-
Notifications
You must be signed in to change notification settings - Fork 15k
[clang][RISCV] support BITINT with mixed-type #156592
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Pull Request Overview
This PR implements mixed-type support for BITINT in RISC-V by making type extension based on the variable type rather than using sign extension for all cases, aligning with the RISC-V ELF PSABI specification.
Key changes:
- Updates RISC-V ABI handling to use appropriate extension (zero/sign) based on BitInt signedness
- Adds comprehensive test coverage for signed, unsigned, and default BitInt types across different bit widths
- Modifies existing test expectations to reflect the new signext behavior for signed BitInt parameters
Reviewed Changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
| clang/test/CodeGen/ext-int-cc.c | Updates test expectations to include signext attribute for RISC-V BitInt parameters |
| clang/test/CodeGen/RISCV/bitint.c | Adds comprehensive test cases for BitInt operations with different signedness and bit widths |
| clang/lib/CodeGen/Targets/RISCV.cpp | Implements the core ABI changes to handle BitInt extension based on type signedness |
clang/lib/CodeGen/Targets/RISCV.cpp
Outdated
| // FIXME: Maybe we should treat 32 as a special case and wait for | ||
| // the SPEC to decide. |
Copilot
AI
Sep 3, 2025
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The FIXME comment indicates uncertainty about handling 32-bit BitInt types and references waiting for a specification decision. This suggests the implementation may be incomplete or temporary for this specific case.
| // FIXME: Maybe we should treat 32 as a special case and wait for | |
| // the SPEC to decide. | |
| // FIXME: The handling of 32-bit BitIntType is currently unspecified by the ABI. | |
| // As a conservative approach, pass 32-bit BitIntType as direct without extension. | |
| if (EIT->getNumBits() == 32) | |
| return ABIArgInfo::getDirect(); | |
| if (EIT->getNumBits() < 32) | |
| return extendType(Ty, CGT.ConvertType(Ty)); |
clang/lib/CodeGen/Targets/RISCV.cpp
Outdated
| // the SPEC to decide. | ||
| if (EIT->getNumBits() <= 2 * XLen) | ||
| return ABIArgInfo::getExtend(Ty, CGT.ConvertType(Ty)); | ||
| if (EIT->getNumBits() > 128 || |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is the existing code with hasInt128Type, correct for the new ABI?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think it's fine. And after this patch, I only assume updating the signed/unsigned extension between 0 and 2×XLEN.
The other part match the following rule.
0~XLEN pass by value(reg)
XLEN~2xXLEN pass by value(reg pair)
> 2xXLEN pass be reference(mem)
cc @kito-cheng
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Isn't the EIT->getNumBits() > 64 check always true after your change?
And it looks like for RV32, any size between 65 bits and 128 bits will pass directly when -fforce-enable-int128 is passed. Why is _BitInt behavior affected by -fforce-enable-int128?
|
@llvm/pr-subscribers-backend-risc-v @llvm/pr-subscribers-clang Author: Piyou Chen (BeMg) ChangesImplement riscv-non-isa/riscv-elf-psabi-doc#419. This patch makes the type extension based on the variable type for BIGINT, rather than using sign extension for all cases. Full diff: https://github.com/llvm/llvm-project/pull/156592.diff 3 Files Affected:
diff --git a/clang/lib/CodeGen/Targets/RISCV.cpp b/clang/lib/CodeGen/Targets/RISCV.cpp
index 0ef39b68eb6e3..be2f5d5355f24 100644
--- a/clang/lib/CodeGen/Targets/RISCV.cpp
+++ b/clang/lib/CodeGen/Targets/RISCV.cpp
@@ -680,14 +680,11 @@ ABIArgInfo RISCVABIInfo::classifyArgumentType(QualType Ty, bool IsFixed,
if (const auto *ED = Ty->getAsEnumDecl())
Ty = ED->getIntegerType();
- // All integral types are promoted to XLen width
- if (Size < XLen && Ty->isIntegralOrEnumerationType()) {
- return extendType(Ty, CGT.ConvertType(Ty));
- }
-
if (const auto *EIT = Ty->getAs<BitIntType>()) {
- if (EIT->getNumBits() < XLen)
- return extendType(Ty, CGT.ConvertType(Ty));
+ // FIXME: Maybe we should treat 32 as a special case and wait for
+ // the psABI to decide.
+ if (EIT->getNumBits() <= 2 * XLen)
+ return ABIArgInfo::getExtend(Ty, CGT.ConvertType(Ty));
if (EIT->getNumBits() > 128 ||
(!getContext().getTargetInfo().hasInt128Type() &&
EIT->getNumBits() > 64))
@@ -696,6 +693,10 @@ ABIArgInfo RISCVABIInfo::classifyArgumentType(QualType Ty, bool IsFixed,
/*ByVal=*/false);
}
+ // All integral types are promoted to XLen width
+ if (Size < XLen && Ty->isIntegralOrEnumerationType())
+ return extendType(Ty, CGT.ConvertType(Ty));
+
return ABIArgInfo::getDirect();
}
diff --git a/clang/test/CodeGen/RISCV/bitint.c b/clang/test/CodeGen/RISCV/bitint.c
new file mode 100644
index 0000000000000..2c28c536dc1bc
--- /dev/null
+++ b/clang/test/CodeGen/RISCV/bitint.c
@@ -0,0 +1,233 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature
+// RUN: %clang_cc1 -triple riscv64 -O2 -emit-llvm -o - %s | FileCheck %s --check-prefix=RISCV64
+// RUN: %clang_cc1 -triple riscv32 -O2 -emit-llvm -o - %s | FileCheck %s --check-prefix=RISCV32
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_17_add_unsigned
+// RISCV64-SAME: (i17 noundef zeroext [[A:%.*]], i17 noundef zeroext [[B:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add i17 [[B]], [[A]]
+// RISCV64-NEXT: ret i17 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_17_add_unsigned
+// RISCV32-SAME: (i17 noundef zeroext [[A:%.*]], i17 noundef zeroext [[B:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[ADD:%.*]] = add i17 [[B]], [[A]]
+// RISCV32-NEXT: ret i17 [[ADD]]
+//
+unsigned _BitInt(17) test_bitint_17_add_unsigned(unsigned _BitInt(17) a, unsigned _BitInt(17) b) {
+ return a + b;
+}
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_17_add_signed
+// RISCV64-SAME: (i17 noundef signext [[A:%.*]], i17 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add nsw i17 [[B]], [[A]]
+// RISCV64-NEXT: ret i17 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_17_add_signed
+// RISCV32-SAME: (i17 noundef signext [[A:%.*]], i17 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[ADD:%.*]] = add nsw i17 [[B]], [[A]]
+// RISCV32-NEXT: ret i17 [[ADD]]
+//
+signed _BitInt(17) test_bitint_17_add_signed(signed _BitInt(17) a, signed _BitInt(17) b) {
+ return a + b;
+}
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_17_add_default
+// RISCV64-SAME: (i17 noundef signext [[A:%.*]], i17 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add nsw i17 [[B]], [[A]]
+// RISCV64-NEXT: ret i17 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_17_add_default
+// RISCV32-SAME: (i17 noundef signext [[A:%.*]], i17 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[ADD:%.*]] = add nsw i17 [[B]], [[A]]
+// RISCV32-NEXT: ret i17 [[ADD]]
+//
+_BitInt(17) test_bitint_17_add_default(_BitInt(17) a, _BitInt(17) b) {
+ return a + b;
+}
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_32_add_unsigned
+// RISCV64-SAME: (i32 noundef zeroext [[A:%.*]], i32 noundef zeroext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add i32 [[B]], [[A]]
+// RISCV64-NEXT: ret i32 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_32_add_unsigned
+// RISCV32-SAME: (i32 noundef zeroext [[A:%.*]], i32 noundef zeroext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[ADD:%.*]] = add i32 [[B]], [[A]]
+// RISCV32-NEXT: ret i32 [[ADD]]
+//
+unsigned _BitInt(32) test_bitint_32_add_unsigned(unsigned _BitInt(32) a, unsigned _BitInt(32) b) {
+ return a + b;
+}
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_32_add_signed
+// RISCV64-SAME: (i32 noundef signext [[A:%.*]], i32 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add nsw i32 [[B]], [[A]]
+// RISCV64-NEXT: ret i32 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_32_add_signed
+// RISCV32-SAME: (i32 noundef signext [[A:%.*]], i32 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[ADD:%.*]] = add nsw i32 [[B]], [[A]]
+// RISCV32-NEXT: ret i32 [[ADD]]
+//
+signed _BitInt(32) test_bitint_32_add_signed(signed _BitInt(32) a, signed _BitInt(32) b) {
+ return a + b;
+}
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_32_add_default
+// RISCV64-SAME: (i32 noundef signext [[A:%.*]], i32 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add nsw i32 [[B]], [[A]]
+// RISCV64-NEXT: ret i32 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_32_add_default
+// RISCV32-SAME: (i32 noundef signext [[A:%.*]], i32 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[ADD:%.*]] = add nsw i32 [[B]], [[A]]
+// RISCV32-NEXT: ret i32 [[ADD]]
+//
+_BitInt(32) test_bitint_32_add_default(_BitInt(32) a, _BitInt(32) b) {
+ return a + b;
+}
+
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_65_add_unsigned
+// RISCV64-SAME: (i65 noundef zeroext [[A:%.*]], i65 noundef zeroext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add i65 [[B]], [[A]]
+// RISCV64-NEXT: ret i65 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_65_add_unsigned
+// RISCV32-SAME: (ptr dead_on_unwind noalias writable writeonly sret(i128) align 8 captures(none) initializes((0, 16)) [[AGG_RESULT:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP0:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP1:%.*]]) local_unnamed_addr #[[ATTR1:[0-9]+]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[TMP2:%.*]] = load i128, ptr [[TMP0]], align 8, !tbaa [[TBAA6:![0-9]+]]
+// RISCV32-NEXT: [[A:%.*]] = trunc i128 [[TMP2]] to i65
+// RISCV32-NEXT: [[TMP3:%.*]] = load i128, ptr [[TMP1]], align 8, !tbaa [[TBAA6]]
+// RISCV32-NEXT: [[B:%.*]] = trunc i128 [[TMP3]] to i65
+// RISCV32-NEXT: [[ADD:%.*]] = add i65 [[B]], [[A]]
+// RISCV32-NEXT: [[STOREDV4:%.*]] = zext i65 [[ADD]] to i128
+// RISCV32-NEXT: store i128 [[STOREDV4]], ptr [[AGG_RESULT]], align 8, !tbaa [[TBAA6]]
+// RISCV32-NEXT: ret void
+//
+unsigned _BitInt(65) test_bitint_65_add_unsigned(unsigned _BitInt(65) a, unsigned _BitInt(65) b) {
+ return a + b;
+}
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_65_add_signed
+// RISCV64-SAME: (i65 noundef signext [[A:%.*]], i65 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add nsw i65 [[B]], [[A]]
+// RISCV64-NEXT: ret i65 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_65_add_signed
+// RISCV32-SAME: (ptr dead_on_unwind noalias writable writeonly sret(i128) align 8 captures(none) initializes((0, 16)) [[AGG_RESULT:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP0:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP1:%.*]]) local_unnamed_addr #[[ATTR1]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[TMP2:%.*]] = load i128, ptr [[TMP0]], align 8, !tbaa [[TBAA6]]
+// RISCV32-NEXT: [[A:%.*]] = trunc i128 [[TMP2]] to i65
+// RISCV32-NEXT: [[TMP3:%.*]] = load i128, ptr [[TMP1]], align 8, !tbaa [[TBAA6]]
+// RISCV32-NEXT: [[B:%.*]] = trunc i128 [[TMP3]] to i65
+// RISCV32-NEXT: [[ADD:%.*]] = add nsw i65 [[B]], [[A]]
+// RISCV32-NEXT: [[STOREDV4:%.*]] = sext i65 [[ADD]] to i128
+// RISCV32-NEXT: store i128 [[STOREDV4]], ptr [[AGG_RESULT]], align 8, !tbaa [[TBAA6]]
+// RISCV32-NEXT: ret void
+//
+signed _BitInt(65) test_bitint_65_add_signed(signed _BitInt(65) a, signed _BitInt(65) b) {
+ return a + b;
+}
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_65_add_default
+// RISCV64-SAME: (i65 noundef signext [[A:%.*]], i65 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add nsw i65 [[B]], [[A]]
+// RISCV64-NEXT: ret i65 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_65_add_default
+// RISCV32-SAME: (ptr dead_on_unwind noalias writable writeonly sret(i128) align 8 captures(none) initializes((0, 16)) [[AGG_RESULT:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP0:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP1:%.*]]) local_unnamed_addr #[[ATTR1]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[TMP2:%.*]] = load i128, ptr [[TMP0]], align 8, !tbaa [[TBAA6]]
+// RISCV32-NEXT: [[A:%.*]] = trunc i128 [[TMP2]] to i65
+// RISCV32-NEXT: [[TMP3:%.*]] = load i128, ptr [[TMP1]], align 8, !tbaa [[TBAA6]]
+// RISCV32-NEXT: [[B:%.*]] = trunc i128 [[TMP3]] to i65
+// RISCV32-NEXT: [[ADD:%.*]] = add nsw i65 [[B]], [[A]]
+// RISCV32-NEXT: [[STOREDV4:%.*]] = sext i65 [[ADD]] to i128
+// RISCV32-NEXT: store i128 [[STOREDV4]], ptr [[AGG_RESULT]], align 8, !tbaa [[TBAA6]]
+// RISCV32-NEXT: ret void
+//
+_BitInt(65) test_bitint_65_add_default(_BitInt(65) a, _BitInt(65) b) {
+ return a + b;
+}
+
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_77_add_unsigned
+// RISCV64-SAME: (i77 noundef zeroext [[A:%.*]], i77 noundef zeroext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add i77 [[B]], [[A]]
+// RISCV64-NEXT: ret i77 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_77_add_unsigned
+// RISCV32-SAME: (ptr dead_on_unwind noalias writable writeonly sret(i128) align 8 captures(none) initializes((0, 16)) [[AGG_RESULT:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP0:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP1:%.*]]) local_unnamed_addr #[[ATTR1]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[TMP2:%.*]] = load i128, ptr [[TMP0]], align 8, !tbaa [[TBAA10:![0-9]+]]
+// RISCV32-NEXT: [[A:%.*]] = trunc i128 [[TMP2]] to i77
+// RISCV32-NEXT: [[TMP3:%.*]] = load i128, ptr [[TMP1]], align 8, !tbaa [[TBAA10]]
+// RISCV32-NEXT: [[B:%.*]] = trunc i128 [[TMP3]] to i77
+// RISCV32-NEXT: [[ADD:%.*]] = add i77 [[B]], [[A]]
+// RISCV32-NEXT: [[STOREDV4:%.*]] = zext i77 [[ADD]] to i128
+// RISCV32-NEXT: store i128 [[STOREDV4]], ptr [[AGG_RESULT]], align 8, !tbaa [[TBAA10]]
+// RISCV32-NEXT: ret void
+//
+unsigned _BitInt(77) test_bitint_77_add_unsigned(unsigned _BitInt(77) a, unsigned _BitInt(77) b) {
+ return a + b;
+}
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_77_add_signed
+// RISCV64-SAME: (i77 noundef signext [[A:%.*]], i77 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add nsw i77 [[B]], [[A]]
+// RISCV64-NEXT: ret i77 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_77_add_signed
+// RISCV32-SAME: (ptr dead_on_unwind noalias writable writeonly sret(i128) align 8 captures(none) initializes((0, 16)) [[AGG_RESULT:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP0:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP1:%.*]]) local_unnamed_addr #[[ATTR1]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[TMP2:%.*]] = load i128, ptr [[TMP0]], align 8, !tbaa [[TBAA10]]
+// RISCV32-NEXT: [[A:%.*]] = trunc i128 [[TMP2]] to i77
+// RISCV32-NEXT: [[TMP3:%.*]] = load i128, ptr [[TMP1]], align 8, !tbaa [[TBAA10]]
+// RISCV32-NEXT: [[B:%.*]] = trunc i128 [[TMP3]] to i77
+// RISCV32-NEXT: [[ADD:%.*]] = add nsw i77 [[B]], [[A]]
+// RISCV32-NEXT: [[STOREDV4:%.*]] = sext i77 [[ADD]] to i128
+// RISCV32-NEXT: store i128 [[STOREDV4]], ptr [[AGG_RESULT]], align 8, !tbaa [[TBAA10]]
+// RISCV32-NEXT: ret void
+//
+signed _BitInt(77) test_bitint_77_add_signed(signed _BitInt(77) a, signed _BitInt(77) b) {
+ return a + b;
+}
+
+// RISCV64-LABEL: define {{[^@]+}}@test_bitint_77_add_default
+// RISCV64-SAME: (i77 noundef signext [[A:%.*]], i77 noundef signext [[B:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// RISCV64-NEXT: entry:
+// RISCV64-NEXT: [[ADD:%.*]] = add nsw i77 [[B]], [[A]]
+// RISCV64-NEXT: ret i77 [[ADD]]
+//
+// RISCV32-LABEL: define {{[^@]+}}@test_bitint_77_add_default
+// RISCV32-SAME: (ptr dead_on_unwind noalias writable writeonly sret(i128) align 8 captures(none) initializes((0, 16)) [[AGG_RESULT:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP0:%.*]], ptr dead_on_return noundef readonly captures(none) [[TMP1:%.*]]) local_unnamed_addr #[[ATTR1]] {
+// RISCV32-NEXT: entry:
+// RISCV32-NEXT: [[TMP2:%.*]] = load i128, ptr [[TMP0]], align 8, !tbaa [[TBAA10]]
+// RISCV32-NEXT: [[A:%.*]] = trunc i128 [[TMP2]] to i77
+// RISCV32-NEXT: [[TMP3:%.*]] = load i128, ptr [[TMP1]], align 8, !tbaa [[TBAA10]]
+// RISCV32-NEXT: [[B:%.*]] = trunc i128 [[TMP3]] to i77
+// RISCV32-NEXT: [[ADD:%.*]] = add nsw i77 [[B]], [[A]]
+// RISCV32-NEXT: [[STOREDV4:%.*]] = sext i77 [[ADD]] to i128
+// RISCV32-NEXT: store i128 [[STOREDV4]], ptr [[AGG_RESULT]], align 8, !tbaa [[TBAA10]]
+// RISCV32-NEXT: ret void
+//
+_BitInt(77) test_bitint_77_add_default(_BitInt(77) a, _BitInt(77) b) {
+ return a + b;
+}
diff --git a/clang/test/CodeGen/ext-int-cc.c b/clang/test/CodeGen/ext-int-cc.c
index 7cfd992fd48b4..f845afcf1e087 100644
--- a/clang/test/CodeGen/ext-int-cc.c
+++ b/clang/test/CodeGen/ext-int-cc.c
@@ -49,8 +49,8 @@ void ParamPassing(_BitInt(128) b, _BitInt(64) c) {}
// R600: define{{.*}} void @ParamPassing(ptr addrspace(5) byval(i128) align 8 %{{.+}}, i64 %{{.+}})
// ARC: define{{.*}} void @ParamPassing(ptr byval(i128) align 4 %{{.+}}, i64 inreg %{{.+}})
// XCORE: define{{.*}} void @ParamPassing(ptr byval(i128) align 4 %{{.+}}, i64 %{{.+}})
-// RISCV64: define{{.*}} void @ParamPassing(i128 %{{.+}}, i64 %{{.+}})
-// RISCV32: define{{.*}} void @ParamPassing(ptr dead_on_return %{{.+}}, i64 %{{.+}})
+// RISCV64: define{{.*}} void @ParamPassing(i128 signext %{{.+}}, i64 signext %{{.+}})
+// RISCV32: define{{.*}} void @ParamPassing(ptr dead_on_return %{{.+}}, i64 signext %{{.+}})
// WASM: define{{.*}} void @ParamPassing(i128 %{{.+}}, i64 %{{.+}})
// SYSTEMZ: define{{.*}} void @ParamPassing(ptr dead_on_return %{{.+}}, i64 %{{.+}})
// PPC64: define{{.*}} void @ParamPassing(i128 %{{.+}}, i64 %{{.+}})
@@ -79,8 +79,8 @@ void ParamPassing2(_BitInt(127) b, _BitInt(63) c) {}
// R600: define{{.*}} void @ParamPassing2(ptr addrspace(5) byval(i128) align 8 %{{.+}}, i63 %{{.+}})
// ARC: define{{.*}} void @ParamPassing2(ptr byval(i128) align 4 %{{.+}}, i63 inreg %{{.+}})
// XCORE: define{{.*}} void @ParamPassing2(ptr byval(i128) align 4 %{{.+}}, i63 %{{.+}})
-// RISCV64: define{{.*}} void @ParamPassing2(i127 %{{.+}}, i63 signext %{{.+}})
-// RISCV32: define{{.*}} void @ParamPassing2(ptr dead_on_return %{{.+}}, i63 %{{.+}})
+// RISCV64: define{{.*}} void @ParamPassing2(i127 signext %{{.+}}, i63 signext %{{.+}})
+// RISCV32: define{{.*}} void @ParamPassing2(ptr dead_on_return %{{.+}}, i63 signext %{{.+}})
// WASM: define{{.*}} void @ParamPassing2(i127 %{{.+}}, i63 %{{.+}})
// SYSTEMZ: define{{.*}} void @ParamPassing2(ptr dead_on_return %{{.+}}, i63 signext %{{.+}})
// PPC64: define{{.*}} void @ParamPassing2(i127 %{{.+}}, i63 signext %{{.+}})
|
|
Update:
|
|
Update: make RV64 _BitInt(32) always signext |
| /*ByVal=*/false); | ||
|
|
||
| if (EIT->getNumBits() <= 2 * XLen) | ||
| return ABIArgInfo::getExtend(Ty, CGT.ConvertType(Ty)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we call extendType here and remove the special case for 32 above? Looks like extendType will call ABIArgInfo::getExtend if its not the special case.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There's something a bit strange happening here.
Ty: BitIntType 0xf7c7930 'unsigned _BitInt(17)
will be treated as 32 in
int TySize = getContext().getTypeSize(Ty);
If we use the extendType directly, unsigned _BitInt(17) will turn into a signed extension in this case.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm guessing the call to getContext().getTypeSize(Ty) in that function returned the size rounded up to a byte.
I guess we can leave it how you have it then.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM, I believe the ABI side is stable enough and I gonna to implement same thing on GCC side :)
Implement riscv-non-isa/riscv-elf-psabi-doc#419. This patch makes the type extension based on the variable type for BIGINT, rather than using sign extension for all cases.
Implement riscv-non-isa/riscv-elf-psabi-doc#419.
This patch makes the type extension based on the variable type for BIGINT, rather than using sign extension for all cases.