| Message ID | 1504960117-24264-1-git-send-email-aford173@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Adam, On Saturday 09 September 2017 05:58 PM, Adam Ford wrote: > When booting with a device tree, message appears: > memcpy is disabled > When reading through the bindings for eDMA3, it shows that > ti,edma-memcpy-channels is an optional property, but the example > binding uses it. > > RFC: I'll be the first to admit that I don't understand what > this is doing, but it seems like a DMA enabled memcpy would > improve performance. Is this appropriate for da850? Since there > are two eDMA3-tpcc controllers, is there a set of channels we can > use for each? > > Signed-off-by: Adam Ford <aford173@gmail.com> This enables DMA-based memcpy (see drivers/dma/dmaetest.c for example). But does not accelerate any memcpy in kernel automatically. If this is done, it should be done in a board specific manner because the available channels you pick up for memcpy support can change from board to board. Thanks, Sekhar
diff --git a/arch/arm/boot/dts/da850.dtsi b/arch/arm/boot/dts/da850.dtsi index af68ef7..eacadb8 100644 --- a/arch/arm/boot/dts/da850.dtsi +++ b/arch/arm/boot/dts/da850.dtsi @@ -263,7 +263,7 @@ interrupts = <11 12>; interrupt-names = "edma3_ccint", "edma3_ccerrint"; #dma-cells = <2>; - + ti,edma-memcpy-channels = <20 21>; ti,tptcs = <&edma0_tptc0 7>, <&edma0_tptc1 0>; }; edma0_tptc0: tptc@8000 {
When booting with a device tree, message appears: memcpy is disabled When reading through the bindings for eDMA3, it shows that ti,edma-memcpy-channels is an optional property, but the example binding uses it. RFC: I'll be the first to admit that I don't understand what this is doing, but it seems like a DMA enabled memcpy would improve performance. Is this appropriate for da850? Since there are two eDMA3-tpcc controllers, is there a set of channels we can use for each? Signed-off-by: Adam Ford <aford173@gmail.com>