JIT, interpreter, AOT preinit, and CoreLib: Saturate float/double conversions to small integral types by Copilot · Pull Request #128604 · dotnet/runtime

Copilot · 2026-05-26T20:10:07Z

Unchecked float/double → small integral type (sbyte, byte, short, ushort, char) conversions did not saturate at the small-type boundary. For example, (short)(double)32768.000000000007 produced -32768 instead of 32767, diverging from the saturating R → int contract introduced in .NET 9.

The JIT expands R → smallT as R → int → smallT. The inner R → int saturates, but the outer int → smallT only truncates low bits. The CoreCLR interpreter and the ILC type preinitializer had analogous bugs and were updated for parity with runtime behavior. In addition, CoreLib TryConvert paths for float/double to small integral types were updated to use direct casts on CoreCLR (while preserving existing MONO behavior under #if MONO).

Summary

Fixes saturating semantics for unchecked float/double → small integral type casts across all supported JIT targets (xarch, x86, arm64, RISC-V64, LoongArch64, ARM32, and WASM), the CoreCLR interpreter, and the NativeAOT type preinitializer, and applies corresponding CoreCLR-side TryConvert simplifications that rely on the corrected cast semantics.

Changes Made

src/coreclr/jit/morph.cpp — fgMorphExpandCast: Added target-specific handling so R -> small saturates at destination bounds across supported architectures:
- FEATURE_HW_INTRINSICS targets use float-domain min/max clamping before existing cast expansion.
- WASM lowers min/max via GT_INTRINSIC.
- ARM32, RISC-V64, and LoongArch64 use integer-domain saturation (R -> int32 followed by NI_PRIMITIVE_SaturateTo{Int8,Int16,UInt8,UInt16}), preserving NaN→0 semantics from the saturating R → int cast.
JIT plumbing for the new NI_PRIMITIVE_SaturateTo{Int8,Int16,UInt8,UInt16} intrinsics and the WASM binary GT_INTRINSIC(MaxNative/MinNative) produced by morph:
- gentree.cpp: added the four SaturateTo* cases to the unary GT_INTRINSIC cost switch in gtSetEvalOrder, added a TARGET_WASM block to the binary GT_INTRINSIC cost switch for MaxNative/MinNative, and added dump labels for the four SaturateTo* IDs (fixes the 'Unknown binary GT_INTRINSIC operator' assert seen on e.g. System.Half:op_Explicit(System.Half):byte).
- importercalls.cpp IsTargetIntrinsic: added the four SaturateTo* IDs to the ARM, RISCV64, and LoongArch64 switches so rationalize.cpp's IsTargetIntrinsic assertion is satisfied.
- valuenum.cpp fgValueNumberIntrinsic: added an explicit branch for the SaturateTo* IDs that gives the node an opaque unary VN, backed by new VNF_SaturateToInt8/Int16/UInt8/UInt16 entries in valuenumfuncs.h, so VN no longer hits the assert(NI_System_Object_GetType) for these intrinsics.
- assertionprop.cpp IntegralRange::ForNode: taught range analysis the exact result range for each SaturateTo* intrinsic ([ByteMin, ByteMax], [ShortMin, ShortMax], [0, UByteMax], [0, UShortMax]).
ARM32 codegen for SaturateTo*: new INS_ssat/INS_usat Thumb-2 instructions in instrsarm.h, encoding/disassembly support in emitarm.cpp, LSRA/codegen wiring in lsraarm.cpp and codegenarmarch.cpp.
RISC-V64 / LoongArch64 codegen for SaturateTo*: branch-based clamp in codegenriscv64.cpp / codegenloongarch64.cpp, with LSRA support in lsrariscv64.cpp / lsraloongarch64.cpp.
src/coreclr/vm/interpexec.cpp — ConvFpHelper: clamps to numeric_limits<TResult> (the destination type) instead of TIntermediate, so floating-point → small-integral conversions now saturate at the destination type's range.
src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/TypePreinit.cs: float-domain conv_i1/conv_i2/conv_u1/conv_u2 cases now explicitly saturate via Math.Clamp (NaN → 0), so AOT-baked statics match runtime behavior regardless of which host JIT runs ILC. Original simple-cast version is kept commented out with a TODO to restore once the fix has propagated through the toolchain.
CoreLib TryConvert simplifications (Double.cs, Single.cs, SByte.cs, Int16.cs): float/double → small-integral TryConvertTo* / TryConvertFrom* paths now use a direct cast on CoreCLR, with the existing manual clamp preserved under #if MONO.
Tests: re-enabled previously-disabled DoubleTests.GenericMath.ConvertToIntegerTest / SingleTests.GenericMath.ConvertToIntegerTest (issue (short)(double)32768.000000000007 produces invalid result #116823 suppressions removed), added a JitBlue regression test for the float/double → small-integral saturation contract, and updated the interpreter test comment/coverage.

Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

dotnet-policy-service · 2026-05-26T20:37:19Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

…ll casts Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

…ing casts Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 20 out of 20 changed files in this pull request and generated 2 comments.

+            // Both operands and result must be of the same floating point type.
            GenTree* op1 = tree->gtGetOp1();
+            GenTree* op2 = tree->gtGetOp2IfPresent();
            assert(varTypeIsFloating(op1));
            assert(op1->TypeGet() == tree->TypeGet());


RISC-V64 and LoongArch64 fmin/fmax implement IEEE 754-2019 "minimumNumber"/"maximumNumber" semantics which replace NaN with the other operand (rather than propagating it). Using MaxNative/MinNative in the float domain would map NaN inputs to smallMin instead of 0. Fix by moving RISC-V64 and LoongArch64 to the same integer-domain saturation approach as ARM32: do R->int first (NaN->0 via the saturating cast), then apply NI_PRIMITIVE_SaturateToInt8/16/UInt8/16 intrinsics that lower to branch-based clamp sequences using bge/mov/instGen_Set_Reg_To_Imm. - morph.cpp: Move RISC-V64/LA64 from float-domain (#if FEATURE_HW_INTRINSICS||WASM) to integer-domain (#elif TARGET_ARM||RISCV64||LOONGARCH64) block - lsrariscv64.cpp: Add saturation intrinsics to GT_INTRINSIC handler; allocate temp register - codegenriscv64.cpp: Implement branch-based clamp in genIntrinsic for saturation intrinsics - lsraloongarch64.cpp: Add saturation intrinsics to GT_INTRINSIC handler; allocate temp register - codegenloongarch64.cpp: Implement branch-based clamp in genIntrinsic for saturation intrinsics Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 22 out of 22 changed files in this pull request and generated no new comments.

Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

tannergooding · 2026-06-05T00:25:17Z

@copilot, some builds are failing with asserts similar to:

#2665 D:\a\_work\1\s\src\coreclr\jit\gentree.cpp (7244) - Assertion failed '!"Unknown binary GT_INTRINSIC operator"' in 'System.Half:op_Explicit(System.Half):byte' during 'Find oper order' (IL size 9; hash 0x039d4a98; FullOpts)

This is because the new intrinsics aren't being handled under all the applicable paths. You should look at something like NI_System_Math_Abs or NI_PRIMITIVE_LeadingZeroCount to find the various places that need the new IDs handled.

Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

Copilot · 2026-06-05T00:50:21Z

@copilot, some builds are failing with asserts similar to:
#2665 D:\a\_work\1\s\src\coreclr\jit\gentree.cpp (7244) - Assertion failed '!"Unknown binary GT_INTRINSIC operator"' in...

Done. Audited the paths used by NI_PRIMITIVE_LeadingZeroCount / NI_System_Math_Abs and added handling for NI_PRIMITIVE_SaturateTo{Int8,Int16,UInt8,UInt16} (and the WASM binary MaxNative/MinNative from morph) in the spots that were asserting:

gentree.cpp gtSetEvalOrder — added the four SaturateTo* cases to the unary GT_INTRINSIC cost switch (next to NI_System_Math_Abs), and added a #elif defined(TARGET_WASM) block to the binary GT_INTRINSIC cost switch for NI_System_Math_MaxNative / MinNative (this was the actual 'Unknown binary GT_INTRINSIC operator' site, since on WASM morph creates a binary GT_INTRINSIC for the float-domain clamp).
gentree.cpp GTF dump — added dump labels for the four SaturateTo* IDs.
importercalls.cpp IsTargetIntrinsic — added the four SaturateTo* IDs to the ARM, RISCV64, and LoongArch64 switches so rationalize.cpp's assert(IsTargetIntrinsic(...)) doesn't fire.
valuenum.cpp fgValueNumberIntrinsic — added an explicit branch for the four SaturateTo* IDs that gives the node an opaque unary VN function (new VNF_SaturateToInt8/Int16/UInt8/UInt16 in valuenumfuncs.h), so the else assert(NI_System_Object_GetType) no longer fires for them.
assertionprop.cpp IntegralRange::ForNode — taught range analysis the exact result range for each SaturateTo* ([ByteMin, ByteMax], [ShortMin, ShortMax], [0, UByteMax], [0, UShortMax]).

Verified by building clr.alljits for x64 (host), and cross-jit builds for loongarch64, riscv64, and the WASM cross-jit (clrjit_universal_wasm_x64); all green with no warnings.

Copilot

Pull request overview

Copilot reviewed 26 out of 26 changed files in this pull request and generated 5 comments.

Copilot · 2026-06-05T03:58:44Z

+        case INS_ssat:
+            // imm1 = shift amount (must be 0 for no shift), imm2 = saturation bits N (1-32)
+            // Encoding: sat_imm field = N-1 stored in bits[4:0]; no shift (sh=0, imm5=0).
+            assert(insDoesNotSetFlags(flags));
+            assert((imm1 == 0) && (imm2 >= 1) && (imm2 <= 32)); // required for encoding
+            imm = (lsb << 5) | (width - 1);                     // lsb=shift=0, width=N -> sat_imm = N-1


Added assert(reg1 != REG_PC) / assert(reg2 != REG_PC) to the INS_ssat case to match the other Thumb-2 D0 encodings in this switch in 71abf12.

Copilot · 2026-06-05T03:58:51Z

+        case INS_usat:
+            // imm1 = shift amount (must be 0 for no shift), imm2 = saturation bits N (0-31)
+            // Encoding: sat_imm field = N stored directly in bits[4:0]; no shift (sh=0, imm5=0).
+            assert(insDoesNotSetFlags(flags));
+            assert((imm1 == 0) && (imm2 >= 0) && (imm2 <= 31)); // required for encoding
+            imm = (lsb << 5) | width;                           // lsb=shift=0, width=N -> sat_imm = N
+


Added the same REG_PC assertions to the INS_usat case in 71abf12.

Copilot · 2026-06-05T03:58:58Z

+                        level++;
+                        break;
+                    }
 #endif // TARGET_RISCV64


Updated the #endif comment to // TARGET_RISCV64 || TARGET_WASM to match the #if/#elif chain in 71abf12.

Copilot · 2026-06-05T03:59:07Z

+            // Copy src to dst (no-op when dst == src).
+            emit->emitIns_R_R(INS_mov, EA_PTRSIZE, dst, src);
+


Replaced the EA_PTRSIZE INS_mov with slli.w rd, rs, 0 (EA_4BYTE) so dst is a sign-extended 32-bit value before the full-register bge compares against the clamp bounds in 71abf12.

Copilot · 2026-06-05T03:59:15Z

+            // Copy src to dst (no-op when dst == src).
+            emit->emitIns_R_R(INS_mov, EA_PTRSIZE, dst, src);
+


Replaced the EA_PTRSIZE INS_mov with INS_sext_w (EA_4BYTE) so the input is normalized to canonical sign-extended form before the full-register bge clamp compares in 71abf12.

…t/usat encoding Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

Initial plan

5f042fc

Copilot AI review requested due to automatic review settings May 26, 2026 20:10

Copilot AI assigned Copilot and tannergooding May 26, 2026

Copilot AI review requested due to automatic review settings May 26, 2026 20:10

Copilot started work on behalf of tannergooding May 26, 2026 20:10 View session

Copilot AI linked an issue May 26, 2026 that may be closed by this pull request

(short)(double)32768.000000000007 produces invalid result #116823

Open

JIT: saturate float/double -> small int casts

0d43c80

Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

Copilot AI requested review from Copilot and removed request for Copilot May 26, 2026 20:30

github-actions Bot added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label May 26, 2026

Add regression test and update interpreter test for saturating R->sma…

30c6aec

…ll casts Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

Copilot AI requested review from Copilot and removed request for Copilot May 26, 2026 20:49

Copilot AI changed the title ~~[WIP] Fix invalid result for short double 32768~~ JIT: Saturate float/double conversions to small integral types May 26, 2026

Copilot finished work on behalf of tannergooding May 26, 2026 20:53

Copilot AI requested a review from tannergooding May 26, 2026 20:53

tannergooding reviewed May 26, 2026

View reviewed changes

Comment thread src/coreclr/jit/morph.cpp Outdated

Copilot started work on behalf of tannergooding May 26, 2026 21:45 View session

JIT: clamp via Min/MaxNative in float domain for R->smallType saturat…

3081861

…ing casts Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

Copilot AI requested review from Copilot and removed request for Copilot May 26, 2026 22:18

Copilot finished work on behalf of tannergooding May 26, 2026 22:20

Copilot AI requested a review from tannergooding May 26, 2026 22:20

tannergooding reviewed May 27, 2026

View reviewed changes

Comment thread src/tests/JIT/interpreter/Interpreter.cs

Copilot started work on behalf of tannergooding May 27, 2026 04:22 View session

Copilot started reviewing on behalf of tannergooding June 3, 2026 15:01 View session

Copilot AI reviewed Jun 3, 2026

View reviewed changes

Copilot started work on behalf of tannergooding June 3, 2026 15:48 View session

Copilot finished work on behalf of tannergooding June 3, 2026 16:11

jkotas reviewed Jun 3, 2026

View reviewed changes

Comment thread src/coreclr/jit/codegenarmarch.cpp Outdated

Apply suggestion from @jkotas

a5c5f12

Copilot AI review requested due to automatic review settings June 3, 2026 23:29

Copilot started reviewing on behalf of jkotas June 3, 2026 23:30 View session

Copilot AI reviewed Jun 3, 2026

View reviewed changes

This was referenced Jun 4, 2026

Unable to pull image from mcr.microsoft.com #117164

Open

Multiple Helix work items fail on maccatalyst/tvos CoreCLR Release #126460

Open

jkotas mentioned this pull request Jun 4, 2026

Make System.Double/Single.ConvertToInteger intrinsic #122424

Closed

tannergooding reviewed Jun 4, 2026

View reviewed changes

Comment thread src/libraries/System.Runtime/tests/System.Runtime.Tests/System/DoubleTests.GenericMath.cs

Copilot started work on behalf of tannergooding June 4, 2026 16:05 View session

wip: investigate primitive ConvertToInteger small-type saturation path

cbcbd69

Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

Copilot AI changed the title ~~JIT, interpreter, AOT preinit: Saturate float/double conversions to small integral types~~ JIT, interpreter, AOT preinit, and CoreLib: Saturate float/double conversions to small integral types Jun 4, 2026

Copilot finished work on behalf of tannergooding June 4, 2026 16:57

Copilot AI requested a review from tannergooding June 4, 2026 16:57

Copilot started work on behalf of tannergooding June 5, 2026 00:25 View session

Handle SaturateTo* intrinsics in remaining JIT paths

5515e6b

Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

Copilot AI requested review from Copilot and removed request for Copilot June 5, 2026 00:50

build-analysis Bot mentioned this pull request Jun 5, 2026

Unable to read data from the transport connection: An existing connection was forcibly closed by the remote host. dotnet/dnceng#5922

Open

3 tasks

Merge branch 'main' into copilot/fix-invalid-result-double-32768

beedfdd

Copilot AI reviewed Jun 5, 2026

View reviewed changes

Address review feedback on saturation intrinsic codegen and ARM32 ssa…

71abf12

…t/usat encoding Co-authored-by: tannergooding <10487869+tannergooding@users.noreply.github.com>

build-analysis Bot mentioned this pull request Jun 5, 2026

iOS tests failing with WORKLOAD TIMED OUT - Killing user command. #108103

Open

		// Copy src to dst (no-op when dst == src).
		emit->emitIns_R_R(INS_mov, EA_PTRSIZE, dst, src);

Conversation

Copilot AI commented May 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changes Made

Uh oh!

dotnet-policy-service Bot commented May 26, 2026

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

tannergooding commented Jun 5, 2026

Uh oh!

Copilot AI commented Jun 5, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Jun 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jun 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jun 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jun 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jun 5, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Copilot AI commented May 26, 2026 •

edited

Loading