summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm')
-rw-r--r--drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm63
1 files changed, 33 insertions, 30 deletions
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 0bb9c577b3a2..6bae2e022c6e 100644
--- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
+++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
@@ -150,10 +150,10 @@ var S_SAVE_SPI_INIT_MTYPE_SHIFT = 28
var S_SAVE_SPI_INIT_FIRST_WAVE_MASK = 0x04000000 //bit[26]: FirstWaveInTG
var S_SAVE_SPI_INIT_FIRST_WAVE_SHIFT = 26
-var S_SAVE_PC_HI_RCNT_SHIFT = 28 //FIXME check with Brian to ensure all fields other than PC[47:0] can be used
-var S_SAVE_PC_HI_RCNT_MASK = 0xF0000000 //FIXME
-var S_SAVE_PC_HI_FIRST_REPLAY_SHIFT = 27 //FIXME
-var S_SAVE_PC_HI_FIRST_REPLAY_MASK = 0x08000000 //FIXME
+var S_SAVE_PC_HI_RCNT_SHIFT = 27 //FIXME check with Brian to ensure all fields other than PC[47:0] can be used
+var S_SAVE_PC_HI_RCNT_MASK = 0xF8000000 //FIXME
+var S_SAVE_PC_HI_FIRST_REPLAY_SHIFT = 26 //FIXME
+var S_SAVE_PC_HI_FIRST_REPLAY_MASK = 0x04000000 //FIXME
var s_save_spi_init_lo = exec_lo
var s_save_spi_init_hi = exec_hi
@@ -162,8 +162,8 @@ var s_save_pc_lo = ttmp0 //{TTMP1, TTMP0} = {3'h0,pc_rewind[3:0], HT[0],tra
var s_save_pc_hi = ttmp1
var s_save_exec_lo = ttmp2
var s_save_exec_hi = ttmp3
-var s_save_tmp = ttmp4
-var s_save_trapsts = ttmp5 //not really used until the end of the SAVE routine
+var s_save_tmp = ttmp14
+var s_save_trapsts = ttmp15 //not really used until the end of the SAVE routine
var s_save_xnack_mask_lo = ttmp6
var s_save_xnack_mask_hi = ttmp7
var s_save_buf_rsrc0 = ttmp8
@@ -171,9 +171,9 @@ var s_save_buf_rsrc1 = ttmp9
var s_save_buf_rsrc2 = ttmp10
var s_save_buf_rsrc3 = ttmp11
var s_save_status = ttmp12
-var s_save_mem_offset = ttmp14
+var s_save_mem_offset = ttmp4
var s_save_alloc_size = s_save_trapsts //conflict
-var s_save_m0 = ttmp15
+var s_save_m0 = ttmp5
var s_save_ttmps_lo = s_save_tmp //no conflict
var s_save_ttmps_hi = s_save_trapsts //no conflict
@@ -207,10 +207,10 @@ var s_restore_mode = ttmp7
var s_restore_pc_lo = ttmp0
var s_restore_pc_hi = ttmp1
-var s_restore_exec_lo = ttmp14
-var s_restore_exec_hi = ttmp15
-var s_restore_status = ttmp4
-var s_restore_trapsts = ttmp5
+var s_restore_exec_lo = ttmp4
+var s_restore_exec_hi = ttmp5
+var s_restore_status = ttmp14
+var s_restore_trapsts = ttmp15
var s_restore_xnack_mask_lo = xnack_mask_lo
var s_restore_xnack_mask_hi = xnack_mask_hi
var s_restore_buf_rsrc0 = ttmp8
@@ -266,10 +266,16 @@ if (!EMU_RUN_HACK)
L_HALT_WAVE:
// If STATUS.HALT is set then this fault must come from SQC instruction fetch.
- // We cannot prevent further faults so just terminate the wavefront.
+ // We cannot prevent further faults. Spin wait until context saved.
s_and_b32 ttmp2, s_save_status, SQ_WAVE_STATUS_HALT_MASK
s_cbranch_scc0 L_NOT_ALREADY_HALTED
- s_endpgm
+
+L_WAIT_CTX_SAVE:
+ s_sleep 0x10
+ s_getreg_b32 ttmp2, hwreg(HW_REG_TRAPSTS)
+ s_and_b32 ttmp2, ttmp2, SQ_WAVE_TRAPSTS_SAVECTX_MASK
+ s_cbranch_scc0 L_WAIT_CTX_SAVE
+
L_NOT_ALREADY_HALTED:
s_or_b32 s_save_status, s_save_status, SQ_WAVE_STATUS_HALT_MASK
@@ -293,12 +299,12 @@ L_FETCH_2ND_TRAP:
// Read second-level TBA/TMA from first-level TMA and jump if available.
// ttmp[2:5] and ttmp12 can be used (others hold SPI-initialized debug data)
// ttmp12 holds SQ_WAVE_STATUS
- s_getreg_b32 ttmp4, hwreg(HW_REG_SQ_SHADER_TMA_LO)
- s_getreg_b32 ttmp5, hwreg(HW_REG_SQ_SHADER_TMA_HI)
- s_lshl_b64 [ttmp4, ttmp5], [ttmp4, ttmp5], 0x8
- s_load_dwordx2 [ttmp2, ttmp3], [ttmp4, ttmp5], 0x0 glc:1 // second-level TBA
+ s_getreg_b32 ttmp14, hwreg(HW_REG_SQ_SHADER_TMA_LO)
+ s_getreg_b32 ttmp15, hwreg(HW_REG_SQ_SHADER_TMA_HI)
+ s_lshl_b64 [ttmp14, ttmp15], [ttmp14, ttmp15], 0x8
+ s_load_dwordx2 [ttmp2, ttmp3], [ttmp14, ttmp15], 0x0 glc:1 // second-level TBA
s_waitcnt lgkmcnt(0)
- s_load_dwordx2 [ttmp4, ttmp5], [ttmp4, ttmp5], 0x8 glc:1 // second-level TMA
+ s_load_dwordx2 [ttmp14, ttmp15], [ttmp14, ttmp15], 0x8 glc:1 // second-level TMA
s_waitcnt lgkmcnt(0)
s_and_b64 [ttmp2, ttmp3], [ttmp2, ttmp3], [ttmp2, ttmp3]
s_cbranch_scc0 L_NO_NEXT_TRAP // second-level trap handler not been set
@@ -405,7 +411,7 @@ end
else
end
- // Save trap temporaries 6-11, 13-15 initialized by SPI debug dispatch logic
+ // Save trap temporaries 4-11, 13 initialized by SPI debug dispatch logic
// ttmp SR memory offset : size(VGPR)+size(SGPR)+0x40
get_vgpr_size_bytes(s_save_ttmps_lo)
get_sgpr_size_bytes(s_save_ttmps_hi)
@@ -413,13 +419,11 @@ end
s_add_u32 s_save_ttmps_lo, s_save_ttmps_lo, s_save_spi_init_lo
s_addc_u32 s_save_ttmps_hi, s_save_spi_init_hi, 0x0
s_and_b32 s_save_ttmps_hi, s_save_ttmps_hi, 0xFFFF
- s_store_dwordx2 [ttmp6, ttmp7], [s_save_ttmps_lo, s_save_ttmps_hi], 0x40 glc:1
- ack_sqc_store_workaround()
- s_store_dwordx4 [ttmp8, ttmp9, ttmp10, ttmp11], [s_save_ttmps_lo, s_save_ttmps_hi], 0x48 glc:1
+ s_store_dwordx4 [ttmp4, ttmp5, ttmp6, ttmp7], [s_save_ttmps_lo, s_save_ttmps_hi], 0x50 glc:1
ack_sqc_store_workaround()
- s_store_dword ttmp13, [s_save_ttmps_lo, s_save_ttmps_hi], 0x58 glc:1
+ s_store_dwordx4 [ttmp8, ttmp9, ttmp10, ttmp11], [s_save_ttmps_lo, s_save_ttmps_hi], 0x60 glc:1
ack_sqc_store_workaround()
- s_store_dwordx2 [ttmp14, ttmp15], [s_save_ttmps_lo, s_save_ttmps_hi], 0x5C glc:1
+ s_store_dword ttmp13, [s_save_ttmps_lo, s_save_ttmps_hi], 0x74 glc:1
ack_sqc_store_workaround()
/* setup Resource Contants */
@@ -1093,7 +1097,7 @@ end
//s_setreg_b32 hwreg(HW_REG_TRAPSTS), s_restore_trapsts //don't overwrite SAVECTX bit as it may be set through external SAVECTX during restore
s_setreg_b32 hwreg(HW_REG_MODE), s_restore_mode
- // Restore trap temporaries 6-11, 13-15 initialized by SPI debug dispatch logic
+ // Restore trap temporaries 4-11, 13 initialized by SPI debug dispatch logic
// ttmp SR memory offset : size(VGPR)+size(SGPR)+0x40
get_vgpr_size_bytes(s_restore_ttmps_lo)
get_sgpr_size_bytes(s_restore_ttmps_hi)
@@ -1101,10 +1105,9 @@ end
s_add_u32 s_restore_ttmps_lo, s_restore_ttmps_lo, s_restore_buf_rsrc0
s_addc_u32 s_restore_ttmps_hi, s_restore_buf_rsrc1, 0x0
s_and_b32 s_restore_ttmps_hi, s_restore_ttmps_hi, 0xFFFF
- s_load_dwordx2 [ttmp6, ttmp7], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x40 glc:1
- s_load_dwordx4 [ttmp8, ttmp9, ttmp10, ttmp11], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x48 glc:1
- s_load_dword ttmp13, [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x58 glc:1
- s_load_dwordx2 [ttmp14, ttmp15], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x5C glc:1
+ s_load_dwordx4 [ttmp4, ttmp5, ttmp6, ttmp7], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x50 glc:1
+ s_load_dwordx4 [ttmp8, ttmp9, ttmp10, ttmp11], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x60 glc:1
+ s_load_dword ttmp13, [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x74 glc:1
s_waitcnt lgkmcnt(0)
//reuse s_restore_m0 as a temp register