Skip to content

B13 Coverage w/ Fixes to Some Handling of Subnorms #110

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
coreyqh merged 4 commits into from
May 8, 2026
Merged

B13 Coverage w/ Fixes to Some Handling of Subnorms #110

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
4eae3f4
Correct Subnormal Exponent Tracking from No-Rounding Cases
fourth-bit May 1, 2026
8fd052f
Add B13 Covergroup
fourth-bit May 1, 2026
f50bec8
Return coverfloat_pkg to correct effective exponent calculation
fourth-bit May 1, 2026
4c67d6f
Handle Subnormal Exponents Correctly across all precisions
fourth-bit May 1, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions config.svh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,11 +26,11 @@
// `define COVER_B8
`define COVER_B9
`define COVER_B10
// `define COVER_B11
`define COVER_B11
`define COVER_B12
// `define COVER_B13
`define COVER_B13
`define COVER_B14
// `define COVER_B15
`define COVER_B15
`define COVER_B16
// `define COVER_B17
// `define COVER_B18
Expand Down
234 changes: 234 additions & 0 deletions coverage/covergroups/B13.svh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,234 @@
// Copyright (C) 2025-26 Harvey Mudd College, Ryan Wolk (rwolk@hmc.edu)
//
// SPDX-License-Identifier: Apache-2.0
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, any work distributed under the
// License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND,
// either express or implied. See the License for the specific language governing permissions
// and limitations under the License.

covergroup B13_cg (virtual coverfloat_interface CFI);
option.per_instance = 0;

// Source Format Helpers

F16_src_fmt: coverpoint (CFI.operandFmt == FMT_HALF) {
type_option.weight = 0;
bins f16 = {1};
}

BF16_src_fmt: coverpoint (CFI.operandFmt == FMT_BF16) {
type_option.weight = 0;
bins bf16 = {1};
}

F32_src_fmt: coverpoint (CFI.operandFmt == FMT_SINGLE) {
type_option.weight = 0;
bins f32 = {1};
}

F64_src_fmt: coverpoint (CFI.operandFmt == FMT_DOUBLE) {
type_option.weight = 0;
bins f64 = {1};
}

F128_src_fmt: coverpoint (CFI.operandFmt == FMT_QUAD) {
type_option.weight = 0;
bins f128 = {1};
}

// RFI Instruction

ADD_SUB_op: coverpoint CFI.op {
type_option.weight = 0;
bins add = { OP_ADD };
bins sub = { OP_SUB };
}

// Cancellation Coverpoints

F32_cancellation: coverpoint
(
(
(CFI.intermX == 0 && CFI.intermM != 0)
? -count_leading_zeros(CFI.intermM[INTERM_M_BITS-1:INTERM_M_BITS-256], 256)
: -F32_M_BITS
)
-
(
(effective_exponent(CFI.a, FMT_SINGLE) > effective_exponent(CFI.b, FMT_SINGLE))
? effective_exponent(CFI.a, FMT_SINGLE) + F32_EXP_BIAS
: effective_exponent(CFI.b, FMT_SINGLE) + F32_EXP_BIAS
)
)
{
type_option.weight = 0;
bins cancel[] = {[-F32_P: 1]};
}

F64_cancellation: coverpoint
(
(
(CFI.intermX == 0 && CFI.intermM != 0)
? -count_leading_zeros(CFI.intermM[INTERM_M_BITS-1:INTERM_M_BITS-256], 256)
: -F64_M_BITS
)
-
(
(effective_exponent(CFI.a, FMT_DOUBLE) > effective_exponent(CFI.b, FMT_DOUBLE))
? effective_exponent(CFI.a, FMT_DOUBLE) + F64_EXP_BIAS
: effective_exponent(CFI.b, FMT_DOUBLE) + F64_EXP_BIAS
)
)
{
type_option.weight = 0;
bins cancel[] = {[-F64_P: 1]};
}

F128_cancellation: coverpoint
(
(
(CFI.intermX == 0 && CFI.intermM != 0)
? -count_leading_zeros(CFI.intermM[INTERM_M_BITS-1:INTERM_M_BITS-256], 256)
: -F128_M_BITS
)
-
(
(effective_exponent(CFI.a, FMT_QUAD) > effective_exponent(CFI.b, FMT_QUAD))
? effective_exponent(CFI.a, FMT_QUAD) + F128_EXP_BIAS
: effective_exponent(CFI.b, FMT_QUAD) + F128_EXP_BIAS
)
)
{
type_option.weight = 0;
bins cancel[] = {[-F128_P: 1]};
}

F16_cancellation: coverpoint
(
(
(CFI.intermX == 0 && CFI.intermM != 0)
? -count_leading_zeros(CFI.intermM[INTERM_M_BITS-1:INTERM_M_BITS-256], 256)
: -F16_M_BITS
)
-
(
(effective_exponent(CFI.a, FMT_HALF) > effective_exponent(CFI.b, FMT_HALF))
? effective_exponent(CFI.a, FMT_HALF) + F16_EXP_BIAS
: effective_exponent(CFI.b, FMT_HALF) + F16_EXP_BIAS
)
)
{
type_option.weight = 0;
bins cancel[] = {[-F16_P: 1]};
}

BF16_cancellation: coverpoint
(
(
(CFI.intermX == 0 && CFI.intermM != 0)
? -count_leading_zeros(CFI.intermM[INTERM_M_BITS-1:INTERM_M_BITS-256], 256)
: -BF16_M_BITS
)
-
(
(effective_exponent(CFI.a, FMT_BF16) > effective_exponent(CFI.b, FMT_BF16))
? effective_exponent(CFI.a, FMT_BF16) + BF16_EXP_BIAS
: effective_exponent(CFI.b, FMT_BF16) + BF16_EXP_BIAS
)
)
{
type_option.weight = 0;
bins cancel[] = {[-BF16_P: 1]};
}

// Subnormal Exponent Coverpoint
F32_subnorm_exp: coverpoint(count_leading_zeros(CFI.result[F32_M_UPPER:0], F32_M_BITS)) {
type_option.weight = 0;
bins negative_exp[] = {[F32_M_BITS-1:0]};
}
F32_subnormal: coverpoint(|CFI.result[F32_E_UPPER:F32_E_LOWER]) {
type_option.weight = 0;
bins is_subnormal = { 0 };
}

F64_subnorm_exp: coverpoint(count_leading_zeros(CFI.result[F64_M_UPPER:0], F64_M_BITS)) {
type_option.weight = 0;
bins negative_exp[] = {[F64_M_BITS-1:0]};
}
F64_subnormal: coverpoint(|CFI.result[F64_E_UPPER:F64_E_LOWER]) {
type_option.weight = 0;
bins is_subnormal = { 0 };
}

F128_subnorm_exp: coverpoint(count_leading_zeros(CFI.result[F128_M_UPPER:0], F128_M_BITS)) {
type_option.weight = 0;
bins negative_exp[] = {[F128_M_BITS-1:0]};
}
F128_subnormal: coverpoint(|CFI.result[F128_E_UPPER:F128_E_LOWER]) {
type_option.weight = 0;
bins is_subnormal = { 0 };
}

F16_subnorm_exp: coverpoint(count_leading_zeros(CFI.result[F16_M_UPPER:0], F16_M_BITS)) {
type_option.weight = 0;
bins negative_exp[] = {[F16_M_BITS-1:0]};
}
F16_subnormal: coverpoint(|CFI.result[F16_E_UPPER:F16_E_LOWER]) {
type_option.weight = 0;
bins is_subnormal = { 0 };
}

BF16_subnorm_exp: coverpoint(count_leading_zeros(CFI.result[BF16_M_UPPER:0], BF16_M_BITS)) {
type_option.weight = 0;
bins negative_exp[] = {[BF16_M_BITS-1:0]};
}
BF16_subnormal: coverpoint(|CFI.result[BF16_E_UPPER:BF16_E_LOWER]) {
type_option.weight = 0;
bins is_subnormal = { 0 };
}

// Main Crosses
`ifdef COVER_F32
B13_F32_subnormal_cancel: cross ADD_SUB_op, F32_src_fmt, F32_cancellation, F32_subnorm_exp, F32_subnormal {
ignore_bins min_subnorm_carry = binsof(F32_subnorm_exp.negative_exp) intersect {F32_M_BITS-1} && binsof(F32_cancellation.cancel) intersect {1};
ignore_bins min_subnorm_no_cancel = binsof(F32_subnorm_exp.negative_exp) intersect {F32_M_BITS-1} && binsof(F32_cancellation.cancel) intersect {0};
}
`endif

`ifdef COVER_F64
B13_F64_subnormal_cancel: cross ADD_SUB_op, F64_src_fmt, F64_cancellation, F64_subnorm_exp, F64_subnormal {
ignore_bins min_subnorm_carry = binsof(F64_subnorm_exp.negative_exp) intersect {F64_M_BITS-1} && binsof(F64_cancellation.cancel) intersect {1};
ignore_bins min_subnorm_no_cancel = binsof(F64_subnorm_exp.negative_exp) intersect {F64_M_BITS-1} && binsof(F64_cancellation.cancel) intersect {0};
}
`endif

`ifdef COVER_F128
B13_F128_subnormal_cancel: cross ADD_SUB_op, F128_src_fmt, F128_cancellation, F128_subnorm_exp, F128_subnormal {
ignore_bins min_subnorm_carry = binsof(F128_subnorm_exp.negative_exp) intersect {F128_M_BITS-1} && binsof(F128_cancellation.cancel) intersect {1};
ignore_bins min_subnorm_no_cancel = binsof(F128_subnorm_exp.negative_exp) intersect {F128_M_BITS-1} && binsof(F128_cancellation.cancel) intersect {0};
}
`endif

`ifdef COVER_F16
B13_F16_subnormal_cancel: cross ADD_SUB_op, F16_src_fmt, F16_cancellation, F16_subnorm_exp, F16_subnormal {
ignore_bins min_subnorm_carry = binsof(F16_subnorm_exp.negative_exp) intersect {F16_M_BITS-1} && binsof(F16_cancellation.cancel) intersect {1};
ignore_bins min_subnorm_no_cancel = binsof(F16_subnorm_exp.negative_exp) intersect {F16_M_BITS-1} && binsof(F16_cancellation.cancel) intersect {0};
}
`endif

`ifdef COVER_BF16
B13_BF16_subnormal_cancel: cross ADD_SUB_op, BF16_src_fmt, BF16_cancellation, BF16_subnorm_exp, BF16_subnormal {
ignore_bins min_subnorm_carry = binsof(BF16_subnorm_exp.negative_exp) intersect {BF16_M_BITS-1} && binsof(BF16_cancellation.cancel) intersect {1};
ignore_bins min_subnorm_no_cancel = binsof(BF16_subnorm_exp.negative_exp) intersect {BF16_M_BITS-1} && binsof(BF16_cancellation.cancel) intersect {0};
}
`endif


endgroup
28 changes: 26 additions & 2 deletions src/c/coverfloat.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2180,6 +2180,12 @@ std::pair<int, std::string> reference_model(

intermResult.exp = exp;

if (exp == 0 && sig != 0) {
// In theory, we place the leading one (zero for subnormals) in the INTERM_SIG - 2 position
// so an effective exponent of 0 means that the leading digit is in the INTERM_SIG - 3 position
intermResult.exp = -(INTERM_SIG_LENGTH - 3 - safe_msb(intermResult.sig));
}

intermResult.sign = signBF16UI(result);
break;
}
Expand All @@ -2196,6 +2202,10 @@ std::pair<int, std::string> reference_model(

intermResult.exp = exp;

if (exp == 0 && sig != 0) {
intermResult.exp = -(INTERM_SIG_LENGTH - 3 - safe_msb(intermResult.sig));
}

intermResult.sign = signF16UI(result);
break;
}
Expand All @@ -2205,11 +2215,16 @@ std::pair<int, std::string> reference_model(

if (exp != 0) {
sig |= (1 << 23);
intermResult.sig = mp::cpp_int(sig) << (INTERM_SIG_LENGTH - 2 - 23);
}

intermResult.sig = mp::cpp_int(sig) << (INTERM_SIG_LENGTH - 2 - 23);

intermResult.exp = exp;

if (exp == 0 && sig != 0) {
intermResult.exp = -(INTERM_SIG_LENGTH - 3 - safe_msb(intermResult.sig));
}

intermResult.sign = signF32UI(result);
break;
}
Expand All @@ -2225,6 +2240,10 @@ std::pair<int, std::string> reference_model(

intermResult.exp = exp;

if (exp == 0 && sig != 0) {
intermResult.exp = -(INTERM_SIG_LENGTH - 3 - safe_msb(intermResult.sig));
}

intermResult.sign = signF64UI(result);
break;
}
Expand All @@ -2244,6 +2263,10 @@ std::pair<int, std::string> reference_model(

intermResult.exp = exp;

if (exp == 0 && sig != 0) {
intermResult.exp = -(INTERM_SIG_LENGTH - 3 - safe_msb(intermResult.sig));
}

intermResult.sign = signF128UI64(result >> 64);
break;
}
Expand Down Expand Up @@ -2398,7 +2421,8 @@ std::pair<int, std::string> reference_model(

// 2. Subnormal Handling
if (intermResult.exp <= 0 && !int_format) {
// See s_roundPackToF32.c for why we add 1. Our exp is +1 theirs
// We add 1 here to account for there being no leading one. Consider intermResult.exp=0,
// in this case, we need to shift the result over by one.
int32_t shift_dist = -intermResult.exp + 1;

// Lets shift right jam ourselves now!
Expand Down