|
|
Descriptionskcpu: sse4.1 floor, f16c f16<->f32
- floor with roundps is about 4.5x faster when available
- f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions:
+0x180 movups (%r15), %xmm1
+0x184 vcvtph2ps (%rbx), %xmm2
+0x189 movaps %xmm1, %xmm3
+0x18c shufps $255, %xmm3, %xmm3
+0x190 movaps %xmm0, %xmm4
+0x193 subps %xmm3, %xmm4
+0x196 mulps %xmm2, %xmm4
+0x199 addps %xmm1, %xmm4
+0x19c vcvtps2ph $0, %xmm4, (%rbx)
+0x1a2 addq $16, %r15
+0x1a6 addq $8, %rbx
+0x1aa decl %r14d
+0x1ad jne +0x180
If we decide to land this it'd be a good idea to convert most or all users of SkFloatToHalf_01 and SkHalfToFloat_01 over to the pointer-based versions.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1891513002
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Committed: https://skia.googlesource.com/skia/+/cbe3c1af987d622ea67ef560d855b41bb14a0ce9
Committed: https://skia.googlesource.com/skia/+/3faf74b8364491ca806f523fbb1d8a97be592663
Committed: https://skia.googlesource.com/skia/+/244a65350e52c9438931ecdc05a4913f29d343bc
Patch Set 1 #Patch Set 2 : more hacking, mem addr versions #Patch Set 3 : rebase #Patch Set 4 : drop expect #Patch Set 5 : simpler #Patch Set 6 : rebase #Patch Set 7 : typo #Patch Set 8 : another typo #Patch Set 9 : another #Patch Set 10 : typo #Patch Set 11 : invert #ifs #
Messages
Total messages: 37 (21 generated)
Description was changed from ========== skcpu: sse4.1 floor, f16c f16<->f32 BUG=skia: ========== to ========== skcpu: sse4.1 floor, f16c f16<->f32 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ==========
Description was changed from ========== skcpu: sse4.1 floor, f16c f16<->f32 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ========== to ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ==========
Description was changed from ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ========== to ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ==========
Description was changed from ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ========== to ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 COMMIT=false Depends on https://codereview.chromium.org/1890483002/ BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ==========
mtklein@chromium.org changed reviewers: + fmalita@chromium.org, herb@google.com, reed@google.com
Here's how we can use SkCpu to detect support for fast floor() and half<->float operations on x86 at runtime. I've been using the benches Sk4f_floor and xferu64_srcover_bw_N_alpha_f16 to test. This is a patch building on https://codereview.chromium.org/1890483002/.
Description was changed from ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 COMMIT=false Depends on https://codereview.chromium.org/1890483002/ BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ========== to ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 If we decide to land this it'd be a good idea to convert most or all users of SkFloatToHalf_01 and SkHalfToFloat_01 over to the pointer-based versions. COMMIT=false Depends on https://codereview.chromium.org/1890483002/ BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ==========
lgtm
lgtm
The CQ bit was checked by mtklein@google.com to run a CQ dry run
Dry run: CQ is trying da patch. Follow status at https://chromium-cq-status.appspot.com/patch-status/1891513002/160001 View timeline at https://chromium-cq-status.appspot.com/patch-timeline/1891513002/160001
The CQ bit was unchecked by commit-bot@chromium.org
Dry run: Try jobs failed on following builders: Build-Win-MSVC-x86-Debug-Trybot on client.skia.compile (JOB_FAILED, http://build.chromium.org/p/client.skia.compile/builders/Build-Win-MSVC-x86-D...) Build-Win-MSVC-x86_64-Debug-Trybot on client.skia.compile (JOB_FAILED, http://build.chromium.org/p/client.skia.compile/builders/Build-Win-MSVC-x86_6...)
The CQ bit was checked by mtklein@google.com to run a CQ dry run
Dry run: CQ is trying da patch. Follow status at https://chromium-cq-status.appspot.com/patch-status/1891513002/180001 View timeline at https://chromium-cq-status.appspot.com/patch-timeline/1891513002/180001
The CQ bit was unchecked by commit-bot@chromium.org
Dry run: This issue passed the CQ dry run.
The CQ bit was checked by mtklein@google.com
The patchset sent to the CQ was uploaded after l-g-t-m from herb@google.com, fmalita@chromium.org Link to the patchset: https://codereview.chromium.org/1891513002/#ps180001 (title: "typo")
The CQ bit was unchecked by commit-bot@chromium.org
COMMIT=false detected. CQ is abandoning the patch.
Description was changed from ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 If we decide to land this it'd be a good idea to convert most or all users of SkFloatToHalf_01 and SkHalfToFloat_01 over to the pointer-based versions. COMMIT=false Depends on https://codereview.chromium.org/1890483002/ BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ========== to ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 If we decide to land this it'd be a good idea to convert most or all users of SkFloatToHalf_01 and SkHalfToFloat_01 over to the pointer-based versions. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ==========
The CQ bit was checked by mtklein@google.com
CQ is trying da patch. Follow status at https://chromium-cq-status.appspot.com/patch-status/1891513002/180001 View timeline at https://chromium-cq-status.appspot.com/patch-timeline/1891513002/180001
Message was sent while issue was closed.
Description was changed from ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 If we decide to land this it'd be a good idea to convert most or all users of SkFloatToHalf_01 and SkHalfToFloat_01 over to the pointer-based versions. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot ========== to ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 If we decide to land this it'd be a good idea to convert most or all users of SkFloatToHalf_01 and SkHalfToFloat_01 over to the pointer-based versions. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot Committed: https://skia.googlesource.com/skia/+/cbe3c1af987d622ea67ef560d855b41bb14a0ce9 ==========
Message was sent while issue was closed.
Committed patchset #10 (id:180001) as https://skia.googlesource.com/skia/+/cbe3c1af987d622ea67ef560d855b41bb14a0ce9
Message was sent while issue was closed.
A revert of this CL (patchset #10 id:180001) has been created in https://codereview.chromium.org/1891993002/ by mtklein@google.com. The reason for reverting is: Need to change around my #if guards so that clang-cl is treated like GCC and Clang, rather than MSVC..
The CQ bit was checked by mtklein@google.com
The patchset sent to the CQ was uploaded after l-g-t-m from herb@google.com, fmalita@chromium.org Link to the patchset: https://codereview.chromium.org/1891513002/#ps200001 (title: "invert #ifs")
CQ is trying da patch. Follow status at https://chromium-cq-status.appspot.com/patch-status/1891513002/200001 View timeline at https://chromium-cq-status.appspot.com/patch-timeline/1891513002/200001
Message was sent while issue was closed.
Description was changed from ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 If we decide to land this it'd be a good idea to convert most or all users of SkFloatToHalf_01 and SkHalfToFloat_01 over to the pointer-based versions. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot Committed: https://skia.googlesource.com/skia/+/cbe3c1af987d622ea67ef560d855b41bb14a0ce9 ========== to ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 If we decide to land this it'd be a good idea to convert most or all users of SkFloatToHalf_01 and SkHalfToFloat_01 over to the pointer-based versions. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot Committed: https://skia.googlesource.com/skia/+/cbe3c1af987d622ea67ef560d855b41bb14a0ce9 Committed: https://skia.googlesource.com/skia/+/3faf74b8364491ca806f523fbb1d8a97be592663 ==========
Message was sent while issue was closed.
Committed patchset #11 (id:200001) as https://skia.googlesource.com/skia/+/3faf74b8364491ca806f523fbb1d8a97be592663
Message was sent while issue was closed.
A revert of this CL (patchset #11 id:200001) has been created in https://codereview.chromium.org/1897433002/ by mtklein@google.com. The reason for reverting is: this depends on a CL I want to revert.
The CQ bit was checked by mtklein@google.com
CQ is trying da patch. Follow status at https://chromium-cq-status.appspot.com/patch-status/1891513002/200001 View timeline at https://chromium-cq-status.appspot.com/patch-timeline/1891513002/200001
Message was sent while issue was closed.
Description was changed from ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 If we decide to land this it'd be a good idea to convert most or all users of SkFloatToHalf_01 and SkHalfToFloat_01 over to the pointer-based versions. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot Committed: https://skia.googlesource.com/skia/+/cbe3c1af987d622ea67ef560d855b41bb14a0ce9 Committed: https://skia.googlesource.com/skia/+/3faf74b8364491ca806f523fbb1d8a97be592663 ========== to ========== skcpu: sse4.1 floor, f16c f16<->f32 - floor with roundps is about 4.5x faster when available - f16 srcover_n is similar to but a little faster than the version in https://codereview.chromium.org/1884683002. This new one fuses the dst load/stores into the f16<->f32 conversions: +0x180 movups (%r15), %xmm1 +0x184 vcvtph2ps (%rbx), %xmm2 +0x189 movaps %xmm1, %xmm3 +0x18c shufps $255, %xmm3, %xmm3 +0x190 movaps %xmm0, %xmm4 +0x193 subps %xmm3, %xmm4 +0x196 mulps %xmm2, %xmm4 +0x199 addps %xmm1, %xmm4 +0x19c vcvtps2ph $0, %xmm4, (%rbx) +0x1a2 addq $16, %r15 +0x1a6 addq $8, %rbx +0x1aa decl %r14d +0x1ad jne +0x180 If we decide to land this it'd be a good idea to convert most or all users of SkFloatToHalf_01 and SkHalfToFloat_01 over to the pointer-based versions. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is... CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot Committed: https://skia.googlesource.com/skia/+/cbe3c1af987d622ea67ef560d855b41bb14a0ce9 Committed: https://skia.googlesource.com/skia/+/3faf74b8364491ca806f523fbb1d8a97be592663 Committed: https://skia.googlesource.com/skia/+/244a65350e52c9438931ecdc05a4913f29d343bc ==========
Message was sent while issue was closed.
Committed patchset #11 (id:200001) as https://skia.googlesource.com/skia/+/244a65350e52c9438931ecdc05a4913f29d343bc |