summaryrefslogtreecommitdiffstats
path: root/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
diff options
context:
space:
mode:
authorMatt Arsenault <Matthew.Arsenault@amd.com>2019-06-05 22:37:50 +0000
committerMatt Arsenault <Matthew.Arsenault@amd.com>2019-06-05 22:37:50 +0000
commit34c8b835b16fb3879f1b9770e91df21883356bb6 (patch)
treeb5f21eeb4c344bebeb9791227d9228c9b18c2b0a /llvm/lib/Target/AMDGPU/SIISelLowering.cpp
parentc72fbe5dc18ed0ffa6a1c67afef90e67c8b55c8b (diff)
downloadbcm5719-llvm-34c8b835b16fb3879f1b9770e91df21883356bb6.tar.gz
bcm5719-llvm-34c8b835b16fb3879f1b9770e91df21883356bb6.zip
AMDGPU: Don't fix emergency stack slot at offset 0
This forced the caller to be aware of this, which is an ugly ABI feature. Partially reverts r295877. The original reasons for doing this are mostly fixed. Alloca is now in a non-0 address space, so it should be OK to have 0 as a valid pointer. Since we treat the absolute address as the pointer value, this part only really needed to apply to kernels. Since r357093, we avoid the need to increment/decrement the offset register in more cases, and since r354816 the scavenger can fail without spilling, so it's less critical that we try to avoid an offset that fits in the MUBUF offset. Restrict to callable functions for now to split this into 2 steps to limit thte number of test updates and in case anything breaks. llvm-svn: 362665
Diffstat (limited to 'llvm/lib/Target/AMDGPU/SIISelLowering.cpp')
-rw-r--r--llvm/lib/Target/AMDGPU/SIISelLowering.cpp10
1 files changed, 0 insertions, 10 deletions
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 8a08bc463da..f5fcb7cdfe0 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -1940,12 +1940,6 @@ SDValue SITargetLowering::LowerFormalArguments(
bool IsKernel = AMDGPU::isKernel(CallConv);
bool IsEntryFunc = AMDGPU::isEntryFunctionCC(CallConv);
- if (!IsEntryFunc) {
- // 4 bytes are reserved at offset 0 for the emergency stack slot. Skip over
- // this when allocating argument fixed offsets.
- CCInfo.AllocateStack(4, 4);
- }
-
if (IsShader) {
processShaderInputArgs(Splits, CallConv, Ins, Skipped, FType, Info);
@@ -2551,7 +2545,6 @@ SDValue SITargetLowering::LowerCall(CallLoweringInfo &CLI,
"unsupported call from graphics shader of function ");
}
- // The first 4 bytes are reserved for the callee's emergency stack slot.
if (IsTailCall) {
IsTailCall = isEligibleForTailCallOptimization(
Callee, CallConv, IsVarArg, Outs, OutVals, Ins, DAG);
@@ -2578,9 +2571,6 @@ SDValue SITargetLowering::LowerCall(CallLoweringInfo &CLI,
CCState CCInfo(CallConv, IsVarArg, MF, ArgLocs, *DAG.getContext());
CCAssignFn *AssignFn = CCAssignFnForCall(CallConv, IsVarArg);
- // The first 4 bytes are reserved for the callee's emergency stack slot.
- CCInfo.AllocateStack(4, 4);
-
CCInfo.AnalyzeCallOperands(Outs, AssignFn);
// Get a count of how many bytes are to be pushed on the stack.
OpenPOWER on IntegriCloud