From 367cfd84c37f7535b7f29e40b76eca01e24bcd60 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault@amd.com>
Date: Mon, 23 Oct 2017 17:09:35 +0000
Subject: [PATCH] AMDGPU: Fix default range in non-kernel functions

The range should be assumed to be the hardware maximum
if a workitem intrinsic is used in a callable function
which does not know the restricted limit of the calling
kernel.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@316346 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/AMDGPU/AMDGPUSubtarget.cpp | 25 +++++++++++++++++++++----
 lib/Target/AMDGPU/AMDGPUSubtarget.h   |  3 +++
 test/CodeGen/AMDGPU/zext-lid.ll       | 22 ++++++++++++++++++++++
 3 files changed, 46 insertions(+), 4 deletions(-)

diff --git a/lib/Target/AMDGPU/AMDGPUSubtarget.cpp b/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
index ddc1cd457b4..06ebff7b394 100644
--- a/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
+++ b/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
@@ -190,14 +190,31 @@ unsigned AMDGPUSubtarget::getOccupancyWithLocalMemSize(uint32_t Bytes,
   return NumWaves;
 }
 
+std::pair<unsigned, unsigned>
+AMDGPUSubtarget::getDefaultFlatWorkGroupSize(CallingConv::ID CC) const {
+  switch (CC) {
+  case CallingConv::AMDGPU_CS:
+  case CallingConv::AMDGPU_KERNEL:
+  case CallingConv::SPIR_KERNEL:
+    return std::make_pair(getWavefrontSize() * 2, getWavefrontSize() * 4);
+  case CallingConv::AMDGPU_VS:
+  case CallingConv::AMDGPU_LS:
+  case CallingConv::AMDGPU_HS:
+  case CallingConv::AMDGPU_ES:
+  case CallingConv::AMDGPU_GS:
+  case CallingConv::AMDGPU_PS:
+    return std::make_pair(1, getWavefrontSize());
+  default:
+    return std::make_pair(1, 16 * getWavefrontSize());
+  }
+}
+
 std::pair<unsigned, unsigned> AMDGPUSubtarget::getFlatWorkGroupSizes(
   const Function &F) const {
+  // FIXME: 1024 if function.
   // Default minimum/maximum flat work group sizes.
   std::pair<unsigned, unsigned> Default =
-    AMDGPU::isCompute(F.getCallingConv()) ?
-      std::pair<unsigned, unsigned>(getWavefrontSize() * 2,
-                                    getWavefrontSize() * 4) :
-      std::pair<unsigned, unsigned>(1, getWavefrontSize());
+    getDefaultFlatWorkGroupSize(F.getCallingConv());
 
   // TODO: Do not process "amdgpu-max-work-group-size" attribute once mesa
   // starts using "amdgpu-flat-work-group-size" attribute.
diff --git a/lib/Target/AMDGPU/AMDGPUSubtarget.h b/lib/Target/AMDGPU/AMDGPUSubtarget.h
index 52e08e538f7..99c525eb9d7 100644
--- a/lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ b/lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -587,6 +587,9 @@ public:
                                                  FlatWorkGroupSize);
   }
 
+  /// \returns Default range flat work group size for a calling convention.
+  std::pair<unsigned, unsigned> getDefaultFlatWorkGroupSize(CallingConv::ID CC) const;
+
   /// \returns Subtarget's default pair of minimum/maximum flat work group sizes
   /// for function \p F, or minimum/maximum flat work group sizes explicitly
   /// requested using "amdgpu-flat-work-group-size" attribute attached to
diff --git a/test/CodeGen/AMDGPU/zext-lid.ll b/test/CodeGen/AMDGPU/zext-lid.ll
index 066f2927727..9a9c1fe7550 100644
--- a/test/CodeGen/AMDGPU/zext-lid.ll
+++ b/test/CodeGen/AMDGPU/zext-lid.ll
@@ -63,6 +63,26 @@ bb:
   ret void
 }
 
+; OPT-LABEL: @func_test_workitem_id_x_known_max_range(
+; OPT: tail call i32 @llvm.amdgcn.workitem.id.x(), !range !0
+define void @func_test_workitem_id_x_known_max_range(i32 addrspace(1)* nocapture %out) #0 {
+entry:
+  %id = tail call i32 @llvm.amdgcn.workitem.id.x()
+  %and = and i32 %id, 1023
+  store i32 %and, i32 addrspace(1)* %out, align 4
+  ret void
+}
+
+; OPT-LABEL: @func_test_workitem_id_x_default_range(
+; OPT: tail call i32 @llvm.amdgcn.workitem.id.x(), !range !6
+define void @func_test_workitem_id_x_default_range(i32 addrspace(1)* nocapture %out) #4 {
+entry:
+  %id = tail call i32 @llvm.amdgcn.workitem.id.x()
+  %and = and i32 %id, 1023
+  store i32 %and, i32 addrspace(1)* %out, align 4
+  ret void
+}
+
 declare i32 @llvm.amdgcn.workitem.id.x() #2
 
 declare i32 @llvm.amdgcn.workitem.id.y() #2
@@ -73,6 +93,7 @@ attributes #0 = { nounwind "amdgpu-flat-work-group-size"="64,128" }
 attributes #1 = { nounwind "amdgpu-flat-work-group-size"="512,512" }
 attributes #2 = { nounwind readnone speculatable }
 attributes #3 = { nounwind readnone }
+attributes #4 = { nounwind }
 
 !0 = !{i32 32, i32 4, i32 1}
 
@@ -82,3 +103,4 @@ attributes #3 = { nounwind readnone }
 ; OPT: !3 = !{i32 0, i32 4}
 ; OPT: !4 = !{i32 0, i32 1}
 ; OPT: !5 = !{i32 0, i32 512}
+; OPT: !6 = !{i32 0, i32 1024}
-- 
2.40.0