aco: add support for device clock on GFX11

According to LLVM, s_sendmsg_rtn(GET_REALTIME) should be used instead
of s_memrealtime.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19267>

diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index 16babf6..88b686b 100644 (file)
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -9020,6 +9020,10 @@ visit_intrinsic(isel_context* ctx, nir_intrinsic_instr* instr)
          /* "((size - 1) << 11) | register" (SHADER_CYCLES is encoded as register 29) */
          Temp clock = bld.sopk(aco_opcode::s_getreg_b32, bld.def(s1), ((20 - 1) << 11) | 29);
          bld.pseudo(aco_opcode::p_create_vector, Definition(dst), clock, Operand::zero());
+      } else if (nir_intrinsic_memory_scope(instr) == NIR_SCOPE_DEVICE &&
+                 ctx->options->gfx_level >= GFX11) {
+         bld.sop1(aco_opcode::s_sendmsg_rtn_b64, Definition(dst),
+                  Operand::c32(sendmsg_rtn_get_realtime));
       } else {
          aco_opcode opcode = nir_intrinsic_memory_scope(instr) == NIR_SCOPE_DEVICE
                                 ? aco_opcode::s_memrealtime