Float Product combine1: Maximum use of data abstraction:
Best: 11.63 (82%), Overall Best: 11.63 40-most: 11.63 cycles/element
Float Product combine2: Take vec_length() out of loop:
Best: 9.70 (2%), Overall Best: 9.85 40-most: 9.85 cycles/element
Float Product combine3: Array reference to vector data:
Best: 3.57 (4%), Overall Best: 3.58 40-most: 3.59 cycles/element
Float Product combine3w: Update *dest within loop only with write:
Best: 3.58 (54%), Overall Best: 3.58 40-most: 3.59 cycles/element
Float Product combine4: Array reference, accumulate in temporary:
Best: 3.57 (2%), Overall Best: 3.58 40-most: 3.60 cycles/element
Float Product combine4b: Include bonds check in loop:
Best: 3.49 (2%), Overall Best: 3.58 40-most: 3.62 cycles/element
Float Product combine4p: Pointer reference, accumulate in temporary:
Best: 3.48 (2%), Overall Best: 3.58 40-most: 3.61 cycles/element
Float Product combine5: Array code, unrolled by 2:
Best: 3.37 (2%), Overall Best: 3.58 40-most: 3.63 cycles/element
Float Product combine5p: Pointer code, unrolled by 2, for loop:
Best: 3.57 (2%), Overall Best: 3.58 40-most: 3.63 cycles/element
Float Product unroll2aw: Array code, unrolled by 2, while loop:
Best: 3.50 (2%), Overall Best: 3.58 40-most: 3.64 cycles/element
Float Product unroll3a: Array code, unrolled by 3:
Best: 3.53 (2%), Overall Best: 3.58 40-most: 3.63 cycles/element
Float Product unroll4a: Array code, unrolled by 4:
Best: 3.48 (2%), Overall Best: 3.58 40-most: 3.63 cycles/element
Float Product unroll5a: Array code, unrolled by 5:
Best: 3.47 (2%), Overall Best: 3.58 40-most: 3.63 cycles/element
Float Product unroll6a: Array code, unrolled by 6:
Best: 3.57 (4%), Overall Best: 3.58 40-most: 3.63 cycles/element
Float Product unroll7a: Array code, unrolled by 7:
Best: 3.54 (4%), Overall Best: 3.58 40-most: 3.62 cycles/element
Float Product unroll8a: Array code, unrolled by 8:
Best: 3.55 (2%), Overall Best: 3.58 40-most: 3.62 cycles/element
Float Product unroll9a: Array code, unrolled by 9:
Best: 3.57 (6%), Overall Best: 3.58 40-most: 3.62 cycles/element
Float Product unroll10a: Array code, unrolled by 10:
Best: 3.54 (2%), Overall Best: 3.57 40-most: 3.61 cycles/element
Float Product unroll16a: Array code, unrolled by 16:
Best: 3.58 (34%), Overall Best: 3.58 40-most: 3.60 cycles/element
Float Product unroll2: Pointer code, unrolled by 2:
Best: 3.45 (2%), Overall Best: 3.58 40-most: 3.59 cycles/element
Float Product unroll3: Pointer code, unrolled by 3:
Best: 3.57 (8%), Overall Best: 3.58 40-most: 3.58 cycles/element
Float Product unroll4: Pointer code, unrolled by 4:
Best: 3.56 (2%), Overall Best: 3.58 40-most: 3.58 cycles/element
Float Product unroll8: Pointer code, unrolled by 8:
Best: 3.55 (2%), Overall Best: 3.58 40-most: 3.58 cycles/element
Float Product unroll16: Pointer code, unrolled by 16:
Best: 3.55 (2%), Overall Best: 3.58 40-most: 3.59 cycles/element
Float Product combine6: Array code, unrolled by 2, Superscalar x2:
Best: 1.78 (4%), Overall Best: 1.79 40-most: 1.79 cycles/element
Float Product unroll4x2a: Array code, unrolled by 4, Superscalar x2:
Best: 1.78 (2%), Overall Best: 1.79 40-most: 1.79 cycles/element
Float Product unroll8x2a: Array code, unrolled by 8, Superscalar x2:
Best: 1.81 (6%), Overall Best: 1.81 40-most: 1.83 cycles/element
Float Product unroll3x3a: Array code, unrolled by 3, Superscalar x3:
Best: 1.19 (82%), Overall Best: 1.20 40-most: 1.19 cycles/element
Float Product unroll4x4a: Array code, unrolled by 4, Superscalar x4:
Best: 0.87 (2%), Overall Best: 0.90 40-most: 0.91 cycles/element
Float Product unroll5x5a: Array code, unrolled by 5, Superscalar x5:
Best: 0.92 (44%), Overall Best: 0.92 40-most: 0.94 cycles/element
Float Product unroll6x6a: Array code, unrolled by 6, Superscalar x6:
Best: 0.92 (2%), Overall Best: 0.93 40-most: 0.97 cycles/element
Float Product unroll7x7a: Array code, unrolled by 7, Superscalar x7:
Best: 0.93 (4%), Overall Best: 0.94 40-most: 0.98 cycles/element
Float Product unroll8x4a: Array code, unrolled by 8, Superscalar x4:
Best: 0.94 (8%), Overall Best: 0.95 40-most: 0.99 cycles/element
Float Product unroll8x8a: Array code, unrolled by 8, Superscalar x8:
Best: 0.93 (2%), Overall Best: 0.95 40-most: 1.00 cycles/element
Float Product unroll9x9a: Array code, unrolled by 9, Superscalar x9:
Best: 0.95 (4%), Overall Best: 0.95 40-most: 1.01 cycles/element
Float Product unroll10x10a: Array code, unrolled by 10, Superscalar x10:
Best: 0.93 (2%), Overall Best: 0.95 40-most: 1.02 cycles/element
Float Product unroll2x6a: Array code, unrolled by 12, Superscalar x6:
Best: 0.87 (2%), Overall Best: 0.96 40-most: 0.98 cycles/element
Float Product unroll12x12a: Array code, unrolled by 12, Superscalar x12:
Best: 0.93 (2%), Overall Best: 0.94 40-most: 1.02 cycles/element
Float Product unroll16x16a: Array code, unrolled by 16, Superscalar x16:
Best: 0.92 (6%), Overall Best: 0.94 40-most: 1.00 cycles/element
Float Product unroll20x20a: Array code, unrolled by 20, Superscalar x20:
Best: 1.11 (2%), Overall Best: 1.18 40-most: 1.23 cycles/element
Float Product unroll8x2: Pointer code, unrolled by 8, Superscalar x2:
Best: 1.76 (2%), Overall Best: 1.78 40-most: 1.85 cycles/element
Float Product unroll8x4: Pointer code, unrolled by 8, Superscalar x4:
Best: 0.92 (2%), Overall Best: 0.94 40-most: 0.98 cycles/element
Float Product unroll8x8: Pointer code, unrolled by 8, Superscalar x8:
Best: 0.91 (2%), Overall Best: 0.93 40-most: 0.98 cycles/element
Float Product unroll9x3: Pointer code, unrolled by 9, Superscalar x3:
Best: 1.19 (2%), Overall Best: 1.21 40-most: 1.25 cycles/element
Float Product unrollx2as: Array code, Unroll x2, Superscalar x2, noninterleaved:
Best: 1.77 (2%), Overall Best: 1.79 40-most: 1.80 cycles/element
Float Product combine7: Array code, unrolled by 2, different associativity:
Best: 1.78 (4%), Overall Best: 1.79 40-most: 1.80 cycles/element
Float Product unroll3aa: Array code, unrolled by 3, Different Associativity:
Best: 1.18 (2%), Overall Best: 1.19 40-most: 1.20 cycles/element
Float Product unroll4aa: Array code, unrolled by 4, Different Associativity:
Best: 0.89 (28%), Overall Best: 0.90 40-most: 0.90 cycles/element
Float Product unroll5aa: Array code, unrolled by 5, Different Associativity:
Best: 0.89 (2%), Overall Best: 0.90 40-most: 0.90 cycles/element
Float Product unroll6aa: Array code, unrolled by 6, Different Associativity:
Best: 0.87 (2%), Overall Best: 0.90 40-most: 0.91 cycles/element
Float Product unroll7aa: Array code, unrolled by 7, Different Associativity:
Best: 0.89 (6%), Overall Best: 0.91 40-most: 0.91 cycles/element
Float Product unroll8aa: Array code, unrolled by 8, Different Associativity:
Best: 0.86 (2%), Overall Best: 0.91 40-most: 0.92 cycles/element
Float Product unroll9aa: Array code, unrolled by 9, Different Associativity:
Best: 0.90 (2%), Overall Best: 0.92 40-most: 0.92 cycles/element
Float Product unroll10aa: Array code, unrolled by 10, Different Associativity:
Best: 0.83 (2%), Overall Best: 0.91 40-most: 0.92 cycles/element
Float Product unroll12aa: Array code, unrolled by 12, Different Associativity:
Best: 0.92 (22%), Overall Best: 0.95 40-most: 0.94 cycles/element
Float Product simd_v1: SSE code, 1*VSIZE-way parallelism:
Best: 7.30 (2%), Overall Best: 7.56 40-most: 7.58 cycles/element
Float Product simd_v2: SSE code, 2*VSIZE-way parallelism:
Best: 5.92 (2%), Overall Best: 6.08 40-most: 6.12 cycles/element
Float Product simd_v4: SSE code, 4*VSIZE-way parallelism:
Best: 2.60 (2%), Overall Best: 2.61 40-most: 2.65 cycles/element
Float Product simd_v8: SSE code, 8*VSIZE-way parallelism:
Best: 1.68 (2%), Overall Best: 1.77 40-most: 1.82 cycles/element
Float Product simd_v10: SSE code, 10*VSIZE-way parallelism:
Best: 1.67 (2%), Overall Best: 1.76 40-most: 1.82 cycles/element
Float Product simd_v12: SSE code, 12*VSIZE-way parallelism:
Best: 1.78 (2%), Overall Best: 1.79 40-most: 1.85 cycles/element
Float Product simd_v2a: SSE code, 2*VSIZE-way parallelism, reassociate:
Best: 5.83 (6%), Overall Best: 5.83 40-most: 5.90 cycles/element
Float Product simd_v4a: SSE code, 4*VSIZE-way parallelism, reassociate:
Best: 2.77 (16%), Overall Best: 2.77 40-most: 2.83 cycles/element
Float Product simd_v8a: SSE code, 8*VSIZE-way parallelism, reassociate:
Best: 1.63 (2%), Overall Best: 1.71 40-most: 1.78 cycles/element