[Know-It-All-Brother]

Part of the Ampere accuracy series (#2040). It depends on #2041 for the test: the commits before this PR's own are #2041's and drop out once it lands, so this PR's own change is the last commit.

What and why

For add.sat.s32 and sub.sat.s32, ptxas emits an IADD3, then two PLOP3.LUT Pd, PT, Ra.SIGN, Rb.SIGN, Rc.SIGN, lut, 0x0 that detect overflow from the operands' and the sum's sign bits, then two SELs that clamp. That PLOP3 form (opcode 0x21f, registers as sources) wasn't implemented, only the predicate-source forms, so a saturating integer add or subtract failed to translate.

APlop3Sign reads each source as R < 0 and runs the existing PLOP3 lookup table. Mesa NAK has no encoding for this form, so the fields come from the Orin's four encodings: the registers sit where the ALU register form has them, the first LUT is bits 72–79, the second LUT is bits 16–23, and the destinations are bits 81–83 and 84–86. The LUT values check out: 0x02 is a=0, b=0, c=1, which is positive overflow on add. It costs three integer compares and the LUT's terms per destination.

Reference entry: "Saturating integer add and subtract (PLOP3 on sign bits)" in docs/reference/shaders.md.

How to verify

DMSH_TEST_XLATE=dm/gfx/tools/dmsh_xlate_bin/dmsh_xlate AMPERE_ORIN_ONLY=iadd_sat,isub_sat python3 dm/gfx/tests/test_dmsh_ampere_orin.py

iadd_sat and isub_sat failed to translate on main and now match the Orin in all 1,024 lanes, so they come out of KNOWN. Both clamps are exercised: the Orin's outputs hold 56 maximum and 57 minimum clamps for iadd_sat, and 45 and 51 for isub_sat. Only the register form is registered, since no other form has been seen.

Checked

• Tests: test_dmsh_ampere_orin.py; every other test_dmsh_* gives the same result as on main (test_dmsh_ampere_mem.py, test_dmsh_meshstage.py and test_dmsh_pairs.py fail on main the same way); test_nsp_chain.py, test_xlate_anycpu.py, test_docs.py. lint_python.py shows no new findings.
• Translator: the dlls are rebuilt from this tree with .NET 10; tool_lock.py --update-lock and dmsh_chain.py --header were run (chain 813df7a0ab79).
• Devices and OS: MacBook Pro M5 Max, macOS 27.
• regress.py: not run. My only local Switch 2 build's shader package comes from an older shader split than main's map, so main's chain can't build it. Ocarina of Time's build already translates every shader, so it can't contain this instruction.
• Games run: none.

Docs

• docs/reference/shaders.md has the entry, and python3 dm/tools/docs_symptoms.py was run
• A chain tool changed: dmsh_chain.py --header was run, and dmsh_chain.h, the translator dlls and the lock are part of the change
• No game files, keys or data extracted from a game