AMDGPU: Partially fix default device for HSA

There are a few different issues, mostly stemming from using
generation based checks for anything instead of subtarget
features. Stop adding flat-address-space as a feature for HSA, as it
should only be a device property. This was incorrectly allowing flat
instructions to select for SI.

Increase the default generation for HSA to avoid the encoding error
when emitting objects. This has some other side effects from various
checks which probably should be separate subtarget features (in the
cost model and for dealing with the DS offset folding issue).

Partial fix for bug 41070. It should probably be an error to try using
amdhsa without flat support.

llvm-svn: 356347

1 files changed, 6 insertions, 4 deletions

diff --git a/llvm/lib/Support/TargetParser.cpp b/llvm/lib/Support/TargetParser.cpp
index 76401ac0a85..59e096b4b2c 100644
--- a/llvm/lib/Support/TargetParser.cpp
+++ b/llvm/lib/Support/TargetParser.cpp
@@ -169,12 +169,14 @@ void AMDGPU::fillValidArchListR600(SmallVectorImpl<StringRef> &Values) {
 }
 
 AMDGPU::IsaVersion AMDGPU::getIsaVersion(StringRef GPU) {
-  if (GPU == "generic")
-    return {7, 0, 0};
-
   AMDGPU::GPUKind AK = parseArchAMDGCN(GPU);
-  if (AK == AMDGPU::GPUKind::GK_NONE)
+  if (AK == AMDGPU::GPUKind::GK_NONE) {
+    if (GPU == "generic-hsa")
+      return {7, 0, 0};
+    if (GPU == "generic")
+      return {6, 0, 0};
     return {0, 0, 0};
+  }
 
   switch (AK) {
   case GK_GFX600: return {6, 0, 0};