a0d699a8e6
This helps to disambiguate accesses in the caller and the callee after LLVM inlining in some apps. I did not see any performance changes, but this is one step towards enabling other optimizations in the apps that I am looking at. The definition of llvm.noalias says: ``` ... indicates that memory locations accessed via pointer values based on the argument or return value are not also accessed, during the execution of the function, via pointer values not based on the argument or return value. This guarantee only holds for memory locations that are modified, by any means, during the execution of the function. ``` I believe this exactly matches Fortran rules for the dummy arguments that are modified during their subprogram execution. I also set llvm.noalias and llvm.nocapture on the !fir.box<> arguments, because the corresponding descriptors cannot be captured and cannot alias anything (not based on them) during the execution of the subprogram.
113 lines
4.8 KiB
Plaintext
113 lines
4.8 KiB
Plaintext
// RUN: tco --target=i386-unknown-linux-gnu %s | FileCheck %s --check-prefix=I32
|
|
// RUN: tco --target=x86_64-unknown-linux-gnu %s | FileCheck %s --check-prefix=X64
|
|
// RUN: tco --target=aarch64-unknown-linux-gnu %s | FileCheck %s --check-prefix=AARCH64
|
|
// RUN: tco --target=powerpc64le-unknown-linux-gnu %s | FileCheck %s --check-prefix=PPC
|
|
|
|
// I32-LABEL: define i64 @gen4()
|
|
// X64-LABEL: define <2 x float> @gen4()
|
|
// AARCH64-LABEL: define { float, float } @gen4()
|
|
// PPC-LABEL: define { float, float } @gen4()
|
|
func.func @gen4() -> complex<f32> {
|
|
%1 = fir.undefined complex<f32>
|
|
%2 = arith.constant 2.0 : f32
|
|
%c0 = arith.constant 0 : i32
|
|
%4 = fir.insert_value %1, %2, [0 : index] : (complex<f32>, f32) -> complex<f32>
|
|
%c1 = arith.constant 1 : i32
|
|
%5 = arith.constant -42.0 : f32
|
|
%6 = fir.insert_value %4, %5, [1 : index] : (complex<f32>, f32) -> complex<f32>
|
|
// I32: store { float, float } { float 2.000000e+00, float -4.200000e+01 }
|
|
// I32: %[[load:.*]] = load i64, ptr
|
|
// I32: ret i64 %[[load]]
|
|
// X64: store { float, float } { float 2.000000e+00, float -4.200000e+01 }
|
|
// X64: %[[load:.*]] = load <2 x float>, ptr
|
|
// X64: ret <2 x float> %[[load]]
|
|
// AARCH64: ret { float, float }
|
|
// PPC: ret { float, float }
|
|
return %6 : complex<f32>
|
|
}
|
|
|
|
// I32-LABEL: define void @gen8(ptr noalias sret({ double, double }) align 4 captures(none) %
|
|
// X64-LABEL: define { double, double } @gen8()
|
|
// AARCH64-LABEL: define { double, double } @gen8()
|
|
// PPC-LABEL: define { double, double } @gen8()
|
|
func.func @gen8() -> complex<f64> {
|
|
%1 = fir.undefined complex<f64>
|
|
%2 = arith.constant 1.0 : f64
|
|
%3 = arith.constant -4.0 : f64
|
|
%c0 = arith.constant 0 : i32
|
|
%4 = fir.insert_value %1, %3, [0 : index] : (complex<f64>, f64) -> complex<f64>
|
|
%c1 = arith.constant 1 : i32
|
|
%5 = fir.insert_value %4, %2, [1 : index] : (complex<f64>, f64) -> complex<f64>
|
|
// I32: store { double, double } { double -4.000000e+00, double 1.000000e+00 }
|
|
// X64: store { double, double } { double -4.000000e+00, double 1.000000e+00 }
|
|
// X64: %[[load:.*]] = load { double, double }
|
|
// X64: ret { double, double } %[[load]]
|
|
// AARCH64: ret { double, double }
|
|
// PPC: ret { double, double }
|
|
return %5 : complex<f64>
|
|
}
|
|
|
|
// I32: declare void @sink4(ptr byval({ float, float }) align 4)
|
|
// X64: declare void @sink4(<2 x float>)
|
|
// AARCH64: declare void @sink4([2 x float])
|
|
// PPC: declare void @sink4(float, float)
|
|
func.func private @sink4(complex<f32>) -> ()
|
|
|
|
// I32: declare void @sink8(ptr byval({ double, double }) align 4)
|
|
// X64: declare void @sink8(double, double)
|
|
// AARCH64: declare void @sink8([2 x double])
|
|
// PPC: declare void @sink8(double, double)
|
|
func.func private @sink8(complex<f64>) -> ()
|
|
|
|
// I32-LABEL: define void @call4()
|
|
// X64-LABEL: define void @call4()
|
|
// AARCH64-LABEL: define void @call4()
|
|
func.func @call4() {
|
|
// I32: = call i64 @gen4()
|
|
// X64: = call <2 x float> @gen4()
|
|
// AARCH64: = call { float, float } @gen4()
|
|
// PPC: = call { float, float } @gen4()
|
|
%1 = fir.call @gen4() : () -> complex<f32>
|
|
// I32: call void @sink4(ptr byval({ float, float }) align 4 %
|
|
// X64: call void @sink4(<2 x float> %
|
|
// AARCH64: call void @sink4([2 x float] %
|
|
// PPC: call void @sink4(float %{{.*}}, float %{{.*}})
|
|
fir.call @sink4(%1) : (complex<f32>) -> ()
|
|
return
|
|
}
|
|
|
|
// I32-LABEL: define void @call8()
|
|
// X64-LABEL: define void @call8()
|
|
// AARCH64-LABEL: define void @call8()
|
|
func.func @call8() {
|
|
// I32: call void @gen8(ptr sret({ double, double }) align 4 %
|
|
// X64: = call { double, double } @gen8()
|
|
// AARCH64: = call { double, double } @gen8()
|
|
// PPC: = call { double, double } @gen8()
|
|
%1 = fir.call @gen8() : () -> complex<f64>
|
|
// I32: call void @sink8(ptr byval({ double, double }) align 4 %
|
|
// X64: call void @sink8(double %{{[0-9]*}}, double %{{[0-9]*}})
|
|
// AARCH64: call void @sink8([2 x double] %
|
|
// PPC: call void @sink8(double %{{.*}}, double %{{.*}})
|
|
fir.call @sink8(%1) : (complex<f64>) -> ()
|
|
return
|
|
}
|
|
|
|
// I32-LABEL: define i64 @char1lensum(ptr {{[^%]*}}%0, ptr {{[^%]*}}%1, i32 %2, i32 %3)
|
|
// X64-LABEL: define i64 @char1lensum(ptr {{[^%]*}}%0, ptr {{[^%]*}}%1, i64 %2, i64 %3)
|
|
// PPC-LABEL: define i64 @char1lensum(ptr {{[^%]*}}%0, ptr {{[^%]*}}%1, i64 %2, i64 %3)
|
|
func.func @char1lensum(%arg0 : !fir.boxchar<1>, %arg1 : !fir.boxchar<1>) -> i64 {
|
|
// X64-DAG: %[[p0:.*]] = insertvalue { ptr, i64 } undef, ptr %1, 0
|
|
// X64-DAG: = insertvalue { ptr, i64 } %[[p0]], i64 %3, 1
|
|
// X64-DAG: %[[p1:.*]] = insertvalue { ptr, i64 } undef, ptr %0, 0
|
|
// X64-DAG: = insertvalue { ptr, i64 } %[[p1]], i64 %2, 1
|
|
%1:2 = fir.unboxchar %arg0 : (!fir.boxchar<1>) -> (!fir.ref<!fir.char<1>>, i64)
|
|
%2:2 = fir.unboxchar %arg1 : (!fir.boxchar<1>) -> (!fir.ref<!fir.char<1>>, i64)
|
|
// I32: %[[add:.*]] = add i64 %
|
|
// X64: %[[add:.*]] = add i64 %
|
|
%3 = arith.addi %1#1, %2#1 : i64
|
|
// I32: ret i64 %[[add]]
|
|
// X64: ret i64 %[[add]]
|
|
return %3 : i64
|
|
}
|