TEST: Layout benchmarks for the i32 kernel too

Results shows the same information we know from packing from the other
implementations (the packing is all shared code and doesn't care about
the element. Only element size and kernel size).

running 8 tests
test layout_i32_032::ccc ... bench:       6,012 ns/iter (+/- 80)
test layout_i32_032::ccf ... bench:       6,010 ns/iter (+/- 68)
test layout_i32_032::cfc ... bench:       6,244 ns/iter (+/- 89)
test layout_i32_032::cff ... bench:       6,257 ns/iter (+/- 240)
test layout_i32_032::fcc ... bench:       5,806 ns/iter (+/- 320)
test layout_i32_032::fcf ... bench:       5,767 ns/iter (+/- 2,013)
test layout_i32_032::ffc ... bench:       6,005 ns/iter (+/- 187)
test layout_i32_032::fff ... bench:       6,031 ns/iter (+/- 187)

benches/benchmarks.rs

@@ -14,9 +14,10 @@ extern crate bencher;
 benchmark_main!(
     mat_mul_f32,
     mat_mul_f64,
+    mat_mul_i32,
     layout_f32_032,
     layout_f64_032,
-    mat_mul_i32
+    layout_i32_032
 );
 
 macro_rules! mat_mul {
@@ -113,20 +114,20 @@ macro_rules! gemm_layout {
 
             fn base(bench: &mut Bencher, al: Layout, bl: Layout, cl: Layout)
             {
-                let a = vec![0.; $m * $m]; 
-                let b = vec![0.; $m * $m];
-                let mut c = vec![0.; $m * $m];
+                let a = vec![0 as _; $m * $m]; 
+                let b = vec![0 as _; $m * $m];
+                let mut c = vec![0 as _; $m * $m];
                 let (rsa, csa) = al.strides($m, 1);
                 let (rsb, csb) = bl.strides($m, 1);
                 let (rsc, csc) = cl.strides($m, 1);
                 bench.iter(|| {
                     unsafe {
                         $gemm(
                             $m, $m, $m,
-                            1.,
+                            1 as _,
                             a.as_ptr(), rsa, csa,
                             b.as_ptr(), rsb, csb,
-                            0.,
+                            0 as _,
                             c.as_mut_ptr(), rsc, csc,
                             )
                     }
@@ -164,6 +165,10 @@ gemm_layout!{layout_f64_032, dgemm,
     (m032, 32)
 }
 
+gemm_layout!{layout_i32_032, igemm,
+    (m032, 32)
+}
+
 
 use std::ops::{Add, Mul};