//===---------------------------------------------------------------------===//
-Dead argument elimination should be enhanced to handle cases when an argument is
-dead to an externally visible function. Though the argument can't be removed
-from the externally visible function, the caller doesn't need to pass it in.
-For example in this testcase:
-
- void foo(int X) __attribute__((noinline));
- void foo(int X) { sideeffect(); }
- void bar(int A) { foo(A+1); }
-
-We compile bar to:
-
-define void @bar(i32 %A) nounwind ssp {
- %0 = add nsw i32 %A, 1 ; <i32> [#uses=1]
- tail call void @foo(i32 %0) nounwind noinline ssp
- ret void
-}
-
-The add is dead, we could pass in 'i32 undef' instead. This occurs for C++
-templates etc, which usually have linkonce_odr/weak_odr linkage, not internal
-linkage.
-
-//===---------------------------------------------------------------------===//
-
With the recent changes to make the implicit def/use set explicit in
machineinstrs, we should change the target descriptions for 'call' instructions
so that the .td files don't list all the call-clobbered registers as implicit
//===---------------------------------------------------------------------===//
-A/B get pinned to the stack because we turn an if/then into a select instead
-of PRE'ing the load/store. This may be fixable in instcombine:
-http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37892
-
-struct X { int i; };
-int foo (int x) {
- struct X a;
- struct X b;
- struct X *p;
- a.i = 1;
- b.i = 2;
- if (x)
- p = &a;
- else
- p = &b;
- return p->i;
-}
-
-//===---------------------------------------------------------------------===//
-
Interesting missed case because of control flow flattening (should be 2 loads):
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26629
With: llvm-gcc t2.c -S -o - -O0 -emit-llvm | llvm-as |
//===---------------------------------------------------------------------===//
-InstCombine should use SimplifyDemandedBits to remove the or instruction:
-
-define i1 @test(i8 %x, i8 %y) {
- %A = or i8 %x, 1
- %B = icmp ugt i8 %A, 3
- ret i1 %B
-}
-
-Currently instcombine calls SimplifyDemandedBits with either all bits or just
-the sign bit, if the comparison is obviously a sign test. In this case, we only
-need all but the bottom two bits from %A, and if we gave that mask to SDB it
-would delete the or instruction for us.
-
-//===---------------------------------------------------------------------===//
-
functionattrs doesn't know much about memcpy/memset. This function should be
marked readnone rather than readonly, since it only twiddles local memory, but
functionattrs doesn't handle memset/memcpy/memmove aggressively:
br label %_ZNSt6vectorIiSaIiEEC1EmRKiRKS0_.exit
This is just the handling the construction of the vector. Most surprising here
-is the fact that all three null stores in %entry are dead, but not eliminated.
+is the fact that all three null stores in %entry are dead (because we do no
+cross-block DSE).
+
Also surprising is that %conv isn't simplified to 0 in %....exit.thread.i.i.
+This is a because the client of LazyValueInfo doesn't simplify all instruction
+operands, just selected ones.
//===---------------------------------------------------------------------===//
clang -O3 -fno-exceptions currently compiles this code:
-void f(int N) {
- std::vector<int> v(N);
- for (int k = 0; k < N; ++k)
- v[k] = 0;
-
- extern void sink(void*); sink(&v);
+void f(char* a, int n) {
+ __builtin_memset(a, 0, n);
+ for (int i = 0; i < n; ++i)
+ a[i] = 0;
}
-into almost the same as the previous note, but replace its final BB with:
+into:
-for.body.lr.ph: ; preds = %cond.true.i.i.i.i
- %mul.i.i.i.i.i = shl i64 %conv, 2
- %call3.i.i.i.i.i = call noalias i8* @_Znwm(i64 %mul.i.i.i.i.i) nounwind
- %0 = bitcast i8* %call3.i.i.i.i.i to i32*
- store i32* %0, i32** %v8.sub, align 8, !tbaa !0
- %add.ptr.i.i.i = getelementptr inbounds i32* %0, i64 %conv
- store i32* %add.ptr.i.i.i, i32** %tmp4.i.i.i.i.i, align 8, !tbaa !0
- call void @llvm.memset.p0i8.i64(i8* %call3.i.i.i.i.i, i8 0, i64 %mul.i.i.i.i.i, i32 4, i1 false)
- store i32* %add.ptr.i.i.i, i32** %tmp3.i.i.i.i.i, align 8, !tbaa !0
- %tmp18 = add i32 %N, -1
- %tmp19 = zext i32 %tmp18 to i64
- %tmp20 = shl i64 %tmp19, 2
- %tmp21 = add i64 %tmp20, 4
- call void @llvm.memset.p0i8.i64(i8* %call3.i.i.i.i.i, i8 0, i64 %tmp21, i32 4, i1 false)
- br label %for.end
+define void @_Z1fPci(i8* nocapture %a, i32 %n) nounwind {
+entry:
+ %conv = sext i32 %n to i64
+ tail call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 %conv, i32 1, i1 false)
+ %cmp8 = icmp sgt i32 %n, 0
+ br i1 %cmp8, label %for.body.lr.ph, label %for.end
+
+for.body.lr.ph: ; preds = %entry
+ %tmp10 = add i32 %n, -1
+ %tmp11 = zext i32 %tmp10 to i64
+ %tmp12 = add i64 %tmp11, 1
+ call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 %tmp12, i32 1, i1 false)
+ ret void
+
+for.end: ; preds = %entry
+ ret void
+}
-First off, why (((zext %N - 1) << 2) + 4) instead of the ((sext %N) << 2) done
-previously? (or better yet, re-use that one?)
+This shouldn't need the ((zext (%n - 1)) + 1) game, and it should ideally fold
+the two memset's together. The issue with %n seems to stem from poor handling
+of the original loop.
-Then, the really painful one is the second memset, of the same memory, to the
-same value.
+To simplify this, we need SCEV to know that "n != 0" because of the dominating
+conditional. That would turn the second memset into a simple memset of 'n'.
//===---------------------------------------------------------------------===//
4 stores on each iteration. If the struct were 8 bytes, this gets turned into
a memset.
+In order to handle this we have to:
+ A) Teach clang to generate metadata for memsets of structs that have holes in
+ them.
+ B) Teach clang to use such a memset for zero init of this struct (since it has
+ a hole), instead of doing elementwise zeroing.
+
//===---------------------------------------------------------------------===//
clang -O3 currently compiles this code:
//===---------------------------------------------------------------------===//
-clang -O3 currently compiles this code:
+We should enhance memcpy/memcpy/memset to allow a metadata node on them
+indicating that some bytes of the transfer are undefined. This is useful for
+frontends like clang when lowering struct copies, when some elements of the
+struct are undefined. Consider something like this:
+
+struct x {
+ char a;
+ int b[4];
+};
+void foo(struct x*P);
+struct x testfunc() {
+ struct x V1, V2;
+ foo(&V1);
+ V2 = V1;
-#include <emmintrin.h>
-int f(double x) { return _mm_cvtsd_si32(_mm_set_sd(x)); }
-int g(double x) { return _mm_cvttsd_si32(_mm_set_sd(x)); }
+ return V2;
+}
-into
+We currently compile this to:
+$ clang t.c -S -o - -O0 -emit-llvm | opt -scalarrepl -S
-define i32 @_Z1fd(double %x) nounwind readnone {
+
+%struct.x = type { i8, [4 x i32] }
+
+define void @testfunc(%struct.x* sret %agg.result) nounwind ssp {
entry:
- %vecinit.i = insertelement <2 x double> undef, double %x, i32 0
- %vecinit1.i = insertelement <2 x double> %vecinit.i, double 0.000000e+00,i32 1
- %0 = tail call i32 @llvm.x86.sse2.cvtsd2si(<2 x double> %vecinit1.i) nounwind
- ret i32 %0
+ %V1 = alloca %struct.x, align 4
+ call void @foo(%struct.x* %V1)
+ %tmp1 = bitcast %struct.x* %V1 to i8*
+ %0 = bitcast %struct.x* %V1 to i160*
+ %srcval1 = load i160* %0, align 4
+ %tmp2 = bitcast %struct.x* %agg.result to i8*
+ %1 = bitcast %struct.x* %agg.result to i160*
+ store i160 %srcval1, i160* %1, align 4
+ ret void
}
-define i32 @_Z1gd(double %x) nounwind readnone {
-entry:
- %conv.i = fptosi double %x to i32
- ret i32 %conv.i
-}
-
-This difference carries over to the assmebly produced, resulting in:
-
-_Z1fd: # @_Z1fd
-# BB#0: # %entry
- pushq %rbp
- movq %rsp, %rbp
- xorps %xmm1, %xmm1
- movsd %xmm0, %xmm1
- cvtsd2sil %xmm1, %eax
- popq %rbp
- ret
-
-_Z1gd: # @_Z1gd
-# BB#0: # %entry
- pushq %rbp
- movq %rsp, %rbp
- cvttsd2si %xmm0, %eax
- popq %rbp
- ret
-
-The problem is that we can't see through the intrinsic call used for cvtsd2si,
-and fold away the unnecessary manipulation of the function parameter. When
-these functions are inlined, it forms a barrier preventing many further
-optimizations. LLVM IR doesn't have a good way to model the logic of
-'cvtsd2si', its only FP -> int conversion path forces truncation. We should add
-a rounding flag onto fptosi so that it can represent this type of rounding
-naturally in the IR rather than using intrinsics. We might need to use a
-'system_rounding_mode' flag to encode that the semantics of the rounding mode
-can be changed by the program, but ideally we could just say that isn't
-supported, and hard code the rounding.
+This happens because SRoA sees that the temp alloca has is being memcpy'd into
+and out of and it has holes and it has to be conservative. If we knew about the
+holes, then this could be much much better.
+
+Having information about these holes would also improve memcpy (etc) lowering at
+llc time when it gets inlined, because we can use smaller transfers. This also
+avoids partial register stalls in some important cases.
//===---------------------------------------------------------------------===//