If we swap its arguments, SkTaskGroup::batch() _is_ sk_parallel_for.

src/core/SkTaskGroup.h

Index: src/core/SkTaskGroup.h
diff --git a/src/core/SkTaskGroup.h b/src/core/SkTaskGroup.h
index d1daa444945c6ada116c155a5c9c6cba38dc17f8..e6c36651fd50772a67b417e2044447e6c120fcdb 100644
--- a/src/core/SkTaskGroup.h
+++ b/src/core/SkTaskGroup.h
@@ -34,7 +34,7 @@ public:
     void add(std::function<void(void)> fn);
 
     // Add a batch of N tasks, all calling fn with different arguments.
-    void batch(std::function<void(int)> fn, int N);
+    void batch(int N, std::function<void(int)> fn);
 
     // Block until all Tasks previously add()ed to this SkTaskGroup have run.
     // You may safely reuse this SkTaskGroup after wait() returns.
@@ -47,48 +47,4 @@ private:
 // Returns best estimate of number of CPU cores available to use.
 int sk_num_cores();
 
-int sk_parallel_for_thread_count();
-
-// Call f(i) for i in [0, end).
-template <typename Func>
-void sk_parallel_for(int end, const Func& f) {
-    if (end <= 0) { return; }
-
-    struct Chunk {
-        const Func* f;
-        int start, end;
-    };
-
-    // TODO(mtklein): this chunking strategy could probably use some tuning.
-    int max_chunks  = sk_num_cores() * 2,
-        stride      = (end + max_chunks - 1 ) / max_chunks,
-        nchunks     = (end + stride - 1 ) / stride;
-    SkASSERT(nchunks <= max_chunks);
-
-#if defined(GOOGLE3)
-    // Stack frame size is limited in GOOGLE3.
-    SkAutoSTMalloc<512, Chunk> chunks(nchunks);
-#else
-    // With the chunking strategy above this won't malloc until we have a machine with >512 cores.
-    SkAutoSTMalloc<1024, Chunk> chunks(nchunks);
-#endif
-
-    for (int i = 0; i < nchunks; i++) {
-        Chunk& c = chunks[i];
-        c.f      = &f;
-        c.start  = i * stride;
-        c.end    = SkTMin(c.start + stride, end);
-        SkASSERT(c.start < c.end);  // Nothing will break if start >= end, but it's a wasted chunk.
-    }
-
-    Chunk* chunkBase = chunks.get();
-    auto run_chunk = [chunkBase](int i) {
-        Chunk& c = chunkBase[i];
-        for (int i = c.start; i < c.end; i++) {
-            (*c.f)(i);
-        }
-    };
-    SkTaskGroup().batch(run_chunk, nchunks);
-}
-
 #endif//SkTaskGroup_DEFINED