Changes in directory llvm/lib/CodeGen:
README.txt updated: 1.4 -> 1.5 --- Log message: Updates. --- Diffs of the changes: (+57 -1) README.txt | 58 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++- 1 files changed, 57 insertions(+), 1 deletion(-) Index: llvm/lib/CodeGen/README.txt diff -u llvm/lib/CodeGen/README.txt:1.4 llvm/lib/CodeGen/README.txt:1.5 --- llvm/lib/CodeGen/README.txt:1.4 Mon Apr 16 13:10:22 2007 +++ llvm/lib/CodeGen/README.txt Mon Apr 30 13:42:09 2007 @@ -85,4 +85,60 @@ //===---------------------------------------------------------------------===// -Re-Materialize load from frame index. +Use local info (i.e. register scavenger) to assign it a free register to allow +reuse: + ldr r3, [sp, #+4] + add r3, r3, #3 + ldr r2, [sp, #+8] + add r2, r2, #2 + ldr r1, [sp, #+4] <== + add r1, r1, #1 + ldr r0, [sp, #+4] + add r0, r0, #2 + +//===---------------------------------------------------------------------===// + +LLVM aggressively lift CSE out of loop. Sometimes this can be negative side- +effects: + +R1 = X + 4 +R2 = X + 7 +R3 = X + 15 + +loop: +load [i + R1] +... +load [i + R2] +... +load [i + R3] + +Suppose there is high register pressure, R1, R2, R3, can be spilled. We need +to implement proper re-materialization to handle this: + +R1 = X + 4 +R2 = X + 7 +R3 = X + 15 + +loop: +R1 = X + 4 @ re-materialized +load [i + R1] +... +R2 = X + 7 @ re-materialized +load [i + R2] +... +R3 = X + 15 @ re-materialized +load [i + R3] + +Furthermore, with re-association, we can enable sharing: + +R1 = X + 4 +R2 = X + 7 +R3 = X + 15 + +loop: +T = i + X +load [T + 4] +... +load [T + 7] +... +load [T + 15] _______________________________________________ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
