[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

PRMT picks bytes from a pair of registers. The translator implemented its default (index) mode and refused the others: F4E, B4E, RC8, ECL, ECR and RC16, which PTX's prmt exposes, so a shader with one failed to translate.

APrmt now does all seven. Byte i of the result is byte k of {c:a}, where s is the selector's low 2 bits and k is computed per mode:

Mode	k
index	selector nibble i (bit 3 repeats the byte's sign, as before)
F4E	s + i
B4E	(s − i) & 7
RC8	s
ECL	max(s, i)
ECR	min(s, i)
RC16	2(s & 1) + (i & 1)

The mode field (bits 72–74) and these formulas come from the Orin's results. My first B4E formula, from memory of PTX's table, was wrong in 1,021 lanes, and the Orin's outputs corrected it. A constant selector still folds at translate time. The index mode's code is unchanged.

Reference entry: "PRMT's byte-select modes" in docs/reference/shaders.md.

How to verify

DMSH_TEST_XLATE=dm/gfx/tools/dmsh_xlate_bin/dmsh_xlate AMPERE_ORIN_ONLY=prmt,prmt_imm python3 dm/gfx/tests/test_dmsh_ampere_orin.py

prmt (one kernel, all seven modes) failed to translate on main and now matches the Orin in all 7,168 lanes, so it comes out of KNOWN. prmt_imm still matches. Only a register selector is in the fixture; the constant-selector path for the six new modes uses the same formulas but hasn't run on the GPU. Mode 7 still fails the shader.

Maxwell's PRMT isn't implemented at all; it only logs "not implemented". That's separate, so I've left it for another change.

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.
• Translator: the dlls are rebuilt from this tree with .NET 10; tool_lock.py --update-lock and dmsh_chain.py --header were run (chain 03721d07b37d).
• 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

lint_python.py shows no new findings.