[karthick-mcw]

Summary

Drop a redundant gather in the refine loop of compute_ideal_weights_for_decimation.

compute_ideal_weights_for_decimation's refine loop gathered both infilled_weights and ei.weights per texel and subtracted them (old_weight - ideal_weight). This pre-subtracts ei.weights from the infilled weight at store time instead, so the refine loop only needs a single gather (weight_diff) for the same result — one fewer gather per texel on the hot path.

The result is bit-identical. The subtraction is reassociated to happen once at infill-store time instead of once per use in the refine loop, so the arithmetic — and therefore the output — doesn't change.

Files: Source/astcenc_ideal_endpoints_and_weights.cpp

Measured

Asset compression workload, gain in % (positive means faster).

Vendor	CPU	OS	Compiler	Single thread	All threads
AMD	Granite 9600	Windows	clang-cl	-0.93	2.2
Intel	Arrow Lake	Windows	clang-cl	0.2	2.31
Apple	Mac Air	macOS	AppleClang	0.58	0.07

Test plan

• Bit-identical output: 6 representative LDR/HDR textures x 6 block sizes (4x4-12x12) x 4 quality presets (144 runs), byte-identical .astc output with and without the patch.
• Unit test: astcenc C++ unit test suite (ctest, -DASTCENC_UNITTEST=ON), 223/223 tests pass on the patched build.
• Functional tests: Test/astc_test_functional.py --encoder avx2 gives the identical result with and without the patch (38/93 pass; the remaining 55 error only on a missing ImageMagick dependency in this test environment, not on the code — zero assertion failures either way).
• Timing: single core and all cores on all three machines (table above).