Rockbox mail archive

• Next message: Chris Pimlott: "Re: some playing around"
• Previous message: Chris Houghten: "Re: OggVorbis for CPUs without FPU"
• Next in thread: Björn Stenberg: "Re: Benchmarks"
• Reply: Björn Stenberg: "Re: Benchmarks"
• Reply: Rob Ward: "RTS8891"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

I've written some simple benchmarks to test the speed of the player-old and recorder I have. The benchmark code is attached. Here is the result:

(Keep in mind that the player and recorder cpus are not clocked at the same frequency. The players run as 12 MHz while the recorders run at 11.0592 MHz. So ~8% difference is normal.)

Test 1: Disk reading. 10 MB sequential disk read.

Player: 1415 KB/s
Recorder: 1264 KB/s (-12%)

Test 2: A simple double busy loop, code in IRAM.

Player: 4540 ms
Recorder: 5030 ms (+9.7%)

Test 3: Same loop as test 2, code in DRAM.

Player: 4680 ms
Recorder: 15340 ms (+69%)

This last number confirms what I have felt (and why I did these test): The Recorder DRAM is *very* slow, seriously affecting overall performance.

We should see if there is anything we can do to improve this. Maybe our dram parameters aren't optimal? Maybe we can use burst mode better? etc.

-- 
Björn

• text/plain attachment: benchmark.patch

• Next message: Chris Pimlott: "Re: some playing around"
• Previous message: Chris Houghten: "Re: OggVorbis for CPUs without FPU"
• Next in thread: Björn Stenberg: "Re: Benchmarks"
• Reply: Björn Stenberg: "Re: Benchmarks"
• Reply: Rob Ward: "RTS8891"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]