<p>also if you power the board on with the debugger connected via SWD, for goddess knows what reasons it usually just doesn&#39;t boot</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@whitequark" class="u-url mention">@<span>whitequark</span></a></span> I asked nxp for a datasheet for one of their imx parts a while ago using my university address, and they told me to fuck off because they considered that datasheet too secret for a pleb like me. I wasn&#39;t very impressed by that stance.</p>

<p>also my only devboard with it emits a loud high-pitched coil whine the entire time it&#39;s turned on. if it blinks a LED the pitch changes slightly. if it processes a command via SWD it also changes</p>

<p>designed by NXP (Freescale (Motorola))</p>

<p>can i please have a normal device _just once_</p>

<p>today in normal microcontrollers (imxrt1052):<br />- no internal flash, requires an external (Q)SPI chip. both NOR and NAND are supported<br />- if you put a SOIC clip onto the chip it stops booting<br />- if you erase the flash it stops being accessible to a debugger<br />- if you assert SYSRESETREQ it stops talking to the debugger for long enough that everything times out. presumably it is looking in the flash for whether debugging is allowed or whatever</p><p>why is it like this ;w;</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.online/@resistor" class="u-url mention">@<span>resistor</span></a></span> <span class="h-card" translate="no"><a href="https://mastodon.social/@barrelshifter" class="u-url mention">@<span>barrelshifter</span></a></span> Either way, I like the way the funnel shift solution worked out. (And will always be grateful for Sanjay doing the legwork!)</p><p>They&#39;re a good normal form, funnel shift&lt;-&gt;rotate (where applicable) is canonical and trivial in both directions, they can be formed early (which avoids destroying the pattern), they&#39;re reasonably common in target ISAs on their own right, and are still pretty straightforward to lower where they&#39;re not available</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.gamedev.place/@rygorous" class="u-url mention">@<span>rygorous</span></a></span> <span class="h-card" translate="no"><a href="https://mastodon.social/@barrelshifter" class="u-url mention">@<span>barrelshifter</span></a></span> I think SimplifyDemandedBits was an area where GCC didn’t do much optimization at the time, so leaning heavily into it made LLVM look more compelling in microbenchmarks.</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.online/@resistor" class="u-url mention">@<span>resistor</span></a></span> <span class="h-card" translate="no"><a href="https://mastodon.social/@barrelshifter" class="u-url mention">@<span>barrelshifter</span></a></span> The ADD-&gt;OR thing is another frequent monkey wrench especially wrt matching things that could be addressing modes, yeah.</p><p>It&#39;s not a worthless transformation for larger-than-native-register integers (where not having cross-limb carries is a solid win) but for &lt;=pointer size, I&#39;d argue it hurts more than it helps.</p>