[AlexFigas]

What

__signbit, __signbitf and __signbitl returned 1 for a negative argument, following FreeBSD's signbit (#3640). The console's versions return the sign bit where it sits in the top 16-bit word of the value: 0x8000 for any argument whose sign bit is set, 0 otherwise. That is also what AnyPS5's _LDsign already returns.

A title that only tests for nonzero is unaffected. A title that compares the result, stores it, or shifts it saw a different value from the console's.

The evidence is the libc.prx bundled with PPSA21564, the library's own code:

Function	NID	Address	Code
__signbit	Rw4J-22tu1U	0x54880	vmovmskpd eax, xmm0; shl rax, 63; shr rax, 48
__signbitf	CjQROLB88a4	0x54890	vmovmskps eax, xmm0; shl eax, 31; shr eax, 16
__signbitl	Cj81LPErPCc	0x548a0	movzx eax, word [rsp+0x10]; and eax, 0x8000

NaNs follow their sign bit like every other value.

Tested

On Ubuntu 24.04 under WSL2 (GCC 13.3, Release), head def9214 (this PR rebased onto main a13ee2d):

• Native comparison. I used the harness from fix(libc): compute acosl with FreeBSD msun's ld80 code as the bundled libc.prx does #11260, with that libc.prx's segments mapped into a Linux process, and ran the binary's three functions next to AnyPS5's. There are 0 mismatches with this PR:

  • __signbitf on all 4,294,967,296 float bit patterns;
  • __signbit on 20,000,000 random double encodings, half of them biased to zero, subnormal, infinite and NaN exponents;
  • __signbitl on 20,000,000 random 80-bit encodings, including unnormals, pseudo-denormals, pseudo-infinities and pseudo-zeros.

  On main, every negative input differs: 2,147,483,648 float patterns, 10,003,161 doubles and 10,002,538 long doubles.

• guest_math passes (3 runs). Its __signbitl checks now expect 0x8000. New checks cover __signbit and __signbitf on signed zeros, a subnormal, infinity and NaNs of both signs.

• Conventions. python3 tools/check_conventions.py --base origin/main passes.

• Overlap. python3 tools/pr_overlap.py --branch fix/libc-signbit-sign-mask reports no line overlap. It could not compare 29 PRs that conflict with main; each gives the same git merge-tree conflicts with this branch as with main.

• Stacking. The branch was stacked on fix(libc): return Dinkumware codes from __fpclassifyf #11277, because __signbitf is the line after the __fpclassifyf that PR rewrites. fix(libc): return Dinkumware codes from __fpclassifyf #11277 has merged, so only this PR's commit is left.

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none (fix(libc): return Dinkumware codes from __fpclassifyf #11277 has merged)
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

[MarcZX10]

Reviewed exact head d58ce8353d3a439117811643cc0cfc1fe6c0a3fb. The 0x8000 sign mask matches the bundled libc.prx instruction sequences described in the report, and the new guest cases cover positive/negative zero, finite values, infinities and NaNs across float, double and long double. The PR reports exhaustive comparisons against the bundled binary; I did not independently rerun that harness. Its dependency on #11277 is recorded. No blocking code finding. Policy and conventions checks pass; full product builds are skipped under the PR path policy.

[github-actions]

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Pull request	Shared files not compared
PR 6085	core/libs/prx/libc/src/Math.cpp
PR 6295	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 8937	core/libs/prx/libc/src/Math.cpp
PR 10253	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10345	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 11892	core/libs/tests/GuestMath.cpp

[MarcZX10]

I found an unreported behavior change in the diff beyond the three sign-mask exports. __fpclassifyf_nid_postfix now returns _FDtest_nid_postfix's internal -2/-1/0/1/2 codes, and CheckFloatClassification was changed to expect those codes. Before this PR, the export mapped those categories to the project's FP_SUBNORMAL/FP_NORMAL/FP_ZERO/FP_INFINITE/FP_NAN values (0x08/0x04/0x10/0x01/0x02). That changes an independent public result for every classification and is not supported by the signbit disassembly or the reported differential tests, which compare only __signbit, __signbitf, and __signbitl.

Please restore the existing __fpclassifyf mapping in this signbit PR, or provide direct evidence for the console's __fpclassifyf return codes and update the scope/report and tests to cover that behavior. My earlier approval did not catch this collateral change; I am correcting it now.

[MarcZX10]

Correction to my immediately preceding review: I attributed the __fpclassifyf change to this signbit commit because GitHub shows the full stack diff against main. The exact commit diff confirms that the classification change is entirely in prerequisite #11277, whose PR provides the libc.prx disassembly and exhaustive float differential; this PR's commit only changes the three sign-mask exports and their tests. #11278 correctly records Depends on: #11277. The concern in my prior review was mistaken, and I withdraw it. The existing signbit evidence, test coverage and completed hosted matrix support approval.

[wachi35676]

Windows run of head d58ce8353 merged onto main 0f34ebb85, WinLibs GCC 15.2.0, Release. The description's runs are under WSL.

• guest_math passes on Windows. With main's Math.cpp it aborts with Guest __fpclassifyf failed for 00000000, the first check of the fix(libc): return Dinkumware codes from __fpclassifyf #11277 commit this branch carries. With only the three signbit returns put back to 1 it still aborts, so the new checks cover this commit on their own.
• No host-side caller of __signbit*_nid_postfix compares the result with 1, so the 0x8000 return reaches only guest code. It merges cleanly once fix(libc): return Dinkumware codes from __fpclassifyf #11277 is in.

AI-assisted: yes (Claude Code); results above were run on this machine.

[babyhuey]

Windows run of this head d58ce8353d3, merged onto main 0a570f617d8 (clean merge, local only), with the selection CI's skipped Windows product job uses: Windows 11, Ryzen 7 5800X, RTX 3080, WinLibs GCC 15.2, Release, BUILD_TESTING=ON. The tree was configured fresh after the merge, then select_tests.py select / ninja -f select.ninja selected / --target libs / select_tests.py ctest --parallel 4 --timeout 120.

• Selected 2547 tests: 1838 pass, 1 fails, guest_raise_exception, which also fails on 0a570f617d8 itself on this machine (crash or timeout under --parallel 4; it passed on 27ad7244d22, and I'm narrowing that down separately), 708 skip. No failures from this PR.

AI-assisted: yes (Claude Code)

[maskym]

Windows result for head d58ce8353d, merged into main 1089f2bbfd (the PR's Tested section names no Windows run).

Host: Windows 11 26300, Intel i7-12700K, WinLibs GCC 15.2.0 (x86_64-ucrt-posix-seh, r6), Release, BUILD_TESTING=ON, Ninja. long double is the 80-bit x87 type on this toolchain, as on the console.

The tree also carried the other open libc PRs #11260, #11262, #11266, #11632, #11634, #11270, #11283, #10929, all merged cleanly, so the run is a joint one; the tests of this PR are guest_math. Every test in the set passed: ctest --test-dir build -R "math|libc|...|stdio" 32/32 and ctest -R "guest_dinkumware_internals|guest_standard_streams" 5/5, guest_math in 0.05 s. Build clean. Not run on a PS5.

AI-assisted: yes (Claude Code), the runs and the numbers above are from this machine.

[JJenkx]

A runtime cross-check on 11 more titles. A scratch loader runs each decrypted libc.prx natively on Windows 11 and calls __signbit (Rw4J-22tu1U), __signbitf (CjQROLB88a4) and __signbitl (Cj81LPErPCc) by NID. The titles are PPSA04158, 05916, 07923, 09614, 13934, 20423, 25636, 26528, 33655, 34008 and 35169.

All 11 export all three, and every one returns:

• 0x8000 (32768): for -2.0, -0.0 and a negative NaN;
• 0: for +2.0.

That matches the description's table. On PPSA07923, 200,000 random double and float arguments (a quarter of them infinities, NaNs, zeros or subnormals, with random signs) gave 0 or 0x8000 from the console every time. Current main gave 1 wherever the console gave 0x8000.

AI-assisted: yes (Claude Code).

[BunnyFl]

Approving head def9214119. I read the diff and ran firmware 13.42's __signbit, __signbitf and __signbitl: the decrypted system libSceLibcInternal.sprx's exports, natively on a PC (not a console). Three identical runs.

For -1, -0, -inf, a negative NaN and the smallest negative subnormal, and for 1, 0, inf, a positive NaN and 1e300, all three return 0x8000 when the sign bit is set and 0 otherwise. That is what this change returns and what the new and updated test rows expect, so the system module agrees with the PPSA21564 libc.prx reading in the description.