err on under-allocations and not on over-allocations (#1277)

* err on under-allocations and not on over-allications

* test code for over-allocated tape

* test allocated tape error too

* 8 byte tape on all -O levels

---------

Co-authored-by: William Moses <gh@wsmoses.com>

enzyme/Enzyme/Enzyme.cpp

@@ -1421,7 +1421,7 @@ class EnzymeBase {
  return true;
  }
  if (tapeType &&
- DL.getTypeSizeInBits(tapeType) < 8 * (size_t)allocatedTapeSize) {
+ DL.getTypeSizeInBits(tapeType) > 8 * (size_t)allocatedTapeSize) {
  auto bytes = DL.getTypeSizeInBits(tapeType) / 8;
  EmitFailure("Insufficient tape allocation size", CI->getDebugLoc(),
  CI, "need ", bytes, " bytes have ", allocatedTapeSize,
@@ -1491,7 +1491,7 @@ class EnzymeBase {
  return true;
  }
  if (tapeType &&
- DL.getTypeSizeInBits(tapeType) < 8 * (size_t)allocatedTapeSize) {
+ DL.getTypeSizeInBits(tapeType) > 8 * (size_t)allocatedTapeSize) {
  auto bytes = DL.getTypeSizeInBits(tapeType) / 8;
  EmitFailure("Insufficient tape allocation size", CI->getDebugLoc(),
  CI, "need ", bytes, " bytes have ", allocatedTapeSize,

enzyme/test/Integration/ReverseMode/allocatedtape.c

@@ -0,0 +1,59 @@
+// This should work on LLVM 7, 8, 9, however in CI the version of clang installed on Ubuntu 18.04 cannot load
+// a clang plugin properly without segfaulting on exit. This is fine on Ubuntu 20.04 or later LLVM versions...
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -O0 %s -S -emit-llvm -o - %loadClangEnzyme | %lli - ; fi
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -O1 %s -S -emit-llvm -o - %loadClangEnzyme | %lli - ; fi
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -O2 %s -S -emit-llvm -o - %loadClangEnzyme | %lli - ; fi
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -O3 %s -S -emit-llvm -o - %loadClangEnzyme | %lli - ; fi
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -O0 %s -S -emit-llvm -o - %loadClangEnzyme -mllvm -enzyme-inline=1 -S | %lli - ; fi
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -O1 %s -S -emit-llvm -o - %loadClangEnzyme -mllvm -enzyme-inline=1 -S | %lli - ; fi
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -O2 %s -S -emit-llvm -o - %loadClangEnzyme -mllvm -enzyme-inline=1 -S | %lli - ; fi
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -O3 %s -S -emit-llvm -o - %loadClangEnzyme -mllvm -enzyme-inline=1 -S | %lli - ; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -O0 %s -S -emit-llvm -o - %newLoadClangEnzyme | %lli - ; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -O1 %s -S -emit-llvm -o - %newLoadClangEnzyme | %lli - ; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -O2 %s -S -emit-llvm -o - %newLoadClangEnzyme | %lli - ; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -O3 %s -S -emit-llvm -o - %newLoadClangEnzyme | %lli - ; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -O0 %s -S -emit-llvm -o - %newLoadClangEnzyme -mllvm -enzyme-inline=1 -S | %lli - ; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -O1 %s -S -emit-llvm -o - %newLoadClangEnzyme -mllvm -enzyme-inline=1 -S | %lli - ; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -O2 %s -S -emit-llvm -o - %newLoadClangEnzyme -mllvm -enzyme-inline=1 -S | %lli - ; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -O3 %s -S -emit-llvm -o - %newLoadClangEnzyme -mllvm -enzyme-inline=1 -S | %lli - ; fi
+
+#include "test_utils.h"
+
+extern int enzyme_dup, enzyme_allocated, enzyme_tape;
+void __enzyme_autodiff(void*, ...);
+void* __enzyme_augmentfwd(void*, ...);
+void __enzyme_reverse(void*, ...);
+
+#define TAPE_SIZE 100 // v0.69 wants 24 bytes, so 100 should be more than enough
+
+// jac(f)(x) is [1 2*x1; 0 -1] as a function of the original inputs
+void f(double* x) {
+ x[0] += x[1]*x[1];
+ x[1] = -x[1];
+}
+
+void f_aug(double* x, double* x_b, void* tape) {
+ __enzyme_augmentfwd(f, enzyme_dup, x, x_b,
+ enzyme_allocated, TAPE_SIZE, enzyme_tape, tape);
+}
+
+void f_rev(double* x, double* x_b, void* tape) {
+ __enzyme_reverse(f, enzyme_dup, x, x_b,
+ enzyme_allocated, TAPE_SIZE, enzyme_tape, tape);
+}
+
+int main() {
+ double x0 = 1.2, x1 = 3.4;
+ double x0_b = 5.6, x1_b = 7.8;
+ double x[2] = {x0, x1};
+ double x_b[2] = {x0_b, x1_b};
+ char tape[TAPE_SIZE];
+
+ f_aug(x, x_b, tape);
+ f_rev(x, x_b, tape);
+
+ APPROX_EQ(x[0], x0 + x1*x1, 1e-10);
+ APPROX_EQ(x[1], -x1, 1e-10);
+ APPROX_EQ(x_b[0], x0_b*(1) + x1_b*(0), 1e-10);
+ APPROX_EQ(x_b[1], x0_b*(2*x1) + x1_b*(-1), 1e-10);
+}

enzyme/test/Integration/ReverseMode/allocatedtape_err.c

@@ -0,0 +1,22 @@
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -g -O0 %s -S -emit-llvm -o - %loadClangEnzyme -Xclang -verify; fi
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -g -O1 %s -S -emit-llvm -o - %loadClangEnzyme -Xclang -verify; fi
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -g -O2 %s -S -emit-llvm -o - %loadClangEnzyme -Xclang -verify; fi
+// RUN: if [ %llvmver -ge 10 ]; then %clang -std=c11 -g -O3 %s -S -emit-llvm -o - %loadClangEnzyme -Xclang -verify; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -g -O0 %s -S -emit-llvm -o - %newLoadClangEnzyme -Xclang -verify; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -g -O1 %s -S -emit-llvm -o - %newLoadClangEnzyme -Xclang -verify; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -g -O2 %s -S -emit-llvm -o - %newLoadClangEnzyme -Xclang -verify; fi
+// RUN: if [ %llvmver -ge 12 ]; then %clang -std=c11 -g -O3 %s -S -emit-llvm -o - %newLoadClangEnzyme -Xclang -verify; fi
+
+#include <math.h>
+
+extern int enzyme_allocated, enzyme_tape;
+void __enzyme_reverse(void*, ...);
+
+void f(double* x) {
+ x[0] = sin(x[0]);
+}
+
+void f_rev(double* x, double* x_b, void* tape) {
+ __enzyme_reverse(f, x, x_b, enzyme_allocated, 1, enzyme_tape, tape); // expected-error {{Enzyme: need 8 bytes have 1 bytes}}
+}
+

enzyme/test/Integration/ReverseMode/test_utils.h

@@ -23,8 +23,9 @@ static inline bool approx_fp_equality_double(double f1, double f2, double thresh
 
 #define APPROX_EQ(LHS, RHS, THRES) \
  { \
- if (__builtin_fabs(LHS - RHS) > THRES) { \
- fprintf(stderr, "Assertion Failed: fabs( [%s = %g] - [%s = %g] ) > %g at %s:%d (%s)\n", #LHS, LHS, #RHS, RHS, THRES, \
+ if (__builtin_fabs((LHS) - (RHS)) > THRES) { \
+ fprintf(stderr, "Assertion Failed: fabs( [%s = %g] - [%s = %g] ) > %g at %s:%d (%s)\n", \
+ #LHS, (double)(LHS), #RHS, (double)(RHS), THRES, \
  __FILE__, __LINE__, __PRETTY_FUNCTION__); \
  abort(); \
  } \