aboutsummaryrefslogtreecommitdiffstats
path: root/flocq/Prop/Relative.v
diff options
context:
space:
mode:
Diffstat (limited to 'flocq/Prop/Relative.v')
-rw-r--r--flocq/Prop/Relative.v983
1 files changed, 983 insertions, 0 deletions
diff --git a/flocq/Prop/Relative.v b/flocq/Prop/Relative.v
new file mode 100644
index 00000000..b936f2f7
--- /dev/null
+++ b/flocq/Prop/Relative.v
@@ -0,0 +1,983 @@
+(**
+This file is part of the Flocq formalization of floating-point
+arithmetic in Coq: http://flocq.gforge.inria.fr/
+
+Copyright (C) 2010-2018 Sylvie Boldo
+#<br />#
+Copyright (C) 2010-2018 Guillaume Melquiond
+
+This library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Lesser General Public
+License as published by the Free Software Foundation; either
+version 3 of the License, or (at your option) any later version.
+
+This library is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+COPYING file for more details.
+*)
+
+(** * Relative error of the roundings *)
+Require Import Core.
+Require Import Psatz. (* for lra *)
+
+Section Fprop_relative.
+
+Variable beta : radix.
+Notation bpow e := (bpow beta e).
+
+Section Fprop_relative_generic.
+
+Variable fexp : Z -> Z.
+Context { prop_exp : Valid_exp fexp }.
+
+Section relative_error_conversion.
+
+Variable rnd : R -> Z.
+Context { valid_rnd : Valid_rnd rnd }.
+
+Lemma relative_error_lt_conversion :
+ forall x b, (0 < b)%R ->
+ (x <> 0 -> Rabs (round beta fexp rnd x - x) < b * Rabs x)%R ->
+ exists eps,
+ (Rabs eps < b)%R /\ round beta fexp rnd x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros x b Hb0 Hxb.
+destruct (Req_dec x 0) as [Hx0|Hx0].
+(* *)
+exists 0%R.
+split.
+now rewrite Rabs_R0.
+rewrite Hx0, Rmult_0_l.
+apply round_0...
+(* *)
+specialize (Hxb Hx0).
+exists ((round beta fexp rnd x - x) / x)%R.
+split. 2: now field.
+unfold Rdiv.
+rewrite Rabs_mult.
+apply Rmult_lt_reg_r with (Rabs x).
+now apply Rabs_pos_lt.
+rewrite Rmult_assoc, <- Rabs_mult.
+rewrite Rinv_l with (1 := Hx0).
+now rewrite Rabs_R1, Rmult_1_r.
+Qed.
+
+Lemma relative_error_le_conversion :
+ forall x b, (0 <= b)%R ->
+ (Rabs (round beta fexp rnd x - x) <= b * Rabs x)%R ->
+ exists eps,
+ (Rabs eps <= b)%R /\ round beta fexp rnd x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros x b Hb0 Hxb.
+destruct (Req_dec x 0) as [Hx0|Hx0].
+(* *)
+exists 0%R.
+split.
+now rewrite Rabs_R0.
+rewrite Hx0, Rmult_0_l.
+apply round_0...
+(* *)
+exists ((round beta fexp rnd x - x) / x)%R.
+split. 2: now field.
+unfold Rdiv.
+rewrite Rabs_mult.
+apply Rmult_le_reg_r with (Rabs x).
+now apply Rabs_pos_lt.
+rewrite Rmult_assoc, <- Rabs_mult.
+rewrite Rinv_l with (1 := Hx0).
+now rewrite Rabs_R1, Rmult_1_r.
+Qed.
+
+Lemma relative_error_le_conversion_inv :
+ forall x b,
+ (exists eps,
+ (Rabs eps <= b)%R /\ round beta fexp rnd x = (x * (1 + eps))%R) ->
+ (Rabs (round beta fexp rnd x - x) <= b * Rabs x)%R.
+Proof with auto with typeclass_instances.
+intros x b (eps, (Beps, Heps)).
+assert (Pb : (0 <= b)%R); [now revert Beps; apply Rle_trans, Rabs_pos|].
+rewrite Heps; replace (_ - _)%R with (eps * x)%R; [|ring].
+now rewrite Rabs_mult; apply Rmult_le_compat_r; [apply Rabs_pos|].
+Qed.
+
+Lemma relative_error_le_conversion_round_inv :
+ forall x b,
+ (exists eps,
+ (Rabs eps <= b)%R /\ x = (round beta fexp rnd x * (1 + eps))%R) ->
+ (Rabs (round beta fexp rnd x - x) <= b * Rabs (round beta fexp rnd x))%R.
+Proof with auto with typeclass_instances.
+intros x b.
+set (rx := round _ _ _ _).
+intros (eps, (Beps, Heps)).
+assert (Pb : (0 <= b)%R); [now revert Beps; apply Rle_trans, Rabs_pos|].
+rewrite Heps; replace (_ - _)%R with (- (eps * rx))%R; [|ring].
+now rewrite Rabs_Ropp, Rabs_mult; apply Rmult_le_compat_r; [apply Rabs_pos|].
+Qed.
+
+End relative_error_conversion.
+
+Variable emin p : Z.
+Hypothesis Hmin : forall k, (emin < k)%Z -> (p <= k - fexp k)%Z.
+
+Variable rnd : R -> Z.
+Context { valid_rnd : Valid_rnd rnd }.
+
+Theorem relative_error :
+ forall x,
+ (bpow emin <= Rabs x)%R ->
+ (Rabs (round beta fexp rnd x - x) < bpow (-p + 1) * Rabs x)%R.
+Proof with auto with typeclass_instances.
+intros x Hx.
+assert (Hx': (x <> 0)%R).
+intros T; contradict Hx; rewrite T, Rabs_R0.
+apply Rlt_not_le, bpow_gt_0.
+apply Rlt_le_trans with (ulp beta fexp x)%R.
+now apply error_lt_ulp...
+rewrite ulp_neq_0; trivial.
+unfold cexp.
+destruct (mag beta x) as (ex, He).
+simpl.
+specialize (He Hx').
+apply Rle_trans with (bpow (-p + 1) * bpow (ex - 1))%R.
+rewrite <- bpow_plus.
+apply bpow_le.
+assert (emin < ex)%Z.
+apply (lt_bpow beta).
+apply Rle_lt_trans with (2 := proj2 He).
+exact Hx.
+generalize (Hmin ex).
+omega.
+apply Rmult_le_compat_l.
+apply bpow_ge_0.
+apply He.
+Qed.
+
+(** 1+#&epsilon;# property in any rounding *)
+Theorem relative_error_ex :
+ forall x,
+ (bpow emin <= Rabs x)%R ->
+ exists eps,
+ (Rabs eps < bpow (-p + 1))%R /\ round beta fexp rnd x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros x Hx.
+apply relative_error_lt_conversion...
+apply bpow_gt_0.
+intros _.
+now apply relative_error.
+Qed.
+
+Theorem relative_error_F2R_emin :
+ forall m, let x := F2R (Float beta m emin) in
+ (x <> 0)%R ->
+ (Rabs (round beta fexp rnd x - x) < bpow (-p + 1) * Rabs x)%R.
+Proof.
+intros m x Hx.
+apply relative_error.
+unfold x.
+rewrite <- F2R_Zabs.
+apply bpow_le_F2R.
+apply lt_F2R with beta emin.
+rewrite F2R_0, F2R_Zabs.
+now apply Rabs_pos_lt.
+Qed.
+
+Theorem relative_error_F2R_emin_ex :
+ forall m, let x := F2R (Float beta m emin) in
+ exists eps,
+ (Rabs eps < bpow (-p + 1))%R /\ round beta fexp rnd x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros m x.
+apply relative_error_lt_conversion...
+apply bpow_gt_0.
+now apply relative_error_F2R_emin.
+Qed.
+
+Theorem relative_error_round :
+ (0 < p)%Z ->
+ forall x,
+ (bpow emin <= Rabs x)%R ->
+ (Rabs (round beta fexp rnd x - x) < bpow (-p + 1) * Rabs (round beta fexp rnd x))%R.
+Proof with auto with typeclass_instances.
+intros Hp x Hx.
+assert (Hx': (x <> 0)%R).
+intros T; contradict Hx; rewrite T, Rabs_R0.
+apply Rlt_not_le, bpow_gt_0.
+apply Rlt_le_trans with (ulp beta fexp x)%R.
+now apply error_lt_ulp.
+rewrite ulp_neq_0; trivial.
+unfold cexp.
+destruct (mag beta x) as (ex, He).
+simpl.
+specialize (He Hx').
+assert (He': (emin < ex)%Z).
+apply (lt_bpow beta).
+apply Rle_lt_trans with (2 := proj2 He).
+exact Hx.
+apply Rle_trans with (bpow (-p + 1) * bpow (ex - 1))%R.
+rewrite <- bpow_plus.
+apply bpow_le.
+generalize (Hmin ex).
+omega.
+apply Rmult_le_compat_l.
+apply bpow_ge_0.
+generalize He.
+apply round_abs_abs...
+clear rnd valid_rnd x Hx Hx' He.
+intros rnd valid_rnd x _ Hx.
+rewrite <- (round_generic beta fexp rnd (bpow (ex - 1))).
+now apply round_le.
+apply generic_format_bpow.
+ring_simplify (ex - 1 + 1)%Z.
+generalize (Hmin ex).
+omega.
+Qed.
+
+Theorem relative_error_round_F2R_emin :
+ (0 < p)%Z ->
+ forall m, let x := F2R (Float beta m emin) in
+ (x <> 0)%R ->
+ (Rabs (round beta fexp rnd x - x) < bpow (-p + 1) * Rabs (round beta fexp rnd x))%R.
+Proof.
+intros Hp m x Hx.
+apply relative_error_round.
+exact Hp.
+unfold x.
+rewrite <- F2R_Zabs.
+apply bpow_le_F2R.
+apply lt_F2R with beta emin.
+rewrite F2R_0, F2R_Zabs.
+now apply Rabs_pos_lt.
+Qed.
+
+Variable choice : Z -> bool.
+
+Theorem relative_error_N :
+ forall x,
+ (bpow emin <= Rabs x)%R ->
+ (Rabs (round beta fexp (Znearest choice) x - x) <= /2 * bpow (-p + 1) * Rabs x)%R.
+Proof.
+intros x Hx.
+apply Rle_trans with (/2 * ulp beta fexp x)%R.
+now apply error_le_half_ulp.
+rewrite Rmult_assoc.
+apply Rmult_le_compat_l.
+apply Rlt_le.
+apply Rinv_0_lt_compat.
+now apply IZR_lt.
+assert (Hx': (x <> 0)%R).
+intros H.
+apply Rlt_not_le with (2 := Hx).
+rewrite H, Rabs_R0.
+apply bpow_gt_0.
+rewrite ulp_neq_0; trivial.
+unfold cexp.
+destruct (mag beta x) as (ex, He).
+simpl.
+specialize (He Hx').
+apply Rle_trans with (bpow (-p + 1) * bpow (ex - 1))%R.
+rewrite <- bpow_plus.
+apply bpow_le.
+assert (emin < ex)%Z.
+apply (lt_bpow beta).
+apply Rle_lt_trans with (2 := proj2 He).
+exact Hx.
+generalize (Hmin ex).
+omega.
+apply Rmult_le_compat_l.
+apply bpow_ge_0.
+apply He.
+Qed.
+
+(** 1+#&epsilon;# property in rounding to nearest *)
+Theorem relative_error_N_ex :
+ forall x,
+ (bpow emin <= Rabs x)%R ->
+ exists eps,
+ (Rabs eps <= /2 * bpow (-p + 1))%R /\ round beta fexp (Znearest choice) x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros x Hx.
+apply relative_error_le_conversion...
+apply Rlt_le.
+apply Rmult_lt_0_compat.
+apply Rinv_0_lt_compat.
+now apply IZR_lt.
+apply bpow_gt_0.
+now apply relative_error_N.
+Qed.
+
+Theorem relative_error_N_F2R_emin :
+ forall m, let x := F2R (Float beta m emin) in
+ (Rabs (round beta fexp (Znearest choice) x - x) <= /2 * bpow (-p + 1) * Rabs x)%R.
+Proof with auto with typeclass_instances.
+intros m x.
+destruct (Req_dec x 0) as [Hx|Hx].
+(* . *)
+rewrite Hx, round_0...
+unfold Rminus.
+rewrite Rplus_0_l, Rabs_Ropp, Rabs_R0.
+rewrite Rmult_0_r.
+apply Rle_refl.
+(* . *)
+apply relative_error_N.
+unfold x.
+rewrite <- F2R_Zabs.
+apply bpow_le_F2R.
+apply lt_F2R with beta emin.
+rewrite F2R_0, F2R_Zabs.
+now apply Rabs_pos_lt.
+Qed.
+
+Theorem relative_error_N_F2R_emin_ex :
+ forall m, let x := F2R (Float beta m emin) in
+ exists eps,
+ (Rabs eps <= /2 * bpow (-p + 1))%R /\ round beta fexp (Znearest choice) x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros m x.
+apply relative_error_le_conversion...
+apply Rlt_le.
+apply Rmult_lt_0_compat.
+apply Rinv_0_lt_compat.
+now apply IZR_lt.
+apply bpow_gt_0.
+now apply relative_error_N_F2R_emin.
+Qed.
+
+Theorem relative_error_N_round :
+ (0 < p)%Z ->
+ forall x,
+ (bpow emin <= Rabs x)%R ->
+ (Rabs (round beta fexp (Znearest choice) x - x) <= /2 * bpow (-p + 1) * Rabs (round beta fexp (Znearest choice) x))%R.
+Proof with auto with typeclass_instances.
+intros Hp x Hx.
+apply Rle_trans with (/2 * ulp beta fexp x)%R.
+now apply error_le_half_ulp.
+rewrite Rmult_assoc.
+apply Rmult_le_compat_l.
+apply Rlt_le.
+apply Rinv_0_lt_compat.
+now apply IZR_lt.
+assert (Hx': (x <> 0)%R).
+intros H.
+apply Rlt_not_le with (2 := Hx).
+rewrite H, Rabs_R0.
+apply bpow_gt_0.
+rewrite ulp_neq_0; trivial.
+unfold cexp.
+destruct (mag beta x) as (ex, He).
+simpl.
+specialize (He Hx').
+assert (He': (emin < ex)%Z).
+apply (lt_bpow beta).
+apply Rle_lt_trans with (2 := proj2 He).
+exact Hx.
+apply Rle_trans with (bpow (-p + 1) * bpow (ex - 1))%R.
+rewrite <- bpow_plus.
+apply bpow_le.
+generalize (Hmin ex).
+omega.
+apply Rmult_le_compat_l.
+apply bpow_ge_0.
+generalize He.
+apply round_abs_abs...
+clear rnd valid_rnd x Hx Hx' He.
+intros rnd valid_rnd x _ Hx.
+rewrite <- (round_generic beta fexp rnd (bpow (ex - 1))).
+now apply round_le.
+apply generic_format_bpow.
+ring_simplify (ex - 1 + 1)%Z.
+generalize (Hmin ex).
+omega.
+Qed.
+
+Theorem relative_error_N_round_F2R_emin :
+ (0 < p)%Z ->
+ forall m, let x := F2R (Float beta m emin) in
+ (Rabs (round beta fexp (Znearest choice) x - x) <= /2 * bpow (-p + 1) * Rabs (round beta fexp (Znearest choice) x))%R.
+Proof with auto with typeclass_instances.
+intros Hp m x.
+destruct (Req_dec x 0) as [Hx|Hx].
+(* . *)
+rewrite Hx, round_0...
+unfold Rminus.
+rewrite Rplus_0_l, Rabs_Ropp, Rabs_R0.
+rewrite Rmult_0_r.
+apply Rle_refl.
+(* . *)
+apply relative_error_N_round with (1 := Hp).
+unfold x.
+rewrite <- F2R_Zabs.
+apply bpow_le_F2R.
+apply lt_F2R with beta emin.
+rewrite F2R_0, F2R_Zabs.
+now apply Rabs_pos_lt.
+Qed.
+
+End Fprop_relative_generic.
+
+Section Fprop_relative_FLX.
+
+Variable prec : Z.
+Variable Hp : Z.lt 0 prec.
+
+Lemma relative_error_FLX_aux :
+ forall k, (prec <= k - FLX_exp prec k)%Z.
+Proof.
+intros k.
+unfold FLX_exp.
+omega.
+Qed.
+
+Variable rnd : R -> Z.
+Context { valid_rnd : Valid_rnd rnd }.
+
+Theorem relative_error_FLX :
+ forall x,
+ (x <> 0)%R ->
+ (Rabs (round beta (FLX_exp prec) rnd x - x) < bpow (-prec + 1) * Rabs x)%R.
+Proof with auto with typeclass_instances.
+intros x Hx.
+destruct (mag beta x) as (ex, He).
+specialize (He Hx).
+apply relative_error with (ex - 1)%Z...
+intros k _.
+apply relative_error_FLX_aux.
+apply He.
+Qed.
+
+(** 1+#&epsilon;# property in any rounding in FLX *)
+Theorem relative_error_FLX_ex :
+ forall x,
+ exists eps,
+ (Rabs eps < bpow (-prec + 1))%R /\ round beta (FLX_exp prec) rnd x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros x.
+apply relative_error_lt_conversion...
+apply bpow_gt_0.
+now apply relative_error_FLX.
+Qed.
+
+Theorem relative_error_FLX_round :
+ forall x,
+ (x <> 0)%R ->
+ (Rabs (round beta (FLX_exp prec) rnd x - x) < bpow (-prec + 1) * Rabs (round beta (FLX_exp prec) rnd x))%R.
+Proof with auto with typeclass_instances.
+intros x Hx.
+destruct (mag beta x) as (ex, He).
+specialize (He Hx).
+apply relative_error_round with (ex - 1)%Z...
+intros k _.
+apply relative_error_FLX_aux.
+apply He.
+Qed.
+
+Variable choice : Z -> bool.
+
+Theorem relative_error_N_FLX :
+ forall x,
+ (Rabs (round beta (FLX_exp prec) (Znearest choice) x - x) <= /2 * bpow (-prec + 1) * Rabs x)%R.
+Proof with auto with typeclass_instances.
+intros x.
+destruct (Req_dec x 0) as [Hx|Hx].
+(* . *)
+rewrite Hx, round_0...
+unfold Rminus.
+rewrite Rplus_0_l, Rabs_Ropp, Rabs_R0.
+rewrite Rmult_0_r.
+apply Rle_refl.
+(* . *)
+destruct (mag beta x) as (ex, He).
+specialize (He Hx).
+apply relative_error_N with (ex - 1)%Z...
+intros k _.
+apply relative_error_FLX_aux.
+apply He.
+Qed.
+
+(** unit roundoff *)
+Definition u_ro := (/2 * bpow (-prec + 1))%R.
+
+Lemma u_ro_pos : (0 <= u_ro)%R.
+Proof. apply Rmult_le_pos; [lra|apply bpow_ge_0]. Qed.
+
+Lemma u_ro_lt_1 : (u_ro < 1)%R.
+Proof.
+unfold u_ro; apply (Rmult_lt_reg_l 2); [lra|].
+rewrite <-Rmult_assoc, Rinv_r, Rmult_1_l, Rmult_1_r; [|lra].
+apply (Rle_lt_trans _ (bpow 0));
+ [apply bpow_le; omega|simpl; lra].
+Qed.
+
+Lemma u_rod1pu_ro_pos : (0 <= u_ro / (1 + u_ro))%R.
+Proof.
+apply Rmult_le_pos; [|apply Rlt_le, Rinv_0_lt_compat];
+assert (H := u_ro_pos); lra.
+Qed.
+
+Lemma u_rod1pu_ro_le_u_ro : (u_ro / (1 + u_ro) <= u_ro)%R.
+Proof.
+assert (Pu_ro := u_ro_pos).
+apply (Rmult_le_reg_r (1 + u_ro)); [lra|].
+unfold Rdiv; rewrite Rmult_assoc, Rinv_l; [|lra].
+assert (0 <= u_ro * u_ro)%R; [apply Rmult_le_pos|]; lra.
+Qed.
+
+Theorem relative_error_N_FLX' :
+ forall x,
+ (Rabs (round beta (FLX_exp prec) (Znearest choice) x - x)
+ <= u_ro / (1 + u_ro) * Rabs x)%R.
+Proof with auto with typeclass_instances.
+intro x.
+assert (Pu_ro : (0 <= u_ro)%R).
+{ apply Rmult_le_pos; [lra|apply bpow_ge_0]. }
+destruct (Req_dec x 0) as [Zx|Nzx].
+{ rewrite Zx, Rabs_R0, Rmult_0_r, round_0...
+ now unfold Rminus; rewrite Rplus_0_l, Rabs_Ropp, Rabs_R0; right. }
+set (ufpx := bpow (mag beta x - 1)%Z).
+set (rx := round _ _ _ _).
+assert (Pufpx : (0 <= ufpx)%R); [now apply bpow_ge_0|].
+assert (H_2_1 : (Rabs (rx - x) <= u_ro * ufpx)%R).
+{ refine (Rle_trans _ _ _ (error_le_half_ulp _ _ _ _) _);
+ [now apply FLX_exp_valid|right].
+ unfold ulp, cexp, FLX_exp, u_ro, ufpx; rewrite (Req_bool_false _ _ Nzx).
+ rewrite Rmult_assoc, <-bpow_plus; do 2 f_equal; ring. }
+assert (H_2_3 : (ufpx + Rabs (rx - x) <= Rabs x)%R).
+{ apply (Rplus_le_reg_r (- ufpx)); ring_simplify.
+ destruct (Rle_or_lt 0 x) as [Sx|Sx].
+ { apply (Rle_trans _ (Rabs (ufpx - x))).
+ { apply round_N_pt; [now apply FLX_exp_valid|].
+ apply generic_format_bpow; unfold FLX_exp; lia. }
+ rewrite Rabs_minus_sym, Rabs_pos_eq.
+ { now rewrite Rabs_pos_eq; [right; ring|]. }
+ apply (Rplus_le_reg_r ufpx); ring_simplify.
+ now rewrite <-(Rabs_pos_eq _ Sx); apply bpow_mag_le. }
+ apply (Rle_trans _ (Rabs (- ufpx - x))).
+ { apply round_N_pt; [now apply FLX_exp_valid|].
+ apply generic_format_opp, generic_format_bpow; unfold FLX_exp; lia. }
+ rewrite Rabs_pos_eq; [now rewrite Rabs_left; [right|]|].
+ apply (Rplus_le_reg_r x); ring_simplify.
+ rewrite <-(Ropp_involutive x); apply Ropp_le_contravar; unfold ufpx.
+ rewrite <-mag_opp, <-Rabs_pos_eq; [apply bpow_mag_le|]; lra. }
+assert (H : (Rabs ((rx - x) / x) <= u_ro / (1 + u_ro))%R).
+{ assert (H : (0 < ufpx + Rabs (rx - x))%R).
+ { apply Rplus_lt_le_0_compat; [apply bpow_gt_0|apply Rabs_pos]. }
+ apply (Rle_trans _ (Rabs (rx - x) / (ufpx + Rabs (rx - x)))).
+ { unfold Rdiv; rewrite Rabs_mult; apply Rmult_le_compat_l; [apply Rabs_pos|].
+ now rewrite (Rabs_Rinv _ Nzx); apply Rinv_le. }
+ apply (Rmult_le_reg_r ((ufpx + Rabs (rx - x)) * (1 + u_ro))).
+ { apply Rmult_lt_0_compat; lra. }
+ field_simplify; [unfold Rdiv; rewrite Rinv_1, !Rmult_1_r| |]; lra. }
+revert H; unfold Rdiv; rewrite Rabs_mult, (Rabs_Rinv _ Nzx); intro H.
+apply (Rmult_le_reg_r (/ Rabs x)); [now apply Rinv_0_lt_compat, Rabs_pos_lt|].
+now apply (Rle_trans _ _ _ H); right; field; split; [apply Rabs_no_R0|lra].
+Qed.
+
+(** 1+#&epsilon;# property in rounding to nearest in FLX *)
+Theorem relative_error_N_FLX_ex :
+ forall x,
+ exists eps,
+ (Rabs eps <= /2 * bpow (-prec + 1))%R /\ round beta (FLX_exp prec) (Znearest choice) x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros x.
+apply relative_error_le_conversion...
+apply Rlt_le.
+apply Rmult_lt_0_compat.
+apply Rinv_0_lt_compat.
+now apply IZR_lt.
+apply bpow_gt_0.
+now apply relative_error_N_FLX.
+Qed.
+
+Theorem relative_error_N_FLX'_ex :
+ forall x,
+ exists eps,
+ (Rabs eps <= u_ro / (1 + u_ro))%R /\
+ round beta (FLX_exp prec) (Znearest choice) x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros x.
+apply relative_error_le_conversion...
+{ apply u_rod1pu_ro_pos. }
+now apply relative_error_N_FLX'.
+Qed.
+
+Lemma relative_error_N_round_ex_derive :
+ forall x rx,
+ (exists eps, (Rabs eps <= u_ro / (1 + u_ro))%R /\ rx = (x * (1 + eps))%R) ->
+ exists eps, (Rabs eps <= u_ro)%R /\ x = (rx * (1 + eps))%R.
+Proof.
+intros x rx (d, (Bd, Hd)).
+assert (Pu_ro := u_ro_pos).
+assert (H := Rabs_le_inv _ _ Bd).
+assert (H' := u_rod1pu_ro_le_u_ro); assert (H'' := u_ro_lt_1).
+destruct (Req_dec rx 0) as [Zfx|Nzfx].
+{ exists 0%R; split; [now rewrite Rabs_R0|].
+ rewrite Rplus_0_r, Rmult_1_r, Zfx.
+ now rewrite Zfx in Hd; destruct (Rmult_integral _ _ (sym_eq Hd)); [|lra]. }
+destruct (Req_dec x 0) as [Zx|Nzx].
+{ now exfalso; revert Hd; rewrite Zx, Rmult_0_l. }
+set (d' := ((x - rx) / rx)%R).
+assert (Hd' : (Rabs d' <= u_ro)%R).
+{ unfold d'; rewrite Hd.
+ replace (_ / _)%R with (- d / (1 + d))%R; [|now field; split; lra].
+ unfold Rdiv; rewrite Rabs_mult, Rabs_Ropp.
+ rewrite (Rabs_pos_eq (/ _)); [|apply Rlt_le, Rinv_0_lt_compat; lra].
+ apply (Rmult_le_reg_r (1 + d)); [lra|].
+ rewrite Rmult_assoc, Rinv_l, Rmult_1_r; [|lra].
+ apply (Rle_trans _ _ _ Bd).
+ unfold Rdiv; apply Rmult_le_compat_l; [now apply u_ro_pos|].
+ apply (Rle_trans _ (1 - u_ro / (1 + u_ro))); [right; field|]; lra. }
+now exists d'; split; [|unfold d'; field].
+Qed.
+
+Theorem relative_error_N_FLX_round_ex :
+ forall x,
+ exists eps,
+ (Rabs eps <= u_ro)%R /\
+ x = (round beta (FLX_exp prec) (Znearest choice) x * (1 + eps))%R.
+Proof.
+intro x; apply relative_error_N_round_ex_derive, relative_error_N_FLX'_ex.
+Qed.
+
+Theorem relative_error_N_FLX_round :
+ forall x,
+ (Rabs (round beta (FLX_exp prec) (Znearest choice) x - x) <= /2 * bpow (-prec + 1) * Rabs(round beta (FLX_exp prec) (Znearest choice) x))%R.
+Proof.
+intro x.
+apply relative_error_le_conversion_round_inv, relative_error_N_FLX_round_ex.
+Qed.
+
+End Fprop_relative_FLX.
+
+Section Fprop_relative_FLT.
+
+Variable emin prec : Z.
+Variable Hp : Z.lt 0 prec.
+
+Lemma relative_error_FLT_aux :
+ forall k, (emin + prec - 1 < k)%Z -> (prec <= k - FLT_exp emin prec k)%Z.
+Proof.
+intros k Hk.
+unfold FLT_exp.
+generalize (Zmax_spec (k - prec) emin).
+omega.
+Qed.
+
+Variable rnd : R -> Z.
+Context { valid_rnd : Valid_rnd rnd }.
+
+Theorem relative_error_FLT :
+ forall x,
+ (bpow (emin + prec - 1) <= Rabs x)%R ->
+ (Rabs (round beta (FLT_exp emin prec) rnd x - x) < bpow (-prec + 1) * Rabs x)%R.
+Proof with auto with typeclass_instances.
+intros x Hx.
+apply relative_error with (emin + prec - 1)%Z...
+apply relative_error_FLT_aux.
+Qed.
+
+Theorem relative_error_FLT_F2R_emin :
+ forall m, let x := F2R (Float beta m emin) in
+ (x <> 0)%R ->
+ (Rabs (round beta (FLT_exp emin prec) rnd x - x) < bpow (-prec + 1) * Rabs x)%R.
+Proof with auto with typeclass_instances.
+intros m x Zx.
+destruct (Rlt_or_le (Rabs x) (bpow (emin + prec - 1))) as [Hx|Hx].
+rewrite round_generic...
+unfold Rminus.
+rewrite Rplus_opp_r, Rabs_R0.
+apply Rmult_lt_0_compat.
+apply bpow_gt_0.
+now apply Rabs_pos_lt.
+apply generic_format_FLT_FIX...
+apply Rlt_le.
+apply Rlt_le_trans with (1 := Hx).
+apply bpow_le.
+apply Zle_pred.
+apply generic_format_FIX.
+now exists (Float beta m emin).
+now apply relative_error_FLT.
+Qed.
+
+Theorem relative_error_FLT_F2R_emin_ex :
+ forall m, let x := F2R (Float beta m emin) in
+ exists eps,
+ (Rabs eps < bpow (-prec + 1))%R /\ round beta (FLT_exp emin prec) rnd x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros m x.
+apply relative_error_lt_conversion...
+apply bpow_gt_0.
+now apply relative_error_FLT_F2R_emin.
+Qed.
+
+(** 1+#&epsilon;# property in any rounding in FLT *)
+Theorem relative_error_FLT_ex :
+ forall x,
+ (bpow (emin + prec - 1) <= Rabs x)%R ->
+ exists eps,
+ (Rabs eps < bpow (-prec + 1))%R /\ round beta (FLT_exp emin prec) rnd x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros x Hx.
+apply relative_error_lt_conversion...
+apply bpow_gt_0.
+intros _; now apply relative_error_FLT.
+Qed.
+
+Variable choice : Z -> bool.
+
+Theorem relative_error_N_FLT :
+ forall x,
+ (bpow (emin + prec - 1) <= Rabs x)%R ->
+ (Rabs (round beta (FLT_exp emin prec) (Znearest choice) x - x) <= /2 * bpow (-prec + 1) * Rabs x)%R.
+Proof with auto with typeclass_instances.
+intros x Hx.
+apply relative_error_N with (emin + prec - 1)%Z...
+apply relative_error_FLT_aux.
+Qed.
+
+(** 1+#&epsilon;# property in rounding to nearest in FLT *)
+Theorem relative_error_N_FLT_ex :
+ forall x,
+ (bpow (emin + prec - 1) <= Rabs x)%R ->
+ exists eps,
+ (Rabs eps <= /2 * bpow (-prec + 1))%R /\ round beta (FLT_exp emin prec) (Znearest choice) x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros x Hx.
+apply relative_error_le_conversion...
+apply Rlt_le.
+apply Rmult_lt_0_compat.
+apply Rinv_0_lt_compat.
+now apply IZR_lt.
+apply bpow_gt_0.
+now apply relative_error_N_FLT.
+Qed.
+
+Theorem relative_error_N_FLT_round :
+ forall x,
+ (bpow (emin + prec - 1) <= Rabs x)%R ->
+ (Rabs (round beta (FLT_exp emin prec) (Znearest choice) x - x) <= /2 * bpow (-prec + 1) * Rabs (round beta (FLT_exp emin prec) (Znearest choice) x))%R.
+Proof with auto with typeclass_instances.
+intros x Hx.
+apply relative_error_N_round with (emin + prec - 1)%Z...
+apply relative_error_FLT_aux.
+Qed.
+
+Theorem relative_error_N_FLT_F2R_emin :
+ forall m, let x := F2R (Float beta m emin) in
+ (Rabs (round beta (FLT_exp emin prec) (Znearest choice) x - x) <= /2 * bpow (-prec + 1) * Rabs x)%R.
+Proof with auto with typeclass_instances.
+intros m x.
+destruct (Rlt_or_le (Rabs x) (bpow (emin + prec - 1))) as [Hx|Hx].
+rewrite round_generic...
+unfold Rminus.
+rewrite Rplus_opp_r, Rabs_R0.
+apply Rmult_le_pos.
+apply Rmult_le_pos.
+apply Rlt_le.
+apply (RinvN_pos 1).
+apply bpow_ge_0.
+apply Rabs_pos.
+apply generic_format_FLT_FIX...
+apply Rlt_le.
+apply Rlt_le_trans with (1 := Hx).
+apply bpow_le.
+apply Zle_pred.
+apply generic_format_FIX.
+now exists (Float beta m emin).
+now apply relative_error_N_FLT.
+Qed.
+
+Theorem relative_error_N_FLT_F2R_emin_ex :
+ forall m, let x := F2R (Float beta m emin) in
+ exists eps,
+ (Rabs eps <= /2 * bpow (-prec + 1))%R /\ round beta (FLT_exp emin prec) (Znearest choice) x = (x * (1 + eps))%R.
+Proof with auto with typeclass_instances.
+intros m x.
+apply relative_error_le_conversion...
+apply Rmult_le_pos.
+apply Rlt_le.
+apply (RinvN_pos 1).
+apply bpow_ge_0.
+now apply relative_error_N_FLT_F2R_emin.
+Qed.
+
+
+Theorem relative_error_N_FLT_round_F2R_emin :
+ forall m, let x := F2R (Float beta m emin) in
+ (Rabs (round beta (FLT_exp emin prec) (Znearest choice) x - x) <= /2 * bpow (-prec + 1) * Rabs (round beta (FLT_exp emin prec) (Znearest choice) x))%R.
+Proof with auto with typeclass_instances.
+intros m x.
+destruct (Rlt_or_le (Rabs x) (bpow (emin + prec - 1))) as [Hx|Hx].
+rewrite round_generic...
+unfold Rminus.
+rewrite Rplus_opp_r, Rabs_R0.
+apply Rmult_le_pos.
+apply Rmult_le_pos.
+apply Rlt_le.
+apply (RinvN_pos 1).
+apply bpow_ge_0.
+apply Rabs_pos.
+apply generic_format_FLT_FIX...
+apply Rlt_le.
+apply Rlt_le_trans with (1 := Hx).
+apply bpow_le.
+apply Zle_pred.
+apply generic_format_FIX.
+now exists (Float beta m emin).
+apply relative_error_N_round with (emin := (emin + prec - 1)%Z)...
+apply relative_error_FLT_aux.
+Qed.
+
+Lemma error_N_FLT_aux :
+ forall x,
+ (0 < x)%R ->
+ exists eps, exists eta,
+ (Rabs eps <= /2 * bpow (-prec + 1))%R /\
+ (Rabs eta <= /2 * bpow (emin))%R /\
+ (eps*eta=0)%R /\
+ round beta (FLT_exp emin prec) (Znearest choice) x = (x * (1 + eps) + eta)%R.
+Proof.
+intros x Hx2.
+case (Rle_or_lt (bpow (emin+prec)) x); intros Hx.
+(* *)
+destruct relative_error_N_ex with (FLT_exp emin prec) (emin+prec)%Z prec choice x
+ as (eps,(Heps1,Heps2)).
+now apply FLT_exp_valid.
+intros; unfold FLT_exp.
+rewrite Zmax_left; omega.
+rewrite Rabs_right;[assumption|apply Rle_ge; now left].
+exists eps; exists 0%R.
+split;[assumption|split].
+rewrite Rabs_R0; apply Rmult_le_pos.
+apply Rlt_le, pos_half_prf.
+apply bpow_ge_0.
+split;[apply Rmult_0_r|idtac].
+now rewrite Rplus_0_r.
+(* *)
+exists 0%R; exists (round beta (FLT_exp emin prec) (Znearest choice) x - x)%R.
+split.
+rewrite Rabs_R0; apply Rmult_le_pos.
+apply Rlt_le, pos_half_prf.
+apply bpow_ge_0.
+split.
+apply Rle_trans with (/2*ulp beta (FLT_exp emin prec) x)%R.
+apply error_le_half_ulp.
+now apply FLT_exp_valid.
+apply Rmult_le_compat_l.
+apply Rlt_le, pos_half_prf.
+rewrite ulp_neq_0.
+2: now apply Rgt_not_eq.
+apply bpow_le.
+unfold FLT_exp, cexp.
+rewrite Zmax_right.
+omega.
+destruct (mag beta x) as (e,He); simpl.
+assert (e-1 < emin+prec)%Z.
+apply (lt_bpow beta).
+apply Rle_lt_trans with (2:=Hx).
+rewrite <- (Rabs_pos_eq x) by now apply Rlt_le.
+now apply He, Rgt_not_eq.
+omega.
+split ; ring.
+Qed.
+
+Theorem relative_error_N_FLT'_ex :
+ forall x,
+ exists eps eta : R,
+ (Rabs eps <= u_ro prec / (1 + u_ro prec))%R /\
+ (Rabs eta <= /2 * bpow emin)%R /\
+ (eps * eta = 0)%R /\
+ round beta (FLT_exp emin prec) (Znearest choice) x
+ = (x * (1 + eps) + eta)%R.
+Proof.
+intro x.
+set (rx := round _ _ _ x).
+assert (Pb := u_rod1pu_ro_pos prec).
+destruct (Rle_or_lt (bpow (emin + prec - 1)) (Rabs x)) as [MX|Mx].
+{ destruct (relative_error_N_FLX'_ex prec Hp choice x) as (d, (Bd, Hd)).
+ exists d, 0%R; split; [exact Bd|]; split.
+ { rewrite Rabs_R0; apply Rmult_le_pos; [lra|apply bpow_ge_0]. }
+ rewrite Rplus_0_r, Rmult_0_r; split; [reflexivity|].
+ now rewrite <- Hd; apply round_FLT_FLX. }
+assert (H : (Rabs (rx - x) <= /2 * bpow emin)%R).
+{ refine (Rle_trans _ _ _ (error_le_half_ulp _ _ _ _) _);
+ [now apply FLT_exp_valid|].
+ rewrite ulp_FLT_small; [now right|now simpl|].
+ apply (Rlt_le_trans _ _ _ Mx), bpow_le; lia. }
+exists 0%R, (rx - x)%R; split; [now rewrite Rabs_R0|]; split; [exact H|].
+now rewrite Rmult_0_l, Rplus_0_r, Rmult_1_r; split; [|ring].
+Qed.
+
+Theorem relative_error_N_FLT'_ex_separate :
+ forall x,
+ exists x' : R,
+ round beta (FLT_exp emin prec) (Znearest choice) x'
+ = round beta (FLT_exp emin prec) (Znearest choice) x /\
+ (exists eta, Rabs eta <= /2 * bpow emin /\ x' = x + eta)%R /\
+ (exists eps, Rabs eps <= u_ro prec / (1 + u_ro prec) /\
+ round beta (FLT_exp emin prec) (Znearest choice) x'
+ = x' * (1 + eps))%R.
+Proof.
+intro x.
+set (rx := round _ _ _ x).
+destruct (relative_error_N_FLT'_ex x) as (d, (e, (Bd, (Be, (Hde0, Hde))))).
+destruct (Rlt_or_le (Rabs (d * x)) (Rabs e)) as [HdxLte|HeLedx].
+{ exists rx; split; [|split].
+ { apply round_generic; [now apply valid_rnd_N|].
+ now apply generic_format_round; [apply FLT_exp_valid|apply valid_rnd_N]. }
+ { exists e; split; [exact Be|].
+ unfold rx; rewrite Hde; destruct (Rmult_integral _ _ Hde0) as [Zd|Ze].
+ { now rewrite Zd, Rplus_0_r, Rmult_1_r. }
+ exfalso; revert HdxLte; rewrite Ze, Rabs_R0; apply Rle_not_lt, Rabs_pos. }
+ exists 0%R; split; [now rewrite Rabs_R0; apply u_rod1pu_ro_pos|].
+ rewrite Rplus_0_r, Rmult_1_r; apply round_generic; [now apply valid_rnd_N|].
+ now apply generic_format_round; [apply FLT_exp_valid|apply valid_rnd_N]. }
+exists x; split; [now simpl|split].
+{ exists 0%R; split;
+ [rewrite Rabs_R0; apply Rmult_le_pos; [lra|apply bpow_ge_0]|ring]. }
+exists d; rewrite Hde; destruct (Rmult_integral _ _ Hde0) as [Zd|Ze].
+{ split; [exact Bd|].
+ assert (Ze : e = 0%R); [|now rewrite Ze, Rplus_0_r].
+ apply Rabs_eq_R0, Rle_antisym; [|now apply Rabs_pos].
+ now revert HeLedx; rewrite Zd, Rmult_0_l, Rabs_R0. }
+now rewrite Ze, Rplus_0_r.
+Qed.
+
+End Fprop_relative_FLT.
+
+Theorem error_N_FLT :
+ forall (emin prec : Z), (0 < prec)%Z ->
+ forall (choice : Z -> bool),
+ forall (x : R),
+ exists eps eta : R,
+ (Rabs eps <= /2 * bpow (-prec + 1))%R /\
+ (Rabs eta <= /2 * bpow emin)%R /\
+ (eps * eta = 0)%R /\
+ (round beta (FLT_exp emin prec) (Znearest choice) x
+ = x * (1 + eps) + eta)%R.
+Proof.
+intros emin prec Pprec choice x.
+destruct (Rtotal_order x 0) as [Nx|[Zx|Px]].
+{ assert (Pmx : (0 < - x)%R).
+ { now rewrite <- Ropp_0; apply Ropp_lt_contravar. }
+ destruct (@error_N_FLT_aux emin prec Pprec
+ (fun t : Z => negb (choice (- (t + 1))%Z))
+ (- x)%R Pmx)
+ as (d,(e,(Hd,(He,(Hde,Hr))))).
+ exists d; exists (- e)%R; split; [exact Hd|split; [|split]].
+ { now rewrite Rabs_Ropp. }
+ { now rewrite Ropp_mult_distr_r_reverse, <- Ropp_0; apply f_equal. }
+ rewrite <- (Ropp_involutive x), round_N_opp.
+ now rewrite Ropp_mult_distr_l_reverse, <- Ropp_plus_distr; apply f_equal. }
+{ assert (Ph2 : (0 <= / 2)%R).
+ { apply (Rmult_le_reg_l 2 _ _ Rlt_0_2).
+ rewrite Rmult_0_r, Rinv_r; [exact Rle_0_1|].
+ apply Rgt_not_eq, Rlt_gt, Rlt_0_2. }
+ exists 0%R; exists 0%R; rewrite Zx; split; [|split; [|split]].
+ { now rewrite Rabs_R0; apply Rmult_le_pos; [|apply bpow_ge_0]. }
+ { now rewrite Rabs_R0; apply Rmult_le_pos; [|apply bpow_ge_0]. }
+ { now rewrite Rmult_0_l. }
+ now rewrite Rmult_0_l, Rplus_0_l, round_0; [|apply valid_rnd_N]. }
+now apply error_N_FLT_aux.
+Qed.
+
+End Fprop_relative.