gstreamer/gst/goom/ppc_drawings.s
Bastien Nocera a7bc7485b1 configure.ac: Add checks for Flex/Yacc/Bison and other furry animals, for the new goom 2k4 based plugin
Original commit message from CVS:
2008-02-23  Bastien Nocera  <hadess@hadess.net>

* configure.ac: Add checks for Flex/Yacc/Bison and other
furry animals, for the new goom 2k4 based plugin

* gst/goom/*: Update to use goom 2k4, uses liboil to detect
CPU optimisations (not working yet), move the old plugin to...

* gst/goom2k1/*: ... here, in case somebody is sick enough

Fixes #515073
2008-02-23 01:51:37 +00:00

381 lines
9.1 KiB
ArmAsm

; PowerPC optimized drawing methods for Goom
; © 2003 Guillaume Borios
; This Source Code is released under the terms of the General Public License
; Change log :
; 30 May 2003 : File creation
; Section definition : We use a read only code section for the whole file
.section __TEXT,__text,regular,pure_instructions
; --------------------------------------------------------------------------------------
; Single 32b pixel drawing macros
; Usage :
; DRAWMETHOD_XXXX_MACRO *pixelIN, *pixelOUT, COLOR, WR1, WR2, WR3, WR4
; Only the work registers (WR) can be touched by the macros
;
; Available methods :
; DRAWMETHOD_DFLT_MACRO : Default drawing method (Actually OVRW)
; DRAWMETHOD_PLUS_MACRO : RVB Saturated per channel addition (SLOWEST)
; DRAWMETHOD_HALF_MACRO : 50% Transparency color drawing
; DRAWMETHOD_OVRW_MACRO : Direct COLOR drawing (FASTEST)
; DRAWMETHOD_B_OR_MACRO : Bitwise OR
; DRAWMETHOD_BAND_MACRO : Bitwise AND
; DRAWMETHOD_BXOR_MACRO : Bitwise XOR
; DRAWMETHOD_BNOT_MACRO : Bitwise NOT
; --------------------------------------------------------------------------------------
.macro DRAWMETHOD_OVRW_MACRO
stw $2,0($1) ;; *$1 <- $2
.endmacro
.macro DRAWMETHOD_B_OR_MACRO
lwz $3,0($0) ;; $3 <- *$0
or $3,$3,$2 ;; $3 <- $3 | $2
stw $3,0($1) ;; *$1 <- $3
.endmacro
.macro DRAWMETHOD_BAND_MACRO
lwz $3,0($0) ;; $3 <- *$0
and $3,$3,$2 ;; $3 <- $3 & $2
stw $3,0($1) ;; *$1 <- $3
.endmacro
.macro DRAWMETHOD_BXOR_MACRO
lwz $3,0($0) ;; $3 <- *$0
xor $3,$3,$2 ;; $3 <- $3 ^ $2
stw $3,0($1) ;; *$1 <- $3
.endmacro
.macro DRAWMETHOD_BNOT_MACRO
lwz $3,0($0) ;; $3 <- *$0
nand $3,$3,$3 ;; $3 <- ~$3
stw $3,0($1) ;; *$1 <- $3
.endmacro
.macro DRAWMETHOD_PLUS_MACRO
lwz $4,0($0) ;; $4 <- *$0
andi. $3,$4,0xFF00 ;; $3 <- $4 & 0x0000FF00
andi. $5,$2,0xFF00 ;; $5 <- $2 & 0x0000FF00
add $3,$3,$5 ;; $3 <- $3 + $5
rlwinm $5,$3,15,0,0 ;; $5 <- 0 | ($3[15] << 15)
srawi $5,$5,23 ;; $5 <- $5 >> 23 (algebraic for sign extension)
or $3,$3,$5 ;; $3 <- $3 | $5
lis $5,0xFF ;; $5 <- 0x00FF00FF
addi $5,$5,0xFF
and $4,$4,$5 ;; $4 <- $4 & $5
and $6,$2,$5 ;; $6 <- $2 & $5
add $4,$4,$6 ;; $4 <- $4 + $6
rlwinm $6,$4,7,0,0 ;; $6 <- 0 | ($4[7] << 7)
srawi $6,$6,15 ;; $6 <- $6 >> 15 (algebraic for sign extension)
rlwinm $5,$4,23,0,0 ;; $5 <- 0 | ($4[23] << 23)
srawi $5,$5,31 ;; $5 <- $5 >> 31 (algebraic for sign extension)
rlwimi $6,$5,0,24,31 ;; $6[24..31] <- $5[24..31]
or $4,$4,$6 ;; $4 <- $4 | $6
rlwimi $4,$3,0,16,23 ;; $4[16..23] <- $3[16..23]
stw $4,0($1) ;; *$1 <- $4
.endmacro
.macro DRAWMETHOD_HALF_MACRO
lwz $4,0($0) ;; $4 <- *$0
andi. $3,$4,0xFF00 ;; $3 <- $4 & 0x0000FF00
andi. $5,$2,0xFF00 ;; $5 <- $2 & 0x0000FF00
add $3,$3,$5 ;; $3 <- $3 + $5
lis $5,0xFF ;; $5 <- 0x00FF00FF
addi $5,$5,0xFF
and $4,$4,$5 ;; $4 <- $4 & $5
and $5,$2,$5 ;; $5 <- $2 & $5
add $4,$4,$5 ;; $4 <- $4 + $5
srwi $4,$4,1 ;; $4 <- $4 >> 1
rlwimi $4,$3,31,16,23 ;; $4[16..23] <- $3[15..22]
stw $4,0($1) ;; *$1 <- $4
.endmacro
.macro DRAWMETHOD_DFLT_MACRO
DRAWMETHOD_PLUS_MACRO
.endmacro
; --------------------------------------------------------------------------------------
; **************************************************************************************
; void DRAWMETHOD_PLUS_PPC(unsigned int * buf, unsigned int _col);
; void DRAWMETHOD_PLUS_2_PPC(unsigned * in, unsigned int * out, unsigned int _col);
; **************************************************************************************
.globl _DRAWMETHOD_PLUS_2_PPC
.align 3
_DRAWMETHOD_PLUS_2_PPC:
DRAWMETHOD_PLUS_MACRO r3,r4,r5,r6,r7,r8,r9
blr ;; return
.globl _DRAWMETHOD_PLUS_PPC
.align 3
_DRAWMETHOD_PLUS_PPC:
DRAWMETHOD_PLUS_MACRO r3,r3,r4,r5,r6,r7,r9
blr ;; return
; **************************************************************************************
; void DRAWMETHOD_HALF_PPC(unsigned int * buf, unsigned int _col);
; void DRAWMETHOD_HALF_2_PPC(unsigned * in, unsigned int * out, unsigned int _col);
; **************************************************************************************
.globl _DRAWMETHOD_HALF_2_PPC
.align 3
_DRAWMETHOD_HALF_2_PPC:
DRAWMETHOD_HALF_MACRO r3,r4,r5,r6,r7,r8
blr ;; return
.globl _DRAWMETHOD_HALF_PPC
.align 3
_DRAWMETHOD_HALF_PPC:
DRAWMETHOD_HALF_MACRO r3,r3,r4,r5,r6,r7
blr ;; return
; **************************************************************************************
; void DRAW_LINE_PPC(unsigned int *data, int x1, int y1, int x2, int y2, unsigned int col,
; unsigned int screenx, unsigned int screeny)
; **************************************************************************************
.globl _DRAW_LINE_PPC
.align 3
_DRAW_LINE_PPC:
;; NOT IMPLEMENTED YET
blr ;; return
; **************************************************************************************
; void _ppc_brightness(Pixel * src, Pixel * dest, unsigned int size, unsigned int coeff)
; **************************************************************************************
.const
.align 4
vectorZERO:
.long 0,0,0,0
.long 0x10101000, 0x10101001, 0x10101002, 0x10101003
.long 0x10101004, 0x10101005, 0x10101006, 0x10101007
.long 0x10101008, 0x10101009, 0x1010100A, 0x1010100B
.long 0x1010100C, 0x1010100D, 0x1010100E, 0x1010100F
.section __TEXT,__text,regular,pure_instructions
.globl _ppc_brightness_G4
.align 3
_ppc_brightness_G4:
;; PowerPC Altivec code
srwi r5,r5,2
mtctr r5
;;vrsave
mfspr r11,256
lis r12,0xCFFC
mtspr 256,r12
mflr r0
bcl 20,31,"L00000000001$pb"
"L00000000001$pb":
mflr r10
mtlr r0
addis r9,r10,ha16(vectorZERO-"L00000000001$pb")
addi r9,r9,lo16(vectorZERO-"L00000000001$pb")
vxor v0,v0,v0 ;; V0 = NULL vector
addi r9,r9,16
lvx v10,0,r9
addi r9,r9,16
lvx v11,0,r9
addi r9,r9,16
lvx v12,0,r9
addi r9,r9,16
lvx v13,0,r9
addis r9,r10,ha16(vectortmpwork-"L00000000001$pb")
addi r9,r9,lo16(vectortmpwork-"L00000000001$pb")
stw r6,0(r9)
li r6,8
stw r6,4(r9)
lvx v9,0,r9
li r9,128
vspltw v8,v9,0
vspltw v9,v9,1
;; elt counter
li r9,0
lis r7,0x0F01
b L7
.align 4
L7:
lvx v1,r9,r3
vperm v4,v1,v0,v10
;*********************
add r10,r9,r3
;*********************
vperm v5,v1,v0,v11
vperm v6,v1,v0,v12
vperm v7,v1,v0,v13
vmulouh v4,v4,v8
;*********************
dst r10,r7,3
;*********************
vmulouh v5,v5,v8
vmulouh v6,v6,v8
vmulouh v7,v7,v8
vsrw v4,v4,v9
vsrw v5,v5,v9
vsrw v6,v6,v9
vsrw v7,v7,v9
vpkuwus v4,v4,v5
vpkuwus v6,v6,v7
vpkuhus v1,v4,v6
stvx v1,r9,r4
addi r9,r9,16
bdnz L7
mtspr 256,r11
blr
.globl _ppc_brightness_G5
.align 3
_ppc_brightness_G5:
;; PowerPC Altivec G5 code
srwi r5,r5,2
mtctr r5
;;vrsave
mfspr r11,256
lis r12,0xCFFC
mtspr 256,r12
mflr r0
bcl 20,31,"L00000000002$pb"
"L00000000002$pb":
mflr r10
mtlr r0
addis r9,r10,ha16(vectorZERO-"L00000000002$pb")
addi r9,r9,lo16(vectorZERO-"L00000000002$pb")
vxor v0,v0,v0 ;; V0 = NULL vector
addi r9,r9,16
lvx v10,0,r9
addi r9,r9,16
lvx v11,0,r9
addi r9,r9,16
lvx v12,0,r9
addi r9,r9,16
lvx v13,0,r9
addis r9,r10,ha16(vectortmpwork-"L00000000002$pb")
addi r9,r9,lo16(vectortmpwork-"L00000000002$pb")
stw r6,0(r9)
li r6,8
stw r6,4(r9)
lvx v9,0,r9
li r9,128
vspltw v8,v9,0
vspltw v9,v9,1
;; elt counter
li r9,0
lis r7,0x0F01
b L6
.align 4
L6:
lvx v1,r9,r3
vperm v4,v1,v0,v10
;*********************
add r10,r9,r3
;*********************
vperm v5,v1,v0,v11
vperm v6,v1,v0,v12
vperm v7,v1,v0,v13
vmulouh v4,v4,v8
vmulouh v5,v5,v8
vmulouh v6,v6,v8
vmulouh v7,v7,v8
vsrw v4,v4,v9
vsrw v5,v5,v9
vsrw v6,v6,v9
vsrw v7,v7,v9
vpkuwus v4,v4,v5
vpkuwus v6,v6,v7
vpkuhus v1,v4,v6
stvx v1,r9,r4
addi r9,r9,16
bdnz L6
mtspr 256,r11
blr
.globl _ppc_brightness_generic
.align 3
_ppc_brightness_generic:
lis r12,0x00FF
ori r12,r12,0x00FF
subi r3,r3,4
subi r4,r4,4
mtctr r5
b L1
.align 4
L1:
lwzu r7,4(r3)
rlwinm r8,r7,16,24,31
rlwinm r9,r7,24,24,31
mullw r8,r8,r6
rlwinm r10,r7,0,24,31
mullw r9,r9,r6
srwi r8,r8,8
mullw r10,r10,r6
srwi r9,r9,8
rlwinm. r11,r8,0,0,23
beq L2
li r8,0xFF
L2:
srwi r10,r10,8
rlwinm. r11,r9,0,0,23
beq L3
li r9,0xFF
L3:
rlwinm r7,r8,16,8,15
rlwinm. r11,r10,0,0,23
beq L4
li r10,0xFF
L4:
rlwimi r7,r9,8,16,23
rlwimi r7,r10,0,24,31
stwu r7,4(r4)
bdnz L1
blr
.static_data
.align 4
vectortmpwork:
.long 0,0,0,0