From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.129.124]) by sourceware.org (Postfix) with ESMTPS id 8A7A13858410 for ; Tue, 1 Nov 2022 12:24:59 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 8A7A13858410 Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=redhat.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=redhat.com DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1667305499; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: in-reply-to:in-reply-to:references:references; bh=S1o3q3WM/b08NuSoBAYRSfc9J1rFj82Km3YQd4G65gE=; b=FqRZCeVHdEqqvgmq/KVoYY9me+t9/bFL15xjHvNgvRMiUwj28No+K7+Hrf5mm22yun4FZ6 JJ13Mj8Z2Ehj6c2grbBizUhE9KFRsdNCggsMOijsO79WGoXQXiro4fFRYx5EYOxGF5kj/U 3xaAVqy6Fe0pV0Fru9W9XeE6zMman3Q= Received: from mail-ej1-f69.google.com (mail-ej1-f69.google.com [209.85.218.69]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.3, cipher=TLS_AES_128_GCM_SHA256) id us-mta-303-SfVPUm1zMau96ApCeH-Uhg-1; Tue, 01 Nov 2022 08:24:57 -0400 X-MC-Unique: SfVPUm1zMau96ApCeH-Uhg-1 Received: by mail-ej1-f69.google.com with SMTP id qk31-20020a1709077f9f00b00791a3e02c80so7849708ejc.21 for ; Tue, 01 Nov 2022 05:24:57 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=cc:to:subject:message-id:date:from:in-reply-to:references :mime-version:x-gm-message-state:from:to:cc:subject:date:message-id :reply-to; bh=S1o3q3WM/b08NuSoBAYRSfc9J1rFj82Km3YQd4G65gE=; b=yEw/zqC/DkncQf+oX6cvFe73M9yvs158UqhooqoqiCFL7W+40755z9S4bj3M1ZT2Y+ FmI79y82wOEFYuV9l9dzaL69r9ltHUyGD17pRPnDWONV1//up0WOgBFPf/vgNIKTtMjt NgmfGGcnkSNiFNUs3j0sNjeE2Fz2XoRydYLBMnqSR7tcstF52hCzq8HvXNQJjKTX7eSV iU9QcdHj0neDvD9HWNEd0e7Pw0pqxJVRq6NjnsflhXbyUdxbQ7ob6tcTtWWr+o4giaCG a6dToZO5lE7keSctEfvzZMMf81tfIMxh0ceWrhxQY2rHPNur3E8nK1hLu861g4eXwySB p0Uw== X-Gm-Message-State: ACrzQf1ZLqskYN/RXju5K7yaAUpaRwC4lFnnsgIkgdP/cLa2UvrUk+5d A/mRe3PdyhdL+l7PdQ+MzrfkrV7hUrnmbtmoZ3jK8CyMNehDKPUhYc8X0eztWXskdLwfaUUju77 qvEc0vAaNaNTrUGL3+AU0lfiMGvHl1Xg= X-Received: by 2002:a17:906:cc58:b0:7a9:6b4e:79e4 with SMTP id mm24-20020a170906cc5800b007a96b4e79e4mr17487793ejb.57.1667305496224; Tue, 01 Nov 2022 05:24:56 -0700 (PDT) X-Google-Smtp-Source: AMsMyM40zjFPc/3U0Ekgd/4zp6cC6TYGOFSOQC9WpCqpMk4dwl6ZUtU1UTGM7UFRQVsYFPxWGMVFgIasFxEcU2pJFrg= X-Received: by 2002:a17:906:cc58:b0:7a9:6b4e:79e4 with SMTP id mm24-20020a170906cc5800b007a96b4e79e4mr17487777ejb.57.1667305495983; Tue, 01 Nov 2022 05:24:55 -0700 (PDT) MIME-Version: 1.0 References: <4ba36955-3f9a-00c5-c406-45821ec2a4db@idea> In-Reply-To: From: Jonathan Wakely Date: Tue, 1 Nov 2022 12:24:45 +0000 Message-ID: Subject: Re: [PATCH] libstdc++: Shortest denormal hex std::to_chars To: Jakub Jelinek Cc: Patrick Palka , gcc-patches@gcc.gnu.org, libstdc++@gcc.gnu.org X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com Content-Type: text/plain; charset="UTF-8" X-Spam-Status: No, score=-7.4 required=5.0 tests=BAYES_00,DKIMWL_WL_HIGH,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,KAM_SHORT,RCVD_IN_DNSWL_NONE,RCVD_IN_MSPIKE_H2,SPF_HELO_NONE,SPF_NONE,TXREP autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org List-Id: On Tue, 1 Nov 2022 at 12:18, Jakub Jelinek wrote: > > On Fri, Oct 28, 2022 at 12:52:44PM -0400, Patrick Palka wrote: > > > The following patch on top of > > > https://gcc.gnu.org/pipermail/libstdc++/2022-October/054849.html > > > adds std::{,b}float16_t support for std::to_chars. > > > When precision is specified (or for std::bfloat16_t for hex mode even if not), > > > I believe we can just use the std::to_chars float (when float is mode > > > compatible with std::float32_t) overloads, both formats are proper subsets > > > of std::float32_t. > > > Unfortunately when precision is not specified and we are supposed to emit > > > shortest string, the std::{,b}float16_t strings are usually much shorter. > > > E.g. 1.e7p-14f16 shortest fixed representation is > > > 0.0001161 and shortest scientific representation is > > > 1.161e-04 while 1.e7p-14f32 (same number promoted to std::float32_t) > > > 0.00011610985 and > > > 1.1610985e-04. > > > Similarly for 1.38p-112bf16, > > > 0.000000000000000000000000000000000235 > > > 2.35e-34 vs. 1.38p-112f32 > > > 0.00000000000000000000000000000000023472271 > > > 2.3472271e-34 > > > For std::float16_t there are differences even in the shortest hex, say: > > > 0.01p-14 vs. 1p-22 > > > but only for denormal std::float16_t values (where all std::float16_t > > > denormals converted to std::float32_t are normal), __FLT16_MIN__ and > > > everything larger in absolute value than that is the same. Unless > > > that is a bug and we should try to discover shorter representations > > > even for denormals... > > > > IIRC for hex formatting of denormals I opted to be consistent with how > > glibc printf formats them, instead of outputting the truly shortest > > form. > > > > I wouldn't be against using the float32 overloads even for shortest hex > > formatting of float16. The output is shorter but equivalent so it > > shouldn't cause any problems. > > The following patch changes the behavior of the shortest hex denormals, > such that they are printed like normals (so for has_implicit_leading_bit > with 1p-149 instead of 0.000002p-126 etc., otherwise (Intel extended) > with the leading digit before dot being [89abcdef]). I think for all the > supported format it is never longer, it can be equal length e.g. for > 0.fffffep-126 vs. 1.fffffcp-127 but fortunately no largest subnormal > in any format has the unbiased exponent like -9, -99, -999, -9999 because > then it would be longer and often it is shorter, sometimes much shorter. > > For the cases with precision it keeps the handling as is. > > While for !has_implicit_leading_bit we for normals or with this patch > even denormals have really shortest representation, for other formats > we sometimes do not, but this patch doesn't deal with that (we > always use 1.NNN while we could use 1.NNN up to f.NNN and by that shortening > by the last hexit if the last hexit doesn't have least significant bit set > and unbiased exponent is not -9, -99, -999 or -9999. > > Tested on x86_64-linux (on top of the 3 to/from_chars {,b}float16_t > patches). This looks good to me. Please give Patrick a chance to comment, but it's approved for trunk unless he objects. Thanks! > > 2022-11-01 Jakub Jelinek > > * src/c++17/floating_to_chars.cc (__floating_to_chars_hex): Drop const > from unbiased_exponent. Canonicalize denormals such that they have > the leading bit set by shifting effective mantissa up and decreasing > unbiased_exponent. > (__floating_to_chars_shortest): Don't instantiate > __floating_to_chars_hex for float16_t either and use float instead. > * testsuite/20_util/to_chars/float.cc (float_to_chars_test_cases): > Adjust testcases for shortest hex denormals. > * testsuite/20_util/to_chars/double.cc (double_to_chars_test_cases): > Likewise. > > --- libstdc++-v3/src/c++17/floating_to_chars.cc.jj 2022-10-31 22:20:35.881121902 +0100 > +++ libstdc++-v3/src/c++17/floating_to_chars.cc 2022-11-01 12:16:14.352652455 +0100 > @@ -844,9 +844,9 @@ template > const bool is_normal_number = (biased_exponent != 0); > > // Calculate the unbiased exponent. > - const int32_t unbiased_exponent = (is_normal_number > - ? biased_exponent - exponent_bias > - : 1 - exponent_bias); > + int32_t unbiased_exponent = (is_normal_number > + ? biased_exponent - exponent_bias > + : 1 - exponent_bias); > > // Shift the mantissa so that its bitwidth is a multiple of 4. > constexpr unsigned rounded_mantissa_bits = (mantissa_bits + 3) / 4 * 4; > @@ -863,6 +863,16 @@ template > __glibcxx_assert(effective_mantissa & (mantissa_t{1} << (mantissa_bits > - 1u))); > } > + else if (!precision.has_value() && effective_mantissa) > + { > + // 1.8p-23 is shorter than 0.00cp-14, so if precision is > + // omitted, try to canonicalize denormals such that they > + // have the leading bit set. > + int width = __bit_width(effective_mantissa); > + int shift = rounded_mantissa_bits - width + has_implicit_leading_bit; > + unbiased_exponent -= shift; > + effective_mantissa <<= shift; > + } > > // Compute the shortest precision needed to print this value exactly, > // disregarding trailing zeros. > @@ -1061,7 +1071,10 @@ template > // std::bfloat16_t has the same exponent range as std::float32_t > // and so we can avoid instantiation of __floating_to_chars_hex > // for bfloat16_t. Shortest hex will be the same as for float. > - if constexpr (is_same_v) > + // When we print shortest form even for denormals, we can do it > + // for std::float16_t as well. > + if constexpr (is_same_v > + || is_same_v) > return __floating_to_chars_hex(first, last, value.x, nullopt); > else > return __floating_to_chars_hex(first, last, value, nullopt); > --- libstdc++-v3/testsuite/20_util/to_chars/float.cc.jj 2022-01-11 22:31:41.605755528 +0100 > +++ libstdc++-v3/testsuite/20_util/to_chars/float.cc 2022-11-01 12:34:21.370882443 +0100 > @@ -521,8 +521,8 @@ inline constexpr float_to_chars_testcase > > // Test hexfloat corner cases. > {0x1.728p+0f, chars_format::hex, "1.728p+0"}, // instead of "2.e5p-1" > - {0x0.000002p-126f, chars_format::hex, "0.000002p-126"}, // instead of "1p-149", min subnormal > - {0x0.fffffep-126f, chars_format::hex, "0.fffffep-126"}, // max subnormal > + {0x0.000002p-126f, chars_format::hex, "1p-149"}, // min subnormal > + {0x0.fffffep-126f, chars_format::hex, "1.fffffcp-127"}, // max subnormal > {0x1p-126f, chars_format::hex, "1p-126"}, // min normal > {0x1.fffffep+127f, chars_format::hex, "1.fffffep+127"}, // max normal > > --- libstdc++-v3/testsuite/20_util/to_chars/double.cc.jj 2022-01-11 22:31:41.604755542 +0100 > +++ libstdc++-v3/testsuite/20_util/to_chars/double.cc 2022-11-01 12:42:39.753112522 +0100 > @@ -2821,8 +2821,8 @@ inline constexpr double_to_chars_testcas > > // Test hexfloat corner cases. > {0x1.728p+0, chars_format::hex, "1.728p+0"}, // instead of "2.e5p-1" > - {0x0.0000000000001p-1022, chars_format::hex, "0.0000000000001p-1022"}, // instead of "1p-1074", min subnormal > - {0x0.fffffffffffffp-1022, chars_format::hex, "0.fffffffffffffp-1022"}, // max subnormal > + {0x0.0000000000001p-1022, chars_format::hex, "1p-1074"}, // min subnormal > + {0x0.fffffffffffffp-1022, chars_format::hex, "1.ffffffffffffep-1023"}, // max subnormal > {0x1p-1022, chars_format::hex, "1p-1022"}, // min normal > {0x1.fffffffffffffp+1023, chars_format::hex, "1.fffffffffffffp+1023"}, // max normal > > > Jakub >