[PATCH] D159459: [AMDGPU] Add ASM and MC updates for preloading kernargs

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

Also diagnose out of bounds offsets?



================
Comment at: llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp:1953
+  case amdhsa::KERNARG_PRELOAD_OFFSET:
+    if (hasKernargPreload()) {
+      using namespace amdhsa;
----------------
I think the disassembler should proceed to print whatever is there regardless of the support. It can skip printing if it's 0


================
Comment at: llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp:2149
+unsigned hasKernargPreload(const MCSubtargetInfo &STI) {
+  return isGFX90A(STI) || isGFX940(STI);
+}
----------------
Probably should define a proper sub target feature for this 


================
Comment at: llvm/test/MC/AMDGPU/user-sgpr-count-diag.s:6
   .amdhsa_user_sgpr_queue_ptr 1
+  .amdhsa_accum_offset 4
   .amdhsa_next_free_vgpr 32
----------------
Why amdhsa_accum_offset? I don't recognize this one


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D159459/new/

https://reviews.llvm.org/D159459