[PATCH] D16698: AMDGPU: Make workgroup id z imply y is enabled

[Matt Arsenault via llvm-commits]

arsenm created this revision.
arsenm added a reviewer: tstellarAMD.
arsenm added a subscriber: llvm-commits.
Herald added a subscriber: arsenm.

According to the HSA ABI document, this is also the case with
the workgroup id as well as the workitem id.

http://reviews.llvm.org/D16698

Files:
  lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
  test/CodeGen/AMDGPU/work-item-intrinsics.ll

Index: test/CodeGen/AMDGPU/work-item-intrinsics.ll
===================================================================
--- test/CodeGen/AMDGPU/work-item-intrinsics.ll
+++ test/CodeGen/AMDGPU/work-item-intrinsics.ll
@@ -175,7 +175,7 @@
 ; FUNC-LABEL: {{^}}tgid_z:
 ; HSA: compute_pgm_rsrc2_user_sgpr = 6
 ; HSA: compute_pgm_rsrc2_tgid_x_en = 1
-; HSA: compute_pgm_rsrc2_tgid_y_en = 0
+; HSA: compute_pgm_rsrc2_tgid_y_en = 1
 ; HSA: compute_pgm_rsrc2_tgid_z_en = 1
 ; HSA: compute_pgm_rsrc2_tg_size_en = 0
 ; HSA: compute_pgm_rsrc2_tidig_comp_cnt = 0
@@ -190,15 +190,15 @@
 ; HSA: enable_sgpr_grid_workgroup_count_y = 0
 ; HSA: enable_sgpr_grid_workgroup_count_z = 0
 
-; GCN-NOHSA: v_mov_b32_e32 [[VVAL:v[0-9]+]], s3{{$}}
-; HSA: v_mov_b32_e32 [[VVAL:v[0-9]+]], s7{{$}}
+; GCN-NOHSA: v_mov_b32_e32 [[VVAL:v[0-9]+]], s4{{$}}
+; HSA: v_mov_b32_e32 [[VVAL:v[0-9]+]], s8{{$}}
 ; GCN-NOHSA: buffer_store_dword [[VVAL]]
 ; HSA: flat_store_dword [[VVAL]]
 
 ; HSA: COMPUTE_PGM_RSRC2:USER_SGPR: 6
 ; GCN-NOHSA: COMPUTE_PGM_RSRC2:USER_SGPR: 2
 ; GCN: COMPUTE_PGM_RSRC2:TGID_X_EN: 1
-; GCN: COMPUTE_PGM_RSRC2:TGID_Y_EN: 0
+; GCN: COMPUTE_PGM_RSRC2:TGID_Y_EN: 1
 ; GCN: COMPUTE_PGM_RSRC2:TGID_Z_EN: 1
 ; GCN: COMPUTE_PGM_RSRC2:TIDIG_COMP_CNT: 0
 define void @tgid_z(i32 addrspace(1)* %out) {
Index: lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
===================================================================
--- lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
+++ lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
@@ -84,8 +84,11 @@
   if (F->hasFnAttribute("amdgpu-work-group-id-y"))
     WorkGroupIDY = true;
 
-  if (F->hasFnAttribute("amdgpu-work-group-id-z"))
+  if (F->hasFnAttribute("amdgpu-work-group-id-z")) {
+    // Z implies Y is enabled.
+    WorkGroupIDY = true;
     WorkGroupIDZ = true;
+  }
 
   if (F->hasFnAttribute("amdgpu-work-item-id-y"))
     WorkItemIDY = true;


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D16698.46325.patch
Type: text/x-patch
Size: 1867 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160129/d9145bc7/attachment.bin>