The BLX instruction is encoded differently than the BL, because why not? In

[oota-llvm.git] / lib / Target / README.txt

diff --git a/lib/Target/README.txt b/lib/Target/README.txt
index 06661c933549c52a8060c20e714c3a5add153edf..38f5af642d7172384cec3e11aef738fd579fe801 100644 (file)
--- a/lib/Target/README.txt
+++ b/lib/Target/README.txt
@@ -938,16 +938,6 @@ The expression should optimize to something like
 
 //===---------------------------------------------------------------------===//
 
-void a(int variable)
-{
- if (variable == 4 || variable == 6)
-   bar();
-}
-This should optimize to "if ((variable | 2) == 6)".  Currently not
-optimized with "clang -emit-llvm-bc | opt -std-compile-opts | llc".
-
-//===---------------------------------------------------------------------===//
-
 unsigned int f(unsigned int i, unsigned int n) {++i; if (i == n) ++i; return
 i;}
 unsigned int f2(unsigned int i, unsigned int n) {++i; i += i == n; return i;}
@@ -1708,28 +1698,35 @@ from gcc.
 //===---------------------------------------------------------------------===//
 
 Missed instcombine transformation:
-define i32 @a(i32 %x) nounwind readnone {
-entry:
-  %rem = srem i32 %x, 32
-  %shl = shl i32 1, %rem
-  ret i32 %shl
-}
 
-The srem can be transformed to an and because if x is negative, the shift is
-undefined. Testcase derived from gcc.
+  %382 = srem i32 %tmp14.i, 64                    ; [#uses=1]
+  %383 = zext i32 %382 to i64                     ; [#uses=1]
+  %384 = shl i64 %381, %383                       ; [#uses=1]
+  %385 = icmp slt i32 %tmp14.i, 64                ; [#uses=1]
+
+The srem can be transformed to an and because if %tmp14.i is negative, the
+shift is undefined.  Testcase derived from 403.gcc.
 
 //===---------------------------------------------------------------------===//
 
-Missed instcombine/dagcombine transformation:
-define i32 @a(i32 %x, i32 %y) nounwind readnone {
-entry:
-  %mul = mul i32 %y, -8
-  %sub = sub i32 %x, %mul
-  ret i32 %sub
-}
+This is a range comparison on a divided result (from 403.gcc):
+
+  %1337 = sdiv i32 %1336, 8                       ; [#uses=1]
+  %.off.i208 = add i32 %1336, 7                   ; [#uses=1]
+  %1338 = icmp ult i32 %.off.i208, 15             ; [#uses=1]
+  
+We already catch this (removing the sdiv) if there isn't an add, we should
+handle the 'add' as well.  This is a common idiom with it's builtin_alloca code.
+C testcase:
+
+int a(int x) { return (unsigned)(x/16+7) < 15; }
 
-Should compile to something like x+y*8, but currently compiles to an
-inefficient result.  Testcase derived from gcc.
+Another similar case involves truncations on 64-bit targets:
+
+  %361 = sdiv i64 %.046, 8                        ; [#uses=1]
+  %362 = trunc i64 %361 to i32                    ; [#uses=2]
+...
+  %367 = icmp eq i32 %362, 0                      ; [#uses=1]
 
 //===---------------------------------------------------------------------===//