[clang][CodeGen] Fix in codegen for __builtin_popcountg/ctzg/clzg #90845
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
bjope
requested review from
nickdesaulniers,
AaronBallman,
efriedma-quic and
Sirraide
llvmbot
added
clang
|
@llvm/pr-subscribers-clang-codegen @llvm/pr-subscribers-clang Author: Björn Pettersson (bjope) ChangesMake sure that the result from the popcnt/ctlz/cttz intrinsics is unsigned casted to int, rather than casted as a signed value, when expanding the __builtin_popcountg/__builtin_ctzg/__builtin_clzg builtins. An example would be Since the input type is generic for those "g" versions of the builtins the intrinsic call may return a value for which the sign bit is set (that could typically for BitInt of size 1 and 2). So we need to emit a zext rather than a sext to avoid negative results. Full diff: https://github.com/llvm/llvm-project/pull/90845.diff 2 Files Affected:
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index a370734e00d3e1..fc1dbb87ed1bf2 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -3239,7 +3239,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
Builder.getInt1(HasFallback || getTarget().isCLZForZeroUndef());
Value *Result = Builder.CreateCall(F, {ArgValue, ZeroUndef});
if (Result->getType() != ResultType)
- Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/true,
+ Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/false,
"cast");
if (!HasFallback)
return RValue::get(Result);
@@ -3271,7 +3271,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
Builder.getInt1(HasFallback || getTarget().isCLZForZeroUndef());
Value *Result = Builder.CreateCall(F, {ArgValue, ZeroUndef});
if (Result->getType() != ResultType)
- Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/true,
+ Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/false,
"cast");
if (!HasFallback)
return RValue::get(Result);
@@ -3351,7 +3351,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
llvm::Type *ResultType = ConvertType(E->getType());
Value *Result = Builder.CreateCall(F, ArgValue);
if (Result->getType() != ResultType)
- Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/true,
+ Result = Builder.CreateIntCast(Result, ResultType, /*isSigned*/false,
"cast");
return RValue::get(Result);
}
diff --git a/clang/test/CodeGen/builtins.c b/clang/test/CodeGen/builtins.c
index 407e0857d22311..b41efb59e61dbc 100644
--- a/clang/test/CodeGen/builtins.c
+++ b/clang/test/CodeGen/builtins.c
@@ -949,12 +949,12 @@ void test_builtin_popcountg(unsigned char uc, unsigned short us,
pop = __builtin_popcountg(uc);
// CHECK: %1 = load i8, ptr %uc.addr, align 1
// CHECK-NEXT: %2 = call i8 @llvm.ctpop.i8(i8 %1)
- // CHECK-NEXT: %cast = sext i8 %2 to i32
+ // CHECK-NEXT: %cast = zext i8 %2 to i32
// CHECK-NEXT: store volatile i32 %cast, ptr %pop, align 4
pop = __builtin_popcountg(us);
// CHECK-NEXT: %3 = load i16, ptr %us.addr, align 2
// CHECK-NEXT: %4 = call i16 @llvm.ctpop.i16(i16 %3)
- // CHECK-NEXT: %cast1 = sext i16 %4 to i32
+ // CHECK-NEXT: %cast1 = zext i16 %4 to i32
// CHECK-NEXT: store volatile i32 %cast1, ptr %pop, align 4
pop = __builtin_popcountg(ui);
// CHECK-NEXT: %5 = load i32, ptr %ui.addr, align 4
@@ -992,12 +992,12 @@ void test_builtin_clzg(unsigned char uc, unsigned short us, unsigned int ui,
lz = __builtin_clzg(uc);
// CHECK: %1 = load i8, ptr %uc.addr, align 1
// CHECK-NEXT: %2 = call i8 @llvm.ctlz.i8(i8 %1, i1 true)
- // CHECK-NEXT: %cast = sext i8 %2 to i32
+ // CHECK-NEXT: %cast = zext i8 %2 to i32
// CHECK-NEXT: store volatile i32 %cast, ptr %lz, align 4
lz = __builtin_clzg(us);
// CHECK-NEXT: %3 = load i16, ptr %us.addr, align 2
// CHECK-NEXT: %4 = call i16 @llvm.ctlz.i16(i16 %3, i1 true)
- // CHECK-NEXT: %cast1 = sext i16 %4 to i32
+ // CHECK-NEXT: %cast1 = zext i16 %4 to i32
// CHECK-NEXT: store volatile i32 %cast1, ptr %lz, align 4
lz = __builtin_clzg(ui);
// CHECK-NEXT: %5 = load i32, ptr %ui.addr, align 4
@@ -1026,7 +1026,7 @@ void test_builtin_clzg(unsigned char uc, unsigned short us, unsigned int ui,
lz = __builtin_clzg(uc, sc);
// CHECK-NEXT: %15 = load i8, ptr %uc.addr, align 1
// CHECK-NEXT: %16 = call i8 @llvm.ctlz.i8(i8 %15, i1 true)
- // CHECK-NEXT: %cast6 = sext i8 %16 to i32
+ // CHECK-NEXT: %cast6 = zext i8 %16 to i32
// CHECK-NEXT: %iszero = icmp eq i8 %15, 0
// CHECK-NEXT: %17 = load i8, ptr %sc.addr, align 1
// CHECK-NEXT: %conv = sext i8 %17 to i32
@@ -1035,7 +1035,7 @@ void test_builtin_clzg(unsigned char uc, unsigned short us, unsigned int ui,
lz = __builtin_clzg(us, uc);
// CHECK-NEXT: %18 = load i16, ptr %us.addr, align 2
// CHECK-NEXT: %19 = call i16 @llvm.ctlz.i16(i16 %18, i1 true)
- // CHECK-NEXT: %cast7 = sext i16 %19 to i32
+ // CHECK-NEXT: %cast7 = zext i16 %19 to i32
// CHECK-NEXT: %iszero8 = icmp eq i16 %18, 0
// CHECK-NEXT: %20 = load i8, ptr %uc.addr, align 1
// CHECK-NEXT: %conv9 = zext i8 %20 to i32
@@ -1094,12 +1094,12 @@ void test_builtin_ctzg(unsigned char uc, unsigned short us, unsigned int ui,
tz = __builtin_ctzg(uc);
// CHECK: %1 = load i8, ptr %uc.addr, align 1
// CHECK-NEXT: %2 = call i8 @llvm.cttz.i8(i8 %1, i1 true)
- // CHECK-NEXT: %cast = sext i8 %2 to i32
+ // CHECK-NEXT: %cast = zext i8 %2 to i32
// CHECK-NEXT: store volatile i32 %cast, ptr %tz, align 4
tz = __builtin_ctzg(us);
// CHECK-NEXT: %3 = load i16, ptr %us.addr, align 2
// CHECK-NEXT: %4 = call i16 @llvm.cttz.i16(i16 %3, i1 true)
- // CHECK-NEXT: %cast1 = sext i16 %4 to i32
+ // CHECK-NEXT: %cast1 = zext i16 %4 to i32
// CHECK-NEXT: store volatile i32 %cast1, ptr %tz, align 4
tz = __builtin_ctzg(ui);
// CHECK-NEXT: %5 = load i32, ptr %ui.addr, align 4
@@ -1128,7 +1128,7 @@ void test_builtin_ctzg(unsigned char uc, unsigned short us, unsigned int ui,
tz = __builtin_ctzg(uc, sc);
// CHECK-NEXT: %15 = load i8, ptr %uc.addr, align 1
// CHECK-NEXT: %16 = call i8 @llvm.cttz.i8(i8 %15, i1 true)
- // CHECK-NEXT: %cast6 = sext i8 %16 to i32
+ // CHECK-NEXT: %cast6 = zext i8 %16 to i32
// CHECK-NEXT: %iszero = icmp eq i8 %15, 0
// CHECK-NEXT: %17 = load i8, ptr %sc.addr, align 1
// CHECK-NEXT: %conv = sext i8 %17 to i32
@@ -1137,7 +1137,7 @@ void test_builtin_ctzg(unsigned char uc, unsigned short us, unsigned int ui,
tz = __builtin_ctzg(us, uc);
// CHECK-NEXT: %18 = load i16, ptr %us.addr, align 2
// CHECK-NEXT: %19 = call i16 @llvm.cttz.i16(i16 %18, i1 true)
- // CHECK-NEXT: %cast7 = sext i16 %19 to i32
+ // CHECK-NEXT: %cast7 = zext i16 %19 to i32
// CHECK-NEXT: %iszero8 = icmp eq i16 %18, 0
// CHECK-NEXT: %20 = load i8, ptr %uc.addr, align 1
// CHECK-NEXT: %conv9 = zext i8 %20 to i32
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
|
Please add a testcase for i1/i2 (which are the only cases where this actually matters). |
I added tests in a new file (builtins-bitint.c) to show that we get expected results for a target that isn't treating a zero input to ctzg/clzg as poison. |
efriedma-quic
approved these changes
May 2, 2024
Make sure that the result from the popcnt/ctlz/cttz intrinsics is unsigned casted to int, rather than casted as a signed value, when expanding the __builtin_popcountg/__builtin_ctzg/__builtin_clzg builtins. An example would be unsigned _BitInt(1) x = ...; int y = __builtin_popcountg(x); which previously was incorrectly expanded to %1 = call i1 @llvm.ctpop.i1(i1 %0) %cast = sext i1 %1 to i32 Since the input type is generic for those "g" versions of the builtins the intrinsic call may return a value for which the sign bit is set (that could typically for BitInt of size 1 and 2). So we need to emit a zext rather than a sext to avoid negative results.
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.
Make sure that the result from the popcnt/ctlz/cttz intrinsics is unsigned casted to int, rather than casted as a signed value, when expanding the __builtin_popcountg/__builtin_ctzg/__builtin_clzg builtins.
An example would be
unsigned _BitInt(1) x = ...;
int y = __builtin_popcountg(x);
which previously was incorrectly expanded to
%1 = call i1 @llvm.ctpop.i1(i1 %0)
%cast = sext i1 %1 to i32
Since the input type is generic for those "g" versions of the builtins the intrinsic call may return a value for which the sign bit is set (that could typically for BitInt of size 1 and 2). So we need to emit a zext rather than a sext to avoid negative results.