------------------------------------------------------------------------
r323369 | aemerson | 2018-01-24 20:59:29 +0100 (Wed, 24 Jan 2018) | 4 lines
[GlobalISel] Don't fall back to FastISel.
Apparently checking the pass structure isn't enough to ensure that we don't fall
back to FastISel, as it's set up as part of the SelectionDAGISel.
------------------------------------------------------------------------
------------------------------------------------------------------------
r323371 | aemerson | 2018-01-24 21:35:37 +0100 (Wed, 24 Jan 2018) | 12 lines
[AArch64][GlobalISel] Fall back during AArch64 isel if we have a volatile load.
The tablegen imported patterns for sext(load(a)) don't check for single uses
of the load or delete the original after matching. As a result two loads are
left in the generated code. This particular issue will be fixed by adding
support for a G_SEXTLOAD opcode in future.
There are however other potential issues around this that wouldn't be fixed by
a G_SEXTLOAD, so until we have a proper solution we don't try to handle volatile
loads at all in the AArch64 selector.
Fixes/works around PR36018.
------------------------------------------------------------------------
git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_60@323434
91177308-0d34-0410-b5e6-
96231b3b80d8
FastISelFailed = false;
// Initialize the Fast-ISel state, if needed.
FastISel *FastIS = nullptr;
- if (TM.Options.EnableFastISel)
+ if (TM.Options.EnableFastISel) {
+ DEBUG(dbgs() << "Enabling fast-isel\n");
FastIS = TLI->createFastISel(*FuncInfo, LibInfo);
+ }
setupSwiftErrorVals(Fn, TLI, FuncInfo);
if (EnableGlobalISel == cl::BOU_TRUE ||
(EnableGlobalISel == cl::BOU_UNSET && isGlobalISelEnabled() &&
EnableFastISelOption != cl::BOU_TRUE)) {
+ TM->setFastISel(false);
+
if (addIRTranslator())
return true;
return false;
}
+ // FIXME: PR36018: Volatile loads in some cases are incorrectly selected by
+ // folding with an extend. Until we have a G_SEXTLOAD solution bail out if
+ // we hit one.
+ if (Opcode == TargetOpcode::G_LOAD && MemOp.isVolatile())
+ return false;
+
const unsigned PtrReg = I.getOperand(1).getReg();
#ifndef NDEBUG
const RegisterBank &PtrRB = *RBI.getRegBank(PtrReg, MRI, TRI);
--- /dev/null
+; RUN: llc -mtriple=aarch64_be-- %s -o /dev/null -debug-only=isel -O0 2>&1 | FileCheck %s
+
+; This test uses big endian in order to force an abort since it's not currently supported for GISel.
+; The purpose is to check that we don't fall back to FastISel. Checking the pass structure is insufficient
+; because the FastISel is set up in the SelectionDAGISel, so it doesn't appear on the pass structure.
+
+; CHECK-NOT: Enabling fast-ise
+define void @empty() {
+ ret void
+}
--- /dev/null
+; RUN: llc -O0 -mtriple=aarch64-apple-ios -o - %s | FileCheck %s
+
+@g = global i16 0, align 2
+declare void @bar(i32)
+
+; Check that only one load is generated. We fall back to
+define hidden void @foo() {
+; CHECK-NOT: ldrh
+; CHECK: ldrsh
+ %1 = load volatile i16, i16* @g, align 2
+ %2 = sext i16 %1 to i32
+ call void @bar(i32 %2)
+ ret void
+}