aco: add ACO_DEBUG=force-waitcnt to emit wait-states
Sounds useful for debugging missing wait-states and for improving detection of the faulty instruction in case of memory violations. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Reviewed-by: Daniel Schürmann <daniel@schuermann.dev> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6386>
This commit is contained in:
parent
258ef4d4a4
commit
f153151730
|
@ -612,6 +612,8 @@ RADV driver environment variables
|
|||
validate register assignment of ACO IR and catches many RA bugs
|
||||
``perfwarn``
|
||||
abort on some suboptimal code generation
|
||||
``force-waitcnt``
|
||||
force emitting waitcnt states if there is something to wait for
|
||||
|
||||
radeonsi driver environment variables
|
||||
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
||||
|
|
|
@ -473,9 +473,32 @@ wait_imm perform_barrier(wait_ctx& ctx, memory_sync_info sync, unsigned semantic
|
|||
return imm;
|
||||
}
|
||||
|
||||
void force_waitcnt(wait_ctx& ctx, wait_imm& imm)
|
||||
{
|
||||
if (ctx.vm_cnt)
|
||||
imm.vm = 0;
|
||||
if (ctx.exp_cnt)
|
||||
imm.exp = 0;
|
||||
if (ctx.lgkm_cnt)
|
||||
imm.lgkm = 0;
|
||||
|
||||
if (ctx.chip_class >= GFX10) {
|
||||
if (ctx.vs_cnt)
|
||||
imm.vs = 0;
|
||||
}
|
||||
}
|
||||
|
||||
wait_imm kill(Instruction* instr, wait_ctx& ctx, memory_sync_info sync_info)
|
||||
{
|
||||
wait_imm imm;
|
||||
|
||||
if (debug_flags & DEBUG_FORCE_WAITCNT) {
|
||||
/* Force emitting waitcnt states right after the instruction if there is
|
||||
* something to wait for.
|
||||
*/
|
||||
force_waitcnt(ctx, imm);
|
||||
}
|
||||
|
||||
if (ctx.exp_cnt || ctx.vm_cnt || ctx.lgkm_cnt)
|
||||
imm.combine(check_instr(instr, ctx));
|
||||
|
||||
|
|
|
@ -34,6 +34,7 @@ static const struct debug_control aco_debug_options[] = {
|
|||
{"validateir", DEBUG_VALIDATE_IR},
|
||||
{"validatera", DEBUG_VALIDATE_RA},
|
||||
{"perfwarn", DEBUG_PERFWARN},
|
||||
{"force-waitcnt", DEBUG_FORCE_WAITCNT},
|
||||
{NULL, 0}
|
||||
};
|
||||
|
||||
|
|
|
@ -51,6 +51,7 @@ enum {
|
|||
DEBUG_VALIDATE_IR = 0x1,
|
||||
DEBUG_VALIDATE_RA = 0x2,
|
||||
DEBUG_PERFWARN = 0x4,
|
||||
DEBUG_FORCE_WAITCNT = 0x8,
|
||||
};
|
||||
|
||||
/**
|
||||
|
|
Loading…
Reference in New Issue