aboutsummaryrefslogtreecommitdiff
path: root/src/qc8-gemm
AgeCommit message (Expand)Author
2022-08-16Rename ARMV6SIMD to ARMSIMD32Marat Dukhan
2022-08-03Return xnn_status instead of hard coded integers in JIT generatorsZhi An Ng
2022-08-02Change JIT generator to return uint8_t instead of xnn_status to remove depend...Zhi An Ng
2022-07-28Remove FP16 includes from all microkernelsMarat Dukhan
2022-07-27Replace FP16 for float<->uint32 bitcasts with internal functionsMarat Dukhan
2022-06-21Fix comment about registers reserved in AArch32Frank Barchard
2022-06-14Work around _mm_loadu_si32 broken on latest gccMarat Dukhan
2022-06-13Fix UBSan failures in x86 and scalar microkernelsMarat Dukhan
2022-06-13QC8/QS8/QU8 GEMM/IGEMM microkernels for ARMv6 SIMDMarat Dukhan
2022-06-09Fix UB due to unaligned loads/stores in quantized GEMM/IGEMM/DWCONV microkernelsMarat Dukhan
2022-06-02Tuned performance 1x8/4x8 GEMM/IGEMM for Cortex A7/A35Frank Barchard
2022-05-16QU8 AArch32 4x8 GEMM microkernel remove stack padding to 16 bytes.Frank Barchard
2022-05-12Q8 1x8 AArch32 GEMM for Cortex A7 / A35Frank Barchard
2022-05-10Fix comment for output_max in Q8 AArch32 GEMM/IGEMM JIT microkernelsFrank Barchard
2022-05-10Fix comment for output_max in Q8 AArch32 GEMM/IGEMM microkernelsFrank Barchard
2022-05-09QC8 4x8 GEMM for Cortex A35Frank Barchard
2022-04-18Pass max_mr to JIT generated microkernelsZhi An Ng
2022-04-11Remove unused MUL16/MUL32 QC8/QS8/QU8 GEMM/IGEMM microkernelsMarat Dukhan
2022-04-01assert(nc_mod_nr < 8) for JIT microkernel generators.Frank Barchard
2022-04-01JIT qs8/qc8 assert(kc != 0) in generatorFrank Barchard
2022-03-15Comment change for register usageFrank Barchard
2022-03-09Apply code generatorsFrank Barchard
2022-02-22White space formatting applied for .S filesFrank Barchard
2022-02-13Remap quantization and weight registers for LD64 GEMM/IGEMM on AArch32 to mat...Frank Barchard
2022-02-09C2S4 GEMM/IGEMM microkernels for SSE2/SSE4.1/AVX/XOPMarat Dukhan
2022-02-07Apply source formatting and template generatorsFrank Barchard
2022-02-04Change JIT generators to take nc % nr instead of nc directlyZhi An Ng
2022-02-03Disable neondot microkernels on iOS 32 bitFrank Barchard
2022-02-03Q8 GEMM for Cortex A7 reduce prefetch to weightsFrank Barchard
2022-02-02Make void* params argument of JIT generators constZhi An Ng
2022-02-02QS8 4x8 lane GEMM AArch32 microkernel for Cortex A7Frank Barchard
2022-02-02QC8 4x8 lane GEMM AArch32 microkernel for Cortex A7Frank Barchard
2022-02-02QS8 4x8 lane GEMM AArch32 microkernel for Cortex A53Frank Barchard
2022-01-31Reoptimize NEON QC8/QS8 GEMM/IGEMM microkernels with SR > 1Marat Dukhan
2022-01-31Reoptimize QS8/QC8 GEMM/IGEMM WAsm SIMD microkernels with swizzleMarat Dukhan
2022-01-31Pad K to a multiple of SR in GEMM/IGEMM microkernelsMarat Dukhan
2022-01-27Remove wb from JIT aarch32 instructions, use mem operand and ++ insteadZhi An Ng
2022-01-26QS8/QC8 4x8 dot product IGEMM AArch32 microkernel for Cortex A55Frank Barchard
2022-01-25Remove 3 blank lines after last jit assembly instruction before end of functionFrank Barchard
2022-01-25Avoid importing the entire xnnpack namespace in aarch32 assemblerZhi An Ng
2022-01-25Update assembly register usage comments.Frank Barchard
2022-01-24QC8 4x8 dot product GEMM AArch32 microkernel for Cortex A55Frank Barchard
2022-01-19QU8 4x8 NEON MLA Lane microkernel AArch32 assembly languageFrank Barchard
2022-01-14QC8 GEMM/IGEMM assembly microkernels for ARMv7 NEONFrank Barchard
2022-01-14Change JIT generator signature to accept nc and kc to specialize on those valuesZhi An Ng
2022-01-11Replicate QC8/QS8/QU8 CONV WAsm SIMD parameters to 64 bit rather than 128 bitMarat Dukhan
2022-01-06Add more QC8 GEMM/IGEMM JIT microkernels.Zhi An Ng
2022-01-06Add qc8 gemm and igemm JIT microkernelsZhi An Ng
2022-01-05QC8 NEON lane microkernelsFrank Barchard
2022-01-04Reoptimize microkernel selection for WAsm 1.0Marat Dukhan