Benchmark Certification Report

Benchmark Information
Benchmark Score #  757  
Suite  TeleBench 1.1  
Type of Platform  Hardware/Production Silicon  
Type of Certification  Optimized  
Certification Date  07/27/06  
Benchmark Run Date  2006-06-14 00:00:00 
Hardware Availability Date  Now 
Software Availability Date  Now 
OS Availability Date  Now 
Benchmark Notes   

Processor & Processor System
Processor Name-Clock IBM 970FX - 2GHz 
Hardware Type Production silicon 
Processor Part Number IBM25PPC970FX6TB348RT 
Vendor Contact 
Public availability date of processor NOW 
Architecture Type RISC 
Native Data Type 64-bit 
L1 Instruction Cache Size (kbyte) 64KB 
L1 Instruction Cache Type Direct map 
L1 Data Cache Size (kbyte) 32 KB 
L1 Data Cache Type 2-way set associative 
On-chip Peripherals Used in Benchmark None 
Board Name Maple D 
Certified Board Serial Number 0003CC30254 
Firmware Version or setup file name PIBS 1.05.0000 
External Data Bus Width (bits) 32 bits 
RAM Amount 2048 MB 
Memory Clock (mhz) 400 MHz 
Type and Size of Other On-chip Memory None 
L2 Cache Clock 2GHz 
L2 Cache Size (kbyte) 512 KB 
L2 Cache Type 8-way set associative 
Ordering Info (Product number) 900-071-1 

Compiler Vendor GCC 
Compiler Model and Version GCC 4.2.0 - autovect 20060523  
Compiler Vendor Public Contact Information 
Libraries Used in Benchmark Standard + Test Harness 
Linker Version 4.2.0 
Floating Point Hardware 

Compiler flags
Code Generation Flags -O3 -ftree-loop-linear -funroll-loops -fvariable-expansion-in-unroller -fgcse -fgcse-las --param max-gcse-passes=3 -fstrict-aliasing -ffast-math -mpowerpc64 -mcpu=power4 -mtune=power4 -maltivec -fcv -ftree-early-loop -fweb --ipa-cp -freorder-blocks -falign-functions=32 -falign-loops=32 -falign-jumps=32 -m64  
Endian Big 

Benchmark Scores


Autocorrelation - Data1 (pulse) 20833333.33 7298 5640 
Autocorrelation - Data2 (sine) 268817.204 7298 7720 
Autocorrelation - Data3 (speech) 276752.768 7298 6808 
Convolutional Encoder - Data1 (xk5r2dt) 29058543.59 8066 6336 
Convolutional Encoder - Data2 (xk4r2dt) 38071065.99 7954 6200 
Convolutional Encoder - Data3 (xk3r2dt) 48543689.32 8082 6072 
Fixed-point Bit Allocation - Data2 (typ) 22014.1341 7,332 2.064 
Fixed-point Bit Allocation - Data3 (step) 215827.338 7,188 1,136 
Fixed Point Bit Allocation - Data6 (pent) 44651.0631 7,556 1,456 
FFT/IFFT - Data1 (pulse) 727826.4976 9634 11944 
FFT/IFFT - Data2 (spn) 728876.7523 9634 11944 
FFT/IFFT - Data3 (sine) 740740.741 9634 9896 
Viterbi Decoder - Data1 (get) 294117.647 8114 7752 
Viterbi Decoder - Data2 (toggle) 294117.647 8114 7752 
Viterbi Decoder - Data3 (ones)  294117.647 8114 7752 
Viterbi Decoder - Data4 (zeros) 294233.4414 8114 7752 
TelemarkTM 1058.7
Highlighted Fields indicate Certified Data 

Optimization Methods
Public  The benchmarks were optimized by recoding the algorithms in the High-Level Interface language for the Vector/SIMD Multimedia Extension Technology (VMX). No assembly language was employed. Autocorrelation --------------- VMX intrinsics are used to vectorize computations. Convolutional Encoder --------------------- Input data is converted to bit stream. VMX intrinsics are used to vectorize computations. Fast Fourier Transform (FFT) ---------------------------- Radix-4, decimation in time FFT is used. VMX intrinsics are used to vectorize computations. Viterbi Decoder --------------- VMX intrinsics are used to vectorize computations. Bit Allocation --------------- VMX intrinsics are used to vectorize computations. Index values computations and array look up are vectorized.  

Certified by the EEMBC Technology Center


© 2020 EEMBC. All rights reserved.