Hi Arnd,
On Fri, Apr 17, 2015 at 2:19 PM, Arnd Bergmann <arnd@xxxxxxxx> wrote:
> On Friday 17 April 2015 01:01:13 Rameshwar Prasad Sahu wrote:
>> v3 changes:
>> * Minor changes in length setting in DMA descriptor
>>
>> v2 changes:
>> * Code cleanup
>> * Changed way of setting DMA descriptors for big-endian
>>
>> This patch fixes compilation sparse warnings like incorrect type in assignment
>> (different base types), cast to restricted __le64, symbol
>> '__UNIQUE_ID_author__COUNTER__' has multiple initializers etc and
>> coccinelle warnings (No need to set .owner here. The core will do it.)
>>
>> This patch is based on slave-dma / for-linus branch.
>> (commit: 9f2fd0dfa594d857fbdaeda523ff7a46f16567f5 [26/28]
>> dmaengine: Add support for APM X-Gene SoC DMA engine driver)
>>
>> Reported-by: kbuild test robot <fengguang.wu@xxxxxxxxx>
>> Signed-off-by: Rameshwar Prasad Sahu <rsahu@xxxxxxx>
>
> Looks good now,
>
> Reviewed-by: Arnd Bergmann <arnd@xxxxxxxx>
>
> There is one small enhancement that you could still do and I'll shut up after
> that ;-)
>
>>
>> -static void *xgene_dma_lookup_ext8(u64 *desc, int idx)
>> +static __le64 *xgene_dma_lookup_ext8(struct xgene_dma_desc_hw *desc, int idx)
>> {
>> - return (idx % 2) ? (desc + idx - 1) : (desc + idx + 1);
>> + switch (idx) {
>> + case 0:
>> + return &desc->m1;
>> + case 1:
>> + return &desc->m0;
>> + case 2:
>> + return &desc->m3;
>> + case 3:
>> + return &desc->m2;
>> + default:
>> + pr_err("Invalid dma descriptor index\n");
>> + }
>> +
>> + return NULL;
>> }
>>
> ...
>
>> /* Set 1st to 5th source addresses */
>> for (i = 0; i < src_cnt; i++) {
>> len = *nbytes;
>> - xgene_dma_set_src_buffer((i == 0) ? (desc1 + 8) :
>> + xgene_dma_set_src_buffer((i == 0) ? &desc1->m1 :
>> xgene_dma_lookup_ext8(desc2, i - 1),
>> &len, &src[i]);
>> - XGENE_DMA_DESC_MULTI_SET(desc1, scf[i], i);
>> + desc1->m2 |= cpu_to_le64((scf[i] << ((i + 1) * 8)));
>> }
>
> If you just unroll this loop, you get code that is smaller and easier to
> understand:
>
> /* Set 1st to 5th source addresses */
> xgene_dma_set_src_buffer(&desc1->m1, &len, &src[0]);
> xgene_dma_set_src_buffer(&desc2->m0, &len, &src[1]);
> xgene_dma_set_src_buffer(&desc2->m3, &len, &src[2]);
> xgene_dma_set_src_buffer(&desc2->m2, &len, &src[3]);
> desc1->m2 |= cpu_to_le64(scf[0] | scf[1] << 8 | scf[2] << 16 | scf[3] << 24);
Actually here, in run time src_cnt value can be 2 or 3 upto 5, so we
can't unroll it.
>
> Arnd
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
