The overhead of a stage today is 3 x86 instructions, typically looking something like this:
- movq (%rdi), %rax // Load the next stage function pointer.
- addq $0x10, %rdi // Step our progress ahead 16 bytes to that next stage.
- jmpq *%rax // Transfer control to that stage.
But if we make sure the pointer's in esi/rsi, we can use lodsd/lodsq to do those first two steps in one instruction:
- lodsq (%rsi), %rax (≈ movq (%rdi), %rax; addq $0x8, %rsi).
- jmpq *%rax
This CL rearranges things so that we can take advantage of this and generally trim off an instruction of overhead. Instead of a vector of {Fn, ctx} pairs, we'll flatten it down into a single interlaced program vector of void*, basically just ommitting any null context pointers. We pass the pointer to program as the second argument to Fn, putting it in rsi. These two changes together make getting the next Fn to call or the current context the same cheap lodsq instruction, encapsulated as load_and_increment().
Here's how the simple "modulate" blend stage changes:
vmulps %ymm4, %ymm0, %ymm0
vmulps %ymm5, %ymm1, %ymm1
vmulps %ymm6, %ymm2, %ymm2
vmulps %ymm7, %ymm3, %ymm3
movq (%rdi), %rax
addq $0x10, %rdi
jmpq *%rax
~~~~~~~~>
vmulps %ymm4, %ymm0, %ymm0
vmulps %ymm5, %ymm1, %ymm1
vmulps %ymm6, %ymm2, %ymm2
vmulps %ymm7, %ymm3, %ymm3
lodsq (%rsi), %rax
jmpq *%rax
This does make getting the current context a one-time, destructive operation. It's switched from referring to ctx as a void* directly to using ctx() as a thunk that returns a void*. No stage so far has ever referred to ctx twice, and it all appears to inline, so this seems harmless. "matrix_2x3" is a good example of what stages that use context pointers end up looking like:
lodsq (%rsi), %rax
vbroadcastss (%rax), %ymm9
vbroadcastss 0x8(%rax), %ymm10
vbroadcastss 0x10(%rax), %ymm8
vfmadd231ps %ymm10, %ymm1, %ymm8
vfmadd231ps %ymm9, %ymm0, %ymm8
vbroadcastss 0x4(%rax), %ymm10
vbroadcastss 0xc(%rax), %ymm11
vbroadcastss 0x14(%rax), %ymm9
vfmadd231ps %ymm11, %ymm1, %ymm9
vfmadd231ps %ymm10, %ymm0, %ymm9
lodsq (%rsi), %rax
vmovaps %ymm8, %ymm0
vmovaps %ymm9, %ymm1
jmpq *%rax
We can't do this with MSVC, as there's no intrinsic for it I can find, and they disallow inline assembly, and rsi is not used to pass arguments to functions there anyway. ARM doesn't need it... it does this in two instructions naturally anyway. We could do this for 32-bit x86 but I'd just rather focus on x86-64.
It's unclear to me that this makes things any faster, but doesn't appear to make things any slower, and makes I think both the code and disassembly simpler.
CQ_INCLUDE_TRYBOTS=skia.primary:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD
Change-Id: Ia7b543a6718c75a33095371924003c5402b3445a
Reviewed-on: https://skia-review.googlesource.com/6271
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reenable test on GTX10700 bot.
BUG=skia:6080
Change-Id: Ieb4292e88fc337c226dad7ac82c6da84879e9522
Reviewed-on: https://skia-review.googlesource.com/6523
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ben Wagner <benjaminwagner@google.com>
abort if incoming data is out of range
TBR=reed@google.com
BUG=676866
Change-Id: I7d4850611654a399e32ea2012b23ca369dc53e70
Reviewed-on: https://skia-review.googlesource.com/6525
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Cary Clark <caryclark@google.com>
Also includes a few additional cleanups.
BUG=skia:
Change-Id: I50899bfef964a3391cc9ddf42c3c5a939e01ceae
Reviewed-on: https://skia-review.googlesource.com/6497
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>
Instead use std::unique_ptr to manage GrOp lifetime.
Change-Id: Ic1dc1e0ffd7254c3994221f498677af5bbf66a71
Reviewed-on: https://skia-review.googlesource.com/6479
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
I was using 'source' where I should have said 'input'. Also, to be
consistent with other image filters, ensure that the input is in the
destination gamut before we start blurring.
BUG=skia:
Change-Id: I751961b42a2a5d110ee8ea8916279c8fe0d5248e
Reviewed-on: https://skia-review.googlesource.com/6486
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
BUG=skia:
Change-Id: Icf616bec73e81aad97815b519566ff5b9db611e3
Reviewed-on: https://skia-review.googlesource.com/6495
Commit-Queue: Hal Canary <halcanary@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
The BCP47FromLanguageID and UnicodeFromMacRoman arrays are logically
const but were not marked as such. Marking them as const lets the
compiler/linker store them in the read-only data segment, which is
strictly better than being in read/write memory. This change moves about
3,000 bytes from the .data to .rdata segment in both chrome.dll and
chrome_child.dll.
BUG=677351
Change-Id: I85ff44d61aa232cf29178833fd2bb2e004032b9e
Reviewed-on: https://skia-review.googlesource.com/6424
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Bruce Dawson <brucedawson@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Share this logic among a couple filters that need it. This also fixes a
bug that showed up in the morhpology GM for GPU color space configs.
Re-land of https://skia-review.googlesource.com/c/6475/
BUG=skia:
Change-Id: If7b8d892cf89fa030bae68bdd3c03118f290f032
Reviewed-on: https://skia-review.googlesource.com/6484
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This reverts commit e02d3caab8.
Reason for revert: New logic triggers out-of-date assert. Reverting until I can fix that, too.
Original change's description:
> Add ImageToColorSpace helper in SkImageFilter
>
> Share this logic among a couple filters that need it. This also fixes a
> bug that showed up in the morhpology GM for GPU color space configs.
>
> BUG=skia:
>
> Change-Id: Ic686b07aff80e58e14a86108703bfbb3cf524979
> Reviewed-on: https://skia-review.googlesource.com/6475
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
>
TBR=bsalomon@google.com,robertphillips@google.com,brianosman@google.com,reviews@skia.org
BUG=skia:
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I00b444dfaaa9b5981f7b33b34419cf9795b52ddb
Reviewed-on: https://skia-review.googlesource.com/6480
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Share this logic among a couple filters that need it. This also fixes a
bug that showed up in the morhpology GM for GPU color space configs.
BUG=skia:
Change-Id: Ic686b07aff80e58e14a86108703bfbb3cf524979
Reviewed-on: https://skia-review.googlesource.com/6475
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This forces all of our testing tools to run with the discrete GPU in
laptop systems that have that option.
BUG=skia:
Change-Id: Ibd7629d6de5f063cdf219b3c7469210af5085d90
Reviewed-on: https://skia-review.googlesource.com/6474
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Brian Osman <brianosman@google.com>
GrMatrixConvolutionFilter is used in two places right now.
The GaussianBlur code ensures that we blur in the space of
the source image, so no conversion is needed. This changes
SkMatrixConvolutionImageFilter to give the same guarantee.
BUG=skia:
Change-Id: I1a6199d4893f2e579dc6fa809c59cd195bd62f90
Reviewed-on: https://skia-review.googlesource.com/6467
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Tag bitmap as sRGB, and adjust gain/bias in color-managed configs to
produce results that are less blown out.
Also added a colorized version of the GM, to validate that gamut
conversion is working.
BUG=skia:
Change-Id: I333988dcdaa1272121e8aa731b4188c942fe19d8
Reviewed-on: https://skia-review.googlesource.com/6466
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
GrRenderTargetOpList now stores the IDs along side each op.
This should put us closer to using proxy IDs and not forcing early render target instantiation as many comments point towards.
Change-Id: I1ee82b01a0818a80d2bcac39fdf3a4ee7dccecc9
Reviewed-on: https://skia-review.googlesource.com/6403
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Correct handling of kGray, k565, k4444 etc. is still a TODO.
SkImage_Generator and SkImage_Gpu are still TODOs.
BUG=skia:6021
Change-Id: Ib53d97d3a866b2b4934fd85c10100855743a8fab
Reviewed-on: https://skia-review.googlesource.com/6396
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Mike Reed <reed@google.com>
GrContext still doesn't convert, but it has the source and destination
color spaces, and call sites are supplying appropriate values where it
makes sense.
BUG=skia:
Change-Id: Ia88733125b8090776cfc9b0dc8030cce365b0b8b
Reviewed-on: https://skia-review.googlesource.com/6400
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Upgrading most Windows bots in the Skolo to version 1607 (build number 14393).
Adding IntelIris540 and GTX1070 bot.
BUG=skia:
Change-Id: Ide7a09a4061d1e387e1d2a22be31b14b13f2fa96
Reviewed-on: https://skia-review.googlesource.com/6357
Reviewed-by: Ravi Mistry <rmistry@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
Since the in/out parameter is a const SkPixmap without the proper
color table, there is no way to tell the client about it without
modifying the const SkPixmap. Rather than cheating, just return
false.
Change-Id: I63fdf57febc59e1ee9af13aa6eb9b253d19bcb17
Reviewed-on: https://skia-review.googlesource.com/6414
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
This reverts commit 9bcda719e8.
Reason for revert: No idea why the bots failed with the original CL. Putting it back in.
Original change's description:
> Revert "Fix issue in SkDebugCanvas where filter canvas prevents GrOp bounds from drawing"
>
> This reverts commit 4bf98e7e80.
>
> Reason for revert: Seems to have caused some compile bots to break-
> https://chromium-swarm.appspot.com/task?id=333c47cadb764f10&refresh=10
>
> Original change's description:
> > Fix issue in SkDebugCanvas where filter canvas prevents GrOp bounds from drawing
> >
> > Change-Id: I3446bfc42c4cf521916a03aa0f367cd38b4fd370
> > Reviewed-on: https://skia-review.googlesource.com/6401
> > Reviewed-by: Ben Wagner <bungeman@google.com>
> > Commit-Queue: Brian Salomon <bsalomon@google.com>
> >
>
> TBR=bsalomon@google.com,bungeman@google.com,reviews@skia.org
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
>
> Change-Id: I0a73fe8c13967233dec950c317693d13d738722a
> Reviewed-on: https://skia-review.googlesource.com/6408
> Commit-Queue: Ravi Mistry <rmistry@google.com>
> Reviewed-by: Ravi Mistry <rmistry@google.com>
>
TBR=bsalomon@google.com,bungeman@google.com,rmistry@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: Iad544f883804d13f04640498b4b63f735d33840e
Reviewed-on: https://skia-review.googlesource.com/6409
Commit-Queue: Ravi Mistry <rmistry@google.com>
Reviewed-by: Ravi Mistry <rmistry@google.com>
This reverts commit 4bf98e7e80.
Reason for revert: Seems to have caused some compile bots to break-
https://chromium-swarm.appspot.com/task?id=333c47cadb764f10&refresh=10
Original change's description:
> Fix issue in SkDebugCanvas where filter canvas prevents GrOp bounds from drawing
>
> Change-Id: I3446bfc42c4cf521916a03aa0f367cd38b4fd370
> Reviewed-on: https://skia-review.googlesource.com/6401
> Reviewed-by: Ben Wagner <bungeman@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
>
TBR=bsalomon@google.com,bungeman@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I0a73fe8c13967233dec950c317693d13d738722a
Reviewed-on: https://skia-review.googlesource.com/6408
Commit-Queue: Ravi Mistry <rmistry@google.com>
Reviewed-by: Ravi Mistry <rmistry@google.com>
- The base class already stores a color space xform, so just forward our
factory argument and use that one.
- We had factories that took arbitrary coefficients, but they were only
used in the unit test. Make the assumption of Mitchell's coefficients
really hard-coded, which lets us skip transposing, storing, comparing,
etc...
BUG=skia:
Change-Id: Ia1d4f72d32c201cba1da24d61b90250f383336d1
Reviewed-on: https://skia-review.googlesource.com/6394
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
99% of the time (or maybe 100%?) morphology will trigger pad_image,
so the input texture will already be in the destimation color space.
If that doesn't happen, then just force the source to be converted,
which keeps the morphology effect and driver code simple.
BUG=skia:
Change-Id: I98876af4f9e9a5da031973213ed76349752ce68f
Reviewed-on: https://skia-review.googlesource.com/6388
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Plumb calls down to SkCodec.
Add a gm
Change-Id: I16da24eb739295ab72f487df02f19968151443f3
Reviewed-on: https://skia-review.googlesource.com/6287
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
This clarifies ownership rules throughout SkFontHost_Mac.cpp, as well as
tidying up various bits of code. Clarifying ownership means replacing
AutoCFRelease with UniqueCFRef which is based on std::unique_ptr. Most of
the cleanup is removing now dead code and properly indenting.
Change-Id: I6d3a225d62b5d0f2f48a9e70c1a24317faa06747
Reviewed-on: https://skia-review.googlesource.com/6297
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Change-Id: I36b8532c57d0b6004a5fd283e30a506df89a4fa6
Reviewed-on: https://skia-review.googlesource.com/6387
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Change-Id: Iedfe5bd019ca1171ab09de569f74c57975aa54c9
Reviewed-on: https://skia-review.googlesource.com/6384
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Change-Id: I80f951976558a284e55386e0a368f08bd835d8ca
Reviewed-on: https://skia-review.googlesource.com/6359
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
std::vector needs to be able to assign objects contained inside it. With
const member variables, this isn't possible. Remove the consts so
SkGIFLZWBlock can be assigned.
BUG=skia:6072
Change-Id: I990dc80fb1c49fbd584712c6d0c1154c2da36e85
Reviewed-on: https://skia-review.googlesource.com/6362
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
