From 90e89339a9bf04fb304129a01ca50f25fdde441d Mon Sep 17 00:00:00 2001
From: Thibaut Goetghebuer-Planchon <thibaut.goetghebuer-planchon@arm.com>
Date: Fri, 7 Aug 2020 16:56:07 +0100
Subject: [PATCH 1/7] Fix potential overflow in 64-bit
MultiplyByQuantizedMultiplier function
---
tensorflow/lite/kernels/internal/common.h | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/tensorflow/lite/kernels/internal/common.h b/tensorflow/lite/kernels/internal/common.h
index 66a2d977f39..c4a620336d3 100644
--- a/tensorflow/lite/kernels/internal/common.h
+++ b/tensorflow/lite/kernels/internal/common.h
@@ -179,7 +179,9 @@ inline int32_t MultiplyByQuantizedMultiplier(int64_t x,
assert(quantized_multiplier >= 0);
assert(shift >= -31 && shift < 8);
- int32_t reduced_multiplier = (quantized_multiplier + (1 << 15)) >> 16;
+ int32_t reduced_multiplier = (quantized_multiplier < 0x7FFF0000)
+ ? ((quantized_multiplier + (1 << 15)) >> 16)
+ : 0x7FFF;
int total_shift = 15 - shift;
x = (x * (int64_t)reduced_multiplier) + ((int64_t)1 << (total_shift - 1));
int32_t result = x >> total_shift;
From 36e7533d3c67dcfd7b23a6a19251676871a37bd7 Mon Sep 17 00:00:00 2001
From: Thibaut Goetghebuer-Planchon <thibaut.goetghebuer-planchon@arm.com>
Date: Wed, 4 Nov 2020 11:05:19 +0000
Subject: [PATCH 2/7] Add tests for MultiplyByQuantizedMultiplier
---
tensorflow/lite/kernels/internal/BUILD | 4 ++
tensorflow/lite/kernels/internal/common.h | 2 +
.../internal/quantization_util_test.cc | 66 +++++++++++++++++++
3 files changed, 72 insertions(+)
diff --git a/tensorflow/lite/kernels/internal/BUILD b/tensorflow/lite/kernels/internal/BUILD
index ad11c06eb37..d1cddfa59ff 100644
--- a/tensorflow/lite/kernels/internal/BUILD
+++ b/tensorflow/lite/kernels/internal/BUILD
@@ -380,6 +380,10 @@ cc_library(
srcs = ["quantization_util.cc"],
hdrs = ["quantization_util.h"],
copts = tflite_copts() + micro_copts(),
+ linkopts = select({
+ "//tensorflow:windows": [],
+ "//conditions:default": ["-lm"],
+ }),
deps = [
":compatibility",
":cppmath",
diff --git a/tensorflow/lite/kernels/internal/common.h b/tensorflow/lite/kernels/internal/common.h
index c4a620336d3..5c5d675928b 100644
--- a/tensorflow/lite/kernels/internal/common.h
+++ b/tensorflow/lite/kernels/internal/common.h
@@ -178,6 +178,8 @@ inline int32_t MultiplyByQuantizedMultiplier(int64_t x,
// - input x is in the range -(1<<47) <= x < (1<<47)
assert(quantized_multiplier >= 0);
assert(shift >= -31 && shift < 8);
+ assert(x >= -(static_cast<int64_t>(1) << 47) &&
+ x < (static_cast<int64_t>(1) << 47) - 1);
int32_t reduced_multiplier = (quantized_multiplier < 0x7FFF0000)
? ((quantized_multiplier + (1 << 15)) >> 16)
diff --git a/tensorflow/lite/kernels/internal/quantization_util_test.cc b/tensorflow/lite/kernels/internal/quantization_util_test.cc
index 053b3116a15..f8de1c345d0 100644
--- a/tensorflow/lite/kernels/internal/quantization_util_test.cc
+++ b/tensorflow/lite/kernels/internal/quantization_util_test.cc
@@ -422,6 +422,72 @@ TEST(QuantizationUtilTest, GetInvSqrtQuantizedMultiplierExp) {
EXPECT_THAT(inv_sqrt(kInt32Max), Pair(189812531, 12));
}
+TEST(QuantizationUtilTest, MultiplyByQuantizedMultiplierInt32) {
+ auto quant_and_multiply = [](int32 x, double multiplier) {
+ int32_t quantized_multiplier;
+ int shift;
+ QuantizeMultiplier(multiplier, &quantized_multiplier, &shift);
+ return MultiplyByQuantizedMultiplier(x, quantized_multiplier, shift);
+ };
+
+ EXPECT_EQ(quant_and_multiply(0, 0.1), 0);
+ EXPECT_EQ(quant_and_multiply(0.1, 0), 0);
+ EXPECT_EQ(quant_and_multiply(10000, 0.00097656), 10);
+ EXPECT_EQ(quant_and_multiply(10000, -0.00097656), -10);
+ EXPECT_EQ(quant_and_multiply(-10000, 0.00097656), -10);
+ EXPECT_EQ(quant_and_multiply(-10000, -0.00097656), 10);
+ EXPECT_EQ(quant_and_multiply(std::numeric_limits<int32_t>::min(), 0.00001),
+ -21475);
+ EXPECT_EQ(quant_and_multiply(std::numeric_limits<int32_t>::min(), -0.00001),
+ 21475);
+ EXPECT_EQ(quant_and_multiply(std::numeric_limits<int32_t>::max(), 0.00001),
+ 21475);
+ EXPECT_EQ(quant_and_multiply(std::numeric_limits<int32_t>::max(), -0.00001),
+ -21475);
+
+ // Test with maximum possible x and quantized_multiplier
+ const int32_t x = std::numeric_limits<int32_t>::max();
+ const int32_t quantized_multiplier = std::numeric_limits<int32_t>::max();
+ const int32_t expected =
+ TfLiteRound(static_cast<int64_t>(x) * quantized_multiplier /
+ static_cast<double>(1ll << 31 - (-3)));
+ EXPECT_EQ(MultiplyByQuantizedMultiplier(x, quantized_multiplier, -3),
+ expected);
+ EXPECT_EQ(MultiplyByQuantizedMultiplier(-x, quantized_multiplier, -3),
+ -expected);
+}
+
+TEST(QuantizationUtilTest, MultiplyByQuantizedMultiplierInt64) {
+ auto quant_and_multiply = [](int64 x, double multiplier) {
+ int32_t quantized_multiplier;
+ int shift;
+ QuantizeMultiplier(multiplier, &quantized_multiplier, &shift);
+ return MultiplyByQuantizedMultiplier(x, quantized_multiplier, shift);
+ };
+
+ EXPECT_EQ(quant_and_multiply(0, 0.1), 0);
+ EXPECT_EQ(quant_and_multiply(0.1, 0), 0);
+ EXPECT_EQ(quant_and_multiply(10000, 0.00097656), 10);
+ EXPECT_EQ(quant_and_multiply(10000, -0.00097656), -10);
+ EXPECT_EQ(quant_and_multiply(-10000, 0.00097656), -10);
+ EXPECT_EQ(quant_and_multiply(-10000, -0.00097656), 10);
+ EXPECT_EQ(quant_and_multiply(-(1ll << 47), 0.00001), -1407385600);
+ EXPECT_EQ(quant_and_multiply(-(1ll << 47), -0.00001), 1407385600);
+ EXPECT_EQ(quant_and_multiply((1ll << 47) - 1, 0.00001), 1407385600);
+ EXPECT_EQ(quant_and_multiply((1ll << 47) - 1, -0.00001), -1407385600);
+
+ // Test with maximum possible x and quantized_multiplier
+ const int64_t x = (1ll << 47) - 1;
+ const int32_t quantized_multiplier = std::numeric_limits<int32_t>::max();
+ // Expected is around 'x * quantized_multiplier / 2**(31 - (-31))' ~= 65536
+ // As there is some rounding error, expected is a bit smaller.
+ const int32_t expected = 65534;
+ EXPECT_EQ(MultiplyByQuantizedMultiplier(x, quantized_multiplier, -31),
+ expected);
+ EXPECT_EQ(MultiplyByQuantizedMultiplier(-x, quantized_multiplier, -31),
+ -expected);
+}
+
TEST(QuantizationUtilTest, PreprocessSoftmaxScaling) {
auto quantize = [](double beta, double scale, int integer_bits) {
int32_t q;
From 291d17a7f5474a743ef2e5b12167f85e12f58220 Mon Sep 17 00:00:00 2001
From: Thibaut Goetghebuer-Planchon <thibaut.goetghebuer-planchon@arm.com>
Date: Tue, 17 Nov 2020 21:39:32 +0000
Subject: [PATCH 3/7] Use intXX_t instead of intxx
---
tensorflow/lite/kernels/internal/quantization_util_test.cc | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/tensorflow/lite/kernels/internal/quantization_util_test.cc b/tensorflow/lite/kernels/internal/quantization_util_test.cc
index f8de1c345d0..65a65fb8ada 100644
--- a/tensorflow/lite/kernels/internal/quantization_util_test.cc
+++ b/tensorflow/lite/kernels/internal/quantization_util_test.cc
@@ -423,7 +423,7 @@ TEST(QuantizationUtilTest, GetInvSqrtQuantizedMultiplierExp) {
}
TEST(QuantizationUtilTest, MultiplyByQuantizedMultiplierInt32) {
- auto quant_and_multiply = [](int32 x, double multiplier) {
+ auto quant_and_multiply = [](int32_t x, double multiplier) {
int32_t quantized_multiplier;
int shift;
QuantizeMultiplier(multiplier, &quantized_multiplier, &shift);
@@ -458,7 +458,7 @@ TEST(QuantizationUtilTest, MultiplyByQuantizedMultiplierInt32) {
}
TEST(QuantizationUtilTest, MultiplyByQuantizedMultiplierInt64) {
- auto quant_and_multiply = [](int64 x, double multiplier) {
+ auto quant_and_multiply = [](int64_t x, double multiplier) {
int32_t quantized_multiplier;
int shift;
QuantizeMultiplier(multiplier, &quantized_multiplier, &shift);
From 0cead482e9973828effe32206a9a749c226e039c Mon Sep 17 00:00:00 2001
From: Thibaut Goetghebuer-Planchon <thibaut.goetghebuer-planchon@arm.com>
Date: Thu, 19 Nov 2020 19:40:03 +0000
Subject: [PATCH 4/7] Fix compilation error in MultiplyByQuantizedMultiplier
test
---
tensorflow/lite/kernels/internal/quantization_util_test.cc | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/tensorflow/lite/kernels/internal/quantization_util_test.cc b/tensorflow/lite/kernels/internal/quantization_util_test.cc
index 65a65fb8ada..8988056ae04 100644
--- a/tensorflow/lite/kernels/internal/quantization_util_test.cc
+++ b/tensorflow/lite/kernels/internal/quantization_util_test.cc
@@ -431,7 +431,7 @@ TEST(QuantizationUtilTest, MultiplyByQuantizedMultiplierInt32) {
};
EXPECT_EQ(quant_and_multiply(0, 0.1), 0);
- EXPECT_EQ(quant_and_multiply(0.1, 0), 0);
+ EXPECT_EQ(quant_and_multiply(1, 0), 0);
EXPECT_EQ(quant_and_multiply(10000, 0.00097656), 10);
EXPECT_EQ(quant_and_multiply(10000, -0.00097656), -10);
EXPECT_EQ(quant_and_multiply(-10000, 0.00097656), -10);
@@ -466,7 +466,7 @@ TEST(QuantizationUtilTest, MultiplyByQuantizedMultiplierInt64) {
};
EXPECT_EQ(quant_and_multiply(0, 0.1), 0);
- EXPECT_EQ(quant_and_multiply(0.1, 0), 0);
+ EXPECT_EQ(quant_and_multiply(1, 0), 0);
EXPECT_EQ(quant_and_multiply(10000, 0.00097656), 10);
EXPECT_EQ(quant_and_multiply(10000, -0.00097656), -10);
EXPECT_EQ(quant_and_multiply(-10000, 0.00097656), -10);
From 100c797f8125c92ef19f7ec65ebc2e9c60c2e616 Mon Sep 17 00:00:00 2001
From: Thibaut Goetghebuer-Planchon <thibaut.goetghebuer-planchon@arm.com>
Date: Wed, 9 Dec 2020 15:15:02 +0000
Subject: [PATCH 5/7] Fix operator precedence warning in
MultiplyByQuantizedMultiplier tests
---
.../kernels/internal/quantization_util_test.cc | 16 +++++++++-------
1 file changed, 9 insertions(+), 7 deletions(-)
diff --git a/tensorflow/lite/kernels/internal/quantization_util_test.cc b/tensorflow/lite/kernels/internal/quantization_util_test.cc
index 8988056ae04..c00593de754 100644
--- a/tensorflow/lite/kernels/internal/quantization_util_test.cc
+++ b/tensorflow/lite/kernels/internal/quantization_util_test.cc
@@ -448,12 +448,13 @@ TEST(QuantizationUtilTest, MultiplyByQuantizedMultiplierInt32) {
// Test with maximum possible x and quantized_multiplier
const int32_t x = std::numeric_limits<int32_t>::max();
const int32_t quantized_multiplier = std::numeric_limits<int32_t>::max();
- const int32_t expected =
+ const int shift = -3;
+ const int32_t expected = static_cast<int32_t>(
TfLiteRound(static_cast<int64_t>(x) * quantized_multiplier /
- static_cast<double>(1ll << 31 - (-3)));
- EXPECT_EQ(MultiplyByQuantizedMultiplier(x, quantized_multiplier, -3),
+ static_cast<double>(1ll << (31 - shift))));
+ EXPECT_EQ(MultiplyByQuantizedMultiplier(x, quantized_multiplier, shift),
expected);
- EXPECT_EQ(MultiplyByQuantizedMultiplier(-x, quantized_multiplier, -3),
+ EXPECT_EQ(MultiplyByQuantizedMultiplier(-x, quantized_multiplier, shift),
-expected);
}
@@ -479,12 +480,13 @@ TEST(QuantizationUtilTest, MultiplyByQuantizedMultiplierInt64) {
// Test with maximum possible x and quantized_multiplier
const int64_t x = (1ll << 47) - 1;
const int32_t quantized_multiplier = std::numeric_limits<int32_t>::max();
- // Expected is around 'x * quantized_multiplier / 2**(31 - (-31))' ~= 65536
+ const int shift = -31;
+ // Expected is around 'x * quantized_multiplier / 2**(31 - shift)' ~= 65536
// As there is some rounding error, expected is a bit smaller.
const int32_t expected = 65534;
- EXPECT_EQ(MultiplyByQuantizedMultiplier(x, quantized_multiplier, -31),
+ EXPECT_EQ(MultiplyByQuantizedMultiplier(x, quantized_multiplier, shift),
expected);
- EXPECT_EQ(MultiplyByQuantizedMultiplier(-x, quantized_multiplier, -31),
+ EXPECT_EQ(MultiplyByQuantizedMultiplier(-x, quantized_multiplier, shift),
-expected);
}
From a66026f713be5f0b737bd6766fe1db6b7ad58156 Mon Sep 17 00:00:00 2001
From: Thibaut Goetghebuer-Planchon <thibaut.goetghebuer-planchon@arm.com>
Date: Thu, 10 Dec 2020 11:02:18 +0000
Subject: [PATCH 6/7] Fix assert in 64-bit MultiplyByQuantizedMultiplier
---
tensorflow/lite/kernels/internal/common.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tensorflow/lite/kernels/internal/common.h b/tensorflow/lite/kernels/internal/common.h
index 9b010603e58..db9cb0ce078 100644
--- a/tensorflow/lite/kernels/internal/common.h
+++ b/tensorflow/lite/kernels/internal/common.h
@@ -179,7 +179,7 @@ inline int32_t MultiplyByQuantizedMultiplier(int64_t x,
assert(quantized_multiplier >= 0);
assert(shift >= -31 && shift < 8);
assert(x >= -(static_cast<int64_t>(1) << 47) &&
- x < (static_cast<int64_t>(1) << 47) - 1);
+ x < (static_cast<int64_t>(1) << 47));
int32_t reduced_multiplier = (quantized_multiplier < 0x7FFF0000)
? ((quantized_multiplier + (1 << 15)) >> 16)
From a8b15d38e47396418c4e15fd4630a68a9762072b Mon Sep 17 00:00:00 2001
From: Thibaut Goetghebuer-Planchon <thibaut.goetghebuer-planchon@arm.com>
Date: Thu, 10 Dec 2020 11:02:39 +0000
Subject: [PATCH 7/7] Remove 64-bit MultiplyByQuantizedMultiplier tests using a
negative multiplier as they are not supported
---
tensorflow/lite/kernels/internal/quantization_util_test.cc | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)
diff --git a/tensorflow/lite/kernels/internal/quantization_util_test.cc b/tensorflow/lite/kernels/internal/quantization_util_test.cc
index c00593de754..b14b7e51007 100644
--- a/tensorflow/lite/kernels/internal/quantization_util_test.cc
+++ b/tensorflow/lite/kernels/internal/quantization_util_test.cc
@@ -466,16 +466,14 @@ TEST(QuantizationUtilTest, MultiplyByQuantizedMultiplierInt64) {
return MultiplyByQuantizedMultiplier(x, quantized_multiplier, shift);
};
+ // Negative multipliers are not supported by the 64-bit
+ // MultiplyByQuantizedMultiplier, only use >= 0 multipliers.
EXPECT_EQ(quant_and_multiply(0, 0.1), 0);
EXPECT_EQ(quant_and_multiply(1, 0), 0);
EXPECT_EQ(quant_and_multiply(10000, 0.00097656), 10);
- EXPECT_EQ(quant_and_multiply(10000, -0.00097656), -10);
EXPECT_EQ(quant_and_multiply(-10000, 0.00097656), -10);
- EXPECT_EQ(quant_and_multiply(-10000, -0.00097656), 10);
EXPECT_EQ(quant_and_multiply(-(1ll << 47), 0.00001), -1407385600);
- EXPECT_EQ(quant_and_multiply(-(1ll << 47), -0.00001), 1407385600);
EXPECT_EQ(quant_and_multiply((1ll << 47) - 1, 0.00001), 1407385600);
- EXPECT_EQ(quant_and_multiply((1ll << 47) - 1, -0.00001), -1407385600);
// Test with maximum possible x and quantized_multiplier
const int64_t x = (1ll << 47) - 1;