make SkJumper stages normal Skia code

Enough clients are using Clang now that we can say, use Clang to build
if you want these software pipeline stages to go fast.

This lets us drop the offline build aspect of SkJumper stages, instead
building as part of Skia using the SkOpts framework.

I think everything should work, except I've (temporarily) removed
AVX-512 support.  I will put this back in a follow up.

I have had to drop Windows down to __vectorcall and our narrower
stage calling convention that keeps the d-registers on the stack.
I tried forcing sysv_abi, but that crashed Clang.  :/

Added a TODO to up the same narrower stage calling convention
for lowp stages... we just *don't* today, for no good reason.

Change-Id: Iaaa792ffe4deab3508d2dc5d0008c163c24b3383
Reviewed-on: https://skia-review.googlesource.com/110641
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>

3 files changed, 32 insertions, 4 deletions

diff --git a/src/core/SkOpts.cpp b/src/core/SkOpts.cpp
index 8a3d30e334..dbf60eaca5 100644
--- a/src/core/SkOpts.cpp
+++ b/src/core/SkOpts.cpp
@@ -40,6 +40,7 @@
 #include "SkBlitRow_opts.h"
 #include "SkChecksum_opts.h"
 #include "SkMorphologyImageFilter_opts.h"
+#include "SkRasterPipeline_opts.h"
 #include "SkSwizzler_opts.h"
 #include "SkUtils_opts.h"
 #include "SkXfermode_opts.h"
@@ -81,11 +82,26 @@ namespace SkOpts {
 
 #undef DEFINE_DEFAULT
 
+#define M(st) (StageFn)SK_OPTS_NS::st,
+    StageFn stages_highp[] = { SK_RASTER_PIPELINE_STAGES(M) };
+    StageFn just_return_highp = (StageFn)SK_OPTS_NS::just_return;
+    void (*start_pipeline_highp)(size_t,size_t,size_t,size_t,void**)
+        = SK_OPTS_NS::start_pipeline;
+#undef M
+
+#define M(st) (StageFn)SK_OPTS_NS::lowp::st,
+    StageFn stages_lowp[] = { SK_RASTER_PIPELINE_STAGES(M) };
+    StageFn just_return_lowp = (StageFn)SK_OPTS_NS::lowp::just_return;
+    void (*start_pipeline_lowp)(size_t,size_t,size_t,size_t,void**)
+        = SK_OPTS_NS::lowp::start_pipeline;
+#undef M
+
     // Each Init_foo() is defined in src/opts/SkOpts_foo.cpp.
     void Init_ssse3();
     void Init_sse41();
     void Init_sse42();
     void Init_avx();
+    void Init_hsw();
     void Init_crc32();
 
     static void init() {
@@ -104,7 +120,8 @@ namespace SkOpts {
         #endif
 
         #if SK_CPU_SSE_LEVEL < SK_CPU_SSE_LEVEL_AVX
-            if (SkCpu::Supports(SkCpu::AVX  )) { Init_avx();   }
+            if (SkCpu::Supports(SkCpu::AVX)) { Init_avx();   }
+            if (SkCpu::Supports(SkCpu::HSW)) { Init_hsw();   }
         #endif
 
     #elif defined(SK_CPU_ARM64)
diff --git a/src/core/SkOpts.h b/src/core/SkOpts.h
index 6e6e16805a..7f4e06613e 100644
--- a/src/core/SkOpts.h
+++ b/src/core/SkOpts.h
@@ -54,6 +54,17 @@ namespace SkOpts {
     static inline uint32_t hash(const void* data, size_t bytes, uint32_t seed=0) {
         return hash_fn(data, bytes, seed);
     }
+
+#define M(st) +1
+    // We can't necessarily express the type of SkJumper stage functions here,
+    // so we just use this void(*)(void) as a stand-in.
+    using StageFn = void(*)(void);
+    extern StageFn stages_highp[SK_RASTER_PIPELINE_STAGES(M)], just_return_highp;
+    extern StageFn stages_lowp [SK_RASTER_PIPELINE_STAGES(M)], just_return_lowp;
+
+    extern void (*start_pipeline_highp)(size_t,size_t,size_t,size_t, void**);
+    extern void (*start_pipeline_lowp )(size_t,size_t,size_t,size_t, void**);
+#undef M
 }
 
 #endif//SkOpts_DEFINED
diff --git a/src/core/SkRasterPipeline.h b/src/core/SkRasterPipeline.h
index 605eebee49..a37ea44e0c 100644
--- a/src/core/SkRasterPipeline.h
+++ b/src/core/SkRasterPipeline.h
@@ -17,8 +17,6 @@
 #include <functional>
 #include <vector>
 
-struct SkJumper_Engine;
-
 /**
  * SkRasterPipeline provides a cheap way to chain together a pixel processing pipeline.
  *
@@ -163,7 +161,9 @@ private:
         void*      ctx;
     };
 
-    const SkJumper_Engine& build_pipeline(void**) const;
+    using StartPipelineFn = void(*)(size_t,size_t,size_t,size_t, void** program);
+    StartPipelineFn build_pipeline(void**) const;
+
     void unchecked_append(StockStage, void*);
 
     SkArenaAlloc* fAlloc;