commit f15315173032714b1cc4c1820e50b95ae29dd88c
author Samuel Pitoiset <samuel.pitoiset@gmail.com> Wed Aug 19 10:40:35 2020 +0200
committer Samuel Pitoiset <samuel.pitoiset@gmail.com> Fri Aug 21 13:22:58 2020 +0200
tree cd0183329460d50d54a2a0b8122a4b51fba0a50c
parent 258ef4d4a4b6c11b141e12e6736b01f408461c1d

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>

docs/envvars.rst

diff --git a/docs/envvars.rst b/docs/envvars.rst
index 4beba25..cee45fb 100644
--- a/docs/envvars.rst
+++ b/docs/envvars.rst
@@ -612,6 +612,8 @@
       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
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

src/amd/compiler/aco_insert_waitcnt.cpp

diff --git a/src/amd/compiler/aco_insert_waitcnt.cpp b/src/amd/compiler/aco_insert_waitcnt.cpp
index db0c7c0..751892e 100644
--- a/src/amd/compiler/aco_insert_waitcnt.cpp
+++ b/src/amd/compiler/aco_insert_waitcnt.cpp
@@ -473,9 +473,32 @@
    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));
 

src/amd/compiler/aco_ir.cpp

diff --git a/src/amd/compiler/aco_ir.cpp b/src/amd/compiler/aco_ir.cpp
index 2a0a500..801fbc6 100644
--- a/src/amd/compiler/aco_ir.cpp
+++ b/src/amd/compiler/aco_ir.cpp
@@ -34,6 +34,7 @@
    {"validateir", DEBUG_VALIDATE_IR},
    {"validatera", DEBUG_VALIDATE_RA},
    {"perfwarn", DEBUG_PERFWARN},
+   {"force-waitcnt", DEBUG_FORCE_WAITCNT},
    {NULL, 0}
 };
 

src/amd/compiler/aco_ir.h

diff --git a/src/amd/compiler/aco_ir.h b/src/amd/compiler/aco_ir.h
index 0b3a776..0e5b69a 100644
--- a/src/amd/compiler/aco_ir.h
+++ b/src/amd/compiler/aco_ir.h
@@ -51,6 +51,7 @@
    DEBUG_VALIDATE_IR = 0x1,
    DEBUG_VALIDATE_RA = 0x2,
    DEBUG_PERFWARN = 0x4,
+   DEBUG_FORCE_WAITCNT = 0x8,
 };
 
 /**