bench: make screening a first-class step with the same caliber as the A/B

Six rounds of A/B cost 13-44 minutes per operator on S5000 (rand, 15 shapes:
13.0 min; sort, 25 shapes: 43.4 min). Screening one costs about 0.8 minutes —
21 operators in 15.0 minutes, which knocked 13 of them out of the expensive
queue and cut it from ~7 hours to ~2.7. That ratio is the whole argument for
having a separate first step.

Screening was being done by ad-hoc scripts living in the results tree, and
those got two things wrong that this codebase already knows better than:

Both sides ran `record`. Each swept autotune independently, so the two sides
compiled different configs and produced different IR for reasons that have
nothing to do with the switch — every operator would screen as "changed".
screen.sh pins configs exactly as round 1 of ab_repeat does: A records, B
replays A.

The criterion was the convert_layout count. Equal counts do not mean equal
generated code, and what the A/B measures is whether the generated code
changed. The ttgir fingerprint decides; the cvt counts are still recorded,
because they explain what kind of change it was.

Four verdicts, and three of them are not "the switch does nothing":
IR_CHANGED (measure it), NO_IR_DIFF (skip it), NO_KERNEL (the operator never
reached Triton on this backend), NO_ROWS (it did not run — usually the dtype).
The last two are runnability problems being mistaken for switch results, which
is how `fft` got written off: the screen forced DTYPES=float16 on every
operator and fft only accepts complex64. OPS entries now carry an optional
per-operator dtype and default to the operator's own set.

ir_fingerprint moves to _ir_fingerprint.sh so screening and the A/B cannot
drift apart on the criterion.

README.md

@@ -137,6 +137,40 @@ ttgir 关键设计：
 - **命名 = 组内有区分度的参数**：同一 (shape, kernel) 组内取值相同的 constexpr 不进文件名；多词参数缩写为首字母（`BLOCK_SIZE_M` → `BSM`，图例见 naming.md）；同名冲突依次用 dtype（`__fp16`）、参数对齐特化（`__EMdiv16`）、hash 前缀消歧。
 - **崩溃也能拿到 IR**：dump 挂在进程 atexit 上，CUDA crash 后仍会落盘已编译部分。
 
+## 两步：先筛选，再实测
+
+跑一个算子的六轮 A/B 要 13~44 分钟（S5000 实测：`rand` 15 shape 13.0 分钟，
+`sort` 25 shape 43.4 分钟）。筛一个算子只要约 0.8 分钟（21 个算子一遍 15.0 分钟）。
+**差 20~50 倍**，所以先筛后测：
+
+```bash
+# 第一步：筛掉不值得排队的算子
+SWITCH=FLAGTREE_MUSA_RLC_ENHANCE ZL_BACKEND=musa FLAGGEMS_DIR=/path/to/FlagGems   OPS="col2im:col2im topk:topk perf_fft:fft" bash screen.sh
+
+# 第二步：只把 IR_CHANGED 的送进来
+SWITCH=FLAGTREE_MUSA_RLC_ENHANCE ROUNDS=6 OPS="topk:topk" bash ab_repeat.sh
+```
+
+`screen.sh` 对每个算子给四种判定之一：
+
+| 判定 | 含义 | 该怎么办 |
+|---|---|---|
+| `IR_CHANGED` | 两侧 ttgir 不同 | 排六轮 A/B |
+| `NO_IR_DIFF` | 逐字节相同（config 已锁死） | 跳过。跑六轮只会测到噪声，并报成一个"无差异"结论 |
+| `NO_KERNEL` | 跑出了行，但没编出 `.ttgir` | 该算子在本后端没走 Triton，编译器开关无从谈起。**不是"开关无作用"** |
+| `NO_ROWS` | 没有测量行 | 先修可运行性：多半是 dtype 不被该算子接受，或该算子在本后端拿不到 config |
+
+三件容易做错的：
+
+- **筛选也是 A/B，口径一样严。** A 侧 record、B 侧 `REPLAY_FROM` 它。两侧各自
+  record 会各自 sweep 出不同 config，IR 自然不同，于是每个算子都判成
+  `IR_CHANGED`——筛了等于没筛。
+- **判据是 ttgir 指纹，不是 `convert_layout` 计数。** cvt 数相同不等于生成代码
+  相同，而要测的是"生成代码变了没有"。cvt 数照样记，它解释的是变化的性质。
+- **不要全局写死一个 `DTYPES`。** `OPS` 元素支持 `OP:OP_FILE:DTYPES`，留空表示用
+  该算子自己的全套。`fft` 只吃 complex64，被一刀切的 `DTYPES=float16` 判死过一次
+  ——那是 harness 的问题，被记成了算子的问题。
+
 ## A/B 对比测试标准流程
 
 ```bash