[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH v2 02/10] target/hexagon: import README for idef-parser
From: |
Alessandro Di Federico |
Subject: |
Re: [PATCH v2 02/10] target/hexagon: import README for idef-parser |
Date: |
Mon, 1 Mar 2021 15:50:35 +0100 |
On Thu, 25 Feb 2021 12:20:53 -0800
Richard Henderson <richard.henderson@linaro.org> wrote:
> This is odd, as is the description of why. Yes, if RdV is read
> without initialization, TCG middle-end will abort (at least with
> --enable-debug-tcg enabling the assertions). But you've just said
> that "no reading" was found.
When we say that no reading was found, we mean that it was not
initialized by the caller. The term "reading" is a leftover from how
the parser input was organized in the first iterations of the base
patchset.
I'll rephrase that.
> So why did you perform this dummy initialization, which will be
> eliminated later?
The initialization is redundant in this specific example, but, in
general, non-initialized values are assumed to be zero-initialized.
For instance when you're writing a 64-bit integer piecewise, by OR-ing
two 32-bit integers, it matters.
In short: useless in this case (but DCE'd by the mid-end), important in
general.
> So, I take it from this that you're emitting tcg directly from within
> the parser, and not generating any kind of abstract syntax tree?
Yes. There a few spots where an AST would have been beneficial, but
overall we deem it would increase the complexity of the parser with
limited return.
--
Alessandro Di Federico
rev.ng
- Re: [PATCH v2 02/10] target/hexagon: import README for idef-parser,
Alessandro Di Federico <=