[Hexagon] Omit calls to specialized {float,fix} routines #117423
Conversation
This file contains 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
|
@llvm/pr-subscribers-backend-hexagon Author: Brian Cain (androm3da) ChangesThese were introduced in 1213a7a (Hexagon backend support, 2011-12-12) but they aren't present in libclangrt.builtins-hexagon. The generic versions of these functions are present in the builtins, though. So it should suffice to call those instead. Full diff: https://github.com/llvm/llvm-project/pull/117423.diff 2 Files Affected:
diff --git a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
index 2c8d141aa21080..5dc2b35616ed2c 100644
--- a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
@@ -1860,13 +1860,6 @@ HexagonTargetLowering::HexagonTargetLowering(const TargetMachine &TM,
setLibcallName(RTLIB::UREM_I32, "__hexagon_umodsi3");
setLibcallName(RTLIB::UREM_I64, "__hexagon_umoddi3");
- setLibcallName(RTLIB::SINTTOFP_I128_F64, "__hexagon_floattidf");
- setLibcallName(RTLIB::SINTTOFP_I128_F32, "__hexagon_floattisf");
- setLibcallName(RTLIB::FPTOUINT_F32_I128, "__hexagon_fixunssfti");
- setLibcallName(RTLIB::FPTOUINT_F64_I128, "__hexagon_fixunsdfti");
- setLibcallName(RTLIB::FPTOSINT_F32_I128, "__hexagon_fixsfti");
- setLibcallName(RTLIB::FPTOSINT_F64_I128, "__hexagon_fixdfti");
-
// This is the only fast library function for sqrtd.
if (FastMath)
setLibcallName(RTLIB::SQRT_F64, "__hexagon_fast2_sqrtdf2");
diff --git a/llvm/test/CodeGen/Hexagon/i128-fpconv-strict.ll b/llvm/test/CodeGen/Hexagon/i128-fpconv-strict.ll
new file mode 100644
index 00000000000000..f643c0f2a5f77a
--- /dev/null
+++ b/llvm/test/CodeGen/Hexagon/i128-fpconv-strict.ll
@@ -0,0 +1,139 @@
+; RUN: llc < %s -mtriple=hexagon-unknown-CHECK-musl \
+; RUN: | FileCheck %s -check-prefix=CHECK
+
+define i64 @double_to_i128(double %d) nounwind strictfp {
+; CHECK-LABEL: double_to_i128:
+; CHECK: // %bb.0:
+; CHECK: call __fixdfti
+; CHECK: dealloc_return
+ %1 = tail call i128 @llvm.experimental.constrained.fptosi.i128.f64(double %d, metadata !"fpexcept.strict")
+ %2 = trunc i128 %1 to i64
+ ret i64 %2
+}
+
+define i64 @double_to_ui128(double %d) nounwind strictfp {
+; CHECK-LABEL: double_to_ui128:
+; CHECK: // %bb.0:
+; CHECK: call __fixunsdfti
+; CHECK: dealloc_return
+ %1 = tail call i128 @llvm.experimental.constrained.fptoui.i128.f64(double %d, metadata !"fpexcept.strict")
+ %2 = trunc i128 %1 to i64
+ ret i64 %2
+}
+
+define i64 @float_to_i128(float %d) nounwind strictfp {
+; CHECK-LABEL: float_to_i128:
+; CHECK: // %bb.0:
+; CHECK: call __fixsfti
+; CHECK: dealloc_return
+ %1 = tail call i128 @llvm.experimental.constrained.fptosi.i128.f32(float %d, metadata !"fpexcept.strict")
+ %2 = trunc i128 %1 to i64
+ ret i64 %2
+}
+
+define i64 @float_to_ui128(float %d) nounwind strictfp {
+; CHECK-LABEL: float_to_ui128:
+; CHECK: // %bb.0:
+; CHECK: call __fixunssfti
+; CHECK: dealloc_return
+ %1 = tail call i128 @llvm.experimental.constrained.fptoui.i128.f32(float %d, metadata !"fpexcept.strict")
+ %2 = trunc i128 %1 to i64
+ ret i64 %2
+}
+
+define i64 @longdouble_to_i128(ptr nocapture readonly %0) nounwind strictfp {
+; CHECK-LABEL: longdouble_to_i128:
+; CHECK: // %bb.0:
+; CHECK: call __fixxfti
+; CHECK: dealloc_return
+ %2 = load x86_fp80, ptr %0, align 16
+ %3 = tail call i128 @llvm.experimental.constrained.fptosi.i128.f80(x86_fp80 %2, metadata !"fpexcept.strict")
+ %4 = trunc i128 %3 to i64
+ ret i64 %4
+}
+
+define i64 @longdouble_to_ui128(ptr nocapture readonly %0) nounwind strictfp {
+; CHECK-LABEL: longdouble_to_ui128:
+; CHECK: // %bb.0:
+; CHECK: call __fixunsxfti
+; CHECK: dealloc_return
+ %2 = load x86_fp80, ptr %0, align 16
+ %3 = tail call i128 @llvm.experimental.constrained.fptoui.i128.f80(x86_fp80 %2, metadata !"fpexcept.strict")
+ %4 = trunc i128 %3 to i64
+ ret i64 %4
+}
+
+define double @i128_to_double(ptr nocapture readonly %0) nounwind strictfp {
+; CHECK-LABEL: i128_to_double:
+; CHECK: // %bb.0:
+; CHECK: call __floattidf
+; CHECK: dealloc_return
+ %2 = load i128, ptr %0, align 16
+ %3 = tail call double @llvm.experimental.constrained.sitofp.f64.i128(i128 %2, metadata !"round.dynamic", metadata !"fpexcept.strict")
+ ret double %3
+}
+
+define double @ui128_to_double(ptr nocapture readonly %0) nounwind strictfp {
+; CHECK-LABEL: ui128_to_double:
+; CHECK: // %bb.0:
+; CHECK: call __floatuntidf
+; CHECK: dealloc_return
+ %2 = load i128, ptr %0, align 16
+ %3 = tail call double @llvm.experimental.constrained.uitofp.f64.i128(i128 %2, metadata !"round.dynamic", metadata !"fpexcept.strict")
+ ret double %3
+}
+
+define float @i128_to_float(ptr nocapture readonly %0) nounwind strictfp {
+; CHECK-LABEL: i128_to_float:
+; CHECK: // %bb.0:
+; CHECK: call __floattisf
+; CHECK: dealloc_return
+ %2 = load i128, ptr %0, align 16
+ %3 = tail call float @llvm.experimental.constrained.sitofp.f32.i128(i128 %2, metadata !"round.dynamic", metadata !"fpexcept.strict")
+ ret float %3
+}
+
+define float @ui128_to_float(ptr nocapture readonly %0) nounwind strictfp {
+; CHECK-LABEL: ui128_to_float:
+; CHECK: // %bb.0:
+; CHECK: call __floatuntisf
+; CHECK: dealloc_return
+ %2 = load i128, ptr %0, align 16
+ %3 = tail call float @llvm.experimental.constrained.uitofp.f32.i128(i128 %2, metadata !"round.dynamic", metadata !"fpexcept.strict")
+ ret float %3
+}
+
+define void @i128_to_longdouble(ptr noalias nocapture sret(x86_fp80) align 16 %agg.result, ptr nocapture readonly %0) nounwind strictfp {
+; CHECK-LABEL: i128_to_longdouble:
+; CHECK: // %bb.0:
+; CHECK: call __floattixf
+; CHECK: dealloc_return
+ %2 = load i128, ptr %0, align 16
+ %3 = tail call x86_fp80 @llvm.experimental.constrained.sitofp.f80.i128(i128 %2, metadata !"round.dynamic", metadata !"fpexcept.strict")
+ store x86_fp80 %3, ptr %agg.result, align 16
+ ret void
+}
+
+define void @ui128_to_longdouble(ptr noalias nocapture sret(x86_fp80) align 16 %agg.result, ptr nocapture readonly %0) nounwind strictfp {
+; CHECK-LABEL: ui128_to_longdouble:
+; CHECK: // %bb.0:
+; CHECK: call __floatuntixf
+; CHECK: dealloc_return
+ %2 = load i128, ptr %0, align 16
+ %3 = tail call x86_fp80 @llvm.experimental.constrained.uitofp.f80.i128(i128 %2, metadata !"round.dynamic", metadata !"fpexcept.strict")
+ store x86_fp80 %3, ptr %agg.result, align 16
+ ret void
+}
+
+declare i128 @llvm.experimental.constrained.fptosi.i128.f64(double, metadata)
+declare i128 @llvm.experimental.constrained.fptoui.i128.f64(double, metadata)
+declare i128 @llvm.experimental.constrained.fptosi.i128.f32(float, metadata)
+declare i128 @llvm.experimental.constrained.fptoui.i128.f32(float, metadata)
+declare i128 @llvm.experimental.constrained.fptosi.i128.f80(x86_fp80, metadata)
+declare i128 @llvm.experimental.constrained.fptoui.i128.f80(x86_fp80, metadata)
+declare double @llvm.experimental.constrained.sitofp.f64.i128(i128, metadata, metadata)
+declare double @llvm.experimental.constrained.uitofp.f64.i128(i128, metadata, metadata)
+declare float @llvm.experimental.constrained.sitofp.f32.i128(i128, metadata, metadata)
+declare float @llvm.experimental.constrained.uitofp.f32.i128(i128, metadata, metadata)
+declare x86_fp80 @llvm.experimental.constrained.sitofp.f80.i128(i128, metadata, metadata)
+declare x86_fp80 @llvm.experimental.constrained.uitofp.f80.i128(i128, metadata, metadata)
|
|
@SundeepKushwaha these functions are provided in Also: I think it's the case that This issue was raised while contributing hexagon specializations for other builtins to zig - see ziglang/zig#22029 |
These were introduced in 1213a7a (Hexagon backend support, 2011-12-12) but they aren't present in libclangrt.builtins-hexagon. The generic versions of these functions are present in the builtins, though. So it should suffice to call those instead.
androm3da
force-pushed
the
bcain/128_fltint_spec
branch
from
47c4f54
to
0580f3f
Compare
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.
These were introduced in 1213a7a (Hexagon backend support, 2011-12-12) but they aren't present in libclangrt.builtins-hexagon. The generic versions of these functions are present in the builtins, though. So it should suffice to call those instead.