aboutsummaryrefslogtreecommitdiffstats
path: root/lib/hipe/rtl/hipe_rtl_binary_match.erl
diff options
context:
space:
mode:
authorKostis Sagonas <[email protected]>2015-10-02 14:50:29 +0200
committerKostis Sagonas <[email protected]>2015-10-02 15:08:00 +0200
commit5aea81c495f06d58d72b5d3dba86ed9326b1eb14 (patch)
treea453e8535236a7344287cdcd89b30a02008bb2e0 /lib/hipe/rtl/hipe_rtl_binary_match.erl
parentc881813b0643138e9b1c1b1a573645460f14404e (diff)
downloadotp-5aea81c495f06d58d72b5d3dba86ed9326b1eb14.tar.gz
otp-5aea81c495f06d58d72b5d3dba86ed9326b1eb14.tar.bz2
otp-5aea81c495f06d58d72b5d3dba86ed9326b1eb14.zip
Fix matching with huge binaries
In certain cases of matching with very big binaries, the HiPE compiler generated code that would fail the match, even in cases that the matching was successful. The problem was more quite noticeable on 32-bit platforms where certain integer quantities would be represented as bignums. Brief summary of changes: * gen_rtl({bs_skip_bits, ...}, ...) could not handle too large constants. Previously the constants were truncated to word size. * hipe_rtl_binary_match:make_size/3 erroneously assumed that the output of first_part/3 would not overflow when multiplied by 8, which is no longer true. To maintain full performance, the overflow test is only performed when BitsVar was a bignum. Thus, the fast path is identical to before. * hipe_rtl_binary_match:set_high/2 was assuming that only bits below bit 27 were ever set in arguments to bs_skip_bits, which is not only false when the arguments are bignums, but also on 64-bit platforms. The commit includes a test taken from the bs_match_bin_SUITE. Most of the credit for finding these HiPE compiler errors and for creating appropriate fixes for them should go to Magnus Lång.
Diffstat (limited to 'lib/hipe/rtl/hipe_rtl_binary_match.erl')
-rw-r--r--lib/hipe/rtl/hipe_rtl_binary_match.erl81
1 files changed, 59 insertions, 22 deletions
diff --git a/lib/hipe/rtl/hipe_rtl_binary_match.erl b/lib/hipe/rtl/hipe_rtl_binary_match.erl
index 37263a951a..51213b71d1 100644
--- a/lib/hipe/rtl/hipe_rtl_binary_match.erl
+++ b/lib/hipe/rtl/hipe_rtl_binary_match.erl
@@ -233,14 +233,26 @@ gen_rtl({bs_skip_bits_all, Unit, _Flags}, Dst, [Ms],
skip_bits_all(Unit, Ms, TrueLblName, FalseLblName);
%% ----- bs_skip_bits -----
gen_rtl({bs_skip_bits, Bits}, Dst, [Ms|Args], TrueLblName, FalseLblName) ->
+ MaxValue = (1 bsl (hipe_rtl_arch:word_size() * ?BYTE_SIZE)),
opt_update_ms(Dst, Ms) ++
- case Args of
- [] ->
- skip_bits2(Ms, hipe_rtl:mk_imm(Bits), TrueLblName, FalseLblName);
- [Arg] ->
- {SizeCode, SizeReg} = make_size(Bits, Arg, FalseLblName),
- InCode = skip_bits2(Ms, SizeReg, TrueLblName, FalseLblName),
- SizeCode ++ InCode
+ case Bits < MaxValue of
+ true ->
+ case Args of
+ [] ->
+ skip_bits2(Ms, hipe_rtl:mk_imm(Bits), TrueLblName, FalseLblName);
+ [Arg] ->
+ {SizeCode, SizeReg} = make_size(Bits, Arg, FalseLblName),
+ InCode = skip_bits2(Ms, SizeReg, TrueLblName, FalseLblName),
+ SizeCode ++ InCode
+ end;
+ false -> % handle overflow case
+ case Args of
+ [] ->
+ [hipe_rtl:mk_goto(FalseLblName)];
+ [Arg] ->
+ [hipe_rtl:mk_branch(Arg, 'eq', hipe_tagscheme:mk_fixnum(0),
+ TrueLblName, FalseLblName, 0.5)]
+ end
end;
%% ----- bs_restore -----
gen_rtl({bs_restore, Slot}, [NewMs], [Ms], TrueLblName, _FalseLblName) ->
@@ -1089,23 +1101,47 @@ create_gcsafe_regs(0) ->
[].
first_part(Var, Register, FalseLblName) ->
- [SuccessLbl1, SuccessLbl2] = create_lbls(2),
- [hipe_tagscheme:test_fixnum(Var, hipe_rtl:label_name(SuccessLbl1),
- FalseLblName, 0.99),
- SuccessLbl1,
- hipe_tagscheme:fixnum_ge(Var, hipe_rtl:mk_imm(hipe_tagscheme:mk_fixnum(0)),
- hipe_rtl:label_name(SuccessLbl2), FalseLblName, 0.99),
- SuccessLbl2,
- hipe_tagscheme:untag_fixnum(Register, Var)].
+ [EndLbl] = create_lbls(1),
+ EndName = hipe_rtl:label_name(EndLbl),
+ first_part(Var, Register, FalseLblName, EndName, EndName, [EndLbl]).
+
+first_part(Var, Register, FalseLblName, TrueLblName, BigLblName, Tail) ->
+ [FixnumLbl, NotFixnumLbl, BignumLbl, SuccessLbl] = create_lbls(4),
+ [hipe_tagscheme:test_fixnum(Var, hipe_rtl:label_name(FixnumLbl),
+ hipe_rtl:label_name(NotFixnumLbl), 0.99),
+ FixnumLbl,
+ hipe_tagscheme:fixnum_ge(Var, hipe_rtl:mk_imm(hipe_tagscheme:mk_fixnum(0)),
+ hipe_rtl:label_name(SuccessLbl), FalseLblName,
+ 0.99),
+ SuccessLbl,
+ hipe_tagscheme:untag_fixnum(Register, Var),
+ hipe_rtl:mk_goto(TrueLblName),
+ NotFixnumLbl,
+ %% Since binaries are not allowed to be larger than 2^wordsize bits
+ %% and since bignum digits are words, we know that a bignum with an
+ %% arity larger than one can't match.
+ hipe_tagscheme:test_pos_bignum_arity(Var, 1, hipe_rtl:label_name(BignumLbl),
+ FalseLblName, 0.99),
+ BignumLbl,
+ hipe_tagscheme:unsafe_get_one_word_pos_bignum(Register, Var),
+ hipe_rtl:mk_goto(BigLblName) | Tail].
make_size(1, BitsVar, FalseLblName) ->
[DstReg] = create_regs(1),
{first_part(BitsVar, DstReg, FalseLblName), DstReg};
make_size(?BYTE_SIZE, BitsVar, FalseLblName) ->
[DstReg] = create_regs(1),
- Code =
- first_part(BitsVar, DstReg, FalseLblName) ++
- [hipe_rtl:mk_alu(DstReg, DstReg, sll, hipe_rtl:mk_imm(?BYTE_SHIFT))],
+ [FixnumLbl, BignumLbl] = create_lbls(2),
+ WordBits = hipe_rtl_arch:word_size() * ?BYTE_SIZE,
+ FixnumLblName = hipe_rtl:label_name(FixnumLbl),
+ Tail = [BignumLbl,
+ hipe_rtl:mk_branch(DstReg, 'ltu',
+ hipe_rtl:mk_imm(1 bsl (WordBits - ?BYTE_SHIFT)),
+ FixnumLblName, FalseLblName, 0.99),
+ FixnumLbl,
+ hipe_rtl:mk_alu(DstReg, DstReg, sll, hipe_rtl:mk_imm(?BYTE_SHIFT))],
+ Code = first_part(BitsVar, DstReg, FalseLblName, FixnumLblName,
+ hipe_rtl:label_name(BignumLbl), Tail),
{Code, DstReg};
make_size(UnitImm, BitsVar, FalseLblName) ->
[DstReg] = create_regs(1),
@@ -1154,12 +1190,13 @@ floorlog2(X) ->
round(math:log(X)/math:log(2)-0.5).
set_high(X) ->
- set_high(X, 0).
+ WordBits = hipe_rtl_arch:word_size() * ?BYTE_SIZE,
+ set_high(min(X, WordBits), WordBits, 0).
-set_high(0, Y) ->
+set_high(0, _, Y) ->
Y;
-set_high(X, Y) ->
- set_high(X-1, Y+(1 bsl (27-X))).
+set_high(X, WordBits, Y) ->
+ set_high(X-1, WordBits, Y+(1 bsl (WordBits-X))).
is_illegal_const(Const) ->
Const >= 1 bsl (hipe_rtl_arch:word_size() * ?BYTE_SIZE) orelse Const < 0.