Actually, burst size is equal to '1 << desc->rqcfg.brst_size'.
we should use burst size, not desc->rqcfg.brst_size.
dma memcpy performance on Rockchip RV1126
@ 1512MHz A7, 1056MHz LPDDR3, 200MHz DMA:
dmatest:
/# echo dma0chan0 > /sys/module/dmatest/parameters/channel
/# echo 4194304 > /sys/module/dmatest/parameters/test_buf_size
/# echo 8 > /sys/module/dmatest/parameters/iterations
/# echo y > /sys/module/dmatest/parameters/norandom
/# echo y > /sys/module/dmatest/parameters/verbose
/# echo 1 > /sys/module/dmatest/parameters/run
dmatest: dma0chan0-copy0: result #1: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
dmatest: dma0chan0-copy0: result #2: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
dmatest: dma0chan0-copy0: result #3: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
dmatest: dma0chan0-copy0: result #4: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
dmatest: dma0chan0-copy0: result #5: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
dmatest: dma0chan0-copy0: result #6: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
dmatest: dma0chan0-copy0: result #7: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
dmatest: dma0chan0-copy0: result #8: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
Before:
dmatest: dma0chan0-copy0: summary 8 tests, 0 failures 48 iops 200338 KB/s (0)
After this patch:
dmatest: dma0chan0-copy0: summary 8 tests, 0 failures 179 iops 734873 KB/s (0)
After this patch and increase dma clk to 400MHz:
dmatest: dma0chan0-copy0: summary 8 tests, 0 failures 259 iops 1062929 KB/s (0)
Signed-off-by: Sugar Zhang <[email protected]>
---
drivers/dma/pl330.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/dma/pl330.c b/drivers/dma/pl330.c
index e9f0101..0f5c193 100644
--- a/drivers/dma/pl330.c
+++ b/drivers/dma/pl330.c
@@ -2799,7 +2799,7 @@ pl330_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dst,
* If burst size is smaller than bus width then make sure we only
* transfer one at a time to avoid a burst stradling an MFIFO entry.
*/
- if (desc->rqcfg.brst_size * 8 < pl330->pcfg.data_bus_width)
+ if (burst * 8 < pl330->pcfg.data_bus_width)
desc->rqcfg.brst_len = 1;
desc->bytes_requested = len;
--
2.7.4
On 14-11-20, 11:55, Sugar Zhang wrote:
> Actually, burst size is equal to '1 << desc->rqcfg.brst_size'.
> we should use burst size, not desc->rqcfg.brst_size.
>
> dma memcpy performance on Rockchip RV1126
> @ 1512MHz A7, 1056MHz LPDDR3, 200MHz DMA:
>
> dmatest:
>
> /# echo dma0chan0 > /sys/module/dmatest/parameters/channel
> /# echo 4194304 > /sys/module/dmatest/parameters/test_buf_size
> /# echo 8 > /sys/module/dmatest/parameters/iterations
> /# echo y > /sys/module/dmatest/parameters/norandom
> /# echo y > /sys/module/dmatest/parameters/verbose
> /# echo 1 > /sys/module/dmatest/parameters/run
>
> dmatest: dma0chan0-copy0: result #1: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
> dmatest: dma0chan0-copy0: result #2: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
> dmatest: dma0chan0-copy0: result #3: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
> dmatest: dma0chan0-copy0: result #4: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
> dmatest: dma0chan0-copy0: result #5: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
> dmatest: dma0chan0-copy0: result #6: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
> dmatest: dma0chan0-copy0: result #7: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
> dmatest: dma0chan0-copy0: result #8: 'test passed' with src_off=0x0 dst_off=0x0 len=0x400000
>
> Before:
>
> dmatest: dma0chan0-copy0: summary 8 tests, 0 failures 48 iops 200338 KB/s (0)
>
> After this patch:
>
> dmatest: dma0chan0-copy0: summary 8 tests, 0 failures 179 iops 734873 KB/s (0)
>
> After this patch and increase dma clk to 400MHz:
>
> dmatest: dma0chan0-copy0: summary 8 tests, 0 failures 259 iops 1062929 KB/s (0)
Applied, thanks
--
~Vinod