From mboxrd@z Thu Jan 1 00:00:00 1970 From: Steven Rostedt Subject: Re: [PATCH] spmi: add command tracepoints for SPMI Date: Wed, 20 May 2015 14:58:07 -0400 Message-ID: <20150520145807.63f62ca7@gandalf.local.home> References: <1431985887-23379-1-git-send-email-ankgupta@codeaurora.org> <555BD719.80508@codeaurora.org> <20150520090808.36cf77c2@ankgupta-lnx.qualcomm.com> <20150520182955.GR31753@codeaurora.org> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20150520182955.GR31753@codeaurora.org> Sender: linux-kernel-owner@vger.kernel.org To: Stephen Boyd Cc: Ankit Gupta , gavidov@codeaurora.org, sdharia@codeaurora.org, mlocke@codeaurora.org, linux-arm-msm@vger.kernel.org, gregkh@linuxfoundation.org, ivan.ivanov@linaro.org, linux-kernel@vger.kernel.org, svarbanov@mm-sol.com, galak@codeaurora.org, agross@codeaurora.org List-Id: linux-arm-msm@vger.kernel.org On Wed, 20 May 2015 11:29:55 -0700 Stephen Boyd wrote: > > I see no reason to spend to 4-8 bytes when spmi spec allows for maximum > > buffer size of 16. Do you suggest changing the API of read_cmd()? > > > > > Is that a maximum buffer size of 16 bytes? I'd prefer consistency > with the API that's being traced, that's all. Changing it to u8 > to save a few bytes probably doesn't make any difference if the > architecture passes function arguments in CPU registers which are > 32 or 64 bits wide or if the function can be inlined enough by > the compiler to where the len parameter is kept in a register. > I believe the worry is about wasting bytes in the ring buffer if not necessary. But we do that in other tracepoints, so it's really up to the maintainer. -- Steve