[PATCH] drm/amdkfd: Use S_ENDPGM_SAVED in trap handler

[Jay Cornwall]

On 1/15/2024 13:07, Jay Cornwall wrote:
> This instruction has no functional difference to S_ENDPGM
> but allows performance counters to track save events correctly.
> 
> Signed-off-by: Jay Cornwall <jay.cornwall at amd.com>
> ---
>  drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h     | 14 +++++++-------
>  .../gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm |  2 +-
>  .../gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm  |  2 +-
>  3 files changed, 9 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
> index df75863393fc..d1caaf0e6a7c 100644
> --- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
> +++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
> @@ -674,7 +674,7 @@ static const uint32_t cwsr_trap_gfx9_hex[] = {
>  	0x86ea6a6a, 0x8f6e837a,
>  	0xb96ee0c2, 0xbf800002,
>  	0xb97a0002, 0xbf8a0000,
> -	0xbe801f6c, 0xbf810000,
> +	0xbe801f6c, 0xbf9b0000,
>  };
>  
>  static const uint32_t cwsr_trap_nv1x_hex[] = {
> @@ -1091,7 +1091,7 @@ static const uint32_t cwsr_trap_nv1x_hex[] = {
>  	0xb9eef807, 0x876dff6d,
>  	0x0000ffff, 0x87fe7e7e,
>  	0x87ea6a6a, 0xb9faf802,
> -	0xbe80226c, 0xbf810000,
> +	0xbe80226c, 0xbf9b0000,
>  	0xbf9f0000, 0xbf9f0000,
>  	0xbf9f0000, 0xbf9f0000,
>  	0xbf9f0000, 0x00000000,
> @@ -1574,7 +1574,7 @@ static const uint32_t cwsr_trap_arcturus_hex[] = {
>  	0x86ea6a6a, 0x8f6e837a,
>  	0xb96ee0c2, 0xbf800002,
>  	0xb97a0002, 0xbf8a0000,
> -	0xbe801f6c, 0xbf810000,
> +	0xbe801f6c, 0xbf9b0000,
>  };
>  
>  static const uint32_t cwsr_trap_aldebaran_hex[] = {
> @@ -2065,7 +2065,7 @@ static const uint32_t cwsr_trap_aldebaran_hex[] = {
>  	0x86ea6a6a, 0x8f6e837a,
>  	0xb96ee0c2, 0xbf800002,
>  	0xb97a0002, 0xbf8a0000,
> -	0xbe801f6c, 0xbf810000,
> +	0xbe801f6c, 0xbf9b0000,
>  };
>  
>  static const uint32_t cwsr_trap_gfx10_hex[] = {
> @@ -2500,7 +2500,7 @@ static const uint32_t cwsr_trap_gfx10_hex[] = {
>  	0x876dff6d, 0x0000ffff,
>  	0x87fe7e7e, 0x87ea6a6a,
>  	0xb9faf802, 0xbe80226c,
> -	0xbf810000, 0xbf9f0000,
> +	0xbf9b0000, 0xbf9f0000,
>  	0xbf9f0000, 0xbf9f0000,
>  	0xbf9f0000, 0xbf9f0000,
>  };
> @@ -2944,7 +2944,7 @@ static const uint32_t cwsr_trap_gfx11_hex[] = {
>  	0xb8eef802, 0xbf0d866e,
>  	0xbfa20002, 0xb97af802,
>  	0xbe80486c, 0xb97af802,
> -	0xbe804a6c, 0xbfb00000,
> +	0xbe804a6c, 0xbfb10000,
>  	0xbf9f0000, 0xbf9f0000,
>  	0xbf9f0000, 0xbf9f0000,
>  	0xbf9f0000, 0x00000000,
> @@ -3436,5 +3436,5 @@ static const uint32_t cwsr_trap_gfx9_4_3_hex[] = {
>  	0x86ea6a6a, 0x8f6e837a,
>  	0xb96ee0c2, 0xbf800002,
>  	0xb97a0002, 0xbf8a0000,
> -	0xbe801f6c, 0xbf810000,
> +	0xbe801f6c, 0xbf9b0000,
>  };
> diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
> index e0140df0b0ec..71b3dc0c7363 100644
> --- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
> +++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
> @@ -1104,7 +1104,7 @@ L_RETURN_WITHOUT_PRIV:
>  	s_rfe_b64	s_restore_pc_lo						//Return to the main shader program and resume execution
>  
>  L_END_PGM:
> -	s_endpgm
> +	s_endpgm_saved
>  end
>  
>  function write_hwreg_to_mem(s, s_rsrc, s_mem_offset)
> diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
> index e506411ad28a..bb26338204f4 100644
> --- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
> +++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
> @@ -921,7 +921,7 @@ L_RESTORE:
>  /*			the END						  */
>  /**************************************************************************/
>  L_END_PGM:
> -    s_endpgm
> +    s_endpgm_saved
>  
>  end
>  

Ping. Patch has been tested and verified, just looking for an Ack.