fix(cmd/gasm): probe the riscv64 operand shapes the encoder accepts

Assisted-by: GLM 5.3 Flash

cmd/gasm/audit.go

@@ -305,22 +305,57 @@ func probeShapes(a arch.Arch) []string {
 			"(X5), X6", "F0, F1, F2", "F0, F1", "p2", "X1, p2", "X0, p2",
 			"X5, X6, p2", "p2(SB)",
 			// AMO atomics: destination, base, source.
-			"R5, (R4), R6", "X5, (X4), X6",
+			"R5, (R4), R6", "X5, (R4), R6", "X5, (X4), X6",
 			// Segment stores take the first vector register aligned
 			// to the segment count, as the toolchain requires.
 			"(X5), X6, V0, V8", "(X5), X6, V0", "(X5), X0, V4",
 			// The FP multiply-add family takes four registers.
 			"F0, F1, F2, F3",
+			// The register banks the conversions and comparisons read:
+			// FCLASS and the float-to-int moves end in an integer
+			// register, the int-to-float forms open with one, and the
+			// FP compares write their verdict into the integer file.
+			"F0, X5", "X5, F0", "F0, F1, X5",
 			// The RVV slice: register, vector-register and vtype forms.
 			"V1, V2, V3", "V1, X5, V2", "V1", "V1, (X5)", "(X5), V1",
 			"$15, V1", "$15", "V1, V2", "V1, X5",
 			"X5, X6, p2", "R5, R6, p2",
 			"X5, E8, M8, TA, MA, X6", "$4, E32, M1, TA, MA, X1",
 			"(X5), X6, V1, V2",
+			// The vector scalar forms: the float scalar in the rs1
+			// field (VFADDVF and the whole VF family, VFMVVF), the
+			// integer scalar spellings (VMVVX, VMVXS, VCPOPM) and the
+			// masked four-operand shapes.
+			"F1, V2, V3", "V1, F2, V3", "F1, V2", "V1, F2",
+			"X1, V2", "V1, X2",
+			"F1, V2, V0, V3", "V1, F2, V0, V3", "V2, F1, V0, V3",
+			// The indexed vector loads and stores name a vector index
+			// register, the strided ones an integer stride.
+			"(X5), V2, V1", "V1, V2, (X5)", "V1, X6, (X5)",
 			// The CSR immediate forms the toolchain's testdata spells:
 			// immediate, CSR name, destination.
 			"$2, TIME, X5",
-			"",
+			// The vector carry and merge forms carry the mandatory V0
+			// mask between the sources and the destination; the
+			// carry-producing ones take the mask-free three-operand
+			// shape, and CANDI's immediate form lands on a prime
+			// register.
+			"V1, V2, V0, V3", "X5, V2, V0, V3", "$15, V2, V0, V3",
+			"$15, V2, V3", "X5, V2, V3", "$-32, X10",
+			// The runtime's bookkeeping pair, the compiler's own
+			// no-operand statements.
+			"$1, $0", "",
+			// The compressed instructions the toolchain's own testdata
+			// spells: the stack-based loads and stores pin their base to
+			// SP, the register-based ones live in the prime registers
+			// X8-X15, and the control forms branch PC-relatively.
+			"20(SP), X10", "24(SP), X10", "32(SP), F10",
+			"X10, 20(SP)", "X10, 24(SP)", "F10, 32(SP)",
+			"20(X10), X11", "24(X10), X11", "32(X10), F11",
+			"X11, 20(X10)", "X11, 24(X10)", "F11, 32(X10)",
+			"1(PC)", "X5, p2", "X10, 1(PC)",
+			"$-32, X5", "$-512, SP", "$4, SP, X10", "$63, X10",
+			"X9, X8", "X6, X5",
 		}
 	case arch.LOONG64:
 		return []string{