mirror of
https://github.com/tiagovignatti/intel-gpu-tools.git
synced 2025-06-10 01:16:18 +00:00
ed816d560c
The gpgpu fill utility functions are used in separate test so it's logical to keep them in separate file. This is similar to what media spin test did in the past. Functionally only gpgpu kernel changed. Send instruction payload size was reduced. Since offset is incremented by 0x10 bytes there is no point in using larger writes. Cc: Thomas Wood <thomas.wood@intel.com> Signed-off-by: Dominik Zeromski <dominik.zeromski@intel.com> [Thomas: Fix typo of gpgpu_fill.h in Makefile.sources] Signed-off-by: Thomas Wood <thomas.wood@intel.com>
40 lines
1.2 KiB
Plaintext
40 lines
1.2 KiB
Plaintext
/*
|
|
* Registers
|
|
* g0 -- header
|
|
* g1 -- constant
|
|
* g2 -- calculate X/Y offset
|
|
* g4-g12 payload for write message
|
|
*/
|
|
define(`ORIG', `g2.0<2,2,1>UD')
|
|
define(`ORIG_X', `g2.0<1>UD')
|
|
define(`ORIG_Y', `g2.4<1>UD')
|
|
define(`COLOR', `g1.0')
|
|
define(`COLORUB', `COLOR<0,1,0>UB')
|
|
define(`COLORUD', `COLOR<0,1,0>UD')
|
|
define(`X', `g0.4<0,1,0>UD')
|
|
define(`Y', `g0.24<0,1,0>UD')
|
|
|
|
mov(4) COLOR<1>UB COLORUB {align1};
|
|
|
|
/* WRITE */
|
|
/* count thread group ID for X/Y offset */
|
|
mul(1) ORIG_X X 0x10UD {align1};
|
|
mov(1) ORIG_Y Y {align1};
|
|
mov(8) g4.0<1>UD g0.0<8,8,1>UD {align1};
|
|
mov(2) g4.0<1>UD ORIG {align1};
|
|
/* Normal mode: for block height 1 row and block width 16 bytes */
|
|
mov(1) g4.8<1>UD 0x0000000fUD {align1};
|
|
|
|
mov(16) g5.0<1>UD COLORUD {align1 compr};
|
|
|
|
/*
|
|
* write(0, 0, 10, 12)
|
|
* 10: media_block_write
|
|
* 12: data cache data port 1
|
|
*/
|
|
send(16) 4 acc0<1>UW null write(0, 0, 10, 12) mlen 3 rlen 0 {align1};
|
|
|
|
/* EOT */
|
|
mov(8) g112.0<1>UD g0.0<8,8,1>UD {align1};
|
|
send(16) 112 null<1>UW null thread_spawner(0, 0, 1) mlen 1 rlen 0 {align1 EOT};
|