[flang] Use correct int extension flags for C-ABI calls on aarch64 #137105
+49
−4
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
llvmbot
added
flang
ashermancinelli
changed the title
|
@llvm/pr-subscribers-flang-fir-hlfir Author: Asher Mancinelli (ashermancinelli) ChangesThe AArch64 procedure call standard does not mandate that the callee extends the return value. Clang does not add signext to functions returning i8 or i16 on linux aarch64, but flang does. This means that runtime routines returning i8's (e.g. MINVAL on an array of INTEGER(KIND=1)) will have signext on the callsite/declaration, but not on the implementation, and the call site will assume the return value has already been sign extended when it has not. Adjust our integer extension flags to match clang and aarch64pcs on linux. The behavior on Darwin should be preserved. This is listed on the apple developer guide as a divergence from aarch64pcs. Full diff: https://github.com/llvm/llvm-project/pull/137105.diff 3 Files Affected:
diff --git a/flang/lib/Optimizer/CodeGen/Target.cpp b/flang/lib/Optimizer/CodeGen/Target.cpp
index 374308fa58947..b2223d95729a0 100644
--- a/flang/lib/Optimizer/CodeGen/Target.cpp
+++ b/flang/lib/Optimizer/CodeGen/Target.cpp
@@ -784,7 +784,7 @@ struct TargetX86_64Win : public GenericTarget<TargetX86_64Win> {
} // namespace
//===----------------------------------------------------------------------===//
-// AArch64 linux target specifics.
+// AArch64 target specifics.
//===----------------------------------------------------------------------===//
namespace {
@@ -810,6 +810,33 @@ struct TargetAArch64 : public GenericTarget<TargetAArch64> {
return marshal;
}
+ CodeGenSpecifics::Marshalling
+ integerArgumentType(mlir::Location loc,
+ mlir::IntegerType argTy) const override {
+ if (argTy.getWidth() < getCIntTypeWidth() && argTy.isSignless()) {
+ AT::IntegerExtension intExt;
+ if (argTy.getWidth() == 1) {
+ // Zero extend for 'i1'.
+ intExt = AT::IntegerExtension::Zero;
+ } else {
+ if (triple.isOSDarwin())
+ // On Darwin, sign extend. The apple developer guide specifies this as
+ // a divergence from the AArch64PCS:
+ // https://developer.apple.com/documentation/xcode/writing-arm64-code-for-apple-platforms#Pass-arguments-to-functions-correctly
+ intExt = AT::IntegerExtension::Sign;
+ else
+ // On linux, pass directly and do not extend.
+ intExt = AT::IntegerExtension::None;
+ }
+ CodeGenSpecifics::Marshalling marshal;
+ marshal.emplace_back(argTy, AT{/*alignment=*/0, /*byval=*/false,
+ /*sret=*/false, /*append=*/false,
+ /*intExt=*/intExt});
+ return marshal;
+ }
+ return GenericTarget::integerArgumentType(loc, argTy);
+ }
+
CodeGenSpecifics::Marshalling
complexReturnType(mlir::Location loc, mlir::Type eleTy) const override {
CodeGenSpecifics::Marshalling marshal;
diff --git a/flang/test/Fir/convert-to-llvm-target.fir b/flang/test/Fir/convert-to-llvm-target.fir
index bb873080072fa..6adc5cd4b737f 100644
--- a/flang/test/Fir/convert-to-llvm-target.fir
+++ b/flang/test/Fir/convert-to-llvm-target.fir
@@ -1,7 +1,7 @@
// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=x86_64-unknown-linux-gnu" %s | FileCheck %s --check-prefix INT64
// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=aarch64-unknown-linux-gnu" %s | FileCheck %s --check-prefixes INT64
// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=i386-unknown-linux-gnu" %s | FileCheck %s --check-prefixes INT32
-// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=powerpc64le-unknown-linux-gn" %s | FileCheck %s --check-prefixes INT64
+// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=powerpc64le-unknown-linux-gnu" %s | FileCheck %s --check-prefixes INT64
//=============================================================================
// SUMMARY: Tests for FIR --> LLVM MLIR conversion that *depend* on the target
diff --git a/flang/test/Fir/target-rewrite-integer.fir b/flang/test/Fir/target-rewrite-integer.fir
index 3be76e21a420f..4cb2026cf9ee0 100644
--- a/flang/test/Fir/target-rewrite-integer.fir
+++ b/flang/test/Fir/target-rewrite-integer.fir
@@ -1,6 +1,8 @@
// RUN: fir-opt --split-input-file --target-rewrite="target=i386-unknown-linux-gnu" %s | FileCheck %s --check-prefixes=I32,ALL
// RUN: fir-opt --split-input-file --target-rewrite="target=x86_64-unknown-linux-gnu" %s | FileCheck %s --check-prefixes=X64,ALL
+// RUN: fir-opt --split-input-file --target-rewrite="target=x86_64-apple-darwin" %s | FileCheck %s --check-prefixes=X64,ALL
// RUN: fir-opt --split-input-file --target-rewrite="target=aarch64-unknown-linux-gnu" %s | FileCheck %s --check-prefixes=AARCH64,ALL
+// RUN: fir-opt --split-input-file --target-rewrite="target=aarch64-apple-darwin" %s | FileCheck %s --check-prefixes=AARCH64DARWIN,ALL
// RUN: fir-opt --split-input-file --target-rewrite="target=powerpc64le-unknown-linux-gnu" %s | FileCheck %s --check-prefixes=PPC,ALL
// RUN: fir-opt --split-input-file --target-rewrite="target=sparc64-unknown-linux-gnu" %s | FileCheck %s --check-prefixes=SPARCV9,ALL
// RUN: fir-opt --split-input-file --target-rewrite="target=sparcv9-sun-solaris2.11" %s | FileCheck %s --check-prefixes=SPARCV9,ALL
@@ -17,6 +19,7 @@
// I32: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
// X64: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
// AARCH64: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
+// AARCH64DARWIN: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
// PPC: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
// SPARCV9: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
// LOONGARCH64: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
@@ -49,6 +52,7 @@ func.func private @_FortranAioEndIoStatement(!fir.ref<i8>) -> i32 attributes {fi
// I32: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
// X64: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
// AARCH64: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
+// AARCH64DARWIN: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
// PPC: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
// SPARCV9: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
// LOONGARCH64: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
@@ -69,6 +73,7 @@ func.func private @_SomeFunc_si1(si1) -> si1 attributes {fir.runtime}
// I32: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
// X64: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
// AARCH64: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
+// AARCH64DARWIN: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
// PPC: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
// SPARCV9: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
// LOONGARCH64: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
@@ -99,8 +104,20 @@ func.func private @_SomeFunc_ui1(ui1) -> ui1 attributes {fir.runtime}
// end subroutine test
// ALL-LABEL: @_QPtest_bindc
-// ALL: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
-// ALL: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// I32: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// I32: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// X64: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// X64: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// AARCH64: func.func private @cfun8(i8) -> i8 attributes {fir.bindc_name = "cfun8"}
+// AARCH64: func.func private @cfun16(i16) -> i16 attributes {fir.bindc_name = "cfun16"}
+// AARCH64DARWIN: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// AARCH64DARWIN: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// PPC: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// PPC: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// SPARCV9: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// SPARCV9: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// LOONGARCH64: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// LOONGARCH64: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
func.func @_QPtest_bindc(%arg0: !fir.ref<i8> {fir.bindc_name = "x"}, %arg1: !fir.ref<i16> {fir.bindc_name = "y"}) {
%0 = fir.load %arg0 : !fir.ref<i8>
%1 = fir.call @cfun8(%0) fastmath<contract> : (i8) -> i8
|
|
@llvm/pr-subscribers-flang-codegen Author: Asher Mancinelli (ashermancinelli) ChangesThe AArch64 procedure call standard does not mandate that the callee extends the return value. Clang does not add signext to functions returning i8 or i16 on linux aarch64, but flang does. This means that runtime routines returning i8's (e.g. MINVAL on an array of INTEGER(KIND=1)) will have signext on the callsite/declaration, but not on the implementation, and the call site will assume the return value has already been sign extended when it has not. Adjust our integer extension flags to match clang and aarch64pcs on linux. The behavior on Darwin should be preserved. This is listed on the apple developer guide as a divergence from aarch64pcs. Full diff: https://github.com/llvm/llvm-project/pull/137105.diff 3 Files Affected:
diff --git a/flang/lib/Optimizer/CodeGen/Target.cpp b/flang/lib/Optimizer/CodeGen/Target.cpp
index 374308fa58947..b2223d95729a0 100644
--- a/flang/lib/Optimizer/CodeGen/Target.cpp
+++ b/flang/lib/Optimizer/CodeGen/Target.cpp
@@ -784,7 +784,7 @@ struct TargetX86_64Win : public GenericTarget<TargetX86_64Win> {
} // namespace
//===----------------------------------------------------------------------===//
-// AArch64 linux target specifics.
+// AArch64 target specifics.
//===----------------------------------------------------------------------===//
namespace {
@@ -810,6 +810,33 @@ struct TargetAArch64 : public GenericTarget<TargetAArch64> {
return marshal;
}
+ CodeGenSpecifics::Marshalling
+ integerArgumentType(mlir::Location loc,
+ mlir::IntegerType argTy) const override {
+ if (argTy.getWidth() < getCIntTypeWidth() && argTy.isSignless()) {
+ AT::IntegerExtension intExt;
+ if (argTy.getWidth() == 1) {
+ // Zero extend for 'i1'.
+ intExt = AT::IntegerExtension::Zero;
+ } else {
+ if (triple.isOSDarwin())
+ // On Darwin, sign extend. The apple developer guide specifies this as
+ // a divergence from the AArch64PCS:
+ // https://developer.apple.com/documentation/xcode/writing-arm64-code-for-apple-platforms#Pass-arguments-to-functions-correctly
+ intExt = AT::IntegerExtension::Sign;
+ else
+ // On linux, pass directly and do not extend.
+ intExt = AT::IntegerExtension::None;
+ }
+ CodeGenSpecifics::Marshalling marshal;
+ marshal.emplace_back(argTy, AT{/*alignment=*/0, /*byval=*/false,
+ /*sret=*/false, /*append=*/false,
+ /*intExt=*/intExt});
+ return marshal;
+ }
+ return GenericTarget::integerArgumentType(loc, argTy);
+ }
+
CodeGenSpecifics::Marshalling
complexReturnType(mlir::Location loc, mlir::Type eleTy) const override {
CodeGenSpecifics::Marshalling marshal;
diff --git a/flang/test/Fir/convert-to-llvm-target.fir b/flang/test/Fir/convert-to-llvm-target.fir
index bb873080072fa..6adc5cd4b737f 100644
--- a/flang/test/Fir/convert-to-llvm-target.fir
+++ b/flang/test/Fir/convert-to-llvm-target.fir
@@ -1,7 +1,7 @@
// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=x86_64-unknown-linux-gnu" %s | FileCheck %s --check-prefix INT64
// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=aarch64-unknown-linux-gnu" %s | FileCheck %s --check-prefixes INT64
// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=i386-unknown-linux-gnu" %s | FileCheck %s --check-prefixes INT32
-// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=powerpc64le-unknown-linux-gn" %s | FileCheck %s --check-prefixes INT64
+// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=powerpc64le-unknown-linux-gnu" %s | FileCheck %s --check-prefixes INT64
//=============================================================================
// SUMMARY: Tests for FIR --> LLVM MLIR conversion that *depend* on the target
diff --git a/flang/test/Fir/target-rewrite-integer.fir b/flang/test/Fir/target-rewrite-integer.fir
index 3be76e21a420f..4cb2026cf9ee0 100644
--- a/flang/test/Fir/target-rewrite-integer.fir
+++ b/flang/test/Fir/target-rewrite-integer.fir
@@ -1,6 +1,8 @@
// RUN: fir-opt --split-input-file --target-rewrite="target=i386-unknown-linux-gnu" %s | FileCheck %s --check-prefixes=I32,ALL
// RUN: fir-opt --split-input-file --target-rewrite="target=x86_64-unknown-linux-gnu" %s | FileCheck %s --check-prefixes=X64,ALL
+// RUN: fir-opt --split-input-file --target-rewrite="target=x86_64-apple-darwin" %s | FileCheck %s --check-prefixes=X64,ALL
// RUN: fir-opt --split-input-file --target-rewrite="target=aarch64-unknown-linux-gnu" %s | FileCheck %s --check-prefixes=AARCH64,ALL
+// RUN: fir-opt --split-input-file --target-rewrite="target=aarch64-apple-darwin" %s | FileCheck %s --check-prefixes=AARCH64DARWIN,ALL
// RUN: fir-opt --split-input-file --target-rewrite="target=powerpc64le-unknown-linux-gnu" %s | FileCheck %s --check-prefixes=PPC,ALL
// RUN: fir-opt --split-input-file --target-rewrite="target=sparc64-unknown-linux-gnu" %s | FileCheck %s --check-prefixes=SPARCV9,ALL
// RUN: fir-opt --split-input-file --target-rewrite="target=sparcv9-sun-solaris2.11" %s | FileCheck %s --check-prefixes=SPARCV9,ALL
@@ -17,6 +19,7 @@
// I32: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
// X64: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
// AARCH64: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
+// AARCH64DARWIN: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
// PPC: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
// SPARCV9: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
// LOONGARCH64: func.func{{.*}}@_FortranAioOutputLogical({{.*}}i1 {llvm.zeroext}) -> (i1 {llvm.zeroext})
@@ -49,6 +52,7 @@ func.func private @_FortranAioEndIoStatement(!fir.ref<i8>) -> i32 attributes {fi
// I32: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
// X64: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
// AARCH64: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
+// AARCH64DARWIN: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
// PPC: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
// SPARCV9: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
// LOONGARCH64: func.func{{.*}}@_SomeFunc_si1(si1 {llvm.signext}) -> (si1 {llvm.signext})
@@ -69,6 +73,7 @@ func.func private @_SomeFunc_si1(si1) -> si1 attributes {fir.runtime}
// I32: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
// X64: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
// AARCH64: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
+// AARCH64DARWIN: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
// PPC: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
// SPARCV9: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
// LOONGARCH64: func.func{{.*}}@_SomeFunc_ui1(ui1 {llvm.zeroext}) -> (ui1 {llvm.zeroext})
@@ -99,8 +104,20 @@ func.func private @_SomeFunc_ui1(ui1) -> ui1 attributes {fir.runtime}
// end subroutine test
// ALL-LABEL: @_QPtest_bindc
-// ALL: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
-// ALL: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// I32: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// I32: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// X64: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// X64: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// AARCH64: func.func private @cfun8(i8) -> i8 attributes {fir.bindc_name = "cfun8"}
+// AARCH64: func.func private @cfun16(i16) -> i16 attributes {fir.bindc_name = "cfun16"}
+// AARCH64DARWIN: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// AARCH64DARWIN: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// PPC: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// PPC: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// SPARCV9: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// SPARCV9: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
+// LOONGARCH64: func.func private @cfun8(i8 {llvm.signext}) -> (i8 {llvm.signext}) attributes {fir.bindc_name = "cfun8"}
+// LOONGARCH64: func.func private @cfun16(i16 {llvm.signext}) -> (i16 {llvm.signext}) attributes {fir.bindc_name = "cfun16"}
func.func @_QPtest_bindc(%arg0: !fir.ref<i8> {fir.bindc_name = "x"}, %arg1: !fir.ref<i16> {fir.bindc_name = "y"}) {
%0 = fir.load %arg0 : !fir.ref<i8>
%1 = fir.call @cfun8(%0) fastmath<contract> : (i8) -> i8
|
The AArch64 procedure call standard does not mandate that the callee extends the return value. Clang does not add signext to functions returning i8 or i16 on aarch64, except on Darwin, but flang does. This means that runtime routines returning i8's (e.g. MINVAL on an array of INTEGER(KIND=1)) will have signext on the callsite/declaration in user code, but not on the implementation, and the call site will assume the return value has already been sign extended when it has not. Adjust our integer extension flags to match clang and aarch64pcs on linux. The behavior on Darwin should be preserved. This is listed on the apple developer guide as a divergence from aarch64pcs.
ashermancinelli
force-pushed
the
ajm/aarch64-int-abi
branch
from
3a32e73 to
e03bac7
Compare
ashermancinelli
changed the title
|
Looks good to me. Thank you, Asher! @kiranchandramohan can you please also review? |
There was a problem hiding this comment.
Thanks @ashermancinelli for the patch. A question and a Nit.
| if (triple.isOSDarwin()) | ||
| // On Darwin, sign extend. The apple developer guide specifies this as | ||
| // a divergence from the AArch64PCS: | ||
| // https://developer.apple.com/documentation/xcode/writing-arm64-code-for-apple-platforms#Pass-arguments-to-functions-correctly |
There was a problem hiding this comment.
Does it talk only about arguments and not return values?
There was a problem hiding this comment.
The apple dev guide does not mention sign/zero extension of returned integers, as far as I can tell. The aa64pcs explains that register types ought to be handled the same way, whether passed or returned:
The manner in which a result is returned from a function is determined by the type of that result:
If the type, T, of the result of a function is such that
void func(T arg)
would require that arg be passed as a value in a register (or set of registers) according to the rules in Parameter passing, then the result is returned in the same registers as would be used for such an argument.Otherwise, the caller shall reserve a block of memory of sufficient size and alignment to hold the result. The address of the memory block shall be passed as an additional argument to the function in x8. The callee may modify the result memory block at any point during the execution of the subroutine (there is no requirement for the callee to preserve the value stored in x8).
Clang does the same sign/zero extension for the return type, which is what I used as my guide. I wouldn't mind if the apple dev guide were updated to explicitly say one way or the other, though.
vzakhari
approved these changes
Apr 24, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The AArch64 procedure call standard does not mandate that the callee extends the return value. Clang does not add signext to functions returning i8 or i16 on linux aarch64, but flang does.
This means that runtime routines returning i8's will have signext on the callsite/declaration, but not on the implementation, and the call site will assume the return value has already been sign extended when it has not. This showed up in a test case calling MINVAL on an array of INTEGER*1.
Adjust our integer extension flags to match clang and aarch64pcs on linux. The behavior on Darwin should be preserved. This is listed on the apple developer guide as a divergence from aarch64pcs.