auto merge of #15471 : erickt/rust/push_all, r=acrichto

llvm is currently not able to conver `Vec::extend` into a memcpy for `Copy` types, which results in methods like `Vec::push_all` to run twice as slow as it should be running. This patch takes the unsafe `Vec::clone` optimization to speed up all the operations that are cloning a slice into a `Vec`.

before:

```
test vec::tests::bench_clone_from_0000_0000                ... bench:        12 ns/iter (+/- 2)
test vec::tests::bench_clone_from_0000_0010                ... bench:       125 ns/iter (+/- 4) = 80 MB/s
test vec::tests::bench_clone_from_0000_0100                ... bench:       360 ns/iter (+/- 33) = 277 MB/s
test vec::tests::bench_clone_from_0000_1000                ... bench:      2601 ns/iter (+/- 175) = 384 MB/s
test vec::tests::bench_clone_from_0010_0000                ... bench:        12 ns/iter (+/- 2)
test vec::tests::bench_clone_from_0010_0010                ... bench:       125 ns/iter (+/- 10) = 80 MB/s
test vec::tests::bench_clone_from_0010_0100                ... bench:       361 ns/iter (+/- 28) = 277 MB/s
test vec::tests::bench_clone_from_0100_0010                ... bench:       131 ns/iter (+/- 13) = 76 MB/s
test vec::tests::bench_clone_from_0100_0100                ... bench:       360 ns/iter (+/- 9) = 277 MB/s
test vec::tests::bench_clone_from_0100_1000                ... bench:      2575 ns/iter (+/- 168) = 388 MB/s
test vec::tests::bench_clone_from_1000_0100                ... bench:       356 ns/iter (+/- 20) = 280 MB/s
test vec::tests::bench_clone_from_1000_1000                ... bench:      2605 ns/iter (+/- 167) = 383 MB/s
test vec::tests::bench_from_slice_0000                     ... bench:        11 ns/iter (+/- 0)
test vec::tests::bench_from_slice_0010                     ... bench:       115 ns/iter (+/- 5) = 86 MB/s
test vec::tests::bench_from_slice_0100                     ... bench:       309 ns/iter (+/- 170) = 323 MB/s
test vec::tests::bench_from_slice_1000                     ... bench:      2065 ns/iter (+/- 198) = 484 MB/s
test vec::tests::bench_push_all_0000_0000                  ... bench:         7 ns/iter (+/- 0)
test vec::tests::bench_push_all_0000_0010                  ... bench:        79 ns/iter (+/- 7) = 126 MB/s
test vec::tests::bench_push_all_0000_0100                  ... bench:       342 ns/iter (+/- 18) = 292 MB/s
test vec::tests::bench_push_all_0000_1000                  ... bench:      2873 ns/iter (+/- 75) = 348 MB/s
test vec::tests::bench_push_all_0010_0010                  ... bench:       154 ns/iter (+/- 8) = 64 MB/s
test vec::tests::bench_push_all_0100_0100                  ... bench:       518 ns/iter (+/- 18) = 193 MB/s
test vec::tests::bench_push_all_1000_1000                  ... bench:      4490 ns/iter (+/- 223) = 222 MB/s
```

after:

```
test vec::tests::bench_clone_from_0000_0000                ... bench:        12 ns/iter (+/- 1)
test vec::tests::bench_clone_from_0000_0010                ... bench:       123 ns/iter (+/- 5) = 81 MB/s
test vec::tests::bench_clone_from_0000_0100                ... bench:       367 ns/iter (+/- 23) = 272 MB/s
test vec::tests::bench_clone_from_0000_1000                ... bench:      2618 ns/iter (+/- 252) = 381 MB/s
test vec::tests::bench_clone_from_0010_0000                ... bench:        12 ns/iter (+/- 1)
test vec::tests::bench_clone_from_0010_0010                ... bench:       124 ns/iter (+/- 7) = 80 MB/s
test vec::tests::bench_clone_from_0010_0100                ... bench:       369 ns/iter (+/- 34) = 271 MB/s
test vec::tests::bench_clone_from_0100_0010                ... bench:       123 ns/iter (+/- 6) = 81 MB/s
test vec::tests::bench_clone_from_0100_0100                ... bench:       371 ns/iter (+/- 25) = 269 MB/s
test vec::tests::bench_clone_from_0100_1000                ... bench:      2713 ns/iter (+/- 532) = 368 MB/s
test vec::tests::bench_clone_from_1000_0100                ... bench:       369 ns/iter (+/- 14) = 271 MB/s
test vec::tests::bench_clone_from_1000_1000                ... bench:      2611 ns/iter (+/- 194) = 382 MB/s
test vec::tests::bench_from_slice_0000                     ... bench:         7 ns/iter (+/- 0)
test vec::tests::bench_from_slice_0010                     ... bench:       108 ns/iter (+/- 4) = 92 MB/s
test vec::tests::bench_from_slice_0100                     ... bench:       235 ns/iter (+/- 24) = 425 MB/s
test vec::tests::bench_from_slice_1000                     ... bench:      1318 ns/iter (+/- 96) = 758 MB/s
test vec::tests::bench_push_all_0000_0000                  ... bench:         7 ns/iter (+/- 0)
test vec::tests::bench_push_all_0000_0010                  ... bench:        70 ns/iter (+/- 4) = 142 MB/s
test vec::tests::bench_push_all_0000_0100                  ... bench:       176 ns/iter (+/- 16) = 568 MB/s
test vec::tests::bench_push_all_0000_1000                  ... bench:      1125 ns/iter (+/- 94) = 888 MB/s
test vec::tests::bench_push_all_0010_0010                  ... bench:       159 ns/iter (+/- 15) = 62 MB/s
test vec::tests::bench_push_all_0100_0100                  ... bench:       363 ns/iter (+/- 12) = 275 MB/s
test vec::tests::bench_push_all_1000_1000                  ... bench:      2860 ns/iter (+/- 415) = 349 MB/s
```

This also includes extra benchmarks for `Vec` and `MemWriter`.

1 files changed, 49 insertions, 5 deletions

diff --git a/src/libstd/io/mem.rs b/src/libstd/io/mem.rs
index 7d293f363f0..7f9a1180785 100644
--- a/src/libstd/io/mem.rs
+++ b/src/libstd/io/mem.rs
@@ -607,18 +607,62 @@ mod test {
         assert_eq!(buf.as_slice(), &[7, 8, 6]);
     }
 
-    #[bench]
-    fn bench_mem_writer(b: &mut Bencher) {
+    fn do_bench_mem_writer(b: &mut Bencher, times: uint, len: uint) {
+        let src: Vec<u8> = Vec::from_elem(len, 5);
+
         b.iter(|| {
             let mut wr = MemWriter::new();
-            for _i in range(0u, 10) {
-                wr.write([5, .. 10]).unwrap();
+            for _ in range(0, times) {
+                wr.write(src.as_slice()).unwrap();
             }
-            assert_eq!(wr.unwrap().as_slice(), [5, .. 100].as_slice());
+
+            let v = wr.unwrap();
+            assert_eq!(v.len(), times * len);
+            assert!(v.iter().all(|x| *x == 5));
         });
     }
 
     #[bench]
+    fn bench_mem_writer_001_0000(b: &mut Bencher) {
+        do_bench_mem_writer(b, 1, 0)
+    }
+
+    #[bench]
+    fn bench_mem_writer_001_0010(b: &mut Bencher) {
+        do_bench_mem_writer(b, 1, 10)
+    }
+
+    #[bench]
+    fn bench_mem_writer_001_0100(b: &mut Bencher) {
+        do_bench_mem_writer(b, 1, 100)
+    }
+
+    #[bench]
+    fn bench_mem_writer_001_1000(b: &mut Bencher) {
+        do_bench_mem_writer(b, 1, 1000)
+    }
+
+    #[bench]
+    fn bench_mem_writer_100_0000(b: &mut Bencher) {
+        do_bench_mem_writer(b, 100, 0)
+    }
+
+    #[bench]
+    fn bench_mem_writer_100_0010(b: &mut Bencher) {
+        do_bench_mem_writer(b, 100, 10)
+    }
+
+    #[bench]
+    fn bench_mem_writer_100_0100(b: &mut Bencher) {
+        do_bench_mem_writer(b, 100, 100)
+    }
+
+    #[bench]
+    fn bench_mem_writer_100_1000(b: &mut Bencher) {
+        do_bench_mem_writer(b, 100, 1000)
+    }
+
+    #[bench]
     fn bench_mem_reader(b: &mut Bencher) {
         b.iter(|| {
             let buf = Vec::from_slice([5 as u8, ..100]);