1, %mm0 /* t154=t156 */
|
|
|
f76d5d |
movq %mm3, %mm1 /* duplicate V60 */
|
|
|
f76d5d |
- pmulhw x539f539f539f539f, %mm1 /* V67 */
|
|
|
f76d5d |
+ pmulhw MUNG(x539f539f539f539f), %mm1 /* V67 */
|
|
|
f76d5d |
movq %mm5, %mm6 /* duplicate V3 */
|
|
|
f76d5d |
psraw $2, %mm4 /* t148=t150 */
|
|
|
f76d5d |
paddsw %mm4, %mm5 /* V61 */
|
|
|
f76d5d |
@@ -212,13 +218,13 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
psllw $1, %mm1 /* t169 */
|
|
|
f76d5d |
paddsw %mm0, %mm5 /* V65 -> result */
|
|
|
f76d5d |
psubsw %mm0, %mm4 /* V64 ; free mm0 */
|
|
|
f76d5d |
- pmulhw x5a825a825a825a82, %mm4 /* V68 */
|
|
|
f76d5d |
+ pmulhw MUNG(x5a825a825a825a82), %mm4 /* V68 */
|
|
|
f76d5d |
psraw $1, %mm3 /* t158 */
|
|
|
f76d5d |
psubsw %mm6, %mm3 /* V66 */
|
|
|
f76d5d |
movq %mm5, %mm2 /* duplicate V65 */
|
|
|
f76d5d |
- pmulhw x61f861f861f861f8, %mm3 /* V70 */
|
|
|
f76d5d |
+ pmulhw MUNG(x61f861f861f861f8), %mm3 /* V70 */
|
|
|
f76d5d |
psllw $1, %mm6 /* t165 */
|
|
|
f76d5d |
- pmulhw x4546454645464546, %mm6 /* V69 */
|
|
|
f76d5d |
+ pmulhw MUNG(x4546454645464546), %mm6 /* V69 */
|
|
|
f76d5d |
psraw $1, %mm2 /* t172 */
|
|
|
f76d5d |
/* moved from next block */
|
|
|
f76d5d |
movq 8*5(%esi), %mm0 /* V56 */
|
|
|
f76d5d |
@@ -343,7 +349,7 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
* movq 8*13(%esi), %mm4 tmt13
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
psubsw %mm4, %mm3 /* V134 */
|
|
|
f76d5d |
- pmulhw x5a825a825a825a82, %mm3 /* 23170 ->V136 */
|
|
|
f76d5d |
+ pmulhw MUNG(x5a825a825a825a82), %mm3 /* 23170 ->V136 */
|
|
|
f76d5d |
movq 8*9(%esi), %mm6 /* tmt9 */
|
|
|
f76d5d |
paddsw %mm4, %mm5 /* V135 ; mm4 free */
|
|
|
f76d5d |
movq %mm0, %mm4 /* duplicate tmt1 */
|
|
|
f76d5d |
@@ -372,17 +378,17 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
psubsw %mm7, %mm0 /* V144 */
|
|
|
f76d5d |
movq %mm0, %mm3 /* duplicate V144 */
|
|
|
f76d5d |
paddsw %mm7, %mm2 /* V147 ; free mm7 */
|
|
|
f76d5d |
- pmulhw x539f539f539f539f, %mm0 /* 21407-> V151 */
|
|
|
f76d5d |
+ pmulhw MUNG(x539f539f539f539f), %mm0 /* 21407-> V151 */
|
|
|
f76d5d |
movq %mm1, %mm7 /* duplicate tmt3 */
|
|
|
f76d5d |
paddsw %mm5, %mm7 /* V145 */
|
|
|
f76d5d |
psubsw %mm5, %mm1 /* V146 ; free mm5 */
|
|
|
f76d5d |
psubsw %mm1, %mm3 /* V150 */
|
|
|
f76d5d |
movq %mm7, %mm5 /* duplicate V145 */
|
|
|
f76d5d |
- pmulhw x4546454645464546, %mm1 /* 17734-> V153 */
|
|
|
f76d5d |
+ pmulhw MUNG(x4546454645464546), %mm1 /* 17734-> V153 */
|
|
|
f76d5d |
psubsw %mm2, %mm5 /* V148 */
|
|
|
f76d5d |
- pmulhw x61f861f861f861f8, %mm3 /* 25080-> V154 */
|
|
|
f76d5d |
+ pmulhw MUNG(x61f861f861f861f8), %mm3 /* 25080-> V154 */
|
|
|
f76d5d |
psllw $2, %mm0 /* t311 */
|
|
|
f76d5d |
- pmulhw x5a825a825a825a82, %mm5 /* 23170-> V152 */
|
|
|
f76d5d |
+ pmulhw MUNG(x5a825a825a825a82), %mm5 /* 23170-> V152 */
|
|
|
f76d5d |
paddsw %mm2, %mm7 /* V149 ; free mm2 */
|
|
|
f76d5d |
psllw $1, %mm1 /* t313 */
|
|
|
f76d5d |
nop /* without the nop - freeze here for one clock */
|
|
|
f76d5d |
@@ -408,7 +414,7 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
paddsw %mm3, %mm6 /* V164 ; free mm3 */
|
|
|
f76d5d |
movq %mm4, %mm3 /* duplicate V142 */
|
|
|
f76d5d |
psubsw %mm5, %mm4 /* V165 ; free mm5 */
|
|
|
f76d5d |
- movq %mm2, scratch7 /* out7 */
|
|
|
f76d5d |
+ movq %mm2, MUNG(scratch7) /* out7 */
|
|
|
f76d5d |
psraw $4, %mm6
|
|
|
f76d5d |
psraw $4, %mm4
|
|
|
f76d5d |
paddsw %mm5, %mm3 /* V162 */
|
|
|
f76d5d |
@@ -419,11 +425,11 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
movq %mm6, 8*9(%esi) /* out9 */
|
|
|
f76d5d |
paddsw %mm1, %mm0 /* V161 */
|
|
|
f76d5d |
- movq %mm3, scratch5 /* out5 */
|
|
|
f76d5d |
+ movq %mm3, MUNG(scratch5) /* out5 */
|
|
|
f76d5d |
psubsw %mm1, %mm5 /* V166 ; free mm1 */
|
|
|
f76d5d |
movq %mm4, 8*11(%esi) /* out11 */
|
|
|
f76d5d |
psraw $4, %mm5
|
|
|
f76d5d |
- movq %mm0, scratch3 /* out3 */
|
|
|
f76d5d |
+ movq %mm0, MUNG(scratch3) /* out3 */
|
|
|
f76d5d |
movq %mm2, %mm4 /* duplicate V140 */
|
|
|
f76d5d |
movq %mm5, 8*13(%esi) /* out13 */
|
|
|
f76d5d |
paddsw %mm7, %mm2 /* V160 */
|
|
|
f76d5d |
@@ -433,7 +439,7 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
/* moved from the next block */
|
|
|
f76d5d |
movq 8*3(%esi), %mm7
|
|
|
f76d5d |
psraw $4, %mm4
|
|
|
f76d5d |
- movq %mm2, scratch1 /* out1 */
|
|
|
f76d5d |
+ movq %mm2, MUNG(scratch1) /* out1 */
|
|
|
f76d5d |
/* moved from the next block */
|
|
|
f76d5d |
movq %mm0, %mm1
|
|
|
f76d5d |
movq %mm4, 8*15(%esi) /* out15 */
|
|
|
f76d5d |
@@ -490,15 +496,15 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
paddsw %mm4, %mm3 /* V113 ; free mm4 */
|
|
|
f76d5d |
movq %mm0, %mm4 /* duplicate V110 */
|
|
|
f76d5d |
paddsw %mm1, %mm2 /* V111 */
|
|
|
f76d5d |
- pmulhw x539f539f539f539f, %mm0 /* 21407-> V117 */
|
|
|
f76d5d |
+ pmulhw MUNG(x539f539f539f539f), %mm0 /* 21407-> V117 */
|
|
|
f76d5d |
psubsw %mm1, %mm5 /* V112 ; free mm1 */
|
|
|
f76d5d |
psubsw %mm5, %mm4 /* V116 */
|
|
|
f76d5d |
movq %mm2, %mm1 /* duplicate V111 */
|
|
|
f76d5d |
- pmulhw x4546454645464546, %mm5 /* 17734-> V119 */
|
|
|
f76d5d |
+ pmulhw MUNG(x4546454645464546), %mm5 /* 17734-> V119 */
|
|
|
f76d5d |
psubsw %mm3, %mm2 /* V114 */
|
|
|
f76d5d |
- pmulhw x61f861f861f861f8, %mm4 /* 25080-> V120 */
|
|
|
f76d5d |
+ pmulhw MUNG(x61f861f861f861f8), %mm4 /* 25080-> V120 */
|
|
|
f76d5d |
paddsw %mm3, %mm1 /* V115 ; free mm3 */
|
|
|
f76d5d |
- pmulhw x5a825a825a825a82, %mm2 /* 23170-> V118 */
|
|
|
f76d5d |
+ pmulhw MUNG(x5a825a825a825a82), %mm2 /* 23170-> V118 */
|
|
|
f76d5d |
psllw $2, %mm0 /* t266 */
|
|
|
f76d5d |
movq %mm1, (%esi) /* save V115 */
|
|
|
f76d5d |
psllw $1, %mm5 /* t268 */
|
|
|
f76d5d |
@@ -516,7 +522,7 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
movq %mm6, %mm3 /* duplicate tmt4 */
|
|
|
f76d5d |
psubsw %mm0, %mm6 /* V100 */
|
|
|
f76d5d |
paddsw %mm0, %mm3 /* V101 ; free mm0 */
|
|
|
f76d5d |
- pmulhw x5a825a825a825a82, %mm6 /* 23170 ->V102 */
|
|
|
f76d5d |
+ pmulhw MUNG(x5a825a825a825a82), %mm6 /* 23170 ->V102 */
|
|
|
f76d5d |
movq %mm7, %mm5 /* duplicate tmt0 */
|
|
|
f76d5d |
movq 8*8(%esi), %mm1 /* tmt8 */
|
|
|
f76d5d |
paddsw %mm1, %mm7 /* V103 */
|
|
|
f76d5d |
@@ -550,10 +556,10 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
movq 8*2(%esi), %mm3 /* V123 */
|
|
|
f76d5d |
paddsw %mm4, %mm7 /* out0 */
|
|
|
f76d5d |
/* moved up from next block */
|
|
|
f76d5d |
- movq scratch3, %mm0
|
|
|
f76d5d |
+ movq MUNG(scratch3), %mm0
|
|
|
f76d5d |
psraw $4, %mm7
|
|
|
f76d5d |
/* moved up from next block */
|
|
|
f76d5d |
- movq scratch5, %mm6
|
|
|
f76d5d |
+ movq MUNG(scratch5), %mm6
|
|
|
f76d5d |
psubsw %mm4, %mm1 /* out14 ; free mm4 */
|
|
|
f76d5d |
paddsw %mm3, %mm5 /* out2 */
|
|
|
f76d5d |
psraw $4, %mm1
|
|
|
f76d5d |
@@ -564,7 +570,7 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
movq %mm5, 8*2(%esi) /* out2 ; free mm5 */
|
|
|
f76d5d |
psraw $4, %mm2
|
|
|
f76d5d |
/* moved up to the prev block */
|
|
|
f76d5d |
- movq scratch7, %mm4
|
|
|
f76d5d |
+ movq MUNG(scratch7), %mm4
|
|
|
f76d5d |
/* moved up to the prev block */
|
|
|
f76d5d |
psraw $4, %mm0
|
|
|
f76d5d |
movq %mm2, 8*12(%esi) /* out12 ; free mm2 */
|
|
|
f76d5d |
@@ -578,7 +584,7 @@ _dv_idct_block_mmx:
|
|
|
f76d5d |
* psraw $4, %mm0
|
|
|
f76d5d |
* psraw $4, %mm6
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
- movq scratch1, %mm1
|
|
|
f76d5d |
+ movq MUNG(scratch1), %mm1
|
|
|
f76d5d |
psraw $4, %mm4
|
|
|
f76d5d |
movq %mm0, 8*3(%esi) /* out3 */
|
|
|
f76d5d |
psraw $4, %mm1
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/parse.c
|
|
|
f76d5d |
+++ libdv-0.104/libdv/parse.c
|
|
|
f76d5d |
@@ -477,6 +477,13 @@ dv_parse_ac_coeffs(dv_videosegment_t *se
|
|
|
f76d5d |
exit(0);
|
|
|
f76d5d |
#endif
|
|
|
f76d5d |
} /* dv_parse_ac_coeffs */
|
|
|
f76d5d |
+#if defined __GNUC__ && __ELF__
|
|
|
f76d5d |
+# define dv_strong_hidden_alias(name, aliasname) \
|
|
|
f76d5d |
+ extern __typeof (name) aliasname __attribute__ ((alias (#name), visibility ("hidden")))
|
|
|
f76d5d |
+dv_strong_hidden_alias(dv_parse_ac_coeffs, asm_dv_parse_ac_coeffs);
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
+int asm_dv_parse_ac_coeffs(dv_videosegment_t *seg) { return dv_parse_ac_coeffs(seg); }
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
|
|
|
f76d5d |
/* ---------------------------------------------------------------------------
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/quant.c
|
|
|
f76d5d |
+++ libdv-0.104/libdv/quant.c
|
|
|
f76d5d |
@@ -144,7 +144,7 @@ uint8_t dv_quant_offset[4] = { 6,3,0,1
|
|
|
f76d5d |
uint32_t dv_quant_248_mul_tab [2] [22] [64];
|
|
|
f76d5d |
uint32_t dv_quant_88_mul_tab [2] [22] [64];
|
|
|
f76d5d |
|
|
|
f76d5d |
-extern void _dv_quant_x86(dv_coeff_t *block,int qno,int klass);
|
|
|
f76d5d |
+extern void _dv_quant_x86(dv_coeff_t *block,int qno,int klass,uint8_t *dv_quant_offset,uint8_t *dv_quant_shifts);
|
|
|
f76d5d |
extern void _dv_quant_x86_64(dv_coeff_t *block,int qno,int klass);
|
|
|
f76d5d |
static void quant_248_inverse_std(dv_coeff_t *block,int qno,int klass,dv_248_coeff_t *co);
|
|
|
f76d5d |
static void quant_248_inverse_mmx(dv_coeff_t *block,int qno,int klass,dv_248_coeff_t *co);
|
|
|
f76d5d |
@@ -210,7 +210,7 @@ void _dv_quant(dv_coeff_t *block,int qno
|
|
|
f76d5d |
_dv_quant_x86_64(block, qno, klass);
|
|
|
f76d5d |
emms();
|
|
|
f76d5d |
#else
|
|
|
f76d5d |
- _dv_quant_x86(block, qno, klass);
|
|
|
f76d5d |
+ _dv_quant_x86(block, qno, klass, dv_quant_offset, dv_quant_shifts);
|
|
|
f76d5d |
emms();
|
|
|
f76d5d |
#endif
|
|
|
f76d5d |
}
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/quant.h
|
|
|
f76d5d |
+++ libdv-0.104/libdv/quant.h
|
|
|
f76d5d |
@@ -27,7 +27,7 @@ extern void _dv_quant(dv_coeff_t *block,
|
|
|
f76d5d |
extern void _dv_quant_88_inverse(dv_coeff_t *block,int qno,int klass);
|
|
|
f76d5d |
extern void (*_dv_quant_248_inverse) (dv_coeff_t *block,int qno,int klass,
|
|
|
f76d5d |
dv_248_coeff_t *co);
|
|
|
f76d5d |
-extern void _dv_quant_88_inverse_x86(dv_coeff_t *block,int qno,int klass);
|
|
|
f76d5d |
+extern void _dv_quant_88_inverse_x86(dv_coeff_t *block,int qno,int klass, uint8_t *offset, uint8_t *shifts);
|
|
|
f76d5d |
extern void _dv_quant_88_inverse_x86_64(dv_coeff_t *block,int qno,int klass);
|
|
|
f76d5d |
extern void dv_quant_init (void);
|
|
|
f76d5d |
#ifdef __cplusplus
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/quant_x86.S
|
|
|
f76d5d |
+++ libdv-0.104/libdv/quant_x86.S
|
|
|
f76d5d |
@@ -73,10 +75,13 @@ _dv_quant_88_inverse_x86:
|
|
|
f76d5d |
|
|
|
f76d5d |
/* pq = dv_quant_shifts[qno + dv_quant_offset[class]]; */
|
|
|
f76d5d |
movl ARGn(1),%eax /* qno */
|
|
|
f76d5d |
+ movl ARGn(3),%ebx /* dv_quant_offset */
|
|
|
f76d5d |
+ addl ARGn(2),%ebx /* class */
|
|
|
f76d5d |
+ movzbl (%ebx),%ecx
|
|
|
f76d5d |
movl ARGn(2),%ebx /* class */
|
|
|
f76d5d |
- movzbl dv_quant_offset(%ebx),%ecx
|
|
|
f76d5d |
addl %ecx,%eax
|
|
|
f76d5d |
- leal dv_quant_shifts(,%eax,4),%edx /* edx is pq */
|
|
|
f76d5d |
+ movl ARGn(4),%edx /* dv_quant_shifts */
|
|
|
f76d5d |
+ leal (%edx,%eax,4),%edx /* edx is pq */
|
|
|
f76d5d |
|
|
|
f76d5d |
/* extra = (class == 3); */
|
|
|
f76d5d |
/* 0 1 2 3 */
|
|
|
f76d5d |
@@ -214,11 +221,13 @@ _dv_quant_x86:
|
|
|
f76d5d |
|
|
|
f76d5d |
/* pq = dv_quant_shifts[qno + dv_quant_offset[class]]; */
|
|
|
f76d5d |
movl ARGn(1),%eax /* qno */
|
|
|
f76d5d |
+ movl ARGn(3),%ebx /* offset */
|
|
|
f76d5d |
+ addl ARGn(2),%ebx /* class */
|
|
|
f76d5d |
+ movzbl (%ebx),%ecx
|
|
|
f76d5d |
movl ARGn(2),%ebx /* class */
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movzbl dv_quant_offset(%ebx),%ecx
|
|
|
f76d5d |
+ movl ARGn(4),%edx /* shifts */
|
|
|
f76d5d |
addl %ecx,%eax
|
|
|
f76d5d |
- leal dv_quant_shifts(,%eax,4),%edx /* edx is pq */
|
|
|
f76d5d |
+ leal (%edx,%eax,4),%edx /* edx is pq */
|
|
|
f76d5d |
|
|
|
f76d5d |
/* extra = (class == 3); */
|
|
|
f76d5d |
/* 0 1 2 3 */
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/rgbtoyuv.S
|
|
|
f76d5d |
+++ libdv-0.104/libdv/rgbtoyuv.S
|
|
|
f76d5d |
@@ -41,9 +41,6 @@
|
|
|
f76d5d |
#define DV_WIDTH_SHORT_HALF 720
|
|
|
f76d5d |
#define DV_WIDTH_BYTE_HALF 360
|
|
|
f76d5d |
|
|
|
f76d5d |
-.global _dv_rgbtoycb_mmx
|
|
|
f76d5d |
-# .global yuvtoycb_mmx
|
|
|
f76d5d |
-
|
|
|
f76d5d |
.data
|
|
|
f76d5d |
|
|
|
f76d5d |
.align 8
|
|
|
f76d5d |
@@ -110,25 +107,26 @@ VR0GR: .long 0,0
|
|
|
f76d5d |
VBG0B: .long 0,0
|
|
|
f76d5d |
|
|
|
f76d5d |
#endif
|
|
|
f76d5d |
-
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+#include "asm_common.S"
|
|
|
f76d5d |
+
|
|
|
f76d5d |
.section .note.GNU-stack, "", @progbits
|
|
|
f76d5d |
|
|
|
f76d5d |
.text
|
|
|
f76d5d |
|
|
|
f76d5d |
-#define _inPtr 8
|
|
|
f76d5d |
-#define _rows 12
|
|
|
f76d5d |
-#define _columns 16
|
|
|
f76d5d |
-#define _outyPtr 20
|
|
|
f76d5d |
-#define _outuPtr 24
|
|
|
f76d5d |
-#define _outvPtr 28
|
|
|
f76d5d |
+#define _inPtr 24+8
|
|
|
f76d5d |
+#define _rows 24+12
|
|
|
f76d5d |
+#define _columns 24+16
|
|
|
f76d5d |
+#define _outyPtr 24+20
|
|
|
f76d5d |
+#define _outuPtr 24+24
|
|
|
f76d5d |
+#define _outvPtr 24+28
|
|
|
f76d5d |
|
|
|
f76d5d |
.global _dv_rgbtoycb_mmx
|
|
|
f76d5d |
.hidden _dv_rgbtoycb_mmx
|
|
|
f76d5d |
.type _dv_rgbtoycb_mmx,@function
|
|
|
f76d5d |
_dv_rgbtoycb_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp, %ebp
|
|
|
f76d5d |
pushl %eax
|
|
|
f76d5d |
pushl %ebx
|
|
|
f76d5d |
pushl %ecx
|
|
|
f76d5d |
@@ -133,46 +134,47 @@ _dv_rgbtoycb_mmx:
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
|
|
|
f76d5d |
- leal ZEROSX, %eax #This section gets around a bug
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ leal MUNG(ZEROSX), %eax #This section gets around a bug
|
|
|
f76d5d |
movq (%eax), %mm0 #unlikely to persist
|
|
|
f76d5d |
- movq %mm0, ZEROS
|
|
|
f76d5d |
- leal OFFSETDX, %eax
|
|
|
f76d5d |
+ movq %mm0, MUNG(ZEROS)
|
|
|
f76d5d |
+ leal MUNG(OFFSETDX), %eax
|
|
|
f76d5d |
movq (%eax), %mm0
|
|
|
f76d5d |
- movq %mm0, OFFSETD
|
|
|
f76d5d |
- leal OFFSETWX, %eax
|
|
|
f76d5d |
+ movq %mm0, MUNG(OFFSETD)
|
|
|
f76d5d |
+ leal MUNG(OFFSETWX), %eax
|
|
|
f76d5d |
movq (%eax), %mm0
|
|
|
f76d5d |
- movq %mm0, OFFSETW
|
|
|
f76d5d |
- leal OFFSETBX, %eax
|
|
|
f76d5d |
+ movq %mm0, MUNG(OFFSETW)
|
|
|
f76d5d |
+ leal MUNG(OFFSETBX), %eax
|
|
|
f76d5d |
movq (%eax), %mm0
|
|
|
f76d5d |
- movq %mm0, OFFSETB
|
|
|
f76d5d |
- leal YR0GRX, %eax
|
|
|
f76d5d |
+ movq %mm0, MUNG(OFFSETB)
|
|
|
f76d5d |
+ leal MUNG(YR0GRX), %eax
|
|
|
f76d5d |
movq (%eax), %mm0
|
|
|
f76d5d |
- movq %mm0, YR0GR
|
|
|
f76d5d |
- leal YBG0BX, %eax
|
|
|
f76d5d |
+ movq %mm0, MUNG(YR0GR)
|
|
|
f76d5d |
+ leal MUNG(YBG0BX), %eax
|
|
|
f76d5d |
movq (%eax), %mm0
|
|
|
f76d5d |
- movq %mm0, YBG0B
|
|
|
f76d5d |
- leal UR0GRX, %eax
|
|
|
f76d5d |
+ movq %mm0, MUNG(YBG0B)
|
|
|
f76d5d |
+ leal MUNG(UR0GRX), %eax
|
|
|
f76d5d |
movq (%eax), %mm0
|
|
|
f76d5d |
- movq %mm0, UR0GR
|
|
|
f76d5d |
- leal UBG0BX, %eax
|
|
|
f76d5d |
+ movq %mm0, MUNG(UR0GR)
|
|
|
f76d5d |
+ leal MUNG(UBG0BX), %eax
|
|
|
f76d5d |
movq (%eax), %mm0
|
|
|
f76d5d |
- movq %mm0, UBG0B
|
|
|
f76d5d |
- leal VR0GRX, %eax
|
|
|
f76d5d |
+ movq %mm0, MUNG(UBG0B)
|
|
|
f76d5d |
+ leal MUNG(VR0GRX), %eax
|
|
|
f76d5d |
movq (%eax), %mm0
|
|
|
f76d5d |
- movq %mm0, VR0GR
|
|
|
f76d5d |
- leal VBG0BX, %eax
|
|
|
f76d5d |
+ movq %mm0, MUNG(VR0GR)
|
|
|
f76d5d |
+ leal MUNG(VBG0BX), %eax
|
|
|
f76d5d |
movq (%eax), %mm0
|
|
|
f76d5d |
- movq %mm0, VBG0B
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movl _rows(%ebp), %eax
|
|
|
f76d5d |
- movl _columns(%ebp), %ebx
|
|
|
f76d5d |
+ movq %mm0, MUNG(VBG0B)
|
|
|
f76d5d |
+ movl _rows(%esp), %eax
|
|
|
f76d5d |
+ movl _columns(%esp), %ebx
|
|
|
f76d5d |
mull %ebx #number pixels
|
|
|
f76d5d |
shrl $3, %eax #number of loops
|
|
|
f76d5d |
movl %eax, %edi #loop counter in edi
|
|
|
f76d5d |
- movl _inPtr(%ebp), %eax
|
|
|
f76d5d |
- movl _outyPtr(%ebp), %ebx
|
|
|
f76d5d |
- movl _outuPtr(%ebp), %ecx
|
|
|
f76d5d |
- movl _outvPtr(%ebp), %edx
|
|
|
f76d5d |
+ movl _inPtr(%esp), %eax
|
|
|
f76d5d |
+ movl _outyPtr(%esp), %ebx
|
|
|
f76d5d |
+ movl _outuPtr(%esp), %ecx
|
|
|
f76d5d |
+ movl _outvPtr(%esp), %edx
|
|
|
f76d5d |
rgbtoycb_mmx_loop:
|
|
|
f76d5d |
movq (%eax), %mm1 #load G2R2B1G1R1B0G0R0
|
|
|
f76d5d |
pxor %mm6, %mm6 #0 -> mm6
|
|
|
f76d5d |
@@ -186,29 +188,29 @@ rgbtoycb_mmx_loop:
|
|
|
f76d5d |
punpcklbw %mm6, %mm1 #B1G1R1B0 -> mm1
|
|
|
f76d5d |
movq %mm0, %mm2 #R1B0G0R0 -> mm2
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd YR0GR, %mm0 #yrR1,ygG0+yrR0 -> mm0
|
|
|
f76d5d |
+ pmaddwd MUNG(YR0GR), %mm0 #yrR1,ygG0+yrR0 -> mm0
|
|
|
f76d5d |
movq %mm1, %mm3 #B1G1R1B0 -> mm3
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd YBG0B, %mm1 #ybB1+ygG1,ybB0 -> mm1
|
|
|
f76d5d |
+ pmaddwd MUNG(YBG0B), %mm1 #ybB1+ygG1,ybB0 -> mm1
|
|
|
f76d5d |
movq %mm2, %mm4 #R1B0G0R0 -> mm4
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd UR0GR, %mm2 #urR1,ugG0+urR0 -> mm2
|
|
|
f76d5d |
+ pmaddwd MUNG(UR0GR), %mm2 #urR1,ugG0+urR0 -> mm2
|
|
|
f76d5d |
movq %mm3, %mm5 #B1G1R1B0 -> mm5
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd UBG0B, %mm3 #ubB1+ugG1,ubB0 -> mm3
|
|
|
f76d5d |
+ pmaddwd MUNG(UBG0B), %mm3 #ubB1+ugG1,ubB0 -> mm3
|
|
|
f76d5d |
punpckhbw %mm6, %mm7 # 00G2R2 -> mm7
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd VR0GR, %mm4 #vrR1,vgG0+vrR0 -> mm4
|
|
|
f76d5d |
+ pmaddwd MUNG(VR0GR), %mm4 #vrR1,vgG0+vrR0 -> mm4
|
|
|
f76d5d |
paddd %mm1, %mm0 #Y1Y0 -> mm0
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd VBG0B, %mm5 #vbB1+vgG1,vbB0 -> mm5
|
|
|
f76d5d |
+ pmaddwd MUNG(VBG0B), %mm5 #vbB1+vgG1,vbB0 -> mm5
|
|
|
f76d5d |
|
|
|
f76d5d |
movq 8(%eax), %mm1 #R5B4G4R4B3G3R3B2 -> mm1
|
|
|
f76d5d |
paddd %mm3, %mm2 #U1U0 -> mm2
|
|
|
f76d5d |
|
|
|
f76d5d |
movq %mm1, %mm6 #R5B4G4R4B3G3R3B2 -> mm6
|
|
|
f76d5d |
|
|
|
f76d5d |
- punpcklbw ZEROS, %mm1 #B3G3R3B2 -> mm1
|
|
|
f76d5d |
+ punpcklbw MUNG(ZEROS), %mm1 #B3G3R3B2 -> mm1
|
|
|
f76d5d |
paddd %mm5, %mm4 #V1V0 -> mm4
|
|
|
f76d5d |
|
|
|
f76d5d |
movq %mm1, %mm5 #B3G3R3B2 -> mm5
|
|
|
f76d5d |
@@ -216,29 +218,29 @@ rgbtoycb_mmx_loop:
|
|
|
f76d5d |
|
|
|
f76d5d |
paddd %mm7, %mm1 #R3B200+00G2R2=R3B2G2R2->mm1
|
|
|
f76d5d |
|
|
|
f76d5d |
- punpckhbw ZEROS, %mm6 #R5B4G4R3 -> mm6
|
|
|
f76d5d |
+ punpckhbw MUNG(ZEROS), %mm6 #R5B4G4R3 -> mm6
|
|
|
f76d5d |
movq %mm1, %mm3 #R3B2G2R2 -> mm3
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd YR0GR, %mm1 #yrR3,ygG2+yrR2 -> mm1
|
|
|
f76d5d |
+ pmaddwd MUNG(YR0GR), %mm1 #yrR3,ygG2+yrR2 -> mm1
|
|
|
f76d5d |
movq %mm5, %mm7 #B3G3R3B2 -> mm7
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd YBG0B, %mm5 #ybB3+ygG3,ybB2 -> mm5
|
|
|
f76d5d |
+ pmaddwd MUNG(YBG0B), %mm5 #ybB3+ygG3,ybB2 -> mm5
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm0 #32-bit scaled Y1Y0 -> mm0
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq %mm6, TEMP0 #R5B4G4R4 -> TEMP0
|
|
|
f76d5d |
+ movq %mm6, MUNG(TEMP0) #R5B4G4R4 -> TEMP0
|
|
|
f76d5d |
movq %mm3, %mm6 #R3B2G2R2 -> mm6
|
|
|
f76d5d |
- pmaddwd UR0GR, %mm6 #urR3,ugG2+urR2 -> mm6
|
|
|
f76d5d |
+ pmaddwd MUNG(UR0GR), %mm6 #urR3,ugG2+urR2 -> mm6
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm2 #32-bit scaled U1U0 -> mm2
|
|
|
f76d5d |
|
|
|
f76d5d |
paddd %mm5, %mm1 #Y3Y2 -> mm1
|
|
|
f76d5d |
movq %mm7, %mm5 #B3G3R3B2 -> mm5
|
|
|
f76d5d |
- pmaddwd UBG0B, %mm7 #ubB3+ugG3,ubB2
|
|
|
f76d5d |
+ pmaddwd MUNG(UBG0B), %mm7 #ubB3+ugG3,ubB2
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm1 #32-bit scaled Y3Y2 -> mm1
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd VR0GR, %mm3 #vrR3,vgG2+vgR2
|
|
|
f76d5d |
+ pmaddwd MUNG(VR0GR), %mm3 #vrR3,vgG2+vgR2
|
|
|
f76d5d |
packssdw %mm1, %mm0 #Y3Y2Y1Y0 -> mm0
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd VBG0B, %mm5 #vbB3+vgG3,vbB2 -> mm5
|
|
|
f76d5d |
+ pmaddwd MUNG(VBG0B), %mm5 #vbB3+vgG3,vbB2 -> mm5
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm4 #32-bit scaled V1V0 -> mm4
|
|
|
f76d5d |
|
|
|
f76d5d |
movq 16(%eax), %mm1 #B7G7R7B6G6R6B5G5 -> mm7
|
|
|
f76d5d |
@@ -253,58 +255,58 @@ rgbtoycb_mmx_loop:
|
|
|
f76d5d |
movq %mm7, %mm5 #R7B6G6R6B5G500 -> mm5
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm3 #32-bit scaled V3V2 -> mm3
|
|
|
f76d5d |
|
|
|
f76d5d |
- paddw OFFSETY, %mm0
|
|
|
f76d5d |
+ paddw MUNG(OFFSETY), %mm0
|
|
|
f76d5d |
movq %mm0, (%ebx) #store Y3Y2Y1Y0
|
|
|
f76d5d |
packssdw %mm6, %mm2 #32-bit scaled U3U2U1U0 -> mm2
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq TEMP0, %mm0 #R5B4G4R4 -> mm0
|
|
|
f76d5d |
+ movq MUNG(TEMP0), %mm0 #R5B4G4R4 -> mm0
|
|
|
f76d5d |
addl $8, %ebx
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- punpcklbw ZEROS, %mm7 #B5G500 -> mm7
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ punpcklbw MUNG(ZEROS), %mm7 #B5G500 -> mm7
|
|
|
f76d5d |
movq %mm0, %mm6 #R5B4G4R4 -> mm6
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq %mm2, TEMPU #32-bit scaled U3U2U1U0 -> TEMPU
|
|
|
f76d5d |
+ movq %mm2, MUNG(TEMPU) #32-bit scaled U3U2U1U0 -> TEMPU
|
|
|
f76d5d |
psrlq $32, %mm0 #00R5B4 -> mm0
|
|
|
f76d5d |
|
|
|
f76d5d |
paddw %mm0, %mm7 #B5G5R5B4 -> mm7
|
|
|
f76d5d |
movq %mm6, %mm2 #B5B4G4R4 -> mm2
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd YR0GR, %mm2 #yrR5,ygG4+yrR4 -> mm2
|
|
|
f76d5d |
+ pmaddwd MUNG(YR0GR), %mm2 #yrR5,ygG4+yrR4 -> mm2
|
|
|
f76d5d |
movq %mm7, %mm0 #B5G5R5B4 -> mm0
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd YBG0B, %mm7 #ybB5+ygG5,ybB4 -> mm7
|
|
|
f76d5d |
+ pmaddwd MUNG(YBG0B), %mm7 #ybB5+ygG5,ybB4 -> mm7
|
|
|
f76d5d |
packssdw %mm3, %mm4 #32-bit scaled V3V2V1V0 -> mm4
|
|
|
f76d5d |
|
|
|
f76d5d |
addl $24, %eax #increment RGB count
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq %mm4, TEMPV #(V3V2V1V0)/256 -> mm4
|
|
|
f76d5d |
+ movq %mm4, MUNG(TEMPV) #(V3V2V1V0)/256 -> mm4
|
|
|
f76d5d |
movq %mm6, %mm4 #B5B4G4R4 -> mm4
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd UR0GR, %mm6 #urR5,ugG4+urR4
|
|
|
f76d5d |
+ pmaddwd MUNG(UR0GR), %mm6 #urR5,ugG4+urR4
|
|
|
f76d5d |
movq %mm0, %mm3 #B5G5R5B4 -> mm0
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd UBG0B, %mm0 #ubB5+ugG5,ubB4
|
|
|
f76d5d |
+ pmaddwd MUNG(UBG0B), %mm0 #ubB5+ugG5,ubB4
|
|
|
f76d5d |
paddd %mm7, %mm2 #Y5Y4 -> mm2
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd VR0GR, %mm4 #vrR5,vgG4+vrR4 -> mm4
|
|
|
f76d5d |
+ pmaddwd MUNG(VR0GR), %mm4 #vrR5,vgG4+vrR4 -> mm4
|
|
|
f76d5d |
pxor %mm7, %mm7 #0 -> mm7
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd VBG0B, %mm3 #vbB5+vgG5,vbB4 -> mm3
|
|
|
f76d5d |
+ pmaddwd MUNG(VBG0B), %mm3 #vbB5+vgG5,vbB4 -> mm3
|
|
|
f76d5d |
punpckhbw %mm7, %mm1 #B7G7R7B6 -> mm1
|
|
|
f76d5d |
|
|
|
f76d5d |
paddd %mm6, %mm0 #U5U4 -> mm0
|
|
|
f76d5d |
movq %mm1, %mm6 #B7G7R7B6 -> mm6
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd YBG0B, %mm6 #ybB7+ygG7,ybB6 -> mm6
|
|
|
f76d5d |
+ pmaddwd MUNG(YBG0B), %mm6 #ybB7+ygG7,ybB6 -> mm6
|
|
|
f76d5d |
punpckhbw %mm7, %mm5 #R7B6G6R6 -> mm5
|
|
|
f76d5d |
|
|
|
f76d5d |
movq %mm5, %mm7 #R7B6G6R6 -> mm7
|
|
|
f76d5d |
paddd %mm4, %mm3 #V5V4 -> mm3
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd YR0GR, %mm5 #yrR7,ygG6+yrR6 -> mm5
|
|
|
f76d5d |
+ pmaddwd MUNG(YR0GR), %mm5 #yrR7,ygG6+yrR6 -> mm5
|
|
|
f76d5d |
movq %mm1, %mm4 #B7G7R7B6 -> mm4
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd UBG0B, %mm4 #ubB7+ugG7,ubB6 -> mm4
|
|
|
f76d5d |
+ pmaddwd MUNG(UBG0B), %mm4 #ubB7+ugG7,ubB6 -> mm4
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm0 #32-bit scaled U5U4 -> mm0
|
|
|
f76d5d |
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm2 #32-bit scaled Y5Y4 -> mm2
|
|
|
f76d5d |
@@ -312,25 +314,25 @@ rgbtoycb_mmx_loop:
|
|
|
f76d5d |
paddd %mm5, %mm6 #Y7Y6 -> mm6
|
|
|
f76d5d |
movq %mm7, %mm5 #R7B6G6R6 -> mm5
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd UR0GR, %mm7 #urR7,ugG6+ugR6 -> mm7
|
|
|
f76d5d |
+ pmaddwd MUNG(UR0GR), %mm7 #urR7,ugG6+ugR6 -> mm7
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm3 #32-bit scaled V5V4 -> mm3
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd VBG0B, %mm1 #vbB7+vgG7,vbB6 -> mm1
|
|
|
f76d5d |
+ pmaddwd MUNG(VBG0B), %mm1 #vbB7+vgG7,vbB6 -> mm1
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm6 #32-bit scaled Y7Y6 -> mm6
|
|
|
f76d5d |
|
|
|
f76d5d |
packssdw %mm6, %mm2 #Y7Y6Y5Y4 -> mm2
|
|
|
f76d5d |
|
|
|
f76d5d |
- pmaddwd VR0GR, %mm5 #vrR7,vgG6+vrR6 -> mm5
|
|
|
f76d5d |
+ pmaddwd MUNG(VR0GR), %mm5 #vrR7,vgG6+vrR6 -> mm5
|
|
|
f76d5d |
paddd %mm4, %mm7 #U7U6 -> mm7
|
|
|
f76d5d |
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm7 #32-bit scaled U7U6 -> mm7
|
|
|
f76d5d |
- paddw OFFSETY, %mm2
|
|
|
f76d5d |
+ paddw MUNG(OFFSETY), %mm2
|
|
|
f76d5d |
movq %mm2, (%ebx) #store Y7Y6Y5Y4
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq ALLONE, %mm6
|
|
|
f76d5d |
+ movq MUNG(ALLONE), %mm6
|
|
|
f76d5d |
packssdw %mm7, %mm0 #32-bit scaled U7U6U5U4 -> mm0
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq TEMPU, %mm4 #32-bit scaled U3U2U1U0 -> mm4
|
|
|
f76d5d |
+ movq MUNG(TEMPU), %mm4 #32-bit scaled U3U2U1U0 -> mm4
|
|
|
f76d5d |
pmaddwd %mm6, %mm0 #U7U6U5U4 averaged -> (U7U6)(U5U4)=UU3 UU2->mm0
|
|
|
f76d5d |
|
|
|
f76d5d |
pmaddwd %mm6, %mm4 #U3U2U1U0 averaged -> (U3U2)(U1U0)=UU1 UU0->mm4
|
|
|
f76d5d |
@@ -340,8 +342,8 @@ rgbtoycb_mmx_loop:
|
|
|
f76d5d |
|
|
|
f76d5d |
psrad $FIXPSHIFT, %mm1 #32-bit scaled V7V6 -> mm1
|
|
|
f76d5d |
psraw $1, %mm4 #divide UU3 UU2 UU1 UU0 by 2 -> mm4
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movq TEMPV, %mm5 #32-bit scaled V3V2V1V0 -> mm5
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movq MUNG(TEMPV), %mm5 #32-bit scaled V3V2V1V0 -> mm5
|
|
|
f76d5d |
|
|
|
f76d5d |
movq %mm4, (%ecx) # store U
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -429,14 +433,15 @@ _dv_ppm_copy_y_block_mmx:
|
|
|
f76d5d |
_dv_pgm_copy_y_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp, %ebp
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movl 8(%ebp), %edi # dest
|
|
|
f76d5d |
- movl 12(%ebp), %esi # src
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq OFFSETY, %mm7
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl 16(%esp), %edi # dest
|
|
|
f76d5d |
+ movl 20(%esp), %esi # src
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movq MUNG(OFFSETY), %mm7
|
|
|
f76d5d |
pxor %mm6, %mm6
|
|
|
f76d5d |
|
|
|
f76d5d |
movq (%esi), %mm0
|
|
|
f76d5d |
@@ -571,14 +578,15 @@ _dv_pgm_copy_y_block_mmx:
|
|
|
f76d5d |
_dv_video_copy_y_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp, %ebp
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movl 8(%ebp), %edi # dest
|
|
|
f76d5d |
- movl 12(%ebp), %esi # src
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq OFFSETBX, %mm7
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl 16(%esp), %edi # dest
|
|
|
f76d5d |
+ movl 20(%esp), %esi # src
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movq MUNG(OFFSETBX), %mm7
|
|
|
f76d5d |
pxor %mm6, %mm6
|
|
|
f76d5d |
|
|
|
f76d5d |
movq (%esi), %mm0
|
|
|
f76d5d |
@@ -859,16 +871,16 @@ _dv_ppm_copy_pal_c_block_mmx:
|
|
|
f76d5d |
_dv_pgm_copy_pal_c_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp, %ebp
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
pushl %ebx
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movl 8(%ebp), %edi # dest
|
|
|
f76d5d |
- movl 12(%ebp), %esi # src
|
|
|
f76d5d |
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl 20(%esp), %edi # dest
|
|
|
f76d5d |
+ movl 24(%esp), %esi # src
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq OFFSETBX, %mm7
|
|
|
f76d5d |
+ movq MUNG(OFFSETBX), %mm7
|
|
|
f76d5d |
pxor %mm6, %mm6
|
|
|
f76d5d |
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -1007,15 +1021,16 @@ _dv_pgm_copy_pal_c_block_mmx:
|
|
|
f76d5d |
_dv_video_copy_pal_c_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp, %ebp
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
pushl %ebx
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movl 8(%ebp), %edi # dest
|
|
|
f76d5d |
- movl 12(%ebp), %esi # src
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq OFFSETBX, %mm7
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl 20(%esp), %edi # dest
|
|
|
f76d5d |
+ movl 24(%esp), %esi # src
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movq MUNG(OFFSETBX), %mm7
|
|
|
f76d5d |
paddw %mm7, %mm7
|
|
|
f76d5d |
pxor %mm6, %mm6
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -1102,18 +1119,18 @@ video_copy_pal_c_block_mmx_loop:
|
|
|
f76d5d |
_dv_ppm_copy_ntsc_c_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp, %ebp
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
pushl %ebx
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movl 8(%ebp), %edi # dest
|
|
|
f76d5d |
- movl 12(%ebp), %esi # src
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl 20(%esp), %edi # dest
|
|
|
f76d5d |
+ movl 24(%esp), %esi # src
|
|
|
f76d5d |
|
|
|
f76d5d |
movl $4, %ebx
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq ALLONE, %mm6
|
|
|
f76d5d |
-
|
|
|
f76d5d |
+ movq MUNG(ALLONE), %mm6
|
|
|
f76d5d |
ppm_copy_ntsc_c_block_mmx_loop:
|
|
|
f76d5d |
|
|
|
f76d5d |
movq (%esi), %mm0
|
|
|
f76d5d |
@@ -1175,14 +1194,15 @@ ppm_copy_ntsc_c_block_mmx_loop:
|
|
|
f76d5d |
_dv_pgm_copy_ntsc_c_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp, %ebp
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movl 8(%ebp), %edi # dest
|
|
|
f76d5d |
- movl 12(%ebp), %esi # src
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq OFFSETBX, %mm7
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl 16(%esp), %edi # dest
|
|
|
f76d5d |
+ movl 20(%esp), %esi # src
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movq MUNG(OFFSETBX), %mm7
|
|
|
f76d5d |
paddw %mm7, %mm7
|
|
|
f76d5d |
pxor %mm6, %mm6
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -1332,15 +1354,16 @@ _dv_pgm_copy_ntsc_c_block_mmx:
|
|
|
f76d5d |
_dv_video_copy_ntsc_c_block_mmx:
|
|
|
f76d5d |
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
- movl %esp, %ebp
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
pushl %ebx
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movl 8(%ebp), %edi # dest
|
|
|
f76d5d |
- movl 12(%ebp), %esi # src
|
|
|
f76d5d |
|
|
|
f76d5d |
- movq OFFSETBX, %mm7
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl 20(%esp), %edi # dest
|
|
|
f76d5d |
+ movl 24(%esp), %esi # src
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movq MUNG(OFFSETBX), %mm7
|
|
|
f76d5d |
paddw %mm7, %mm7
|
|
|
f76d5d |
pxor %mm6, %mm6
|
|
|
f76d5d |
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/rgbtoyuv_x86_64.S
|
|
|
f76d5d |
+++ libdv-0.104/libdv/rgbtoyuv_x86_64.S
|
|
|
f76d5d |
@@ -41,9 +41,6 @@
|
|
|
f76d5d |
#define DV_WIDTH_SHORT_HALF 720
|
|
|
f76d5d |
#define DV_WIDTH_BYTE_HALF 360
|
|
|
f76d5d |
|
|
|
f76d5d |
-.global _dv_rgbtoycb_mmx_x86_64
|
|
|
f76d5d |
-# .global yuvtoycb_mmx_x86_64
|
|
|
f76d5d |
-
|
|
|
f76d5d |
.data
|
|
|
f76d5d |
|
|
|
f76d5d |
.align 8
|
|
|
f76d5d |
--- libdv-0.104-old/libdv/vlc_x86.S
|
|
|
f76d5d |
+++ libdv-0.104/libdv/vlc_x86.S
|
|
|
f76d5d |
@@ -1,31 +1,39 @@
|
|
|
f76d5d |
#include "asmoff.h"
|
|
|
f76d5d |
.section .note.GNU-stack, "", @progbits
|
|
|
f76d5d |
+ #include "asm_common.S"
|
|
|
f76d5d |
|
|
|
f76d5d |
.text
|
|
|
f76d5d |
.align 4
|
|
|
f76d5d |
.globl dv_decode_vlc
|
|
|
f76d5d |
+.globl asm_dv_decode_vlc
|
|
|
f76d5d |
+.hidden asm_dv_decode_vlc
|
|
|
f76d5d |
+asm_dv_decode_vlc = dv_decode_vlc
|
|
|
f76d5d |
+
|
|
|
f76d5d |
.type dv_decode_vlc,@function
|
|
|
f76d5d |
dv_decode_vlc:
|
|
|
f76d5d |
pushl %ebx
|
|
|
f76d5d |
+ pushl %ebp
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
|
|
|
f76d5d |
- /* Args are at 8(%esp). */
|
|
|
f76d5d |
- movl 8(%esp),%eax /* %eax is bits */
|
|
|
f76d5d |
- movl 12(%esp),%ebx /* %ebx is maxbits */
|
|
|
f76d5d |
+ /* Args are at 12(%esp). */
|
|
|
f76d5d |
+ movl 12(%esp),%eax /* %eax is bits */
|
|
|
f76d5d |
+ movl 16(%esp),%ebx /* %ebx is maxbits */
|
|
|
f76d5d |
andl $0x3f,%ebx /* limit index range STL*/
|
|
|
f76d5d |
|
|
|
f76d5d |
- movl dv_vlc_class_index_mask(,%ebx,4),%edx
|
|
|
f76d5d |
+ movl MUNG_ARR(dv_vlc_class_index_mask,%ebx,4),%edx
|
|
|
f76d5d |
andl %eax,%edx
|
|
|
f76d5d |
- movl dv_vlc_class_index_rshift(,%ebx,4),%ecx
|
|
|
f76d5d |
+ movl MUNG_ARR(dv_vlc_class_index_rshift,%ebx,4),%ecx
|
|
|
f76d5d |
sarl %cl,%edx
|
|
|
f76d5d |
- movl dv_vlc_classes(,%ebx,4),%ecx
|
|
|
f76d5d |
+ movl MUNG_ARR(dv_vlc_classes,%ebx,4),%ecx
|
|
|
f76d5d |
movsbl (%ecx,%edx,1),%edx /* %edx is class */
|
|
|
f76d5d |
|
|
|
f76d5d |
- movl dv_vlc_index_mask(,%edx,4),%ebx
|
|
|
f76d5d |
- movl dv_vlc_index_rshift(,%edx,4),%ecx
|
|
|
f76d5d |
+ movl MUNG_ARR(dv_vlc_index_mask,%edx,4),%ebx
|
|
|
f76d5d |
+ movl MUNG_ARR(dv_vlc_index_rshift,%edx,4),%ecx
|
|
|
f76d5d |
andl %eax,%ebx
|
|
|
f76d5d |
sarl %cl,%ebx
|
|
|
f76d5d |
|
|
|
f76d5d |
- movl dv_vlc_lookups(,%edx,4),%edx
|
|
|
f76d5d |
+ movl MUNG_ARR(dv_vlc_lookups,%edx,4),%edx
|
|
|
f76d5d |
movl (%edx,%ebx,4),%edx
|
|
|
f76d5d |
|
|
|
f76d5d |
/* Now %edx holds result, like this:
|
|
|
f76d5d |
@@ -43,7 +52,7 @@ dv_decode_vlc:
|
|
|
f76d5d |
movl %edx,%ecx
|
|
|
f76d5d |
sarl $8,%ecx
|
|
|
f76d5d |
andl $0xff,%ecx
|
|
|
f76d5d |
- movl sign_mask(,%ecx,4),%ebx
|
|
|
f76d5d |
+ movl MUNG_ARR(sign_mask,%ecx,4),%ebx
|
|
|
f76d5d |
andl %ebx,%eax
|
|
|
f76d5d |
negl %eax
|
|
|
f76d5d |
sarl $31,%eax
|
|
|
f76d5d |
@@ -64,14 +73,14 @@ dv_decode_vlc:
|
|
|
f76d5d |
*result = broken;
|
|
|
f76d5d |
Note that the 'broken' pattern is all ones (i.e. 0xffffffff)
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
- movl 12(%esp),%ebx /* %ebx is maxbits */
|
|
|
f76d5d |
+ movl 16(%esp),%ebx /* %ebx is maxbits */
|
|
|
f76d5d |
subl %ecx,%ebx
|
|
|
f76d5d |
sbbl %ebx,%ebx
|
|
|
f76d5d |
orl %ebx,%edx
|
|
|
f76d5d |
|
|
|
f76d5d |
- movl 16(%esp),%eax
|
|
|
f76d5d |
+ movl 20(%esp),%eax
|
|
|
f76d5d |
movl %edx,(%eax)
|
|
|
f76d5d |
-
|
|
|
f76d5d |
+ popl %ebp
|
|
|
f76d5d |
popl %ebx
|
|
|
f76d5d |
ret
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -81,21 +90,28 @@ dv_decode_vlc:
|
|
|
f76d5d |
.type __dv_decode_vlc,@function
|
|
|
f76d5d |
__dv_decode_vlc:
|
|
|
f76d5d |
pushl %ebx
|
|
|
f76d5d |
+ pushl %ebp
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ LOAD_PIC_REG(bp)
|
|
|
f76d5d |
|
|
|
f76d5d |
- /* Args are at 8(%esp). */
|
|
|
f76d5d |
- movl 8(%esp),%eax /* %eax is bits */
|
|
|
f76d5d |
+ /* Args are at 12(%esp). */
|
|
|
f76d5d |
+ movl 12(%esp),%eax /* %eax is bits */
|
|
|
f76d5d |
|
|
|
f76d5d |
movl %eax,%edx /* %edx is class */
|
|
|
f76d5d |
andl $0xfe00,%edx
|
|
|
f76d5d |
sarl $9,%edx
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movsbl dv_vlc_class_lookup5@GOTOFF(%ebp,%edx),%edx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movsbl dv_vlc_class_lookup5(%edx),%edx
|
|
|
f76d5d |
-
|
|
|
f76d5d |
- movl dv_vlc_index_mask(,%edx,4),%ebx
|
|
|
f76d5d |
- movl dv_vlc_index_rshift(,%edx,4),%ecx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl MUNG_ARR(dv_vlc_index_mask,%edx,4),%ebx
|
|
|
f76d5d |
+ movl MUNG_ARR(dv_vlc_index_rshift,%edx,4),%ecx
|
|
|
f76d5d |
andl %eax,%ebx
|
|
|
f76d5d |
sarl %cl,%ebx
|
|
|
f76d5d |
|
|
|
f76d5d |
- movl dv_vlc_lookups(,%edx,4),%edx
|
|
|
f76d5d |
+ movl MUNG_ARR(dv_vlc_lookups,%edx,4),%edx
|
|
|
f76d5d |
movl (%edx,%ebx,4),%edx
|
|
|
f76d5d |
|
|
|
f76d5d |
/* Now %edx holds result, like this:
|
|
|
f76d5d |
@@ -113,7 +129,7 @@ __dv_decode_vlc:
|
|
|
f76d5d |
movl %edx,%ecx
|
|
|
f76d5d |
sarl $8,%ecx
|
|
|
f76d5d |
andl $0xff,%ecx
|
|
|
f76d5d |
- movl sign_mask(,%ecx,4),%ecx
|
|
|
f76d5d |
+ movl MUNG_ARR(sign_mask,%ecx,4),%ecx
|
|
|
f76d5d |
andl %ecx,%eax
|
|
|
f76d5d |
negl %eax
|
|
|
f76d5d |
sarl $31,%eax
|
|
|
f76d5d |
@@ -128,9 +144,9 @@ __dv_decode_vlc:
|
|
|
f76d5d |
xorl %eax,%edx
|
|
|
f76d5d |
subl %eax,%edx
|
|
|
f76d5d |
|
|
|
f76d5d |
- movl 12(%esp),%eax
|
|
|
f76d5d |
+ movl 16(%esp),%eax
|
|
|
f76d5d |
movl %edx,(%eax)
|
|
|
f76d5d |
-
|
|
|
f76d5d |
+ popl %ebp
|
|
|
f76d5d |
popl %ebx
|
|
|
f76d5d |
ret
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -141,14 +157,20 @@ void dv_parse_ac_coeffs_pass0(bitstream_
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
.text
|
|
|
f76d5d |
.align 4
|
|
|
f76d5d |
+.globl asm_dv_parse_ac_coeffs_pass0
|
|
|
f76d5d |
+.hidden asm_dv_parse_ac_coeffs_pass0
|
|
|
f76d5d |
+ asm_dv_parse_ac_coeffs_pass0 = dv_parse_ac_coeffs_pass0
|
|
|
f76d5d |
+
|
|
|
f76d5d |
.globl dv_parse_ac_coeffs_pass0
|
|
|
f76d5d |
.type dv_parse_ac_coeffs_pass0,@function
|
|
|
f76d5d |
dv_parse_ac_coeffs_pass0:
|
|
|
f76d5d |
pushl %ebx
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
|
|
|
f76d5d |
+ LOAD_PIC_REG(si)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
#define ARGn(N) (20+(4*(N)))(%esp)
|
|
|
f76d5d |
|
|
|
f76d5d |
/*
|
|
|
f76d5d |
@@ -160,8 +183,10 @@ dv_parse_ac_coeffs_pass0:
|
|
|
f76d5d |
ebp bl
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
movl ARGn(2),%ebp
|
|
|
f76d5d |
+#ifndef __PIC__
|
|
|
f76d5d |
movl ARGn(0),%esi
|
|
|
f76d5d |
movl bitstream_t_buf(%esi),%esi
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
movl dv_block_t_offset(%ebp),%edi
|
|
|
f76d5d |
movl dv_block_t_reorder(%ebp),%ebx
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -171,7 +196,11 @@ dv_parse_ac_coeffs_pass0:
|
|
|
f76d5d |
|
|
|
f76d5d |
movq dv_block_t_coeffs(%ebp),%mm1
|
|
|
f76d5d |
pxor %mm0,%mm0
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ pand const_f_0_0_0@GOTOFF(%esi),%mm1
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
pand const_f_0_0_0,%mm1
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
movq %mm1,dv_block_t_coeffs(%ebp)
|
|
|
f76d5d |
movq %mm0,(dv_block_t_coeffs + 8)(%ebp)
|
|
|
f76d5d |
movq %mm0,(dv_block_t_coeffs + 16)(%ebp)
|
|
|
f76d5d |
@@ -192,9 +221,17 @@ dv_parse_ac_coeffs_pass0:
|
|
|
f76d5d |
readloop:
|
|
|
f76d5d |
movl %edi,%ecx
|
|
|
f76d5d |
shrl $3,%ecx
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl ARGn(0),%eax
|
|
|
f76d5d |
+ addl bitstream_t_buf(%eax),%ecx
|
|
|
f76d5d |
+ movzbl (%ecx),%eax
|
|
|
f76d5d |
+ movzbl 1(%ecx),%edx
|
|
|
f76d5d |
+ movzbl 2(%ecx),%ecx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movzbl (%esi,%ecx,1),%eax
|
|
|
f76d5d |
movzbl 1(%esi,%ecx,1),%edx
|
|
|
f76d5d |
movzbl 2(%esi,%ecx,1),%ecx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
shll $16,%eax
|
|
|
f76d5d |
shll $8,%edx
|
|
|
f76d5d |
orl %ecx,%eax
|
|
|
f76d5d |
@@ -218,7 +255,11 @@ readloop:
|
|
|
f76d5d |
|
|
|
f76d5d |
/* Attempt to use the shortcut first. If it hits, then
|
|
|
f76d5d |
this vlc term has been decoded. */
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl dv_vlc_class1_shortcut@GOTOFF(%esi,%ecx,4),%edx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl dv_vlc_class1_shortcut(,%ecx,4),%edx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
test $0x80,%edx
|
|
|
f76d5d |
je done_decode
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -229,12 +270,19 @@ readloop:
|
|
|
f76d5d |
movl %ebx,dv_block_t_reorder(%ebp)
|
|
|
f76d5d |
|
|
|
f76d5d |
/* %eax is bits */
|
|
|
f76d5d |
-
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movsbl dv_vlc_class_lookup5@GOTOFF(%esi,%ecx),%ecx
|
|
|
f76d5d |
+
|
|
|
f76d5d |
+ movl dv_vlc_index_mask@GOTOFF(%esi,%ecx,4),%ebx
|
|
|
f76d5d |
+ movl dv_vlc_lookups@GOTOFF(%esi,%ecx,4),%edx
|
|
|
f76d5d |
+ movl dv_vlc_index_rshift@GOTOFF(%esi,%ecx,4),%ecx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movsbl dv_vlc_class_lookup5(%ecx),%ecx
|
|
|
f76d5d |
|
|
|
f76d5d |
movl dv_vlc_index_mask(,%ecx,4),%ebx
|
|
|
f76d5d |
movl dv_vlc_lookups(,%ecx,4),%edx
|
|
|
f76d5d |
movl dv_vlc_index_rshift(,%ecx,4),%ecx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
andl %eax,%ebx
|
|
|
f76d5d |
sarl %cl,%ebx
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -257,7 +305,11 @@ readloop:
|
|
|
f76d5d |
movl %edx,%ecx
|
|
|
f76d5d |
sarl $8,%ecx
|
|
|
f76d5d |
andl $0xff,%ecx
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl sign_mask@GOTOFF(%esi,%ecx,4),%ecx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl sign_mask(,%ecx,4),%ecx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
andl %ecx,%eax
|
|
|
f76d5d |
negl %eax
|
|
|
f76d5d |
sarl $31,%eax
|
|
|
f76d5d |
@@ -327,10 +379,16 @@ alldone:
|
|
|
f76d5d |
|
|
|
f76d5d |
slowpath:
|
|
|
f76d5d |
/* slow path: use dv_decode_vlc */;
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ pushl %esi
|
|
|
f76d5d |
+ leal vlc@GOTOFF(%esi),%esi
|
|
|
f76d5d |
+ xchgl %esi,(%esp) /* last parameter is &vlc */
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
pushl $vlc /* last parameter is &vlc */
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
pushl %edx /* bits_left */
|
|
|
f76d5d |
pushl %eax /* bits */
|
|
|
f76d5d |
- call dv_decode_vlc
|
|
|
f76d5d |
+ call asm_dv_decode_vlc
|
|
|
f76d5d |
addl $12,%esp
|
|
|
f76d5d |
test $0x80,%edx /* If (vlc.run < 0) break */
|
|
|
f76d5d |
jne escape
|
|
|
f76d5d |
@@ -367,6 +425,8 @@ show16:
|
|
|
f76d5d |
pushl %esi
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
|
|
|
f76d5d |
+ LOAD_PIC_REG(si)
|
|
|
f76d5d |
+
|
|
|
f76d5d |
#define ARGn(N) (20+(4*(N)))(%esp)
|
|
|
f76d5d |
|
|
|
f76d5d |
movl ARGn(1),%eax /* quality */
|
|
|
f76d5d |
@@ -374,7 +435,11 @@ dv_parse_video_segment:
|
|
|
f76d5d |
jz its_mono
|
|
|
f76d5d |
movl $6,%ebx
|
|
|
f76d5d |
its_mono:
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl %ebx,n_blocks@GOTOFF(%esi)
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl %ebx,n_blocks
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
|
|
|
f76d5d |
/*
|
|
|
f76d5d |
* ebx seg/b
|
|
|
f76d5d |
@@ -385,15 +450,22 @@ its_mono:
|
|
|
f76d5d |
* ebp bl
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
movl ARGn(0),%ebx
|
|
|
f76d5d |
+#ifndef __PIC__
|
|
|
f76d5d |
movl dv_videosegment_t_bs(%ebx),%esi
|
|
|
f76d5d |
movl bitstream_t_buf(%esi),%esi
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
leal dv_videosegment_t_mb(%ebx),%edi
|
|
|
f76d5d |
|
|
|
f76d5d |
movl $0,%eax
|
|
|
f76d5d |
movl $0,%ecx
|
|
|
f76d5d |
macloop:
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl %eax,m@GOTOFF(%esi)
|
|
|
f76d5d |
+ movl %ecx,mb_start@GOTOFF(%esi)
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl %eax,m
|
|
|
f76d5d |
movl %ecx,mb_start
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
|
|
|
f76d5d |
movl ARGn(0),%ebx
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -401,7 +473,13 @@ macloop:
|
|
|
f76d5d |
/* mb->qno = bitstream_get(bs,4); */
|
|
|
f76d5d |
movl %ecx,%edx
|
|
|
f76d5d |
shr $3,%edx
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl dv_videosegment_t_bs(%ebx),%ecx
|
|
|
f76d5d |
+ movl bitstream_t_buf(%ecx),%ecx
|
|
|
f76d5d |
+ movzbl 3(%ecx,%edx,1),%edx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movzbl 3(%esi,%edx,1),%edx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
andl $0xf,%edx
|
|
|
f76d5d |
movl %edx,dv_macroblock_t_qno(%edi)
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -412,7 +490,11 @@ macloop:
|
|
|
f76d5d |
movl %edx,dv_macroblock_t_eob_count(%edi)
|
|
|
f76d5d |
|
|
|
f76d5d |
/* mb->i = (seg->i + dv_super_map_vertical[m]) % (seg->isPAL?12:10); */
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl dv_super_map_vertical@GOTOFF(%esi,%eax,4),%edx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl dv_super_map_vertical(,%eax,4),%edx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
movl dv_videosegment_t_i(%ebx),%ecx
|
|
|
f76d5d |
addl %ecx,%edx
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -423,11 +505,20 @@ skarly:
|
|
|
f76d5d |
andl $1,%ecx
|
|
|
f76d5d |
shll $5,%ecx /* ecx = (isPAL ? 32 : 0) */
|
|
|
f76d5d |
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ leal mod_10@GOTOFF(%esi),%edx
|
|
|
f76d5d |
+ movzbl (%edx,%ecx,1),%edx /* uses mod_12 for PAL */
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movzbl mod_10(%edx,%ecx,1),%edx /* uses mod_12 for PAL */
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
movl %edx,dv_macroblock_t_i(%edi)
|
|
|
f76d5d |
|
|
|
f76d5d |
/* mb->j = dv_super_map_horizontal[m]; */
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl dv_super_map_horizontal@GOTOFF(%esi,%eax,4),%edx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl dv_super_map_horizontal(,%eax,4),%edx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
movl %edx,dv_macroblock_t_j(%edi)
|
|
|
f76d5d |
|
|
|
f76d5d |
/* mb->k = seg->k; */
|
|
|
f76d5d |
@@ -446,12 +537,28 @@ blkloop:
|
|
|
f76d5d |
+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+
|
|
|
f76d5d |
*/
|
|
|
f76d5d |
/* dc = bitstream_get(bs,9); */
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl mb_start@GOTOFF(%esi),%ecx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl mb_start,%ecx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
shr $3,%ecx
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movzbl blk_start@GOTOFF(%esi,%ebx),%edx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movzbl blk_start(%ebx),%edx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
addl %ecx,%edx
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl ARGn(0),%ecx
|
|
|
f76d5d |
+ movl dv_videosegment_t_bs(%ecx),%ecx
|
|
|
f76d5d |
+ movl bitstream_t_buf(%ecx),%ecx
|
|
|
f76d5d |
+ movzbl (%ecx,%edx,1),%eax /* hi byte */
|
|
|
f76d5d |
+ movzbl 1(%ecx,%edx,1),%ecx /* lo byte */
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movzbl (%esi,%edx,1),%eax /* hi byte */
|
|
|
f76d5d |
movzbl 1(%esi,%edx,1),%ecx /* lo byte */
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
shll $8,%eax
|
|
|
f76d5d |
orl %ecx,%eax
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -478,7 +585,11 @@ blkloop:
|
|
|
f76d5d |
|
|
|
f76d5d |
/* bl->reorder = &dv_reorder[bl->dct_mode][1]; */
|
|
|
f76d5d |
shll $6,%eax
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ leal dv_reorder@GOTOFF+1(%esi,%eax),%eax
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
addl $(dv_reorder+1),%eax
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
movl %eax,dv_block_t_reorder(%ebp)
|
|
|
f76d5d |
|
|
|
f76d5d |
/* bl->reorder_sentinel = bl->reorder + 63; */
|
|
|
f76d5d |
@@ -486,13 +597,22 @@ blkloop:
|
|
|
f76d5d |
movl %eax,dv_block_t_reorder_sentinel(%ebp)
|
|
|
f76d5d |
|
|
|
f76d5d |
/* bl->offset= mb_start + dv_parse_bit_start[b]; */
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl mb_start@GOTOFF(%esi),%ecx
|
|
|
f76d5d |
+ movl dv_parse_bit_start@GOTOFF(%esi,%ebx,4),%eax
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl mb_start,%ecx
|
|
|
f76d5d |
movl dv_parse_bit_start(,%ebx,4),%eax
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
addl %ecx,%eax
|
|
|
f76d5d |
movl %eax,dv_block_t_offset(%ebp)
|
|
|
f76d5d |
|
|
|
f76d5d |
/* bl->end= mb_start + dv_parse_bit_end[b]; */
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl dv_parse_bit_end@GOTOFF(%esi,%ebx,4),%eax
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl dv_parse_bit_end(,%ebx,4),%eax
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
addl %ecx,%eax
|
|
|
f76d5d |
movl %eax,dv_block_t_end(%ebp)
|
|
|
f76d5d |
|
|
|
f76d5d |
@@ -504,7 +624,11 @@ blkloop:
|
|
|
f76d5d |
/* no AC pass. Just zero out the remaining coeffs */
|
|
|
f76d5d |
movq dv_block_t_coeffs(%ebp),%mm1
|
|
|
f76d5d |
pxor %mm0,%mm0
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ pand const_f_0_0_0@GOTOFF(%esi),%mm1
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
pand const_f_0_0_0,%mm1
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
movq %mm1,dv_block_t_coeffs(%ebp)
|
|
|
f76d5d |
movq %mm0,(dv_block_t_coeffs + 8)(%ebp)
|
|
|
f76d5d |
movq %mm0,(dv_block_t_coeffs + 16)(%ebp)
|
|
|
f76d5d |
@@ -529,18 +653,27 @@ do_ac_pass:
|
|
|
f76d5d |
pushl %ebp
|
|
|
f76d5d |
pushl %edi
|
|
|
f76d5d |
pushl %eax
|
|
|
f76d5d |
- call dv_parse_ac_coeffs_pass0
|
|
|
f76d5d |
+ call asm_dv_parse_ac_coeffs_pass0
|
|
|
f76d5d |
addl $12,%esp
|
|
|
f76d5d |
done_ac:
|
|
|
f76d5d |
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl n_blocks@GOTOFF(%esi),%eax
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl n_blocks,%eax
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
addl $dv_block_t_size,%ebp
|
|
|
f76d5d |
incl %ebx
|
|
|
f76d5d |
cmpl %eax,%ebx
|
|
|
f76d5d |
jnz blkloop
|
|
|
f76d5d |
|
|
|
f76d5d |
+#ifdef __PIC__
|
|
|
f76d5d |
+ movl m@GOTOFF(%esi),%eax
|
|
|
f76d5d |
+ movl mb_start@GOTOFF(%esi),%ecx
|
|
|
f76d5d |
+#else
|
|
|
f76d5d |
movl m,%eax
|
|
|
f76d5d |
movl mb_start,%ecx
|
|
|
f76d5d |
+#endif
|
|
|
f76d5d |
addl $(8 * 80),%ecx
|
|
|
f76d5d |
addl $dv_macroblock_t_size,%edi
|
|
|
f76d5d |
incl %eax
|
|
|
f76d5d |
@@ -558,7 +691,7 @@ done_ac:
|
|
|
f76d5d |
|
|
|
f76d5d |
andl $DV_QUALITY_AC_MASK,%eax
|
|
|
f76d5d |
cmpl $DV_QUALITY_AC_2,%eax
|
|
|
f76d5d |
- jz dv_parse_ac_coeffs
|
|
|
f76d5d |
+ jz asm_dv_parse_ac_coeffs
|
|
|
f76d5d |
movl $0,%eax
|
|
|
f76d5d |
ret
|
|
|
f76d5d |
|