From: Benjamin Kramer Date: Mon, 25 Jul 2011 22:30:00 +0000 (+0000) Subject: Add a note about efficient codegen for binary log. X-Git-Url: http://plrg.eecs.uci.edu/git/?a=commitdiff_plain;h=a3d782f40c8c8ca3d1e2bf24c07e64b48e235336;p=oota-llvm.git Add a note about efficient codegen for binary log. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@135996 91177308-0d34-0410-b5e6-96231b3b80d8 --- diff --git a/lib/Target/X86/README.txt b/lib/Target/X86/README.txt index 560947a4a04..d6ceaa7ed4b 100644 --- a/lib/Target/X86/README.txt +++ b/lib/Target/X86/README.txt @@ -2066,3 +2066,51 @@ The trick is to match "fetch_and_add(X, -C) == C". //===---------------------------------------------------------------------===// +unsigned log2(unsigned x) { + return x > 1 ? 32-__builtin_clz(x-1) : 0; +} + +generates (x86_64): + xorl %eax, %eax + cmpl $2, %edi + jb LBB0_2 +## BB#1: + decl %edi + movl $63, %eax + bsrl %edi, %ecx + cmovel %eax, %ecx + xorl $31, %ecx + movl $32, %eax + subl %ecx, %eax +LBB0_2: + ret + +The cmov and the early test are redundant: + xorl %eax, %eax + cmpl $2, %edi + jb LBB0_2 +## BB#1: + decl %edi + bsrl %edi, %ecx + xorl $31, %ecx + movl $32, %eax + subl %ecx, %eax +LBB0_2: + ret + +If we want to get really fancy we could use some two's complement magic: + xorl %eax, %eax + cmpl $2, %edi + jb LBB0_2 +## BB#1: + decl %edi + bsrl %edi, %ecx + xorl $-32, %ecx + leal 33(%ecx), %eax +LBB0_2: + ret + +This is only useful on targets that can't encode the first operand of a sub +directly. The rule is C1 - (X^C2) -> (C1+1) + (X^~C2). + +//===---------------------------------------------------------------------===//