From nobody Sun Apr 12 04:37:22 2026 Delivered-To: importer@patchew.org Authentication-Results: mx.zohomail.com; dkim=pass; spf=pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) smtp.mailfrom=qemu-devel-bounces+importer=patchew.org@nongnu.org; dmarc=pass(p=none dis=none) header.from=linaro.org ARC-Seal: i=1; a=rsa-sha256; t=1771911963; cv=none; d=zohomail.com; s=zohoarc; b=lh7lLlX7Jn/XYYcz5COJ01N0DLhwajFbMZaSSuRW5Xpmp8PKXIFXEAepp3detgFGvQWVFTW1irmis6cBWWt0mlEiu0OFeRfrt7Z9/OzE/MN3DIUdsc2FpCm24yrnCJEDCdeZvrm4X0Pb4tHyD/CBEH+bTbvTuSSZ9gylIFj1NZ0= ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=zohomail.com; s=zohoarc; t=1771911963; h=Content-Transfer-Encoding:Cc:Cc:Date:Date:From:From:In-Reply-To:List-Subscribe:List-Post:List-Id:List-Archive:List-Help:List-Unsubscribe:MIME-Version:Message-ID:References:Sender:Subject:Subject:To:To:Message-Id:Reply-To; bh=Vey5Rh3waHmkn33K/lS5TwQ0lXIHYR1iFIEDzRn1i+k=; b=RaBPWjA3y1bWr60oUIVdFxM0L9oEr3NNEZWUjrUVz629m9m9Ff4zip5sOd5aDOQbwIhwGFoaQDUc7WHrDNIAYa2ONsgCul5TguMDuqJDUJjVvjfDaX8OLzttbxPjc1g9DyXr6UOljl7mBJuR9b8t8lg7Wt8yPyoOwZk2PA771SI= ARC-Authentication-Results: i=1; mx.zohomail.com; dkim=pass; spf=pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) smtp.mailfrom=qemu-devel-bounces+importer=patchew.org@nongnu.org; dmarc=pass header.from= (p=none dis=none) Return-Path: Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) by mx.zohomail.com with SMTPS id 1771911963788472.9613963860262; Mon, 23 Feb 2026 21:46:03 -0800 (PST) Received: from localhost ([::1] helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1vulF3-0004Z8-O7; Tue, 24 Feb 2026 00:45:53 -0500 Received: from eggs.gnu.org ([2001:470:142:3::10]) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1vulEK-000415-OE for qemu-devel@nongnu.org; Tue, 24 Feb 2026 00:45:17 -0500 Received: from mail-dy1-x1335.google.com ([2607:f8b0:4864:20::1335]) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_128_GCM_SHA256:128) (Exim 4.90_1) (envelope-from ) id 1vulEG-0002Nb-UA for qemu-devel@nongnu.org; Tue, 24 Feb 2026 00:45:07 -0500 Received: by mail-dy1-x1335.google.com with SMTP id 5a478bee46e88-2ba88e53b46so2793618eec.1 for ; Mon, 23 Feb 2026 21:45:03 -0800 (PST) Received: from stoup.. ([172.56.177.223]) by smtp.gmail.com with ESMTPSA id 5a478bee46e88-2bd7daa37c7sm5958093eec.12.2026.02.23.21.44.59 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Mon, 23 Feb 2026 21:45:02 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; t=1771911902; x=1772516702; darn=nongnu.org; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:from:to:cc:subject:date :message-id:reply-to; bh=Vey5Rh3waHmkn33K/lS5TwQ0lXIHYR1iFIEDzRn1i+k=; b=B5+SrTUKjU3bb/J2wxftBTUSd+1h3fNTDR91QOVcXAAnFQVo+IAfCI7DDnGhyHm2ib A8XPIZBD+0kzmbrltFf9yFNtOwlFSusPKm05CUItSWjZ/m1wv+MpantCQFLNapz355Ia 09P7hx6CMT8EES0ObCc01nLiStW+o+7XcRjKLGkzhicmxpJ1+F+gUy9e/mAED7fiK+9N qw/njoC0iP0aMaKTAo93kS1/E6yEyOmt8S6PR42+cvCWc+A0EdZtQs8I8e68+0jRRR0Y 3k4/nS3/8RAjg4zWLiDO/dm+h4dMruI3B27Ivf6Ymcq9FYMUbaO7lNy3UE3RDyacmc3a DQNQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1771911902; x=1772516702; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:x-gm-gg:x-gm-message-state:from :to:cc:subject:date:message-id:reply-to; bh=Vey5Rh3waHmkn33K/lS5TwQ0lXIHYR1iFIEDzRn1i+k=; b=UmRO+uEvihtbGrDomqRxKbDCDMWd2qDbykAFYTkLu9EFAf20l6ZhWr3SMyk7HFMAZY Np2sYCPS+SWRmG2Wuy8w0v0IZkTqUoio0PqaklYd8Z2SwXPwwIPO4Fr1Suul19vPoIzX A76600mlAtwvfE7usVCWLxKO932orYe913Z9vv8mO0I/vUVTTDaBC3d2iGoIT/V7A/RJ RIBwQDc7vMuUmTMlUmWTDaUvIzJYtfFQ6uOzWNonUQhruwHrZFcEjtxlGMz/QgyYO9kT x+o373bMmGrpaD9WzLOtSte6Cu7gkFhHefFM4Sf08+hcGWgqPBmh0SuXOMuz3o1aTEas B32g== X-Gm-Message-State: AOJu0YzX5tsRa+18Jeb7Jc7ELXSwkv4WEc4Bd4rsNobODXTzpe/I/kxl C/4ohoQUx7JS0OnNmBTyopo0T3VCdDkQz6Ega6qQRbyp+maHpbVCGaXwFOPErEvrEGenE1pjg/m 8hMFvOl0= X-Gm-Gg: ATEYQzyfmfXE4xo5RA0xdt80Luq1clIhjGwiqvpvK9t+YXD1kckuxYpV25dzQ8mj9In WcF8CrIbDZGfes+5Mk8ZlYBHHER5PmqPzntfx7FK9ZeAo/4kTtKG/Z4plN/YnR1CS3ai1hukVlh PEen2Fhy2gBZJhts2+r3ECFBARuSrwXooiy8QuGzx2K2YJiOeNGlsoag6u6V8pP4wTVOonZA8W9 cTH8b2ZAP2SS9556x96SSipoSuFiujMCiiH4SjXPvu+0wuCtgzA2OLQluJJDfWq7yZsaYdJ1eqy vOmnqA45/ItcwuwLlhMk9ogXEoPVxC7i2fbIUBwZgDy3ykMax3irsszfYvKIiSYVbRo20B/B7P2 Dop2xvANlmdkCxgE2GA6wo7MjvWx2G9jdofDj/08VOrJcAV02Zvf6jdPdjpInmmbQJRtTup56pB 2SvRYiW3SFTZlzQunwxiwz6lRip13PgDsvjfJ8h1jbmtTVDiHwHdEz X-Received: by 2002:a05:7301:6505:b0:2ba:7c92:82fb with SMTP id 5a478bee46e88-2bd60769b84mr6846251eec.5.1771911902343; Mon, 23 Feb 2026 21:45:02 -0800 (PST) From: Richard Henderson To: qemu-devel@nongnu.org Cc: Max Chou , Chao Liu Subject: [PULL 8/9] fpu: Add conversion routines for OCP FP8 E4M3 Date: Tue, 24 Feb 2026 16:44:16 +1100 Message-ID: <20260224054417.32430-9-richard.henderson@linaro.org> X-Mailer: git-send-email 2.43.0 In-Reply-To: <20260224054417.32430-1-richard.henderson@linaro.org> References: <20260224054417.32430-1-richard.henderson@linaro.org> MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Received-SPF: pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) client-ip=209.51.188.17; envelope-from=qemu-devel-bounces+importer=patchew.org@nongnu.org; helo=lists.gnu.org; Received-SPF: pass client-ip=2607:f8b0:4864:20::1335; envelope-from=richard.henderson@linaro.org; helo=mail-dy1-x1335.google.com X-Spam_score_int: -20 X-Spam_score: -2.1 X-Spam_bar: -- X-Spam_report: (-2.1 / 5.0 requ) BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_NONE=-0.0001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001 autolearn=ham autolearn_force=no X-Spam_action: no action X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: qemu development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: qemu-devel-bounces+importer=patchew.org@nongnu.org Sender: qemu-devel-bounces+importer=patchew.org@nongnu.org X-ZohoMail-DKIM: pass (identity @linaro.org) X-ZM-MESSAGEID: 1771911965912158500 Content-Type: text/plain; charset="utf-8" From: Max Chou Reviewed-by: Chao Liu Signed-off-by: Max Chou [rth: Split out of a larger patch; adjust overflow detection.] Signed-off-by: Richard Henderson --- include/fpu/softfloat-types.h | 1 + include/fpu/softfloat.h | 4 +++ fpu/softfloat.c | 62 +++++++++++++++++++++++++++++++++++ fpu/softfloat-parts.c.inc | 45 +++++++++++++++++++++++-- 4 files changed, 109 insertions(+), 3 deletions(-) diff --git a/include/fpu/softfloat-types.h b/include/fpu/softfloat-types.h index be7e2de6e3..9c84a101e5 100644 --- a/include/fpu/softfloat-types.h +++ b/include/fpu/softfloat-types.h @@ -122,6 +122,7 @@ typedef uint16_t bfloat16; /* * Open Compute Project (OCP) Microscaling Formats */ +typedef uint8_t float8_e4m3; typedef uint8_t float8_e5m2; =20 /* diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h index 4385462992..31d3f76d3f 100644 --- a/include/fpu/softfloat.h +++ b/include/fpu/softfloat.h @@ -193,6 +193,10 @@ float128 uint128_to_float128(Int128, float_status *sta= tus); | OCP FP8 conversion routines. *-------------------------------------------------------------------------= ---*/ =20 +bfloat16 float8_e4m3_to_bfloat16(float8_e4m3, float_status *status); +float8_e4m3 bfloat16_to_float8_e4m3(bfloat16, bool sat, float_status *stat= us); +float8_e4m3 float32_to_float8_e4m3(float32, bool sat, float_status *status= ); + bfloat16 float8_e5m2_to_bfloat16(float8_e5m2, float_status *status); float8_e5m2 bfloat16_to_float8_e5m2(bfloat16, bool sat, float_status *stat= us); float8_e5m2 float32_to_float8_e5m2(float32, bool sat, float_status *status= ); diff --git a/fpu/softfloat.c b/fpu/softfloat.c index 0dc769283d..6e21882ab2 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -528,6 +528,8 @@ typedef enum __attribute__((__packed__)) { float_expmax_ieee, /* exp=3D=3Dmax is a normal number; no infinity or nan representation.= */ float_expmax_normal, + /* exp=3D=3Dmax, frac=3D=3Dmax ? nan : normal; no infinity representat= ion. */ + float_expmax_e4m3, } FloatFmtExpMaxKind; =20 /* @@ -572,6 +574,14 @@ typedef struct { .frac_shift =3D (-F - 1) & 63, \ .round_mask =3D (1ull << ((-F - 1) & 63)) - 1 =20 +static const FloatFmt float8_e4m3_params =3D { + FLOAT_PARAMS(4, 3), + .exp_max_kind =3D float_expmax_e4m3 +}; + +/* 110 << frac_shift, with the implicit bit set */ +#define E4M3_NORMAL_FRAC_MAX 0xe000000000000000ull + static const FloatFmt float8_e5m2_params =3D { FLOAT_PARAMS(5, 2) }; @@ -631,6 +641,11 @@ static void unpack_raw64(FloatParts64 *r, const FloatF= mt *fmt, uint64_t raw) }; } =20 +static void QEMU_FLATTEN float8_e4m3_unpack_raw(FloatParts64 *p, float8_e4= m3 f) +{ + unpack_raw64(p, &float8_e4m3_params, f); +} + static void QEMU_FLATTEN float8_e5m2_unpack_raw(FloatParts64 *p, float8_e5= m2 f) { unpack_raw64(p, &float8_e5m2_params, f); @@ -693,6 +708,11 @@ static uint64_t pack_raw64(const FloatParts64 *p, cons= t FloatFmt *fmt) return ret; } =20 +static float8_e4m3 QEMU_FLATTEN float8_e4m3_pack_raw(const FloatParts64 *p) +{ + return pack_raw64(p, &float8_e4m3_params); +} + static float8_e5m2 QEMU_FLATTEN float8_e5m2_pack_raw(const FloatParts64 *p) { return pack_raw64(p, &float8_e5m2_params); @@ -1689,6 +1709,13 @@ static const uint16_t rsqrt_tab[128] =3D { * Pack/unpack routines with a specific FloatFmt. */ =20 +static void float8_e4m3_unpack_canonical(FloatParts64 *p, float8_e4m3 f, + float_status *s) +{ + float8_e4m3_unpack_raw(p, f); + parts_canonicalize(p, s, &float8_e4m3_params); +} + static void float8_e5m2_unpack_canonical(FloatParts64 *p, float8_e5m2 f, float_status *s) { @@ -1716,6 +1743,14 @@ static void bfloat16_unpack_canonical(FloatParts64 *= p, bfloat16 f, parts_canonicalize(p, s, &bfloat16_params); } =20 +static float8_e4m3 float8_e4m3_round_pack_canonical(FloatParts64 *p, + float_status *s, + bool saturate) +{ + parts_uncanon(p, s, &float8_e4m3_params, saturate); + return float8_e4m3_pack_raw(p); +} + static float8_e5m2 float8_e5m2_round_pack_canonical(FloatParts64 *p, float_status *s, bool saturate) @@ -2894,6 +2929,15 @@ static void parts_float_to_float_widen(FloatParts128= *a, FloatParts64 *b, } } =20 +bfloat16 float8_e4m3_to_bfloat16(float8_e4m3 a, float_status *s) +{ + FloatParts64 p; + + float8_e4m3_unpack_canonical(&p, a, s); + parts_float_to_float(&p, s); + return bfloat16_round_pack_canonical(&p, s); +} + bfloat16 float8_e5m2_to_bfloat16(float8_e5m2 a, float_status *s) { FloatParts64 p; @@ -2923,6 +2967,15 @@ float64 float16_to_float64(float16 a, bool ieee, flo= at_status *s) return float64_round_pack_canonical(&p, s); } =20 +float8_e4m3 float32_to_float8_e4m3(float32 a, bool saturate, float_status = *s) +{ + FloatParts64 p; + + float32_unpack_canonical(&p, a, s); + parts_float_to_float(&p, s); + return float8_e4m3_round_pack_canonical(&p, s, saturate); +} + float8_e5m2 float32_to_float8_e5m2(float32 a, bool saturate, float_status = *s) { FloatParts64 p; @@ -2999,6 +3052,15 @@ float32 float64_to_float32(float64 a, float_status *= s) return float32_round_pack_canonical(&p, s); } =20 +float8_e4m3 bfloat16_to_float8_e4m3(bfloat16 a, bool saturate, float_statu= s *s) +{ + FloatParts64 p; + + bfloat16_unpack_canonical(&p, a, s); + parts_float_to_float(&p, s); + return float8_e4m3_round_pack_canonical(&p, s, saturate); +} + float8_e5m2 bfloat16_to_float8_e5m2(bfloat16 a, bool saturate, float_statu= s *s) { FloatParts64 p; diff --git a/fpu/softfloat-parts.c.inc b/fpu/softfloat-parts.c.inc index 09be686645..61b07307bf 100644 --- a/fpu/softfloat-parts.c.inc +++ b/fpu/softfloat-parts.c.inc @@ -242,6 +242,15 @@ static void partsN(canonicalize)(FloatPartsN *p, float= _status *status, return; case float_expmax_normal: break; + case float_expmax_e4m3: + if (p->frac_hi =3D=3D 0b111) { + frac_shl(p, fmt->frac_shift); + p->cls =3D (parts_is_snan_frac(p->frac_hi, status) + ? float_class_snan : float_class_qnan); + return; + } + /* otherwise normal */ + break; default: g_assert_not_reached(); } @@ -262,6 +271,21 @@ static void partsN(canonicalize)(FloatPartsN *p, float= _status *status, * The saturate parameter controls saturation behavior for formats that * support it -- when true, overflow produces max normal instead of infini= ty. */ + +/* Helper for uncanon_normal and uncanon, for FP8 E4M3. */ +static void partsN(uncanon_e4m3_overflow)(FloatPartsN *p, float_status *s, + const FloatFmt *fmt, bool satura= te) +{ + assert(N =3D=3D 64); + float_raise(float_flag_overflow | float_flag_inexact, s); + if (saturate) { + p->exp =3D fmt->exp_max; + p->frac_hi =3D E4M3_NORMAL_FRAC_MAX; + } else { + parts_default_nan(p, s); + } +} + static void partsN(uncanon_normal)(FloatPartsN *p, float_status *s, const FloatFmt *fmt, bool saturate) { @@ -360,6 +384,12 @@ static void partsN(uncanon_normal)(FloatPartsN *p, flo= at_status *s, } break; =20 + case float_expmax_e4m3: + if (exp > exp_max || p->frac_hi > E4M3_NORMAL_FRAC_MAX) { + partsN(uncanon_e4m3_overflow)(p, s, fmt, overflow_norm= ); + } + break; + default: g_assert_not_reached(); } @@ -459,9 +489,18 @@ static void partsN(uncanon)(FloatPartsN *p, float_stat= us *s, frac_clear(p); return; case float_class_inf: - assert(fmt->exp_max_kind =3D=3D float_expmax_ieee); - p->exp =3D fmt->exp_max; - frac_clear(p); + switch (fmt->exp_max_kind) { + case float_expmax_ieee: + p->exp =3D fmt->exp_max; + frac_clear(p); + break; + case float_expmax_e4m3: + partsN(uncanon_e4m3_overflow)(p, s, fmt, saturate); + break; + case float_expmax_normal: + default: + g_assert_not_reached(); + } return; case float_class_qnan: case float_class_snan: --=20 2.43.0