2
* All Video Processing kernels
3
* Copyright © <2010>, Intel Corporation.
5
* This program is licensed under the terms and conditions of the
6
* Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
7
* http://www.opensource.org/licenses/eclipse-1.0.php.
11
//---------- PA_Scaling.asm ----------
12
#include "Scaling.inc"
14
// Build 16 elements ramp in float32 and normalized it
15
// mov (8) SAMPLER_RAMP(0)<1> 0x76543210:v
16
// add (8) SAMPLER_RAMP(1)<1> SAMPLER_RAMP(0) 8.0:f
17
mov (4) SAMPLER_RAMP(0)<1> 0x48403000:vf //3, 2, 1, 0 in float vector
18
mov (4) SAMPLER_RAMP(0,4)<1> 0x5C585450:vf //7, 6, 5, 4 in float vector
19
add (8) SAMPLER_RAMP(1)<1> SAMPLER_RAMP(0) 8.0:f
21
//Module: PrepareScaleCoord.asm
23
// Setup for sampler msg hdr
24
mov (2) rMSGSRC.0<1>:ud 0:ud { NoDDClr } // Unused fields
25
mov (1) rMSGSRC.2<1>:ud 0:ud { NoDDChk } // Write and offset
27
// Calculate 16 v based on the step Y and vertical origin
28
mov (16) mfMSGPAYLOAD(2)<1> fSRC_VID_V_ORI<0;1,0>:f
29
mov (16) SCALE_COORD_Y<1>:f fSRC_VID_V_ORI<0;1,0>:f
31
// Calculate 16 u based on the step X and hori origin
32
// line (16) mfMSGPAYLOAD(0)<1> SCALE_STEP_X<0;1,0>:f SAMPLER_RAMP(0) // Assign to mrf directly
33
mov (16) acc0:f fSRC_VID_H_ORI<0;1,0>:f { Compr }
34
mac (16) mfMSGPAYLOAD(0)<1> fVIDEO_STEP_X<0;1,0>:f SAMPLER_RAMP(0) { Compr }
36
//Setup the constants for line instruction
37
mov (1) SCALE_LINE_P255<1>:f 255.0:f { NoDDClr } //{ NoDDClr, NoDDChk }
38
mov (1) SCALE_LINE_P0_5<1>:f 0.5:f { NoDDChk }
40
//------------------------------------------------------------------------------
42
$for (0; <nY_NUM_OF_ROWS; 1) {
44
// Read 16 sampled pixels and store them in float32 in 8 GRFs in the order of BGRA (VYUA).
45
mov (8) MSGHDR_SCALE.0:ud rMSGSRC.0<8;8,1>:ud // Copy msg header and payload mirrors to MRFs
46
send (16) SCALE_RESPONSE_YW(0)<1> MSGHDR_SCALE udDUMMY_NULL nSMPL_ENGINE SMPLR_MSG_DSC+nSI_SRC_SIMD16_YUV+nBI_CURRENT_SRC_YUV
48
// Calculate 16 v for next line
49
add (16) mfMSGPAYLOAD(2)<1> SCALE_COORD_Y<8;8,1>:f fVIDEO_STEP_Y<0;1,0>:f // Assign to mrf directly
50
add (16) SCALE_COORD_Y<1>:f SCALE_COORD_Y<8;8,1>:f fVIDEO_STEP_Y<0;1,0>:f // Assign to mrf directly
52
// Scale back to [0, 255], convert f to ud
53
line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(0) { Compr } // Process B, V
54
mov (16) SCALE_RESPONSE_YD(0)<1> acc0:f { Compr }
56
line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(2) { Compr } // Process B, V
57
mov (16) SCALE_RESPONSE_YD(2)<1> acc0:f { Compr }
59
line (16) acc0:f SCALE_LINE_P255<0;1,0>:f SCALE_RESPONSE_YF(4) { Compr } // Process B, V
60
mov (16) SCALE_RESPONSE_YD(4)<1> acc0:f { Compr }
62
mov (16) DEST_V(%1)<1> SCALE_RESPONSE_YB(0) //possible error due to truncation - vK
63
mov (16) DEST_Y(%1)<1> SCALE_RESPONSE_YB(2) //possible error due to truncation - vK
64
mov (16) DEST_U(%1)<1> SCALE_RESPONSE_YB(4) //possible error due to truncation - vK
68
#define nSRC_REGION nREGION_1
70
//------------------------------------------------------------------------------