SASS evidence coverage
This report shows the recorded GPU checks, PTX compilation probes and reference cases for SASS instructions, alongside which instructions our generated Ghidra languages can model. Modeled support and evidence are counted separately.
Worked example: reading F2F coverage
The cards below count an opcode as covered when any of its forms has semantics; their percentages do not measure complete support. The report uses the exact floating-point language variants. Generated by sass-refresh-coverage.
Search by opcode name, or choose an evidence level to keep rows with matching evidence in at least one cell (in the selected architecture’s cell when an architecture filter is active). Evidence filters apply to modeled forms; “modeled, no evidence” finds gaps in their checks. Dimmed cells do not match the evidence filter. Click an architecture heading or card to require that its column contain the opcode; other columns stay visible for comparison. Click again to clear that selection. Category counts follow the visible rows; cards always summarize the full architecture. Hatched cells mean the opcode is absent from that architecture’s machine description. Click a cell for details; click it again or press Escape to close.
No opcodes match.