ErrorBounds.v 19.1 KB
Newer Older
1
(**
Heiko Becker's avatar
Heiko Becker committed
2 3
Proofs of general bounds on the error of arithmetic expressions.
This shortens soundness proofs later.
4
Bounds are explained in section 5, Deriving Computable Error Bounds
5
**)
6
Require Import Coq.Reals.Reals Coq.micromega.Psatz Coq.QArith.QArith Coq.QArith.Qreals.
Heiko Becker's avatar
Heiko Becker committed
7 8
Require Import Daisy.Infra.Abbrevs Daisy.Infra.RationalSimps Daisy.Infra.RealSimps Daisy.Infra.RealRationalProps.
Require Import Daisy.Environments Daisy.Infra.ExpressionAbbrevs.
Heiko Becker's avatar
Heiko Becker committed
9

10
Lemma const_abs_err_bounded (n:R) (nR:R) (nF:R) (E1 E2:env) (absenv:analysisResult) (m:mType):
11
  eval_exp E1 (Const M0 n) nR M0 ->
12
  eval_exp E2 (Const m n) nF m ->
13
  (Rabs (nR - nF) <= Rabs n * (Q2R (meps m)))%R.
14
Proof.
Heiko Becker's avatar
Heiko Becker committed
15
  intros eval_real eval_float.
16
  inversion eval_real; subst.
17
  rewrite delta_0_deterministic; auto.
18 19
  inversion eval_float; subst.
  unfold perturb; simpl.
20
  rewrite Rabs_err_simpl, Rabs_mult.
21
  apply Rmult_le_compat_l; [apply Rabs_pos | auto].
22 23
  simpl (meps M0) in *.
  apply (Rle_trans _ (Q2R 0) _); try auto.
24
  rewrite Q2R0_is_0; lra.
25 26
Qed.

27 28 29 30 31
(*
Lemma param_abs_err_bounded (P:precond) (n:nat) (nR:R) (nF:R) (E1 E2:env) (absenv:analysisResult):
  eval_exp 0%R E1 P (Param R n) nR ->
  eval_exp (Q2R machineEpsilon) E2 P (Param R n) nF ->
  (Rabs (nR - nF) <=  * (Q2R machineEpsilon))%R.
32
Proof.
Heiko Becker's avatar
Heiko Becker committed
33
  intros eval_real eval_float.
34
  inversion eval_real; subst.
35
  rewrite delta_0_deterministic; auto.
36 37
  inversion eval_float; subst.
  unfold perturb; simpl.
38 39 40
  exists v; split; try auto.

  rewrite H3 in H8; inversion H8.
41 42 43 44
  rewrite Rabs_err_simpl.
  repeat rewrite Rabs_mult.
  apply Rmult_le_compat_l; [ apply Rabs_pos | auto].
Qed.
45
*)
46

47
Lemma add_abs_err_bounded (e1:exp Q) (e1R:R) (e1F:R) (e2:exp Q) (e2R:R) (e2F:R)
48
      (vR:R) (vF:R) (E1 E2:env) (err1 err2 :Q) (m m1 m2:mType):
49
  eval_exp E1 (toREval (toRExp e1)) e1R M0 ->
50
  eval_exp E2 (toRExp e1) e1F m1->
51
  eval_exp E1 (toREval (toRExp e2)) e2R M0 ->
52
  eval_exp E2 (toRExp e2) e2F m2 ->
53
  eval_exp E1 (toREval (Binop Plus (toRExp e1) (toRExp e2))) vR M0 ->
54
  eval_exp (updEnv 2 m2 e2F (updEnv 1 m1 e1F emptyEnv)) (Binop Plus (Var R m1 1) (Var R m2 2)) vF m->
Heiko Becker's avatar
Heiko Becker committed
55 56
  (Rabs (e1R - e1F) <= Q2R err1)%R ->
  (Rabs (e2R - e2F) <= Q2R err2)%R ->
57
  (Rabs (vR - vF) <= Q2R err1 + Q2R err2 + (Rabs (e1F + e2F) * (Q2R (meps m))))%R.
58
Proof.
Heiko Becker's avatar
Heiko Becker committed
59
  intros e1_real e1_float e2_real e2_float plus_real plus_float bound_e1 bound_e2.
60
  (* Prove that e1R and e2R are the correct values and that vR is e1R + e2R *)
61 62 63
  inversion plus_real; subst.
  destruct m0; destruct m3; inversion H2;
      simpl in H4; rewrite Q2R0_is_0 in H4; auto.
64
  rewrite delta_0_deterministic in plus_real; auto.
65 66
  rewrite (delta_0_deterministic (evalBinop Plus v1 v2) delta); auto.
  unfold evalBinop in *; simpl in *.
67 68 69 70 71 72
  clear delta H4.
  rewrite (meps_0_deterministic (toRExp e1) H5 e1_real);
    rewrite (meps_0_deterministic (toRExp e2) H6 e2_real).
  rewrite (meps_0_deterministic (toRExp e1) H5 e1_real) in plus_real.
  rewrite (meps_0_deterministic (toRExp e2) H6 e2_real) in plus_real.
  clear H5 H6 v1 v2.
73 74 75
  (* Now unfold the float valued evaluation to get the deltas we need for the inequality *)
  inversion plus_float; subst.
  unfold perturb; simpl.
76
  inversion H6; subst; inversion H7; subst.
77
  unfold updEnv; simpl.
78 79 80
  unfold updEnv in H5,H8; simpl in *.
  symmetry in H5,H8.
  inversion H5; inversion H8; subst.
81
  (* We have now obtained all necessary values from the evaluations --> remove them for readability *)
82
  clear plus_float H7 H8 plus_real e1_real e1_float e2_real e2_float H5 H8.
83 84 85 86 87 88 89 90 91 92 93 94
  repeat rewrite Rmult_plus_distr_l.
  rewrite Rmult_1_r.
  rewrite Rsub_eq_Ropp_Rplus.
  repeat rewrite Ropp_plus_distr.
  rewrite plus_bounds_simplify.
  pose proof (Rabs_triang (e1R + - e1F) ((e2R + - e2F) + - ((e1F + e2F) * delta))).
  rewrite Rplus_assoc.
  eapply Rle_trans.
  apply H.
  pose proof (Rabs_triang (e2R + - e2F) (- ((e1F + e2F) * delta))).
  pose proof (Rplus_le_compat_l (Rabs (e1R + - e1F)) _ _ H0).
  eapply Rle_trans.
95
  apply H1.
96 97 98 99 100 101 102 103 104
  rewrite <- Rplus_assoc.
  repeat rewrite <- Rsub_eq_Ropp_Rplus.
  rewrite Rabs_Ropp.
  eapply Rplus_le_compat.
  - eapply Rplus_le_compat; auto.
  - rewrite Rabs_mult.
    eapply Rle_trans.
    eapply Rmult_le_compat_l.
    apply Rabs_pos.
105
    apply H3.
106 107 108 109 110 111
    apply Req_le; auto.
Qed.

(**
  Copy-Paste proof with minor differences, was easier then manipulating the evaluations and then applying the lemma
**)
112
Lemma subtract_abs_err_bounded (e1:exp Q) (e1R:R) (e1F:R) (e2:exp Q) (e2R:R)
113
      (e2F:R) (vR:R) (vF:R) (E1 E2:env) err1 err2 (m m1 m2:mType):
114
  eval_exp E1 (toREval (toRExp e1)) e1R M0 ->
115
  eval_exp E2 (toRExp e1) e1F m1 ->
116
  eval_exp E1 (toREval (toRExp e2)) e2R M0 ->
117
  eval_exp E2 (toRExp e2) e2F m2 ->
118
  eval_exp E1 (toREval (Binop Sub (toRExp e1) (toRExp e2))) vR M0 ->
119
  eval_exp (updEnv 2 m2 e2F (updEnv 1 m1 e1F emptyEnv)) (Binop Sub (Var R m1 1) (Var R m2 2)) vF m ->
Heiko Becker's avatar
Heiko Becker committed
120 121
  (Rabs (e1R - e1F) <= Q2R err1)%R ->
  (Rabs (e2R - e2F) <= Q2R err2)%R ->
122
  (Rabs (vR - vF) <= Q2R err1 + Q2R err2 + ((Rabs (e1F - e2F)) * (Q2R (meps m))))%R.
123
Proof.
Heiko Becker's avatar
Heiko Becker committed
124
  intros e1_real e1_float e2_real e2_float sub_real sub_float bound_e1 bound_e2.
125
  (* Prove that e1R and e2R are the correct values and that vR is e1R + e2R *)
126
  inversion sub_real; subst;
127 128
  destruct m0; destruct m3; inversion H2;
      simpl in H4; rewrite Q2R0_is_0 in H4; auto.
129 130
  rewrite delta_0_deterministic in sub_real; auto.
  rewrite delta_0_deterministic; auto.
131
  unfold evalBinop in *; simpl in *.
132 133 134 135 136 137
  clear delta H4.
  rewrite (meps_0_deterministic (toRExp e1) H5 e1_real);
    rewrite (meps_0_deterministic (toRExp e2) H6 e2_real).
  rewrite (meps_0_deterministic (toRExp e1) H5 e1_real) in sub_real.
  rewrite (meps_0_deterministic (toRExp e2) H6 e2_real) in sub_real.
  clear H5 H6 v1 v2.
138 139 140
  (* Now unfold the float valued evaluation to get the deltas we need for the inequality *)
  inversion sub_float; subst.
  unfold perturb; simpl.
141
  inversion H6; subst; inversion H7; subst.
142
  unfold updEnv; simpl.
143 144 145
  symmetry in H5, H8.
  unfold updEnv in H5, H8; simpl in H5, H8.
  inversion H5; inversion H8; subst.
146
  (* We have now obtained all necessary values from the evaluations --> remove them for readability *)
147
  clear sub_float H7 H8 sub_real e1_real e1_float e2_real e2_float H5 H8.
148 149 150 151 152
  repeat rewrite Rmult_plus_distr_l.
  rewrite Rmult_1_r.
  repeat rewrite Rsub_eq_Ropp_Rplus.
  repeat rewrite Ropp_plus_distr.
  rewrite plus_bounds_simplify.
153
  rewrite Ropp_involutive.
154 155
  rewrite Rplus_assoc.
  eapply Rle_trans.
156
  apply Rabs_triang.
157
  eapply Rle_trans.
158 159
  eapply Rplus_le_compat_l.
  apply Rabs_triang.
160
  rewrite <- Rplus_assoc.
161
  setoid_rewrite Rplus_comm at 4.
162 163
  repeat rewrite <- Rsub_eq_Ropp_Rplus.
  rewrite Rabs_Ropp.
164 165 166 167
  rewrite Rabs_minus_sym in bound_e2.
  apply Rplus_le_compat; [apply Rplus_le_compat; auto | ].
  rewrite Rabs_mult.
  eapply Rmult_le_compat_l; [apply Rabs_pos | auto].
168 169
Qed.

170
Lemma mult_abs_err_bounded (e1:exp Q) (e1R:R) (e1F:R) (e2:exp Q) (e2R:R) (e2F:R)
171
      (vR:R) (vF:R) (E1 E2:env) (m m1 m2:mType):
172
  eval_exp E1 (toREval (toRExp e1)) e1R M0 ->
173
  eval_exp E2 (toRExp e1) e1F m1 ->
174
  eval_exp E1 (toREval (toRExp e2)) e2R M0 ->
175
  eval_exp E2 (toRExp e2) e2F m2 ->
176
  eval_exp E1 (toREval (Binop Mult (toRExp e1) (toRExp e2))) vR M0 ->
177
  eval_exp (updEnv 2 m2 e2F (updEnv 1 m1 e1F emptyEnv)) (Binop Mult (Var R m1 1) (Var R m2 2)) vF m->
178
  (Rabs (vR - vF) <= Rabs (e1R * e2R - e1F * e2F) + Rabs (e1F * e2F) * (Q2R (meps m)))%R.
179
Proof.
Heiko Becker's avatar
Heiko Becker committed
180
  intros e1_real e1_float e2_real e2_float mult_real mult_float.
181
  (* Prove that e1R and e2R are the correct values and that vR is e1R * e2R *)
182
  inversion mult_real; subst;
183 184
    destruct m0; destruct m3; inversion H2;
      simpl in H4; rewrite Q2R0_is_0 in H4; auto.
185 186
  rewrite delta_0_deterministic in mult_real; auto.
  rewrite delta_0_deterministic; auto.
187
  unfold evalBinop in *; simpl in *.
188 189 190 191 192 193
  clear delta H4.
  rewrite (meps_0_deterministic (toRExp e1) H5 e1_real);
    rewrite (meps_0_deterministic (toRExp e2) H6 e2_real).
  rewrite (meps_0_deterministic (toRExp e1) H5 e1_real) in mult_real.
  rewrite (meps_0_deterministic (toRExp e2) H6 e2_real) in mult_real.
  clear H5 H6 v1 v2.
194 195 196
  (* Now unfold the float valued evaluation to get the deltas we need for the inequality *)
  inversion mult_float; subst.
  unfold perturb; simpl.
197 198
  inversion H6; subst; inversion H7; subst.
  symmetry in H5, H8;
199
    unfold updEnv in *; simpl in *.
200
  inversion H5; inversion H8; subst.
201
  (* We have now obtained all necessary values from the evaluations --> remove them for readability *)
202
  clear mult_float H7 H8 mult_real e1_real e1_float e2_real e2_float H5 H8.
203 204 205 206
  repeat rewrite Rmult_plus_distr_l.
  rewrite Rmult_1_r.
  rewrite Rsub_eq_Ropp_Rplus.
  rewrite Ropp_plus_distr.
Heiko Becker's avatar
Heiko Becker committed
207 208
  rewrite <- Rplus_assoc.
  setoid_rewrite <- Rsub_eq_Ropp_Rplus at 2.
209 210 211
  eapply Rle_trans.
  eapply Rabs_triang.
  eapply Rplus_le_compat_l.
Heiko Becker's avatar
Heiko Becker committed
212
  rewrite Rabs_Ropp.
213
  repeat rewrite Rabs_mult.
Heiko Becker's avatar
Heiko Becker committed
214
  eapply Rmult_le_compat_l; auto.
215 216
  rewrite <- Rabs_mult.
  apply Rabs_pos.
217 218
Qed.

219
Lemma div_abs_err_bounded (e1:exp Q) (e1R:R) (e1F:R) (e2:exp Q) (e2R:R) (e2F:R)
220
      (vR:R) (vF:R) (E1 E2:env) (m m1 m2:mType):
221
  eval_exp E1 (toREval (toRExp e1)) e1R M0 ->
222
  eval_exp E2 (toRExp e1) e1F m1 ->
223
  eval_exp E1 (toREval (toRExp e2)) e2R M0 ->
224
  eval_exp E2 (toRExp e2) e2F m2 ->
225
  eval_exp E1 (toREval (Binop Div (toRExp e1) (toRExp e2))) vR M0 ->
226
  eval_exp (updEnv 2 m2 e2F (updEnv 1 m1 e1F emptyEnv)) (Binop Div (Var R m1 1) (Var R m2 2)) vF m ->
227
  (Rabs (vR - vF) <= Rabs (e1R / e2R - e1F / e2F) + Rabs (e1F / e2F) * (Q2R (meps m)))%R.
228
Proof.
Heiko Becker's avatar
Heiko Becker committed
229
  intros e1_real e1_float e2_real e2_float div_real div_float.
230
  (* Prove that e1R and e2R are the correct values and that vR is e1R * e2R *)
231
  inversion div_real; subst;
232 233
  destruct m0; destruct m3; inversion H2;
    simpl in H4; rewrite Q2R0_is_0 in H4; auto.
234 235
  rewrite delta_0_deterministic in div_real; auto.
  rewrite delta_0_deterministic; auto.
236
  unfold evalBinop in *; simpl in *.
237 238 239 240 241 242
  clear delta H4.
  rewrite (meps_0_deterministic (toRExp e1) H5 e1_real);
    rewrite (meps_0_deterministic (toRExp e2) H6 e2_real).
  rewrite (meps_0_deterministic (toRExp e1) H5 e1_real) in div_real.
  rewrite (meps_0_deterministic (toRExp e2) H6 e2_real) in div_real.
  clear H5 H6 v1 v2.
243 244 245
  (* Now unfold the float valued evaluation to get the deltas we need for the inequality *)
  inversion div_float; subst.
  unfold perturb; simpl.
246 247
  inversion H6; subst; inversion H7; subst.
  symmetry in H5, H8;
248
    unfold updEnv in *; simpl in *.
249
  inversion H5; inversion H8; subst.
250
  (* We have now obtained all necessary values from the evaluations --> remove them for readability *)
251
  clear div_float H7 H8 div_real e1_real e1_float e2_real e2_float H5 H8.
252 253 254 255 256 257 258 259 260 261 262 263 264
  repeat rewrite Rmult_plus_distr_l.
  rewrite Rmult_1_r.
  rewrite Rsub_eq_Ropp_Rplus.
  rewrite Ropp_plus_distr.
  rewrite <- Rplus_assoc.
  setoid_rewrite <- Rsub_eq_Ropp_Rplus at 2.
  eapply Rle_trans.
  eapply Rabs_triang.
  eapply Rplus_le_compat_l.
  rewrite Rabs_Ropp.
  repeat rewrite Rabs_mult.
  eapply Rmult_le_compat_l; auto.
  apply Rabs_pos.
265 266
Qed.

267 268 269 270 271 272 273 274
Lemma err_prop_inversion_pos_real nF nR err elo ehi
      (float_iv_pos : (0 < elo - err)%R)
      (real_iv_pos : (0 < elo)%R)
      (err_bounded : (Rabs (nR - nF) <= err)%R)
      (valid_bounds_e2 : (elo <= nR <= ehi)%R)
      (valid_bounds_e2_err : (elo - err <= nF <= ehi + err)%R)
      (err_pos : (0 <= err)%R):
  (Rabs (/nR - / nF) <= err * / ((elo - err) * (elo- err)))%R.
275
Proof.
276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405
  rewrite Rabs_case_inverted in err_bounded.
  assert (0 < nF)%R as nF_pos by lra.
  destruct err_bounded as [ [diff_pos err_bounded] | [diff_neg err_bounded]].
  - cut (0 < /nF - / nR)%R.
    + intros abs_neg.
      rewrite Rabs_left; try lra.
      rewrite Rsub_eq_Ropp_Rplus, Ropp_plus_distr, Ropp_involutive.
      rewrite Ropp_inv_permute; try lra.
      apply (Rle_trans _ (/ - nR + / (nR - err))).
      * apply Rplus_le_compat_l.
        apply Rinv_le_contravar; lra.
      * rewrite equal_naming_inv; try lra.
        assert (- nR + (nR - err) = - err)%R as simplify_up by lra.
        rewrite simplify_up.
        unfold Rdiv.
        repeat(rewrite <- Ropp_mult_distr_l); rewrite <- Ropp_inv_permute.
        {
          rewrite <- Ropp_mult_distr_r, Ropp_involutive.
          apply Rmult_le_compat_l; try lra.
          apply Rinv_le_contravar.
          - apply Rmult_0_lt_preserving; lra.
          - apply Rmult_le_compat; lra.
        }
        { assert (0 < nR * (nR - err))%R by (apply Rmult_0_lt_preserving; lra); lra. }
    + cut (/ nR < /nF)%R.
      * intros; lra.
      * apply Rinv_lt_contravar; try lra.
        apply Rmult_0_lt_preserving; lra.
  - cut (0 <= /nR - /nF)%R.
    + intros abs_pos.
      rewrite Rabs_right; try lra.
      rewrite Rsub_eq_Ropp_Rplus, Ropp_plus_distr, Ropp_involutive in err_bounded.
      rewrite Rsub_eq_Ropp_Rplus.
      apply (Rle_trans _ (/nR - / (nR + err))).
      * apply Rplus_le_compat_l.
        apply Ropp_le_contravar.
        apply Rinv_le_contravar; lra.
      * rewrite Rsub_eq_Ropp_Rplus, Ropp_inv_permute; try lra.
        rewrite equal_naming_inv; try lra.
        assert (nR + - (nR + err) = - err)%R as simpl_up by lra.
        rewrite simpl_up.
        unfold Rdiv.
        rewrite <- Ropp_mult_distr_l, <- Ropp_mult_distr_r, <- Ropp_inv_permute.
        { rewrite <- Ropp_mult_distr_r. rewrite Ropp_involutive.
          apply Rmult_le_compat_l; try auto.
          apply Rinv_le_contravar.
          - apply Rmult_0_lt_preserving; lra.
          - apply Rmult_le_compat; lra.
        }
        { assert (0 < nR * (nR + err))%R by (apply Rmult_0_lt_preserving; lra); lra. }
    + cut (/nF <= /nR)%R.
      * intros; lra.
      * apply Rinv_le_contravar; try lra.
Qed.

Lemma err_prop_inversion_pos nF nR err (elo ehi:Q)
      (float_iv_pos : (Q2R 0 < Q2R (elo - err))%R)
      (real_iv_pos : (Q2R 0 < Q2R elo)%R)
      (err_bounded : (Rabs (nR - nF) <= Q2R err)%R)
      (valid_bounds_e2 : (Q2R elo <= nR <= Q2R ehi)%R)
      (valid_bounds_e2_err : (Q2R elo - Q2R err <= nF <= Q2R ehi + Q2R err)%R)
      (err_pos : (0 <= Q2R err)%R):
  (Rabs (/nR - / nF) <= Q2R err * / ((Q2R elo- Q2R err) * (Q2R elo- Q2R err)))%R.
Proof.
  eapply err_prop_inversion_pos_real; try rewrite <- Q2R0_is_0; eauto.
  rewrite <- Q2R_minus; auto.
  rewrite Q2R0_is_0; auto.
Qed.

Lemma err_prop_inversion_neg_real nF nR err elo ehi
      (float_iv_neg : (ehi + err < 0)%R)
      (real_iv_neg : (ehi < 0)%R)
      (err_bounded : (Rabs (nR - nF) <= err)%R)
      (valid_bounds_e : (elo <= nR <= ehi)%R)
      (valid_bounds_e_err : (elo - err <= nF <= ehi + err)%R)
      (err_pos : (0 <= err)%R):
  (Rabs (/nR - / nF) <= err * / ((ehi + err) * (ehi + err)))%R.
Proof.
  rewrite Rabs_case_inverted in err_bounded.
  assert (nF < 0)%R as nF_neg by lra.
  destruct err_bounded as [ [diff_pos err_bounded] | [diff_neg err_bounded]].
  - cut (0 < /nF - / nR)%R.
    + intros abs_neg.
      rewrite Rabs_left; try lra.
      rewrite Rsub_eq_Ropp_Rplus, Ropp_plus_distr, Ropp_involutive.
      rewrite Ropp_inv_permute; try lra.
      apply (Rle_trans _ (/ - nR + / (nR - err))).
      * apply Rplus_le_compat_l.
        assert (0 < - nF)%R by lra.
        assert (0 < - (nR - err))%R by lra.
        assert (nR - err <= nF)%R as nR_lower by lra.
        apply Ropp_le_contravar in nR_lower.
        apply Rinv_le_contravar in nR_lower; try lra.
        repeat (rewrite <- Ropp_inv_permute in nR_lower; try lra).
      * rewrite equal_naming_inv; try lra.
        assert (- nR + (nR - err) = - err)%R as simplify_up by lra.
        rewrite simplify_up.
        unfold Rdiv.
        repeat(rewrite <- Ropp_mult_distr_l); rewrite <- Ropp_inv_permute.
        {
          rewrite <- Ropp_mult_distr_r, Ropp_involutive.
          apply Rmult_le_compat_l; try lra.
          apply Rinv_le_contravar.
          - apply Rmult_lt_0_inverting; lra.
          - eapply Rle_trans.
            eapply Rmult_le_compat_neg_l; try lra.
            instantiate (1 := (nR - err)%R); try lra.
            setoid_rewrite Rmult_comm.
            eapply Rmult_le_compat_neg_l; lra.
        }
        { assert (0 < nR * (nR - err))%R by (apply Rmult_lt_0_inverting; lra); lra. }
    + cut (/ nR < /nF)%R.
      * intros; lra.
      * apply Rinv_lt_contravar; try lra.
        apply Rmult_lt_0_inverting; lra.
  - cut (0 <= /nR - /nF)%R.
    + intros abs_pos.
      rewrite Rabs_right; try lra.
      rewrite Rsub_eq_Ropp_Rplus, Ropp_plus_distr, Ropp_involutive in err_bounded.
      rewrite Rsub_eq_Ropp_Rplus.
      apply (Rle_trans _ (/nR - / (nR + err))).
      * apply Rplus_le_compat_l.
        apply Ropp_le_contravar.
        assert (0 < - nF)%R by lra.
        assert (0 < - (nR + err))%R by lra.
        assert (nF <= nR + err)%R as nR_upper by lra.
        apply Ropp_le_contravar in nR_upper.
        apply Rinv_le_contravar in nR_upper; try lra.
        repeat (rewrite <- Ropp_inv_permute in nR_upper; try lra).
      * rewrite Rsub_eq_Ropp_Rplus, Ropp_inv_permute; try lra.
406
        rewrite equal_naming_inv; try lra.
407 408
        assert (nR + - (nR + err) = - err)%R as simpl_up by lra.
        rewrite simpl_up.
409
        unfold Rdiv.
410
        rewrite <- Ropp_mult_distr_l, <- Ropp_mult_distr_r, <- Ropp_inv_permute.
411 412 413
        { rewrite <- Ropp_mult_distr_r. rewrite Ropp_involutive.
          apply Rmult_le_compat_l; try auto.
          apply Rinv_le_contravar.
414 415 416 417 418 419 420 421 422 423 424 425 426 427
          - apply Rmult_lt_0_inverting; lra.
          - eapply Rle_trans.
            eapply Rmult_le_compat_neg_l; try lra.
            instantiate (1:= (nR + err)%R); try lra.
            setoid_rewrite Rmult_comm.
            eapply Rmult_le_compat_neg_l; lra.
        }
        { assert (0 < nR * (nR + err))%R by (apply Rmult_lt_0_inverting; lra); lra. }
    + cut (/nF <= /nR)%R.
      * intros; lra.
      * assert (nR <= nF)%R by lra.
        assert (- nF <= - nR)%R as le_inv by lra.
        apply Rinv_le_contravar in le_inv; try lra.
        repeat (rewrite <- Ropp_inv_permute in le_inv; try lra).
428 429
Qed.

430 431 432 433 434 435 436 437
Lemma err_prop_inversion_neg nF nR err (elo ehi:Q)
      (float_iv_neg : (Q2R (ehi + err) < Q2R 0)%R)
      (real_iv_neg : (Q2R ehi < Q2R 0)%R)
      (err_bounded : (Rabs (nR - nF) <= Q2R err)%R)
      (valid_bounds_e : (Q2R elo <= nR <= Q2R ehi)%R)
      (valid_bounds_e_err : (Q2R elo - Q2R err <= nF <= Q2R ehi + Q2R err)%R)
      (err_pos : (0 <= Q2R err)%R):
  (Rabs (/nR - / nF) <= Q2R err * / ((Q2R ehi + Q2R err) * (Q2R ehi + Q2R err)))%R.
438
Proof.
439 440 441 442 443 444
  eapply err_prop_inversion_neg_real; try rewrite <- Q2R0_is_0; try lra.
  rewrite <- Q2R_plus ; auto.
  apply valid_bounds_e.
  auto.
  rewrite Q2R0_is_0; auto.
Qed.
445

446 447 448 449
Lemma round_abs_err_bounded (e:exp R) (nR nF1 nF:R) (E1 E2: env) (err:R) (machineEpsilon m:mType):
  eval_exp E1 (toREval e) nR M0 ->
  eval_exp E2 e nF1 m ->
  eval_exp (updEnv 1 m nF1 emptyEnv) (toRExp (Downcast machineEpsilon (Var Q m 1))) nF machineEpsilon->
450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470
  (Rabs (nR - nF1) <= err)%R ->
  (Rabs (nR - nF) <= err + (Rabs nF1) * Q2R (meps machineEpsilon))%R.
Proof.
  intros eval_real eval_float eval_float_rnd err_bounded.
  replace (nR - nF)%R with ((nR - nF1) + (nF1 - nF))%R by lra.
  eapply Rle_trans.
  apply (Rabs_triang (nR - nF1) (nF1 - nF)).
  apply (Rle_trans _ (err + Rabs (nF1 - nF))  _).
  - apply Rplus_le_compat_r; assumption.
  - apply Rplus_le_compat_l.
    inversion eval_float_rnd; subst.
    inversion H5; subst.
    inversion H7.
    unfold perturb; simpl.
    replace (v1 - v1 * (1 + delta))%R with (- (v1 * delta))%R by lra.
    replace (Rabs (-(v1*delta))) with (Rabs (v1*delta)) by (symmetry; apply Rabs_Ropp).
    rewrite Rabs_mult.
    apply Rmult_le_compat_l.
    + apply Rabs_pos.
    + auto. 
Qed.