From e7ea878dbacee5f1f2c84fbfef50158edbf4c418 Mon Sep 17 00:00:00 2001 From: maochen Date: Sat, 29 Mar 2025 16:05:28 +0800 Subject: [PATCH] =?UTF-8?q?=E3=80=90cleancode=E3=80=91=E3=80=90=E5=AE=89?= =?UTF-8?q?=E5=85=A8=E3=80=91=E6=9B=BF=E6=8D=A2=E9=9D=9E=E5=AE=89=E5=85=A8?= =?UTF-8?q?=E5=87=BD=E6=95=B0memcpy=E7=9A=84=E4=BD=BF=E7=94=A8?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../msprobe/ccsrc/core/AclTensor.cpp | 92 ++++++++++++++----- .../msprobe/ccsrc/utils/DataUtils.cpp | 13 ++- 2 files changed, 78 insertions(+), 27 deletions(-) diff --git a/debug/accuracy_tools/msprobe/ccsrc/core/AclTensor.cpp b/debug/accuracy_tools/msprobe/ccsrc/core/AclTensor.cpp index 4a5ec4c555..968c46a402 100644 --- a/debug/accuracy_tools/msprobe/ccsrc/core/AclTensor.cpp +++ b/debug/accuracy_tools/msprobe/ccsrc/core/AclTensor.cpp @@ -394,8 +394,9 @@ static DebuggerErrno FRAC_Z_TO_NCHW_WITH_GROUPS(AclTensorInfo& tensor) auto coutOpt = AlignCeil(eMult * coutOri, kCubeSize); auto c1Dim = cinOpt / cubeK; const uint8_t* src = tensor.aclData; - uint8_t* dst = tensor.transBuf.data(); + auto dst = tensor.transBuf.begin(); auto dtypeSize = SizeOfAclDType(tensor); + auto dstSize = tensor.transBuf.size(); for (int64_t g = 0; g < groups; ++g) { for (int64_t c = 0; c < cDim; ++c) { @@ -411,8 +412,13 @@ static DebuggerErrno FRAC_Z_TO_NCHW_WITH_GROUPS(AclTensorInfo& tensor) (dstCi / cubeK) * hDim * wDim * coutOpt * cubeK + h * wDim * coutOpt * cubeK + w * coutOpt * cubeK + dstCo * cubeK + temporary; int64_t hstIdx = srcCo * cDim * hDim * wDim + c * hDim * wDim + h * wDim + w; - /* 此处由偏移计算逻辑保障不会越界读写 */ - std::memcpy(dst + hstIdx * dtypeSize, src + devIdx * dtypeSize, dtypeSize); + int64_t devOffset = devIdx * dtypeSize; + int64_t hstOffset = hstIdx * dtypeSize; + if (hstOffset + dtypeSize > dstSize) { + return DebuggerErrno::ERROR_INVALID_VALUE; + } + std::copy(src + devOffset, src + devOffset + dtypeSize, + dst + hstOffset); } } } @@ -450,8 +456,9 @@ static DebuggerErrno FRAC_Z_TO_NCHW(AclTensorInfo& tensor) } const uint8_t* src = tensor.aclData; - uint8_t* dst = tensor.transBuf.data(); + auto dst = tensor.transBuf.begin(); auto dtypeSize = SizeOfAclDType(tensor); + auto dstSize = tensor.transBuf.size(); for (int64_t nIdx = 0; nIdx < n; nIdx++) { int64_t nHeadAddr = nIdx * chw; for (int64_t cIdx = 0; cIdx < c; cIdx++) { @@ -464,8 +471,13 @@ static DebuggerErrno FRAC_Z_TO_NCHW(AclTensorInfo& tensor) auto c0Idx = cIdx % c0; auto ncIdx = nIdx; auto srcIdx = c1Idx * hwncc0 + hIdx * wncc0 + wIdx * ncc0 + ncIdx * c0 + c0Idx; - /* 此处由偏移计算逻辑保障不会越界读写 */ - std::memcpy(dst + dstIdx * dtypeSize, src + srcIdx * dtypeSize, dtypeSize); + auto dstOffset = dstIdx * dtypeSize; + auto srcOffset = srcIdx * dtypeSize; + if (dstOffset + dtypeSize > dstSize) { + return DebuggerErrno::ERROR_INVALID_VALUE; + } + std::copy(src + srcOffset, src + srcOffset + dtypeSize, + dst + dstOffset); } } } @@ -518,8 +530,9 @@ static DebuggerErrno FRAC_NZ_TO_NCHW(AclTensorInfo& tensor) auto numW1 = w / w0; const uint8_t* src = tensor.aclData; - uint8_t* dst = tensor.transBuf.data(); + auto dst = tensor.transBuf.begin(); auto dtypeSize = SizeOfAclDType(tensor); + auto dstSize = tensor.transBuf.size(); for (int64_t timesIdx = 0; timesIdx < times; timesIdx++) { auto timesHead = timesIdx * w1h1h0w0; @@ -531,8 +544,13 @@ static DebuggerErrno FRAC_NZ_TO_NCHW(AclTensorInfo& tensor) for (int64_t i = 0; i < w0; ++i) { int64_t srcIdx = h1h0Head + w1Idx * h1h0w0 + i; int64_t dstIdx = srcHHead + w1Idx * w0 + i; - /* 此处由偏移计算逻辑保障不会越界读写 */ - std::memcpy(dst + dstIdx * dtypeSize, src + srcIdx * dtypeSize, dtypeSize); + int64_t dstOffset = dstIdx * dtypeSize; + int64_t srcOffset = srcIdx * dtypeSize; + if (dstOffset + dtypeSize > dstSize) { + return DebuggerErrno::ERROR_INVALID_VALUE; + } + std::copy(src + srcOffset, src + srcOffset + dtypeSize, + dst + dstOffset); } } auto w1Head = numW1 * w0; @@ -540,8 +558,12 @@ static DebuggerErrno FRAC_NZ_TO_NCHW(AclTensorInfo& tensor) auto srcWIdx = w1Head + w0Idx; int64_t srcIdx = h1h0Head + numW1 * h1h0w0 + w0Idx; int64_t dstIdx = srcHHead + srcWIdx; - /* 此处由偏移计算逻辑保障不会越界读写 */ - std::memcpy(dst + dstIdx * dtypeSize, src + srcIdx * dtypeSize, dtypeSize); + int64_t dstOffset = dstIdx * dtypeSize; + int64_t srcOffset = srcIdx * dtypeSize; + if (dstOffset + dtypeSize > dstSize) { + return DebuggerErrno::ERROR_INVALID_VALUE; + } + std::copy(src + srcOffset, src + srcOffset + dtypeSize, dst + dstOffset); } } } @@ -568,8 +590,9 @@ static DebuggerErrno NC1HWC0_TO_NCHW(AclTensorInfo& tensor) auto c1hwc0 = c1 * hwc0; const uint8_t* src = tensor.aclData; - uint8_t* dst = tensor.transBuf.data(); + auto dst = tensor.transBuf.begin(); auto dtypeSize = SizeOfAclDType(tensor); + auto dstSize = tensor.transBuf.size(); for (int64_t nIndex = 0; nIndex < n; nIndex++) { int64_t nHeadAddr = nIndex * chw; for (int64_t cIndex = 0; cIndex < c; cIndex++) { @@ -581,8 +604,13 @@ static DebuggerErrno NC1HWC0_TO_NCHW(AclTensorInfo& tensor) int64_t c1Index = cIndex / c0; int64_t c0Index = cIndex % c0; int64_t srcIdx = nIndex * c1hwc0 + c1Index * hwc0 + hIndex * wc0 + wIndex * c0 + c0Index; - /* 此处由偏移计算逻辑保障不会越界读写 */ - std::memcpy(dst + dstIdx * dtypeSize, src + srcIdx * dtypeSize, dtypeSize); + int64_t dstOffset = dstIdx * dtypeSize; + int64_t srcOffset = srcIdx * dtypeSize; + if (dstOffset + dtypeSize > dstSize) { + return DebuggerErrno::ERROR_INVALID_VALUE; + } + std::copy(src + srcOffset, src + srcOffset + dtypeSize, + dst + dstOffset); } } } @@ -613,8 +641,9 @@ static DebuggerErrno NDC1HWC0_TO_NCDHW(AclTensorInfo& tensor) const int64_t wc0 = w * c0; const uint8_t* src = tensor.aclData; - uint8_t* dst = tensor.transBuf.data(); + auto dst = tensor.transBuf.begin(); auto dtypeSize = SizeOfAclDType(tensor); + auto dstSize = tensor.transBuf.size(); for (int64_t nIndex = 0; nIndex < n; nIndex++) { int64_t nHead = nIndex * cdhw; for (int64_t cIndex = 0; cIndex < c; cIndex++) { @@ -629,8 +658,13 @@ static DebuggerErrno NDC1HWC0_TO_NCDHW(AclTensorInfo& tensor) int64_t c0Index = cIndex % c0; auto srcIdx = nIndex * dc1hwc0 + dIndex * c1hwc0 + c1Index * hwc0 + hIndex * wc0 + wIndex * c0 + c0Index; - /* 此处由偏移计算逻辑保障不会越界读写 */ - std::memcpy(dst + dstIdx * dtypeSize, src + srcIdx * dtypeSize, dtypeSize); + int64_t dstOffset = dstIdx * dtypeSize; + int64_t srcOffset = srcIdx * dtypeSize; + if (dstOffset + dtypeSize > dstSize) { + return DebuggerErrno::ERROR_INVALID_VALUE; + } + std::copy(src + srcOffset, src + srcOffset + dtypeSize, + dst + dstOffset); } } } @@ -656,8 +690,9 @@ static DebuggerErrno C1HWNCoC0_TO_NCHW(AclTensorInfo& tensor) auto cubeK = GetCubeSizeByType(tensor.dtype); const uint8_t* src = tensor.aclData; - uint8_t* dst = tensor.transBuf.data(); + auto dst = tensor.transBuf.begin(); auto dtypeSize = SizeOfAclDType(tensor); + auto dstSize = tensor.transBuf.size(); for (int64_t nIndex = 0; nIndex < n; nIndex++) { for (int64_t cIndex = 0; cIndex < c; cIndex++) { for (int64_t hIndex = 0; hIndex < h; hIndex++) { @@ -668,8 +703,13 @@ static DebuggerErrno C1HWNCoC0_TO_NCHW(AclTensorInfo& tensor) int64_t coIndex = c0Index; int64_t srcIdx = c1Index * h * w * n * co * c0 + hIndex * w * n * co * c0 + wIndex * n * co * c0 + nIndex * co * c0 + coIndex * c0 + c0Index; - /* 此处由偏移计算逻辑保障不会越界读写 */ - std::memcpy(dst + dstIdx * dtypeSize, src + srcIdx * dtypeSize, dtypeSize); + int64_t dstOffset = dstIdx * dtypeSize; + int64_t srcOffset = srcIdx * dtypeSize; + if (dstOffset + dtypeSize > dstSize) { + return DebuggerErrno::ERROR_INVALID_VALUE; + } + std::copy(src + srcOffset, src + srcOffset + dtypeSize, + dst + dstOffset); } } } @@ -708,8 +748,9 @@ static DebuggerErrno FRAC_Z3D_TO_NCDHW(AclTensorInfo& tensor) auto cdhw = c * dhw; const uint8_t* src = tensor.aclData; - uint8_t* dst = tensor.transBuf.data(); + auto dst = tensor.transBuf.begin(); auto dtypeSize = SizeOfAclDType(tensor); + auto dstSize = tensor.transBuf.size(); for (int64_t nIdx = 0; nIdx < n; nIdx++) { int64_t nHead = nIdx * cdhw; for (int64_t cIdx = 0; cIdx < c; cIdx++) { @@ -725,8 +766,13 @@ static DebuggerErrno FRAC_Z3D_TO_NCDHW(AclTensorInfo& tensor) int64_t ncIdx = nIdx; int64_t srcIdx = dIdx * c1hwn1n0c0 + c1I * c1hwn1n0c0 + hIdx * wn1n0c0 + wI * n1n0c0 + ncIdx * c0 + c0I; - /* 此处由偏移计算逻辑保障不会越界读写 */ - std::memcpy(dst + dstIdx * dtypeSize, src + srcIdx * dtypeSize, dtypeSize); + int64_t dstOffset = dstIdx * dtypeSize; + int64_t srcOffset = srcIdx * dtypeSize; + if (dstOffset + dtypeSize > dstSize) { + return DebuggerErrno::ERROR_INVALID_VALUE; + } + std::copy(src + srcOffset, src + srcOffset + dtypeSize, + dst + dstOffset); } } } diff --git a/debug/accuracy_tools/msprobe/ccsrc/utils/DataUtils.cpp b/debug/accuracy_tools/msprobe/ccsrc/utils/DataUtils.cpp index c2d7df8529..23088d48e3 100644 --- a/debug/accuracy_tools/msprobe/ccsrc/utils/DataUtils.cpp +++ b/debug/accuracy_tools/msprobe/ccsrc/utils/DataUtils.cpp @@ -56,10 +56,15 @@ BFloat16::BFloat16(float f32) BFloat16::operator float() const { - float f32 = 0; - uint32_t tmp = value_; - tmp <<= 16; - std::memcpy(&f32, &tmp, sizeof(f32)); + /* 为了兼容性，不要用c++20的bit_cast */ + union + { + float f32; + uint32_t ui32; + }; + + ui32 = static_cast(value_); + ui32 <<= 16; // 将ui32左移16位 return f32; } -- Gitee