mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-11-27 12:11:13 +00:00
394 lines
9.7 KiB
ArmAsm
394 lines
9.7 KiB
ArmAsm
; PowerPC optimized drawing methods for Goom
|
|
; © 2003 Guillaume Borios
|
|
; This library is free software; you can redistribute it and/or
|
|
; modify it under the terms of the GNU Library General Public
|
|
; License as published by the Free Software Foundation; either
|
|
; version 2 of the License, or (at your option) any later version.
|
|
;
|
|
; This library is distributed in the hope that it will be useful,
|
|
; but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
; Library General Public License for more details.
|
|
;
|
|
; You should have received a copy of the GNU Library General Public
|
|
; License along with this library; if not, write to the
|
|
; Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
|
|
; Boston, MA 02110-1301, USA.
|
|
|
|
; 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
|
|
|