Bug #1357: Fixing variables into specific CPU registers deemed overrated & may disturb compilers' optimizers

• Status: Closed
• Assignee: ko1 (Koichi Sasada)
• Target version: -
• ruby -v: ruby 1.9.2dev (2009-03-31) [i386-darwin9.6.0]

[roberto (Ollivier Robert)]

=begin
I do not think that fixing specific variable into CPU registers is really reflected in performance gains (benchmarks welcome!). It will also generally disturb compilers' handling of optimization (especially on register-starved CPUs like IA-32/i386) and may generate less efficient code. Remove the i386/amd64-specific code.

As a side-effect, it enables successful compilation with recent versions of clang/llvm.
=end

[nobu (Nobuyoshi Nakada)]

=begin
Hi,

At Mon, 6 Apr 2009 17:43:20 +0900,
Ollivier Robert wrote in [ruby-core:23132]:

As a side-effect, it enables successful compilation with
recent versions of clang/llvm.

Does it have a macro to tell if it's clang/llvm?

--
Nobu Nakada

=end

[roberto (Ollivier Robert)]

=begin
clang uses clang as main #define. It also aims to support most GNUC constructs & defines.
=end

[nobu (Nobuyoshi Nakada)]

=begin
Hi,

At Mon, 6 Apr 2009 19:38:09 +0900,
Ollivier Robert wrote in [ruby-core:23137]:

clang uses clang as main #define. It also aims to
support most GNUC constructs & defines.

Also GNUC and i386?

I'd thought llvm generates code for it's own VM.

--
Nobu Nakada

=end

[yugui (Yuki Sonoda)]

• Status changed from Open to Feedback
• Assignee set to ko1 (Koichi Sasada)

=begin

=end

[naruse (Yui NARUSE)]

• Status changed from Feedback to Closed

=begin
r25504 fixes this.
=end