From cfa2bba98169cb5ab1945462514be1b6badf7d98 Mon Sep 17 00:00:00 2001
From: Georgios Pinitas <georgios.pinitas@arm.com>
Date: Thu, 27 Jun 2019 17:00:52 +0100
Subject: COMPMID-2178: Update GEMM assembly code.

Perform offset reduction and requantization within the assembly wrapper.

Change-Id: I5d5b3e1f6f9ef4c71805362c57f88ff199c027a3
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1541
Comments-Addressed: Pablo Marquez <pablo.tello@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
---
 src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp | 37 ++++++++++++++++++++++-----
 1 file changed, 30 insertions(+), 7 deletions(-)

(limited to 'src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp')

diff --git a/src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp b/src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp
index 9321bfccfd..84a45f700e 100644
--- a/src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp
+++ b/src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp
@@ -34,18 +34,28 @@
 #include "kernels/a64_gemm_u8_12x8.hpp"
 #include "kernels/a64_gemm_u8_4x4.hpp"
 #include "kernels/a64_hybrid_u8u32_dot_16x4.hpp"
+#include "kernels/a64_smallK_hybrid_u8u32_dot_4x6.hpp"
+#include "kernels/a64_smallK_hybrid_u8u32_dot_4x8.hpp"
 #include "kernels/sve_hybrid_u8u32_dot_4VLx4.hpp"
 #include "kernels/sve_interleaved_u8u32_dot_3VLx8.hpp"
 #include "kernels/sve_native_u8u32_dot_4VLx4.hpp"
+#include "kernels/sve_smallK_hybrid_u8u32_dot_1VLx8.hpp"
 
 namespace arm_gemm {
 
 static const GemmImplementation<uint8_t, uint32_t> gemm_u8_methods[] = {
 #ifdef __ARM_FEATURE_SVE
+{
+    GemmMethod::GEMM_HYBRID,
+    "smallK_hybrid_u8u32_dot_1VLx8",
+    [](const GemmArgs<uint32_t> &args) { return args._Ksize<=64 && args._alpha==1 && args._beta==0 && !args._trA && args._pretransposed_hint; },
+    nullptr,
+    [](const GemmArgs<uint32_t> &args) { return new GemmHybrid<smallK_hybrid_u8u32_dot_1VLx8, uint8_t, uint32_t>(args); }
+},
 {
     GemmMethod::GEMM_HYBRID,
     "hybrid_u8u32_dot_4VLx4",
-    [](const GemmArgs<uint32_t> &args) { return args._Ksize>=16 && args._alpha==1 && !args._trA && !args._trB && args._pretransposed_hint; },
+    [](const GemmArgs<uint32_t> &args) { return args._Ksize>=16 && args._alpha==1 && !args._trA && args._pretransposed_hint; },
     [](const GemmArgs<uint32_t> &args) { return ((args._Ksize <= 128) && (args._Nsize <= 128)) || ((args._nmulti > 1) && ((args._Msize / args._maxthreads) < 8)); },
     [](const GemmArgs<uint32_t> &args) { return new GemmHybrid<hybrid_u8u32_dot_4VLx4, uint8_t, uint32_t>(args); }
 },
@@ -59,15 +69,29 @@ static const GemmImplementation<uint8_t, uint32_t> gemm_u8_methods[] = {
 {
     GemmMethod::GEMM_INTERLEAVED,
     "interleaved_u8u32_dot_3VLx8",
-    [](const GemmArgs<uint32_t> &args) { return (args._Ksize>4); },
+    [](const GemmArgs<uint32_t> &args) { return (args._Ksize>4) && args._alpha==1 && (args._beta==0 || args._beta==1); },
     nullptr,
     [](const GemmArgs<uint32_t> &args) { return new GemmInterleaved<interleaved_u8u32_dot_3VLx8, uint8_t, uint32_t>(args); }
 },
 #endif
+{
+    GemmMethod::GEMM_HYBRID,
+    "smallK_hybrid_u8u32_dot_4x8",
+    [](const GemmArgs<uint32_t> &args) { return args._ci->has_dotprod() && (args._Nsize % 4 == 0) && (args._Ksize<=32) && args._alpha==1 && args._beta==0 && !args._trA && args._pretransposed_hint; },
+    nullptr,
+    [](const GemmArgs<uint32_t> &args) { return new GemmHybrid<smallK_hybrid_u8u32_dot_4x8, uint8_t, uint32_t>(args); }
+},
+{
+    GemmMethod::GEMM_HYBRID,
+    "smallK_hybrid_u8u32_dot_4x6",
+    [](const GemmArgs<uint32_t> &args) { return args._ci->has_dotprod() && (args._Nsize % 4 == 0) && (args._Ksize>32) && (args._Ksize<=64) && args._alpha==1 && args._beta==0 && !args._trA && args._pretransposed_hint; },
+    nullptr,
+    [](const GemmArgs<uint32_t> &args) { return new GemmHybrid<smallK_hybrid_u8u32_dot_4x6, uint8_t, uint32_t>(args); }
+},
 {
     GemmMethod::GEMM_HYBRID,
     "hybrid_u8u32_dot_16x4",
-    [](const GemmArgs<uint32_t> &args) { return args._ci->has_dotprod() && args._Ksize>=16 && !args._trA && !args._trB && args._pretransposed_hint; },
+    [](const GemmArgs<uint32_t> &args) { return args._ci->has_dotprod() && args._Ksize>=16 && args._alpha==1 && !args._trA && !args._trB && args._pretransposed_hint; },
     [](const GemmArgs<uint32_t> &args) { return args._Nsize<=256 && args._Ksize>128; },
     [](const GemmArgs<uint32_t> &args) { return new GemmHybrid<hybrid_u8u32_dot_16x4, uint8_t, uint32_t>(args); }
 },
@@ -100,10 +124,9 @@ const GemmImplementation<uint8_t, uint32_t> *gemm_implementation_list<uint8_t, u
 }
 
 /* Explicitly instantiate the external functions for these types. */
-template UniqueGemmCommon<uint8_t, uint32_t> gemm<uint8_t, uint32_t>(const GemmArgs<uint32_t> &args);
-template KernelDescription get_gemm_method<uint8_t, uint32_t>(const GemmArgs<uint32_t> &args);
-template bool method_is_compatible<uint8_t, uint32_t>(GemmMethod method, const GemmArgs<uint32_t> &args);
-template std::vector<KernelDescription> get_compatible_kernels<uint8_t, uint32_t> (const GemmArgs<uint32_t> &args);
+template UniqueGemmCommon<uint8_t, uint32_t> gemm<uint8_t, uint32_t, Nothing>(const GemmArgs<uint32_t> &args, const Nothing &);
+template KernelDescription get_gemm_method<uint8_t, uint32_t, Nothing>(const GemmArgs<uint32_t> &args, const Nothing &);
+template std::vector<KernelDescription> get_compatible_kernels<uint8_t, uint32_t, Nothing> (const GemmArgs<uint32_t> &args, const Nothing &);
 
 } // namespace arm_gemm
 
-- 
cgit v1.2.1