<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On Fri, Apr 20, 2018 at 4:30 PM, Ian Romanick <<a href="mailto:idr@freedesktop.org" target="_blank">idr@freedesktop.org</a>> wrote: <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On 04/20/2018 11:56 AM, Jason Ekstrand wrote: > Ugh... I was actually thinking about this the other day. How did you > come up with your list of instructions? Is there some algorithmic way > to determine it such as sends with no header? I just looked through the generator for pseudo-ops that emit explicit reads of g0. Since it was a manual process, I may have missed some. :( I don't know of a way to automate this, but... it may be possible. Not sure. <div class="gmail-HOEnZb"><div class="gmail-h5"></div></div></blockquote><div> </div><div>We ran into a similar issue with Dota 2 on Vulkan which was fixed by ff4726077d86800d33520581f154a27dac408588 </div><div>For vec4, this may be a reasonable solution. I don't like list of instructions in general but I'm a bit more ok with duct tape in vec4. Adding matt & Ken in case they have opinions. </div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div class="gmail-HOEnZb"><div class="gmail-h5"> > On April 20, 2018 08:52:50 "Ian Romanick" <<a href="mailto:idr@freedesktop.org">idr@freedesktop.org</a>> wrote: > >> From: Ian Romanick <<a href="mailto:ian.d.romanick@intel.com">ian.d.romanick@intel.com</a>> >> >> Otherwise the scheduler can move the writes after the reads. >> >> Signed-off-by: Ian Romanick <<a href="mailto:ian.d.romanick@intel.com">ian.d.romanick@intel.com</a>> >> Bugzilla: <a href="https://bugs.freedesktop.org/show_bug.cgi?id=95012" rel="noreferrer" target="_blank">https://bugs.freedesktop.org/show_bug.cgi?id=95012</a> >> Cc: Mark Janes <<a href="mailto:mark.a.janes@intel.com">mark.a.janes@intel.com</a>> >> Cc: Clayton A Craft <<a href="mailto:clayton.a.craft@intel.com">clayton.a.craft@intel.com</a>> >> --- >> I'd really like to have this run on the CI with the tests from bugzill >> #95012 re-enabled. I don't know how to do this on my own. This series >> is also available at: >> >> <a href="https://cgit.freedesktop.org/~idr/mesa/log/?h=bug-95012" rel="noreferrer" target="_blank">https://cgit.freedesktop.org/~idr/mesa/log/?h=bug-95012</a>. >> >> src/intel/compiler/brw_ir_vec4.h | 25 >> ++++++++++++++++++++++++ >> src/intel/compiler/brw_schedule_instructions.cpp | 3 +++ >> 2 files changed, 28 insertions(+) >> >> diff --git a/src/intel/compiler/brw_ir_vec4.h >> b/src/intel/compiler/brw_ir_vec4.h >> index 95c5119c6c0..e401d8b4d16 100644 >> --- a/src/intel/compiler/brw_ir_vec4.h >> +++ b/src/intel/compiler/brw_ir_vec4.h >> @@ -334,6 +334,31 @@ public: >> opcode != BRW_OPCODE_IF && >> opcode != BRW_OPCODE_WHILE)); >> } >> + >> + bool reads_g0_implicitly() const >> + { >> + switch (opcode) { >> + case SHADER_OPCODE_TEX: >> + case SHADER_OPCODE_TXL: >> + case SHADER_OPCODE_TXD: >> + case SHADER_OPCODE_TXF: >> + case SHADER_OPCODE_TXF_CMS_W: >> + case SHADER_OPCODE_TXF_CMS: >> + case SHADER_OPCODE_TXF_MCS: >> + case SHADER_OPCODE_TXS: >> + case SHADER_OPCODE_TG4: >> + case SHADER_OPCODE_TG4_OFFSET: >> + case SHADER_OPCODE_SAMPLEINFO: >> + case VS_OPCODE_PULL_CONSTANT_LOAD: >> + case GS_OPCODE_SET_PRIMITIVE_ID: >> + case GS_OPCODE_GET_INSTANCE_ID: >> + case SHADER_OPCODE_GEN4_SCRATCH_READ: >> + case SHADER_OPCODE_GEN4_SCRATCH_WRITE: >> + return true; >> + default: >> + return false; >> + } >> + } >> }; >> >> /** >> diff --git a/src/intel/compiler/brw_schedule_instructions.cpp >> b/src/intel/compiler/brw_schedule_instructions.cpp >> index fa85045de74..f817142a8b5 100644 >> --- a/src/intel/compiler/brw_schedule_instructions.cpp >> +++ b/src/intel/compiler/brw_schedule_instructions.cpp >> @@ -1267,6 +1267,9 @@ vec4_instruction_scheduler::calculate_deps() >> } >> } >> >> + if (inst->reads_g0_implicitly()) >> + add_dep(last_fixed_grf_write, n); >> + >> if (!inst->is_send_from_grf()) { >> for (int i = 0; i < inst->mlen; i++) { >> /* It looks like the MRF regs are released in the send >> -- >> 2.14.3 >> >> _______________________________________________ >> mesa-dev mailing list >> <a href="mailto:mesa-dev@lists.freedesktop.org">mesa-dev@lists.freedesktop.org</a> >> <a href="https://lists.freedesktop.org/mailman/listinfo/mesa-dev" rel="noreferrer" target="_blank">https://lists.freedesktop.org/mailman/listinfo/mesa-dev</a> </div></div></blockquote></div> </div></div>