On 6/21/25 16:50, Richard Henderson wrote:
+uint64_t HELPER(sve2p1_cntp_c)(uint32_t png, uint32_t desc)
+{
+ int pl = FIELD_EX32(desc, PREDDESC, OPRSZ);
+ int vl = pl * 8;
+ unsigned v_esz = FIELD_EX32(desc, PREDDESC, ESZ);
+ int lg2_width = FIELD_EX32(desc, PREDDESC, DATA) + 1;
+ unsigned p_esz;
+ int p_count, maxelem;
+ bool p_invert;
+
+ /* C.f. Arm pseudocode CounterToPredicate. */
+ if ((png & 0xf) == 0) {
+ /* Canonical false predicate. */
+ return 0;
+ }
+ p_esz = ctz32(png);
+
+ /*
+ * maxbit = log2(pl * 4)
+ * = log2(vl / 8 * 4)
+ * = log2(vl / 2)
+ * = log2(vl) - 1
+ * maxbit_mask = ones<maxbit:0>
+ * = (1 << (maxbit + 1)) - 1
+ * = (1 << (log2(vl) - 1 + 1)) - 1
+ * = (1 << log2(vl)) - 1
+ * = pow2ceil(vl) - 1
+ * Note that we keep count in bytes, not elements.
+ */
+ p_count = (png & (pow2ceil(vl) - 1)) >> 1;
This is too clever for it's own good, and misses masking out the esz bit we located via
ctz above. All of the predicate-as-counter insns suffer the same error.
I'll put all of the counter parsing into some helper functions.
r~
